Sea Surface Temperature (SST), due to its influence on air-sea interactions, is a critical input into weather models. While physics-based ocean models are continually improving to better represent SST in weather models, data-driven methods offer a promising alternative. In this work, we present an implementation of nonlinear operator inference on a satellite-based SST field (10 km spatial resolution, 1 day temporal resolution) in the northern Indian Ocean, which is known to significantly impact the Indian monsoon. For the prediction of SST, a reduced-order model with a polynomial structure is built non-intrusively from satellite data over a 30-day training period, showing the first five proper orthogonal decomposition modes to capture the SST evolution. A moving-window assimilation scheme utilises the reduced-order model adjoint to correct the prior state, enforcing the model equations over the assimilation window with state observations. Results show that this framework corrects drift, extending the prediction horizon from one week to twenty days. We demonstrate the efficacy of the discovered models using error metrics and their ability to accurately capture lateral SST gradients. Importantly, the inferred operators from the reduced-order model enable the derivation of an explicit adjoint directly from the data, overcoming the computational constraints of General Circulation Models that prohibit rapid adjoint-based assimilation. The performance of the reduced-order model over multiple seasons will also be presented, including the effects of training with data from several years.

How to cite: Madduri, S. H. Y. K., Mathur, M., and Kalur, A.: Data-Driven Modelling and Assimilation of the Sub-Seasonal Evolution of Sea Surface Temperature, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-707, https://doi.org/10.5194/egusphere-egu26-707, 2026.

This study investigated forest edge areas adjacent to a residential road in a hilly area of Nagano Prefecture, Japan, to examine the impact of Cervus nippon (hereafter referred to as “deer”) movements on physical erosion and changes in infiltration capacity of forest soils. The survey area included the edges of cypress and larch forests bordering a residential road west of the Mochizuki Highland Ranch in Mochizuki-machi, Saku City, Nagano Prefecture. Soil erosion was assessed by measuring the height and direction of exposed roots at multiple points. Analysis of root system exposure height (Rh) revealed higher values in the larch forest than in the Japanese cypress forest. Furthermore, the polar coordinate distribution of exposed roots indicated predominant exposure in the steepest slope direction, with some deviations, suggesting that slope angle influences deer movement patterns. Comparisons of cumulative infiltration capacity showed lower values in the cypress forest compared to the larch forest. Soil with clear deer hoof prints exhibited lower infiltration capacity in both areas. The unsaturated hydraulic conductivity (K) for disturbed soil along the deer migration route was approximately half that for natural soil, and in soil with clear deer hoof prints, it decreased to about 1/10 that for natural soil. These findings demonstrate that deer traffic significantly reduces soil infiltration capacity. The results indicated that in forested areas with high levels of deer traffic, K may decrease to 1/2 to 1/10 of normal levels, highlighting the substantial impact of deer activity on forest soil properties.

How to cite: Akita, H., Yusa, S., Yokoyama, H., Kawasaki, M., Kamida, K., Usuda, Y., and Ikeda, M.: Impact of deer traffic on physical soil erosion and changes in infiltration capacity at forest edges, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1331, https://doi.org/10.5194/egusphere-egu26-1331, 2026.

Urban renewal is not only a transformation in urban development models but also a shift in urban governance approaches. Implementing urban renewal initiatives is a crucial component of the new urbanization strategy. After experiencing rapid urbanization characterized primarily by "extensive expansion," Chinese cities are gradually shifting toward "intensive development," entering a stage of optimizing existing urban stock through renewal. As a new engine for promoting high-quality urban development, urban renewal is increasingly becoming a key force in optimizing urban spaces and enhancing people's quality of life. It serves as a vital means to advance modernization and achieve the construction of livable cities. Clarifying the thermal environmental effects of urban renewal and their driving mechanisms can provide targeted management strategies for improving urban thermal environments and enhancing livability.

This study focuses on renewal areas within Fuzhou's built-up zones where significant changes have occurred in building structures while the underlying surfaces remain impervious. We analyzed the spatiotemporal distribution characteristics of heat island intensity at key time nodes and the changes in heat island patterns within the renewal area. Additionally, the differences in thermal environmental effects across different types of urban renewal areas at the block scale have been quantified. On this basis, we explored the driving mechanisms of these thermal environmental effects.

The main findings are as follows: (1) From 2000 to 2022, the urban renewal area of Fuzhou City covered approximately 67 km², with the renewal zone concentrated in the old urban area. Renewal during this period mainly focused on the transformation from high-density mid-to-low-rise buildings to low-density mid-to-high-rise buildings, as well as the transformation of industrial sites.

(2) The spatial distribution of changes in urban heat island intensity aligns closely with urban development types. Areas where heat island intensity weakens are mainly concentrated in urban renewal zones, while areas where it strengthens appear in urban expansion zones. The distribution of extremely strong heat islands shows a migration trend from northwest to southeast, consistent with Fuzhou’s urban development strategy.

(3) Overall, urban renewal has improved the thermal environment of Fuzhou. The average intensity of the urban heat island in the updated area decreased by 1.00°C. The primary change in heat island intensity was the transition from extremely strong heat islands to lower intensity categories, effectively mitigating extreme thermal risks.

(4) The analysis of driving mechanisms shows that the thermal environmental effects of urban renewal are driven by the interaction of the water vapor index (NDMI), vegetation index (NDVI), bare soil index (BSI), building coverage rate (BCR), building height (BH), POI mixture degree, and distance to adjacent green spaces and factories. Among these, BSI and BCR are the main driving forces for the increase in heat island intensity, while BH, POI mixture degree, and distance to adjacent factories are the primary factors driving the decrease in heat island intensity.

How to cite: Liu, Z.: Urban Renewal Makes Cities More Livable-An Empirical Study of Fuzhou City from the Perspective of Thermal Environment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1620, https://doi.org/10.5194/egusphere-egu26-1620, 2026.

In the context of a rapidly changing climate, there is a growing need to assess the impacts of climate change on natural systems, infrastructure, and human activities. Arctic regions are particularly vulnerable, as climate-driven changes extend beyond environmental degradation to significantly affect multiple socioeconomic dimensions. Therefore, there is an increasing need for holistic frameworks capable of capturing and analysing the socioeconomic impacts of climate change on local Arctic communities. In this regard, recent advances in geoinformation technologies - particularly Earth Observation (EO), cloud computing, Geographic Information Systems (GIS), and WebGIS platforms - offer unprecedented opportunities for Arctic climate change research. Nevertheless, a notable gap remains in existing methodological approaches for the effective integration of geoinformatics with socioeconomic studies. This study aims to provide an overview of the EO-PERSIST project, an EU-funded project under the MSCA Staff Exchanges scheme, which aims at developing a cloud-based geospatial platform for understanding the socioeconomic impacts of climate change on Arctic communities. In addition, this study presents the proposed methodological frameworks integrating socioeconomic and geoinformation data developed under EO-PERSIST project, alongside key results from the socioeconomic modeling and the project’s Use Cases. Overall, this work highlights the need for an interdisciplinary and integrated approach that combines EO data, geospatial technologies, and socioeconomic analysis to support informed decision-making in Arctic regions. The EO-PERSIST geospatial platform contributes to this effort by providing key research outputs and methodological approaches that support adaptation strategies and policy development, ultimately enhancing resilience in Arctic permafrost environments.

Keywords: GIS; Earth Observation; Geoinformatics; EO-PERSIST, Cloud Platform, Arctic, Socioeconomic Impact; Acknowledgement The present research study is supported by the project “EO-PERSIST”, funded by the European Union’s Horizon Europe research and innovation program (HORIZON-MSCA-2021-SE-01-01, under grant agreement no. 101086386

How to cite: Starakis, M., Myofa, N., Volianaki, E., Tselos, G. N., Petropoulou, K., Detsikas, S. E., Litke, A., and Petropoulos, G. P.: Assessing Socio-Economic Impacts of Climate Change in the Arctic through Geoinformatics: the contribution of EO-PERSIST project , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1736, https://doi.org/10.5194/egusphere-egu26-1736, 2026.

Climate adaptation is critical for the functionality and quality of life in urban areas under more frequent and severe extreme weather events, such as heatwaves, droughts, and floods. Smaller towns, however, may struggle to adapt because of funding issues, administrative burdens and difficulties using environmental data. This study presents EcoScapes, a decision-support framework to enhance LLM advisory with local Earth observation data. EcoScapes integrates three key components: automated acquisition and preprocessing of Sentinel-2 imagery; Vision Language Models (VLMs) for structured interpretation of satellite-derived representations; and a downstream knowledge-based advisory workflow inspired by prior work.

Given a user-provided town or city name, EcoScapes geocodes the location and retrieves Sentinel-2 imagery for a 5 km bounding box around the urban center. To minimize cloud interference, we use a 1% cloud cover filter, which enables usability but might bias towards drier conditions and miss seasonal water bodies. EcoScapes processes satellite data rendered by the Sentinel-2 API, which includes RGB, water, and moisture views. The system uses a modular analytical pipeline, with an RGB analysis module employing a VLM to describe urban structures, like built-up areas, green spaces, and roads, via focused prompts. This approach reduces hallucinations and ensures more accurate analyses. Separate water and moisture modules analyze the outputs. Water analysis removes small, likely irrelevant features before an RGB-based step connects identified water bodies to their environment and infrastructure. Moisture analysis is used to find heat islands. Finally, a local small language model combines outputs into a single “Climate Report”. This report is subsequently used as context for a ChatClimate-style [1] system that is grounded in the IPCC AR6. This enables a comparison between a baseline advisory system relying on the knowledge base alone and the same system augmented with EcoScapes’ local report.

Since EcoScapes generates varied text outputs, we qualitatively assess its performance using two contrasting case studies: Roßtal, a small rural community of 10,000 people, and Erlangen, a medium-sized city with a population exceeding 100,000. The results indicate EcoScapes can provide useful local context where pre-existing model knowledge is limited. EcoScapes’ report made Roßtal’s adaptation recommendations more relevant and usable, correcting geographically inaccurate suggestions in the baseline. However, EcoScapes’ own inconsistencies and occasional hallucinations remain a limiting factor. The downstream recommendations were affected by errors in interpreting water data in Erlangen, relative to the baseline system, which was more familiar with the city because of its training data. EcoScapes demonstrates Sentinel-2 data’s potential to improve climate advice in smaller towns. Achieving generalization will require improved multimodal reasoning and higher resolution images, while broader evaluation is necessary to determine whether such generalization holds.

More information can be found at our GitHub repository (https://github.com/Photon-GitHub/EcoScapes) and the corresponding paper on arXiv (https://arxiv.org/abs/2512.14373).

References

[1] S. Vaghefi et al., “Chatclimate: Grounding conversational ai in climate science,” Communications Earth & Environment, vol. 4, no. 1, pp. 480, 2023

How to cite: Röhn, M., Gourmelon, N., and Christlein, V.: EcoScapes: LLM-Powered Advice for Crafting Sustainable Cities, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3515, https://doi.org/10.5194/egusphere-egu26-3515, 2026.

There is a pressing need for developing pedagogical frameworks that respond to the damaged, uneven, and entangled planetary conditions of the Anthropocene. I propose “patch-based learning” as a new pedagogical concept, in order to engage learners with the deep predicaments of the Anthropocene. The case study focuses on Yongsan in central Seoul, South Korea—a site marked by layered histories of militarization, displacement, and environmental degradation. Attending to ferality, terrestrial traceability, and denizenship as guiding vectors for traversing Yongsan, I explore ways of reading the site as Anthropocene patches and consider the pedagogical significance of such a reading. I argue that patch-based learning may offer a way to work with the ruptures, leaks, and feral dynamics that characterize planetary landscapes in the Anthropocene.

How to cite: Ahn, S.: How to Reimagine Education in the Anthropocene: Patch-based Learning of Feral Beings and Effects, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4187, https://doi.org/10.5194/egusphere-egu26-4187, 2026.

Calcareous nannofossils represent a key proxy for biostratigraphy and paleoenvironmental reconstructions, due to their high abundance, widespread distribution and rapid evolutionary turnover. However, conventional taxonomic identification under optical or electron microscopy remains time-consuming and strongly dependent on expert interpretation, especially when working with large datasets and heterogeneous assemblages. This limitation is critical for high-resolution stratigraphic studies in complex sedimentary settings where reworking, redeposition and tectonic transport may generate mixed-age associations.

This poster focuses on qualitative and quantitative investigations of Quaternary calcareous nannofossils based on microscopic analyses and the development of an automated taxonomic identification workflow. We propose a deep learning approach using a convolutional neural network (CNN) trained on curated image catalogues of nannofossil taxa, aiming to achieve end-to-end classification of microfossil imagery. The targeted temporal interval spans approximately on the last 25,000 years (since the LGM – Last Glacial Maximum), focused on samples from the NW Black Sea cores.

Beyond accelerating routine identifications, automated classification has the potential to provide more objective and reproducible taxonomic assignments, enabling consistent quantitative counting and supporting multidisciplinary analyses linking nannofossil variability to paleoenvironmental controls such as salinity, nutrient input and temperature. The proposed workflow represents a step toward scalable microfossil taxonomy, supporting robust stratigraphic correlations and palaeoceanographic interpretations in Quaternary successions.

Keywords: nannofossils, neural networks, image recognition

How to cite: Cudalbu, C., Cudalbu, B., and Melinte - Dobrinescu, M.: Automated Taxonomic Identification of Calcareous Nannofossils from Microscopic Imagery Using Convolutional Neural Networks, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7766, https://doi.org/10.5194/egusphere-egu26-7766, 2026.

In May 2024, Samcheok Blue Power Unit 1, a coal-fired thermal power plant in Samcheok, Gangwon Province, South Korea, began commercial operation. Together with Unit 1 (completed in October 2023) and Unit 2 (scheduled for completion by the end of 2025), the Samcheok Blue Power complex will reach an installed capacity of 2,100 MW. Considering that fossil fuels are a decisive contributor to climate change and that South Korea has officially pledged to phase out coal by 2050, the continued construction and operation of a new coal plant in 2024 appears paradoxical. This puzzle becomes even more striking given that Samcheok has been widely recognized as a region with a successful anti-nuclear movement, suggesting the presence of an active environmental politics and a history of resistance to energy megaprojects.

To explore this contradiction, this research investigates how fossil fuel infrastructure is sustained through intertwined “circuits” of capital, material, and affect. In doing so, the study engages with debates on the technosphere, understood as a global assemblage of energy systems, infrastructures, institutions, and material interdependencies that shape, and often constrain, social and ecological futures. Rather than treating infrastructure as a self-contained system with clear boundaries, the study proposes the concept of infra-circuits. This concept emphasizes that infrastructures function as nodal points within circuits that are simultaneously connected and closed: they enable specific forms of connection while restricting others, much like electronic circuits that allow flow only through certain configured pathways. Infra-circuits are also chained, meaning that if one link is disrupted, the stability of the entire configuration is threatened unless alternative routes can be mobilized.

Importantly, infra-circuits are not only spatial but also temporal. They operate through inherited material pathways, regulatory arrangements, financial instruments, and labor regimes that bind present energy decisions to past investments and future obligations. While this resonates with socio-technical systems theory and its emphasis on path dependence, the concept of infra-circuits allows for analytical dimensions that remain underdeveloped in conventional approaches to technological adoption and innovation. Specifically, it draws attention to how infrastructures endure by assembling heterogeneous circuits of matter, finance, and affect, thereby revealing the intimate relationship between fossil development and patterned forms of public sentiment, attachment, fear, and aspiration.

By highlighting the chained and temporally extended nature of these circuits, this study argues that fossil infrastructure persists not merely due to economic rationality or policy failure, but because it is embedded in technospheric arrangements that stabilize particular futures while foreclosing others. Ultimately, the concept of infra-circuits offers a framework for rethinking fossil energy infrastructure as a material and affective formation situated at the apex of ecological crises in the Anthropocene.

How to cite: Kim, J.: Infra-circuits of fossil capital and Technosphere: More-than-human politics of the Samcheok thermal power plant, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8729, https://doi.org/10.5194/egusphere-egu26-8729, 2026.

How, and to what extent, can societal norms legitimately enter scientific knowledge-making, or can science intervene in societal norm-making? This question has become a key matter in defining and studying the Anthropocene as a new geological epoch. This paper aims to enrich the discussion by examining how two kinds of norms – one operating primarily within the boundary of science and the other originating from broader societal concerns – came to intersect in the debate over formalizing the Anthropocene as a new geological epoch. The first part of the paper traces the historical development of the GSSP practice as the central normative backbone of modern chronostratigraphy. Drawing on archival documents from the International Subcommission on Stratigraphic Classification (ISSC), in which the concept of GSSP was first debated and negotiated, it shows how the classification of geological time became a GSSP-based institutional practice through specific procedures, standards, and conventions for recognizing particular stratigraphic signals as valid evidence for defining geological time. Against this historical backdrop, the second part points out that, from its inception, the Anthropocene has carried the reflexive mode of thinking about the consequences of human activities, such as climate change, biodiversity loss, and habitability, hence calling for planetary stewardship. Since a new geological epoch can only be ratified through the acceptance of a specific GSSP proposal, formalizing the Anthropocene became a site at which the scientific norms constructed in the late 20^th century for the development of GSSP are brought into contact with the 21^st-century societal norms embedded in the concept of a human-driven Earth-system change. In a nutshell, the very term “Anthropocene” connotes both descriptive and prescriptive practices.

In 2023, the Anthropocene Working Group (AWG) submitted a GSSP proposal identifying plutonium-239 fallout from the mid-20^th-century nuclear testing as a globally synchronous marker, supported by multiple auxiliary stratigraphic proxies. As maintained by Skelton and Noone (2025) and the members of the AWG, this proposal has met the formal GSSP requirements with evidential robustness exceeding those of many previously ratified epochs. Nevertheless, the Subcommission on Quaternary Stratigraphy (SQS) voted to reject the proposal. This paper argues that the difficulties surrounding the formalization of the Anthropocene do not stem from matters of empirical evidence, but from matters of normative science: i.e., how existing scientific norms are to be interpreted, negotiated, and sometimes reconstructed when they encounter the pressure of societal imperatives to address planetary transformations. The paper thus asks how scientists should navigate the deeply humanistic implications of their stratigraphic decision about the Anthropocene.

How to cite: Koh, K. and Park, B. S.: Formalizing the Anthropocene: an interplay between normative knowledge-making and societal norm-making, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8747, https://doi.org/10.5194/egusphere-egu26-8747, 2026.

Urban-rural interdependencies from an Earth system’s view

Global demand for resources such as food, building materials and water is rising, while land take —driven significantly by urbanization—is accelerating and has become a critical factor. This surge in demand is accompanied by the spatial decoupling of production and consumption regions, leading to unevenly distributed environmental damage. Consequently, issues like soil degradation, water pollution, and greenhouse gas emissions are externalized and cause the deterioration of natural conditions in the hinterland or in teleconnected rural areas. Accordingly, sustainability balancing ecological, social and economic aspects can hardly be achieved.

While the Earth system sciences in the Anthropocene also deal with the cumulative effects of human activity on environmental change, research on urban-rural interdependencies in the context of global sustainability remains rare. However, compliance with Earth system boundaries requires integrated approaches across resources, sectors and spatial scales. This necessitates rethinking urban-rural relationships beyond the traditional dichotomy of producers and consumers and instead views them as cooperative socio-ecological systems.

Based on the thematic examples of food, material, water and land use, we highlight regional approaches and derive three fundamental principles—‘circularity’, ‘spatial justice’, and ‘participation’—alongside with two heuristic perspectives: ‘socio-ecological systems thinking’ and ‘framing and governance'. hey are used to propose an advanced research agenda covering (i) an integrated framework for system knowledge on the complex and dynamic urban-rural interdependencies, (ii) scientific references for regional target knowledge informed by Earth boundaries, and (iii) the examination of governance structures as transformation knowledge to enable cross-regional design and implementation.

How to cite: Warner, B. and Müller-Petke, M.: Urban-rural interdependencies from an Earth system’s view, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9385, https://doi.org/10.5194/egusphere-egu26-9385, 2026.

Climate models are not designed to provide detailed information on local rainfall that may trigger an outbreak of diarrhoea, but are nevertheless able to reproduce large-scale climatic conditions, processes, and phenomena. Hence, they have a minimum skillful scale, and downscaling makes use of skilfully simulated large-scale aspects in addition to information about how local rainfall depends on those larger scale conditions. The SPRINGS project studies the link between climate change and diarrhoea outbreak through a chain of models, where one stage provides input to the next. It’s important to design such model chains so that they provide a flow of salient and relevant information. This framework also needs to ensure robust results, as different global climate model simulations may give a different regional outlooks. It also needs to involve proper evaluation, and it's important that it is designed for both how the end-results are being used in decision-making, and that the end-results are correctly interpreted in terms of what they really represent. Here, such a framework used in SPRINGS is presented.

How to cite: Benestad, R.: Using global climate model simulations for outlooks on how climate change affects future diarrhoea risks, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9608, https://doi.org/10.5194/egusphere-egu26-9608, 2026.

Rising extreme temperatures driven by climate change are expected to significantly degrade outdoor thermal conditions, stretching the day of extreme heat and leaving fewer hours for comfortable and safe outdoor activity, while increasing the health risks associated with outdoor exposure. This study investigates the impact of climate change on thermal discomfort across the Kingdom of Saudi Arabia. Projections from two CMIP6 models, at 6-hour temporal resolution, were used to compute the Discomfort Index (DI) based on dry-bulb temperature and relative humidity, and to assess diurnal variations in thermal stress at 03 UTC (06:00 AST), 09 UTC (12:00 AST), 15 UTC (18:00 AST), and 21 UTC (00:00 AST) under SSP1-2.6, SSP2-4.5, and SSP5-8.5 scenarios. Changes were evaluated for the near (2021–2040), mid (2041–2060), and far future (2081–2100). Thermal discomfort across Saudi Arabia intensifies progressively from the historical period to the far future, exhibiting pronounced spatial and diurnal variability. Historically, daytime discomfort (09–15 UTC) had a mean DI value of ~25.4 °C, corresponding to uncomfortable conditions across most regions, with some areas, particularly in the southeast and coastal regions, reaching very uncomfortable conditions. Early morning and evening hours (03–21 UTC) were slightly lower, with mean DI values around 22.8–23.4 °C, corresponding to slightly uncomfortable conditions. Future projections indicate a substantial increase in discomfort magnitude, particularly in coastal and southeastern areas. In the near-future (2021–2040), mean DI values increase to ~25–26 °C during daytime and ~23 °C during early morning and evening hours. By the mid-future (2041–2060), at 09 UTC (12:00 AST), the southeast and coastal regions are very uncomfortable and can reach extremely uncomfortable conditions under SSP5-8.5, reflecting peak thermal stress during the day. In the far-future period (2081–2100), at 09 UTC (12:00 AST), mean DI values reach ~27–28.6 °C under SSP2-4.5 and SSP5-8.5 scenarios, with maximum values exceeding 32 °C in the southeast region under SSP5-8.5, corresponding to dangerous conditions, highlighting the severity of midday thermal stress and its potential impacts on outdoor activities and urban livability. Evening and early morning mean DI values also rise substantially compared to historical conditions, reaching ~25–27 °C (uncomfortable), with some regions, particularly in the southeast, reaching up to ~30.5 °C (extremely uncomfortable), indicating that nighttime relief is markedly reduced and thermal discomfort persists even outside peak daytime hours. These findings emphasize the necessity of adaptive strategies to ensure the resilience, safety, and comfort of outdoor environments under increasing heat stress.

How to cite: Abuwaer, N., Vinodhkumar, B., and Al-Ghamdi, S.: Clocking the Heat: Projected Diurnal Patterns of Thermal Discomfort Across Saudi Arabia Under Future Climate Scenarios, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10525, https://doi.org/10.5194/egusphere-egu26-10525, 2026.

High-quality, temporally consistent training samples are the cornerstone of accurate long-term urban Land Use/Land Cover (LULC) mapping. However, traditional sample generation relies heavily on labor-intensive manual interpretation and often lacks reproducibility. To address this, we developed PRTS-AI (Primary Regulated Time-series Sampling), an open-source system that integrates OpenStreetMap (OSM) data extraction, Large Language Model (LLM)-driven semantic classification, and LandTrendr-based temporal filtering into an automated workflow. By leveraging generative AI (e.g., DeepSeek/ChatGPT/Gemini) to interpret polygon attributes and using POI-based consistency checks, the system significantly reduces manual workload while ensuring semantic accuracy.

The PRTS-AI system integrates multi-source spatial and temporal data into a streamlined workflow, including:

(1) extraction of OpenStreetMap (OSM) features for user-defined study areas;

(2) semantic classification of polygon features using large language models;

(3) detection and filtering of change pixels using the LandTrendr time-series algorithm;

(4) recommendation of city-specific sampling parameters based on a six-dimensional urban typology framework.

This system enables reproducible multi-temporal sample generation, spatial heterogeneity validation, and fine-scale classification support across diverse urban settings. Furthermore, this system can operate in parallel with the usual land cover sample selection and subsequent classification processes.

We applied PRTS-AI to map the urban evolution of diverse cities in Liaoning and Shandong provinces, China, from 2000 to 2020. The framework achieved an overall mapping accuracy of ~80%, with residential categories reaching 90%. Beyond mapping, we utilized the fine-grained Local Climate Zone (LCZ) metrics generated by the system to investigate the transferability of samples. Through Principal Component Analysis (PCA) of residential morphologies, we quantitatively identified that cities cluster into distinct typologies driven by macro-factors (e.g., coastal vs. resource-based industrial cities) rather than administrative hierarchies. These findings challenge the assumption of universal sample transferability, suggesting that sample migration is most effective within specific urban typologies. Consequently, PRTS-AI incorporates a typology-based parameter recommendation module to guide city-specific sampling. This study presents a scalable, AI-empowered solution for urban mapping and offers new insights into the spatiotemporal heterogeneity of urban forms.

However, limited sample transferability may still be achieved between cities with similar characteristics, based on a preliminary six-dimensional classification framework.

PRTS-AI provides a lightweight, reproducible, and extensible solution for urban LULC research, supporting both academic investigations and practical urban planning applications.

How to cite: Tian, T., Yu, L., Chen, B., and Gong, P.: From Generative Sampling to Urban Typology: A PRTS-AI Supported Framework for Multi-Decadal Urban LULC Mapping and Cross-City Transferability Analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13942, https://doi.org/10.5194/egusphere-egu26-13942, 2026.

The net zero (NZ) agenda is one of the foremost planetary challenges facing urban policymakers - presenting both localised impacts, along with transnational co-operation and governance challenges. Data is fundamental to measuring policy success and failure and taking informed intervention decisions, and global Earth Observation data has enhanced evidence bases in terms of where local actions are needed. In the face of evolving national politics, it is common for city-regions to lead on NZ policy. However, the distribution of multi-level powers and resources fundamentally shapes what urban leaders can do and who they need to work with to respond to the climate emergency. Given this complex policy architecture, progress towards NZ is dependent on the effective use of data. Many intermediate city-regions need support to build capacity and marshal data effectively, and questions about which data sources to deploy at specific contexts can be difficult to resolve. This leads to the possibility for a gap between the sophistication of data which may be able to support policymakers – increasingly available from breakthrough techniques and modelling – and capability, governance and communication issues in subnational policymakers’ ability to act. Starting with the end users of data at city-region level, we explore the need for better understanding between the policy and data/science communities.

How to cite: Allan, G., Oda, T., and Waite, D.: How does the growing availability of novel data interact with the uses of data by policymakers in city-regions on their journey to net zero?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14570, https://doi.org/10.5194/egusphere-egu26-14570, 2026.

Coastal communities in Bangladesh are increasingly exposed to a range of natural hazards due to their low elevation, the dynamic nature of river systems, and environmental changes driven by climate. This study presents an integrated geospatial framework for assessing multi-hazard vulnerability and mapping community resources in Dakhin Bedkashi Union, Koyra Upazila, a coastal administrative unit bordering the Sundarbans mangrove forest. The research addresses six key hazards that impact the region: riverbank erosion, cyclones, flooding and tidal surges, waterlogging and salinity intrusion, drought, and earthquakes. This study employed a mixed-methods approach combining remote sensing analysis, GIS-based spatial modeling, and participatory assessment techniques. Temporal analysis of riverbank erosion was conducted using Normalized Difference Water Index (NDWI) derived from Landsat imagery (1990–2022) processed in Google Earth Engine. Cyclone exposure was evaluated through historical track digitization (1990–2022) and network analysis to determine shelter accessibility within 500m, 1000m, and 1500m service areas. Flood susceptibility, earthquake risk zonation, and seasonal drought patterns were mapped using datasets from the Bangladesh Agricultural Research Council and Space Research (BRAC) and Space Research and Remote Sensing Organization (SPARRSO). Primary data collection included three Focus Group Discussions (n=47 participants), two Key Informant Interviews, and GPS-based ground truthing of critical infrastructure. Results indicate that river erosion and tidal flooding pose the highest risks to the study area, followed by cyclone exposure and waterlogging. The NDWI time-series reveals progressive land loss along the Kopotakkho River, exacerbated by inadequate embankment construction and proliferation of informal sluice gates for shrimp aquaculture. Network analysis demonstrates that residents in peripheral wards must travel over 45 minutes on foot to reach cyclone shelters, with accessibility further constrained by predominantly unpaved road networks. The area falls within earthquake Zone III (moderate risk) but remains vulnerable to potential tsunami-induced coastal inundation. Community consultations revealed that while cyclone impacts have decreased due to improved early warning systems, chronic hazards including erosion, salinity intrusion, and waterlogging increasingly threaten livelihoods and freshwater security. The resource mapping component identified critical gaps in disaster response infrastructure: only four cyclone shelters and one health facility serve a population exceeding 16,000. Housing vulnerability is acute, with 98% of structures classified as non-permanent (kaccha) construction. This research demonstrates how combining top-down remote sensing with bottom-up community knowledge can expose the hidden spatial dimensions of socioeconomic vulnerability in climate-threatened deltas.

How to cite: Rabbi, N. F., Tazwar, M., Mahmud, S. A., and Muna, T. S.: Integrating Remote Sensing and Participatory Assessment Techniques to Map Multi-Hazard Vulnerability and Resource Gaps: A Geospatial Study of Socioeconomic Inequity of Coastal Bangladesh, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15302, https://doi.org/10.5194/egusphere-egu26-15302, 2026.

As human exploration advances into increasingly hostile and isolated environments, such as extraterrestrial habitats on Mars, the Moon, or deep-sea stations, the concept of resilience must evolve beyond its traditional technical and physiological dimensions.

This need becomes particularly critical in contexts of long-duration habitation, where survival alone is insufficient to guarantee long-term operational stability and human wellbeing.

Central to this assertion is the recognition that resilience entails examining construction in relation to permanence, which may also be understood as a sense of feeling at home, shifting resilience from a purely performance-based concept to a relational and experiential condition.

This perspective requires redirecting science, technology, and design toward the conditions that enable habitation to become sustainable, meaningful, and socially durable.

This includes environmental adaptation, understood as the strategic use of local raw materials and regenerative systems, reducing dependency on external supply chains and increasing environmental compatibility, as well as the processes accompanying construction, which involve the complex relationships between these local materials, the tools, crafts, and other elements that make construction possible.

These construction–material ecologies play a decisive role in transforming temporary shelters into places of permanence.

Finally, it encompasses cultural embeddedness, which acknowledges the importance of cultural identity, symbolic practices, and sensory experiences that converge in the creation of an atmosphere of resilience, influencing perception of safety, cohesion, and long-term habitability.

The literature on this concept is fragmented due to the complexity and interdisciplinary nature of the aspects involved in the state of feeling at home.

Architecture, design, sociology and anthropology, nutrition, indoor environmental quality (thermal, acoustic, lighting, olfactory), tactile experience, physical activity, structural safety, and risk perception all contribute to this condition, yet are rarely addressed within a unified framework.

A common view across disciplines is missing in the related literature, yet it is of fundamental importance to understand and to design the future of resilient spatial architecture, both in extraterrestrial settings and in climate-stressed environments on Earth.

This abstract proposes a theoretical framework for understanding resilience in these terms, emphasizing the integration of cultural, psychological, material, and collaborative factors in the sustainable design of long-term human settlements in hostile environments.

By reframing resilience as the capacity to sustain a sense of “being at home”, the framework offers a shared conceptual ground for interdisciplinary dialogue across environmental sciences, engineering, architecture, and the social sciences.

 It challenges the prevailing techno-centric framing of resilience in extreme environments, arguing instead for a holistic approach that embraces human complexity, cultural roots, and collaborative innovation, with direct implications for climate adaptation, remote communities, and future off-Earth settlements.

How to cite: Alcindor, M., Salese, F., Sangiorgio, V., Araújo, A. M., Rodrigues, P. F. S., and Simão, E.: Feeling at Home as a Dimension of Resilience in Architecture for Extreme Environments , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16322, https://doi.org/10.5194/egusphere-egu26-16322, 2026.

India, with its rapid urbanisation, faces high pollution levels and continues to fail to meet World Health Organisation (WHO) standards, accounting for 17 of the 30 most polluted cities globally.  The annual economic losses incurred due to its polluted air are equivalent to almost 3 per cent of the nation’s GDP.  Effective air pollution management requires adequate budgetary support and resource allocation. To address this, the National Clean Air Programme (NCAP), launched in 2019, is India’s flagship programme aimed at achieving a 40 per cent reduction in particulate concentration by 2026 in 130 non-attainment cities. NCAP implementation is supported through multiple funding streams, including convergence of existing national schemes like Smart Cities Mission, Swachh Bharat Mission, etc., as well as Fifteenth Finance Commission (XV-FC) and NCAP grants. Through this, India established a framework for financing clean air action, but challenges related to capital absorption and impact persist. As of October 2025, only 59.15 per cent of the NCAP funds and 77 per cent of XV-FC funds have been utilised, and by 2024-25, only 25 out of 130 cities have reduced PM 10 levels by 40 per cent.

This study critically examines the evolution of the fund disbursal mechanism (pre-requisites, performance assessment criteria and disbursement) over the years, by tracing the fund flow mechanism and Portal for Regulation of Air pollution in Non-Attainment cities (PRANA) records. Furthermore, it compares allocation versus absorption and assesses structural and operational complexities that limit the impact of fund utilisation and overall cost-effectiveness. This study leverages a mixed-methods approach, integrating insights from secondary literature and city-level field consultations. The analysis identifies a set of design and implementation constraints, including limited mechanisms for assessing the effectiveness of fund utilisation, sectoral prioritisation that is not consistently aligned with air quality outcomes, weak interdepartmental coordination and capacity limitations at the city level. It also highlights inadequate recognition of city-level initiatives within performance assessment frameworks, the absence of a sufficiently targeted and results-oriented approach, and delays in state-level financial systems that affect the timeliness of fund disbursal, and in turn, the overall progress of the programme. In addition, issues pertaining to data availability, pollution monitoring representativeness, and operation and maintenance requirements continue to influence programme performance. The study emphasises the value of integrating procedural and statutory costs and considerations into financial planning processes, strengthening institutional capacities and promoting effective fund utilisation. The findings aim to inform policy deliberations on air quality governance and financing in India. 

How to cite: Tiwari, A., Srivastava, R., and Goel, U.: Fund Flows and Absorption Challenges under India’s National Clean Air Programme (NCAP) — Evidence from public financial management systems and city-level consultations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17008, https://doi.org/10.5194/egusphere-egu26-17008, 2026.

Climate-sensitive vector-borne diseases are increasingly influenced by environmental and climatic variability, posing growing challenges for public health preparedness under climate change. Within the Planet4Health project, an Early Warning System (EWS) is being developed to support anticipatory decision-making for climate-sensitive diseases by integrating climate, environmental, and epidemiological information into operational risk products.

This contribution presents an EWS focused on the sand fly vector (Phlebotomus spp.) and leishmaniasis over the Iberian Peninsula, using machine learning–based modelling approaches. The system integrates high-resolution climate data, climate-derived indicators (e.g. temperature, humidity, and precipitation-related indices), land and environmental variables, and vector presence information to model conditions favourable for sand fly activity and disease transmission. The modelling strategy prioritises interpretable machine learning techniques to ensure transparency and usability for public health and veterinary stakeholders.

The EWS operates across multiple temporal scales, addressing short-term and seasonal forecasts, while also incorporating climate projections to assess potential future changes in  environmental suitability for sand flies and associated disease risk. Machine learning models are trained and evaluated using historical climate and entomological data, capturing non-linear relationships between environmental drivers and vector presence while explicitly accounting for uncertainty. Model outputs are translated into spatially explicit risk maps and alert-oriented indicators designed to support operational surveillance and decision-making.

Results from the Iberian sand fly–leishmaniosis case study demonstrate that the EWS successfully reproduces known spatial patterns of vector suitability and seasonal dynamics across the Peninsula, as well as interannual variability linked to climatic anomalies. The modular and data-driven design of the system supports adaptation of the framework to other regions and climate-sensitive diseases, in line with the broader objectives of Planet4Health.

Funding: The PLANET4HEALTH consortium is funded by the European Commission grant 101136652. The five Horizon Europe projects, GO GREEN NEXT, MOSAIC, PLANET4HEALTH, SPRINGS, and TULIP, form the Planetary Health Cluster. The data for EDENext was obtained from the Palebludata website (https://www.palebludata.com). The data for Vectornet was obtained from the ECDC.

How to cite: Natal, S., San-Martín, D., Maia, C., Marme, R., Berriatua, E., Verdú-Serrano, E., Risueño, J., Pérez-Cutillas, P., JImenez, M., and Molina, R.: An Early Warning System for sand fly-borne diseases in the Iberian Peninsula, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17834, https://doi.org/10.5194/egusphere-egu26-17834, 2026.

Mediterranean territories are increasingly exposed to growing environmental fragility, risks linked to climate change and associated environmental disasters, as well as persistent socioeconomic challenges exacerbated by long-established patterns of urbanization. In this context, nature-based solutions (NBS) have been promoted by European policy frameworks as key tools for addressing these challenges. However, despite their growing political relevance, NBS often encounter barriers to implementation related to low public acceptance, limited social legitimacy, concerns about environmental and social justice, and insufficient integration of local knowledge and everyday practices.

This study addresses this gap by examining how local communities perceive, interpret, and interact with NBS in Mediterranean contexts through public participation processes in urban environments. The analysis focuses on several case studies located in Italy (Catania and Ferla, Demo 4), France (Saint-Jérôme and Saint-Charles, Marseille, Demo 5), Spain (Zaragoza, Demo 6), and Cyprus (Nicosia, Demo 9), within the CARDIMED project. These cases include various implementations of NBS, such as rain gardens, vertical green walls, green facades with vertical gardening and hydroponic systems, photobioreactor systems, biological drainage channels, and other nature-based interventions.

The study is theoretically grounded in socio-ecological governance and sustainability transition theories, conceptualizing NBS not only as technical measures but as relational and well-being-oriented solutions capable of reshaping human-environment relationships and strengthening social cohesion The participatory methodology draws on behavioral economics principles to analyze the underlying human behaviors, attitudes, and perceptions that condition NBS acceptance. To explore these dynamics, structured focus groups were conducted with key community representatives  (5 - 14 participants per group) to investigate shared perceptions, experiences, and concerns towards NBS, as well as their role in shaping narratives on water conservation, climate resilience, and sustainable land-use practices. The qualitative data were then analyzed using content analysis and ATLAS.ti software.

The results indicate that participatory processes play a decisive role in improving the awareness, legitimacy, and long-term governance of NBS, while revealing the structural and institutional constraints that risk undermining their transformative potential. These findings provide critical insights and pave the way for further investigation into justice-based and socially rooted NBS implementation pathways, supporting greater societal acceptance and strengthening collective ownership.

How to cite: Giuffrida, E. R., Sciuto, L., Cirelli, G. L., Quina Gomez, A., Muñoz Gonzalez, D. B., Fernandez, B. G., Zamudio Correa, A. F., and Licciardello, F.: Towards a citizen-based green transition: Nature-Based Solutions in mediterranean areas: CARDIMED project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19141, https://doi.org/10.5194/egusphere-egu26-19141, 2026.

Climate change adaptation and mitigation take place in a complex world associated with deep uncertainties related to external factors, among others population growth, new technologies, socio-economic developments and their subsequent impacts (Haasnoot et al., 2013, 2024). Therefore, there is a need for flexible framework that can respond to these challenges, bridge the social and environmental sciences and support climate change mitigation. Scenario planning can assist in developing integrative mental models to deliver pathways of change while incorporating alternative policies, evolving innovative practices and management options (Sroufe and Watts, 2022). The fundamental strength of the pathways approach is their ability to deal with uncertainty by assessing possible future impacts and navigating across multiple future trajectories. Pathways are designed to achieve future vision and assess whether the desired objectives have been accomplished (Coulter, 2019). Specifically, this approach can help to explore potential future trajectories, investigate innovation for carbon sequestering and more sustainable agriculture (Sroufe and Watts, 2022). So far, limited literature concerning development of pathways approach to GHG mitigation in agriculture exists.

Our approach aims to combine SSPs (Shared-socioeconomic pathways) downscaled for the Czech Republic within AdAgriF project with Mitigation pathways developed for Czech agriculture. Such integration enables us to assess the full potential of particular SSP-pathway combinations while considering future uncertainties. These SSP-independent pathways are not tied to a single SSP storyline, but instead each pathway is assessed for robustness across SSPs. This approach avoids over-commitment to one socio-economic future and highlights no-regret and robust mitigation pathways (bundles of measures).

This presentation highlights the process of interdisciplinary cooperation in order to support the pathway co-development, which involves exploring potential trajectories of pathways and their mitigation measures as well as SSPs with modelling using various agro-ecosystem simulation models that will be applied.

References:

Haasnoot, M., Di Fant, V., Kwakkel, J., & Lawrence, J. (2024). Lessons from a decade of adaptive pathways studies for climate adaptation. Global Environmental Change, 88, 102907. https://doi.org/10.1016/j.gloenvcha.2024.102907

Haasnoot, M., Kwakkel, J. H., Walker, W. E., & Ter Maat, J. (2013). Dynamic adaptive policy pathways: A method for crafting robust decisions for a deeply uncertain world. Global Environmental Change, 23(2), 485–498. https://doi.org/10.1016/j.gloenvcha.2012.12.006

Sroufe, R., & Watts, A. (2022). Pathways to Agricultural Decarbonization: Climate Change Obstacles and Opportunities in the US. Resources, Conservation and Recycling, 182, 106276. https://doi.org/10.1016/j.resconrec.2022.106276

How to cite: Krkoška Lorencová, E., Suchá, L., Koudelková, M., and Harmáčková, Z.: Designing mitigation pathways in Czech agriculture, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20080, https://doi.org/10.5194/egusphere-egu26-20080, 2026.

The most critical blind spot in contemporary climate crisis response is the reliance on standardized macro-metrics, which obscure the specific reality of individual suffering. Just as economics uses consumer sentiment to capture household realities and meteorology uses apparent temperature to reflect physiological truths, occupational safety must transition toward integrating perceived risks that exist beyond mere numerical thresholds. This study argues that human perception functions as a high-fidelity biological integration of environmental stressors and conceptualizes it as a Perceptual Trigger: an embodied risk signal with diagnostic and policy relevance. The 2023 fatality of a young logistics worker in Korea illustrates the lethal failure of current systems; while sensors recorded ambient conditions within regulatory thresholds, the system failed to register the worker’s chest tightness—a critical physiological survival signal.

To bridge this gap, a Living Lab for Heatwave Adaptation was implemented in August 2025, engaging 30 port workers from Incheon and 6 technicians from a specialized manufacturer of surface treatment additives as active co-creators. In this study, workers were not treated as mere subjects for data extraction but were empowered as epistemic agents who fundamentally identified and defined hazards within their real-world micro-climates. This study employed the Living Lab methodology as a requisite mechanism to derive Worker Perception Data, which can only be captured within the complex real-world context of the field. Through the systematic qualitative analysis of this co-creation process, the researcher demonstrated that complex heat risks—such as localized radiant heat, engine emissions, and entrapped micro-climates—which are systematically overlooked by standardized sensor arrays, can be effectively rendered into data via worker perception.

The core contribution of this research lies in its translational process: converting Worker Perception Data into systematic risk signals (Information), consolidating them into collectively validated Evidence, and establishing the Policy Grounds for the right to stop work. The researcher proposes a Complementary Governance Model that precisely fills the blind spots of technical sensor monitoring with the acute sensitivity of worker perception data, arguing that this model is a vital mechanism for ensuring site-specific climate adaptation. By framing the datafication of lived experience as an act of Industrial Democracy, this approach serves as an essential interface for connecting grassroots experience with institutional decision-making.

How to cite: Kim, J.: Beyond Metric-Centric Adaptation: Redefining Occupational Heatwave Governance through Living Lab Co-creation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20084, https://doi.org/10.5194/egusphere-egu26-20084, 2026.

We present the progress of the project Twin-ER: Pilot Digital Twin for Earthquake Risk. The goal of the project is the integration of digital models of the city and the Earth into the structure of a digital twin, focused on seismic risk.

The Earth model includes the generation of new seismic source models based on maps correlating surface deformation and seismic activity rates. Deformation maps will be determined through the analysis of GNSS time series and InSAR images for several dates. Seismic activity rates will be calculated by combining statistical analyses of the seismic catalog with mechanical analyses of earthquake-related stress changes in the crust. The derived maps will show location-, magnitude-, and time-dependent activity rates. Seismic source models will form the basis for the development of seismic hazard maps and constitute the main component of the Earth model.

The city model integrates innovative exposure models based on Cadastral data, enhanced with machine learning and deep learning algorithms to identify building typologies and their seismic vulnerability. These analyses will incorporate data of different nature, such as cadastral reference value or exposure time to high temperatures, with the aim of extending the exposure to a multi-hazard and multi-risk context. The exposure and vulnerability models constitute the main component of the city model.

By combining seismic hazard models on one hand, and exposure and vulnerability models on the other, the seismic risk model will be obtained. This model represents the expected damage and losses in a city in the event of an earthquake. Therefore, it is a crucial piece of information for proposing risk mitigation measures and planning emergency response.

Both Earth and city models are embebed into the digital twin seismic risk. This digital twin is conceived in a pilot phase. The model will be fed with the results of risk simulations, which can be visualized in a web environment, leaving aspects of data loading automation from updated sensors or external servers and subsequent simulations with that updated data for future developments.

The project is applied in two study areas of similar size but different, complementary characteristics. One is southeastern Spain, where (1) seismic activity is moderate, and major earthquakes occur rarely, (2) cities have a relatively old building stock and are more vulnerable to earthquakes, and (3) the availability and accessibility to cadastral data are optimal. The other study area is El Salvador, where (1) there is high seismic activity with frequent large earthquakes, (2) cities have a relatively modern building stock with abundant informal construction, and (3) there is no free access to cadastral data.

 The advances presented here include the UML model of the entire digital twin, the seismic activity and deformation maps in SE Spain, and the city 3D models of two scenarios of application.

How to cite: Staller, A., Gaspar-Escribano, J., Torres, Y., Martínez-Cuevas, S., Arranz, J. J., García-Aranda, C., Iturrioz, T., and Pallero, J. L. G. and the Twin-ER Team: Progress of the Twin-ER project: pilot digital twin for earthquake risk, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20253, https://doi.org/10.5194/egusphere-egu26-20253, 2026.

Planetary Boundaries (PBs) define biophysical limits that safeguard Earth system stability. Exceeding these limits undermines ecosystem services, food security, economic stability, and climate resilience. Humanity is currently transgressing several of the PBs, demanding integrative and transformative research approaches that connect biophysical monitoring, sustainability targets, and societal decision-making. Despite its conceptual strength, the PB framework remains difficult to operationalize for regional agricultural systems. Global-scale assessments obscure the pronounced spatial heterogeneity of farming landscapes, where localized exceedances in nitrogen cycling, freshwater use, climate sensitivity, and biosphere integrity accumulate to drive broader Earth system risks. Consequently, there are limited guidance on where, how, and under which biophysical constraints agriculture can remain productive without breaching local environmental limits. This study proposes an integrated monitoring and modelling paradigm to assess regional agricultural production within planetary boundaries.

Our method moves beyond static, indicator-based assessments toward a dynamic, process-aware evaluation of local biophysical variables. We integrate high-resolution climate, soil, and land-use data with a spatially explicit crop model (SIMPLACE) to define regional control variables, including yield thresholds, nitrate leaching, and water-stress limits. To address structural uncertainties and capture non-linear climate-crop-soil interaction, we develop a hybrid modelling approach that couples SIMPLACE with machine learning algorithm (XGBoost).

Using SSP5-8.5 projections, we quantify specific yield and environmental constraints for Winter Wheat and Silage Maize in the Berlin–Brandenburg region in Germany. Hybrid simulations significantly outperform standalone process-based models, reducing mean absolute percentage error by ~9% for Winter Wheat and yielding consistently higher skill for Silage Maize. Our results reveal that emerging local boundaries are increasingly governed by compound climate extremes, particularly heat stress and precipitation deficits during flowering and early grain filling.

By framing PBs at the regional scale, hybrid modelling approaches enable the identification of conditions under which agricultural productivity, climate adaptation, and environmental integrity remain compatible—and where biophysical limits impose fundamental constraints. This approach offers a transferable pathway for embedding planetary stewardship into regional agricultural planning, climate adaptation strategies, and land-system governance.

How to cite: Halder, K., Srivastava, A. K., Almeida, B., Nowak, L., Awoke, M. D., Stuckas, H., Fritz, S., Helming, K., and Ewert, F.: Assessing Agricultural Production within Planetary Boundaries using an Integrated Monitoring and Hybrid Modelling Approach, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20658, https://doi.org/10.5194/egusphere-egu26-20658, 2026.

Climate change contributes to the emergence of multiple hazards, including zoonotic diseases whose transmission dynamics are closely linked to environmental and socio-ecological transformations. Following the Covid-19 pandemic, a re-emergence of CCHF was observed in Senegal, particularly in rural areas where livestock farming plays a central role. This emerging zoonosis, transmitted mainly through ticks and infected cattle, remains poorly understood by the general population and disproportionately affects women involved in agro-pastoral activities. While epidemiological responses currently use a One Health framework, the approach often lacks community inclusion and adequate consideration of mental health. Previous global health emergencies (Ebola and Covid-19) have led to social, psychological, and emotional disruptions, causing fear and reinforcing misconceptions about health measures and denial of disease, particularly in contexts where cultural beliefs and mistrust hinder public health interventions. This study analyses the psychosocial effects associated with the emergence of CCHF in order to identify key challenges for epidemic interventions within a broader context of climate-related health risks. A mixed-methods approach was conducted across eight regions of Senegal, combining surveys, observations, and in-depth interviews (IDIs). Quantitative surveys were administered to 434 livestock keepers at the household level, alongside interviews with 6 farmers to assess knowledge of zoonotic diseases and risk perception. In 2023, field observations focused on surveillance activities, followed in 2024 by IDIs with 10 directly affected individuals, including bereaved families, and 6 health professionals involved in case management. The findings reveal limited knowledge and low risk perception of zoonotic diseases among livestock keepers, who often rely on informal practices for disease management. High levels of psychological distress, including fear, panic, insomnia, and social stigma, were reported among patients, relatives, and communities. Isolation measures and restrictions on visits intensified suffering, eroded trust in response teams, and in some cases triggered hostility toward intervention actors. Health professionals experienced ethical dilemmas between their duty of care and fear of infection, exacerbated by harsh climatic conditions. The study highlights the need for systemic and multidisciplinary risk-reduction strategies that extend beyond biomedical control. This call for Integrating structured psychosocial support, community engagement, and culturally sensitive communication. Strengthening the links between environmental change, disease emergence, mental health, and social behaviour is essential to enhancing resilience and preparedness for future epidemics in climate-vulnerable contexts.
Keywords: emerging zoonotic diseases, CCHF, climate-related health risks, risk perception, psychosocial effects, epidemic intervention, Senegal.

How to cite: Ndoye, F., Sène, M., Ndione, A. G., Sow, A., Tine, J. A., Leneman, M., Boersma, K., Ndour, A. P., and Ngowi, H. A.: Psychosocial effects and intervention challenges during the re-emergence of Crimean–Congo Hemorrhagic Fever (CCHF) in Senegal, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21166, https://doi.org/10.5194/egusphere-egu26-21166, 2026.

Small Island Developing States (SIDS) experience disproportionate vulnerability to natural and climate related hazards driven by geographic constraints, demographic trends, limited economic diversification and growing development pressures. In the Caribbean, flooding is one of the region’s most devastating and recurrent hazards, contributing to substantial socio-economic losses. Despite frequent events, many SIDS lack the long-term datasets needed to characterize flood behavior, particularly for coastal compound flooding, involving the interaction of multiple drivers such as storm surge, waves, tides, precipitation, runoff and river discharge. Climate change, including sea level rise, is expected to alter these processes and increase uncertainty in both magnitude and frequency.

Coastal ecosystems such as mangrove forests are increasingly recognized for their potential as Nature-based Coastal Solutions (NBCS), offering coastal protection alongside social, environmental and economic co-benefits. However, key gaps remain, including limited understanding of their flood mitigative properties across varying hydrodynamic conditions and stages of ecosystem maturity and health. Although numerical models are widely used to assess flood hazards, their ability to represent multiple interacting drivers and incorporate NBCS remains limited, a challenge that is particularly pronounced in data-sparse regions. Addressing these limitations requires field data to develop numerical models.

The relevance of these challenges becomes particularly clear in Trinidad’s South Oropouche River Basin (SORB), a low lying and highly flood prone watershed on the southwest coast that includes mangrove areas within the Godineau Swamp. This study therefore centers on collecting the necessary datasets and integrating them into the numerical modelling needed to characterize compound flooding in this basin. Field monitoring in SORB includes weather stations, water level loggers, short-term ADCP deployments, and a paired camera and water level logger system designed to capture flood depth and extent at a high resolution. Additional measurements including water quality parameters and vegetation characteristics from field surveys and satellite imagery, will support the mangrove related parameterization.

The modelling will be forced primarily using open-source datasets, with field observations used to assess their performance and suitability. Comparison of radar rainfall with in-situ measurements will enable the development of a bias-corrected relationship, allowing long-term radar datasets to be translated into site-specific rainfall inputs for compound flood modelling. These observations will be supplemented by historical datasets, including river discharge, Intensity–Duration–Frequency (IDF) curves, bathymetry and land cover. Thus, the numerical model will simulate the key hydrodynamic processes driving compound flooding while mangrove influences will be represented using vegetation-drag formulations to capture momentum dissipation and associated reductions in inundation. Field observations will be used to calibrate and validate the model, enabling spatial estimates of flood depth and extent under different forcing scenarios.

Field monitoring in SORB is expected to provide new insights into how flood drivers interact to generate inundation, as well as emerging trends and patterns, while deterministic modelling will quantify the degree to which mangroves mitigate flooding. Together, the data-collection and modelling approaches offer a practical means of improving compound flood assessment in regions with limited long-term observations and support a more holistic evaluation of NBCS for SIDS.

How to cite: Williams, A.: Field Data Collection to Support the Numerical Modelling of Mangrove Contributions to Compound Flood Mitigation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1439, https://doi.org/10.5194/egusphere-egu26-1439, 2026.

Mangrove forests provide critical shoreline protection in tropical and sub-tropical regions through wave attenuation, soil accretion and floodwater storage. These protective mechanisms relate to both ecosystem functionality and persistence (Lovelock et al. 2024). Multiple studies over the past decades have effectively shown that mangrove forest extent can lead to reduced wave heights between 50-99%, with vegetative characteristics slowly being introduced as a critical element (McIvor et al. 2013). Increasing evidence has identified that the eco-geomorphological conditions shape the consistency and scale of protection but have not been properly considered. Ecological, hydrodynamic and geomorphological processes which occur at various temporal and spatial scales influence species-specific interactions, functional type formations and habitat structure (Gijsman et al. 2021). Mangrove forests can develop into distinct ecotypes over time (Twilley and Rivera-Monroy 2009), directly influenced by tidal exchanges between the mangrove forests and nearshore environments, affecting the level of productivity within the mangroves (Mitsch and Gosselink 2015). These interactions influence the mangrove forest structure through variability in sediment deposition rates, biomass accumulation, seedling recruitment and overall forest productivity (van Hespen et al. 2023).

Since variations in eco-geomorphological features affect mangrove functionality and persistence at multiple scales, this research will investigate how these differences can affect the ability of mangroves to provide consistent coastal protection. Building on existing modelling approaches (Beselly, van Der Wegen, and Roelvink 2025), the aim is to design an ideal model capable of capturing nuanced interactions between mangrove ecosystems and the geomorphological features. For instance, predictive models (WAPROMAN), designed to capture wave propagation through a uniform forest, utilised drag coefficients (McIvor et al. 2013), while a measure of mangrove forest extent seaward followed a mechanistic approach using the window of opportunity for seedling establishment predictions (van Hespen et al. 2023).

The current workflow will identify and isolate the key drivers and traits of crucial mangrove forests that affect mangrove functionality and persistence, for parameterisation. As a preliminary approach, these parameters will be integrated into a numerical model, incorporating elements from previous mechanistic and empirical approaches, modified to ascertain and accommodate the variability in mangrove eco-geomorphology and sediment dynamics (Gijsman et al. 2021). This process can facilitate the quantification of impact and identify key thresholds that these selected attributes of mangrove forests have on the function and persistence related to long-term coastal protection. Through this integration of multiple layers of eco-geomorphological variability, this work offers insights into how mangrove systems work as Nature-based Solutions and how they thrive within our changing climate.

How to cite: Emmanuel, S.: Mangrove traits influencing coastal protection under varying environmental and eco-geomorphic conditions. , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1457, https://doi.org/10.5194/egusphere-egu26-1457, 2026.

Mythogenic Mountain Landscapes and Shakta Sacred Geographies: Cultural Memory of Geodynamic Processes in the Indian Subcontinent

Nigam Dave, Shrishti Kushwah, Ankita Srivastava, Dharmanshu Vaidya

Mountain landscapes of India are characterised by active tectonics, complex relief, and frequent exposure to earthquakes, landslides, and hydrological disasters. While geospatial hazard research models these processes using physical datasets, culturally grounded responses to long-term environmental instability remain less expolored within landscape-based analyses. This paper examines mythogenic mountain landscapes by analysing how Shakta sacred geographies function as spatial expressions of cultural memory associated with geodynamic processes.

The study focuses on selected Shakta-associated sacred sites situated in tectonically and geomorphically dynamic regions, including Kamakhya (Nilachal Hill, Assam), Jwalamukhi/Jwala Devi (Kangra Valley, Himachal Pradesh), Naina Devi and Chintpurni (Shivalik foothills, Himachal Pradesh), and Jayanti at Nartiang (Jaintia Hills, Meghalaya). Using GIS-based spatial profiling, site locations are analyzed in relation to relief, drainage corridors, and regional deformation zones. We also comparatively interpret recurring mythic motifs and ritual-temporal practices.

The analysis reveals patterned concentrations of sacred sites along mountain–plain transitions and structurally complex landscapes associated with environmental volatility. By situating landscape-scale patterning rather than site-specific belief, the study invites cross-disciplinary discussion on the role of geomythology in geoheritage interpretation and risk awareness. Recognising such mythogenic landscapes suggests culturally grounded perspectives for disaster-risk communication in regions facing increasing multi-hazard pressures.

How to cite: Dave, N., Kushwah, S., Srivastava, A., and Vaidya, D.: Mythogenic Mountain Landscapes and Shakta Sacred Geographies: Cultural Memory of Geodynamic Processes in the Indian Subcontinent, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2312, https://doi.org/10.5194/egusphere-egu26-2312, 2026.

Coastal cities in fragile and conflict-affected states face unprecedented challenges in maintaining infrastructure and protecting ecosystems. In Sudan, Port Sudan has recently emerged as the temporary administrative capital, experiencing rapid urban pressure alongside heightened climate vulnerability. This research evaluates the integration of Nature-based Coastal Solutions (NBCS), such as coral reef and mangrove preservation, into the city’s urban recovery framework. Utilizing GIS and satellite-based geoscience monitoring, the study assesses the current state of coastal assets and their protective capacity. A major barrier to implementing these solutions is the financing gap and high perceived risk in fragile economies. This study explores innovative financial frameworks, specifically the role of Development Finance Institutions (DFIs) in providing 'patient capital' and de-risking investments for sustainable coastal infrastructure. By combining interdisciplinary financial modeling with environmental assessment, the research proposes a strategic roadmap for financing resilient coastal protection. The findings demonstrate that NBCS can significantly reduce infrastructure restoration costs while serving as a vital catalyst for long-term economic stability and post-conflict recovery.

Final results, including a detailed comparative cost-benefit analysis and quantified financial projections, will be presented at the conference. This will provide a rigorous evidence-based framework for integrating Nature-based Solutions into Port Sudan’s post-conflict urban recovery.

How to cite: Ahmed, M.: Resilient Recovery: Financing Nature-based Coastal Solutions for Port Sudan’s Urban Infrastructure., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2636, https://doi.org/10.5194/egusphere-egu26-2636, 2026.

Strengthening water resilience in Europe requires the widespread adoption of Nature-Based Solutions (NbS) that are easily understood, trusted and supported by citizens and local stakeholders. This study focuses on the development of an Augmented Reality (AR) engagement system designed to communicate how different NbSs function in real-world scenarios and address water-related challenges. The AR experiences were co-created with local communities through dedicated focus groups, co-design workshops and structured discussions with key stakeholders, ensuring that the content reflects local priorities and practical needs at each pilot location.

The AR system brings together a set of NbS demonstrations into a unified series of interactive experiences. These include: (i) soil restoration and small-scale water retention measures in dry island landscapes that can reduce runoff, prevent erosion and enhance soil water storage for agricultural resilience; (ii) green walls that can treat greywater within a public building, enabling its safe reuse for non-potable applications such as toilet flushing; (iii) urban NbSs (including pocket forests, bioswales, permeable surfaces and soil improvement) that can mitigate flooding, reduce urban heat stress, and enhance environmental quality; and (iv) hydroponic wall systems that support urban gardening by combining seasonal planting, traditional knowledge and water-efficient practices.

The AR campaigns integrate maps, 3D models, photographs and explanatory narratives to guide users through each process step-by-step (e.g. users can follow the flow of greywater through a treatment system or observe the gradual transformation of degraded land as NbS are applied). By making otherwise invisible processes tangible and spatially explicit, the AR mobile application enhances understanding of how NbS improve water availability, reduce flood risks, support local food production and contribute to healthier and more resilient living environments.

The next phase of the work focuses on real-world validation across various pilot areas, involving diverse user groups (including residents, farmers, students, local authorities, and planners) to interact with the AR experiences on site and obtain their feedback to refine content clarity, usability and relevance for local planning processes and everyday decision-making.

The use of the AR mobile application demonstrates how visual storytelling combined with participatory design and field-based feedback can enhance awareness, build trust and support the mainstreaming of NbSs, contributing to strengthened water resilience across Mediterranean and broader European contexts.

 Acknowledgement:

This research has been funded by European Union’s Horizon Europe research and innovation programme under CARDIMED project (Grant Agreement No. 101112731) (Climate Adaptation and Resilience Demonstrated in the MEDiterranean region).

How to cite: Katika, T., Koukoudis, K., Touramanis, A., Michalis, P., and Amditis, A.: From Co-Design to Mainstreaming: Using Augmented Reality to Communicate Nature-Based Solutions for Water Resilience, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3848, https://doi.org/10.5194/egusphere-egu26-3848, 2026.

How to cite: Naskou, K., Katika, T., Touramanis, A., Koukoudis, K., and Amditis, A.: Immersive Citizen Engagement for Climate-Resilient Rural–Urban Interfaces, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3947, https://doi.org/10.5194/egusphere-egu26-3947, 2026.

Mountain districts within biodiversity hotspots often experience increasing ecological pressure despite retaining extensive forest cover. In the Western Ghats of India, Idukki district has undergone rapid tourism expansion, infrastructure development, and land-use reconfiguration over the past decade. This study assesses how changes in urban nature accessibility and population demand influence ecosystem service distribution and habitat vulnerability using the InVEST modelling framework. Urban Nature Access and balance indicators accessibility, per-capita balance, and total population balance were evaluated alongside a Habitat Risk Assessment for 2011 and 2025. The results indicate a growing spatial mismatch between population demand and accessible natural spaces, with strongly negative urban nature balance values expanding across central and southern Idukki by 2025. Accessibility and population pressure have become increasingly concentrated along valley floors, plantation belts, and transport corridors, while large forested areas remain functionally inaccessible. Habitat Risk Assessment results show that human-modified land-cover classes experience disproportionately higher risk, with built-up areas exhibiting the highest mean risk (R̄ = 0.42), followed by plantations (R̄ = 0.38) and croplands (R̄ = 0.34). Deciduous forests display lower vulnerability (R̄ = 0.22), and water bodies remain largely unaffected (R̄ = 0.05). More than one-third of built-up and plantation landscapes fall within medium to high habitat risk categories. High-risk zones identified by the model spatially coincide with landslide-prone regions that experienced repeated slope failures during extreme monsoon years (2018–2020), particularly in tourism-intensive areas such as Munnar, Adimali, and Peermade. These patterns indicate that ecological vulnerability in Idukki is driven less by absolute forest loss than by accessibility-induced concentration of human activities within steep, geophysically fragile landscapes. The findings emphasize the importance of integrating accessibility-aware ecosystem service assessments with hazard-sensitive nature-based land-use planning to reduce ecological degradation and disaster risk while supporting sustainable tourism and development in the Western Ghats.

How to cite: Jalaja, D. and Singh, S.: Accessibility-driven habitat vulnerability in the tropical mountain landscape of Idukki district, India, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5175, https://doi.org/10.5194/egusphere-egu26-5175, 2026.

As climate change impacts intensify, cities and regions are increasingly required to address adaptation and mitigation in parallel. In practice, however, these two dimensions are often planned and implemented separately, leading to missed co-benefits or unintended trade-offs. Thus, there is a growing need for traceable and operational methods capable of revealing, assessing, and integrating the interdependencies between adaptation and mitigation across sectors and spatial scales. To address this gap, this paper introduces the Indicator Service Framework (ISF), produced in the context of the ClimEmpower project (EU Horizon 2020) This methodological approach translates climate indicators into actionable insights, bridging the two fields of study to improve spatial analysis and local-to-regional decision-making.

The ISF operationalizes climate science by translating robust climate indicators into actionable policy insights. Its design is deliberately anchored in three core principles: multi-scale applicability, ensuring relevance from local to regional levels; data-agnostic design, allowing compatibility with any data source derived from hazard, exposure, and vulnerability assessments; and explicitness of decision logic. A central element of the ISF is the focus on identifying the most appropriate indicators for specific policy objectives, clearly establishing their relationship to the underlying climate risks and local conditions.

The framework employs a streamlined two-step process: first, indicator values are rigorously classified according to their scientific meaning,or against a defined benchmark (e.g., a European average or median value), which subsequently establishes the threshold for policy recommendations; second, they are standardized into harmonized classes. This standardization is crucial, as it enables systematic comparability across regions and facilitates the mapping of results to tailored recommendations. This mechanism is key to identifying concrete opportunities for co-benefits, such as mobility policies that simultaneously reduce emissions and enhance urban thermal comfort.

By structuring a clear pathway from climate data to policy decisions, the ISF functions as more than just a tool; it provides a clear strategic "reading frame" upon which climate actions can be anchored. This approach ensures that the resulting recommendations are systematically adapted to foster the overarching objective of 'climate resilient development' (IPCC 2022). The framework offers a practical contribution to integrated climate governance, enhancing stakeholder awareness and supporting more coherent, resilient, and sustainable strategies under conditions of multi-sectoral complexity.

How to cite: Murano, I., D'Angelo, G., Pavone, V., Del Prete, P., and Zuccaro, G.: An Indicator Service Framework for assessing and integrating climate adaptation–mitigation interdependencies across spatial scales, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5739, https://doi.org/10.5194/egusphere-egu26-5739, 2026.

 In arid and semi-arid landscapes like many areas in Morocco, addressing water scarcity requires innovative nature-based solutions (NbS). Fog and Low Stratus (FLS) clouds constitute a major atmospheric feature in Morocco, simultaneously representing a significant hazard for air, maritime, and road transportation and a valuable nature-based water resource for arid and semi-arid ecosystems through fog-water harvesting. However, effective implementation of such NbS depends on precise identification of viable locations and optimal collection periods. In a country characterized by strong climatic heterogeneity and limited ground-based observations, satellite remote sensing provides a critical means for assessing the spatial and temporal availability of this underutilized water source under current and future climate variability. This study introduces a novel nighttime FLS detection algorithm specifically designed for Morocco’s diverse climatic regimes, using only infrared observations from the Meteosat Second Generation (MSG) SEVIRI instrument. Hourly satellite data spanning 2020–2024 were processed to produce the first high-resolution, national-scale climatology of FLS occurrence over Morocco. Designed for the region's heterogeneous climates, the tool provides essential monitoring for assessing NbS potential. The algorithm was systematically validated using coincident hourly SYNOP observations from the Moroccan Directorate General of Meteorology network. Validation results demonstrate reliable performance, with a probability of detection exceeding 54%, a false alarm ratio close to 45%, and a frequency bias generally within 1.4. The resulting climatology reveals two major coastal hotspots of persistent FLS occurrence along Morocco’s Atlantic façade, in the Northwest and Southwest, both exhibiting pronounced seasonal and diurnal cycles. These regions coincide with areas of high potential for fog-water harvesting, offering a climate-resilient, nature-based solution to enhance water availability in water-stressed environments. These findings directly inform hydrological planning by pinpointing areas where fog harvesting projects are most likely to be effective and resilient. By providing spatially explicit and operationally robust information on FLS occurrence, this study supports the integration of satellite-based monitoring into the planning and upscaling of fog-water harvesting systems. The results contribute to broader NbS strategies aimed at improving water security, supporting ecosystem services, and strengthening climate adaptation in arid and semi-arid landscapes.

How to cite: Mouhtadi, A., Bari, D., and Mordane, S.: A Satellite-Based Climatology of Fog and Low Stratus to Support Nature-Based Water Harvesting in Arid Areas of Morocco, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6983, https://doi.org/10.5194/egusphere-egu26-6983, 2026.

The Atlantic Meridional Overturning Circulation (AMOC) is a key component of the Earth’s climate system, and has been suggested to have multiple stable states. Critical slowing down (CSD) can detect stability changes in Earth system components, and has been found in sea-surface temperature (SST) based fingerprints of the AMOC. Here, we look for CSD in historical simulations from 27 models from the sixth Climate Model Intercomparison Project (CMIP6). We calculate three different CSD indicators for the AMOC streamfunction strengths at 26.5°N and 35°N, as well as for a previously suggested SST-based AMOC index (ASSTI) based on averaging SSTs in the subpolar gyre region. No model shows CSD in the ASSTI, which is in marked disagreement with the real-world. This lack of CSD is reflected in the AMOC streamfunctions in most models, although individual ensemble members in some models do show signs of CSD even under a conservative significance calculation. We thus conclude that: 1) The historical AMOC in CMIP6 models is not losing stability, 2) studies of AMOC stability must consider an ensemble of realisations, 3) no other physical process in the 1850-2014 period causes signs of CSD in North-Atlantic SSTs, and thus the CSD in the observed ASSTI is likely a sign of a change in the AMOC. This final result suggests that observed changes in the ASSTI could indicate a loss of stability in the real-world AMOC.

How to cite: Ben Yami, M., Blaschke, L., Bathiany, S., and Boers, N.: No critical slowing down in the Atlantic Overturning Circulation in historical CMIP6 simulations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7797, https://doi.org/10.5194/egusphere-egu26-7797, 2026.

Temporary storage areas (TSAs) are a nature-based solution for attenuating flood peaks through the temporary detention of floodwaters in small (up to 10,000 m³) storage ponds on hillslopes or floodplains. Despite their increasing prevalence as part of Natural Flood Management (NFM) schemes in the UK, empirical evidence demonstrating their capability to mitigate flooding at catchment scales is limited. Addressing this evidence gap is a key priority for informing future flood risk management policies.

In this study, we intensively monitored a prominent NFM scheme in the Littlestock Brook, a lowland rural sub-catchment (6.4 km²) of the River Evenlode in England. Ten TSAs providing a combined 25,000 m³ of flood storage were implemented between 2018 and 2020 to protect a flood-prone settlement. Measurements of river discharge (5 min), TSA stored volume (5 min), and precipitation (10 min) enabled the filling and drainage dynamics of individual TSAs to be quantified. The monitoring period (2019-2021) captured several notable storm events, including one with an estimated return period of 1 in 37 years. 

To quantify the aggregated impact of multiple TSAs on flood hydrographs at the catchment scale, observed TSA inflows and river discharge were used within a time-of-travel based hydrograph reconstruction approach to enable the estimation of downstream discharge in the absence of TSAs. Comparison of observed (with TSAs) and reconstructed (without TSAs) hydrographs indicate a 23% reduction in peak discharge for a 1 in 16-year return period storm. Furthermore, analysis of individual TSAs revealed substantial variation in storage utilisation and drainage during and after storms. These results provide quantitative evidence of how TSAs function both individually and in combination. The potential effectiveness of TSAs as a sustainable Natural Flood Management intervention will be discussed.

How to cite: Bishop, J., Old, G., Rameshwaran, P., Wade, A., Robotham, J., Gasca-Tucker, D., Berkeley, A., Old, J., and McKnight, D.: The effectiveness of Temporary Storage Areas for Natural Flood Management: Empirical evidence from a lowland catchment, UK, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8304, https://doi.org/10.5194/egusphere-egu26-8304, 2026.

Deforestation-Driven Surface Warming and Heat Exposure in a Tropical Dry Forest District

Deforestation is widely understood as an important driver of local-scale climate warming in tropical regions, yet its consequences for human heat exposure and associated health risks remain poorly quantified at fine spatial scales. Forest cover regulates land surface temperature through canopy shading and evapotranspiration, suggesting that forest loss may amplify near-surface warming and intensify heat stress beyond background climate change. While global and regional studies have documented warming associated with deforestation, most analyses are conducted at coarse spatial scales and offer limited insight into district-level impacts relevant for human exposure. This gap is particularly evident in tropical dry deciduous forest regions, which experience pronounced seasonal heat stress and support populations heavily dependent on outdoor labor. In India, this type of landscape is widespread, yet fine-resolution assessments linking forest-cover change to heat exposure remain scarce.

This study proposes a district-level investigation of deforestation-driven warming and heat exposure in a district of Jharkhand, which is an ecologically stressed dry tropical forest region characterized by forest degradation and extreme summer temperatures. Forest-cover change since 2000 is quantified using Landsat-based Hansen Global Forest Change data, while land surface temperature patterns are examined using MODIS daytime LST observations. Hourly temperature and humidity fields from ERA5 reanalysis are used to reconstruct diurnal heat exposure and derive heat-stress indicators relevant to outdoor working conditions. Population-weighted exposure metrics and established temperature–health response functions from global burden datasets are employed to explore potential implications for heat-related mortality and losses in safe working hours.

By integrating high-resolution forest, climate, and population datasets, this work aims to isolate the contribution of local forest loss to heat exposure beyond broader regional warming trends. The analysis is expected to provide early evidence of how deforestation can intensify heat risks in vulnerable rural districts, with direct relevance for heat-adaptation planning, forest conservation priorities, and occupational health policies. These insights can inform district-level climate action plans, guide nature-based cooling strategies, and also support targeted interventions to reduce heat exposure among outdoor workers and farmers in tropical dry forest regions.

How to cite: Rani, S. and Sen, S.: Deforestation-Driven Surface Warming and Heat Exposure in a Tropical Dry Forest District., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9676, https://doi.org/10.5194/egusphere-egu26-9676, 2026.

Accurate simulation of irrigation water use is essential for quantifying human impacts on the global water cycle. Given that continuous large-scale in situ monitoring of irrigation is scarce, the fidelity of irrigation estimates relies heavily on how models represent soil-moisture deficits and management targets. In many global hydrological models (e.g., H08), irrigation demand is commonly computed using a soil-moisture deficit approach: water is applied to refill the soil when moisture levels fall below a prescribed target. However, this target is typically implemented as a static, empirically specified parameter. While computationally efficient, this practice introduces substantial uncertainty into simulated irrigation water use.

Here, we develop a satellite-based framework that utilizes observed surface soil moisture to constrain irrigation demand in hydrological models. We first construct a day-of-year climatology of satellite-derived surface soil moisture to capture multi-year mean irrigation conditions and management requirements. Subsequently, we employ a vertical extrapolation strategy to translate satellite-derived surface targets into a root-zone proxy compatible with the H08 model. We validate this strategy in non-irrigated regions before applying it to irrigated areas to enable dynamic, observation-constrained irrigation targets. Preliminary diagnostics indicate that this framework offers a practical pathway for integrating satellite soil-moisture data into H08, improving the spatial realism of irrigation demand and facilitating more consistent evaluations against independent benchmarks.

How to cite: Huang, X., He, Q., Hanasaki, N., and Oki, T.: Constraining irrigation simulation in Global Hydrological Model H08 using satellite-derived dynamic targets, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16078, https://doi.org/10.5194/egusphere-egu26-16078, 2026.

Virtual water trade (VWT) redistributes water embodied in agricultural commodities across borders and thereby shapes global interdependence between water resources and food security. Recent studies have increasingly used integrated assessment models (IAMs)—including GCAM, a partial-equilibrium IAM—to project future agricultural production and trade balances under future climate and socio-economic change and to infer virtual water transfer flows(e.g., Graham et al., 2020). However, such approaches assume that commodities are traded in a single global markets, making it difficult to explicitly quantify bilateral exporter–importer dependency structures.
In this study, we develop a scenario-based framework to estimate bilateral virtual water trade of rice and wheat toward 2100 by combining projections of harvested area (land-use), climate-driven yield changes, and population dynamics with an extrapolation of current trade structures. Using baseline bilateral trade matrices from FAOSTAT, we assume that (i) exporter-specific allocation to destination countries and (ii) national export-to-production ratios remain fixed, and we scale bilateral trade volumes in accordance with scenario-driven changes in production and demand. We then compute bilateral VWT by linking projected crop flows with crop- and location-specific water-use coefficients. The analysis focuses on SSP2 as the primary scenario, with additional SSP comparison(SSP126 and SSP585). This framework enables assessment of how future VWT magnitude and bilateral dependency patterns may evolve differently between rice—characterized by relatively thin international markets—and wheat, which is traded in thicker global markets, providing insights for water–food security assessment under future climate and socio-economic change.

How to cite: Tsuda, K., Sano, T., Oki, T., and Iizumi, T.: Projecting bilateral virtual water trade of rice and wheat toward 2100 under different SSP scenarios, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16183, https://doi.org/10.5194/egusphere-egu26-16183, 2026.

Increasing temperatures create more challenges for outdoor elite sports, particularly high-intensity tournaments such as the Australian Open, where players frequently experience high thermal stress. This study investigates the impact of environmental heat stress on professional tennis performance using high-resolution data from professional tennis matches with environmental performance diagnostics. To quantify these impacts, ATP and WTA singles matches played at various Australian Open tournaments have been analysed in conjunction with ERA5-Land reanalysis data averaged per hour, covering air temperature, relative humidity, global radiation, and wind speed. Heat stress was computed using the Wet Bulb Globe Temperature index and categorised into heat danger levels according to the heat danger classification of Sports Medicine Australia. A hypothesis-driven, uncertainty-aware statistical framework was employed, utilising robust non-parametric tests, trend analyses, and Spearman rank correlations to evaluate the sensitivity of key performance metrics to escalating levels of heat stress. Overall, the results indicate that severe heat stress conditions negatively affect the efficiency of serve and return, the number of unforced errors, the level of performance variability, and the length of a match in ATP and WTA events. More specifically, aggressive serve-related variables, such as aces, demonstrate a partial level of resilience in severe heat, while rally complexity, shot variety, and return length decrease with increased levels of heat stress. When analysed by set status, the results further suggest that while one of the most elite players controls their playstyle in severe heat conditions, the lower-seeded players take more risks and tend to make errors. Taken together, these findings provide large-scale empirical evidence of the impacts of environmental stress during the Australian Open tournament games. In light of these findings, the Australian Open tournament should adjust its schedule to prioritise tennis players’ health, and future tournaments should be scheduled more precisely according to reports from climate scientists and data-informed schedules.

How to cite: Kahraman, G., Turp, M. T., and An, N.: Heat Stress Impacts on Elite Tennis Performance: Evidence from the Australian Open, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16816, https://doi.org/10.5194/egusphere-egu26-16816, 2026.

Drought remains a pervasive environmental and socio-economic challenge across developing countries, with rural and semi-arid regions such as South Africa’s particularly vulnerable. In recent decades, climate variability has exacerbated the frequency, severity, and duration of droughts, prompting an expanding body of literature on resilience and adaptation. Traditional monitoring tools such as the Standardized Precipitation Index, Standardized Streamflow Index, and NDVI provide valuable biophysical insights but often fail to capture the socio-economic dimensions that shape community vulnerability and response. This review explores the evolution and application of the socio-ecological systems (SES) framework in drought resilience research within developing contexts. The SES approach offers a holistic lens to understand the complex interplay between environmental stressors, livelihoods, governance, and social systems. Emerging literature highlights the growing use of SES yet also reveals persistent gaps including weak integration between quantitative climate data and qualitative social insights, limited longitudinal studies, and inadequate incorporation of local knowledge. Drawing on studies from sub-Saharan Africa and other Global South regions, this review synthesizes key trends, methodological advancements, and research gaps in SES-informed drought resilience. It underscores the need for interdisciplinary, participatory, and context-sensitive approaches to support equitable and sustainable adaptation strategies aligned with global frameworks such as SDG 13 and the Sendai Framework.

How to cite: Mothapo, K., Mathivha, F., Chikoore, H., and Krueger, E.: Integrating drought indices and socio-ecological theory to analyze long-term drought impacts: A review of South Africa’s rural communities., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16979, https://doi.org/10.5194/egusphere-egu26-16979, 2026.

Responding to the challenges of a changing climate requires information that is relevant and actionable at the local scale where adaptation actions take place. To address these needs within Denmark, Klimaatlas, the Danish National Climate Atlas, was developed to provide information to ministries, regional authorities, businesses and citizens about climate change in Denmark.  Here we present the lessons learnt since the inception of the project in 2018, with a focus on those that are relevant to the development of similar tools in other regions. We will examine issues around the conception and setup of the climate service, particularly the need to identify users, work with champions and set limits. Communication is a critical aspect of such a service and we will discuss our approach of communicating on multiple levels, and taking up the challenge of uncertainty. Updatability, maintenance and operationalisation are also key, and the merits of the “rolling-releases” model used by Klimaatlas will be discussed, together with our efforts to open our codebase via the KAPy project. Finally, we discuss issues around future maintenance and possible expansions of Klimaatlas, including the use of convection permitting simulations, incorporation of compound events, updates between IPCC cycles and extensions to new sectors.

How to cite: Payne, M. R.: Lessons in climate service development from Klimaatlas, the Danish National Climate Atlas., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17146, https://doi.org/10.5194/egusphere-egu26-17146, 2026.

The rising temperatures and intensification of global heat hazards have evolved beyond occasional or seasonal heatwaves into a frequent state of chronic heat stress, amplifying both the duration and impact of extreme heat events. Driven by rising temperatures, humidity, rapid industrialization, and urbanization, the South Asian region, specifically India, faces escalating vulnerability to this compound hazard, which threatens public health, livelihoods, economic productivity, ecosystem balance, and overall quality of life.
India’s institutional response began with Ahmedabad’s pioneering 2013 Heat Action Plan (HAP), which catalysed the adoption of city- and state-level HAPs nationwide. To understand this evolution, 'content analysis' was conducted for 40 Heat Action Plans of 17 Indian states, available officially and publicly, founded against the National Disaster Management Authority’s (NDMA) 2019 guidelines and a global standards study from the WHO and UNDRR. The 23 heatwave-prone states were identified since 2013, only 18 currently have an official HAP.

This review evaluates document structure, its regional contextualization, accessibility to data, and institutional framework. While the NDMA’s (2019) heatwave framework has enabled widespread adoption, it is heatwave-centric and would benefit from explicitly incorporating heat stress through a nationally identified temperature–humidity index, as experimentally presented by IMD in 2023. Although the NOAA Heat Index is frequently cited in HAP documents, it is not suited to Indian conditions, as it does not reliably capture the extreme temperature–humidity regimes prevalent across the country. Furthermore, less than 50% of HAPs include localized vulnerability assessments, which should ideally contextualize physiological and social intricacies, regionally.

Additionally funding ambiguity is another persistent challenge, with most plans lacking identified financial sources or budgetary commitments. Communication gaps are evident, as less than 10% of HAPs provide materials in regional languages, constraining access to vulnerable populations in terms of educational limitation. Although, Ahmedabad’s evolving model remains the most comprehensive in this context. Notably, over 35% of HAPs fail to address land-use land-cover change, urban development plans, or localized climate-resilient design, despite strong links between the built environment and rising heat exposure. Data limitations, fragmented institutional accountability, and the lack of regional context with multi-sector actionability further weaken adaptive governance.
Altogether, these findings highlight the urgent need to move from fragmented, reactive heat responses toward anticipatory, multi-sectoral resilience planning. While the efficacy of HAPs depends on regional contextuality, this diversity must be supported by a replicable national framework guide that acknowledges heat stress while enabling inter-regional comparability. HAPs are primarily action-oriented instruments, this should reflect in the accessibility through local language translations, simplified formats with infographic tools, alongside comprehensive technical format that addresses meteorological services, health surveillance, funding mechanisms, and urban planning and design.

Resilience shouldn’t wait for the next disaster. The global shift toward proactive disaster risk management and the legacy of Ahmedabad’s 2010 heat-related mortality should motivate preparedness over response. Institutionalizing and updating HAPs primarily across all heatwave-prone states followed by the rest is central to embedding preparedness within India’s climate governance and recognizing heat as a structural climate–development challenge, rather than a seasonal hazard.

How to cite: Deshpande, S. and Mukherjee, M.: Gaps, Challenges, and Priorities for Future Adaptation of Heat Action Plans in India, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18217, https://doi.org/10.5194/egusphere-egu26-18217, 2026.

In the context of escalating global population, rapid economic development, and ongoing climate change, water resource management is confronted with a multitude of challenges. The North China Plain (NCP), as the economic powerhouse of China, is facing a multifaceted set of water-related issues, including inefficient water use under persistent scarcity, complex virtual water trade flows, and the increasing pressure on allocating water resource among cities through water diversion projects. Traditional water resource models often overlook the two-way feedbacks between water supply sources and demand sectors, therefore may not adequately represent the real-world water resilience dynamics. To address these challenges, this study constructs a System Dynamic (SD) model in NCP, building on water supply and demand statistics from local governmental reports. Different from previous SD-based water models for this region, we explicitly consider the roles of different water supply sources and municipal emergency water reserves. This provides a unique advantage for assessing urban water system resilience under extreme climate conditions.  In this presentation, we will first show the validation of our model in the historical period (2000-2020) compared to water agency statistics. We will also illustrate how the interactions between each urban water system components may change under different future climate scenarios. By investigating the  dynamic feedbacks between the natural and anthropogenic water cycles, our model is set to provide a scientific reference for governments to plan flexible and adaptive water resource management strategies.

Key word: Water Management; System Dynamic Model; North China Plain.

How to cite: Junkun, L., Qing, H., Xizhu, H., Hui, L., and Taikan, O.: A System Dynamics Model to Assess Water Resilience in the North China Plain, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18945, https://doi.org/10.5194/egusphere-egu26-18945, 2026.

The Challenge: 

Establishing a viable and systematic approach to measure the volume of stormwater runoff that can be captured to replenish aquifers and enhance climate resilience. Droughts, floods, erosion, heat domes, and crop failures are interconnected issues related to water, food, climate, and economics. Scaling up science-based methods across large areas presents challenges. 

Overview: 

Water is a common thread in climate change manifestation. Anthropological land use changes have transformed hydrology in various regions. Opportunities exist to integrate stormwater capture into water and climate management. It is important to consider rainwater as a valuable resource rather than something that is discarded. Conventional infrastructure drains rainwater excessively from agricultural, forested, and urban lands, wasting resources and threatening ecosystem stability and biodiversity. 

Solutions: 

Solution explores stakeholder-supported volumetric stormwater capture projects to deliver net positive water resource benefits, enhance climate resilience, and provide multiple co-benefits. This integration leads to financial returns and improved community satisfaction. A new water paradigm can help restore water resources on land. 

Case Study: In Slovakia, a new water paradigm approach has emerged over the last three decades, focusing on critical rainwater management. The authors discuss their experience in implementing past projects and their positive impact on the community, detailing the new Košice Region restoration plan in Slovakia. The new water paradigm approach attracted the attention of the UN Foresight Brief and UN Decade on Ecosystem Restoration and within the EU Climate-ADAPT framework.

How to cite: Kravčík, M. and Mulkerin, Z.: Climate Resiliency through Restoration using New Water Paradigm Methods, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19386, https://doi.org/10.5194/egusphere-egu26-19386, 2026.

Recent research indicates that the risk of collapse of the Atlantic Meridional Overturning Circulation (AMOC) this century is far higher than previously assumed, with severe consequences expected. Yet country-specific pathways and perspectives of impact remain underexplored. This study examines Sweden’s vulnerability to food insecurity under a plausible AMOC-collapse scenario, with particular attention to adaptive capacity assessed through social work and allied civil society agencies as the last line of defence in food security. Using a vulnerability framework (exposure, sensitivity, adaptive capacity), we conducted semi-structured interviews with 10 climate scientists, 10 agricultural scientists, and 10 Swedish social workers engaged in food support. Experts agreed that an AMOC collapse would substantially disrupt Sweden’s climate (colder, longer winters; greater variability), producing cascading effects on agriculture, imports, and availability that threaten national food supply. While climate and agricultural experts identified catastrophic risks, suggesting famine may be a plausible outcome, social workers reported minimal preparedness, no protocols or training for food-shortage response, and limited ability to respond to sudden increases in need, despite their central role in supporting vulnerable groups. Findings highlight both the scale of the threat and the absence of attention to AMOC-related risks within Swedish agriculture, food security, and welfare systems. The study maps an impact pathway, identifies the gaps in preparedness, as well as demonstrating a replicable approach for linking climate hazards to social-system readiness. The results suggest urgent action is needed to integrate AMOC-type risks into Swedish and international food-security policy, planning, and training in order to mitigate potentially catastrophic human impacts.

How to cite: Rost, S.: Sweden’s Food System Vulnerability to AMOC Collapse through Climate, Agricultural, and Social Work Perspectives, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21355, https://doi.org/10.5194/egusphere-egu26-21355, 2026.

Uncertainty and perceived lack of quantifiability in the evaluation of nature-based solution (NbS) benefits relating to non-market ecosystem services remains a barrier to the ready adoption of NbS as water resilience projects. We aim to bridge this gap for coastal and riverine NbS by creating a framework to improve inclusion of the entire range of ecosystem services provided by NbS in cost-benefit analysis of water resilience project alternatives. We have conducted a literature review of NbS and natural and nature-based feature (NNBF) literature and case studies to determine which ecosystem services are associated with wetlands, dunes and beaches, seagrass meadows, barrier islands, and forested ecosystems. Through the review, we have identified ecological and environmental, carbon capture, coastal land loss reduction, hazard risk reduction, socio-economic and cultural, and economic and financial services of each NbS type, along with the range of metrics currently used to evaluate project output of these benefits. We created a fully cited framework detailing the benefits and metrics for each NbS type, and implemented it in both a knowledge graph and interactive radial graph formats. The interactive radial graph provides support for human user exploration of the framework and cited literature and case studies. The knowledge graph will serve to support retrieval-augmented generative agent tools in the future. In future work, we will improve on the framework with inclusion of cost and limitation information, as well as a basic method for estimating market values of non-market benefits based on those of market benefits. 

How to cite: Noggle, R., Akter, D., Rahim, M. A., and Mostafiz, R. B.: A framework to facilitate inclusion of NbS ecosystem service benefits in cost-benefit analysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22058, https://doi.org/10.5194/egusphere-egu26-22058, 2026.

Rapid and accurate estimates of ground-motion intensity measures are critical for seismic hazard assessment and disaster response. Empirical ground-motion models provide fast predictions, but suffer from large uncertainties, especially in regions with sparse observations. Physics-based simulations offer physically consistent shaking intensity estimates but remain computationally prohibitive for real-time applications and large-scale scenario analyses. We present a machine-learning framework that predicts high-resolution ground-motion intensity maps conditioned on earthquake source parameters, combining physics-consistent predictions with near-instantaneous inference. The framework predicts a suite of intensity measures widely used in seismic hazard and earthquake-engineering studies -- including peak ground velocity (PGV), peak ground acceleration (PGA), and response spectra -- for arbitrary double-couple sources embedded in a realistic 3D medium, inherently capturing complex geological and topographic effects.

Our approach leverages two complementary training datasets obtained from waveform simulations: spatially sparse shaking intensity maps generated via reciprocity methods and spatially dense intensity maps from forward simulations. A conditioned U-Net is first pretrained on abundant spatially sparse maps to learn global spatial features, subsequently fine-tuned using a limited set of spatially dense maps. This staged training strategy significantly reduces training data requirements while maintaining high predictive accuracy. Applied to the San Francisco Bay Area and Wellington, New Zealand, the framework produces physics-consistent intensity maps with speedups of 6–7 orders of magnitude compared to traditional wave-propagation simulations. This enables scalable, near-instantaneous hazard assessment for both rapid disaster response and comprehensive scenario-based analyses.

How to cite: Ramadan, F., Nissen-Meyer, T., Koelemeijer, P., and Fry, B.: Near-Instantaneous Physics-Based Ground-Motion Maps Using Sparse-to-Dense Deep Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-734, https://doi.org/10.5194/egusphere-egu26-734, 2026.

The south of the Iberian Peninsula, particularly the Baetic System, is one of the most seismically active regions of the Iberian Peninsula. Its complex seismotectonic configuration causes recurrent moderate to strong earthquakes, posing a significant hazard to society and the built environment, requiring rapid and accurate post-event assessment of ground-motion intensity.  These high-risk areas coincide with densely populated areas of Murcia, such as Lorca, or the province of Almeria. In addition, population dynamics vary significantly between summer and winter, due to seasonal tourism and residential tourism, which increases vulnerability and the need for rapid and accurate assessments following an earthquake. To address this need, the Machine Learning Estimator for Ground Shaking Maps (MLESmap) was developed as a rapid-response framework that combines high-quality physics-based simulations with Machine Learning techniques to infer spatially distributed ground-motion intensity measures within seconds after earthquake initiation. Trained on a large ensemble of synthetic seismic scenarios, MLESmap provides near real-time predictions of ground-motion intensity fields, such as acceleration levels and shaking patterns.

Our methodology incorporates both offline and online phases in a comprehensive workflow. It begins with the generation of a synthetic training data set generated by the CyberShake platform. Then  predictor characteristics are extracted before the validation and learning stages. The result is a model that can be used for fast inference validated with start-of-art methodologies and available real data  .

To evaluate the influence of surface representation on model performance, synthetic simulations are carried out using both 1D and 3D seismic velocity models, allowing for a systematic comparison of their impact on training and prediction accuracy. In addition, different learning strategies are explored, as for example, multi-objective approaches that allow for the simultaneous estimation of multiple measures of ground motion intensity. These analyses quantify the influence of velocity model dimensionality and training strategy on the performance of MLESmap predictions

Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Spain, Italy, Iceland, Germany, Norway, France, Finland and Croatia under grant agreement No 101093038, ChEESE-2P, project PCI2022-134980-2 funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR)

How to cite: Blanco-Prieto, R., Zamora, N., Monterrubio-Velasco, M., and de la Puente, J.: From physics-based simulation to ground motion models using Machine-Learning Estimator for Ground Shaking Map, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1865, https://doi.org/10.5194/egusphere-egu26-1865, 2026.

Soil moisture dynamics play a critical role in slope stability, especially for rainfall-induced group-occurring landslides. With the growing availability of remote sensing–derived soil moisture products, there is increasing potential to improve landslide susceptibility assessment. However, few studies have explicitly incorporated both the spatial and temporal dynamics of soil moisture into susceptibility modeling. This study introduces a novel framework that integrates a Residual-Sparse Autoencoder (ResSAE) with Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) algorithms to enhance landslide susceptibility prediction using remotely sensed soil moisture data. Spatio-temporal soil moisture information for the study area in Nanping, China, is obtained from three open-access datasets: SMCI1.0, ERA5-Land, and SMAP-L4. Results show that antecedent soil moisture features extracted by ResSAE substantially improve prediction accuracy. The influence of rainfall, antecedent period length, and dataset source is further evaluated. Further analysis reveals that antecedent soil moisture over the prior seven days captures most of the hydrological memory relevant for slope failure, while additional rainfall data contribute only marginal gains. Optimal performance is achieved with ERA5-Land for RF, SMAP-L4 for SVM, and SMCI1.0 for ANN.Overall, the study highlights the importance of incorporating spatio-temporal soil moisture into susceptibility assessment. The proposed approach enables efficient and cost-effective predictions, supporting near-real-time applications and offering potential to strengthen regional to global rainfall-induced landslide prevention and mitigation strategies.

How to cite: An, N. and Xie, E.: Enhancing landslide hazard assessment by considering spatio-temporal soil moisture dynamics, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3068, https://doi.org/10.5194/egusphere-egu26-3068, 2026.

Urban flooding is emerging as an increasingly severe global challenge due to climate change and urbanization. Although machine learning offers numerous solutions for urban flood forecasting, its application remains constrained. Existing research remains constrained by the scarcity of traditional hydrological monitoring data, and the absence of systematic comparisons across multiple models creates uncertainty when selecting the most suitable algorithms and features, making the decision-making mechanisms for selecting the most suitable algorithms and features remains unclear. To address these challenges, social media data was adopted as the sole basis in this study to evaluate and compare the performance of seven typical machine learning algorithms in urban flood forecasting. The Shapley Additive exPlanations (SHAP) framework was established, investigating the adaptability of the selected algorithms based on a multidimensional feature system while elucidating the decision-making mechanisms for selecting the most suitable algorithms and features. The results suggest that: (1) Social media data can serve as the sole source for precise urban flood identification, overcoming the real-time and spatial coverage limitations of traditional methods. (2) Different machine learning models show significant performance heterogeneity; reliance on a single model risks systematic bias, whereas ensemble tree models demonstrate superior predictive performance. (3) Feature importance is highly model-dependent, exhibiting contextual sensitivity and interactive influence mechanisms. Therefore, feature engineering should be based on multi-model consensus, prioritizing features with significant differences such as natural characteristics and risk exposure.

How to cite: Miao, R., Huang, R., and Zheng, J.: Adaptability of Multiple Social Media Data Integrated Machine Learning Algorithms in Urban Flood Forecasting using the SHAP Framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3659, https://doi.org/10.5194/egusphere-egu26-3659, 2026.

Mining-induced geological hazards in the mountainous regions are often characterized by wide impact zones and complex subsurface structures, which pose significant challenges for the precise identification of landslides and the analysis of their formation mechanisms. To address this issue, we takes the Leji landslide in southwestern China  as a case study and integrates SBAS-InSAR technology, two-dimensional decomposition modeling, UAV photogrammetry, field geological investigation, and AMT sounding to establish a multidimensional “Space-Air-Ground-Subsurface” detection strategy. This multi-dimensional framework enables the systematic acquisition of both surface deformation and subsurface structural information of the Leji landslide, thereby elucidating its controlling factors and causative mechanisms. The results reveal that the central parts of Landslide I and Landslide II exhibit the most significant deformation, with surface displacement dominated by downslope subsidence. The maximum annual average subsidence rates range between –60 mm/y and –80 mm/y. The cumulative deformation zones retrieved by SBAS-InSAR closely coincide with the mining areas detected by AMT. Through data fusion, the boundary angles of the mining areas were determined as 77° in the upslope direction and 48° in the downslope direction along the dip, and 77° and 55° in the strike direction. Comprehensive analysis indicates that the Leji landslide is a Quaternary soil creep landslide formed under the combined influence of fault–fold structures, frequent heavy rainfall, and both open-pit and underground mining activities, and it remains in an active state. This study demonstrates that the “Space-Air-Ground-Subsurface” collaborative observation system effectively overcomes the limitations of single techniques in landslide mechanism research, providing a reliable technical pathway and scientific basis for understanding the development mechanisms and disaster risk mitigation of mining-induced landslides.

How to cite: Zhang, Y. and Zhu, Y.: Using SBAS-InSAR and Audio-Magnetotelluric Sounding to Characterize Two-Dimensional Deformation and Failure Mechanisms in Mining-Induced Landslides, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3666, https://doi.org/10.5194/egusphere-egu26-3666, 2026.

How to cite: Gatti, F., Perrone, N., Lehmann, F., and Fresca, S.: Shake Anywhere: a simulation-free AI-based earthquake ground motion generator for any source/any geology., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3994, https://doi.org/10.5194/egusphere-egu26-3994, 2026.

In recent years, the Yizhong River basin in Deqin County, Yunnan Province, has experienced frequent debris flow events, posing a significant threat to surrounding residential areas and infrastructure. This study aims to investigate the hydrodynamic characteristics and hazard risk of debris flows in this basin under extreme rainfall conditions, providing a scientific basis for disaster risk reduction and prevention. The research employed UAV aerial photogrammetry, field investigations, and numerical simulation techniques to construct a high-resolution 3D terrain model of the Yizhong River basin. Using the continuum mechanics method based on deep integration and embedded by Physics-Informed Neural Networks (PINNs), the movement processes of flood-type and landslide-type debris flows were simulated under two extreme rainfall frequencies: 1% and 0.5%. Simulation results reveal that the frequent initiation of debris flows in the Yizhong River basin is influenced by multiple factors, including topography, material source conditions, and rainfall intensity. Under the 1% rainfall frequency, both types of debris flows trigger slope instability along the channel, leading to the entrainment of additional source material and enlarging the affected area. At the 0.5% rainfall frequency, the drainage channels within Deqin County were completely overwhelmed, with major transport arteries largely blocked. Substantial volumes of debris flow material entered the Zhiqu River, with overflow even burying Deqin No. 1 Middle School. Risk assessments for single-channel debris flow in the Yizhong River basin revealed that at the 0.5% rainfall frequency, debris flows within the Yizhong River channel reached an extremely high risk level, highlighting the inadequate capacity of the existing protection measures. Consequently, urgent attention must be given to stabilising unstable slopes along both banks of the Yizhong River basin and constructing drainage and diversion facilities within Deqin County. Future regional disaster prevention and mitigation efforts should prioritise curbing the frequent occurrence of debris flow disasters in the Yizhong River basin at their source.

How to cite: Wang, W., Zhu, S., Zhang, Y., and Chen, C.: Dynamic characteristics and risk assessment of debris flow under extreme rainfall in Yizhong River Basin of Deqin, Yunnan Province, from numerical simulation to PINN model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4409, https://doi.org/10.5194/egusphere-egu26-4409, 2026.

The EuroHPC Center of Excellence for Exascale in Solid Earth (ChEESE CoE, 2018–2026, DOI: 10.3030/101093038) is preparing 11 community-driven, open-source codes to run optimally on large accelerated supercomputing infrastructures (Leonardo, LUMI, MareNostrum-5). The CoE works with flagship codes in different areas of geophysics (earthquakes, tsunamis, volcanoes, magnetohydrodynamics, geodynamics, and glacier modelling), focusing on performance, scalability, CI/CD on EuroHPC systems, and portability across current and emerging hardware architectures. During 2025, the CoE has been awarded more than 1 million node-hours on GPU-accelerated systems such as Leonardo (4 NVIDIA A100 per node), LUMI-G (8 AMD MI250X per node), and MareNostrum-5 (4 NVIDIA H100 per node). The resulting simulations and use cases are being stored in data lakes together with their metadata for use by the scientific community, for example to train AI models or to be accessed through the European Plate Observing System (EPOS). All codes and applications under the ChEESE umbrella are available in open GitLab/GitHub repositories and undergo an SQAaaS process to obtain software quality badges. In addition, the project aims to enable urgent supercomputing services for emergencies during high-impact events (earthquakes, tsunamis, and volcanoes), including the associated technical challenges and recommendations on access policies. This is done in collaboration with end users such as civil protection agencies in various European countries. For example, ChEESE researchers tested an urgent supercomputing service for earthquakes during the September 19th 2025 drill, in collaboration with the Mexican Seismological Service (SSM).

Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Spain, Italy, Iceland, Germany, Norway, France, Finland and Croatia under grant agreement No 101093038, ChEESE-2P, project PCI2022-134973-2 funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR.

How to cite: Folch, A.: ChEESE: the European Center of Excellence for supercomputing in geosciences, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5866, https://doi.org/10.5194/egusphere-egu26-5866, 2026.

Driven by global climate change, extreme weather events leading to short-duration heavy rainfall have emerged as a primary challenge for urban disaster prevention and resilience. Frequent and intense rainfall not only significantly increases the risk of urban pluvial flooding but also disrupts the stable operation of public infrastructure. Traditional drainage system designs often rely on static solutions that are inadequate for coping with the rapid intensity changes and high uncertainty of extreme rainfall, further exacerbating disaster risks in urban areas.

This study integrates advanced data analytics with machine learning to propose a rainfall and flood risk prediction system based on Self-Organizing Maps (SOM) and Long Short-Term Memory (LSTM). Leveraging Internet of Things (IoT) technology, the study incorporates high-resolution data (10-minute intervals) from flood-prone communities in Taipei City between 2015 and 2021. The multi-source dataset includes radar reflectivity, meteorological observations, sewer water level monitoring, and historical flood records to build a hydro-meteorological model with strong spatial and temporal representation. Preliminary results indicate that incorporating wind speed and direction data significantly enhances prediction accuracy and reduces uncertainty. Through SOM technology, the system performs refined classification of high-dimensional meteorological data, excelling in identifying extreme rainfall patterns. Combined with LSTM’s capability to capture temporal sequence characteristics, the system predicts rainfall and water level fluctuations. Furthermore, through a monitoring mechanism for sewer water level rise rates, integrating terrain and sewer spatial characteristics to provide localized, dynamic notifications and tailored response recommendations.

By combining AI-driven uncertainty analysis with real-time hydrological monitoring, this research strengthens flood forecasting capabilities under diverse wind field conditions, providing a science-based decision-support framework. The application of this model not only enhances the precision of community-scale flood prevention planning but also offers an adaptive regional warning strategy for urban climate adaptation. Ultimately, this system will effectively bolster urban disaster resilience and provide local governments with robust decision-support tools to achieve long-term sustainable development goals.

How to cite: Chen, L.-Y., Pan, T.-Y., Lai, J.-S., and Chang, M.-J.: Development of a Urban Flood Prediction Model Using SOM-LSTM: Integrating Environmental IoT and Sewer Water Level Rising Rates, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7239, https://doi.org/10.5194/egusphere-egu26-7239, 2026.

High-performance computing (HPC) plays a central role in advancing AI-based approaches for time-critical natural hazard applications, especially in regions where observational data are limited. In seismology, the scarcity of strong-motion records for large earthquakes poses a major challenge for the development of purely data-driven ground motion models. Here, we highlight the use of HPC to generate large, high-fidelity synthetic earthquake datasets specifically tailored for training machine-learning (ML) models for rapid ground motion forecasting in Southern Iceland.

Using the CyberShake workflow on HPC systems, we compute an unprecedented ensemble of approximately 100,000 physics-based earthquake scenarios, spanning magnitudes Mw 5.0–7.4, at 350 synthetic stations across the Southern Iceland Seismic Zone and the Reykjanes Peninsula Oblique Rift. Seismic wave propagation is simulated deterministically up to 2 Hz using three alternative Earth velocity models, allowing us to systematically investigate how subsurface velocity heterogeneity influences ground motion. By exploiting seismic reciprocity, the computational cost scales with the number of virtual recording sites rather than with the number of earthquakes, making it feasible to explore tens of thousands of rupture scenarios on Tier-0 HPC systems. The resulting simulations combine multiple velocity models, dense site coverage, and designed magnitude distributions, forming a comprehensive and carefully curated training dataset.

This large HPC-generated database is then used to train machine-learning surrogate models within the Machine Learning Estimator for Ground Shaking Maps (MLESmap) framework, including both tree-based ensembles and deep neural networks. Although these ML models provide near-instantaneous predictions of ground motion intensity measures during post-event response, their reliability ultimately depends on the quality, diversity, and physical realism of the underlying training data.

Our results show that HPC-driven simulation workflows can effectively close the data gap in regions with limited observations, delivering physically grounded datasets that support robust AI models for time-critical seismic hazard assessment. More broadly, this work underscores the role of HPC not only as a computational tool for modeling extreme events, but as a cornerstone of next-generation AI-driven systems for hazard forecasting and emergency response.

How to cite: Monterrubio-Velasco, M., Zamora, N., Blanco-Prieto, R., Riaño, A. C., Vázquez, F., Chapagain, B., and de la Puente, J.: HPC-enabled large-scale physics-based seismic simulations as training data for AI-driven ground motion forecasting in Southern Iceland, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7756, https://doi.org/10.5194/egusphere-egu26-7756, 2026.

Droughts are hydroclimatic anomalies driven by precipitation deficits and increased evapotranspiration, posing an escalating threat under a warming Mediterranean climate. Assessing drought risk remains challenging due to the complex interactions between biophysical conditions and human systems, as well as limitations in impact reporting. Moreover, drought impacts are highly heterogeneous across sectors, as different types of drought affect socio-environmental systems differently. In this context, we develop a spatio-temporal deep learning framework to predict sector-specific drought impacts and identify the environmental and climatic drivers of these impacts.

We combine two primary data sources: the European Drought Impact Database (EDID), which contains above 13,000 georeferenced drought impact reports spanning 1970 to 2023 and aggregated into four sectors (agriculture, ecosystem, energy, and socio-economic); and a set of physical drivers, including precipitation, temperature, drought indices, vegetation indices, and population density, derived from various sources for the period 2001–2021.

The prediction task is formulated as a spatio-temporal segmentation problem using a 3D U-Net architecture to capture dependencies in climate and environmental conditions over a one-year period. The preprocessing workflow harmonizes all variables to a spatial resolution of 0.25° and an 8-day time step. Seasonally varying predictors are transformed into anomalies, and all variables are normalized. Input samples are arranged as tensors with shape 36×48×16×16 (C×T×H×W), representing one year of conditions, while the target consists of a binary impact map (1×1×16×16) corresponding to the subsequent 8-day period. The training dataset is balanced through equal sampling of impact and no-impact cases. Consequently, the model learns to use one year of spatio-temporal context to predict drought-affected areas at the next time step.

Initial results for the agricultural sector indicate that traditional drought indices have limited predictive skill for drought impacts. A baseline evaluation of the Standardized Precipitation-Evapotranspiration Index (SPEI) across multiple thresholds shows that the 1-month SPEI achieves a PR-AUC of 0.13 and an ROC-AUC of 0.32 for the impact class over the 2018-2020 test period, with similarly low performance for the 3-, 6-, and 12-month variants. In contrast, preliminary model experiments demonstrate a substantial improvement over the baseline, achieving an F1 score of 0.43, a PR-AUC of 0.41, and a ROC-AUC of 0.71, despite remaining limitations in predictive performance.

These limitations are primarily attributed to noise and spatial uncertainty in the ground-truth labels, as EDID impacts are reported at coarse administrative units (NUTS3) and uniformly assigned to all grid cells within each region, constraining pixel-level learning. In addition, drought impacts are influenced by large-scale atmospheric circulation patterns and remote climate teleconnections (e.g., ENSO and NAO) that are not explicitly represented in the current feature set. Future work will address these limitations by incorporating large-scale circulation and teleconnection indicators, developing strategies to mitigate label noise, and extending the modelling framework to additional sectors. Once predictive performance is optimized, explainable AI methods based on Integrated Gradients will be applied to identify the most influential climatic and environmental drivers, enabling sector-specific interpretation of drought impact mechanisms and their temporal dynamics.

How to cite: Sapena, M., Papadopoulos, N., Athanasiou, G., Papoutsis, I., and Camps-Valls, G.: Predicting multi-sectoral drought impacts in the Mediterranean with spatio-temporal deep learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7822, https://doi.org/10.5194/egusphere-egu26-7822, 2026.

The GANANA project is an EU–India initiative that builds on three pillars: geohazards, weather and climate and life sciences, each linked to a EuroHPC Center of Excellence (CoE). In particular, the ChEESE-2P CoE pillar  advances the use of High-Performance Computing (HPC) for geophysical hazard assessment and risk mitigation. It harnesses flagship HPC codes to deliver integrated, physics-based and data-driven solutions for earthquakes, tsunamis, smoke dispersion, and cascading hazards, with a strong focus on urgent computing,operational readiness and rapid response. We present GANANA’s high-level framework and first results across three core geophysical hazard domains. For earthquakes, urgent computing workflows enable near-real-time ground-shaking simulations using physics-based solvers, supporting rapid impact assessment for civil protection. These workflows are complemented by Artificial Intelligence / ML techniques  for seismic data monitoring, where deep-learning pipelines automate event detection, phase picking, and magnitude estimation, and are tightly integrated with physics-based simulations to enhance robustness in data-scarce and tectonically complex regions. For tsunamis, GANANA extends established HPC workflows for rapid forecasting and high-resolution inundation mapping, triggered by seismic events, with particular emphasis on operational applicability and transferability to new coastal regions. 

The workflow focused on wildfire spread and smoke dispersion, aims to develop an integrated forecasting system for urgent computing applications built upon expertise on the development of HPC codes for Numerical Weather Prediction (NWP) and atmospheric dispersion models. A defining feature of GANANA is its structured, bidirectional exchange of codes, expertise, and operational practices between Europe and India, enabling the adaptation, validation, and deployment of advanced HPC technologies in diverse geographical and institutional contexts. 

A key aspect of the project is also the cascading hazard - framework. Preliminary demonstrators show that this exchange significantly improves model performance, interoperability, and time-to-solution, while simultaneously fostering capacity building and shared ownership of advanced HPC tools. GANANA thus illustrates how sustained international collaboration can transform mature exascale-ready codes into scalable, user-oriented systems for geophysical hazard forecasting and early warning.

How to cite: Zamora, N., Srivastava, N., Sánchez, C., Mingari, L., Folch, A., Macías, J., Monterrubio-Velasco, M., Diez-Ventura, G., Esquivel-Mendiola, L. I., Vázquez-Novoa, F., Badia, R. M., and de la Puente, J.: Physics-Based and AI-Driven HPC Workflows for Geophysical Hazards in GANANA project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7839, https://doi.org/10.5194/egusphere-egu26-7839, 2026.

Heavy precipitation is a major hazard associated with tropical cyclones, often causing substantial economic losses and casualties through secondary disasters such as floods, landslides, and debris flows. The southeastern coast of China is one of the region most severely impacted by tropical cyclones. Under the context of global warming, the risks posed by tropical cyclone precipitation are expected to increase further. Accurate simulation of tropical cyclone rainfall is crucial for assessing flood hazards and provides a scientific basis for regional disaster risk mitigation policies. In this study, based on MSWEP precipitation data and tropical cyclone track data, we developed a China-focused tropical cyclone precipitation simulation model using the XGBoost algorithm reconstructed the precipitation field of TCs from 2000~2020. First, based on the tropical cyclone best-track data provided by the China Meteorological Administration, a rainfall field was constructed as a collection of 100 km × 100 km grid cells, forming an approximately circular domain with a radius of about 1000 km centered on the tropical cyclone. Mean precipitation for each grid cell was then extracted from the MSWEP dataset. Fifteen predictor variables were selected, including cyclone center latitude and longitude, grid center latitude and longitude, distance and azimuth between grid center and cyclone center, elevation, slope, aspect, wind speed and direction, cyclone forward direction, distance to land, season, and whether the cyclone center was over land. Based in MSWEP data from 2000 to 2020, a model was trained to predict precipitation in each grid using XGBoost algorithm. Based on this model, a reconstructed dataset of tropical cyclone rainfall for 2000–2020 was generated and evaluated. The main results indicate that, for a 70:30 train-test split, the model achieved RMSE=173.768mm, MAE= 85.504mm, and R²=0.674, demonstrating good performance. The simulated data effectively reproduce the spatial distribution of total tropical cyclone precipitation. Comparison of precipitation distribution maps based on MSWEP and simulated data further confirms that the model captures the spatial characteristics of total tropical cyclone rainfall with reasonable accuracy.

How to cite: Tao, K. and Xu, W.: A Machine Learning-Based Tropical Cyclone Precipitation Simulation in China, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8510, https://doi.org/10.5194/egusphere-egu26-8510, 2026.

Hernandez, E.¹, Folch, A¹, Mingari, L.¹, Stramondo, S.², Trasatti, E.², Ganci, G.², Corradini, S.², Gonçalves, P.³, Brenot, H.⁴, Pacini, F. ³

• Geociencias Barcelona (GEO3BCN), CSIC, Barcelona, Spain
• Instituto Nazionale di Geofisica e Vulcanologia (INGV), Sezione di Bologna, Bologna, Italy
• Terradue, Roma, Italy
• Royal Belgian Institute for Space Aeronomy (BIRA), Brussels, Belgium

The ESA Geohazards Early Digital Twin Component (GET-it) project aims to deliver interactive, scenario-based tools for decision-making during geohazard crises. Within this framework, we present recent advancements in volcanic ash and gas dispersion modeling through the integration of satellite data assimilation into the FALL3D model. The main innovation consists of assimilating SEVIRI-derived SO₂ mass loading during the 2021 Cumbre Vieja eruption (La Palma) and volcanic ash during the 2018 Mount Etna eruption. These enhancements significantly improve the accuracy of quantitative forecasts of volcanic clouds, which are critical for aviation safety and public health.

The assimilation system implemented in FALL3D is based on the Local Ensemble Transform Kalman Filter (LETKF), an ensemble-based technique with localization designed to run efficiently on parallel computing platforms. The observation operator maps the model state to satellite retrievals, enabling sequential assimilation cycles. After each cycle, the corrected 3D concentration field initializes a new forecast, reducing uncertainty in cloud position and concentration. For La Palma, three assimilation steps were performed at 3-hour intervals using SEVIRI SO₂ retrievals, improving consistency with independent observations of cloud height.

To enable operational use, these simulations have been deployed on the Geohazards Thematic Exploitation Platform (TEP) by Terradue. The implementation leverages Common Workflow Language (CWL) workflows and Docker containers, ensuring reproducibility and scalability. The platform provides interactive visualization of eruption scenarios, including maps and time series, and allows users to modify key eruption source parameters (e.g., column height, intensity) through predefined scenarios (low, medium, high).

This work demonstrates the potential of combining Earth observation data with advanced numerical modeling in a cloud-based environment to deliver actionable information for crisis management. Future developments will focus on extending these capabilities to other geohazards and enhancing real-time operational readiness.

How to cite: Hernandez Plaza, E.: Advancing Volcanic Crisis Management through Satellite Data Assimilation in FALL3D within the ESA GET-it Digital Twin Framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9859, https://doi.org/10.5194/egusphere-egu26-9859, 2026.

Tsunami Early Warning Systems (TEWS) in the NEAM region (North-East Atlantic, the Mediterranean, and connected seas) operate under strict time constraints, particularly for near-field events where coastal impact may occur within a few minutes. In the NEAM region, operational chains typically use decision matrices and precomputed scenario databases. In Spain, the TEWS is operated by the Instituto Geográfico Nacional (IGN), and this work is carried out jointly with IGN to support operational decision-making. These established tools can be reinforced with rapid products that provide early indicators of coastal impact within some minutes or even seconds of the first source estimate. One example is the use of Faster-Than-Real-Time (FTRT) simulations, already implemented in the current system.

Here we present a workflow in which neural-network surrogates are trained on large sets of physics-based tsunami scenarios, enabling fast inference of coastal impact metrics. The Tsunami-HySEA code is used to generate large-scale simulation sets, providing the data required by models designed for near-instant inference on standard CPUs. The surrogates models learn to map solid Earth earthquake source descriptors (capturing some uncertainty in fault parameters) to warning-relevant coastal metrics, focusing on maximum wave height and first-arrival time at multiple sites. Once trained, the models deliver predictions within seconds, facilitating rapid updates as source estimates evolve. Model interpretability is assessed using SHAP values, confirming how each input influences the predictions. The results confirm that the patterns follow the physical principles of tsunami generation and propagation. In an operational workflow, model results are fed into an automated reporting layer that produces tables, maps and graphics for Civil Protection within seconds, enabling rapid situational updates as source estimates evolve.

We first report initial results for Atlantic sources affecting SW Spain. Approximately 250,000 HySEA simulations covering multiple Atlantic fault segments, focal mechanisms and magnitudes were used to train models. The results for forecast points along the Huelva–Cádiz coast show good agreement with observed patterns of maximum wave height and meet operational speed requirements, with errors remaining within the acceptable range for TEWS procedures. We then describe the extension of the methodology to the Western Mediterranean, covering the Spanish Mediterranean coast and the Balearic Islands. This extension involves defining and parameterising multiple tsunamigenic fault systems, assembling and controlling the quality of high-resolution topo-bathymetric datasets, and designing robust training and validation strategies.

A practical limitation is that, despite comprehensive coverage of the targeted fault systems, rare source realisations or parameter combinations may fall outside the effective support of the training distribution, which can reduce reliability of point predictions. To handle such cases in operations, we complement deterministic estimates with threshold exceedance probabilities, enabling risk-aware decisions while preserving consistency with established TEWS procedures.

How to cite: Rodríguez Gálvez, J. F., Macías Sánchez, J., Gaite Castrillo, B., Sánchez Linares, C., González del Pino, A., Castro Díaz, M. J., Cantavella Nadal, J. V., and Puertas González, L. C.: AI- and HPC–Driven Tsunami Decision Support for the Spanish TEWS: Atlantic Results and Western Mediterranean Extension, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10252, https://doi.org/10.5194/egusphere-egu26-10252, 2026.

The Lower Mekong River Basin (LMRB) is a flood-prone region experiencing increasing flood risk due to climate change and human activities. This growing challenge underscores the need for robust hydrological models capable of accurate flood prediction. Although purely deep learning approaches have demonstrated strong predictive performance, their data-driven nature does not explicitly represent the underlying physical mechanisms, which limits their interpretability.

In this study, we develop an interpretable deep learning framework based on a Long Short-Term Memory (LSTM) model to predict river discharge across multiple subbasins in the LMRB, with post-hoc interpretation provided by SHapley Additive exPlanations (SHAP). Feature contributions and dominant flood drivers are analysed using SHAP, enabling transparent interpretation of the model’s predictions. The LSTM model demonstrates high predictive performance, achieving Nash–Sutcliffe Efficiency values above 0.9 across all subbasins, although the largest flood peaks are slightly underestimated in midstream subbasins during extreme rainfall events. SHAP analysis indicates that soil-related variables are predominant contributors to discharge prediction, and their influence is partially mediated through interactions with precipitation and runoff. Furthermore, the relative importance of contributing variables changes over time: soil and vegetation-related variables dominate in earlier periods from 2013 to 2017, whereas hydrometeorological variables are more influential after 2017.

Overall, this study highlights the potential of post-hoc interpretable techniques applied to deep learning models for identifying the main contributing variables for discharge prediction and the drivers of flood events across the subbasins of the LMRB, providing valuable insights to support improved flood risk management.

How to cite: Qiu, Y., Shi, X., and He, X.: An interpretable deep learning framework for flood prediction in the Lower Mekong River Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14089, https://doi.org/10.5194/egusphere-egu26-14089, 2026.

Groundwater plays an important role in flood formation yet, flood forecasting in coastal basins is often limited by inadequate representation of surface and groundwater interactions. In this study, we use a Graph Neural Network (GNN) to evaluate the added value of incorporating hourly groundwater information for short-term flood forecasting. Harris County, Texas is considered as a case study. The region is monitored by an extensive network of rainfall and channel-level sensors, supplemented by United States Geological Survey (USGS) wells providing hourly groundwater level data. Within the GNN framework, the sensor network is represented as a graph, where nodes correspond to monitoring areas and edges represent learned hydrological influence paths. Node inputs include recent precipitation, recent streamflow level changes, and normalized groundwater hydraulic load anomalies derived from Harvey Hurricane (2017) and post-Harvey flood events from 2018 to 2023. Results show that including a single groundwater-based prediction variable improves prediction ability by approximately 20% compared to precipitation and level-based reference models. This gain is strongest in areas with continuous groundwater withdrawal and accelerated recharge, where enhanced hydraulic gradients can intensify coastal storage exchange and enhance hydrogeological memory. The learned graph also provides an interpretable, directed interaction structure that supports data-driven causal hypotheses about network connectivity. Furthermore, we estimated the time delay dependency associated with the lag between two stations in our study area, which form a head-to-tail pair. The learned delay between these two stations is sub-daily, with a magnitude of ~0.5 to 0.7 days, corresponding to roughly 12 to 17 hours. This information can guide the parameterization of lag in rainfall-runoff modeling workflows. The results indicate that shallow groundwater dynamics can act as an important regulator of short-term urban flood response in coastal basins. When designing next-generation warning systems for Harris County and similar regions, groundwater levels and rainfall effects should be considered together.

How to cite: N Geykli, A., Gul, E., and Hassanzadeh, E.: Enhancing short-lead flood forecasting by integrated modeling of surface and groundwater , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15179, https://doi.org/10.5194/egusphere-egu26-15179, 2026.

Extreme climate change intensifies the spatiotemporal variability of soil moisture and temperature fields, thereby increasing the frequency and uncertainty of hydrogeological hazards such as floods, landslides, and droughts. These processes are governed by highly nonlinear water–heat coupling in unsaturated soil, where state variables and constitutive parameters are strongly interdependent. This complexity poses significant challenges for conventional physics-based numerical models due to difficulties in parameterization and uncertainty in boundary conditions, while purely data-driven models often lack physical consistency and interpretability. To address these limitations, this study proposes a hybrid modeling framework that integrates physical mechanisms with deep learning by embedding constitutive relationships and physical constraints derived from water–heat transport equations in unsaturated soil into a deep neural network. The proposed approach enables accurate prediction of the spatiotemporal evolution of soil moisture and temperature while preserving physical consistency. Numerical experiments were conducted for multiple soil types and boundary conditions, and the effects of data sparsity and noise on model performance were systematically evaluated. The results demonstrate that the hybrid model significantly outperforms purely data-driven approaches in terms of prediction accuracy and generalization capability, particularly in capturing localized moisture transport fronts and nonlinear dynamic behaviors. Further validation using bench-scale laboratory water–heat coupling experiments demonstrates that the proposed framework not only reconstructs key hydrothermal constitutive parameters but also successfully reproduces the temporal evolution of volumetric water content and temperature in unsaturated soil. Overall, this study provides a robust hybrid modeling strategy for simulating coupled water–vapor–heat processes in unsaturated soil. The proposed framework highlights the potential of physics-constrained deep learning for complex hydrological processes and supports its application in hydrogeological hazard analysis and risk assessment.

How to cite: Tian, S., Wang, Q., Lu, Y., Su, W., and Liu, Y.: Physics-Constrained Artificial Intelligence for Modeling Water–Heat Processes in Unsaturated Soil under Climate Change, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15418, https://doi.org/10.5194/egusphere-egu26-15418, 2026.

Timely detection of landslide precursors is essential for life-saving early warnings, yet remains challenging due to the subtle, non-linear nature of pre-failure ground motion and the computational intensity of processing SAR time series. To address this, we present an operational automated InSAR framework, co-developed under India’s National Supercomputing Mission and the India-EU GANANA HPC collaboration, that processes multi-temporal SAR data optimized on HPC infrastructure (AIRAWAT) to enable long-term satellite-based monitoring large areas for landslide hazard assessment and early warning.

The system ingests Sentinel-1 SLC, IW data and ancillary geospatial layers (DEM and historical landslide inventories). Using GMTSAR-automated workflows, it generates displacement time series and LOS velocity maps across large, landslide-prone regions. These outputs are analysed to identify accelerated displacement trends for known landslides. Threshold values are identified based on the movement signatures, key precursors to slope failure, days to weeks before catastrophic events.

Critically, the entire pipeline, from SAR data ingestion to risk classification, is optimized for low-latency execution on HPC, enabling updates within 24–48 hours of new satellite acquisitions. Outputs are translated into a dynamic risk alert system (Green–Red) and delivered via an interactive dashboard with API access, designed for integration into national disaster response workflows.

Currently piloted in the Himalayas and Western Ghats, this framework demonstrates a scalable, HPC-driven paradigm for time-critical geo-hazard monitoring directly supporting rapid situational awareness and proactive evacuation decisions. The architecture is extensible to other InSAR-monitored hazards (e.g., subsidence, volcanic unrest).

The framework was tested with the well-documented Nepal Earthquake (7.8 M) on 25 April 2015, which triggered more than 47,200 co-seismic landslides. The displacement and coherence time-series were plotted at the crown points and centroids of the landslide polygon. The time-series plots show prominent trends towards the event date. Significant peak was observed in the displacement derived from 08 February 2015 and 21 April 2015 (Sentinel-1 Ascending) interferogram, which may be used as an early warning precursor.

How to cite: Singh, Y. K., Prabhu, T. S. M., Sidar, V. K., and Khare, M. K.: Towards an Operational InSAR Framework on HPC for Time-Critical Landslide Precursor Detection and Early Warning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16040, https://doi.org/10.5194/egusphere-egu26-16040, 2026.

Our study investigates the dynamics of the Gross Glienicker Lake, a groundwater fed lake in the Berlin-Brandenburg region of Germany. This lake (similarly to many others in the region) is experiencing a significant water decline mainly driven by the climate, loosing more than 2 meters of its water levels since the 1970s. To understand the hydrogeological system better, and to identify potential mitigation measures we applied a coupled groundwater-surface water model using HydroGeoSphere (HGS). This 3-D model simulates the hydrological processes of the catchment with high spatial and temporal resolution, incorporating all available geological and hydrological data from the area.

The model was mainly created to evaluate the impacts of different future climate projections on the water levels. We have investigated 3 different RCP scenarios using 43 different climate projection simulations. We have employed machine learning tools to fill in any future data gaps, for example future levels of a river boundary condition and future groundwater extraction rates given population growth trends. To access the uncertainties originating from the HGS model, we have used a meta modeling framework. Meta modeling uses a machine learning based surrogate model (an LSTM in this case), to emulate the input-output numerical relationship of the HGS model in a computationally efficient way. Once trained, the meta model can emulate an HGS model run accurately in a couple of seconds. We fed the meta model with thousands of perturbed climate inputs, showing that the model output is robust even under extreme climatic conditions.

Our results showed that the lake is highly sensitive to precipitation variability, therefore future projections diverge significantly given the scenarios. Except for the wet scenario, all predictions show further water level decrease and they also reveal a strong shift in the seasonal dynamics.

How to cite: Somogyvári, M., Mahmoodi, N., Ölmez, C., Tügel, F., Schneider, M., and Merz, C.: Meta modeling: using machine learning to assess the model uncertainty of a high-resolution groundwater model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16429, https://doi.org/10.5194/egusphere-egu26-16429, 2026.

Severe droughts in the Mekong Delta have exerted profound social and economic impacts in recent decades, underscoring the need for advanced predictive tools to enhance drought mitigation and preparedness. This study presents an AI-based framework that integrates precipitation moisture diagnostics with deep learning to significantly improve drought prediction in the Vietnamese Mekong Delta (VMD). First, moisture source contributions were quantified by using the Water Accounting Model-2layers (WAM-2layers), a moisture tracking tool with the ERA5 reanalysis data as inputs, revealing that over 60% of VMD precipitation originates from upwind source regions, with humidity and wind speed identified as dominant causal drivers of drought-period deficits. Building on this physical insight, a Convolutional Gated Recurrent Unit (ConvGRU) model was employed and explicitly trained with these external atmospheric variables. The model demonstrated robust multi-type drought forecasting skill at a 3-month lead, accurately detecting ~90% of meteorological and ~80% of agricultural droughts with low false-alarm rates (<10%), and reliably reconstructing major historical drought events. This work establishes a synergistic methodology, in which process-based diagnostics inform and validate an AI-driven prediction system, directly contributing to more reliable, physically interpretable early warning and supporting agricultural resilience and economic stability in this climate-sensitive delta.

How to cite: Shi, J. X. and Zhou, K.: AI-Enhanced Drought Forecasting: Fusing Moisture Source Diagnostics and Deep Learning in the Mekong Delta, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17624, https://doi.org/10.5194/egusphere-egu26-17624, 2026.

We present a Digital Twin that tracks the evolution of unrest caused by dike intrusion at volcanoes, leveraging HPC computational models and Artificial Intelligence algorithms to combine real-time monitoring data and physics-based predictions.

The Digital Twin includes three main components. A preliminary, offline scenario database is produced by simulating ground deformation due to dike
intrusion using the finite element HPC software GALES. The model calculates the three-dimensional elastostatic response induced by overpressurized  dikes within a spectrum of geometries, positions and orien. The computational domain can include DEM topography and heterogeneous rock properties, e.g. from seismic tomography surveys. Scenarios are used to train a machine learning module that reconstructs the source of observed deformation patterns. The source is identified in terms of dike size, position, orientation and intensity (dike opening). An auto-encoder, trained on multi-parametric observational time series, detects unrest by identifying variations from the long-term trends at multiple stations. As unrest is detected, inversion of the observed deformation is performed by the trained ML module, providing the location and size of dike intrusion. The geodetic dataset is updated in near-real-time, providing the ability to model dike evolution as it rises towards the surface.

The Digital Twin has been applied restrospectively to the December 2018 dike intrusion at Mount Etna, tailoring ground deformation simulations to the specifics of the volcano, including observed distribution of dike properties. Results show the ability of the Digital Twin to identify unrest and track the evolution of the dike towards the surface to the eruptive vent.

The Digital Twin is available through a dedicated GitLab repository for the EU-funded DT-GEO project, including the case study application. 

How to cite: Montagna, C. P., Bruni, R., De Paolo, E., Allegra, M., Garg, D., Cannavò, F., and Papale, P.: Near-real-time detection of dike-intrusion-indiced unrest: a Digital Twinfor Mount Etna volcano, Italy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18997, https://doi.org/10.5194/egusphere-egu26-18997, 2026.

The processing and analysis of the large volumes of data generated by Interferometric Synthetic Aperture Radar (InSAR) require a significant investment of time, particularly in regions with complex geodynamic behavior. While InSAR presents notable advantages in terms of spatial coverage, precision, or data acquisition speed, traditional analytical methods can be insufficient to fully capture the complexity of deformation patterns or to efficiently manage the increasing amount of available data.

Integrating machine learning techniques into the InSAR computations and interpretation workflow enhances efficiency and automation. These methods enable automated detection of deformation patterns, improved separation of geophysical signals from atmospheric or orbital noise, and the identification of subtle or non‑linear ground motion that may be overlooked by conventional approaches. Such capabilities provide a more robust, reproducible, and sensitive framework for deformation analysis, which is essential for subsequent inversion procedures.

We describe in this presentation first results obtained in the Guadalentin Basin (SE Spain) using all these combined methodologies, as well as the comparison with previous studies for the area.

How to cite: Carrillo, R., Núñez, D., Pardo, E., and Fernández, J.: A New Machine Learning Method for Advanced Treatment of InSAR Deformation Data: Preliminary Results from the Guadalentín Basin (Spain), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19335, https://doi.org/10.5194/egusphere-egu26-19335, 2026.

Effective crisis management requires timely and accurate decision support, leveraging advanced computational methods and geoscientific insights. This study focuses on enhancing decision support for flood and landslide scenarios by integrating geoscientific concepts into prognosis modeling within immersive situation representation frameworks. Building upon the experiences and outcomes of the oKat-SIM project (optimized disaster response through simulation), we demonstrate how coupling high-performance computing with Geographic Information Systems (GIS) can improve real-time response capabilities in civil protection. The project aligns with the foundational goal stated in the Leopoldina report, emphasizing the significance of geoscientific process understanding in decision-making to prepare for, mitigate, and manage natural disasters effectively.

Our approach transcends traditional mapping by utilizing immersive and dynamic 3D representations through synchronized augmented reality (AR) glasses, allowing crisis management teams to maintain interpersonal communication while interacting with floating 3D scenario displays. This integration augments situational awareness and facilitates decision-making in high-pressure environments, such as crisis management centers. The involvement of end-users - both, operational and administrative personnel from municipalities and regional authorities - is crucial throughout the process of application development, allowing iterative improvements driven by real-world feedback.

Technical building aspects include: real-time landslide susceptibility and run-out modelling tightly coupled with GIS-based preprocessing and executed inside a Unity-based immersive runtime, enabling near-real-time scenario updates driven by HPC- and AI-assisted workflows. Advanced rendering techniques such as Gaussian Splatting, multi-resolution terrain streaming, and federated data fusion are leveraged to efficiently integrate remote sensing data, simulation outputs, and uncertainty layers into synchronized AR/3D views, providing scalable, low-latency situational awareness and decision support for time-critical crisis management.

Our case studies demonstrate the effective visualization of historical and potential disaster scenarios, fostering deeper understanding of complex interdependencies and enabling faster, informed decision-making. This interdisciplinary effort bridges geoscience and computational technologies, advancing operational platforms for flood and landslide preparedness and response, and fostering collaborative advancements for modern crisis management.

How to cite: Zeilinger, G., Kauling, S., Oswald, O., Prasannan, A., and Conrad, F.: Integrating Geoscientific Concepts in Prognostic Modeling for Immersive Situation Representation: Enhancements from the oKat-SIM Project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19582, https://doi.org/10.5194/egusphere-egu26-19582, 2026.

In recent years, technological advances in the use of geospatial data (such as satellite images, anthropogenic and/or environmental raster and vector open data, etc.) for hydrogeological risk assessment, combined with advanced analysis techniques (e.g., machine learning), have become increasingly valuable. These technologies can be utilized by local and national authorities for land planning and emergency management to better understand the dynamics associated with climate change. This understanding can help guide actions aimed at safeguarding not only environmental resources but also socio-economic assets and citizens’ lives.

In pursuit of this goal, a partnership has been established between the Po River Basin District Authority (AdBPo), the Italian Space Agency (ASI), and academic and research institutions such as the University of Bologna (UNIBO), the University of Modena and Reggio Emilia (UNIMORE), the University of Padova (UNIPD), and the Institute of Environmental Geology and Geoengineering of the National Research Council of Italy (CNR-IGAG). The aim is to implement a downstream service for monitoring landscape evolution related to fluvial systems (geomorphological classification), and slope dynamics (including landslides and rock glaciers) and to quantitatively evaluate the exposed assets.

The PARACELSO project (Predictive Analysis, MonitoRing, and mAnagement of Climate change Effects Leveraging Satellite Observations) aims to develop a modular and interoperable GIS cloud-based platform that supports the analysis of natural phenomena (such as fluvial hydrodynamics, landslides, and rock glaciers) using multi-sensor satellite data imagery provided by: 

• DIAS platforms deployed by the Copernicus Programme (e.g., Sentinel 1-2),
• ASI missions such as CosmoSkyMed, PRISMA, and SAOCOM.

Furthermore, a methodology integrating Earth Observation and geospatial data analysis, to evaluate the exposed assets, has been implemented using open-source libraries.

To facilitate this, within the Big Data HPC MarghERita infrastructure— a supercomputing system named in honor of the scientist Margherita Hack and provided by the Emilia-Romagna Region — computational resources are employed for the high-performance processing, analysis, and storage of large volumes of acquired satellite imagery, as well as additional geospatial datasets. The platform executes the project-developed algorithms to investigate the temporal evolution of fluvial and slope systems. Furthermore, the infrastructure supports the access, visualization, and sharing of the processed and analyzed data.

The project has received funding from ASI through the “I4DP_PA (Innovation for Downstream Preparation for Public Administrations)” Call for Ideas.

How to cite: Hammouti, M., Zazzeri, M., Sterlacchini, S., Correa Da Mota, T., Mazzanti, M., Pancaldi, M., Agostini, M., Bizzi, S., Cecchetto, M., Berti, M., Brardinoni, F., Corsini, A., Tondo, M., Critelli, V., Mulas, M., Candela, L., D'Amato, L., and Simonelli, T.: Cloud-Based GIS Platform for the Management of Hydrogeological Risks in the Po Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19618, https://doi.org/10.5194/egusphere-egu26-19618, 2026.

Data-Driven Prediction of Peak Ground Acceleration from Seismic Waveforms
The identification and rapid estimation of earthquake parameters, such as Peak Ground
Acceleration (PGA), are critical components of earthquake monitoring and Earthquake Early
Warning (EEW) systems. As seismic waves propagate through the geological media, their
interaction with subsurface layers possessing varying elastic and damping properties leads to
significant variability in observed ground motion. These local site effects strongly influence
PGA values, for instance if the site is composed of soft-sediments the amplification within the
ground motion is more prominent than that of a rocky terrain or very firm sediments.
In this study, we investigate the application of deep learning techniques to model the nonlinear
relationships between incoming seismic signals and the resulting PGA. The proposed model
architecture may be considered a prototype that can be integrated into operational EEW
systems, enhancing the timeliness and accuracy of ground motion predictions and thereby
supporting more effective emergency response and risk mitigation strategies.

How to cite: Srivastava, N., Faber, J., Sandeep, S., and Yadav, M.: Data-Driven Prediction of Peak Ground Acceleration from Seismic Waveforms, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19763, https://doi.org/10.5194/egusphere-egu26-19763, 2026.

Thermokarst activity is intensifying under a warming climate, and retrogressive thaw slumps (RTSs) in the Beiluhe region of the Qinghai–Tibet Plateau represent one of the most active examples. To produce regional, multi-year RTS inventories, we applied a domain-adaptation AI approach to improve model transferability across optical remote-sensing imagery acquired under diverse illumination conditions. From 2019 to 2022, the number of mapped slumps increased from 803 to 885, and the total affected area expanded from 1,727 ha to 2,329 ha. Despite these rapid changes, how hydroclimatic forcing, especially precipitation and land surface temperature (LST), jointly influences slump-related ground deformation remains unclear. Here, we analyze InSAR-derived surface deformation in relation to precipitation across different LST regimes. RTSs exhibiting larger seasonal deformation amplitudes also show higher subsidence rates. When LST is below ~0 °C, greater annual subsidence is associated with drier years; when LST is above 0 °C, greater subsidence more often occurs in wetter years. These results highlight precipitation and temperature controls on RTS deformation and emphasize the need to consider combined hydroclimatic conditions when interpreting remote-sensing deformation signals in permafrost terrain.

How to cite: Hu, X., Zhou, Y., Lin, Y., and Song, Y.: Hydroclimatic Controls on Thaw Slump Deformation on the Qinghai–Tibet Plateau, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21782, https://doi.org/10.5194/egusphere-egu26-21782, 2026.

Reservoir modelling in heterogeneous carbonate systems is often constrained by sparse well control and labor-intensive interpretation, which increases uncertainty when extrapolating between wells. We present an enhanced Pix2Geomodel.v2 workflow that reframes facies and petrophysical modelling as paired image-to-image translation. Facies and petrophysical properties are exported from a reference reservoir model, converted into paired 2D training images, and used to train a Pix2Pix-style conditional generative adversarial network (cGAN). The architecture couples a U-Net generator with a PatchGAN discriminator, enabling the model to learn spatial relationships directly from examples. To reduce data requirements while retaining geological heterogeneity, the workflow operates on a streamlined grid of 54 vertical layers and targets complex facies distributions. Preliminary results show stable training and predictions that reproduce the main geological patterns of the reference data. In facies-to-property translation, the network learns meaningful mappings to porosity, permeability, and volume of shale.

How to cite: Al-Fakih, A., Hanafy, S., Saraih, N., Koeshidayatullah, A., and Kaka, S.: Data-efficient enhanced Pix2Geomodel.v2 for complex facies settings, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2220, https://doi.org/10.5194/egusphere-egu26-2220, 2026.

Capturing cross-property correlations while preserving spatial continuity is essential for reliable reservoir characterization, especially in heterogeneous reservoirs where facies architecture controls petrophysical variability. In this study, we evaluate Pix2Geomodel.v2 as a bidirectional image-to-image translation framework that learns mappings between facies and petrophysical properties using paired 2D slices exported from a reference reservoir model. To reduce data demands while maintaining geological complexity, the workflow operates on a streamlined grid of 54 vertical layers, enabling efficient training and rapid experimentation without removing key stratigraphic and facies patterns. The approach is based on a conditional generative adversarial learning strategy. A U-Net generator is trained to synthesize target facies or property maps from input images, while a PatchGAN discriminator encourages locally realistic textures and geologically plausible transitions. The paired-slice formulation allows the model to learn both large-scale structural organization and fine-scale heterogeneity directly from examples. We investigate two complementary directions: (i) facies-to-property translation, where facies maps are used to predict continuous property fields such as porosity and permeability, and (ii) property-to-facies translation, where petrophysical images are used to reconstruct discrete facies distributions. Beyond conventional forward mapping, the reverse translation experiments are particularly informative because they test whether the model captures meaningful cross-property dependencies rather than superficial patterns. The reconstructed facies maps recover coherent large-scale facies trends and geologically consistent connectivity, indicating that the learned representation encodes relationships between depositional architecture and petrophysical response. Spatial realism is further examined using experimental variograms, providing a continuity-based check that generated outputs qualitatively align with the reference model in terms of spatial correlation structure. Overall, the results suggest a data-efficient route to robust forward and reverse translations that can support faster reservoir model prototyping, property population guided by facies, and consistency checking between facies and petrophysical interpretations.

How to cite: Kaka, S., Al-Fakih, A., Saraih, N., Koeshidayatullah, A., and Hanafy, S.: Bidirectional translation + spatial continuity validation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2222, https://doi.org/10.5194/egusphere-egu26-2222, 2026.

Data generated in the field of geoscience has unique properties that can be characterized by high complexity, sparsity, and site-specific variability. Owing to the unique characteristics, applying general artificial intelligence frameworks and achieving model generalization in geoscience still remains a challenging problem. In this work, we introduce the KIGAM GeoAI Platform, an integrated AI environment designed to bridge the gap between advancing AI technology and practical geoscience research. The platform supports the entire research workflow through a user-friendly, web-based interface, systematically covering essential stages: data uploading, preprocessing, model development, testing, validation, and the final deployment of analytical applications. By providing a centralized online environment for collaborative research, the platform aims to reduce technical entry barriers for geoscientists who may not be AI experts, while establishing a robust foundation for data-driven cooperation. We plan to continue improving and scaling this platform to ensure it remains a stable, accessible, and high-performance tool for both domestic and international geoscience communities. Through these efforts, the KIGAM GeoAI Platform is expected to accelerate digital transformation and foster a more integrated global research ecosystem in the field of geoscience.

How to cite: Kwon, J.:  Innovate with Ease: Introducing the KIGAM GeoAI Platform, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2290, https://doi.org/10.5194/egusphere-egu26-2290, 2026.

Artificial intelligence (AI) tools increasingly enhance the efficiency and consistency of seismic interpretation, particularly in structurally complex areas or areas where data quality is reduced by acquisition limitations. As a result, interpretations can become difficult and time-consuming, especially in the context of structural interpretation and fault tracking. To evaluate the performance of AI-based fault detection, we applied Geoplat AI software to a 3D seismic volume from the Drmno Basin, located at the southeastern margin of the Pannonian SuperBasin in Serbia.
A conventional structural interpretation was first performed by mapping the major fault systems, then minor fault systems, generating fault sticks and polygons for all visible faults and developing a structural model to illustrate the basin's opening and evolution. Subsequently, AI-based workflows were applied in order to enhance the quality of the seismic data. This involved removing noise, restoring reflections, highlighting fault zones, and applying smoothing filters. The final step was the utilization of a fault tracking tool that segments the seismic data, recognizes fault zones, traces them, identifies structural patterns, and calculates a probability field. The AI-derived fault interpretation was then compared with the manual interpretation.
The results indicate that the Drmno basin was developed under an extensional tectonic regime during the Early Miocene, which formed a large Morava detachment fault and opened accommodation of the basin. The basin itself has complex architecture in the syn-rift phase, with many synthetic and few antithetic faults, oriented from the east to the west. During the stage of the rift climax, the dominant fault systems remained consistent, with most syn-rift structures continuing to accommodate the subsidence formed by the Morava detachment. The shift in the tectonic conditions in the post-rift stage leads to the formation of systems of parallel faults in the younger sediments, adjusting strike-slip movements in a compressional tectonic field. The younger structures are dominantly oriented in the north-south direction, or reactivated older fault structures.
The AI tool effectively interpreted fault systems in the younger geological units, benefiting from higher data quality, and clearly indicated younger fault systems with a high level of certainty. However, in the lower part of the seismic cube, the basement structures remain unclear or unrecognized. Reactivated fault surfaces and a significant fault zone are evident in the interpretation. In areas with low-quality seismic data, the AI tool struggled to trace faults accurately, resulting in geologically inconsistent fault patterns.
Overall, the AI-based 3D fault tracking tool proved effective in resolving the main structural framework of the basin. The dominant fault directions are clearly identifiable, and the main geological structures have been mapped with reasonable precision. The AI-supported interpretation successfully captures the main structural trends and provides a solid basis for evaluating the tectonic evolution. This case study demonstrates the potential of AI to support structural interpretation and tectonic analysis of complex sedimentary basins.

How to cite: Ninić, A., Radivojević, D., and Đurić, D.: The application of artificial intelligence in fault tracking on 3D seismic data – A case study from Drmno Basin (SE Serbia), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2380, https://doi.org/10.5194/egusphere-egu26-2380, 2026.

Climate Service Recipes (HACID-CSR) is an agentic system designed to assist providers of climate services in developing their advice for a wide range of clients. HACID-CSR guides providers by navigating the large and ever-increasing corpus of knowledge as well as an area without established standards and with limited access to scientific experts. It automatically generates detailed workflows (or “recipes”) by leveraging both a large language model’s internal reasoning and contextual knowledge from a domain knowledge graph for climate services (CS-DKG). The CS-DKG is an expert-curated ontology of climate service concepts with mapped relationships between climate variables, emission scenarios, indices, hazards, sectors, and key datasets (CORDEX, CMIP5, UKCP18), built as part of the Horizon Europe-funded HACID project (Hybrid Human Artificial Collective Intelligence in Open-Ended Decision Making).

The HACID-CSR architecture consists of a memory-enabled supervisor agent orchestrating multiple specialised agents. A planning agent first proposes an initial workflow outline, and a preliminary recipe agent uses only the LLM’s knowledge to draft answers to key workflow steps. The system then engages a knowledge graph retrieval sequence: a class selection agent identifies relevant classes in the CS-DKG, an instance selection agent finds specific instances (entries) highly relevant to the query within those classes following a two-stage selection process, i.e. semantic similarity based pre-selection and LLM-enabled refined selection, and a subgraph extraction agent retrieves the corresponding subgraph of related knowledge entities. Next, a recipe generation agent creates each step of the workflow by combining the LLM’s reasoning with the retrieved graph context using graph retrieval-augmented generation (GraphRAG). Finally, a recipe refinement agent compares the preliminary LLM-only solution with the knowledge-enhanced solution and refines the output, yielding a diverse and context-aware workflow.

By using this multi-agent approach, HACID-CSR increases the diversity of solutions and fills the knowledge gap between climate information and domain specific applications, helping experts to identify suitable methodologies and datasets. The resulting workflows are more traceable and transparent, improving user trust compared to answers from a general-purpose chatbot. We have also developed a bespoke automatic evaluation method to complement human expert validation of the generated recipes. We highlight the potential of the HACID-CSR approach for multi-hazard climate service design, and discuss remaining challenges and opportunities for further refinement of this agentic LLM-based system.

How to cite: Abele, A., Xie, H., Biswas, A., Dong, H., Fung, F., and Williams, H.: Climate Service Recipes: automatic multi-hazard climate information workflow generation using agentic Large Language Models (LLMs) and knowledge graphs, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2819, https://doi.org/10.5194/egusphere-egu26-2819, 2026.

Earth system science relies on the integration of knowledge from many branches of geoscience, including climate dynamics, hydrology, ecology, land use and biogeochemical cycles. However, the scientific literature informing these domains has become vast and increasingly difficult to navigate due to its rapid development and disciplinary spread. This complexity makes it difficult to maintain an integrated overview of relevant findings and to identify scientific connections in a systematic manner. Recent advances in generative artificial intelligence (AI) and large language models (LLMs) provide opportunities to support these tasks, particularly when combined with retrieval methods and transparent source attribution.

Here we propose a retrieval-augmented AI platform designed to assist scientific knowledge integration in Earth system science. The platform is conceived as a living system, built on a continuously expanding and updateable knowledge base that aggregates scholarly literature from major scientific databases. User queries initiate targeted retrieval of relevant documents followed by the generation of concise, source-linked summaries using locally hosted open-weighted LLMs. By explicitly grounding outputs in retrieved literature, the platform alleviates the need for manual screening and limits hallucination risks that currently constrain the use of general-purpose LLMs in geoscientific research.

Evaluation of the initial prototype demonstrates that domain-specific retrieval-augmented generation systems can provide reliable, traceable synthesis of Earth system knowledge and help address the growing gap between accelerating publication rates and the need for timely, verifiable scientific assessment.

How to cite: Kart Tokmak, Ö., Caesar, L., and Sakschewski, B.: A Living AI Platform for the Earth System Science, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3316, https://doi.org/10.5194/egusphere-egu26-3316, 2026.

This study investigates the domain adaptation of the Vision-Language Model (VLM) for road damage assessment, focusing on a fine-tuning strategy optimized for resource-constrained engineering environments. Unlike conventional object detection models that operate within fixed label spaces, VLMs provide superior semantic understanding and generalization in complex scenarios. To facilitate practical deployment, this research systematically analyzes key variables of Parameter-Efficient Fine-Tuning (PEFT) to mitigate the high computational demands inherent in large-scale VLMs.

In the experimental phase, hyperparameter tuning was conducted using the Low-Rank Adaptation (LoRA) technique. The primary variables included LoRA ranks (16, 32, 64, and 96), training data scale, and image resolutions (1,024ⅹ28ⅹ28 vs. 1,536ⅹ28ⅹ28). A comprehensive dataset of 26,796 images comprising six damage categories and negative samples was established, utilizing a 7n sampling strategy (n=500, 750, 1,000) to address class imbalance. The impact of data volume was evaluated by augmenting the 7,000-sample set (corresponding to n=1,000) to match the full dataset size of 26,796, with zero-shot inference serving as the performance baseline.

Experimental results demonstrated substantial improvements over zero-shot inference, indicating that performance positively correlates with increased data scale with augmentation and higher image resolution, while lower LoRA ranks (16, 32) proved most effective for this domain. Furthermore, the introduction of specialized ad-hoc metrics, M_mAP and M_F1, verified a stable trade-off between False Positives and False Negatives. Notably, to minimize safety-critical False Negatives, a prompt engineering-based 'Double Check' mechanism and multi-turn interactions were utilized. This approach successfully leveraged the model’s inherent reasoning capabilities to refine damage identification through iterative feedback.

Acknowledgements This research was supported by Basic Science Research Program through the National Research Foundation of Korea(NRF) funded by the Ministry of Education(RS-2025-25437298)

How to cite: Kim, D. and Youn, H.: Optimizing Vision-Language Model for Robust Road Damage Assessment via Parameter-Efficient Fine-Tuning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3674, https://doi.org/10.5194/egusphere-egu26-3674, 2026.

Nitrate contamination of groundwater is often addressed as a diffuse agricultural management problem, yet monitoring in Denmark depicts that exceedance risk at abstraction and observation wells is spatially structured and closely linked to surrounding land-use composition and configuration. This suggests a land-use policy opportunity: if landscape fractions and fragmentation patterns help drive nitrate vulnerability, interventions could be spatially targeted and tailored rather than uniformly applied. In this study, we present a scenario-based planning framework for policy appraisal, enabling regulators, municipalities, and water utilities to test alternative policy packages and targeting rules and quantify their expected effects on groundwater nitrate hotspot risk. The system operates a predictive pipeline that relates nitrate outcomes to land-use fractions and landscape configuration metrics computed within configurable protection zones. Model outputs are formulated as a binary hotspot classification (hotspot vs. non-hotspot) based on exceedance of a drinking-water nitrate threshold, producing vulnerability maps to prioritize locations for intervention and prevention. The core functionality is a “what-if” engine built on an AI-based ensemble that generates a baseline nitrate-risk probability map and re-predicts risk under user-defined scenarios. Scenario levers are organized into two policy bundles: (i) land-use policy and management, implemented as controlled reallocations among land-cover fractions (e.g., reducing large contiguous cropland blocks, increasing wetland/riparian woodland cover, restricting impervious expansion) while enforcing feasibility constraints; and (ii) agricultural management, implemented as proportional reductions or caps on nitrogen surplus and fertilizer inputs. For each scenario, the system outputs an updated probability map and a different map relative to baseline, supporting spatial prioritization, instrument design, and transparent justification of differential targeting. By combining ex-ante scenario testing with ex-post monitoring of hotspot transitions after implementation, the framework supports adaptive groundwater governance and moves from risk mapping toward operational, spatially explicit nitrate-reduction policy design.

How to cite: Naghibi, A., Ahmadi, K., and Berndtsson, R.: Policy-oriented Land-Use and Agricultural Management Scenarios for Groundwater Nitrate Hotspot Mitigation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4407, https://doi.org/10.5194/egusphere-egu26-4407, 2026.

AI-driven weather and climate prediction has become highly visible in recent years, and most Earth scientists are now familiar with learned forecast models. Far fewer are aware that the same advances in artificial intelligence are producing general-purpose systems that can autonomously review literature, write and debug code, design experiments, and carry out extended research tasks with minimal supervision. These capabilities may ultimately have a greater impact on everyday scientific practice than any single prediction model.

AI-based forecasting represents only a narrow entry point into a broader transformation driven by hybrid intelligence, in which domain-specific Earth system models are combined with general AI systems such as large language and multimodal models and autonomous agents. In practice, this hybrid intelligence already spans simulation, data assimilation, downscaling, and analysis, while general AI systems increasingly handle coding, synthesis, and workflow orchestration. Together, these systems function less as isolated tools and more as adaptive research partners. Drawing on examples from NVIDIA’s Earth-2 research program and related international efforts, this talk examines how this shift reconfigures the human role toward problem formulation, validation, interpretation, and ethical governance, and highlights practical AI-assisted workflows already reshaping research productivity. Framing AI for environmental prediction within this wider context invites a broader discussion of how hybrid intelligence should be integrated thoughtfully into future Earth system science.

How to cite: Hall, D.: Beyond the Forecast: Hybrid Intelligence as a Force Multiplier for Earth Science, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5926, https://doi.org/10.5194/egusphere-egu26-5926, 2026.

The Yellow River Basin (YRB) serves as a critical repository of literature for understanding human-earth systems, yet existing automated metadata-level review methods suffer from deep semantic loss and deficiencies in spatial representation: They neither capture fine-grained logic chains from full texts nor possess the capability to extract the spatial and hierarchical attributes of geographic entities. However, rapid developments in Large Language Models (LLMs) provide a technological opportunity for the automated extraction of full-text knowledge. To this end, this study proposes the Geo-Knowledge Infused Reasoning Framework (GK-IRF), coupling full-text semantics with multi-level spatial indexing. Methodologically, we first construct an ontology-based full-text parsing mechanism based on 8,493 YRB-related papers (2015-2024), utilizing LLMs to accurately extract structured semantic triplets. Simultaneously, we introduce an adaptive multi-level GeoHash indexing model to map textual toponyms into hierarchically nested grid sets, reconstructing the spatial coverage and multi-scale associations of geographic entities. Validations against a manually annotated dataset indicate that GK-IRF achieves an F1-score comparable to human performance in full-granularity semantic extraction; furthermore, the Spatial Coverage Accuracy of the multi-level grids for the YRB substantially outperforms traditional geocoding methods, effectively resolving the challenge of multi-scale coverage representation.

How to cite: Wu, S. and Wang, H.: Coupling Full-Text Semantics with Multi-Level Spatial Indexing: A Knowledge Representation Framework for Yellow River Basin Literature, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6201, https://doi.org/10.5194/egusphere-egu26-6201, 2026.

Seismic interpretation of mass transport deposits (MTDs) relies heavily on expert knowledge and conceptual reasoning yet remains difficult to formalize and scale. While recent artificial intelligence (AI) methods have shown strong capabilities in seismic pattern recognition, most approaches operate as black boxes and remain poorly aligned with the interpretative frameworks used by geoscientists, limiting transparency and trust.

This study proposes a geoscience-aware hybrid intelligence framework that integrates expert knowledge graphs (KGs) with large language models (LLMs) to support interpretable seismic interpretation of MTDs. The approach builds upon the conceptual methodology of Le Bouteiller et al. (2019), which organizes MTD interpretation through causal relationships linking environmental controls, mass transport properties, and observable seismic descriptors across trigger, transport, and post-deposition phases.

The KG provides a structured reference for interpretation that constrains vocabulary, causal direction, and temporal logic. Our workflow reads scientific papers, identifies relevant descriptors and processes, checks them with LLMs, and evaluates how well they support interpretation. In this setup, seismic descriptors give different levels of support (weak to strong) for geological processes, like how experts reason under uncertainty.

Preliminary results show that ~68% of expert defined concepts are recovered in the inferred graph, with a semantic validation score of 0.73, indicating good conceptual alignment. However, descriptor matching based on textual similarity remains difficult, with average scores around 0.41. This gap highlights the difference between semantic agreement (conceptually correct) and textual agreement (exact wording), mainly due to synonymy and variable phrasing in the literature. We plan to address this by using domain-specific LLMs and ontology-based synonym expansion to improve semantic matching in future iterations

How to cite: Talbi, F. B., Armitage, J., Charléty, J., Rabaute, A., Bouziat, A., Vittaut, J.-N., and Leroy, S.: Geoscience-Aware AI for Interpretable Seismic Interpretation of Mass Transport Deposits Using Knowledge Graphs and Large Language Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6459, https://doi.org/10.5194/egusphere-egu26-6459, 2026.

The assessment of environmental and resource performance of energy transition technologies relies on quantitative information scattered across heterogeneous sources, including scientific articles, patents, and industrial reports, such as ESG (Environmental, Social, and Governance) disclosures. These documents contain key data for Life Cycle Inventory (LCI) and Material Flow Analysis (MFA), such as material and energy intensities, water consumption, mining production volumes, emissions, and technological descriptors. However, this information is predominantly embedded in unstructured PDF documents optimized for human reading, making large-scale, traceable data aggregation difficult and costly when performed manually.

This work presents an automated and modular methodology designed to extract and contextualize quantitative LCI and MFA data from three major categories of technical documentation. The approach combines large-scale document collection, relevance screening, and multimodal artificial intelligence within a reproducible and auditable workflow.

• Scientific Articles

Peer-reviewed articles are collected through automated scraping workflows based on structured search outputs. Documents are screened for LCI/MFA relevance using domain-specific keywords, methodological markers, and quantitative signal density. Relevant articles are then processed using a multimodal AI-based extraction core in which each page is analyzed through a combined text and image input. This enables robust extraction of numerical values from tables, text and figures while preserving contextual information such as units, methodological assumptions, and source location.

• Patents

Patent documents contain information about future trends on technologies and metal uses. Patents are collected via dedicated scraping pipelines and processed separately from scientific articles. The workflow focuses on extracting and structuring patent metadata, including publication year, country, and technology class, in order to characterize technological activity related to energy transition technologies. While quantitative LCI/MFA extraction from patents is not yet systematically performed, the pipeline enables descriptive statistical analyses of patent dynamics, including temporal trends and geographical patterns of technological development.

• Mining technical and ESG Reports

Official mining companies reports, with a specific focus on ESG ones, are processed through a screening module acting as a gatekeeper. The screening relies on sequential text parsing and, when necessary, geometric reconstruction of tables to identify reports containing sufficiently granular and structured quantitative information. Following human validation of the screening results, selected reports are analyzed using a IA-multimodal vision–language model combining page images and extracted text, enabling structured extraction of industrial metrics with associated context and traceability.

This automated methodology addresses one of the core challenges of data collection and significantly improves the granularity, consistency, and verifiability of LCI datasets and MFA inputs. The application of methodology is illustrated through examples related to battery and hydrogen technologies based on scientific articles and patents, and through case studies on copper and nickel production with a focus on mining based on industrial report. Although applied for LCA and MFA, the approach can also support the extraction of other types of data and indicators relevant to environmental and resource analyses. The tool provides automated and reliable support for researchers aiming to extract comprehensive foundational data from heterogeneous sources.

How to cite: Bejjit, C. E., Monfort, D., Muller, S., Lai, F., Beylot, A., and Hennioui, D.: Mining and raw materials sector: Automated Data Extraction and Contextualization for Life Cycle Inventory (LCI) and Material Flow Analysis (MFA) Across Scientific Articles, Patents and Mining companies Reports, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6501, https://doi.org/10.5194/egusphere-egu26-6501, 2026.

Compound drought and heat extremes (CDHEs) exert impacts that exceed the sum of their individual components. With global warming amplifying the associated risks, CDHEs have become a critical threat to agricultural production. Thus, identifying and monitoring CDHEs in cropland systems is key for food security. As CDHEs formation and evolution are shaped by climatic factors, hydrological cycles, and ecosystem feedbacks, their fine- and large-scale identification in agricultural areas poses substantial challenges.

Our study reviews existing methods for identifying CDHEs, including combined threshold approaches, comprehensive index methods, traditional machine learning techniques, and improved mechanistic modeling. We summarize the current limitations of these methods as follows: (1) Combined threshold and comprehensive index methods often focus on a single aspect of CDHEs, failing to systematically describe the complex processes of compound events. (2) While traditional machine learning methods attempt to integrate characteristics of the hazard-bearing body, disaster-causing factors, and hazard-inducing environment to establish complex nonlinear relationships between multiple elements and compound event indices, their "black-box" nature lacks mechanistic interpretability. Furthermore, these methods rely heavily on large volumes of high-quality samples to achieve satisfactory accuracy. (3) Improved mechanistic models, typically based on classical agricultural process models such as APSIM and AquaCrop, introduce CDHE impact modules to address the oversimplification of these effects in original models. Nevertheless, these mechanistic models require extensive input parameters, and their calibration processes depend on substantial amounts of measured data. Additionally, the computational resources needed for simulations are considerable, making the cost of analyzing CDHEs over large farmland areas under various future climate scenarios prohibitive for individual researchers.

To address these challenges, this study highlights the potential of physics-informed neural network models for identifying compound events and proposes future research directions regarding mechanistic constraints, neural network architecture design, and experimental plans: (1) Farmland CDHEs are essentially phenomena of water and heat imbalance within the soil-crop-atmosphere (SCA) system. Utilizing the Richards equation and the Penman-Monteith formula can characterize this process by constraining the water and heat environmental factors at the two key interfaces: root-soil and leaf-atmosphere. (2) Solar-induced chlorophyll fluorescence (SIF), a byproduct of vegetation photosynthesis closely related to GPP, responds rapidly to physiological damage caused by stress. Utilizing multi-band SIF data can provide a detailed depiction of crop physiological responses to stress from the perspective of the hazard-affected body. (3) Automated design of model architectures incorporating mechanistic information for farmland compound events can be achieved through distillation learning. (4) Future work should integrate ground-based water and heat control experiments with site-specific hyperspectral SIF observation data. Through continuous combinatorial experimental design, this approach can lead to the development of accurate and efficient physics-informed neural networks. Coupled with large-scale satellite and reanalysis data products, this framework aims to enable the large-area identification of farmland CDHEs under future climate scenarios.

How to cite: Xia, H., Wu, J., Zhou, L., and Du, R.: Towards a Mechanism-Informed Intelligent Framework for Identification of Compound Drought-Heat Extremes in Croplands, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8579, https://doi.org/10.5194/egusphere-egu26-8579, 2026.

Electromagnetic methods are among the most widely used techniques in the geophysical exploration industry due to their efficiency and non-invasive nature. However, their data processing workflows are highly time-consuming and strongly dependent on expert intervention. With the rapid and broad success of deep learning, applying deep learning techniques to electromagnetic methods to overcome the limitations of traditional approaches has become an active area of research. The effectiveness of deep learning methods, however, largely depends on the quality of the dataset, which directly influences model performance and generalization capability. Existing applications typically rely on self-constructed datasets composed of randomly generated one-dimensional models or structurally simple three-dimensional models, which fail to capture the complexity of realistic geological environments. Moreover, the absence of a unified and publicly available three-dimensional geoelectrical model repository has further constrained the development of deep learning for three-dimensional electromagnetic exploration. To address these challenges, we introduce OpenEM, a large-scale, multi-structural three-dimensional geoelectrical model repository that incorporates a wide range of geologically plausible subsurface structures.

OpenEM comprises nine categories of geoelectrical models, encompassing a wide spectrum of subsurface structures ranging from simple to complex. These include models of homogeneous half-spaces with embedded anomalous bodies, as well as configurations featuring flat stratigraphy, curved stratigraphy, planar faults, curved faults, and their variants containing anomalous bodies. The resistivity values span from 1 to 2000 Ω·m, with the number of layers ranging from three to seven. In models containing anomalous bodies, the number of anomalies varies from one to five, and both regular and irregular geometries are considered to enhance dataset diversity and realistic representativeness. In addition, OpenEM is accompanied by a three-dimensional model generator that enables fully controllable model construction, allowing users to customize structural configurations, including resistivity magnitudes, fault geometries and locations, as well as the size, shape, and placement of anomalous bodies.

OpenEM provides a unified, comprehensive, and large-scale dataset for common electromagnetic exploration systems, thereby promoting the application of deep learning methods in electromagnetic prospecting.

How to cite: Wang, S., Wang, X., Deng, F., and Jiang, P.: OpenEM: Large-scale multi-structure 3D dataset for electromagnetic methods, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8976, https://doi.org/10.5194/egusphere-egu26-8976, 2026.

Typically, to start working on a remote sensing–based application, various analyses and insights are needed from domain experts. A significant amount of time and effort goes into preprocessing, structuring, and analyzing the data, which can be a repetitive task, especially when a multi-sensor approach is involved. This often takes away time that could otherwise be invested in innovation or research. To address this, training an LLM to understand and process the context of remote sensing tasks can improve efficiency and reduce human-induced errors.

In this work, we develop an AI agent that can reason and think like a remote sensing expert. This agent uses a RAG-based foundational model (FM) and is equipped with various image processing tools to complete a task. We use gpt-4.1-mini as the FM and the Agno framework to deploy the agent. The knowledge base provided to this agent is specially curated with relevant research articles, books, and remote sensing methodologies. This knowledge base helps the model break down a problem into logical steps that can be performed using the tools available within the agent.

These tools can download data, process it, and provide relevant statistics and visualizations. The user can prompt the agent to download multi-sensor (optical and SAR) data, perform time-series analysis for forest monitoring, and identify deforestation hotspots. The agent can fetch data from Google Earth Engine (GEE), plan processing workflows, dynamically generate Python code, and complete the prompted tasks. This approach highlights the feasibility of integrating LLMs with domain-specific knowledge bases and geospatial processing tools to create autonomous, context-aware systems. Figure 1 depicts the overall workflow of the proposed agentic system, illustrating the interaction between the user, the knowledge base, the foundational model, and the integrated processing tools. The framework is directly usable for operational forest monitoring applications and can be further fine-tuned and extended to support a broader range of environmental monitoring and geospatial analytics use cases.

                                        Figure 1: Workflow of the Agentic AI system

How to cite: Jain, A. and Sabir, A.: Development of a Context-Aware AI Agent for Forest Applications Using Multi-Sensor Data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10390, https://doi.org/10.5194/egusphere-egu26-10390, 2026.

Multi-agent GIS systems are increasingly emerging as a general paradigm for complex geospatial tasks. However, many existing approaches rely on text-only large language models (LLMs) as the primary reasoning substrate. In the absence of explicit geometric constraints and verifiable evidence, spatial relations are often indirectly represented through linguistic statistical correlations. This makes LLMs prone to inconsistency when interpreting and inferring topological, directional, and distance relations in geospatial data, and leads to error accumulation across multi-step tool invocations and long-horizon decision-making, ultimately degrading the accuracy and efficiency of task reasoning and execution. In this work, we propose VisCritic-GIS, a multi-agent framework for geospatial task reasoning and execution driven by visualized evidence review. VisCritic-GIS introduces a Visualization Generation Agent and a Visualization Critic Agent into conventional multi-agent GIS pipelines. The generation agent renders key spatial data and intermediate results into 2D maps, explicitly externalizing spatial relations in a visual form. The critic agent leverages multimodal LLMs to read and critically review these map-based evidence, producing textual feedback on spatial relations, anomalous results, and reasoning deviations, which constrains and drives iterative refinement of other agents’ reasoning trajectories and toolchain configurations. We build evaluation protocols over representative remote sensing and geospatial tasks, and systematically demonstrate that VisCritic-GIS improves task accuracy, execution efficiency, and interpretability. Overall, our framework provides a mechanism for shifting geospatial reasoning from “text-only probabilistic completion” toward “visually grounded, verifiable inference,” thereby strengthening the robustness of spatial relation understanding in multi-agent GIS systems.

How to cite: Lan, Q., Hu, L., Wu, S., and Du, Z.: VisCritic-GIS: A Visualization-Critic–Empowered Framework for Multi-Agent Geospatial Task Reasoning and Execution, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10473, https://doi.org/10.5194/egusphere-egu26-10473, 2026.

We present SeaScope, an explainable AI agent that accelerates interaction with complex Earth Observation (EO) workflows. Users express analytical questions in natural language, which are transformed into transparent, executable EO analyses. By combining generative AI, vision–language models, and Retrieval-Augmented Generation (RAG), SeaScope links scientific literature, satellite data descriptions, and validated analysis methods to automatically generate, execute, and explain EO workflows. For example, a query such as “Detect vessel activity and possible oil spills in May 2025” triggers dataset selection, code generation, cloud execution, and map outputs with traceable reasoning.

SeaScope is designed as a geoscience-specific AI agent that supports both rapid decision-making and accelerated research. Non-technical users can obtain EO-based insights in time-critical situations without continuous involvement of expert programmers, while researchers benefit from faster hypothesis testing, automated pipeline generation, and reproducible workflows. Human expertise remains central: users inspect retrieved sources, review generated code, and validate analytical steps, ensuring scientific control and accountability. This setup combines domain knowledge with AI-driven scalability, addressing challenges such as sensor-specific scripts and fragmented tools.

As a pilot use case, SeaScope is applied to maritime EO in the Mediterranean region, supporting environmental monitoring and marine activity analysis using satellite data. Beyond the application, the project delivers research insights on generative and vision-based AI for EO, including lessons learned from benchmarking LLMs for code generation, evaluating vision-language models for image understanding, and comparing different RAG and knowledge ingestion strategies. The findings highlight practical trade-offs in accuracy, robustness, explainability, and user validation in real-world workflows.

How to cite: Sekas, C., Philippopoulos, K., Agathangelidis, I., Cartalis, C., Neophytides, S., and Mavrovouniotis, M.: Agentic AI for Earth-Observation-Driven Maritime Monitoring - the SeaScope Project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13454, https://doi.org/10.5194/egusphere-egu26-13454, 2026.

Large Language Models (LLMs) have emerged as powerful tools for text and data processing, with potential extending far beyond conversational interfaces. We demonstrate that integrating LLMs into agentic workflows enables automated climate and oceanographic data analysis while minimizing hallucinations through strict reliance on real data sources.

ClimSight combines LLMs with climate model data to deliver localized climate insights for decision-making. Specialized agents consult external databases, extract variables from climate models, generate Python scripts for post-processing, and validate outputs through visual analysis. The workflow iteratively corrects errors until reliable results are achieved.

PANGAEA GPT enhances accessibility to the PANGAEA data repository through a supervisor agent that interprets queries, delegates tasks to domain-specific subagents, and coordinates data extraction, statistical analysis, and visualization of oceanographic and atmospheric datasets.

Both systems leverage automatic Python execution and image analysis for quality control. By constraining outputs to verifiable data sources and implementing multi-agent verification, we demonstrate that LLMs can play a significant role in geoscientific data pipelines and automated research workflows.

How to cite: Kuznetsov, I., Pantiukhin, D., Grassi, J., Shapkin, B., Jung, T., and Koldunov, N.: Integrating Large Language Models into Climate and Geoscientific Data Workflows, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13612, https://doi.org/10.5194/egusphere-egu26-13612, 2026.

Generative AI is now woven into the daily study practices of geoscience students, often more deeply than educators acknowledge. This study examines how bachelor students in Earth Sciences (GEOL1008, NTNU) and master students in Engineering Geology (TGB4200, NTNU) use AI tools to understand literature, analyse data, synthesise research findings, and prepare written and oral assignments. The analysis draws on two structured surveys designed to map the extent and character of AI use in both cohorts.

Preliminary results indicate that AI has become the default support tool. Students turn to it to decode complex concepts, troubleshoot coding tasks, analyze data, structure reports, and polish presentations. Many see little distinction between traditional digital tools and generative AI, and the boundary between personal work and AI-augmented work is increasingly blurred. At the same time, students express uncertainty and worry about ethical expectations, disclosure practices, and the legitimacy of relying heavily on AI in academic work.

These trends have immediate consequences for assessment. Home exams, reports, and pre-prepared presentations no longer reliably reveal individual understanding, since nearly all students now use AI during preparation. Emerging evidence from portfolio-based courses suggests grade inflation and reduced differentiation between students, not because learning outcomes have improved, but because AI elevates the baseline quality of submitted work. In practice, written in-person exams and oral examinations remain among the few ways to assess unassisted reasoning.

The findings underscore a need to rethink teaching and assessment in geoscience education. AI is not a future challenge but a present reality, and universities must adapt if they aim to evaluate what students actually know rather than what their tools can produce.

How to cite: Fredin, O.: Geoscience Education in the Age of Generative AI: What Do Students Actually Learn?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16861, https://doi.org/10.5194/egusphere-egu26-16861, 2026.

Large Language Models (LLMs), a class of contemporary artificial intelligence systems, are increasingly used in scientific practice to support research workflows, accelerate discovery, and automate routine administrative tasks. This contribution identifies and analyzes three underexplored aspects of LLM adoption in scientific research. The first aspect concerns the uneven adoption of LLMs among scientists and the inconsistent application of established best practices. The second examines how LLMs can be employed to improve the robustness and reproducibility of scientific practices. The third addresses institutional strategies by which large scientific organizations—such as universities and research networks—can reduce dependence on commercial technology providers while increasing trust in LLM-based systems.

The findings found in our contribution are the partly summarization of István Bozsó’s experiences serving in the role of an “AI ambassador” at the Institute of Earth Phyisics and Space Science (EPSS) of the Hungarian Research Network (HUN-REN).

In our experience many scientists are still skeptical of using LLMs in any capacity or lack the time to invest in learning these technologies. These barriers are primarily sociotechnical rather than purely technical in nature, and they require, on one hand materials that teach best-practices and show motivating examples for using LLMs, on the other hand services provided by research organisations.

Recent advances in open-weight LLMs enable self-hosting within institutional computing infrastructures, which means research institutes can run these models on their own hardware and thereby ensuring that sensitive data and research materials remain within the organization’s controlled digital environment. This also ensures that the LLM usage stays independent of Large Technology corporations and builds trust with colleagues.

Regarding motivating examples, we wish to focus on two areas which can be addressed with the help of LLMs. One area is scientific communication. LLMs can easily generate materials (primarily text, sound and video) which can be used to inform the wider public on new scientific discoveries and push back against misinformation and disinformation campaigns. The involment of scientists is paramount in the review and finalization of such materials to ensure they represent accurate scientific information.

The other area is scientific programming. Many scientists are not trained as professional software engineers and often lack the time and background to apply software development best practices. In many cases this results in software artifacts that are fragile, difficult to reproduce, and challenging to maintain and usually only work on the machine of the researcher who developed the package. LLMs can help out in these situations by suggesting and even implementing best practices and giving programming advice to the researcher during the development of the scientific code.

The common theme in these examples is that the LLM is not meant to replace the scientist but enhance their capabilities with the goal of increasing the robustness, transparency, and sustainability of the scientific research process.

How to cite: Bozsó, I., Horváth, A., and Kuslits, L.: Bridging the gap between scientists and large language models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17176, https://doi.org/10.5194/egusphere-egu26-17176, 2026.

Urban public spaces are highly dynamic systems where traffic patterns, pedestrian flows, and human activities vary strongly across temporal scales. Capturing these dynamics at high temporal resolution remains challenging, particularly using low-cost and reproducible observation methods. In this study, we present an automated workflow for continuous urban activity monitoring based on publicly available webcam imagery and deep learning–based object detection.

A public webcam overlooking Augustusplatz, a central urban square in Leipzig (Germany), is continuously accessed, and still frames are extracted from the video stream at one-minute intervals. Each frame is processed using the YOLO11 object detection model to identify and count relevant object classes, including passenger vehicles and pedestrians. The detection results are converted into structured JSON records and enriched with metadata such as timestamp and geographic location. All data are stored in an InfluxDB time-series database and visualized and statistically analyzed using Grafana.

This setup enables near-real-time and long-term analysis of urban activity patterns across multiple temporal scales. Distinct signatures of recurring and episodic events can be identified, including daily commuting cycles, evening rush hours, road closures, public celebrations, and large seasonal events such as Christmas markets. The minute-scale resolution allows for detailed investigation of short-term dynamics, while continuous operation over longer periods enables comparative and trend analyses.

The presented approach demonstrates how publicly available visual data and open-source tools can be combined into a scalable and transferable framework for urban monitoring. Potential applications include event detection, urban mobility analysis, validation of traffic models, assessment of public space usage, and integration with other environmental or socio-economic datasets. The method provides a cost-efficient complement to traditional urban sensing infrastructures and offers new opportunities for data-driven urban and environmental research.

How to cite: Oesen, B., Wagner, R., and Goblirsch, T.: High-Temporal-Resolution Urban Activity Monitoring Using Public Webcams and Deep Learning: A Case Study from Leipzig, Germany, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17514, https://doi.org/10.5194/egusphere-egu26-17514, 2026.

Climate change, land-use change, and increasing socio-economic pressures are reshaping water and environmental systems, while the volume and heterogeneity of available data—from in situ observations to reanalysis products, remote sensing, and citizen-generated sources—continue to grow. Machine learning (ML) has become an important component of hydro-environmental modelling for forecasting, classification, and pattern discovery. However, in practice, many ML applications remain highly case-specific and dependent on implicit expert decisions related to problem formulation, predictor selection, validation design, and interpretation, which are rarely made explicit or transferable across regions and users.

This contribution presents a human-in-the-loop hybrid intelligence framework that integrates ML workflows with Large Language Models (LLMs) to support structured reasoning during environmental model development and evaluation. Rather than using LLMs for automated optimisation or model selection, the framework positions them as a guidance and scaffolding layer that helps make modelling assumptions, choices, and limitations explicit and traceable, while retaining expert control over all final decisions.

Methodologically, the framework combines (i) hands-on ML pipelines, ranging from baseline statistical models to more advanced learning algorithms for forecasting and classification, and (ii) an LLM-based guidance layer that structures expert reasoning through prompts, checklists, and decision logs. This guidance supports key stages of the modelling process, including the definition of modelling objectives, assessment of data quality, selection of environmentally meaningful predictors, and the design of validation strategies. Particular emphasis is placed on encouraging validation schemes that account for temporal dependence and spatial heterogeneity, such as blocked or spatial cross-validation, rather than default random data splits.

The framework is currently being developed and iteratively evaluated through expert-led case studies using real hydro-environmental datasets, rather than through formal classroom deployment. Initial applications focus on groundwater level analysis and hydro-environmental forecasting problems in Greece, including collaborative work in Crete, where the framework has been used to structure modelling choices and interpret model behaviour under non-stationary conditions. Additional exploratory applications using existing datasets have been used to stress-test the transferability of the workflow across contrasting environmental settings. Ongoing extensions include the application of the framework within coastal erosion modelling activities currently being developed in Colombia.

The LLM layer supports explicit reasoning about why a model performs well or poorly under specific conditions, how assumptions propagate into uncertainty, and where data-driven learning diverges from physical expectations. This reflective use of hybrid intelligence helps expose failure modes and modelling sensitivities that are often hidden in automated pipelines.

Results from the expert-led evaluations indicate that the proposed framework improves the transparency and reproducibility of modelling decisions, facilitates comparison across case studies, and supports more consistent interpretation of ML results across regions and scales. At the same time, the approach lowers the entry barrier for non-specialists without removing expert oversight or domain judgement.

The framework is being developed within the context of the Erasmus+ AI-LEARN project (Project reference: 2025-1-NL01-KA220-HED-000355215), where it serves as a methodological backbone for future training and capacity-building activities in water and environmental intelligence.

How to cite: Corzo P, G. A., Varouchakis, E., Kamińska-Chuchmała, A., Agioutanti, R., and Dominguez, V.: Integrating Machine Learning and Large Language Models for Next-Generation Water & Environmental Intelligence, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18439, https://doi.org/10.5194/egusphere-egu26-18439, 2026.

The increasing availability of satellite remote sensing data has made automatic land cover change detection a persistent research focus. However, real-world applications show that single AI models struggle to cope with the combined challenges of spatial-temporal complexity, feature diversity, and evolving engineering requirements. Consequently, the accuracy of automatically extracted land cover changes is often compromised, making the results insufficient for direct engineering application. Guided by practical application need, this paper focuses on how to utilize satellite remote sensing data, various knowledge and AI technologies to improve the accuracy and efficiency of automatic land cover extraction. This research focuses on the key technologies involved in the complete land cover monitoring process. Central to this study is the proposal of a progressive intelligent change detection technology for satellite remote sensing, characterized by a “identify all, discriminate precisely, refine extraction” workflow. Specifically, the “identify all” step extracts all potential change patches using models such as generic binary change detection. Building on these results, the “discriminate precisely” step filters out patches that are not of current interest. Finally, the “refine extraction” step employs models like semantic segmentation to further screen the results and enhance overall accuracy. An application demonstration in Shanxi Province, China, for new PV facilities, buildings, and roads demonstrated a recall rate of 89.3% for automatic extraction. The high-quality outputs confirm the practical applicability of the results. Consequently, this research affirms the technology as both a valuable and transferable solution for land cover monitoring.

How to cite: You, S., Du, L., He, Y., and Ye, F.: Research on Key Technologies for High-Precision Land Cover Change Monitoring Using Satellite Data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19954, https://doi.org/10.5194/egusphere-egu26-19954, 2026.

This study explores the integration of EVE (Earth Virtual Expert), a Large Language Model specialized in Earth Observation (EO) and Earth Sciences, developed under ESA’s Φ-lab #AI4EO initiative in collaboration with Pi School. The primary objective is to enable EVE to connect ESA’s EO platforms and data clusters, creating an integrated ecosystem for the community. This approach leverages agentic capabilities, allowing EVE to dynamically interact with EO tools, databases, and APIs to reason and act autonomously.
To demonstrate this concept, we present a use case where EVE operates within an agentic framework to interact with the EO Dashboard, a joint initiative by ESA, NASA, and JAXA that provides global indicators and narratives derived from multi-mission EO data. Using the MCP protocol, this work enables dynamic connectivity between EVE and the Dashboard, allowing the model to interpret and summarize narratives, extend insights with additional context, and facilitate advanced information retrieval across datasets and stories. In addition, the study considers potential directions for agentic behaviors, assessing early-stage possibilities and limitations for features such as autonomous task chaining. These capabilities enable EVE to perform multi-step reasoning, for example, by interpreting quantitative trends in dashboard indicators such as air quality changes, greenhouse gas concentrations, or land cover dynamics. This links EVE to underlying datasets and enables the generation of scientifically grounded responses. This proof-of-concept demonstrates EVE’s potential to foster interoperability and accelerate Earth system science by improving knowledge accessibility and enabling more effective use of EO data resources.

How to cite: Gmelich Meijling, E., D'Ercole, R., Anghelea, A., Cocchiara, C. M., and Longepe, N.: Building Connected Earth Observation Ecosystems with Agentic AI using EVE, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20206, https://doi.org/10.5194/egusphere-egu26-20206, 2026.

The scale and heterogeneity of modern geospatial datasets, coupled with expanding suites of statistical and dynamical models, produce analysis outputs that are increasingly difficult to navigate and synthesise. We present a practical case study on using a large language model (LLM)-assisted coding tool (GitHub Copilot with GPT-5 mini within Visual Studio Code) to accelerate the development of a lightweight, HTML-based platform that visualises results from pre-calculated climate indicators.

Our starting point was a dataset comprising more than 130 climate indicators derived from gridded observations spanning over 60 years. These indicators originate from multiple meteorological variable groups (e.g., temperature, precipitation) and are aggregated at several temporal resolutions (e.g., annual, seasonal). Downstream analyses include spatiotemporal  statistics, extreme value analyses and statistical significance testing, yielding hundreds of figures that are difficult to navigate and analyse. To make these outputs tractable, we prompted Copilot to generate a simple web application for visualisation and analysis purposes. The pre-generated plots from the climate indicator workflow were displayed there in an organised way, allowing for quick filtering through all indicators and different temporal resolutions, comparing different plots next to each other and using a subpage to concisely display aggregated group plots.

The platform is embedded and deployed via a GitLab CI pipeline, ensuring reproducible updates and immediate web accessibility for collaborators and users, thereby enabling rapid and easy access to vasts amount of output results. Our process of prompting a LLM to generate a visualisation platform offers a convenient and transferable workflow to aid geospatial data analysis.

How to cite: Lehner, S. and Schlögl, M.: Using copilot for the rapid generation of a visualisation platform to aid geospatial analyses, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20992, https://doi.org/10.5194/egusphere-egu26-20992, 2026.

Sub-grid parameterisations in climate models are traditionally static and tuned offline, limiting adaptability to evolving states. This work introduces FedRAIN-Lite, a federated reinforcement learning (FedRL) framework that mirrors the spatial decomposition used in general circulation models (GCMs) by assigning agents to latitude bands, enabling local parameter learning with periodic global aggregation. Using a hierarchy of simplified energy-balance climate models, from a single-agent baseline (ebm-v1) to multi-agent ensemble (ebm-v2) and GCM-like (ebm-v3) setups, we benchmark three RL algorithms under different FedRL configurations. Results show that Deep Deterministic Policy Gradient (DDPG) consistently outperforms both static and single-agent baselines, with faster convergence and lower area-weighted RMSE in tropical and mid-latitude zones across both ebm-v2 and ebm-v3 setups. DDPG's ability to transfer across hyperparameters and low computational cost make it well-suited for geographically adaptive parameter learning. This capability offers a scalable pathway towards high-complexity GCMs and provides a prototype for physically aligned, online-learning climate models that can evolve with a changing climate. Code accessible at https://github.com/p3jitnath/climate-rl-fedRL.

How to cite: Nath, P., Schemm, S., Moss, H., Haynes, P., Shuckburgh, E., and Webb, M.: FedRAIN-Lite: Federated Reinforcement Algorithms for Improving Idealised Numerical Weather and Climate Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-351, https://doi.org/10.5194/egusphere-egu26-351, 2026.

This study presents a data-driven parameterization of deep convection, implemented and tested within the global climate model ARP-GEM at 50 km resolution. Initially, a 'naive' neural network was used to replace ARP-GEM's traditional physical parameterization. A 30-year simulation with this data-driven approach revealed significant biases, particularly in the representation of high clouds.
To adress these biases, we developed a two-fold neural network architecture: one component responsible for detecting the triggering of the convection and another responsible for computing convective tendency terms. This refined parameterization substantially improved performance compared to the initial version. Furthermore, the enhanced parameterization was evaluated under warmer climate conditions, demonstrating online stability and consistent overall fidelity.

How to cite: Balogh, B., Germain, H., Geoffroy, O., and Saint-Martin, D.: Online test of a data-driven parameterization of deep-convection: evaluation in present and future climate, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1118, https://doi.org/10.5194/egusphere-egu26-1118, 2026.

Earth System Models of Intermediate Complexity (EMICs) are essential tools for investigating climate dynamics on millennial to orbital time scales, which are computationally prohibitive for high-resolution CMIP-class models. The computational efficiency of EMICs is primarily achieved by reduced spatial resolution of the atmosphere and ocean components. However, EMICs often couple ice-sheet and terrestrial vegetation components, which require much higher spatial resolution. The coupling of these components therefore remains a major challenge and often results in inadequate climatic forcing for these sub-modules, particularly regarding precipitation patterns. Generative machine learning, specifically diffusion models and their variants, has emerged as a powerful technique to bridge this resolution gap. Here, we present the integration of a consistency model-based approach to facilitate efficient, online downscaling of temperature and precipitation within the Bern3D EMIC with negligible computational overhead.

To achieve this, the consistency model was trained on monthly ERA5 ensemble output to learn the mapping from the coarse Bern3D grid to high-resolution fields.  This approach successfully reconstructs high-resolution spatial variability while maintaining inference speeds compatible with long model integration times, effectively avoiding additional runtime costs. This framework therefore allows for the representation of small-scale heterogeneity in surface boundary conditions which is critical for realistic ice sheet and vegetation dynamics. 

Ultimately, this approach opens new avenues to investigate complex climate-ice-vegetation feedback on orbital time scales, such as during the Last Glacial Cycle or the Mid-Pleistocene Transition.

How to cite: Wirths, C., Hofmann Elizondo, U., Hess, P., and Pöppelmeier, F.: Towards Generative Machine Learning-based Downscaling for Atmosphere-Surface Coupling in the Bern3D EMIC, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2982, https://doi.org/10.5194/egusphere-egu26-2982, 2026.

In weather and climate models, momentum, heat, humidity and tracer fluxes between the Earth’s surface and atmosphere strongly depend on surface roughness. The roughness length depends on space and time-dependent surface properties over ocean, sea-ice and land. For example, surface winds impact wave height over sea-ice free oceans; vegetation and orography determine roughness length over land, where its effect on near-surface turbulence strongly impacts the surface fluxes. Here, we present a set of machine learning models trained on reanalysis data to predict surface roughness over both land and ocean grid cells in SpeedyWeather, a Julia-based climate model. More accurately representing the surface roughness has been shown to significantly improve model bias against observations over a range of variables such as surface air temperatures and near-surface wind speed. We explore the downstream impacts of using this parameterisation in the climate model, and test the generalisability of an offline-learned surface roughness scheme in future climates with reduced sea ice and land-use change. Spatial generalisation is achieved through surface roughness being a function of local variables only. We discuss efficient inference on CPU and GPU for every grid cell on each integration time-step. So-called model distillation via symbolic regression minimises the trade-off between speed versus accuracy, enabling another route to rapid inference on a grid-cell basis. Further, we investigate online learning through differentiable physics parameterisations to calibrate the learned parameterisation to surface variables from ERA5 reanalysis. We generally propose machine-learned schemes of individual climate processes towards interpretable, data-driven climate modelling. 

How to cite: Munday, G., Klöwer, M., Mansfield, L., and Gelbrecht, M.: A learned surface roughness scheme for climate prediction, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4851, https://doi.org/10.5194/egusphere-egu26-4851, 2026.

Hybrid Earth system models (ESMs) that combine physical laws with machine learning (ML) demonstrate great potential to reduce uncertainties in climate projections, particularly for subgrid processes like clouds. However, widespread adoption faces critical challenges: deep learning "black boxes" often lack interpretability and physical consistency, and coupling them with standard ESMs remains difficult due to stability issues and the need for complex re-calibration. Here, a two-step method is presented to improve a climate model with data-driven parameterizations. First, we incorporate a physically consistent cloud cover parameterization, derived from storm-resolving simulations via symbolic regression, into the ICON atmospheric climate model. We refer to this hybrid configuration, which retains the interpretability and efficiency of the traditional model, as ICON-A-MLe. Second, we address the coupling and tuning bottleneck by introducing an automated, gradient-free calibration procedure based on the Nelder-Mead algorithm. This method efficiently calibrates ICON-A-MLe without requiring differentiable physical components, making it easily extendable to other ESMs. Our results show that the tuned ICON-A-MLe substantially reduces long-standing biases. Specifically, it reduces cloud cover errors over the Southern Ocean by 75% and in subtropical stratocumulus regions by 44%. These improvements also lead to a better top-of-atmosphere radiative budget. Crucially, the model demonstrates strong generalization capabilities: it remains robust and physically consistent under significantly warmer climate scenarios. These results demonstrate that interpretable machine-learned parameterizations, paired with practical tuning, can efficiently and transparently strengthen ESM fidelity.

How to cite: Grundner, A., Beucler, T., Savre, J., Lauer, A., Schlund, M., and Eyring, V.: Reduced cloud cover errors in a hybrid AI-climate model through equation discovery and automatic tuning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5525, https://doi.org/10.5194/egusphere-egu26-5525, 2026.

Accurately simulating the terrestrial carbon cycle remains a major challenge in climate science, due in part to uncertainties in how slow-varying land-surface boundaries and fast-varying biophysical states are represented and coupled in Earth-system models.  We introduce a unified data-driven framework designed to generate high-resolution (1 km) historical reconstructions and future projections of Land Use (LU), Land Cover (LC), and Leaf Area Index (LAI) for real-time coupling with digital twin platforms, such as those deployed in the Destination Earth framework.

Moving beyond sequential downscaling, this framework treats the generation of boundary conditions as a cohesive multi-task learning problem. We benchmark two distinct modeling strategies: (1) Architectures trained from scratch, where we compare the performance of convolutional baselines (U-Net) against attention-based Vision Transformers (ViT) in capturing spatial heterogeneity; and (2) Foundation Model (FM) Adaptation, where we leverage state-of-the-art Earth FMs (such as  TerraMind and Prithvi) as backbones. Within this second strategy, we evaluate the trade-offs between full fine-tuning, parameter-efficient techniques using adapters, and models trained from scratch.

By integrating static geophysical features with high-frequency climate reanalysis (ERA5) and atmospheric CO2​ concentrations, the framework ensures that vegetation dynamics remain phenologically consistent with environmental forcing. We assess these approaches based on their computational efficiency, generalization across sparse data regimes, and physical consistency between categorical (LU/LC) and continuous (LAI) variables. The final output is a suite of open-source interoperable emulators designed to act as dynamic, on-demand boundary condition generators. 

How to cite: Duarte, A., Mozaffari, A., Castaño, M., Materia, S., and Castrillo Melguizo, M.: A Unified Data-Driven Framework for High-Resolution Land Surface Boundary Conditions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9688, https://doi.org/10.5194/egusphere-egu26-9688, 2026.

Global land surface and hydrological models are crucial components of Earth System Models (ESMs). In addition to providing realistic boundary conditions for the atmosphere and ocean components, they also play a key role in understanding Earth’s changing energy imbalance and the response of the terrestrial carbon and water cycles to anthropogenic climate change. The land surface components of most ESMs typically rely on reduced-complexity parameterizations of land processes in order to efficiently resolve the transient coupling of the land surface to the atmosphere at global scales. The complexity of such models is therefore limited by the coarse spatial resolution of the atmosphere and thus they are not easily constrained by in situ and remote sensing observations of land surface parameters. As a result, offline downscaled and bias-corrected climate models and reanalysis products are often used as forcings when calibrating land surface and hydrological models at local and regional scales. We argue that this lack of online coupling in the downscaling step is one of many factors contributing to persistent biases in modern ESMs. As such, there is a need for a new generation of land models which can support more flexible coupling with the atmosphere as well as the incorporation of data-driven components. Here we present Terrarium.jl, a Julia-based land modeling framework for GPU-accelerated and automatically differentiable simulations of soil, snow, and vegetation dynamics, along with their corresponding land-atmosphere exchange fluxes. We demonstrate the value of GPU acceleration and differentiability through a series of performance benchmarks and sensitivity analyses. We further present our initial experiments in achieving stable coupling to a reduced-complexity atmosphere model, SpeedyWeather.jl, as well as a proof-of-concept for online downscaling from the scale of an intermediate-complexity ESM (~5°) to that of ERA5 (~0.25°). We discuss the main challenges encountered thus far and outline a roadmap for future development.

How to cite: Groenke, B., Badri, M., Lin, Y., Gelbrecht, M., and Boers, N.: Terrarium.jl: A framework for fully differentiable and GPU-accelerated land modeling to enable online downscaling in coarse-scale ESMs, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11235, https://doi.org/10.5194/egusphere-egu26-11235, 2026.

Subgrid-scale (SGS) parameterizations remain a leading source of uncertainty in weather and climate models, where they represent the effects of unresolved processes occurring at scales smaller than the model’s grid resolution on the resolved fields. Similar closure problems arise in computational fluid dynamics (CFD), where turbulence models are needed to represent the impact of unresolved scales on the resolved flow. Despite recent progress toward differentiable, hybrid climate models enabled by automatic differentiation, most operational Earth system models (ESMs) remain effectively non-differentiable, limiting systematic online calibration and training.

While data-driven closures trained offline can perform well a priori, their performance often deteriorates a posteriori, once coupled to the solver because the coupled setting introduces feedbacks that are absent during offline training. In this study, we treat a controlled CFD turbulence setting as a benchmark for climate-relevant SGS learning and compare two classes of online training strategies for data-driven closures in non-differentiable models: (i) gradient-free ensemble Kalman inversion (EKI), leveraging the robustness and parallelism of ensemble-based inverse methods, and (ii) gradient-based optimization enabled by a learned differentiable emulator. For the emulator, we train a fast neural ODE surrogate of the forward model dynamics that preserves its structure and is differentiable by construction, enabling gradient-based training without modifying the original solver. We then evaluate both approaches using metrics such as accuracy, computational cost, and scalability.

How to cite: Badri, M., Groenke, B., Gelbrecht, M., and Boers, N.: Comparison of Online Training Methods for Data-Driven Subgrid-Scale Parameterizations in Non-Differentiable Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11475, https://doi.org/10.5194/egusphere-egu26-11475, 2026.

The Pelagic Interaction Scheme for Carbon and Ecosystem Studies (PISCES) is a marine biogeochemical model that is used in several IPCC-Class Earth System models. PISCES simulates the distribution of nutrients (four macronutrients and one micronutrient) that regulate the growth of two phytoplankton classes (nanophytoplankton and diatoms). It also simulates the ocean carbon cycle with a complete representation of the marine carbonate systems. PISCES includes 24 state variables, and increases the runtime of NEMO, the physical ocean model with which it is coupled, by a factor of 3.4, indicating a high computational cost.

PISCES-AI has been developed as a U-Net based machine learning PISCES emulator, which takes a small number of input variables (TOS, ZOS, SOS, PAR, atmospheric CO₂), and predicts two output variables: surface chlorophyll and the difference in partial pressure of CO₂ between the atmosphere and the ocean. These are the only outputs which have a direct influence on climate simulations by Earth system models. Previous work has shown the predictive power of PISCES-AI across multiple timescales, and in an out-of-domain setting.

In this work, we couple the AI emulator of PISCES to NEMO, using Eophis and Morays for Python-Fortran interaction. We evaluate its performance, as well as its computational efficiency, to give a holistic picture of the challenges and opportunites for AI emulation of ocean biogeochemistry. With a particular interest in the computational speed, we find that inference for a single time-step to be around 10ms, with a much larger preliminary bottleneck due to CPU-GPU transfer (200ms per timestep). Even with this bottleneck, with our implementation we obtain a speed-up of factor 3 compared to PISCES, and we explore ways in which the data transfer bottleneck could be reduced.

How to cite: Gow-Smith, E. and Séférian, R.: Coupling of NEMO to a neural network emulator of PISCES, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12174, https://doi.org/10.5194/egusphere-egu26-12174, 2026.

Reliable high resolution precipitation fields are essential for hydrology flood risk management agriculture and climate impact assessment yet remain difficult to reconstruct from sparse and irregular rain gauge networks. Reanalysis products such as ERA5 provide physically consistent estimates but are constrained by coarse effective resolution temporal smoothing and weak local observational constraints. By formulating interpolation of spatiotemporal precipitation fields as a probabilistic context to target regression problem using neural process (NP) models, this study assesses whether NP-based approaches can outperform reanalysis and classical interpolation for local to regional rainfall reconstruction. Using high quality UK rain gauge observations combined with gridded auxiliary variables from ERA5 we implement convolutional NPs within the DeepSensor framework and compare them with a transformer based NP variant.

Models are jointly conditioned on dense meteorological fields and sparse precipitation observations and output full predictive distributions using a Bernoulli–Gamma likelihood to capture intermittency and extremes. Training is performed using random sensor masking to enforce location agnostic learning and enable zero shot prediction at unseen coordinates. Performance is evaluated against ERA5 and Kriging using identical data splits with emphasis on interpolation accuracy as well as calibration robustness to sensor sparsity. Generalisation is further assessed through few shot and zero shot transfer across regions with contrasting regimes including England, Scotland and selected GHCN domains in the US.

Using NPs, this work aims to recover sharper spatial structure with improved uncertainty calibration and higher frequency precipitation estimates relative to ERA5 under sparse observation scenarios and also evaluates their potential as robust uncertainty aware additions to physics-based models for high resolution environmental monitoring.

How to cite: Pazola, A., Ladopoulou, D., Morris-Wiltshire, C., Nath, P., and Coca-Castro, A.: Learning Spatiotemporal Precipitation Fields with Probabilistic Neural Processes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12294, https://doi.org/10.5194/egusphere-egu26-12294, 2026.

The integration of machine learning parameterisations within climate models is paving the way for the next generation of Earth System models. Machine learning parameterisations are being developed to represent ocean and atmosphere processes such as turbulence, vertical mixing, and cloud and precipitation. These parameterisations typically require large volumes of high-resolution data for their training. This training data is often derived from the same numerical model that the parameterisation is intended for. This has the advantage that the machine learning model is only exposed to one set of numerical discretisation schemes.

Recently, global km-scale models have been introduced that simulate climate processes at remarkable detail. Explicitly resolving mesoscale and sub-mesoscale eddies and filaments enables these models to capture heat, carbon, and salt fluxes without the need for parameterisations. Global km-scale models are therefore promising training data sets for machine learning parameterisations.

In this work we intend to examine two global km-scale models that could be employed for oceanic turbulence parameterisations: NEMO ORCA36 and ICON-O. The ORCA36 model uses the tripolar grid and ICON-O uses an icosahedral grid. The question is, can either model be used to inform new ML parameterisations that can be employed in any numerical model? Therefore, a key assessment of these models will be done by exploring and contrasting their energetics, as well as the heat, salt, and carbon transports. This work will take the first step towards model-independent machine learning parameterisation development, while facilitating further cross modelling centre collaboration.

How to cite: Wilder, T. and Li, H.: Towards model-independent machine learning parameterisations of meso-scale eddies, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12770, https://doi.org/10.5194/egusphere-egu26-12770, 2026.

Representing and quantifying uncertainty in physical parameterisations is a central challenge in weather and climate modelling, and approaches are often developed separately for different timescales. Here, we consider the separation of uncertainty by source using machine learning frameworks for subgrid-scale parameterisations. In this context, aleatoric uncertainty arises from internal variability in the training data, and epistemic uncertainty, arises from poorly constrained parameters during training. Using the Lorenz 1996 system as a testbed for simplified chaotic dynamics, we deal with uncertainties through a unified framework using Bayesian Neural Networks, to explore how the different sources of uncertainty evolve over different prediction timescales.

How to cite: Mansfield, L. and Christensen, H.: Separating Epistemic and Aleatoric Uncertainties in Weather and Climate Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13183, https://doi.org/10.5194/egusphere-egu26-13183, 2026.

In an Atmospheric General Circulation Model (AGCM), the representation of subgrid-scale physical phenomena, also referred to as physical parameterizations, requires computational time which constrains model numerical efficiency. However, the development of emulators based on Machine Learning offers a promising alternative to traditional approaches.

We have developed offline emulators of the atmospheric component named ICOLMDZ (for DYNAMICO and LMDZ) of the IPSL climate model, in an idealized aquaplanet configuration, with the aim of emulating all the parameterizations, i.e. the LMDZ atmospheric physics component. While the results are quite promising, some fundamental questions are raised, particularly in terms of the generalization of the emulation process to meteorological conditions not seen by the emulator. This step is important for adopting the emulator as a substitute for traditional parameterizations.

This question of generalization, which relates to the ability of emulators to infer and adapt to new system states, has been studied in experiments linked to climate change. Indeed, we first investigated the performance of our emulators, trained on an aquaplanet configuration, in extrapolating the emulation process to new aquaplanets where boundary conditions are modified in order to simulate climates that are warmer and colder than the climate on which emulators are trained. The results reveal the potential of our aquaplanet emulators to reproduce the physical parameterizations of new climates. However, we also showed the limitations of these aquaplanet emulators since they encountered difficulties to generalize on a realistic configuration, i.e. when continents, topography and sea ice area are included.

This study encourages the coupling of emulators with the dynamic parts called DYNAMICO in order to better assess the relevance of the learning process, while analyzing the stability of the simulations obtained.

How to cite: Crossouard, S., Kageyama, M., Vrac, M., Dubos, T., Thao, S., and Meurdesoif, Y.: Generalization ability of emulators in reproducing the physical parameterizations of the IPSL model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17216, https://doi.org/10.5194/egusphere-egu26-17216, 2026.

Atmospheric ozone is a crucial absorber of solar radiation and an important greenhouse gas. However, most climate models participating in the Coupled Model Intercomparison Project (CMIP) still lack an interactive representation of ozone due to the high computational costs of atmospheric chemistry schemes. Here, we introduce a machine learning parameterization (mloz) to interactively model daily ozone variability and trends across the troposphere and stratosphere in common CMIP simulations, including pre-industrial, abrupt-4xCO₂(Ma et al. 2025), historical and future Shared Socioeconomic Pathway (SSP) scenarios simulations. We demonstrate its high fidelity on decadal timescales and its flexible use online across two different climate models -- the UK Earth System Model (UKESM) and the German ICOsahedral Nonhydrostatic (ICON) model. With meteorological variables and forcing data as inputs, mloz produces stable ozone predictions around 31 times faster than the chemistry scheme in UKESM, contributing less than 4% of the respective total climate model runtimes. In particular, we also demonstrate its transferability to different climate models without chemistry schemes by transferring the parameterization from UKESM to ICON in standard climate sensitivity simulations. This highlights mloz’s potential for widespread adoption in CMIP-level climate models that lack interactive chemistry for future climate change assessments, where ozone trends and variability will significantly modulate atmospheric feedback processes.

Reference:
Ma Y, Abraham N L, Versick S, et al. mloz: A Highly Efficient Machine Learning-Based Ozone Parameterization for Climate Sensitivity Simulations[J]. arXiv preprint arXiv:2509.20422, 2025.

How to cite: Ma, Y., Abraham, N. L., Versick, S., Ruhnke, R., Schneidereit, A., Niemeier, U., Back, F., Braesicke, P., and Nowack, P.: mloz: A Highly Efficient Machine Learning-Based Ozone Parameterization for CMIP Simulations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17576, https://doi.org/10.5194/egusphere-egu26-17576, 2026.

Are AI-powered climate models intrinsically more efficient than traditional climate models?

While progress is still needed before they become operational, hybrid AI-physics climate models and AI emulators of climate models have the potential to sharply reduce inference cost relative to traditional CPU-based models, allowing larger ensembles to explore different scenarios and sharpen uncertainty estimation. Yet this apparent efficiency becomes less obvious when the comparison includes GPU-ported dynamical climate models, and when efficiency is assessed against the effective complexity of the simulated climate system.

As a first step, recognizing that a perfect apple-to-apple comparison is rarely possible from reported configurations, we synthesize reported performance for leading AI climate model emulators (e.g., ACE2, CAMulator), hybrid AI-physics models (e.g., CliMA, NeuralGCM), and GPU-accelerated traditional models (e.g., SCREAM, ICON). We examine two complementary scaling views. The first compares throughput (simulated years per day) per accelerator (GPUs or TPUs) and per prognostic variable, as a function of horizontal grid spacing. The second compares the same normalized throughput against an effective complexity proxy, defined as the number of vertical levels divided by the product of the time step and the squared horizontal grid spacing, to account for the simulated vertical structure and, importantly, time-step constraints imposed by numerical stability.

We find that AI-powered models can show favorable apparent scaling with horizontal resolution in raw throughput, but that the advantage becomes modest once effective complexity is accounted for: at comparable complexity, AI climate models do not appear intrinsically more efficient than GPU-ported dynamical models. Hybrid approaches occupy a distinct middle ground: their acceleration and added value come primarily from learned parameterizations that improve the representation of unresolved processes while the overall model retains a physically-based dynamical core, including explicit conservation laws. AI climate model emulators, by contrast, offer their clearest computational advantage through task-targeted prediction, where a limited set of climate-relevant variables can be directly simulated on the grid of interest. This avoids integrating the full high-frequency, multivariate state at the short time step traditionally required for numerical stability, which is especially advantageous when emulating a fine-resolution reference model with a coarser emulator. Diverse downscaling or targeted post-processing strategies can further substitute for explicit fine-scale resolution when observations are available, enabling inexpensive local or hazard-specific risk assessment at decadal to multi-decadal time horizons.

How to cite: Beucler, T., Neelin, D., Su, H., Bretherton, C., Chapman, W., Christopoulos, C., Grover, A., Lopez-Gomez, I., Schneider, T., Subel, A., Watt-Meyer, O., and Zanna, L.: Reassessing the Scaling of AI-Powered Climate Models Against Dynamical Counterparts, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18440, https://doi.org/10.5194/egusphere-egu26-18440, 2026.

Persistent systematic errors in Earth system models (ESMs) arise from difficulties in representing the full diversity of subgrid, multiscale atmospheric convection and turbulence. Machine learning (ML) parameterizations trained on short high-resolution simulations show strong potential to reduce these errors. However, stable long-term atmospheric simulations with hybrid (physics + ML) ESMs remain difficult, as neural networks (NNs) trained offline often destabilize online runs. Training convection parameterizations directly on coarse-grained data is challenging, notably because scales cannot be cleanly separated. This issue is mitigated using data from superparameterized simulations, which provide clearer scale separation. Yet, transferring a parameterization from one ESM to another remains difficult due to distribution shifts that induce large inference errors. Here, we present a proof-of-concept where a ClimSim-trained, physics-informed NN convection parameterization is successfully transferred to ICON-A. The scheme is (a) trained on adjusted ClimSim data with subtracted radiative tendencies, and (b) integrated into ICON-A. The NN parameterization predicts its own error, enabling mixing with a conventional convection scheme when confidence is low, thus making the hybrid AI-physics model tunable with respect to observations and reanalysis through mixing parameters. This improves process understanding by constraining convective tendencies across column water vapor, lower-tropospheric stability, and geographical conditions, yielding interpretable regime behavior. In AMIP-style setups, several hybrid configurations outperform the default convection scheme (e.g., improved precipitation statistics). With additive input noise during training, both hybrid and pure-ML schemes lead to stable simulations and remain physically consistent for at least 20 years, demonstrating inter-ESM transferability and advancing long-term integrability.

How to cite: Heuer, H., Beucler, T., Schwabe, M., Savre, J., Schlund, M., and Eyring, V.: Beyond the Training Data: Confidence-Guided Mixing of Parameterizations in a Hybrid AI-Climate Model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18986, https://doi.org/10.5194/egusphere-egu26-18986, 2026.

Ensuring that deep learning models generalize across distinct regimes remains a fundamental challenge in Earth system modeling. Due to the inherent violation of the independent and identically distributed (i.i.d.) assumption, models optimized for local conditions rarely exhibit robust performance on unseen domains. While Unsupervised Domain Adaptation (UDA) is a well-established technique for mitigating such distribution shifts in computer vision, its application to Earth system modeling remains underexplored. In this study we investigate the efficacy of UDA for the super-resolution of atmospheric fields, utilizing kilometer-scale COSMO simulations [1] and the RainShift benchmark dataset [2] to quantify model robustness across different regions. We apply residual learning to jointly super-resolve precipitation and surface pressure, incorporating static predictors such as topography. To quantify transferability, we propose a systematic framework that trains on source domains and evaluates on unseen target domains, treating spatial transfer as a proxy for model robustness under distribution shifts. We introduce a consistency metric to evaluate model adaptation by comparing mean performance on seen versus unseen domains. We assess a hierarchy of adaptation methods, ranging from simple regularization to physics-informed approaches. These include domain-specific regularization and distribution alignment methods, domain adversarial training, and geometry-robust training via group-equivariant convolutions. Preliminary results on the COSMO simulations demonstrate that even elementary adaptation strategies, such as dropout and data augmentation, improve cross-domain consistency. This work establishes a controlled setup for benchmarking generalization, suggesting that UDA offers a viable pathway to bridge the gap between locally trained models and global applicability.

[1]: Cui, R., Thurnherr, I., Velasquez, P., Brennan, K. P., Leclair, M., Mazzoleni, A., et al. (2025). A European hail and lightning climatology from an 11-year kilometer-scale regional climate simulation. Journal of Geophysical Research: Atmospheres, 130, e2024JD042828. https://doi.org/10.1029/2024JD042828

[2]: Paula Harder et al. RainShift: A Benchmark for Precipitation Downscaling Across Geographies. 2025. arXiv: 2507.04930 [cs.CV]. url: https://arxiv.org/abs/2507.04930.

How to cite: Quarenghi, F. and Beucler, T.: Transferring knowledge across regions: unsupervised domain adaptation for km-scale super-resolution, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19396, https://doi.org/10.5194/egusphere-egu26-19396, 2026.

Data driven parameterisations for sub-grid processes unlocks the ability to surpass the current computational constraints of Earth system models. However, machine learning (ML) can be brittle. State-of-the-art ML approaches can reliably perform on in-distribution data, exceeding human ability across a diverse range of tasks. Yet, when faced with shifts in data distribution, performance degrades. In climate modelling, when the task is predicting the state of a non-stationary system, this is evidently a substantial issue. We illustrate this with the ClimSim dataset, forming spatio-temporal groups and quantitatively show how even small shifts in distribution affect performance.

Next, we use the theory of compositional generalisation to build models which are less susceptible to these shifts in distribution. Compositional generalisation is the formation of novel combinations of observed elementary components. That is, the ability to decompose data into building blocks that are reused across both the in- and shifted-domains, such that a model can capture a domain shifted state through a set of in-domain, learnt abstractions. Inspired by these concepts we propose various architectural and regularisation changes to standard ML parameterisations to improve generalisation. Preliminary results in sub-grid process emulators suggest new insights into if and how CG can reduce model sensitivity to domain shifts.

How to cite: Stanley-Clamp, B., Posner, I., and Christensen, H.: Beyond In-Distribution Skill: Towards Robust ML Parameterisations for Non-Stationary Climate Systems, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20353, https://doi.org/10.5194/egusphere-egu26-20353, 2026.

Terrestrial Water Storage (TWS) represents all forms of water on land, including the cryosphere (polar ice sheets and mountain glaciers), the biosphere (canopies), soil and subsurface water (groundwater), and other inland water bodies (reservoirs, rivers, lakes, and wetlands). Modelling TWS remains a challenge because of difficulties in representing the water cycle on land. Furthermore, TWS is the source of mass-driven sea level change, an increasingly important contributor to sea level variation across the globe in the 20^th and 21^st centuries, with significant implications for coastal areas.

Here, we leverage the potential of machine learning and propose a Physics-Informed Neural Networks (PINNs) framework for modeling and predicting TWS and its associated sea level impacts. Because TWS varies in space and time, we build our framework based on convLSTM, an architecture that is suitable for serially-correlated “two-dimensional images” of data. The physical constraint for our PINNs is provided by the physics of continental-ocean mass redistribution, i.e., the sea level component, as described by the gravitationally self-consistent methodology of the sea level equation. The sea level equation connects TWS and sea level change by considering the gravitational, rotational, and deformational feedbacks caused by TWS components, particularly the cryosphere.  

We train and test our PINNs based on global TWS data from 1900 up to the end of 2018 (1900-2001 for training; 2002-2018 for testing). The data have a temporal resolution of 1 year and a spatial resolution of , and were derived from an assimilation of models and satellite gravimetry observations (in the time period 2002-2018). We perform various tests and discuss the advantages and shortcomings of our PINNs framework for modeling and predicting TWS. First, we show that TWS and sea level rise can be predicted reasonably well up to 10 years ahead (relative error of less than 30% on a global scale). This might prove useful for studies of sea level rise in coastal areas. Second, we compare our predictions with those of the Ice Sheet Model Intercomparison Project (ISMIP) in CMIP6 climate models, and satellite observations of Gravity Recovery and Climate Experiment (GRACE; in the range 2015-2024). We demonstrate that our predictions are closer to GRACE observations and, therefore, more accurate (up to 40% for the lead horizon of 10 years) compared to ISMIP projections under high and low emission pathways. Finally, we discuss how the predictions could be further improved by using probabilistic deep learning approaches, particularly  so-called deep ensembles. Our results show that once trained, PINNs can provide predictions orders of magnitude faster than climate models and with better accuracy.

How to cite: Kiani Shahvandi, M., Gasparini, B., and Voigt, A.: Physics-informed neural networks predict changes in terrestrial water storage and sea level, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-594, https://doi.org/10.5194/egusphere-egu26-594, 2026.

Generating high-resolution (HR) weather and climate information at ~10 km or finer across the Himalayan regions remains a major challenge due to extremely high computational cost of forecasting models and complexity of atmospheric processes. Most operational global weather prediction systems run at low-resolution (LR) of ~25 km or coarser, that are inadequate for impact-based analyses of highly localized extreme weather events common to these regions. To bridge this gap, downscaling is essential for producing climate information at impact-relevant scales, with both statistical and dynamical approaches remaining widely used despite major shortcomings. The former is computationally efficient but often fail under future climate non-stationarity, while the latter, though physically consistent, is computationally expensive and constrained by domain-resolution trade-offs. Currently, there is no efficient data-driven approach that can produce regional-model-scale precipitation fields for the Himalayan region. This work presents WGAN, a deterministic deep neural generative adversarial network (GAN)-based emulator of the Weather Research and Forecasting (WRF) model for HR precipitation downscaling over the Himalayan region. The model is conditioned on LR meteorological variables from the European Centre for Medium-Range Weather Forecasts Re-Analysis version 5 (ERA5; 0.25°×0.25°) as input and is trained against HR precipitation from WRF (0.1°×0.1°), which uses ERA5 as boundary conditions. The architecture uses Wasserstein-1 distance (WGAN) in the generator and critic value functions with a gradient penalty for stable training. WGAN demonstrated the ability to generate fine-scale precipitation fields that closely matches WRF’s outputs, accurately capturing spatial patterns and the mean values. Incorporating terrain and an extreme aware-weighting MSE (Mean Squared Error) loss function in the model further improves precipitation magnitude representation, reduces biases, and yield ~29% reduction in RMSE in the upper decile. The model effectively captured low-frequency (large-scale) variability and better matches WRF’s power spectrum at mid-high frequency (short-scale) variability. This raises the probability of detection and lowers the false alarm rate across thresholds. With a case study, WGAN showed the ability to capture the fine-scale spatial distribution of precipitation in the mountains and foothills, at both extreme precipitation day and dry conditions, outperforming CNN-based precipitation output. These results underscore the capability of WGAN as a fast and efficient tool for precipitation downscaling for the Himalayan region, operating at only a fraction of the computational cost. The model has strong potential for operational use in early warning, risk assessment, vulnerability analysis, disaster management, and other sectors that rely on localized climate information, ultimately supporting the preparedness of communities living in and around these mountains.

How to cite: Lyngwa, R., Nikumbh, A., and Ghosh, S.: A Generative-driven Model for Precipitation Downscaling Over Himalayan Region, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-821, https://doi.org/10.5194/egusphere-egu26-821, 2026.

Exploring uncertainty and internal variability across future emission pathways remains computationally demanding with state-of-the-art Earth system models (ESMs). We present a diffusion-based machine-learning emulator trained on output from the CESM2 large ensemble dataset to reproduce absolute annual-mean temperature and year to year variability,  conditioned on anthropogenic co2 and sulfate emisisson from ssp3-7.0 scenario. The emulator employs a three-dimensional UNet architecture that learns the spatiotemporal distribution of global temperature fields in latitude–longitude–time space. Conditioning variables include cumulative CO₂ and aerosol emissions, enabling the generation of physically consistent climate responses under arbitrary emission trajectories.To enhance physical interpretability, we integrate explainable AI (XAI) methods, including gradient-based attribution and sensitivity analyses, to quantify how emission-related conditioning variables influence regional temperature responses. The emulator reduces computational cost by several orders of magnitude compared to full ESM simulations, enabling rapid scenario exploration and uncertainty assessment. This framework aims provides a scalable and interpretable pathway for fast climate response emulation

How to cite: Nordling, K.: Emulating absolute annual temperatures and variability  from the CESM2 Large Ensemble using a diffusion model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2500, https://doi.org/10.5194/egusphere-egu26-2500, 2026.

Earth system models (ESMs) are key tools in projecting the reponse of the Earth's climate and ecosystems to anthropogenic forcing in terms of increasing greenhouse gas concentrations and resulting temperature increases, as well as land use change. However, ESMs continue to suffer from prononced biases when compared to observations, and exhibit limited horizontal resolution due to computational constraints, mking reliable impact assessment challenging. Generative machine learning methods, such as Generative Adversarial Networks or Diffusion models, have shown great success in bias correcting and downscaling Earth system model output [1,2]. However, so far these approaches have been applied only as a postprocessing. After summarizing advances in this context, I will present recent work addressing conceptual and technical challenges in incorporating (generative) machine learning inside the architectures of process-based ESMs. These include the need for automatic differentiability of all ESM components [3], as well as physical constraints to assure that dynamics learned by machine learning components fulfills, for example, physical conservation laws [4].  

[1] P. Hess, M. Drüke, F. Strnad, S. Petri, N. Boers: Physically constrained generative adversarial networks for improving precipitation fields from Earth system models, Nature Machine Intelligence 4, 828-839 (2022)

[2] P. Hess, M. Aich, B. Pan, N. Boers: Fast, scale-Adaptive, and uncertainty-aware downscaling of Earth system model fields with generative machine learning, Nature Machine Intelligence 7, 363–373 (2025)

[3] M. Gelbrecht, A. White, S. Bathiany, N. Boers: Differentiable Programming for Earth System Modelling, Geoscientific Model Development 16, 3123–3135 (2023)

[4] A. White, N. Kilbertus, M. Gelbrecht, N. Boers: Stabilized Neural Differential Equations, NeurIPS (2023)

How to cite: Boers, N.: Machine learning for hybrid Earth system modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3829, https://doi.org/10.5194/egusphere-egu26-3829, 2026.

Uncertainty in projections of future regional climate change remains large, driven by structural differences among Earth System Models and the influence of internal climate variability. Existing uncertainty-reduction approaches, including emergent constraints and Bayesian variants, primarily focus on forced climate responses derived from simple aggregate metrics, thereby requiring strong assumptions and exploiting only low-dimensional climate information. Here we propose a data-driven deep-learning framework that directly forecasts spatially and monthly resolved decadal mean climatologies of surface temperature anomalies from the 2030s to the 2090s, using only recent monthly trajectories spanning 1980-2025. The training ensemble contains 265 historical+SSP2-4.5 simulations, distributed across 40 ESMs from 25 different families (i.e., modelling centers) over which the cross validation is performed. The architecture couples pluri-annual to multi-decadal temporal convolutions with a spatial U-Net encoder-decoder and is evaluated on CMIP6 simulations using a leave-one-model-family-out cross-validation (LOMFO-CV) design to ensure generalisation across separately developed ESMs. Predictive uncertainty is quantified via LOMFO-CV errors, yielding conservative and reliable ranges that incorporate irreducible internal variability and systematic model shifts.

To further evaluate the predictive capacity beyond the CMIP6 distribution, we evaluated the network on historical+SSP2-4.5 simulations from a recent HadGEM3-GC5 model hierarchy developed within the European Eddy-Rich ESMs (EERIE) project, the European contribution to HighResMIP2 for CMIP7. In particular, the eddy-rich GC5-HH configuration explicitly simulates mesoscale ocean dynamics that are absent in CMIP6-type models, providing a rigorous test of generalisation to richer and more realistic physical representations. Despite these substantial differences, the network successfully reproduces warming trajectories and future climate patterns for all three model configurations (GC5-LL, GC5-MM, GC5-HH), with forecast errors largely contained within empirically calibrated uncertainty bounds from the LOMFO-CV, both globally and locally. These results, notably for GC5-HH and its more realistic physics, strengthens confidence in the applicability of the framework to real-world data.

When applied to observations, the extracted end-of-century global-mean surface temperature and its uncertainty range are consistent with prior estimates from Bayesian frameworks. At local scales, the network reduces uncertainty by 40% (2030s) to 30% (2090s) on average, and by up to 75% in some regions for all future decades. Importantly, these uncertainty estimates account not only for uncertainty in the forced response (as emergent constraint methods do), but also for errors associated with predicting different realisations of internal variability, providing a physically meaningful reduction of local and global climate uncertainty.

How to cite: Michel, S., Strommen, K., and Christensen, H.: Short- to long-range climate forecasts with deep learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4158, https://doi.org/10.5194/egusphere-egu26-4158, 2026.

Regional Climate Models (RCMs) are vital for capturing mesoscale variability, however remain too coarse for impact assessments in complex topographies like Switzerland. In this study, we bridge the "km-scale gap" by introducing a generative super resolution pipeline to downscale EURO-CORDEX ensemble to a 1 km grid over Switzerland.

We establish the added value of a deterministic residual U-Net, pixel-based as well as generative residual Latent Diffusion over operational baselines and conventional bias correction (BC) methods such as Cumulative Distribution Function - transform (CDF-t), Empirical Quantile Mapping (EQM) and dynamical Optimal Transport Correction (dOTC). Our results demonstrate that super resolved fields have superior distributional skill, better visual fidelity of fields, shows improved  trend preservation and representation of interannual variability across diverse biogeographical regions  and major population centres such as Bern, Zurich and Locarno. Further, as demonstrated by a marked reduction in bias for  20-, 50-, and 100-year return levels of multi-day precipitation totals, super resolution (SR) also complements BC for improved representation of extremes in our km-scale downscaled EUROCORDEX. Our findings establish that while BC methods remain essential for distributional fidelity, residual generative models offer a potent, actionable pathway for producing high-resolution climate information from coarse climate fields.

How to cite: Asthana, S., Koch, E., Kotlarski, S., and beucler, T.: Next-Generation Climate Projections: Insights from Blending Bias Correction with Super Resolution over Complex Terrain, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4507, https://doi.org/10.5194/egusphere-egu26-4507, 2026.

Machine learning-based weather prediction is revolutionizing weather forecasting by learning from present-day climate. However, generalization to other climates remains a major challenge. With melting sea ice, land-use change and increasing ocean temperatures, boundary conditions are changing. Therefore, generalization in time will likely only be possible if generalization in space is also given. The physics of the atmosphere is invariant in space, and as such, a model should demonstrate the same to accurately represent the real world.

Here, we present three test cases to evaluate whether machine learning-based weather and climate models generalize spatially and apply them to multiple AI weather models. The tests consist of reversing the entirety of the input data and boundary conditions in latitude (Test 1), reversing them in longitude (Test 2), as well as rotating them by 180˚ in longitude (Test 3), while keeping all aspects of the simulation physically consistent. For a deterministic model that generalizes in space, each of these test cases yields the same predictions as the baseline case, only subject to a rounding error. With these test cases, we investigate whether data-driven models hardcode representations of spatial relationships in the training data into their latent space. We show that currently, both fully data-driven and hybrid general circulation models do not pass these tests, instead performing poorly with unphysical results. This implies that they have likely not learned underlying atmospheric physics principles, but instead local spatial relationships statistically dependent on geographical location. This calls into question the ability of such models to simulate a changing regional climate. As such, we propose that machine learning-based climate models be evaluated using our spatial tests during model development to reduce overfitting on present-day regional climate.

How to cite: Höver, M., Klöwer, M., Schroeder de Witt, C., and Christensen, H. M.: Spatial Generalization Tests for Machine Learning-based Weather Models as a Requirement for Climate Predictions, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4512, https://doi.org/10.5194/egusphere-egu26-4512, 2026.

Understanding long-term trends in stratospheric species is vital for evaluating the success of the Montreal Protocol and its amendments. However, reliable trend estimation remains challenging due to the sparse spatial and temporal coverage of high-quality observations, such as those from the Atmospheric Chemistry Experiment–Fourier Transform Spectrometer (ACE-FTS).

How to cite: Dhomse, S. and Chipperfield, M.:   Machine Learning For Atmospheric Chemistry: Creating Global, Gap-Free Stratospheric Datasets for Montreal Protocol Assessments, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4514, https://doi.org/10.5194/egusphere-egu26-4514, 2026.

Machine learning models have shown great success in predicting weather up to two weeks ahead, outperforming process-based benchmarks. However, existing approaches mostly focus on the prediction task, and do not incorporate the necessary data assimilation. Moreover, these models often suffer from long-term error accumulation, limiting their applicability to seasonal predictions and climate projections. Here, we introduce Generative Assimilation and Prediction (GAP), a unified deep generative framework for assimilation and prediction of both weather and climate. By learning to quantify the probabilistic distribution of atmospheric states under observational, predictive, and external forcing constraints, GAP excels in a broad range of weather-climate related tasks, including data assimilation, seamless prediction, and climate simulation. In particular, GAP delivers probabilistic weather forecasts competitive with state-of-the-art forecasting systems, while using its own assimilated initial states from a small fraction of observations. Also, it provides seasonal predictions with skill comparable to leading operational system. Finally, GAP produces stable millennial-scale climate simulations that capture variability from daily weather fluctuations to decadal oscillations.

How to cite: Yang, S., Nai, C., Boers, N., Yuan, H., and Pan, B.: GAP: a unified deep generative framework for emulating weather and climate, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4964, https://doi.org/10.5194/egusphere-egu26-4964, 2026.

A pressing problem exacerbated by climate change is the inability to prepare for extreme climate and weather events due to the limited historical record of observed extremes. While crucial for risk assessment and informed policy-making, a better representation of the distribution of "feasible" outcomes remains largely uncertain, with  predictions ranging at variously defined confidence levels that remain sensitive to the choice of metrics and physical assumptions. This question naturally lends itself to investigating how we can engender plausible realizations of extreme events, and thereby allow for mitigation efforts, before communities are forced to confront destructive realities. We present a time-conditioned generative framework based on a computer-vision-aided diffusion model trained on 1km $\times$ 1km precipitation fields and their trajectories over time. The output of this model is n future potential realizations of possible storm events that may unfold over the San Jacinto river basin in the south coast of Texas.

Beyond unconditional sampling, we introduce control variables that make generation decision-relevant: the model is trained to be conditional on a (duration, intensity) pair, enabling users to request ensembles spanning targeted severity regimes (e.g., short–extreme vs. long–moderate) while preserving realistic spatiotemporal structure. This yields a family of distributions over storm trajectories indexed by interpretable controls, allowing systematic stress testing of infrastructure and emergency-response plans under plausible but high-impact scenarios. 

We separate the evaluation of our approach into two complementary perspectives: (i) distribution matching for in-sample generations, and (ii) physics-based alignment with storm-based properties for out-of-sample generations. Spatiotemporal structure of storms is also benchmarked against strong baselines like the analog ensemble method, quantifying the performance of our model to realistically capture intense rainfalls. To extract evolving storm geometries, we employ a kNN-based (k-nearest neighbors) computer-vision algorithm that dynamically identifies storm shapes across time steps. Due to the probabilistic nature of diffusion models, more comprehensive envelopes of the storm intensity and trajectory can be obtained for uncertainty quantification purposes. 

Finally, we introduce a metric that jointly measures physical plausibility through features like intensity–duration structure and scaling, as well as novelty relative to the raw training data. This metric works by penalizing overfitting patterns while rewarding those that respect feasible dynamics, allowing us to define a principled way to compare generative models for extremes. Therefore, we can determine not only how realistic our generated storms are, but also how much physical diversity they contribute beyond the observed data. We present an open evaluation suite for controllable storm generation, including storm-tracking, intensity–duration diagnostics, and physical-novelty scoring.

How to cite: Tsao, V., Zaniolo, M., and Veveakis, M.: Synthetic Physics-Aware Storm Generation via Diffusion Models for Risk Analysis of Catastrophic Events, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5928, https://doi.org/10.5194/egusphere-egu26-5928, 2026.

Various pollutants pose significant threats to river ecosystems. This issue is particularly critical in Taiwan, where the unique geography of short, rapid rivers makes water retention difficult, necessitating rigorous water quality monitoring. Given the complex, non-linear correlations between water quality and meteorological parameters, this study investigates the impact of different feature selection techniques and predictive models on water quality forecasting for eight rivers in Taoyuan. We utilized 14 meteorological and water quality inputs to predict six key indicators, including COD, DO, EC, NH3-N, ORP, and SS. The methodology compared four feature selection strategies—Pearson Correlation, Entropy Weight Method (EWM), Combined Weights, and Mutual Information—alongside four forecasting models: Seq2Seq LSTM, ANFIS, MLP, and Transformer.The feature selection results reveal that the Entropy Weight Method yielded the highest precision (R^2 =0.9336), surpassing the Pearson method (R^2 =0.9161). This indicates that prioritizing features based on information entropy effectively minimizes information loss during screening. Regarding predictive modeling, the Transformer model demonstrated superior stability and accuracy. While other models fluctuated, the Transformer consistently achieved the best performance with an MSE of approximately 14.86 (RMSE=3.855) and an accuracy of 82.52%, significantly outperforming the MLP and ANFIS models. This study concludes that integrating entropy-based feature selection with the Transformer model establishes a superior and highly accurate framework for water quality forecasting in Taoyuan's rivers.

How to cite: Lu, Y.-Y. and Lin, Y.-C.: Integrated Analysis of Feature Selection and Deep Learning Models for Water Quality Forecasting Based on Meteorological Parameters, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6357, https://doi.org/10.5194/egusphere-egu26-6357, 2026.

Air pollution has emerged as one of the most critical environmental health hazards globally. According to statistics from the World Health Organization (WHO) and the Global Burden of Disease Study, approximately 7 million premature deaths occur annually due to air pollution. Fine particulate matter (PM2.5), capable of penetrating deep into the lungs and entering the bloodstream, has been confirmed to be highly correlated with ischemic heart disease, stroke, chronic obstructive pulmonary disease (COPD), and lung cancer. Given its serious threat to public health, establishing high-precision PM2.5 prediction models is critical for early warning systems and health protection.

Addressing the common issue of missing values in environmental monitoring data, this study proposes a data preprocessing framework that combines Principal Component Analysis (PCA) for feature dimensionality reduction with the GRU-D model for time-series imputation. Testing confirms that this method effectively reconstructs data features without causing excessive smoothing. In terms of predictive modeling, this study incorporates East Asian-scale atmospheric pressure field data as a key environmental variable to capture the impact of large-scale weather systems on local air pollution. The performance of three advanced deep learning models—LSTM+CNN, PatchTST, and iTransformer—is evaluated and compared.

The results indicate that, when considering multivariate factors and long- and short-term dependencies, the iTransformer model demonstrates superior predictive performance with an R² of 0.91, exhibiting exceptional non-linear feature extraction capabilities. In comparison, both the LSTM+CNN and PatchTST models achieved an R² of approximately 0.86. Based on the iTransformer's advantages in handling large-scale meteorological features and high-dimensional time-series data, this study employs it as the core model to further extend PM2.5 concentration predictions across Taiwan, aiming to provide a valuable scientific reference for regional air quality management.

How to cite: Chuang, Y. and Lin, Y.-C.: Spatiotemporal Prediction of PM2.5 in Taiwan Using iTransformer and Large-Scale Atmospheric Pressure Features, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6363, https://doi.org/10.5194/egusphere-egu26-6363, 2026.

In recent years, climate change has led to a clear increase in both the frequency and intensity of extreme weather events. Taiwan lies along major typhoon tracks in the western North Pacific, where typhoons represent one of the most significant natural hazards. The strong winds and heavy rainfall associated with typhoons frequently cause flooding, agricultural losses, and damage to critical infrastructure. In practice, however, the severity of typhoon-related disasters does not always correspond to traditional typhoon intensity classifications based primarily on central pressure and wind speed, indicating that wind-based classifications alone may not adequately represent actual disaster impacts.

This study utilizes hourly meteorological station observations to investigate the wind and rainfall characteristics of historical typhoon events in Taiwan. Multiple machine learning and regression models are applied, together with residual analysis, to quantify typhoon characteristics and construct a Typhoon Type Index (TTI). Based on the relative behavior of wind and rainfall during individual events, different typhoon types are further examined to identify their occurrence patterns and characteristic differences across historical cases.

The results indicate that the TTI derived from machine learning–based classification models can effectively improve upon previous TTI formulations established using regression models alone. Moreover, typhoons with different wind–rainfall characteristics are associated with distinct patterns of disaster impacts, and in some cases, rainfall intensity better reflects disaster severity than wind speed. By offering an alternative perspective to conventional intensity-based classifications, this study contributes to improved typhoon disaster risk assessment and provides useful insights for future disaster mitigation and preparedness strategies.

How to cite: Huang, S.-H. and Lin, Y.-C.: Improving the Typhoon Type Index by Integrating Strong Wind and Heavy Rainfall Using Machine Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6364, https://doi.org/10.5194/egusphere-egu26-6364, 2026.

We introduce Field-Space Attention, a novel, scalable, interpretable, and flexible attention module designed for Earth system machine learning models. The key concept involves computing attention directly within physical space on the HEALPix sphere. This approach ensures that all intermediate states remain as globally defined geophysical fields rather than as abstract latent tokens. This field-centric design maintains the physical meaning of internal representations, renders layer-wise updates interpretable, and offers a simple interface for integrating scientific constraints and prior knowledge throughout the network (see Figure). Field-Space Attention is based on a fixed, non-learned, multiscale, spherical decomposition. It learns structure-preserving deformations that coherently couple information across coarse and fine scales. This enables global context without sacrificing local detail.

We demonstrate the module's effectiveness in representative Earth system learning experiments on spherical grids. We focus on global near-surface temperature super-resolution on a HEALPix grid using ERA5 reanalysis data and benchmark it against widely used Vision Transformer and U-Net–style baselines. Our Field-Space Transformer model trains more stably, converge faster, achieve strong accuracy with substantially fewer parameters, and yield physically interpretable intermediate fields.

By keeping computation in field space and explicitly separating scales, Field-Space Attention is particularly well-suited for high-resolution Earth system modeling. It supports scale-aware inductive biases, principled cross-scale consistency, and the efficient coupling of large-scale dynamics with fine-scale variability. These properties position Field-Space Attention as a compact building block for next-generation, high-resolution Earth system prediction and generative modeling. This includes downscaling, spatiotemporal forecasting, infilling, and data assimilation under stronger physical constraints.

How to cite: Witte, M., Meuer, J., Plésiat, É., and Kadow, C.: Field-Space Attention for Structure-Preserving Earth System Transformers, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7473, https://doi.org/10.5194/egusphere-egu26-7473, 2026.

The impact of droughts on vegetation is commonly assessed through correlational analysis of satellite-derived variables, such as NDVI, precipitation anomalies, soil moisture, and more (Hao & Singh 2015, Park et al. 2016, Joiner et al. 2018). However, these correlation-based approaches cannot disentangle the true causal drivers from their confounded associations (Zhang et al. 2022). This limits our ability to understand and attribute the scale of vegetation stress to specific drought mechanisms (e.g. soil moisture deficits versus irrigation resilience), and our ability to design effective interventions that address the primary drivers.

As such, we propose a novel causal inference framework to estimate the impact of drought on vegetation health using satellite time-series data, and demonstrate its application to the Iberian Peninsula. We firstly define a graphical causal model based on established eco-hydrological pathways, and then integrate multi-sensor remote sensing data (MODIS NDVI, SPEI, etc.) and climate reanalysis (ERA5). By extending traditional causal inference methods for georeferenced time-series raster data and controlling for well-established confounding variables (temperature, solar radiation, precipitation, soil moisture, land cover, and irrigation), we isolate the effect of drought severity on vegetation. We also implement novel visualisation methods to display these causal influence estimates.

While causal inference allows us to move beyond correlation and understand the impact on vegetation from each of these key variables, counterfactual intervention is also essential to understand how varying conditions would otherwise change the outcome (Schölkopf et al. 2021), i.e. the severity of the drought impact. Therefore, by leveraging these interventions, our results go from descriptive analytics to actionable insights on drought severity under the changing climate. This enables more effective drought impact assessment for scientists, policymakers, and industry experts.

References:
1. Hao, Z. and Singh, V.P., 2015. Drought characterization from a multivariate perspective: A review. Journal of Hydrology, 527, pp.668-678.
2. Park, S., Im, J., Jang, E. and Rhee, J., 2016. Drought assessment and monitoring through blending of multi-sensor indices using machine learning approaches for different climate regions. Agricultural and forest meteorology, 216, pp.157-169.
3. Joiner, J., Yoshida, Y., Anderson, M., Holmes, T., Hain, C., Reichle, R., Koster, R., Middleton, E. and Zeng, F.W., 2018. Global relationships among traditional reflectance vegetation indices (NDVI and NDII), evapotranspiration (ET), and soil moisture variability on weekly timescales. Remote Sensing of Environment, 219, pp.339-352.
4. Zhang, X., Hao, Z., Singh, V.P., Zhang, Y., Feng, S., Xu, Y. and Hao, F., 2022. Drought propagation under global warming: Characteristics, approaches, processes, and controlling factors. Science of the Total Environment, 838, p.156021.
5. Schölkopf, B., Locatello, F., Bauer, S., Ke, N.R., Kalchbrenner, N., Goyal, A. and Bengio, Y., 2021. Toward causal representation learning. Proceedings of the IEEE, 109(5), pp.612-634.

How to cite: Jancauskas, V., Garske, S., and Espinoza Molina, D.: A Causal Inference Framework for Analysing Drought Drivers, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7552, https://doi.org/10.5194/egusphere-egu26-7552, 2026.

Earth system model (ESM) intercomparison is essential for assessing model performance and identifying future challenges in climate modeling. The Taylor diagram [1] is one of the most widely used tools for this purpose, as it provides an intuitive summary of standard evaluation metrics — such as correlation, root-mean-square error, and standard deviation — by comparing multiple simulated datasets against a reference, typically observations or a ground truth, within a single plot.

However, in several relevant applications, including the development of new ESM parameterizations, the comparison of conceptual models, or the evaluation of simulated statistical distributions, classic linear correlation and RMSE metrics may be insufficient. Here, we propose a set of extensions to the Taylor diagram based on a generalization of cross-covariance using kernels, allowing both nonlinear relationships and distributional aspects of similarity to be taken into account. Nonlinear similarity is characterized through a kernel-space analogue of rotational alignment, while distributional similarity can be quantified using metrics such as maximum mean discrepancy, as originally introduced in [2], as well as alternative kernel-based measures. Using controlled synthetic experiments, we show that the proposed kernel Taylor diagrams can resolve differences in model skill that remain indistinguishable under the classical Taylor diagram. These results indicate that the kernel-based extensions provide complementary diagnostic information to standard metrics and can support more informative Earth system model evaluation and development.

[1] Taylor, K. E. (2001), Summarizing multiple aspects of model performance in a single diagram, J. Geophys. Res., 106(D7), 7183–7192, doi:10.1029/2000JD900719.

[2] Wickstrøm, K., Johnson, J. E., Løkse, S., Camps-Valls, G., Mikalsen, K. Ø., Kampffmeyer, M., & Jenssen, R. (2022). The Kernelized Taylor Diagram. doi:10.48550/arXiv.2205.08864

How to cite: Gavrilov, A., Mankovich, N., Link, M., Huang, F., and Camps-Valls, G.: Kernel Taylor Diagram for Earth System Model Evaluation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7678, https://doi.org/10.5194/egusphere-egu26-7678, 2026.

Understanding the driving forces behind mesoscale cloud organization is fundamental to reducing uncertainties in cloud climate feedbacks. Traditional climate models cannot explicitly resolve mesoscale cloud structures due to their limited resolution, leading to large uncertainties in cloud climate feedback estimates. Storm-resolving models that simulate the atmosphere at kilometre resolution have the potential to reduce these uncertainties. Yet, these models are still biased in their organizational structure when compared to satellite observations. Approaches constraining cloud feedbacks directly from the satellite records are promising but often rely on manually chosen cloud controlling factors (CCFs) that do not necessarily capture all the information necessary to explain mesoscale organizational structures and generally only utilise linear models to predict cloud radiative properties from CCFs.

We present CloudDiff, a probabilistic machine learning model that generates mesoscale cloud structures at kilometre resolution conditioned on environmental conditions in the atmosphere, namely the temperature and humidity profiles as well as vertical and horizontal winds. The model is trained on MODIS Level 1 satellite data and environmental conditions from ECMWF ERA5 reanalysis data. CloudDiff is able to reconstruct realistic MODIS observations from matching ERA5 environmental conditions and achieves a lower reconstruction error compared to generating MODIS observations solely from pre-defined CCFs. In CloudDiff’s generation stage, the environmental conditions are compressed into a latent representation using an attention mechanism. This latent representation can be interpreted as a set of CCFs that have been learned purely from data. We’ll discuss the properties of the learned CCFs including how they relate to existing CCFs, their geographical distribution, and their predictive power of the radiative properties of cloud fields.

How to cite: Reichelt, T. and Stier, P.: CloudDiff: A Conditional Diffusion Model to Generate Mesoscale Cloud Structures, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7971, https://doi.org/10.5194/egusphere-egu26-7971, 2026.

How to cite: Dewey, M., Wilcox, L., Samset, B., and Ekman, A.: Deep-AeroGP: deep kernel learning for projecting the regional climate response to anthropogenic aerosol emission changes , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8307, https://doi.org/10.5194/egusphere-egu26-8307, 2026.

Accurate initiation of deep convection remains a persistent challenge in weather and climate models. Most general circulation models (GCMs) operate at coarse resolution and therefore cannot explicitly resolve convective events; instead, they rely on convective parameterizations in which triggering is diagnosed from environmental thresholds, commonly based on convective available potential energy (CAPE). Convection-permitting models (CPMs) alleviate some of these structural limitations by resolving grid-scale convective spectrum while leaving behind sub-grid scale events. On the other hand, machine learning (ML)-based convection trigger functions have emerged, but still with uncertainty, whose causes are rarely examined. Here, we diagnose the atmospheric states associated with “blind spots” in ML predictors of deep convection initiation, leveraging the Department of Energy Atmospheric Radiation Measurement constrained variational analysis (VARANAL) product and the CPM-based CONUS404 hydroclimate dataset over the Southern Great Plains (SGP). We train a conventional artificial neural network (ANN) and a controlled abstention network (CAN), evaluate their skill in identifying deep convection, and use CAN to quantitatively isolate low-confidence samples while understanding the associated physical conditions in which the models are least reliable. ANN and CAN show comparable baseline performance, and for both models, skill increases when low-confidence samples are excluded, indicating that abstention identifies systematically difficult conditions rather than random noise. Across both VARANAL and CONUS404 datasets, low-confidence samples preferentially occur under weak-to-moderately negative mid-level vertical velocity (−10 to −5 hPa hr⁻¹) and dynamic generation rate of CAPE (dCAPE; 0–200 J kg⁻¹ hr⁻¹). Additionally, these cases are dominated by short, convective episodes that persist for only a few hours, dominantly occurring during the afternoon. These abstention samples also exhibit locally forced, non-equilibrium environments characterized by larger convective adjustment time (τ), consistent with reduced predictability relative to regimes controlled by broader synoptic forcing with smaller τ. Collectively, our results quantitatively identify the regimes and associated physical mechanisms in which ML-based convection predictors are least robust, providing actionable guidance for operational forecasters to treat predictions with greater caution when these low-confidence conditions are present.

How to cite: Bhattarai, A. and Zheng, Y.: Uncertainty-Aware Machine Learning for Deep Convection Initiation: Insights from ARM Observations and Kilometer-Scale Hydroclimate Reanalysis, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8710, https://doi.org/10.5194/egusphere-egu26-8710, 2026.

The emergence of global km-scale climate models challenges traditional model evaluation approaches, which typically rely on long climatological averages. The substantial computational costs and enormous data volumes associated with km-scale simulations often constrain simulation length, limiting the availability of long-term averages. As a result, conventional analysis methods become less practical and less informative when assessing short, high-frequency model output that is potentially dominated by internal variability. At the same time, recent advances in machine learning (ML), particularly in deep neural networks, offer new and innovative ways to efficiently extract information from large climate datasets. Building on this progress, we present an ML-based framework for evaluating climate models on a regional scale over short periods, focusing on daily near-surface air temperature fields over Europe.

We train a convolutional neural network (CNN) to distinguish spatial temperature fields from a large set of climate models. We employ 28 regional simulations from EURO-CORDEX and two global km-scale models from nextGEMS and Destination Earth. Beyond the classification based on climate model simulations, the pre-trained CNN is applied to observation-based test datasets. This setup allows us to build towards an evaluation metric, as the model, the observation-based datasets are more frequently assigned to, might be considered most similar to observed climate. Despite the regional focus of EURO-CORDEX, observation-based samples are most frequently classified as the global km-scale model IFS-FESOM. This suggests that this global km-scale model may capture regional temperature patterns more accurately than regional climate model simulations. Although our results are consistent with traditional metrics in identifying IFS-FESOM as the best-performing model, they also indicate that CNN-based evaluation provides additional information about the similarity between models and observations. 

To better understand which spatial features influence the CNN’s classification for observation-based samples, we apply explainable artificial intelligence (XAI) methods, specifically layerwise relevance propagation (LRP), to the classification outcomes. The resulting relevance patterns indicate that static features such as orography and coastlines, as well as relevance hotspots potentially linked to regions of dynamic variability, play a dominant role in the classification. This highlights that the CNN is sensitive to physically meaningful structures that define model-specific spatial fingerprints.

Using our ML-based framework, we show that a CNN can robustly distinguish between climate models on regional and short time scales as well as identify the model closest to observations. More broadly, we demonstrate that ML, combined with XAI, offers a scalable and physically interpretable approach for evaluating high-resolution climate models, thereby complementing established evaluation frameworks.

How to cite: Meindl, M., Kornblueh, M., Brunner, L., and Voigt, A.: Using Explainable AI to uncover physically meaningful features in km-scale climate models on a regional scale, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9427, https://doi.org/10.5194/egusphere-egu26-9427, 2026.

Following recent advances of foundation models in natural language processing and computer vision, there is growing interest in leveraging geospatial foundation models (GFMs) for Earth system monitoring and climate-relevant applications. In particular, GFMs promise to support large-scale observation of climate-driven extreme events such as wildfires, floods and landslides. However, despite strong benchmark results, recent studies indicate that GFMs for land-cover modelling and hazard mapping models can behave unreliably under real-world conditions. Pretraining datasets often underrepresent rare or extreme environmental regimes, leading to degraded model performance precisely in situations where robust predictions are most critical for climate risk assessment and disaster response. Furthermore, GFMs are often surpassed by simple supervised baselines, highlighting the need for systematic reliability analysis, including out-of-distribution (OOD) detection and uncertainty quantification.

We present SHRUG-FM (systematic handling of real-world uncertainty in geospatial foundation models), a reliability-aware prediction framework that integrates three complementary signals: (1) OOD detection in the input space, (2) OOD detection in the embedding space and (3) task-specific predictive uncertainty obtained from decoder ensembles. We evaluate SHRUG-FM on climate-relevant extreme-event applications, including burn-scar, flood and landslide segmentation. Our results show that elevated OOD scores consistently co-locate with degraded model performance, while uncertainty-based indicators successfully capture many low-confidence and erroneous predictions. By linking these reliability signals to hydro-environmental descriptors from HydroATLAS, we further demonstrate that model failures cluster in distinct geographic and hydroclimatic regimes, revealing interpretable gaps in the pretraining distribution and guiding future dataset design.

SHRUG-FM delivers practical, operationally relevant diagnostics for Earth system monitoring and prediction. It enables selective prediction, rejection strategies, and reliability-aware quality control. These capabilities are essential for integrating GFMs into real-world workflows for climate impact assessment, hazard monitoring and early warning systems. Future work will extend the framework to additional foundation models and climate-driven hazards.

How to cite: Cohrs, K.-H., Gonzalez-Calabuig, M., Nedungadi, V., Osika, Z., Cartuyvels, R., Knoblauch, S., Massant, J., Nath, S., Ebel, P., and Sitokonstantinou, V.: SHRUG-FM: Reliability-Aware Foundation Models for Earth Observation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9940, https://doi.org/10.5194/egusphere-egu26-9940, 2026.

Statistical causality methods are becoming increasingly widespread in climate teleconnection analysis, but they typically require a prior reduction of high-dimensional, multivariate climate fields. Most common aggregation techniques, such as spatial averaging or Principal Component Analysis (PCA) (largely known as Empirical Orthogonal Functions, EOF, in the climate community) [1], are not designed to preserve causal structure and can mask spatially complex or low-variance causal signals.

We introduce Granger PCA [2], a novel dimensionality reduction method that explicitly extracts components that are influenced by a causal driver. Instead of maximizing variance, Granger PCA identifies spatial patterns whose associated time series are maximally Granger caused by an external variable, such as a large-scale climate mode. This is achieved by optimizing spatial weights to maximize the Granger causality F-statistic and yields a low-dimensional representation that captures the Granger causal information present in the field.

The method is particularly effective in cases where causal effects are spatially heterogeneous, have low variance, or are hidden by strong local autocorrelation. In such cases, variance-based methods can fail even when robust causal influence exists.

We apply Granger PCA to several teleconnection problems, including the influence of the North Atlantic Oscillation on precipitation and the impact of ENSO on vegetation variability. Granger PCA recovers physically interpretable patterns that are not captured by PCA or correlation-based approaches.

In summary, Granger PCA provides a simple and interpretable framework for causally oriented dimensionality reduction and offers a new tool for teleconnection analysis in climate science.

References

• [1] A. Hannachi, I. T. Jolliffe, D. B. Stephenson et al., “Empirical orthogonal functions and related techniques in atmospheric science: A review,” International Journal of Climatology, vol. 27, no. 9, pp. 1119–1152, 2007.
• [2] G. Varando, M.-Á. Fernández-Torres, J. Muñoz-Marí, and G. Camps-Valls, “Learning causal representations with Granger PCA,” in UAI 2022 Workshop on Causal Representation Learning, 2022.

How to cite: Durand, H., Varando, G., and Camps-Valls, G.: Granger PCA: Extracting Granger-causal patterns in climate fields, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10055, https://doi.org/10.5194/egusphere-egu26-10055, 2026.

How to cite: Bouabid, S., Womack, C., Flierl, G., Selin, N., Ferrari, R., Souza, A., Giani, P., and Lutjens, B.: Advances in generative climate emulation to support impact-assessment, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10825, https://doi.org/10.5194/egusphere-egu26-10825, 2026.

Africa and South America together account for more than 70 % of the global burned area representing nearly 65 % of global fire-related carbon emissions (van der Werf et al., 2017). Beyond carbon release, wildfires emit large amounts of dust and aerosols that influence regional climate through radiative processes. More generally, wildfires strongly modify land surface properties, including vegetation composition, soil carbon stocks, or surface albedo, with far-reaching consequences for regional carbon, water, and energy cycles.

In the ISBA land surface model (Delire et al., 2020), burned area is currently parameterized using grid-cell surface characteristics, a fire-resistance coefficient, soil moisture, and available biomass. While computationally efficient, this simplified formulation may contribute to persistent regional biases in simulated fire activity. To overcome these limitations, we develop a data-driven fire modeling framework based on two artificial neural network architectures: one addressing a regression task and the other a classification task. The models use meteorological conditions, vegetation states, and anthropogenic factors to estimate the daily burned area fraction.

The proposed framework reproduces the spatiotemporal variability of burned areas with some fidelity. It is specially the case in important areas such as Africa, South America, and Australia. These results highlight the potential of deep learning approaches to enhance wildfire representation and prediction in Earth system models. That would be the very future of our research project.

How to cite: Rougier, H., Decharme, B., and Mallet, M.: Modeling of burned areas on a global scale using statistical learning methods, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11185, https://doi.org/10.5194/egusphere-egu26-11185, 2026.

Application of deep learning has proved useful in many scientific domains and has also gained increased interest as a tool for weather and climate modeling in recent years. Deep Learning weather models have already demonstrated competitive prediction performance to state-of-the-art methods while hybrid models and emulators have shown some promise for climate simulation. However, the realism of simulated climate variability, and climate modes of pure deep learning models trained only on observational or reanalysis data, has not received as much attention.
As one example of these models, we investigate DLESyM, an autoregressive deep learning model based on the U-Net architecture and originally trained on ERA5 reanalysis data from 1981 to 2017 (REF1). Unlike many weather-generating deep learning models, DLESyM does not draw on sea-surface temperatures as boundary conditions, but learns to generate ocean surface patterns. Its applications could, therefore, extend to free-running simulations. The original authors showed its ability to generate stable climate simulations for time-spans up to three millenia, with the absence of spurious drifts and unphysical smoothing in the annual cycle. Here we test how realistic the simulated climate variability of DLESyM is, focusing on interannual to centennial spatio-temporal modes of internal climate variability. We seek to identify whether it is able to generalize to the underlying physical processes of the climate system, or if it is only capable of reproducing spatio-temporal statistical patterns of its training data. We compare the unforced variability of the deep learning model to that in equilibrium simulations out of General Circulation Models out of the Coupled Model Intercomparison Project phase 6 (CMIP6 GCMs), and palaeoclimate reconstructions (REF2). We focus on regional and global power spectra of surface temperatures, and gradients between land and ocean, tropics and extratropics, as well as the high latitudes. To assess the model’s ability to generalize outside the distribution of the training data we perform simulations from varying initial conditions, and comparing them with the output of CMIP6 GCMs. Based on this we discuss potentials and limitations of such a purely data-driven model for climate simulations and future climate risk assessment, where characteristics beyond mean state and slow changes become relevant.

REF1 Cresswell-Clay, N., Liu, B., Durran, D. R., Liu, Z., Espinosa, Z. I., Moreno, R. A., & Karlbauer, M. (2025). A deep learning Earth system model for efficient simulation of the observed climate. AGU Advances, 6, e2025AV001706. https://doi.org/10.1029/2025AV001706

REF2 Laepple, T., Ziegler, E., Weitzel, N. et al. (2023) Regional but not global temperature variability underestimated by climate models at supradecadal timescales. Nat. Geosci., 16, 958–966. https://doi.org/10.1038/s41561-023-01299-9

How to cite: Jansen, H., Racky, M., and Rehfeld, K.: Testing the realism of interannual to centennial climate variability in a generative coupled atmosphere-ocean deep learning model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11509, https://doi.org/10.5194/egusphere-egu26-11509, 2026.

Numerical Earth system model (ESM) simulations require bias correction and downscaling to assess regional climate impacts due to their coarse resolution (50-100km) and systematic errors. Recent generative machine learning-based downscaling methods show promise in capturing small-scale spatial patterns, as well as multivariate and temporal dependencies [1,2,3]. However, making these approaches efficient and scalable to high resolutions globally remains challenging.

Here, we present a generative machine learning method for multivariate and temporally consistent downscaling of global climate fields at daily and 0.25° spatial resolution.  An autoregressive consistency model [4] is trained using Patch Diffusion [5] as an efficient probabilistic emulator of the ERA5 reanalysis and applied to downscale 8 key climate impact variables, including precipitation, temperature, wind speed, and radiation.
We downscale five 100-year simulations per ESM, including pre-industrial control,  historical, and 2K warming scenarios with and without tipping of the Atlantic meridional overturning circulation and the Amazon rainforest, from three CMIP6-class ESMs (MPI-ESM1-2-HR, HadGEM3-GC31-MM, and CESM1-CAM5).

The approach accurately reproduces small-scale variability and extremes, outperforms statistical baselines, substantially reduces biases, and preserves the large-scale response of the tipping dynamics in the ESMs.

   
[1] Mardani, M., Brenowitz, N., Cohen, Y., Pathak, J., Chen, C. Y., Liu, C. C., ... & Pritchard, M., Residual corrective diffusion modeling for km-scale atmospheric downscaling, Communications Earth & Environment, 6(1), 124, 2025. 
[2] Schmidt, J., Schmidt, L., Strnad, F. M., Ludwig, N., & Hennig, P., A generative framework for probabilistic, spatiotemporally coherent downscaling of climate simulation. npj Climate and Atmospheric Science, 8(1), 270, 2025.
[3] Hess, P., Aich, M., Pan, B., & Boers, N.,  Fast, scale-adaptive and uncertainty-aware downscaling of Earth system model fields with generative machine learning, Nature Machine Intelligence, 1-11, 2025.
[4] Wang, Z., Jiang, Y., Zheng, H., Wang, P., He, P., Wang, Z., ... & Zhou, M., Patch diffusion: Faster and more data-efficient training of diffusion models, Advances in neural information processing systems, 36, 72137-72154, 2023. 
[5] Song, Y., & Dhariwal, P., Improved techniques for training consistency models, In The Twelfth International Conference on Learning Representations, 2024.

How to cite: Hess, P., Bathiany, S., and Boers, N.: Generative Machine Learning for Dynamically Consistent Multivariate Downscaling of Tipping Point Simulations from Global Earth System Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11688, https://doi.org/10.5194/egusphere-egu26-11688, 2026.

Air pollution, particularly fine particulate matter (PM2.5), poses a significant risk to public health, necessitating accurate high-resolution monitoring. While global Chemical Transport Models (CTMs) like the Copernicus Atmosphere Monitoring Service (CAMS) provide continuous worldwide coverage, their coarse spatial resolution (~40 km) limits their utility for assessing local exposure relative to regional models (~10 km) that are restricted to specific domains, such as Europe. To bridge this gap, we present a novel deep learning approach for global downscaling of pollutant concentrations based on the Adaptive Fourier Neural Operator (AFNO), benchmarking its performance against a standard U-Net baseline.

We adapted the Modulated AFNO architecture for spatial super-resolution, using low-resolution CAMS Global PM2.5 and dynamic meteorological fields (wind, temperature, dew point, boundary layer height). A key innovation is integrating these inputs with high-resolution static data: orography and population density. We demonstrate that directly inputting static features into the network backbone outperforms separate spatial conditioning, effectively leveraging the Fast Fourier Transform to capture long-range dependencies while respecting local physical constraints.

The model was developed using daily forecasts from 2020 to mid-2025. Training used a sequential split into 2021–2024, preserving 2020 (COVID-19 anomalies) and 2025 as a held-out test set. The model effectively reconstructed fine-scale details and corrected global model biases. Verification against European Environment Agency observations (2020) confirmed performance comparable to high-resolution CAMS Europe regional forecasts. Crucially, the AFNO model consistently outperformed the U-Net baseline and traditional linear interpolation in spatial correlations and error rates. Finally, transferability tests in North America (AirNow data) confirmed the model generalizes effectively to unseen regions, maintaining lower errors than both the original global forecast and the baseline.

How to cite: Monsalvez-Pozo, K., Granell-Haro, F., Martinez-Roig, M., Galván Fraile, V., Plaza-Martín, N. P., Paul Ramacher, M. O., Bieser, J., Flemming, J., Razinger, M., Harder, P., Azorin-Molina, C., and Camps-Valls, G.: AFNO-based downscaling of global air pollution fields, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12008, https://doi.org/10.5194/egusphere-egu26-12008, 2026.

Providing high-resolution climate information by downscaling future climate projections from the Coupled Model Intercomparison Project (CMIP6) remains a central challenge for the regional climate modeling community. CMIP6 includes a wide range of global climate model (GCM) simulations across multiple Shared Socioeconomic Pathways (SSPs), resulting in substantial computational demand for dynamical downscaling if each member is to be fully regionalized. To address this challenge, we propose a computationally efficient statistical downscaling framework based on a U-Net architecture trained over Europe. The model learns high-resolution spatial mappings directly from reanalysis data, offering a low-cost complement to regional climate models (RCMs) for large-ensemble downscaling.

We demonstrate that the climate downscaling U-Net achieves performance comparable to the HCLIM RCM when applied to unbiased EC-Earth3-Veg simulations for both the historical period and the low-emission SSP1-2.6 scenario up to 2100. The model captures spatial temperature patterns, seasonal variability, and the amplitude of warming remarkably well in these cases, providing confidence in its ability to translate GCM-scale information into higher regional climate scales.

When the U-Net is trained exclusively on reanalysis data, its extrapolation behavior under stronger forcing scenarios becomes an important aspect to evaluate. In the high-emission SSP3-7.0 scenario, after the regional climate warms by approximately +2.0 °C beyond the conditions represented in the training data, typically during 2060-2080, the model begins to diverge modestly from the warming magnitude simulated by both the driving GCM and the HCLIM downscaling. This divergence is most pronounced during summer months, while winter temperature trends remain in close agreement. These deviations are not presented as shortcomings of the method, but rather as a clear illustration of the limits of extrapolation when statistical models are trained solely on historical climate states. Highlighting these limits is essential for understanding the robustness of statistical downscaling within and beyond the training domain, particularly for applications involving strong climate-change signals.

Finally, we investigate how the model’s capabilities evolve when future regional climate information is included in the training set. Incorporating a subset of future data markedly improves the extrapolation performance, enabling the U-Net to recover long-term warming trends and seasonal patterns consistent with HCLIM even under strong forcing. This demonstrates that the U-Net architecture can effectively learn and generalize high-resolution climate transformations when provided with an extended training domain. Overall, our findings underscore the potential of deep-learning-based downscaling for scalable, ensemble-wide applications while also clarifying the conditions under which historical-only statistical training remains reliable.

How to cite: Ivanov, M., Fuentes Franco, R., and Koenigk, T.: Can ML-based statistical downscaling models reliably extrapolate into the future?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12050, https://doi.org/10.5194/egusphere-egu26-12050, 2026.

How to cite: Rodriguez Pinilla, M., Benitez Benavides, M., Exarchou, E., Margalef, T., and Panadero, J.: Deep Learning Downscaling of Precipitation and Temperature Climate Data for Future Wildfire Risk Assessment , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12404, https://doi.org/10.5194/egusphere-egu26-12404, 2026.

Climate models typically operate at coarse spatial resolution (~100 km) due to computational constraints, yet many climate-change impact assessments require fine-scale information (<10 km). In this study, we systematically benchmark three state-of-the-art machine-learning approaches for statistical downscaling, using the storm-resolving ICON NextGEMS dataset as reference. All methods take coarse-resolution climate fields as input and generate physically plausible high-resolution predictions. We compare: (1) UNet, a deterministic encoder–decoder architecture; (2) CorrDiff, which augments the UNet backbone with a diffusion model to produce probabilistic ensembles; and (3) CorrDiff++, which replaces diffusion with flow-matching to improve sampling efficiency. We perform 10× downscaling (0.56° to 0.056°) over central Europe for six surface variables, including temperature, wind, and precipitation. The models are evaluated along multiple dimensions: deterministic accuracy (bias, correlation), probabilistic skill (ensemble reliability and sharpness), and physical realism (energy spectra, temporal coherence, representation of extremes). Our results highlight fundamental trade-offs between computational cost, physical consistency, and uncertainty quantification. These insights provide guidance on when the additional complexity of generative models is justified for climate science applications.

How to cite: Debeire, K., Eyring, V., and Thuerey, N.: Benchmarking Deterministic and Generative Machine Learning Models for Statistical Climate Downscaling over Europe, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12407, https://doi.org/10.5194/egusphere-egu26-12407, 2026.

The timing of the first ice-free Arctic summer is a key indicator of climate change, yet projections remain highly uncertain due to inter-model spread, internal variability, and systematic model biases. We develop a prototype framework that combines machine-learning-based methods with causal diagnostics to assess how different bias-correction and emulation approaches influence projections of the first year of ice-free Arctic conditions. Linear scaling is used as a statistical baseline to provide a transparent reference for evaluating more complex machine-learning-based approaches.

Building on recent analyses of the drivers of summer Arctic sea-ice extent at the interannual time scale, we analyse CMIP6 multi-model large ensembles to quantify relationships between September Arctic sea-ice extent and its dominant drivers, including preceding winter sea-ice volume, Arctic near-surface air temperature, and ocean heat transport. Machine-learning-based regression and emulation models are applied to refine model output, while causal diagnostics based on information flow are used to evaluate the physical consistency of inferred driver–response relationships.

We focus on two CMIP6 large ensembles with contrasting historical Arctic temperature biases over 1980–2014. Ensemble uncertainty is explored by partitioning ensemble members into bias-based subsets to assess the sensitivity of projected ice-free timing and inferred driver relationships. Results show that linear scaling shifts projected timing without altering causal structure, whereas machine-learning-based methods can modify ice-free year distributions and induce state-dependent changes in inferred causal relationships. These findings highlight the value of causal diagnostics for interpreting machine-learning-based climate projections and underscore the need for physically interpretable frameworks when applying data-driven methods to critical Arctic climate transitions.

How to cite: Tian, T., Richards, B., and Docquier, D.: Toward more reliable projections of an ice-free Arctic: Integrating machine learning and causal diagnostics in CMIP6 ensembles, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12525, https://doi.org/10.5194/egusphere-egu26-12525, 2026.

Probabilistic risk assessment requires large ensembles of high-resolution climate scenarios, yet generating such data is often computationally intractable. This study introduces a scalable generative framework designed to overcome the scarcity of high-fidelity climate data. We introduce the Field-Space Autoencoder, a geometric compression model that preserves the causal structure of atmospheric fields without forcing them onto regular lat-lon grids. Unlike standard deep learning approaches fixed to a single resolution, our method utilizes a multi-scale decomposition that stores a resolution-invariant latent representation. This flexibility unlocks a novel hybrid training strategy for generative diffusion: we combine the statistical robustness of multi-century, low-resolution simulations with the structural precision of limited high-resolution datasets. The resulting Compressed Field Diffusion model is capable of synthesizing atmospheric states that inherit the internal variability of the large ensemble and the spectral sharpness of the high-res ground truth. By bridging these data sources, we present a pathway to democratizing access to exascale-quality climate data through efficient, physically consistent emulation.

How to cite: Meuer, J., Witte, M., Plésiat, É., and Kadow, C.: Generative Emulation on the Sphere: Bridging the Resolution Gap with Field-Space Diffusion, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12592, https://doi.org/10.5194/egusphere-egu26-12592, 2026.

Cloud feedback is one of the key sources of uncertainty in the sensitivity of climate projections to anthropogenic forcing in Earth system models (ESMs). Improving its representation remains challenging because clouds sit at the intersection of radiation, dynamics, and microphysics, and small errors in any of these can strongly affect climate sensitivity.. Consequently, analysing and understanding errors in simulated cloud feedback, evaluated against observations, is essential for advancing cloud parameterizations in ESMs.

In this work, we explore methodological frameworks for evaluating cloud feedback in climate models that move beyond simple model–observation comparisons toward physically interpretable insights into model properties and dynamics. We propose two advances: (1) improved cloud regime identification by extending standard k-means clustering to Wasserstein k-means, and (2) the use of explainable machine-learning methods to evaluate the extent the ESMs capture the realistic sensitivity between the cloud radiative anomalies and key cloud-controlling factors. We demonstrate these approaches by evaluating different versions of the HadGEM model in AMIP experiments against observations, illustrating their potential to support more physically grounded diagnosis of cloud-feedback behaviour in climate models.

How to cite: Mankovich, N., Gavrilov, A., Huang, F., Camps-Valls, G., Lan, F., and Bodas-Salcedo, A.: Explainable Cloud Feedback Evaluation in Earth System Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13326, https://doi.org/10.5194/egusphere-egu26-13326, 2026.

Reliable causal inference is a central challenge in Earth and climate sciences: observational records are limited, interventions are rare or impossible, and process representations in models rely on parametrizations that can introduce strong asymmetries between variables and the causal mechanisms [1,2]. Leveraging these asymmetries, rather than treating them as nuisances, can offer a principled route to causal discovery that is directly aligned with scientific modeling practice [2].

We address bivariate causal discovery from the standpoint of equation discovery using the Bayesian Machine Scientist (BMS) framework [3]. Our key contribution is to formalize the theoretical link between Symbolic Regression (SR) and Algorithmic Information Theory (AIT) via the Minimum Description Length (MDL) principle: the more plausible causal direction is the one that admits a shorter joint description in terms of a mechanism plus independent inputs [4]. Building on this connection, we characterize the mathematical properties of the resulting causal criterion, including identifiability and asymptotic consistency, and we analyze the role of core assumptions—most notably the Principle of Independent Causal Mechanisms (ICM)—in the context of geophysical data and climate-model parametrizations [5].

We demonstrate the approach on simulated benchmarks and on real Earth-system examples covering both i.i.d. settings and time-series climate data. The results illustrate when and why asymmetric parametrizations help disambiguate causal direction, and they provide a practical pathway to turn discovered governing equations into testable causal hypotheses for Earth and climate science.

References

[1] Jonas Peters, Dominik Janzing, and Bernhard Schölkopf. Elements of causal inference: foundations and learning algorithms. The MIT Press, 2017.

[2] Gustau Camps-Valls, Andreas Gerhardus, Urmi Ninad, Gherardo Varando, Georg Martius, Emili Balaguer-Ballester, Ricardo Vinuesa, Emiliano Diaz, Laure Zanna, and Jakob Runge. Discovering causal relations and equations from data. Physics Reports, 1044:1–68, 2023.

[3] Roger Guimera, Ignasi Reichardt, Antoni Aguilar-Mogas, Francesco A. Massucci, Manuel Miranda, Jordi Pallares y Marta Sales-Pardo. A Bayesian machine scientist to aid in the solution of challenging scientific problems. Science Advances, 6(5):eaav6971, 2020.

[4] Dominik Janzing, Joris Mooij, Kun Zhang, Jan Lemeire, Jakob Zscheischler, Povilas Daniūsis, Bastian Steudel und Bernhard Schölkopf. Information-geometric approach to inferring causal directions. Artificial Intelligence, 182:1–31, 2012.

[5] Sascha Xu, Sarah Mameche, and Jilles Vreeken. Information-theoretic causal discovery in topological order. In The 28th International Conference on Artificial Intelligence and Statistics, 2025.

How to cite: Camps-Valls, G., Guimerà, R., Varando, G., Diaz, E., Cohrs, K.-H., and Sales-Pardo, M.: Causal discovery from equation discovery, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13632, https://doi.org/10.5194/egusphere-egu26-13632, 2026.

Drought is a slowly developing natural hazard that can affect all climatic zones and is commonly defined as a temporary but significant decrease in water availability. In Europe alone, drought impacts over the last decades have generated very large economic losses, and recent summer events have been exceptional in a long-term historical perspective. Despite extensive research on drought monitoring and management, accurately characterizing how drought drivers evolve into impacts is still a key unresolved challenge, especially when impacts result from the cumulative and interacting effects of multiple hydroclimatic anomalies rather than a single precursor.

In this work, we introduce a machine learning procedure named DRIER (Drought Detection via Regression-based Interpretable Extraction and Causal Relationships) to develop interpretable, impact-based drought indices. Unlike traditional indices that primarily look at meteorological anomalies (e.g., precipitation deficits), DRIER is designed to capture the compound nature of drought impacts, such as prolonged dry periods occurring alongside high temperatures and reduced snowpack. DRIER is a fully data-driven and automated framework that integrates: (i) non-linear feature aggregation for dimensionality reduction to preserve an interpretable representation of candidate hydro-meteorological predictors, while reducing their dimension; (ii) conditional mutual information-based feature selection to identify the most informative drought drivers; (iii) multi-task linear regression to upscale learning across multiple sub-regions, leveraging shared drought processes while preserving local heterogeneity; (iv) causal validation using the Transfer Entropy Feature Selection algorithm to confirm that the relationships identified between hydroclimatic variables and drought impacts are not merely correlative but grounded in robust causal mechanisms.

We demonstrate DRIER in the Po River Basin (Italy) by considering 10 sub-basins and using vegetation stress quantified through the Vegetation Health Index (VHI) as an impact proxy. The application shows that DRIER can capture spatially heterogeneous drought–impact relationships across sub-regions while benefiting from multi-task learning to share information where responses are correlated. Importantly, because the framework is interpretable end-to-end, the resulting impact-based index is not a black-box score: each step produces transparent, auditable outputs that identify the key hydroclimatic drivers, how they are aggregated into the index, and how they contribute (in sign and magnitude) to vegetation stress. The integrated causal discovery component further strengthens confidence in real-world use by privileging predictors consistent with robust physical mechanisms, reducing the influence of spurious correlations and supporting transferability across space and time.

How to cite: Bonetti, P., Giuliani, M., Bucci, T., Cardigliano, V., Metelli, A. M., Restelli, M., and Castelletti, A.: Impact-based drought detection via Interpretable Machine Learning and Causal Discovery, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13676, https://doi.org/10.5194/egusphere-egu26-13676, 2026.

Rapid climate scenario exploration remains constrained by a fundamental tension: General Circulation Models and Earth System Models provide comprehensive representations of atmosphere-ocean-carbon interactions but impose computational demands prohibitive for iterative policy evaluation, while Energy Balance Models offer tractability at significant cost to predictive fidelity. Conventional machine learning approaches, though computationally efficient, exhibit excessive data dependence and lack the mechanistic transparency essential for regulatory compliance and evidence-based climate policy. This methodological gap motivates our development of a scientific machine learning framework that augments coupled climate-carbon dynamics through Universal Differential Equations (UDEs), achieving simultaneous forecasting accuracy and interpretability for rapid scenario assessment.

We formulate a three-state coupled dynamical system governing surface temperature anomaly, deep ocean temperature anomaly, and atmospheric CO₂ concentration, incorporating radiative forcing, ocean-atmosphere heat exchange, and temperature-dependent carbon uptake feedback mechanisms. Our investigation proceeds through systematic experimental evaluation. First, we assess Neural Ordinary Differential Equations (Neural ODEs) as black-box dynamical system learners across three random initializations under 1% observational noise. Neural ODEs exhibit substantial forecasting errors—12.45% for surface temperature, 64.08% for ocean temperature, and 5.17% for CO₂ concentration at t=50 years—with progressive error amplification throughout the forecast horizon, demonstrating fundamental limitations in capturing climate dynamics without physical constraints.

Subsequently, we construct a UDE architecture that preserves known energy balance and carbon cycle physics while replacing the temperature-dependent carbon uptake term (βTC) with a neural network component. This hybrid formulation achieves forecasting errors below 0.2% across all climate variables for three distinct initializations, representing order-of-magnitude improvement over Neural ODEs while requiring 57.5% fewer training iterations. Comprehensive robustness analysis across six noise levels (1–25%) demonstrates exceptional stability, with percentage errors remaining below 0.74% up to 20% observational noise, degrading catastrophically only at the 25% threshold.

To ensure mechanistic transparency—critical for climate policy applications—we employ Sparse Identification of Nonlinear Dynamics (SINDy) for symbolic regression on learned neural network outputs. SINDy successfully recovers the correct functional form β·T·C across all noise regimes up to 20%, achieving 100% functional form recovery rate with average relative error of 25.22% at 1% noise. Performance metrics degrade systematically with increasing noise: R² decreases from 0.9985 (1% noise) to 0.7812 (20% noise), with complete interpretability breakdown at 25% noise (R²=0.4028). This characterizes operational bounds for symbolic recovery under realistic measurement uncertainty.

Comparative benchmarking against statistical baselines—Vector Autoregression (VAR) and Autoregressive Integrated Moving Average (ARIMA)—confirms UDE superiority in data-scarce regimes with known physical constraints. While VAR and ARIMA exhibit computational parsimony (21 and 10 parameters respectively versus 8,577 for UDE), they incur prediction errors exceeding 19% for temperature variables, rendering them unsuitable for high-fidelity forecasting. The UDE framework uniquely achieves the accuracy-efficiency-interpretability tradeoff essential for climate scenario exploration, enabling policymakers to evaluate interventions through mechanistically transparent simulations satisfying quantitative risk assessment requirements.

Our results establish that physics-informed machine learning enables accurate climate trajectory prediction while symbolic regression maintains interpretability, yielding a computationally efficient framework for rapid exploration of emission scenarios, carbon taxation policies, and adaptation strategies with explicit uncertainty quantification.

How to cite: Pani, T., Dinesh Joshi, P., Abhijit Dandekar, R., Dandekar, R., and Panat, S.: CLIMASIM — Climate Simulation with Scientific Machine Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13744, https://doi.org/10.5194/egusphere-egu26-13744, 2026.

As climate change amplifies precipitation extremes and their societal and economic impacts, downscaling precipitation provides valuable local-scale information for risk assessment and adaptation planning.
However, deep-learning based statistical downscaling methods typically rely on high-resolution training data (e.g., radar observations), which are scarce and unevenly distributed globally, making geographic generalization a central challenge. Prior work shows large performance drops of deep-learning based downscaling models under geographic distribution shifts -- effects that remain even when considerably increasing the training data [1].
We view the geographic distribution shift as a form of subpopulation shift, where training and target samples are drawn from the same set of geographic domains but differ in their sampling frequencies. Consequently, the shift is driven primarily by changes in the prevalence of climatic regimes, rather than by changes in the conditional relationship between predictors and targets.
To improve robustness under cross-region transfer, we inject additional geographic context through Earth embeddings from geospatial foundation models (e.g., SatCLIP [2]). Potential strategies for integrating these embeddings into diffusion-based downscaling models include attention-based conditioning, feature modulation, and auxiliary conditioning networks.

[1] Harder, P., Schmidt, L., Pelletier, F., Ludwig, N., Chantry, M., Lessig, C., Hernandez-Garcia, A. and
Rolnick, D. [2025], ‘Rainshift: A benchmark for precipitation downscaling across geographies’, arXiv
preprint arXiv:2507.04930 .

[2] Klemmer, K., Rolf, E., Robinson, C., Mackey, L. and Rußwurm, M. [2025], Satclip: Global, general-
purpose location embeddings with satellite imagery, in ‘Proceedings of the AAAI Conference on Artificial
Intelligence’, Vol. 39, pp. 4347–4355.

How to cite: Schmidt, L., Lemaire, P.-L., Ludwig, N., Hernandez-Garcia, A., and Rolnick, D.: Leveraging Earth Embeddings for Generalizable Precipitation Downscaling Across Geographies, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14198, https://doi.org/10.5194/egusphere-egu26-14198, 2026.

Many climate variables are naturally defined on the sphere and exhibit strong anisotropy and directionality (e.g., fronts, jets, boundary currents). Yet most deep-learning forecasting models still rely on planar projections and Euclidean convolutions, which introduce geometric distortions and artificial discontinuities. Graph-based spherical models alleviate some of these issues, but typically remain isotropic and do not explicitly represent local orientation, a key ingredient to model directional transport-like patterns.

Here we introduce and evaluate a gauge-equivariant spherical U-Net implemented directly on the HEALPix grid, designed to encode local orientation consistently across the sphere. Our approach leverages gauge-equivariant convolutions that transform predictably under changes of local reference frame, allowing the network to learn directional filters while preserving spherical geometry. This provides a principled alternative to both planar U-Nets (with longitude-periodic padding) and graph U-Nets, and addresses a core limitation of most spherical models: the lack of explicit orientation handling. This work benchmarks this model against two strong baselines: a planar U-Net with longitude-periodic padding and a spherical graph U-Net defined on the same HEALPix discretization.

We apply this architecture to multi-horizon forecasting of global sea-surface temperature (SST) anomalies at NSIDE=32, using a controlled experimental design with matched training protocols and comparable parameter budgets, with emphasis on low-capacity regimes relevant to data-limited climate settings (≈30–40 years of monthly observations). We report quantitative metrics across horizons and analyze qualitative error modes, showing how gauge-equivariant spherical convolutions mitigate projection artefacts while enabling orientation-aware feature extraction on the sphere. Our results highlight when and why encoding orientation through gauge equivariance provides added value beyond “spherical-but-isotropic” baselines, and offer practical guidance for deploying spherical equivariant models in climate forecasting pipelines.

How to cite: Delouis, J.-M., Odaka, T., and Tétaud, S.: Gauge-Equivariant Spherical U-Nets on HEALPix for Global SST Forecasting: Encoding local orientation on the sphere, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14352, https://doi.org/10.5194/egusphere-egu26-14352, 2026.

Tropical cyclones (TCs) are among the most destructive natural hazards worldwide. While several decades of satellite and reanalysis products now provide relatively large observational datasets of TCs, these datasets remain small by modern deep-learning standards and, crucially, are extremely imbalanced and do not sufficiently cover the tails of the distribution, with Category 5 cyclones being several orders of magnitude rarer than tropical storms. This severe data scarcity and imbalance poses fundamental limitations for supervised learning approaches to tasks such as intensity estimation, rapid intensification forecasting, or impact modeling, where performance on extremes is often the primary objective.

In this context, generative artificial intelligence offers a promising alternative. Diffusion models, in particular, have recently demonstrated state-of-the-art performance in modeling complex, high-dimensional data distributions. By learning the full probability distribution of TC-related fields rather than a single conditional mapping, diffusion models have the potential to generate physically plausible samples across the entire intensity spectrum, including rare but high-impact extremes. However, most existing applications of diffusion models—both within and outside the geosciences—are evaluated using perceptual or distributional metrics originally developed for natural images, such as visual inspection or feature-space distances. These metrics are poorly aligned with the physical constraints and scientific objectives that govern atmospheric phenomena, and may obscure important deficiencies in dynamical or thermodynamical realism.

Here, we present a diffusion-based generative framework for tropical cyclone spatial fields and propose a comprehensive evaluation strategy grounded in physically meaningful diagnostics. Rather than relying on perception-oriented scores, we assess generated samples using a suite of metrics designed to capture key aspects of TC structure and behavior, including radial symmetry, intensity–structure relationships, spatial gradients, and consistency with known climatological distributions across intensity classes. This allows us to directly interrogate whether the model reproduces physically coherent storm morphologies, particularly in the poorly sampled tails of the distribution. Beyond evaluation, we also explore multiple strategies for embedding physical realism directly into the model design. Together, these results highlight both the opportunities and the limitations of diffusion models as scientific tools for tropical cyclone research, and provide a framework for using generative AI not merely as a data-augmentation device, but as a principled instrument for studying rare and extreme atmospheric phenomena.

How to cite: Ascenso, G., Scoccimarro, E., and Castelletti, A.: Assessing physical realism in diffusion models for tropical cyclones, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15130, https://doi.org/10.5194/egusphere-egu26-15130, 2026.

Accurate forecasting of precipitation remains a central challenge in climate science, primarily due to the strong temporal and spatial variability of rainfall, a difficulty that is further intensified by the ongoing impacts of climate change. Recent developments in machine learning have facilitated the design of more accurate and robust predictive frameworks. In this context, the present study implements and evaluates three deep learning architectures; Long Short-Term Memory (LSTM), Gated Recurrent Unit (GRU), and Extended Long Short-Term Memory (xLSTM); to forecast monthly precipitation at 27 meteorological stations distributed across Morocco, for lead times ranging from 1 to 4 months. The models are trained using a heterogeneous set of large-scale climatic predictors, including sea surface temperature (SST) over the Atlantic Ocean and the Mediterranean Sea, the East Atlantic pattern (EA), the Madden–Julian Oscillation (MJO), the El Niño–Southern Oscillation (ENSO), the Mediterranean Oscillation (MO), the North Atlantic Oscillation (NAO), and the Western Mediterranean Oscillation (WeMO). To identify the most influential predictors at each station, a principal component analysis (PCA)-based feature selection procedure is implemented. The results indicate that precipitation variability across the study area is predominantly controlled by the MO, NAO, and WeMO indices. Probabilistic forecasts are then generated using Monte Carlo dropout, enabling the networks to approximate Bayesian inference and thereby quantify predictive uncertainty and associated confidence intervals. Relative to conventional LSTM and GRU configurations, the xLSTM architecture exhibits superior predictive performance across all stations and lead times, with notably reduced uncertainty, particularly in the representation of extreme precipitation events. Overall, the models demonstrate robust skill in northern Morocco, with coefficients of determination (R²) ranging from 0.82 to 0.96 for a 1‑month lead time. However, predictive skill degrades toward the southern region, characterized by arid to semi-arid climatic conditions, where R² values decrease to 0.36–0.86. These results indicate that xLSTM effectively captures long-range temporal dependencies and low-frequency, high-intensity rainfall events, thereby representing a promising framework for improving probabilistic monthly precipitation forecasts in climatically heterogeneous regions such as Morocco.

How to cite: Zellou, B., Agdoud, F., and Ouatiki, H.: Probabilistic Monthly Precipitation Forecasting over Morocco Using xLSTM and Large-Scale Climate Predictors, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15554, https://doi.org/10.5194/egusphere-egu26-15554, 2026.

Machine learning–based weather prediction models have recently surpassed traditional numerical weather prediction systems on many skill metrics at regional and global scales, yet there is limited progress towards models operating on hectometric-scale resolutions. This setting is challenging both due to the cost of generating high-quality training data and the complex dynamics of important small-scale processes.

We introduce a graph neural network with Large-Eddy Simulation (LES) capabilities, to operate at hectometer horizontal resolution and sub-hourly time steps. Using 42 days of high-resolution realistic model output for the trade-wind regime over the western Atlantic, we train and evaluate the network on its ability to reproduce key mesoscale processes, with particular emphasis on cold-pool dynamics and convective triggering.

Cold pools are a crucial driver of low-level thermodynamic variability and cloudiness, and thus provide a stringent physical consistency test for models targeting hectometer scales, as they require accurate coupling between the cloud layer and the surface. Through a targeted ablation study, we quantify the relative importance of different input variables for reproducing surface temperature perturbations associated with cold pools, offering guidance for future parameterization and data selection strategies.

Finally, we show that the model can deterministically predict the evolution of cold pools over multiple successive generations, indicating that graph-based LES emulators can robustly capture the nonlinear feedbacks governing mesoscale organization in shallow convective regimes.

How to cite: Schulz, H., Oskarsson, J., and Denby, L.: ML-LES: Modeling cold-pool dynamics with graph-based neural network at hecto-meter grid-spacings, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16552, https://doi.org/10.5194/egusphere-egu26-16552, 2026.

The majority of supervised machine learning pipelines, particularly in the popular domains of natural language processing and computer vision, rely on manually annotated data. In geoscience applications, however, reference data are not necessarily derived from human annotation but could come as the output of explicit physical models or algorithms. These algorithms typically rely on simplifying hypotheses about the underlying physical processes and may be computationally expensive or applicable only to a limited subset of observations. In such circumstances, machine learning can be used to emulate explicit algorithms, with the objective of reproducing their outputs while potentially exploiting wider information pathways present in the data.

Beyond computational considerations, this hypothesis-light, data-driven framework allows for counterfactual testing by selectively removing input information and evaluating the model’s ability to recover similar predictions. For instance, in computer vision, color information can be removed by averaging RGB channels, while semantic or contextual information can be limited by progressively reducing the input patch size or by exploiting the inductive biases of different neural network architectures. In this way, one can identify additional cues in the input data linked to the physical property of interest, but also assess whether the model reproduces biases inherent to the reference algorithm. 

We explore this approach for high-resolution cloud-top height (CTH) estimation within the Clouds Decoded project, which uses Sentinel-2 (S2) multispectral observations (originally intended for land monitoring) to retrieve cloud properties. CTH can be estimated from S2 imagery using a stereo-based method that leverages the instrument’s geometry and inter-band delays. While effective, this approach is computationally demanding and relies on assumptions that restrict its applicability across diverse cloud scenes. We assess whether a neural network can learn to approximate this stereo-based CTH retrieval and analyse which textural, spectral, high-level semantic, or even geolocation-related cues the model might use to infer cloud height.

How to cite: Borne--Pons, P., Francis, A., Czerkawski, M., Campbell, J., and Bertozzi, B.: Machine learning emulation of stereo-based cloud-top height retrieval from Sentinel-2 , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18374, https://doi.org/10.5194/egusphere-egu26-18374, 2026.

Machine learning based dynamic numerical climate multi-model ensemble
weighting for high impact weather affecting the energy infrastructure
The infrastructure for renewable energy production and electrical grid itself is affected
by weather and climate conditions and vulnerable to high impact weather and
cascading events. A reliable representation of the meteorological conditions leading to
such events including their uncertainty is therefore needed for both weather and climate
time scales. Individual numerical weather and climate models exhibit systematic
strengths and weaknesses across scales, and geographic regions, despite the
differences in model physics and parametrizations. One way to tackle this and avoid
running single-model ensemble climate simulations are multi-model or poor-man
ensembles consisting of a set of different climate or weather models combined.
Ensemble mixing offers a way to mitigate these weaknesses while providing uncertainty
quantification. Simply ensemble averaging can dilute forecast and climate signals and
penalize outliers and rare extremes. Different approaches have been proposed to tackle
this problem by assigning non-uniform weights to individual model fields and
parameters, however, these methods often rely on domain knowledge such as model
dependencies [1,2].

Here, we propose a machine learning-based multi-model ensemble model-mixing
framework that is domain-agnostic and assigns spatially and temporally dynamic
weights, in addition to an error metric. The domain of interest is the Alps, which exhibit
challenging terrain and localized extreme events, e.g. precipitation extremes that are
difficult to capture in conventional climate models. The CERRA reanalysis data at ~5.5
km resolution serves as the target grid. We build a multi-model ensemble by combining
dynamically downscaled simulations of 2 m air temperature, precipitation, and wind
speed from COSMO-CLM (6 km) and WRF (10 km). Each regional model is driven by two
CMIP6 global climate models (MPI-ESM and EC-Earth) under two scenarios (SSP1-2.6
and SSP5-8.5), with an additional historical period used for training. Static information
such as orography and seasonal dependencies are considered. We evaluate the
ensemble’s performance on selected extreme events (e.g., heavy precipitation,
windstorms, heatwaves) that can (and did) harm energy infrastructure, such as the
European derecho 2022.

[1] Christensen, Jh, E Kjellström, F Giorgi, G Lenderink, and M Rummukainen. 2010.
‘Weight Assignment in Regional Climate Models’. Climate Research 44 (2–3): 179–94.
https://doi.org/10.3354/cr00916.

[2] Merrifield, Anna Louise, Lukas Brunner, Ruth Lorenz, Iselin Medhaug, and Reto
Knutti. 2020. ‘An Investigation of Weighting Schemes Suitable for Incorporating Large
Ensembles into Multi-Model Ensembles’. Earth System Dynamics 11 (3): 807–34.
https://doi.org/10.5194/esd-11-807-2020.

How to cite: Thiele, P., Baier, K., Hasel, K., Schellander-Gorgas, T., Lehner, S., Spiekermann, R., Lampert, J., Lexner, A., and Schicker, I.: Machine learning based dynamic numerical climate multi-model ensemble weighting for high impact weather affecting the energy infrastructure, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18613, https://doi.org/10.5194/egusphere-egu26-18613, 2026.

 Ocean physical and biogeochemical extremes, such as marine heatwaves (MHWs), deoxygenation, and acidification events have significant impacts on the marine environment, ecosystems, and economic livelihoods. In recent decades, the frequency, intensity and spatial extent of these extremes have been amplified (Capotondi et al., 2024; Shu et al., 2025; Gruber et al., 2021). Hence, a deeper understanding of the processes and precursors leading to extreme events remains crucial for improving and forecasting risk assessment.

In this study, we apply interpretable machine learning approaches to investigate which oceanic and atmospheric variables, as well as their lag effects, are most relevant for the extreme events in the North Atlantic, a relevant region for their occurrence in recent decades (England et al., 2025). Our framework combines high-resolution ocean model simulations with explainable artificial intelligence (XAI) techniques (He et al., 2024, Camps-Valls, 2025), allowing us to examine where, when, and which model variables are more important when identifying extreme events.

Rather than focusing on predictive skill, the emphasis of this study lies on identifying the underlying physics of precursor patterns leading to ocean extremes across different spatial and temporal scales. By integrating XAI into the analysis, this approach provides a more transparent and interpretable perspective on the decision-making processes of machine learning models, offering insights into the key variables and structures associated with the occurrence of ocean extremes. The outcomes of this study improve the interpretable assessment of potential precursors of MHWs, ocean deoxygenation and acidification extremes.

Camps-Valls, G., Fernández-Torres, M. Á., Cohrs, K. H., et al. (2025). Artificial intelligence for modeling and understanding extreme weather and climate events. Nature Communications, 16, 1919. https://doi.org/10.1038/s41467-025-56573-8

Capotondi, A., Rodrigues, R. R., Sen Gupta, A., et al. (2024). A global overview of marine heatwaves in a changing climate. Communications Earth & Environment, 5, 701. https://doi.org/10.1038/s43247-024-01806-9

England, M. H., Li, Z., Huguenin, M. F., et al. (2025). Drivers of the extreme North Atlantic marine heatwave during 2023. Nature, 642, 636–643. https://doi.org/10.1038/s41586-025-08903-5

Gruber, N., Boyd, P. W., Frölicher, T. L., et al. (2021). Biogeochemical extremes and compound events in the ocean. Nature, 600, 395–407. https://doi.org/10.1038/s41586-021-03981-7

He, Q., Zhu, Z., Zhao, D., Song, W., & Huang, D. (2024). An interpretable deep learning approach for detecting marine heatwave patterns. Applied Sciences, 14(2), 601. https://doi.org/10.3390/app14020601

Shu, R., Wu, H., Gao, Y., et al. (2025). Advanced forecasts of global extreme marine heatwaves through a physics-guided data-driven approach. Environmental Research Letters, 20(4). https://doi.org/10.1088/1748-9326/adbddd

How to cite: Radin, C., Mathis, M., Li, H., and Ilyina, T.: Explainable AI for Identifying Precursors of Extreme Oceanic Events in the North Atlantic, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19617, https://doi.org/10.5194/egusphere-egu26-19617, 2026.

High-resolution gridded climate datasets are essential for Earth system modelling and impact assessments, yet generating them from sparse, irregularly distributed station networks remains a significant challenge, particularly in regions with complex topography. This study evaluates the Spatial Multi-Attention Conditional Neural Process (SMACNP), a probabilistic deep learning framework, for the daily spatial interpolation of air temperature and precipitation, marking the first application of its localized encoder variant to the challenge of gridding climate data from a sparse station network. We investigate two distinct encoder configurations—Global and Localized—to determine the optimal structural prior for capturing spatial dependencies in data-scarce regimes. The models were developed and evaluated using data from a sparse network of meteorological stations in Romania from 2020 to 2023. To ensure applicability for long-term historical reconstruction, the input features were restricted to static topographic predictors derived from a Digital Elevation Model (DEM). Performance was benchmarked against Regression Kriging (RK), a standard geostatistical baseline that incorporates these same topographic covariates. Results demonstrate that the SMACNP architectures substantially outperform the RK baseline for both variables.

The SMACNP (Localized) configuration, which utilizes an attention mechanism, emerged as the most robust model, achieving the lowest Mean Absolute Error (MAE) and the highest correlation across the majority of seasons. The performance gains were particularly pronounced for precipitation, where the deep learning models effectively captured fine-scale spatial heterogeneity and non-linearities that traditional methods tended to over-smooth. These findings indicate that localized neural process-based models offer a powerful, scalable, and physically plausible alternative to geostatistical methods for generating high-quality gridded climate datasets in complex, data-sparse environments.

This research was supported by the project “Cross-sectoral Framework for Socio-Economic Resilience to Climate Change and Extreme Events in Europe (CROSSEU)” funded by the European Union Horizon Europe Programme, under Grant agreement n° 101081377.

How to cite: Dumitrescu, A.: A deep learning framework for gridding daily climate variables from a sparse station network, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19697, https://doi.org/10.5194/egusphere-egu26-19697, 2026.

How to cite: Miranda, V., Castro, M., Paixão, J., Girão, I., Marques, B., Magnus Koktvedgaard Zeitzen, R., Cunha, R., Fonteneles, C., Pereira, É., Khudynian, M., Thejll, P., Jomo Danielsen Sørup, H., Paletta, Q., and Oliveira, A. P.: CLIM4cities - from Citizen Science, Machine Learning and Earth Observation towards Urban Climate Services , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19718, https://doi.org/10.5194/egusphere-egu26-19718, 2026.

Global Climate Models (GCMs) provide critical insights into future climate variability, yet their coarse spatial resolution limits their utility for regional and local-scale impact assessments. AI-driven downscaling techniques have emerged in the last few years as a cost-effective and viable alternative to traditional methods to enhance the spatial resolution of climate projections. Nevertheless, establishing their reliability in unseen climate states remains a priority. This study applies and evaluates a deep generative Latent Diffusion Model, leveraging a residual approach (LDM_res, Tomasi et al., 2025) to downscale GCM outputs (~1°) to high-resolution (~4 km) 6-hourly precipitation and 2-m minimum and maximum temperature fields.

The LDM is developed as an emulator of the COSMO-CLM dynamical model, trained on VHR-REA_IT data (Raffa et al., 2021 - a dynamical downscaling of ERA5). By using aggregated ERA5 data as low-resolution predictors (along with high-resolution static data), the LDM_res model is required to learn to mimic the computationally expensive physics of dynamical downscaling. The model, trained over the past 40 years, is subsequently applied to generate high-resolution climate projections based on the input from four selected CMIP6 GCMs across four different emission scenarios. This modeling chain establishes a hybrid ML-Physics-based system to provide impact assessors with cost-effective, high-resolution climate information.

A central challenge addressed in this work is the evaluation of the model's out-of-distribution generalization—specifically its ability to perform in unseen future climate states and under predictor configurations characteristic of CMIP6 projections. We evaluate the emulator's reliability by comparing its outputs against VHR-PRO_IT, a "twin" dataset of VHR-REA_IT produced using COSMO_CLM to dynamically downscale projections (Raffa et al., 2023), providing a rigorous test of the ML system’s reliability in out-of-domain scenarios.

Furthermore, we compare the LDM_res against traditional statistical (e.g., quantile mapping) and dynamical approaches. Comparative results over the Italian peninsula indicate that while the LDM preserves large-scale seasonal signals from CMIP6 models, it significantly enhances spatial realism and local variability in topographically complex areas. Unlike purely statistical methods, the hybrid ML approach demonstrates superior ability to represent fine-scale heterogeneity in mountainous and coastal regions while maintaining consistency with the original signal.

How to cite: Tomasi, E., Franch, G., Tomezzoli, G., Calmanti, S., and Cristoforetti, M.: Deep learning for high-resolution climate projections: a Latent Diffusion Model emulating dynamical downscaling over Italy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19912, https://doi.org/10.5194/egusphere-egu26-19912, 2026.

The accurate assessment of extreme flood events and their associated losses requires massive sample sizes (e.g., 50,000+ years of weather data) for statistical robustness and a comprehensive coverage of event characteristics. Generating such a large dataset using dynamical Earth System Models would be extremely computationally intensive, so instead, we propose a lightweight and computationally efficient climate emulator built upon a video diffusion architecture. 

The model is trained to reproduce the statistical properties and physical dynamics of the Community Earth System Model version 2 (CESM2) over Europe. It operates autoregressively to generate synthetic, multivariate, daily atmospheric data (including temperature, specific humidity, wind vectors, and surface pressure) at ~100 km resolution. The model utilizes a U-Net architecture that is conditioned on previous time-steps to produce and evolve weather patterns with spatial and temporal consistency. To enhance the stability of long-term generation and improve the faithful reproduction of extremes, we employ a seasonality-aware standardization scheme, training the model to learn the dynamics in anomaly space rather than physical space.

We demonstrate that this approach successfully reproduces the complex spatiotemporal dependencies within CESM2, captures atmospheric dynamics, including the frequency and persistence of dominant circulation types, and can maintain stability over multi-decadal generation windows. Furthermore, the output of this emulator can be fed into existing downscaling models to produce higher resolution multivariate meteorological data fields to drive downstream impact models. We validate this full modeling chain by demonstrating that the resulting hydrological statistics exhibit physical characteristics consistent with the CESM2-driven benchmark.

This computationally efficient generative model offers a pathway to generating thousands of years of physically consistent flood events. 

How to cite: Marshall, A., Lucas, C., Addor, N., Lord, N., Moraga, J. S., Hoch, J., and Wing, O.: Fast emulation of climate models for precipitation and flood impact modelling using autoregressive video diffusion, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19946, https://doi.org/10.5194/egusphere-egu26-19946, 2026.

Remote sensing datasets for land cover classification are mostly distributed in UTM projection which introduce significant geometric distortions—particularly at high latitudes—and fail to respect the spherical geometry of Earth. These distortions propagate into deep learning models trained on such data, leading to latitude-dependent biases, edge artifacts in tile-based processing, and poor generalization across geographic boundaries. While convolutional neural networks (CNNs) have achieved state-of-the-art performance on benchmark datasets like BigEarthNet, they operate on Euclidean grids and cannot naturally handle the structure of a sphere.

Here we introduce a comprehensive pipeline for transforming the BigEarthNet dataset—comprising 549,488 multispectral image patches from its original UTM projection into the HEALPix (Hierarchical Equal Area isoLatitude Pixelization) representation. HEALPix, originally developed for cosmic microwave background analysis, offers equal-area partitioning of the sphere, ensuring uniform statistical treatment of pixels regardless of latitude, and provides a natural hierarchical structure for multi-resolution analysis.

We implement and evaluate spherical CNNs architectures designed for data on spherical manifolds—against traditional planar CNN baselines (Unet/Resnet50) trained on the HEALPix-transformed data, benchmarking classification performance for multi-label land cover prediction using the 19-class BigEarthNet nomenclature with metrics suited to imbalanced settings (F1-macro/micro, precision, recall, average precision).

This work represents the first large-scale application of HEALPix projection to Remote Sensing classification and validates the effectiveness of spherical deep learning for real-world remote sensing beyond traditional climate science domains. Our experimental design employs matched training protocols and comparable model capacities, demonstrating that spherical representations eliminate projection-induced artifacts, enable seamless cross-boundary analysis, and provide rotation equivariance that reduces the need for extensive spatial data augmentation—key advantages for global-scale Earth observation applications.

How to cite: Tétaud, S. and Delouis, J. M.: BigEarthNet-HEALPix: Spherical CNNs for Land Cover Classificatiom, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20756, https://doi.org/10.5194/egusphere-egu26-20756, 2026.

Surface methane emissions can be estimated from atmospheric observations using inverse modelling systems, which often rely on Lagrangian Particle Dispersion Models (LPDMs) to simulate how the gas is transported through the atmosphere using meteorological fields. However, LPDM-based techniques struggle to scale to the size of modern satellite datasets, as one LPDM run is needed for each observation, taking on the order of 10 CPU-minutes to complete. Previously, we introduced the Machine Learning model GATES (Graph-Neural-Network Atmospheric Transport Emulation System), which can replicate LPDM outputs 1000x faster than the physics-based model, and demonstrated its application to infer emissions over South America. Training GATES over other world regions and comparing cross-regional performance shows that the learnt transport is domain-specific, consistent with the strong heterogeneity in wind patterns and topography across continents. In this presentation, we discuss transfer learning techniques and characterisation of regional differences in wind patterns, topography, data availability and the shape and magnitude of LPDM outputs, to increase transfer learning performance. This work builds capabilities towards efficient, global methane emissions emulation. 

How to cite: Clark, J., Fillola, E., Keshtmand, N., Santos-Rodriguez, R., and Rigby, M.: From regional to global emulation: characterising regional differences to increase transfer learning performance  , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20966, https://doi.org/10.5194/egusphere-egu26-20966, 2026.

Deep convection exhibits substantial variability even under fixed large-scale forcing, challenging deterministic descriptions of convective organization. Using idealized radiative–convective equilibrium simulations with imposed low-level shear, we quantify this intrinsic variability through a reduced-order stochastic framework. Convective transport is characterized by isentropic mass flux and embedded in a low-dimensional latent space using a variational autoencoder. The temporal evolution of convection in this space is modeled as a Markov chain, yielding a data-driven representation of convective states and their transition probabilities.

This framework demonstrates that internal feedbacks alone generate a broad ensemble of admissible convective trajectories within a single environment, which we interpret as the system’s intrinsic stochasticity. The leading latent dimensions correspond to the convective life cycle and degree of organization, while state transitions identify the constrained pathways through which organized convection emerges and evolves. Comparison of individual storm trajectories in latent space exposes systematic differences in dynamical behavior that are difficult to diagnose in physical space. However, departures from strictly Markovian behavior indicate that the instantaneous state representation does not fully capture slow memory effects associated with convective organization, which likely condition transition probabilities.

These results show that organized convection is best understood as one realization drawn from a constrained distribution of possible trajectories and establish a general machine-learning-enabled framework for quantifying variability and limits of predictability in multiscale atmospheric systems.

How to cite: Abramian, S., Olivier, P., and Pierre, G.: How Organized Convection Evolves in Latent Space, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21859, https://doi.org/10.5194/egusphere-egu26-21859, 2026.

The Marmara Sea, covering approximately 11,350 km² in northwestern Turkey, links the Black Sea and the Aegean Sea via the Bosporus and Dardanelles straits. It is bordered by densely populated and industrialized cities such as Istanbul. The Marmara Sea is facing eutrophication and mucilage outbreaks, necessitating the monitoring of key indicators, including chlorophyll-a, which serves as an indicator of phytoplankton abundance. Atmospheric dust deposition can play a significant role in providing nutrients such as nitrogen, phosphorus, silica, and iron to the surface ocean, thereby affecting phytoplankton growth. Excessive phytoplankton growth and the accumulation of organic matter trigger mucilage formation under suitable conditions. The region is influenced by dust transported from regional and distant sources, such as the Sahara Desert.

In this study, spatio-temporal dynamics of chlorophyll-a (Chl-a), Aerosol Optical Depth (AOD), Sea Surface Temperature (SST), Particulate Organic Carbon (POC), Photosynthetically Active Radiation (PAR), and precipitation were investigated on a monthly scale using MODIS-derived products from 2005 to 2020. Time series analysis and machine learning models such as HGB (Histogram Gradient Boosting), Random Forest, and Multiple Linear Regression were performed for exploring temporal patterns, relationships, and modeling Chl-a, respectively. Chl-a showed a moderate negative correlation with SST (r = –0.52) and a strong positive correlation with POC (r = 0.80), while its relationship with AOD was negligible. It should be noted that during desert dust episodes, a significant lagged correlation was observed between Chl-a and AOD. The observed Chl-a values ranged between 0.6 and 19.50 mg/m³ over the study period, with the highest values observed in April and the lowest values occurring between June and November. Modeling Chl-a based on satellite-derived environmental variables showed that the Histogram Gradient Boosting algorithm achieved the highest performance, yielding r = 0.807, R² = 0.645, RMSE = 1.870, MAE = 1.218, and MBE = 0.062. These results highlighted the strong influence of SST and POC on Chl-a variability, while AOD appears to have minimal direct impact. Further investigation of the impact of the high dust deposition periods during dust storm events is suggested for the Marmara Sea.   

How to cite: Demir, B., Aydin, Y., and Olgun, N.: Machine-Learning Assessment of Chlorophyll-a Responses to Atmospheric Dust and Environmental Factors Using Remote Sensing Data in the Marmara Sea, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1199, https://doi.org/10.5194/egusphere-egu26-1199, 2026.

Hybrid modeling integrates data-driven Machine Learning (ML) components, such as Neural Networks (NN), into physics-based numerical models to improve the accuracy, stability, and adaptability of dynamical simulations. Rather than replacing established physical laws, hybrid models augment them by learning corrections that compensate for unresolved processes, reduce systematic biases, or dynamically calibrate uncertain parameters.

In oceanic and atmospheric numerical models, unresolved dynamics are represented through sub-grid-scale (SGS) parameterizations coupled to the Navier–Stokes equations. As these parameterizations constitute a major source of uncertainty, recent work has increasingly explored Artificial Intelligence (AI) to improve their modeling and constraint. A particularly promising strategy is online learning, in which the AI model is embedded within the numerical solver and trained while interacting with the evolving system dynamics. This setup allows the model to learn temporal dependencies across multiple solver steps and to optimize long-term behavior. Although online learning has demonstrated improved forecast skill and stability over long horizons compared to the more widely used offline learning strategy, its application to high-dimensional ocean models is limited by two key challenges: the requirement for fully differentiable solvers and the high computational and memory costs associated with backpropagation through long trajectories.

To overcome these limitations, we introduce a new family of gradient-approximation methods that selectively simplify intermediate Jacobians in the backpropagation chain. The resulting gradients closely approximate the exact full gradients over long trajectories, preserving the dominant sensitivities required for effective online learning and substantially reducing computational and memory overhead.

We evaluate the proposed methods using two case studies of increasing complexity. We first consider a hybrid neural–Lorenz-63 model in which an AI component compensates for missing dynamics. The framework is then extended to a semi-realistic hybrid quasi-geostrophic model of the Northwestern Mediterranean Sea, demonstrating two complementary enhancement strategies: the calibration of a biased physical parameter (bottom drag) and a NN-based correction of bottom-layer momentum tendencies. Together, these experiments show that our Jacobian-approximation strategies enable stable and efficient online learning across both low-dimensional chaotic systems and high-dimensional ocean models. Although our configurations remain simpler than fully operational ocean models, our results provide a foundation for scaling online learning to realistic ocean applications and, ultimately, for integrating AI-based corrections into next-generation forecasting systems.

How to cite: González Zamora, E., Ouala, S., and Tandeo, P.: Efficient Gradient-Approximation Methods for Online Learning in Hybrid Neural–Physical Ocean Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2346, https://doi.org/10.5194/egusphere-egu26-2346, 2026.

We present the development of a next-generation family of Lat–Lon–Cap (LLC) global ocean simulations, culminating in LLC8640, a 1/96^◦ (≈ 1 km) realistic global 'nature run’ that, once complete, will represent the highest-resolution global ocean model produced under realistic conditions. This effort advances well beyond the widely used LLC4320 configuration by addressing long-standing dynamical biases through coordinated improvements in resolution, physical formulation, and forcing.

Key advances include increased vertical and horizontal resolution, updated global bathymetry, non-linear free surface, explicit ice-shelf cavities around Greenland and Antarctica, hourly atmospheric forcing, realistic river discharge, and improved astronomical tidal forcing. Together, these developments directly target deficiencies in earlier LLC models, including a misplaced Gulf Stream, a crude representation of Antarctic shelf circulation, and weak tropical instability waves. Particular emphasis is placed on the equatorial ocean, where Green’s-function-based approaches are used to optimize turbulence parameterizations and reduce persistent discrepancies between global models and observations. Early results from the ongoing lower-resolution spin-up already demonstrate markedly improved realism, including a more accurate Gulf Stream path and a strengthened, more realistic equatorial undercurrent.

The modeling strategy employs a staged spin-up across resolutions: a multi-year 1/12^◦ (LLC1080 ) integration to equilibrate large-scale circulation and kinetic energy; a subsequent 1/48^◦ (LLC4320 ) phase to sharpen mesoscale and submesoscale dynamics; and a final month-long 1/96^◦ (LLC8640 ) integration producing several petabytes of hourly three-dimensional velocity, temperature, and salinity fields. The resulting dataset will provide an unprecedented global benchmark for studies of internal tides and waves, submesoscale turbulence and mixing parameterizations, and SWOT-era sea-surface height variability.

How to cite: Momeni, K., Menemenlis, D., Zhang, K. Q., and Peltier, W. R.: A Trailblazing Global Ocean Simulation in the Time of Wide Swath Altimetry, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2382, https://doi.org/10.5194/egusphere-egu26-2382, 2026.

The ocean floor is littered with seamounts, most of which are volcanic in origin. Seamounts are important in the marine geosciences because they are oceanographic ‘dip-sticks’, biodiversity hotspots, scatterers of tsunami waves, and hazards for navigation. Research ships with single beam echo-sounders have discovered many small seamounts and some large ones while satellite altimetry has led to discovery of many large seamounts and some small ones. The exact number of seamounts in the world’s ocean basins and their margins remains, however, unknown.  Here we use machine learning in an attempt to locate all seamounts, to estimate their height and volume and to speculate on their origin. We use the seamounts found by Hillier & Watts (2007) along ship track from single beam echo-sounder data acquired on 5585 individual research cruises during 1950 to 2002 as a ‘training’ data set and the SRTM15+V2.7 (GEBCO 2025) topographic grid that combines shipboard single beam and multibeam (swath) bathymetry data acquired on 2154 individual research cruises during 1980 to 2024 with predicted bathymetry from satellite altimeter data in regions of sparse ship tracks to determine the 6 main attributes (channels) of seamounts, 4 of which refer to their slopes. We then use the SRTM15+V2.7 (GEBCO 2025) topographic grid together with machine learning to update the global seamount census of Hillier & Watts (2007). Preliminary results in two pilot study areas on old and young oceanic crust in the Pacific Ocean indicate that machine learning yields up to a factor of 2 more seamounts than were identified in the training data set. The implications of these results are examined for volcanism on Earth and on other terrestrial planets.

References:

Hillier, J.K., Watts, A.B., 2007. Global distribution of seamounts from ship-track bathymetry data. Geophys. Res. Letts. 34, 1-5, doi:10.1029/2007GL029874.

Tozer, B., Sandwell, D.T., Smith, W.H.F., Olson, C., Beale, J.R., and Wessel, P., 2019 Global Bathymetry and Topography at 15 Arc Sec: SRTM15+. Earth and Space Science 6, doi:10.1029/2019EA000658

How to cite: Wang, Z. and Watts, A. B.: Global distribution of seamounts from machine learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2457, https://doi.org/10.5194/egusphere-egu26-2457, 2026.

The unprecedented 2025 coral bleaching event in the Ningaloo Coast World Heritage Area highlights that even climate refugia are vulnerable to severe thermal stress. Here we use a deep learning approach to downscale sea surface temperature (SST) from five select CMIP6 models to a 10 km resolution. The quantile delta mapping method is used to correct ~1°C warm bias in SST from climate models determined by satellite observations. We used the NOAA SST satellite observations as the training dataset to resolve coastal regions in the Ningaloo and the Exmouth Gulf. We compare the results with those obtained using the SST training data from an eddy resolving ocean model. Our projections show that SST in the Ningaloo will warm by about 0.5°C at a 1.5°C global warming level, increasing to over 1.5°C at a 3°C level. This projected warming leads to a substantial increase in Degree Heating Weeks (DHWs), suggesting that the coral bleaching event of 2025 will likely become more common in the future.

How to cite: Sun, C., Cyriac, A., Copcutt, M., Matear, R., and Tatylor, J.: Deep learning to downscale future climate projections to assess future coral bleaching risks for the Ningaloo Reef, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4195, https://doi.org/10.5194/egusphere-egu26-4195, 2026.

In this study, a hybrid architecture combining convolutional neural networks for spatial reconstruction and long short-term memory networks for temporal forecasting is used to predict sea-level variations in the German Bight. This new framework is applied to a series of sea level data ranging from academic to realistic data. Experiments with monochromatic waves demonstrate the model’s ability to deliver accurate short-term forecasts with minimal errors. Forecasts of real tidal constituents, including M2 and the sum of M2 and M4 tides, confirm robust model performance over lead times up to 48 h. A key result is that deep learning can reconstruct basin-wide sea level from a limited number of coastal gauge stations. Therefore, in the forecast experiments, adding data from coastal observations (mimicking data assimilation) significantly improves prediction accuracy. The study highlights the potential of deep learning to supplement traditional numerical models, particularly in regions with dense observational coverage. Key factors influencing model performance are identified, among them spatial signal complexity and steepness of gradients. An overall result is that deep learning can complement numerical models in operational ocean forecasting and provide a valuable tool for evidence-based coastal management in data-rich regions.

How to cite: Yildiz, I., Stanev, E. V., and Staneva, J.: From monochromatic waves to realistic tides: deep learning for short-term forecasting of coastal ocean, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5209, https://doi.org/10.5194/egusphere-egu26-5209, 2026.

Incorporating physical laws into neural networks has long been a central topic in geophysical machine learning. While purely data-driven approaches can achieve strong prediction skill, they often lack physical consistency and degrade under sparse observations or long lead times. In this study, we impose a simple yet fundamental constraint, global volume conservation, by introducing a dedicated volume-conserving layer into neural networks. We apply this volume-conserved network in both an idealized shallow-water model and a realistic global sea level anomaly prediction task, and show systematic improvements in prediction skill, reaching up to 25%. The improvement increases as observation points decreasing and leading time increasing, and the predictions follow physical laws strictly. In addition, although post-processing also enforce physical consistency, the constrained model achieves substantially lower prediction errors, with reductions of up to 15%. These results demonstrate the effectiveness of embedding hard physical constraints as network layers for improving both accuracy and physical fidelity.

How to cite: Li, Y. and Tang, Y.: Improving Global Sea Level Prediction with Hard Physical Constraints in Neural Networks, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6777, https://doi.org/10.5194/egusphere-egu26-6777, 2026.

How to cite: Carsey, S., Hornschild, A., and Saynisch-Wagner, J.: Development of an Observational Encoder for Data Assimilation in the Latent Space of a Variational Autoencoder (with Sea Surface Temperature) , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-6860, https://doi.org/10.5194/egusphere-egu26-6860, 2026.

Growing evidence suggests that the present-day Atlantic Meridional Overturning Circulation (AMOC) operates in a bistable regime and may transition to a weakened or collapsed (“OFF”) state under climate change forcing, with severe global climate impacts. In addition to deterministic forcing, stochastic variability can induce noise-induced transitions between stable AMOC states. Quantifying the probability and pathways of such transitions is therefore critical.

Previous work (Soons et al., 2024) applied Large Deviation Theory (LDT) to a stochastic box ocean model (Wood et al., 2019) to estimate the most probable pathways for noise-induced AMOC collapse and recovery. While effective, this approach requires explicit knowledge of system properties, such as the Jacobian, limiting its applicability to higher-dimensional, more complex climate models.

Here, we adapt a recently proposed deep reinforcement learning framework (Lin et al., 2025) to compute most probable transition pathways in stochastic dynamical systems without prior knowledge of the governing equations. Applied to the stochastic box ocean model, the method robustly identifies physically consistent collapse and recovery pathways, comparable to those obtained using LDT. Finally, we demonstrate the feasibility of this framework in a more complex ocean model.

How to cite: Guardamagna, F. and Dijkstra, H.: Estimating Most Probable AMOC Collapse and Recovery Pathways Using Deep Reinforcement Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7643, https://doi.org/10.5194/egusphere-egu26-7643, 2026.

Sea surface temperature (SST) is critically important for understanding ocean dynamics and supporting various marine activities, making accurate short-term SST forecasting highly significant. However, accurately modeling the multi-scale variability of SST remains challenging for existing deep learning (DL) models. This study introduces the coupled Transformer–CNN network (CoTCN), a hybrid architecture designed to leverage the multiscale variability of SST. The CoTCN combines the strengths of Transformers and convolutional neural networks (CNNs), significantly enhancing SST forecasts’ spatial continuity and predictive accuracy. Compared to five state-of-the-art DL models based on Transformers or CNNs that include convolutional long short-term memory (ConvLSTM), ConvGRU, adaptive Fourier neural operator (AFNO), PredRNN, and SwinLSTM, the CoTCN demonstrates superior performance in global and local areas of SST forecasting. At 1-day lead time, the CoTCN reduces the global average root-mean-square error (RMSE) by over 15%, with forecast errors ranging from 0.20 °C to 0.53 °C across 1–10-day lead times. Moreover, the CoTCN effectively mitigates the checkerboard artifacts inherent to the Vision Transformer (ViT) architecture. These findings highlight the effectiveness of the CoTCN in capturing SST’s multiscale features and underscore the promising potential of hybrid architectures for future DL models.

How to cite: Zhang, T., Lin, P., Liu, H., Wang, P., Wang, Y., Xu, K., Zheng, W., Li, Y., Jiang, J., Zhao, L., and Chen, J.: A Coupled Transformer-CNN Network: Advancing Sea Surface Temperature Forecast Accuracy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8845, https://doi.org/10.5194/egusphere-egu26-8845, 2026.

Sea surface height (SSH) derived from satellite altimetry is essential for oceanographic research and marine monitoring. To improve SSH prediction accuracy, we propose a set of physics-informed methods based on neural networks (NNs). The main strategies include: (1) integrating a geostrophic constraint (GC) into the loss function; (2) incorporating land mask information (MI) to mitigate artifacts introduced by the land points in ocean data.

Utilizing altimeter satellite gridded absolute dynamic topography data, we evaluate three mainstream spatiotemporal predictive NNs—SimVPv2 (SV), PredRNNv2 (PR), and PredFormer (PF)—each exhibiting distinct inductive biases inherent to their architectures, to assess their performance under the proposed strategies. The results indicate that both strategies can significantly improve SSH prediction, though their effects vary across architectures. While SV shows limited improvement from MI, PR benefits the most, which can likely be attributed to its gating mechanism and recurrent architecture. In contrast, GC enhances the performance of SV more effectively than that of PR. However, both strategies degrade the performance of PF, a Vision Transformer (ViT)-based model that differs fundamentally from SV and PR. To our knowledge, this study is the first to identify land-induced artifacts in spatiotemporal predictive NNs and to implement a land mask input strategy to mitigate their impact on ocean forecasting.

Building upon these findings, we further explored the potential of multivariable inputs. Contrary to expectations, our experiments of concatenating wind speed with SSH as inputs reveal that directly combining heterogeneous oceanic variables is suboptimal. This finding highlights a broader multimodal integration problem in applying NNs to oceanography, which remains an open challenge.

How to cite: Huang, L., Shu, Y., Yao, J., and Liu, D.: Investigation of Physics-Informed Methods for Improving Sea Surface Height Prediction Based on Neural Networks , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9123, https://doi.org/10.5194/egusphere-egu26-9123, 2026.

Harmful algal blooms (HABs) pose a persistent challenge to coastal ecosystems, fisheries, and public health, particularly in urbanized coastal regions subject to strong hydrodynamic forcing and meteorological variability. HAB dynamics emerge from the interaction of biologically driven growth processes and physically governed transport and dispersion, operating across disparate spatial and temporal scales. However, most existing data-driven forecasting approaches treat these processes implicitly and holistically, limiting physical interpretability, robustness under nonstationary forcing, and the ability to represent forecast uncertainty.

This study proposes a physics-informed diffusion-based framework for HAB forecasting in coastal environments, with the objective of explicitly separating biological and physical drivers within a generative probabilistic model. The central hypothesis is that decoupling meteorological and hydrodynamic forces can improve the physical consistency and generalizability of HAB forecasts while enabling uncertainty-aware prediction. To this end, future HAB states are formulated as conditional samples generated through a reverse diffusion process guided by physically meaningful environmental inputs.

The proposed framework adopts a dual-forcing architecture. A meteorological branch encodes atmospheric variables—including air temperature, precipitation, wind speed, and radiative forcing—that primarily regulate phytoplankton growth potential and bloom initiation. In parallel, a hydrodynamic branch incorporates tidal dynamics and wave-related information to represent advection, mixing, and dispersion processes governing the spatial evolution of algal biomass in coastal waters. Physical consistency is promoted by embedding the advection–diffusion equation as a soft constraint within the hydrodynamic latent space, encouraging mass-conserving and physically plausible transport behavior without imposing a fully deterministic dynamical model.

By leveraging diffusion probabilistic modeling, the framework generates ensemble-based forecasts that characterize the conditional probability distribution of future HAB states rather than single deterministic trajectories. Forecast outputs are formulated in terms of a probabilistic HAB severity index, facilitating interpretable, risk-informed early warning analogous to probabilistic weather forecasting systems. Model development is designed to integrate multi-source environmental datasets, including high-frequency meteorological observations, wave and tidal records, and routine coastal water-quality monitoring.

The framework is developed with a focus on tidally energetic coastal systems, with the Hong Kong coastal region serving as a representative application domain. Overall, this study outlines a physically interpretable and uncertainty-aware modeling paradigm for HAB forecasting and provides a conceptual foundation for next-generation early-warning systems in coastal environments.

How to cite: Liu, Z.: Physics-Informed Diffusion Model for HAB Forecasting in Hong Kong Coastal Waters, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9163, https://doi.org/10.5194/egusphere-egu26-9163, 2026.

The SWOT (Surface Water Ocean Topography) mission is currently providing unpreceded high-resolution measurements of Sea Surface Height (SSH), revealing ocean features at finer scales. Nevertheless, the two-dimensional observations of KaRIn altimeter of SWOT suffer from instrumental and geophysical correction errors. This noise degradation is polluting the high frequencies of SWOT signal, thus hiding the submesoscale dynamics from oceanographers. For this reason, Tréboutte et al. (2023) has developed a convolutional neural network (CNN) based on UNet architecture to separate the noise from the physical signals contained in the SSH. This work has already demonstrated great results on SWOT measurements. However, last version of the algorithm delivers poor performance in certain oceanic conditions. Therefore, we modify the training procedure to obtain a more robust version of the algorithm. We show that we manage to mitigate these issues significantly, avoiding biases and artefacts in the denoised observations.

This data is also incomplete. SWOT measurements are sometimes distorted by various factors, such as rain cells, boats, icebergs, etc. To address these errors, editing is applied to remove erroneous pixels from the data. However, this lost data is valuable to many users. That is why we have also developed a deep learning inpainting methodology using a CNN to retrieve the missing physical information. We demonstrate that it is possible to accurately restore measurements lost after the editing step, better than classical interpolation approaches.

How to cite: Meis, G., Tréboutte, A., Pujol, M.-I., Ballarotta, M., and Dibarboure, G.: Enhancing two-dimensional SWOT oceanic measurements using deep learning approaches for denoising and inpainting, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11229, https://doi.org/10.5194/egusphere-egu26-11229, 2026.

Global ocean circulation has a significant impact on climate variability, where ~80% of the ocean energy transfer occurs in small-scale processes. While the existing record of altimetry goes back thirty years and has enabled the assimilation of gridded sea surface height maps, their operational resolution of 0.25° is not high enough to study these mesoscale eddies, and we are therefore in need of methods that can improve their resolution globally. 

The recently launched SWOT satellite with ~2km resolution now offers the first data record with sufficient resolution to reveal these processes in observations, and offers the possibility of drastically improving sea surface state maps. However, its sparse temporal and spatial record brings challenges for global assimilation. 

We propose a generative machine learning approach to downscale existing gridded Level 4 sea surface height to the fine resolution of SWOT. Our methodology involves two steps: first, training a conditional diffusion downscaling model on high resolution simulated data as a prior joint distribution over sea state observations, including height, temperature and salinity. Secondly, a data assimilation scheme via a Bayesian posterior formulation that generates high resolution sea surface state maps assimilated with a set of observations. We evaluate our methodology both in simulated and observing system experiments that demonstrate the efficacy of our approach as well as their scalability to global context in evaluations of major currents. Under the Bayesian formulation we also find that the diffusion model produces well calibrated predictive uncertainty estimates, which further underlines the applicability of diffusion models as a computationally efficient method in this domain. Our high resolution sea surface height maps open up new insights into mesoscale eddies.

How to cite: Lehmann, N., Shah, A., Bamber, J., and Zhu, X.: OceanBottle: Sea Surface State Data Assimilation and Downscaling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11308, https://doi.org/10.5194/egusphere-egu26-11308, 2026.

Exchange of water masses across the Greenland-Scotland ridge is an important part of the AMOC. The complicated bathymetry of the ridge is not properly resolved in standard CMIP models with around 1 degree resolution and this reduces confidence in simulated exchanges and their variability. Previously, a nested-domain approach with finer resolution in selected areas has been applied. Here, we perform a pilot study of the alternative approach of a fine-resolution ocean emulator. We use daily 3D salinity and temperature fields of the GLORYS reanalysis in original (target) and 4x reduced (input) resolution and demonstrate that a fine-resolution emulator consisting of a simple U-net architecture trained on 10 years of  input/target can be used to reconstruct the target field from the input fields outside the training period with a significant skill compared to simple interpolation. We apply temporal and spatial scrambling to assess input feature importance. Our study suggests that the fine resolution model in a nested setup can be replaced with an ocean emulator leading to substantial gains in overall execution speed.

How to cite: Schmith, T., Beauchamp, M., Devilliers, M., Gierisch, A., and Olsen, S.: Fine-resolution ocean emulator for the Greenland-Scotland Ridge, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11454, https://doi.org/10.5194/egusphere-egu26-11454, 2026.

The Argo Program is a global network of 4,000 autonomous drifting floats that provide essential, real-time data on the upper 2,000 meters of the ocean. By measuring temperature and salinity, Argo has become the primary source of information for monitoring ocean warming, sea-level rise, and climate variability. However, the massive volume of data generated—totaling millions of profiles—presents a significant challenge for Quality Control (QC).

Traditionally, delayed-mode quality control has relied heavily on human expertise and the "trained eye" of scientists to identify instrumental drifts and sensor malfunctions. To address the cost and limitations of manual inspection, we introduce Argo-YOLO, an innovative approach that transposes computer vision techniques into the field of physical oceanography.

By converting oceanographic profiles into graphical representations, our system utilizes the YOLO (You Only Look Once) deep learning architecture to "scan" the data, mimicking the visual diagnostic capabilities of expert oceanographers. This method enables high-speed, systematic detection of instrumental drifts, sensor malfunctions, and profile anomalies across the entire Argo dataset while maintaining the nuanced precision of human analysis.

Initial results demonstrate that Argo-YOLO faithfully reproduces expert visual diagnostics with high performance: 97% accuracy in identifying valid profiles with only 3% false alarms, and 96% success in detecting anomalous profiles with 4% missed detections.

These results confirm the viability of computer vision for operational oceanographic quality control.

Argo-YOLO demonstrates how computer vision can be successfully adapted to oceanographic challenges, representing a major step toward automated, scalable quality control in global ocean observing systems and ensuring the integrity of long-term climate records in an era of "Big Data" oceanography.

How to cite: Carval, T., Tosello, V., Dobler, D., and Lebeaud, A.: Argo-YOLO: Leveraging Computer Vision for Automated Quality Control of Argo Ocean Profiles, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11527, https://doi.org/10.5194/egusphere-egu26-11527, 2026.

Decadal climate predictions are essential for climate adaptation, yet remain challenging due to the complex interplay of initial conditions and external forcings. A key factor in achieving skillful forecasts is the upper ocean, which plays a central role in modulating decadal-scale climate variability, including phenomena such as ENSO, the Atlantic Multidecadal Variability, and the Indian Ocean Dipole. Accurately capturing the ocean’s memory is therefore critical, but traditional numerical models are computationally demanding and often exhibit systematic biases. While machine learning has shown promise in improving medium-range weather forecasts, its application to decadal climate prediction remains limited.

This work explores the feasibility of using machine learning to predict sea surface temperature (SST) and ocean heat content (OHC) on decadal timescales. We develop an autoregressive model based on a UNet-like convolutional neural network, trained on 1,000 years of data from a pre-industrial control run from the fully coupled MPI-ESM. This simulation provides a controlled environment to study predictability arising from internal ocean dynamics. Inputs include SST, OHC, a land-sea mask, and top-of-atmosphere solar radiation to encode the seasonal cycle. We conduct a systematic study of input design, to assess how the representation of past states influences model stability and predictive skill. Our results suggest that machine learning can be a viable and flexible approach for decadal ocean prediction. Additionally, we find that longer input windows and coarser resolution may improve long-term stability, potentially offering new insights into how climate memory is encoded.

How to cite: Meyer, F., Kadow, C., and Baehr, J.: Machine Learning for Decadal Ocean Prediction - Exploring the Feasibility of Capturing Climate Memory in the Upper Ocean, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12894, https://doi.org/10.5194/egusphere-egu26-12894, 2026.

Substantial natural variability can obscure the detection of anthropogenic long-term trends in the marine carbonate system (e.g., ocean acidification). Yet the magnitude of the trends and variability remains poorly constrained due to limited marine carbonate system observations. Here, we use a Bayesian machine-learning approach based on Gaussian Process Regression (GPR) to decompose total variability of ocean acidification-related variables into seasonal, interannual and long-term components. The method is first applied to three decades of observations from the Line P carbon program, the longest marine carbonate system timeseries in the Northeast Pacific (1990-2019), typically taking samples three times per year. We found that over the period from 1990 to 2019, the local oceanic uptake of anthropogenic carbon dioxide from the atmosphere was the main driver of long-term changes in the marine carbonate system, including acidification. The seasonal cycle of dissolved inorganic carbon and the aragonite saturation state (both indicators of ocean acidification) was the dominant contributor to total variability in the top 60-70 m of the water column, with a mean surface seasonal amplitude of 35 ± 3 µmol kg−1 and 0.31 ± 0.04, respectively. In this depth range, the magnitude of the interannual variability was at least half of the seasonal variability for most variables. We then apply GPR to output from a global ocean biogeochemical model subsampled as per availability of observations, to assess the observational effort required to detect future ocean carbon trends, with a particular focus on detecting signals related to potential marine carbon dioxide removal interventions.

How to cite: Franco, A. C., Monahan, A. H., Ianson, D., and Bernardello, R.: Using Gaussian Process Regression to disentangle marine carbonate system trends and variability, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12925, https://doi.org/10.5194/egusphere-egu26-12925, 2026.

Climate change is reshaping ocean ecosystems faster than we can monitor them. Predicting shifts in productivity, carbon uptake, and oxygen levels requires forecasting interacting biogeochemical variables, a task where traditional process-based models struggle with computational cost and parameter uncertainty.  

BG4Sea is a machine-learned seasonal forecast that was trained on Mercator Océan's operational biogeochemical analysis. The model can generate skillful seasonal predictions of the carbon cycle, nutrients, oxygen, pH, chlorophyll, and plankton dynamics at a fraction of the computational cost, all while remaining competitive even at longer forecasting horizons. However, while the model demonstrates skill when evaluated against reanalysis data, this is likely to share the parametrization assumptions and constraints that are characteristic of process-based models.

This contribution explores strategies for evaluating against real-world measurements and for using observations to guide and constrain the model. We investigate “global-first” approaches, which prioritize remote-sensing data, as well as “regional-first” approaches, which use the model’s grid-independent structure to produce region-specific updates from in-situ stations.

How to cite: Martinez Balbontin, G., Charantonis, A., Bereziat, D., and Ciavatta, S.: Guiding Machine-Learned Biogeochemical Forecasts with Observations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13007, https://doi.org/10.5194/egusphere-egu26-13007, 2026.

Abstract:  Argo floats provide a global dataset of subsurface  temperature and salinity profiles but lack direct velocity observations. This limits the reconstruction of Lagrangian ocean transport using the Argo data. We propose a physics-informed machine learning emulator that infers latent horizontal velocity fields from Argo hydrographic observations. The model learns a neural velocity representation using 3D temperature–salinity gradients, which is constrained by advection–diffusion equations. This approach implicitly recovers flow patterns that are consistent with the observed changes in properties and enables the simulation of synthetic trajectory without the input of explicit velocity data. Sparse years are handled via physics-based self-supervision and spatio-temporal regularization. Preliminary experiments in the Mediterranean Sea demonstrate that the learned velocities reproduce qualitatively the known major gyres and boundary currents, achieving realistic float displacements and energy spectra that are comparable to those in reanalysis fields. This framework offers a new way to reconstruct Lagrangian dynamics directly from hydrography, providing an efficient, observation-driven alternative to numerical trajectory modeling.

How to cite: Gasana Elysee, M., Pirro, A., Poulain, P.-M., Mauri, E., Manzoni, L., and Menna, M.: Learning Implicit Subsurface Velocity Fields from Argo Hydrography Using Physics-Informed Neural Emulation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13260, https://doi.org/10.5194/egusphere-egu26-13260, 2026.

Accurately forecasting Sea Surface Temperature (SST) is critical for understanding ocean dynamics, climate change impacts, and marine ecosystem management (Brito-Morales et al., 2020; Gattuso et al., 2018). In recent years, Graph Neural Networks (GNNs) have emerged as a powerful tool for spatiotemporal oceanographic forecasting, offering advantages over traditional Euclidean deep learning models by operating on unstructured grids (Liang et al., 2023; Zhang et al., 2025). However, the transition from structured satellite-derived data to mesh-based representations often introduces numerical artifacts, particularly due to the grid-to-mesh coupling mechanisms (Cuervo-Londoño et al., 2026; Cuervo-Londoño, Sánchez, et al., 2025).

This study investigates the origin of "Voronoi-induced artifacts" in GNN architectures applied to SST forecasting in the Northwest African region and the Canary Islands. We demonstrate that the grid-to-mesh association is algebraically equivalent to an order-k Voronoi partition (Cuervo-Londoño, Reyes, et al., 2025; Okabe et al., 2000), implying that the way nodes are distributed and how they associate with the underlying data grid significantly influences the quality of the predictions. To address these issues, we propose and evaluate four different mesh configurations: structured quadrangular meshes (Holmberg et al., 2024; Lam et al., 2023) and three unstructured approaches, including novel bathymetry-aware meshes.

Our findings reveal that connectivity plays a decisive role in mitigating artifact formation. Specifically, using approximately four connections per node under optimized grid-to-mesh association rules significantly reduces errors. Furthermore, the results show that densifying the node distribution according to the seabed’s topography (bathymetry) not only reduce spatial artifacts but also increases forecast accuracy. The bathymetry-based meshes with optimized connectivity (3-4 connections) achieved a 30% improvement in performance compared to traditional structured mesh baselines. These insights suggest that incorporating geographical and topological priors into GNN design is essential for developing robust and reliable machine-learning surrogates for physical oceanography (Reichstein et al., 2019).

Acknowledgments: This work was supported by the projects SIRENA and SIRENA 2, funded by the collaboration of the Biodiversity Foundation of the Ministry for the Ecological Transition and the Demographic Challenge, through the Pleamar Program, and are co-financed by the European Union through the EMFAF (European Maritime, Fisheries and Aquaculture Fund).

References

Cuervo-Londoño, G. A., Reyes, J. G., Rodríguez-Santana, Á., & Sánchez, J. (2025). Voronoi-Induced Artifacts from Grid-to-Mesh Coupling and Bathymetry-Aware Meshes in Graph Neural Networks for Sea Surface Temperature Forecasting. Electronics, 14(24), 4841. https://doi.org/10.3390/electronics14244841

Cuervo-Londoño, G. A., Sánchez, J., & Rodríguez-Santana, Á. (2025). Deep Learning Weather Models for Subregional Ocean Forecasting: A Case Study on the Canary Current Upwelling System (No. arXiv:2505.24429). arXiv.https://doi.org/10.48550/arXiv.2505.24429

Cuervo-Londoño, G. A., Sánchez, J., & Rodríguez-Santana, Á. (2026). Forecasting Sea Surface Temperature from Satellite Images with Graph Neural Networks. In M. Castrillón-Santana, C. M. Travieso-González, O. Deniz Suarez, D. Freire-Obregón, D. Hernández-Sosa, J. Lorenzo-Navarro, & O. J. Santana (Eds.), Computer Analysis of Images and Patterns (pp. 329–339). Springer Nature Switzerland. https://doi.org/10.1007/978-3-032-05060-1_28

How to cite: Cuervo Londoño, G. A., Rodríguez Santana, Á., and Sánchez, J.: Mitigating Voronoi-induced artifacts in GNN-based sea surface temperature forecasting using bathymetry-aware adaptive meshes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15146, https://doi.org/10.5194/egusphere-egu26-15146, 2026.

Artificial Neural Networks (ANN) are applied to estimate the interannual variability of monthly-mean satellite-derived chlorophyll-a (CHL) at a global scale in the 1997-2025 period, as function of different physical variables and climate teleconnection indices. Among other variables, satellite-derived sea-surface height (SSH) proved to be a good single predictor for the CHL, showing significant CHL-SSH correlation in most of the world ocean between 60°S and 60°N (where the most continuous data series are available). This correlation, generally low for a linear estimation, opens the possibility to CHL reconstruction using higher-performance non-linear techniques like ANN. The ANN-model successfully reproduces the CHL interannual variability: 59% of the modeled CHL present correlations > 0.90. Then, the ANN-model can be used to predict CHL beyond the training period, showing a good predictability at least one season ahead. On the other hand, a similar exercise for the reconstruction/predictability of CHL is subsequently carried out using selected teleconnection indices as predictors, presenting an alternative simpler method to estimate the CHL variability in key regions along the world ocean. Thus, the proposed methods open the possibility to predict not only CHL but other related biogeochemical variables.

How to cite: Rivas, D.: Estimation of global satellite-derived chlorophyll-a as function of physical drivers using shallow neural networks, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15274, https://doi.org/10.5194/egusphere-egu26-15274, 2026.

Real-time, high-fidelity ocean state estimation is a prerequisite for Earth system digital twins, yet faces a dilemma between the computational bottlenecks of traditional assimilation and the grid-based fidelity losses of deep learning. Here we present ADAF-Ocean, a geometry-agnostic framework that resolves this by assimilating multi-source observations directly at their original resolutions. Leveraging a neural process-based architecture, our approach learns a continuous mapping from heterogeneous inputs, such as sparse profiles and satellite imagery, thereby maximizing information extraction while enforcing multivariate physical consistency. Although purely data-driven, ADAF-Ocean is capable of implicitly learning the coupling patterns between thermodynamic and kinematic variables directly from high-fidelity datasets. Evaluations show that superior analysis accuracy gives rise to emergent physical coherence.  Serving as superior initial conditions for a DL forecast model, these coherent fields sustain a significant forecast skill advantage for up to 20 days. Furthermore, by quantifying the contribution of individual observational sources, this framework establishes a trustworthy pathway for AI-driven oceanography, bridging data-driven efficiency with the rigorous standards of Earth system monitoring.

How to cite: Xiang, Y.: Advancing Ocean State Estimation with efficient and scalable AI, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15437, https://doi.org/10.5194/egusphere-egu26-15437, 2026.

The Surface Water and Ocean Topography (SWOT) satellite mission, launched in December 2022, provides revolutionary measurements of the sea surface height (SSH) variations with unprecedented spatial resolution down to Ο(1 km). As a result, SWOT products have significant potential in monitoring ocean dynamics down to the submesoscale. However, the repeat cycle of 21 days introduces a barrier to fully capture these dynamics as they vary on the order of days. To fully exploit the potential of the satellite mission and simplify processing requirements for potential users, we propose a physics-informed neural network (PINN) to generate gridded SSH products from SWOT L3 along-track snapshots. The neural network has a U-Net-like architecture combined with residual learning to consider the spatial variations of the SSH field, and takes time, geolocations, and gridded SSH from conventional altimetry missions as input features, while the SWOT observations serve as ground truth. In addition to the classical data loss, the PINN model applies direct constraints on the model's trainable parameters by forcing them to fulfill the next-order correction of the quasi-geostrophic theory (SQG⁺¹), which has been demonstrated to be able to capture cyclogeostrophic balance and frontogenesis attributed to submesoscale dynamics. To this end, the high resolution of SWOT observations is kept, while the velocities and pressure fields associated with the SQG⁺¹ theory are predicted. We conducted experiments using both simulated data and real-world data. Both experiments demonstrate the benefits of incorporating physical loss to achieve higher generalizability, thereby filling the gaps between SWOT tracks reasonably. Based on the real-world data, 2-km gridded SSH products with a temporal resolution of five days are achieved. The proposed method shows promising potential for generating high-resolution gridded products while considering physical constraints. The product will be beneficial for the community to analyze mesoscale to submesoscale ocean dynamics, and compare with other sources of surface and in-situ data in the upper ocean.

How to cite: Gou, J., Dù, R. S., Smith, K. S., Soja, B., and Bodner, A.: Physics-informed neural network for gridded SSH from SWOT observations considering the next-order balanced model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15749, https://doi.org/10.5194/egusphere-egu26-15749, 2026.

The Kuroshio Current is a western boundary current in the northwestern Pacific, and its transport and path variability significantly affect air-sea interactions, thus modulating North Pacific climate, as well as ecosystems. The Japan Meteorological Agency (JMA) 137°E repeat hydrographic section, occupied every winter from 1967 to 2023 (57 years), provides a long and consistent benchmark for diagnosing the variability of the Kuroshio Current system. Here, we analyze these repeated occupations and derive the vertical structure of zonal geostrophic velocity and associated transport. Our analysis reveals that the Kuroshio Current system exhibits substantial variability and intrinsic asymmetry in its transport, axis position, and vertical hydrographic structure. To capture the asymmetric hydrographic patterns associated with these transport fluctuations, we extract leading variability in the vertical structure using empirical orthogonal functions and apply a 1×5 self-organizing map (SOM) to classify distinct circulation patterns. The SOM yields five physically interpretable nodes: two large-meander (LM) nodes (moderate and extreme) and three distinct non-LM nodes. The extreme LM node features a southward shift of the Kuroshio axis to around 30°N accompanied by a significant weakening of the recirculation gyre. Moderate LM events exhibit a less pronounced southward shift near 31°N. The non-LM nodes can be characterized by (i) strengthened recirculation with near-normal net transport, (ii) enhanced net eastward transport, and (iii) reduced net transport. The heaving of isopycnal lines mostly accounts for thermohaline anomalies throughout the nodes, whereas spicing plays a partial role only in the extreme LM node. This study argues that variation in the thickness of the Subtropical Mode Water (STMW) accounts for upper ocean heat content and consequently for volume transport, underpinning STMW thickness as a metric integrating variability across the Kuroshio Current system along the 137°E section.

How to cite: Park, H.-J., Kim, Y. S., and Na, H.: Nodes of the Kuroshio Current system from multidecadal repeat observations along 137°E, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16459, https://doi.org/10.5194/egusphere-egu26-16459, 2026.

How to cite: Tromp, W., Zhao, J., and Verlaan, M.: Machine learning emulators for predicting storm surges in the North Sea , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16754, https://doi.org/10.5194/egusphere-egu26-16754, 2026.

Harmful algal blooms (HABs) are massive proliferations of microalgae in aquatic ecosystems that may be harmful to the ecosystems or to society. Predicting HABs in spatially complex coastal environments requires understanding the potential environmental drivers that may determine microalgal population dynamics. When considering the study of HABs we may evaluate if these processes are spatially invariant or if they demonstrate site-specific dynamics. Machine learning models often achieve high training performance but fail when extrapolating to unseen locations due to site-specific overfitting. We developed a methodological framework integrating hierarchical modelling, spatially explicit machine learning, and interpretable AI techniques to quantify spatial heterogeneity in HAB environmental drivers.

Gambierdiscus spp is a genus of benthic marine microalgae (dinoflagellate) that are found in coastal areas and that produce potent marine toxins which are transferred mainly to fish. We analysed 348 observations of Gambierdiscus spp. abundances across 32 sites in the Balearic Islands (2021-2024), integrating field abundance data with satellite-derived oceanographic variables (temperature, nutrients, hydrodynamics) from Copernicus Marine Service. Seven modelling approaches were compared: Generalized Additive Mixed Models (GAMM), Generalized Additive Models (GAM), Geographically Weighted Regression (GWR), Random Forest (RF), Geographic Random Forest (GRF), XGBoost, and Geographic XGBoost. A three-phase feature selection procedure (temporal lag optimization, collinearity removal via VIF, LASSO regularization) reduced 61 candidate predictors to 12 ecologically interpretable variables optimized for spatial modelling.

Model validation employed Leave-One-Out Cross-Validation (LOO-CV) to assess true spatial generalization rather than interpolation. Machine learning models achieved high training performance (R²=0.75-0.85) but collapsed under spatial extrapolation (R²_LOO=0.30-0.40). In contrast, GAMM demonstrated superior spatial transferability (R²_LOO=0.47), attributable to its explicit separation of fixed environmental relationships from hierarchical site-specific random effects. SHAP (SHapley Additive exPlanations) analysis on island-stratified Random Forest models quantified spatial non-stationarity: temperature importance varied 13-fold across islands (SHAP: 0.05-0.64), while phosphate exhibited 2.6-fold consistency (SHAP: 0.10-0.26). Partial dependence plots verify that drivers operate through fundamentally different mechanisms across the archipelago.

Significant spatial clustering (Moran's I=0.346, p<0.001) with persistent hotspots and coldspots validated non-stationarity. Phosphate emerged as the only universal driver, while temperature, substrate, and hydrodynamics exhibited location-dependent effects. Our findings demonstrate that interpretable ML combined with spatial cross-validation effectively diagnoses when environmental relationships transfer versus when they require location-specific calibration, providing a generalizable framework for spatial prediction in heterogeneous ocean systems.

How to cite: Dorado Guerra, D. Y., Gimeno Monforte, S., Alcaraz Cazorla, C., and Diogène Fadini, J.: Spatial Non-Stationarity in Harmful Algal Bloom Drivers for the benthic dinoflagellate Gambieridscus spp in the Balearic Islands, Revealed Through Interpretable Machine Learning and Hierarchical Modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17203, https://doi.org/10.5194/egusphere-egu26-17203, 2026.

Marine research and operational services require accurate sea-surface current (SSC) data. Because direct observations are sparse and spatially incomplete, spatially consistent SSC fields are most commonly obtained from numerical ocean models. These models are physically comprehensive but computationally expensive, as they integrate the full three-dimensional ocean state even when only surface currents are required. This makes their routine use inefficient for applications that primarily need surface information.

Here we develop a convolutional U-shaped neural network to partially emulate daily-mean SSC variability in the Baltic Sea. The emulator is formulated as a one-day state-update operator that predicts next-day zonal and meridional SSC components from the previous-day SSC field and prescribed near-surface atmospheric forcing. The network is trained on nine years (2015–2023) of SSC fields from the Copernicus Marine Service Baltic Sea Physical Reanalysis, together with near-surface atmospheric forcing from the fifth-generation European Centre for Medium-Range Weather Forecasts (ECMWF) Reanalysis (ERA5), interpolated to the SSC grid. Both datasets are used at 1-nautical-mile spatial resolution and daily temporal resolution. Predictive performance is evaluated on an independent test year (2024).

Occlusion sensitivity-based input selection indicates that SSC persistence (SSC on day t) and near-surface wind forcing (wind on day t+1) capture the dominant controls on day-to-day SSC evolution (SSC on day t+1), allowing the input space to be reduced to four channels by excluding additional atmospheric variables. One-day emulation achieves high skill across most of the basin, with spatially averaged vector errors of 2.4–2.6 cm s⁻¹ and correlations exceeding 0.9. When deployed in an autoregressive mode, errors increase smoothly with lead time and correlations decrease to approximately 0.65 by day 21. However, large parts of the coastal and interior Baltic Sea retain correlations above 0.9 and vector errors below 10 cm s⁻¹ even at multi-week lead times, indicating stable and spatially localized error growth.

To interpret the learned dynamics, we apply two explainability analyses: layer-wise relevance propagation (LRP) and diagonal Jacobian elasticity (DJE). LRP identifies which input information supports the formation of the forecast by propagating the predicted output backward through the network and assigning each input grid point a relevance score that reflects its contribution to the forward computation, independent of local sensitivity or numerical scaling. DJE, which we term here, characterizes how the forecast responds to small input perturbations by using the model’s Jacobian—the set of partial derivatives linking outputs to inputs—to quantify local, co-located sensitivities. The results show that SSC persistence provides the primary structural support for predictions in energetic boundary and strait regions, while wind forcing dominates the local sensitivity of predicted SSC over the interior basin and offshore waters. These diagnostics indicate that the network learns physically plausible state-memory and wind-driven adjustment patterns rather than relying on diffuse, non-local correlations.

How to cite: Barzandeh, A., Manß, C., Stahl, F., Maljutenko, I., Rikka, S., and Raudsepp, U.: Partial Emulation of Simulated Sea-Surface Currents in the Baltic Sea: An Assessment of Explainability and Potential Forecast Skill, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17936, https://doi.org/10.5194/egusphere-egu26-17936, 2026.

Marine biogeochemical models typically contain tens to hundreds of parameters and are notoriously challenging to tune to sparse and noisy observations, in particular for specific regional conditions. While ensemble-based methods can automate this process and are also used for data assimilation, they do not scale well to large numbers of unknown parameters. Gradient-based methods, on the other hand, scale well with high dimensionalities but require adjoint models. However, state-of-the-art differentiable programming frameworks such as PyTorch eliminate the need for manual adjoint implementations through automatic differentiation, that is, by using the chain rule to automatically compute analytic derivatives.

We introduce a fully differentiable framework for tracer transport and marine biogeochemical (BGC) processes in PyTorch. We implement advection and diffusion operators based on popular models written in Fortran, e.g. the General Ocean Turbulence Model (GOTM) for water columns. As GOTM's vertical mixing formulation requires implicit time stepping, we provide efficient differentiable solvers for batched tridiagonal systems with custom backward methods derived by implicit differentiation. Furthermore, our framework includes a PyTorch base class for differentiable BGC models with an interface similar to the Framework for Aquatic Biogeochemical Models (FABM). We provide several examples, including a re-implementation of the popular ecosystem model ECOSMO. As our operators are implemented in PyTorch, they can easily be combined with established neural network layers and optimizers.

We demonstrate our framework by performing model tuning and data assimilation in BGC models using 4DVar on sparse and noisy observations. We investigate the scaling behaviour of our tridiagonal solver for various batch and system sizes with both GPU and CPU computation. Our contribution has the potential to enhance data assimilation, speed up parameter tuning workflows and improve the accuracy of biogeochemical modelling.

How to cite: Nimtz, P. R., Demir, K. T., Zinchenko, V., Frion, A., and Greenberg, D. S.: Fully differentiable transport operators enable gradient-based parameter tuning and data assimilation of marine biogeochemical models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19761, https://doi.org/10.5194/egusphere-egu26-19761, 2026.

Anthropogenic climate change induces sea level changes (SLC) that must be accurately estimated to improve understanding of both past and future changes, facilitate timely adaptation and mitigate coastal risk. The rate and acceleration of global mean sea level and the associated uncertainty has been thoroughly assessed for the period since 1900. For the period since 1993, regional assessments have been produced, leveraging tide gauge records and satellite altimetry, allowing nations to understand and adapt more appropriately to local sea-level changes. However, improved regional timeseries are needed to robustly detect potential accelerations in local SLC.

This study proposes a novel data-driven approach for reconstructing regional SLC from tide gauges. We use a Reduced-Space Ensemble Kalman Smoother associated with the statistical Analog Prediction. This method, named RedAnDA, has been previously applied to reconstruct past temperature and salinity fields in the tropical Pacific, with good results. In this work, RedAnDA derives empirical orthogonal functions from satellite altimetry to extrapolate spatial features of the variability, as well as Analogs to predict monthly SLC associated with interannual-to-decadal variability. The uncertainty is quantified from the spread within the ensemble and takes various components into consideration, such as non-linearity in the dynamics or sampling issues. Tide gauge and altimetry input datasets are pre-processed (for instance for vertical land motion) using state-of-the-art methods. 

The RedAnDA performance is assessed by comparing the reconstruction to altimetry and existing tide gauge reconstructions, to evaluate our results over the recent period. In comparison to other reconstruction methods, RedAnDA can assess monthly changes associated with interannual variability over the 20th century, relying only on observational-based information. We further test the method by doing reconstructions which only assimilate 50-75% of the tide gauges, using the remaining ones for validation. These different tests show that RedAnDA can provide important additional regional information on SLC in the 20th century, including new estimates of the acceleration in regional SLC. 

How to cite: Oulhen, E., Slangen, A. B. A., and Palmer, M. D.: Reconstructing regional 20th century sea level changes from tide gauges using the Analog Method, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20090, https://doi.org/10.5194/egusphere-egu26-20090, 2026.

Data-driven models promise higher-fidelity Earth system forecasts at a fraction of the computational cost of numerical models, enabling the use of large ensembles for more robust statistics. Consequently, the number of purely data-driven atmospheric models has grown explosively in recent years. However, the sheer diversity of architectures and the absence of a clear "winner" pose a significant design challenge for those seeking to replicate these successes in oceanography.

GraphCast was one of the first models in this arena and remains state-of-the-art. Based on graph neural networks, it lacks specific atmospheric inductive biases, such as fixed physical dimensions, conservation laws, or explicit evolution equations. Its only relational inductive bias is the physical proximity between interacting elements. When provided with an appropriate graph, this principle should hold equally well for the ocean, making GraphCast an ideal candidate for cross-domain application.

To test this hypothesis, we introduce ARCO-OCEAN: a new Analysis-Ready, Cloud-Optimized curated dataset designed for training such models. ARCO-OCEAN contains global reanalyses and hindcasts of multiple Earth system components, including ocean physical state, waves, sea ice, and atmospheric/hydrological forcing. Widely available through the AWS Open Data program, this dataset decouples AI/ML-related methodological development from domain-specific scientific knowledge (e.g., variable selection, spatial and temporal resolution) and data engineering (e.g., choice of format, chunking), relieving data scientists of the heavy burden of data preparation.

We detail the specific design choices of ARCO-OCEAN intended for coupled atmosphere-ocean modeling at subseasonal-to-seasonal timescales. Finally, by equipping GraphCast with land-masking capabilities and a global ocean mesh graph, we present preliminary results on its training performance within the ocean domain.

How to cite: Campanella, S., Salon, S., Querin, S., and Bortolussi, L.: Can GraphCast learn skillful subseasonal-to-seasonal global ocean forecasting using the ARCO-OCEAN testbed?, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20245, https://doi.org/10.5194/egusphere-egu26-20245, 2026.

Data-driven approaches, particularly deep learning, are rapidly transforming earth system modeling. OceanBench has established a standardized benchmark for global short-range data-driven ocean forecasting, providing operationally consistent datasets and evaluation protocols that support reproducible development and assessment of ML-based ocean forecasting systems.

Building on this foundation, we introduce new extensions to OceanBench that broaden its accessibility and applicability under realistic computational constraints. These include the integration of coarser-resolution (~1°) global models, enabling computationally efficient experimentation, regional evaluation capabilities, and the inclusion of new candidate models spanning both physics-based and machine-learning approaches. By supporting multiple resolutions and modeling paradigms, the extended OceanBench framework enables more flexible and application-relevant assessment of ocean forecasts, accelerating research and operational adoption of data-driven and hybrid ocean modeling systems.

How to cite: El Aouni, A., Gaudel, Q., Aissa-Abdi, Z., Bricaud, C., and Ruggiero, G.: OceanBench: A Benchmark for Data-Driven Global Ocean Forecasting systems, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20905, https://doi.org/10.5194/egusphere-egu26-20905, 2026.

Sea surface temperature (SST) is a fundamental variable influencing  the variability of the ocean and atmosphere on synoptic, decadal and climate timescales. Satellites play a major role in its estimation and particularly measurements from infrared (IR) radiometers, which provide high-resolution observations of SST. However, IR retrievals are contaminated by  the presence of clouds that are therefore removed resulting in gaps in the retrieved fields. Because many applications rely on a gap-free SST field, including  marine heatwaves studies and ocean reanalysis, a high-quality reconstruction of missing SST is required.

Traditional techniques to address this issue include Empirical orthogonal functions (EOFs) and Optimal interpolation (OI). However, those techniques often result in over-smoothing, even where observations are present. Recently, deep learning (DL) techniques have been employed, leveraging their capacity of capturing non-linearities to better reconstruct data with gaps. 

Recently Asperti et al. (2025) developed DL models based on U-Net and transformer architectures with several configurations implemented in the Italian Seas to reconstruct SST using Level 3 products. The results show that DL based models are promising to reconstruct SST fields even close to complex coastlines. In this work, we extend the methodology introduced in their study to the entire Mediterranean Sea, starting from the best performing configuration, based on U-Net architecture. Here the method used to train the neural network is to add an additional cloud mask from a randomly picked day, to the input SST, in order to have a ground truth to use for the loss computation. The extended Mediterranean Sea model skill is comparable to the model in Asperti et al. (2025) on the overlapping regions. Since the modulation of observed fields is negligible by U-Net, our model shows better skill compared to the Level 4 products based on OI. Finally, we will also present results from an independent validation against in-situ drifter SST observations that are mainly located in the western Mediterranean basin. Level 3 SST products show discrepancies relative to drifters in terms of both overall error and mean bias, which are preserved by the U-Net in cloud-free regions. In reconstructed regions, only a modest degradation in skill relative to drifter observations is observed, indicating that the reconstruction introduces limited additional error.

Deep Learning for Sea Surface Temperature Reconstruction under Cloud Occlusion by Asperti et al. 2025. Applied Ocean Research. In review. https://arxiv.org/abs/2412.03413

How to cite: Tartufoli, B., Aydogdu, A., Pinardi, N., Asperti, A., and Oddo, P.: Sea surface temperature reconstruction in the Mediterranean Sea using deep learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21240, https://doi.org/10.5194/egusphere-egu26-21240, 2026.

Short-term forecasting of Arctic essential climate variables (ECVs) requires methods that can exploit the growing diversity of forthcoming satellite observations while remaining robust to sparse and heterogeneous sampling. This study targets sea-ice concentration (SIC) and sea-ice thickness (SIT) forecasting using observations from the new Copernicus Sentinel Expansion missions: ROSE-L providing high-resolution (≈500 m) SIC, CIMR delivering intermediate-resolution (≈5 km) SIC and thin sea ice thickness, and CRISTAL altimeter supplying SIT and sea surface height at similar scales. We propose an online multiresolution neural forecasting framework designed to ingest irregular satellite swaths across resolutions and sensor types, and to produce observation-conditioned nowcasts compatible with operational constraints. The model combines multiscale forecast architectures to explicitly handle intermittency, scale disparities, and sensor-dependent information content. Beyond its operational relevance, the framework is used as a research tool to investigate predictability across scales, enabling a systematic analysis of how submesoscale ice processes impact short-term forecast skill at coarser resolutions. Forecast performance is assessed using resolution-aware metrics, revealing scale-dependent gains in ice-edge sharpness, thin-ice variability, and short-lead SIT evolution compared to baseline methods. By explicitly combining ROSE-L, CIMR, and CRISTAL observations within a unified multiresolution framework, this work enables a direct assessment of how high-resolution sea-ice variability propagates across scales and impacts short-term predictability in operational Arctic ECV forecasts.

How to cite: Beauchamp, M., de Nailly, P., le Guillouzic, M., Singha, S., Rasmussen, T., Sievers, I., and Fablet, R.: Multiscale data-driven forecasting of Sea Ice Essential Climate Variables, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21323, https://doi.org/10.5194/egusphere-egu26-21323, 2026.

End-to-end neural schemes have become state-of-the-art approaches for the reconstruction of ocean variables from irregularly-sampled observations, especially for sea surface dynamics (e.g., SST, SLA, ocean colour…).While most studies rely on the direct application of state-of-the-art architectures developed in imaging science, especially Unets, a class of approaches explicitly leverage state-space formulation and generalize in a neural fashion established data assimilation schemes such as 4DVar algorithms and EnKF schemes. Most of these approaches have been demonstrated for toy examples or intermediate-complexity case-studies. Here, we focus on 4DVarNet architectures which generalizes weak-constraint 4DVar solvers. Drawing inspirations from unrolled neural architectures used in computational imaging, especially in diffusion and flow matching models, we extend the original 4DVarNet architectures to a broader class of unrolled architectures which differ according to the specific parameterization of the considered iterative residual update. Leveraging diffusion-based Unet schemes with time embedding blocks, the resulting 4DVarNet schemes range from 1-million-parameter configurations to 50-million-parameter ones. Through an application to satellite altimetry and Sea Level Anomaly mapping, we assess the performance of the proposed architectures. Our contributions are three-fold: (i) we report state-of-the-art performance of considered neural global SLA mapping schemes compared to the state-of-the-art (eg, MIOST, NeuROST); (ii) unrolled architectures with just very few iterations, typically 5 to 10, reach the best mapping performance, (iii) the best unrolled architecture explicitly benefits from the knowledge conveyed by the underlying variational representation of the mapping problem. We discuss how these results could pave the way towards at-scale demonstrations of end-to-end neural DA schemes for the reconstruction of global ocean states from partial observations, including uncertainty quantification issues.

How to cite: Fablet, R., Zhu, D., de Nauily, P., Botvynko, D., and le Sommer, J.: Scaling End-to-end neural DA up to real-world problems: a case study for global-scale SLAmapping and 4DVarNets, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21696, https://doi.org/10.5194/egusphere-egu26-21696, 2026.

The global ocean carbon sink is a critical component of the Earth’s climate but current models are limited in their predictive capability because of the high computational cost that a biogeochemical model, required to simulate air-sea CO₂ fluxes, entails. In this study, we investigate the ability of deep learning (DL) methods to reconstruct monthly air-sea CO₂ fluxes using only the physical output of an ocean circulation model, thereby exploring a data-driven alternative to a costly biogeochemical model. We used a collection of global simulations from the ocean biogeochemistry model NEMO-MOPS at a horizontal resolution of 0.25°, which differ in their atmospheric forcing components. The simulations span 61 years (1958-2018), providing a long, high-resolution dataset that captures substantial interannual to decadal variability. Our objectives are threefold: (1) to assess how accurately DL models can reconstruct CO₂ fluxes from physical variables alone, (2) to evaluate the generalization of these models across unseen years and forcing regimes, and (3) to identify the relative importance of physical drivers and their temporal lags in predicting air-sea CO₂ exchange. To this end, we train a point-wise Long Short-Term Memory (LSTM) network augmented with a temporal attention mechanism, which enables dynamically weight information from different time steps, to predict present-month CO₂ fluxes. To this end, we use eight physical predictors from the current month and the preceding five months. Standard regression metrics indicate an overall accurate reconstruction even though extreme CO₂ outgassing events are often underestimated. Seasonal and interannual variations are mostly well reconstructed across different ocean regimes. Spatial patterns are also well reconstructed, even though the DL model is trained only with local features (not including latitude and longitude information). This is a promising result in terms of generalizing to other physical settings, which we aim to test in future experiments. We finally interpret the learned relationships, by computing the Shapley values to quantify the contribution of each physical driver across time lags. Overall, our work highlights the potential of combining DL based techniques and explainable AI as a scalable and transparent complement to traditional Earth system modeling for studying ocean carbon cycle dynamics.

How to cite: Mohanty, S., Patara, L., Rath, W., Kazempour, D., and Kröger, P.: Reconstructing surface ocean carbon flux from physical parameters using deep learning methods, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-22647, https://doi.org/10.5194/egusphere-egu26-22647, 2026.

Land degradation, deforestation and climate change have exacerbated droughts in Ethiopia, severely threatening its agriculture dependent economy. This led to large-scale restoration initiatives such as Sustainable Land Management Program (SLMP), Reduction of Emission from Deforestation and Forest Degradation Plus (REDD⁺) and the Green Legacy Initiative (GLI). GLI reported planting 32 billion trees since 2019, yet evidence remains limited. Here, we developed a deep learning framework robust to geolocation errors to monitor nationwide canopy height dynamic at 10m resolution to conduct intervention specific outcome assessments. We found a net gain of 23,537 km² in tree cover with trees above 8m height over the period 2019-2024. The large gain in young trees offsetting loss of tall trees is attributed to recent tree planting initiatives such as the GLI, REDD⁺, SLMP and expansion of commercial plantation by the small landholder farmers. SLMP and REDD⁺ interventions yielded the largest mean canopy height gains albeit in smaller areas.  Our results demonstrate measurable evidence that large-scale restoration interventions in Ethiopia are reversing the long-standing deforestation trends in the country.

How to cite: Workie, T., Brandt, M., Ciais, P., Gaber, M., Pellikka, P., and Gonsamo, A.: Satellite observations reveal large-scale restoration interventions reversing deforestation in Ethiopia, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-2837, https://doi.org/10.5194/egusphere-egu26-2837, 2026.

Tree-based ecosystems play a crucial role in climate change mitigation by sequestering atmospheric CO₂. However, tree resource monitoring practices are often inconsistent, biased, and fail to account for trees outside forests, limiting the effectiveness of carbon credit systems and restoration strategies. This talk presents recent advances in large-scale tree ecosystem monitoring enabled by machine learning and remote sensing [1]. We demonstrate methods for estimating tree biomass and carbon stocks at continental and national scales based on high-resolution satellite imagery and LiDAR data using deep neural networks. Case studies include mapping 9.9 billion trees across African drylands [5], nationwide tree mapping and carbon stock estimation in Rwanda supporting efforts to achieve net-zero emissions [3], and assessing the overlooked contribution of trees outside forests in Europe [2]. We present an application of 3D point cloud deep neural networks to predicting vegetation biomass from airborne LiDAR [4]. Furthermore, we introduce an approach for predicting vertical vegetation structure from Sentinel-2 and spaceborne LiDAR (GEDI) data at 10 meter resolution, potentially providing insights into biodiversity, biomass, and human interventions [6]. These developments pave the way for accurate, high-resolution, and unbiased monitoring of tree biomass, supporting carbon cycle modelling and informing carbon market policies.

[1] Brandt et al. High-resolution sensors and deep learning models for tree resource monitoring. Nature Reviews Electrical Engineering, 2025

[2] Liu et al. The overlooked contribution of trees outside forests to tree cover and woody biomass across Europe. Science Advances, 2023

[3] Mugabowindekwe et al. Trees on smallholder farms and forest restoration are critical for Rwanda to achieve net zero emissions. Communications Earth & Environment , 2024

[4] Oehmcke et al. Deep point cloud regression for above-ground forest biomass estimation from airborne LiDAR. Remote Sensing of Environment, 2024

[5] Tucker et al. Towards continental scale monitoring of carbon stocks of individual trees in African dryland. Nature, 2023

[6] Zhang et al. A Vertical Vegetation Structure Model of Europe. Advances in Representation Learning for Earth Observation at EURIPS, 2025

How to cite: Igel, C.: Machine Learning and Remote Sensing for Monitoring Tree Biomass, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7566, https://doi.org/10.5194/egusphere-egu26-7566, 2026.

Land use change is a significant source of anthropogenic carbon emissions, making it a critical yet often underrepresented component in climate projections. As next-generation Earth System Models move toward kilometer-scale resolutions to capture fine-scale land-atmosphere interactions, existing land use projections (typically provided at ≈30 km resolution) are insufficient to represent the spatial heterogeneity these models require.

Relying on coarse datasets can result in a loss of 31–54% of spatial information, introducing substantial biases in simulated terrestrial carbon sequestration and surface fluxes. To address this, we present a deep learning framework designed to downscale coarse Land-Use Harmonization 2 (LUH2) data into high-resolution 1 km mosaics covering the historical and future period from 1850 to 2100.

Our methodology employs a U-Net architecture to integrate transient anthropogenic drivers from LUH2, high-resolution environmental conditions using Köppen-Geiger climate classifications, and high-resolution population density with a suite of high-resolution static geophysical features (elevation, 2D depth-weighted soil composition, terrain characteristics). 

A key technical advancement is our distributed inference pipeline using Gaussian-weighted patch aggregation. By normalizing overlapping predictions, this approach eliminates blockiness and edge artifacts, ensuring seamless global transitions across the 1 km mosaic. Validation against the HILDA+ dataset demonstrates high fidelity, achieving a global accuracy of 94.5% and a mean Intersection over Union (mIoU) of 0.799 for primary land use classes. These results provide a continuous boundary condition that enhances the realism of carbon, water, and energy fluxes in next-generation climate simulations and digital twin infrastructures.

How to cite: Castaño, M., Mozaffari, A., Materia, S., and Duarte, A.: Global 1 km Reconstruction of Historical and Future Land Use with Machine Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10039, https://doi.org/10.5194/egusphere-egu26-10039, 2026.

Robust carbon cycle science and effective carbon market governance depend on accurate monitoring, transparent modelling and credible representation of climate–economic feedbacks. Integrated Assessment Models (IAMs) such as RICE provide a long-standing framework for linking carbon emissions, climate dynamics and economic development and are widely used to inform mitigation pathways, carbon pricing and international climate policy. However, traditional IAMs rely on hand-calibrated parameters, simplified damage functions and fixed ethical assumptions, limiting their ability to integrate observational data, quantify uncertainty and support evidence-based carbon management. We build on recent advances in machine learning for climate policy and introduce RICE-N-JAX, a fully differentiable implementation of the multi-region RICE-N model (Zhang et al., 2025). RICE-N extends classical IAMs with multi-agent reinforcement learning to model strategic interactions and international climate negotiations. Our JAX-based reimplementation makes the entire climate–economic simulation fast and differentiable, including carbon emissions, climate response, production, trade, mitigation decisions and negotiation dynamics. Differentiability enables a new class of hybrid, data-driven climate–economic models. Our current research focuses on two key directions. First, we develop non-parametric hybrid damage functions in which the traditional analytical damage formulation is replaced by neural or spline-based surrogates trained on empirical and scenario data. This allows the damage–temperature relationship to be learned directly from data. Second, we perform inverse modelling of ethical and behavioural parameters, such as regional risk aversion, time preferences and mitigation bias, by calibrating the model against emissions, GDP and temperature trajectories from the Shared Socioeconomic Pathways (SSPs). This enables the recovery of latent normative assumptions embedded in scenario narratives and provides a data-informed basis for policy analysis. Finally, differentiability supports gradient-based calibration, uncertainty quantification, and sensitivity analysis of carbon price trajectories, mitigation pathways, and long-term climate impacts. We demonstrate a proof-of-concept end-to-end calibration of climate damage functions and show how parameter uncertainty propagates into future economic and emissions outcomes. By bridging process-based climate–economic theory with hybrid, knowledge-guided machine learning, RICE-N-JAX provides a foundation for fast and data-driven carbon-cycle modelling. The framework supports policy-relevant applications ranging from carbon pricing and climate clubs to carbon market design, illustrating how hybrid ML can strengthen the scientific basis of carbon management and climate mitigation.

References: Zhang, T., Williams, A. R., Wozny, P., Cohrs, K.-H., Ponse, K., Jiralerspong, M., Phade, S. R., Srinivasa, S., Li, L., Zhang, Y., Gupta, P., Acar, E., Rish, I., Bengio, Y., and Zheng, S.: AI for global climate cooperation: Modeling global climate negotiations, agreements, and long-term cooperation in RICE-N, Proceedings of the 42nd International Conference on Machine Learning (ICML 2025), 2025

How to cite: Ponse, K., Cohrs, K.-H., Wozny, P., Williams, A. R., Zhang, T., Acar, E., Bengio, Y., Plaat, A., Moerland, T., Gentine, P., and Camps-Valls, G.: Leveraging Differentiable Climate-Economy Models for Hybrid Modeling and Inverse Problems, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-11690, https://doi.org/10.5194/egusphere-egu26-11690, 2026.

How to cite: Cuba, M. D. and Black, H.: A Hybrid Sampling-Modelling Approach using Direct Measurement and Remote Sensing to Optimise the Cost-Uncertainty Balance in Large Scale Carbon Monitoring and Carbon Farming Projects.  , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12017, https://doi.org/10.5194/egusphere-egu26-12017, 2026.

Urban vegetation mitigates carbon and provides ecosystem services. Quantifying these benefits relies on land surface models like JSBACH, but high-resolution long-term simulations are computationally heavy and too complex for practical applications. Machine learning emulators offer a computationally efficient alternative. Here, we present daily and monthly emulators for gross primary production (GPP) and net ecosystem exchange (NEE) of CO₂ for different plant functional types (PFTs) in Helsinki: deciduous and coniferous trees, lawn, and crops represented by 50/50 weight of cereal and agricultural grass. The emulators are trained on JSBACH simulations for 1991-2015 and evaluated for 2016-2024. Predictor variables are derived from daily air temperature, precipitation, and shortwave radiation.

The emulators are based on gradient boosting models with automated hyperparameter optimization. We trained separate models for each target variable and PFT. To estimate the total value of a target variable for each 50 m × 50 m pixel in Helsinki, we combined PFT specific predictions weighted by the fractional coverage of each vegetation type within the pixel.

Emulator performance was high across all plant functional types and for both carbon fluxes. The monthly emulator outperformed the daily emulator consistently, as demonstrated by higher explained variance and lower errors for both GPP and NEE. Although the monthly emulator smoothed out short-term variability, it still reproduced total annual GPP and NEE with a level of accuracy almost matching that of the daily emulator. 

The two machine learning emulators developed in this study achieved high levels of accuracy, enabling faster simulations than the original land surface model. The daily emulator provided more detailed information on how vegetation responds to different meteorological conditions. In contrast, the monthly emulator was better suited to urban planning, offering fast and reliable information on the carbon sequestration of various PFTs over extended periods, while reducing simulation time by over 95% compared to the daily emulator.

How to cite: Vasenkari, V., Backman, L., Leskinen, J., Lindqvist, H., Pihlatie, M., Järvi, L., and Kulmala, L.: A Machine-Learning Emulator of the land surface model JSBACH for High-Resolution Urban Biogenic CO2 Fluxes, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12145, https://doi.org/10.5194/egusphere-egu26-12145, 2026.

Terrestrial ecosystems have cumulatively sequestered 24% of anthropogenic carbon dioxide (CO₂) emissions since 1850 and are critical for mitigating future climate change. However, current Earth System Models (ESMs) remain highly uncertain in projecting future trajectories of this carbon sink capacity, hampering our predictive understanding of climate mitigation potential and impeding effective climate and carbon management policies. This study develops a novel framework that harnesses deep-learning (DL) to constrain uncertainties of ESM-projected Gross Primary Production (GPP) and Net Ecosystem Production (NEP) through 2100. Specifically, we apply DL to characterize the “offset” between ESM-simulated output (using CMIP6 models) and best-available observational products (top-down, bottom-up). This offset is treated as unresolved processes by current ESMs that could be effectively resolved by DL, which, once trained during historical periods, can be applied to adjust CMIP6 projections of the future. We find that DL significantly reduces the inter-model spread of GPP by ~56% and NEP by ~66% across the CMIP6 ESM ensemble . Under the medium emission scenario (SSP 245), the ensemble mean for NEP in 2100 is much weaker, 2.42 ± 1.16 PgC yr⁻¹ compared to 5.52 ± 3.45 PgC yr⁻¹ in the raw CMIP6 projections, suggesting a current overestimation of future carbon sequestration capability. Interestingly, DL revealed a slower trajectory of NEP growth compared to the raw CMIP6 projection. Beyond curbing the uncertainties of CMIP6 projections, DL also captures key environmental sensitivities of carbon cycle processes such as CO₂ fertilization and sensitivity to warming. These findings demonstrate the power of DL in effectively curbing ESMs projection uncertainties and suggest that relying solely on natural terrestrial carbon sinks for climate mitigation is unlikely to slow down climate warming.

How to cite: Wu, Z., Zhu, Q., Lehner, F., Sun, W., Terrer, C., Cambron, T. W., Norby, R. J., Smith, W. K., Wen, J., Luo, Y., Tao, F., Wei, N., Albertson, J. D., Fu, Y., Ma, P., Luo, X., Fan, J., Gomes, C. P., and Sun, Y.: Artificial Intelligence Reveals a Weaker CMIP6 Terrestrial Carbon Sink with Reduced Uncertainty, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-12172, https://doi.org/10.5194/egusphere-egu26-12172, 2026.

High-resolution Greenhouse Gas (GHG) estimation is critical for verifying emissions inventories and informing climate policy. Current state-of-the-art estimates rely on "bottom-up" inventories, which are expensive to maintain, subject to reporting lags, and sensitive to inconsistent data supply chains. Conversely, "top-down" global reanalysis products, such as CarbonTracker, offer high quality but lack the spatial resolution required for actionable local policy, and high accuracy estimation of individual large polluters.

To bridge this gap, we present a deepC, a method that leverages high-resolution simulation data to inform a generative prior while assimilating diverse ground-truth observations. We learn a patch-based diffusion prior from multi-resolution simulations of regional and global carbon transport to model the joint distribution of winds, surface fluxes, column concentrations, and emissions. We then apply a Bayesian posterior formulation to guide the generation process using sparse observations from six satellite missions, ground stations, and coarse global reanalysis. To ensure consistency over large regions, we employ a novel spatio-temporal Markov blanket scheme during posterior sampling, producing carbon emissions estimates at 1km resolution.

We demonstrate the model's efficacy in CONUS and Western Europe, achieving stable emissions trajectories with low error relative to high-quality ground sensor and TCCON data. Early experiments suggest that conditioning the prior on embeddings from remote sensing foundation models significantly improves generalization to unseen domains. Furthermore, the model is robust to distribution shifts -- maintaining coherence under simulated future background CO₂​ levels. Finally, our approach yields well-calibrated uncertainty quantification at high inference speeds with ensemble generation, highlighting its potential for rapid, transparent emissions stocktaking, and lag-free policymaking.

How to cite: Shah, A., Lehman, N., Hess, P., Cohen, R. C., and Chuang, J.: deepC: High-Resolution Carbon Emissions Monitoring via Spatio-Temporal Generative Data Assimilation, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13689, https://doi.org/10.5194/egusphere-egu26-13689, 2026.

Robust carbon monitoring is fundamental to the credibility of climate mitigation strategies, including carbon markets, nature-based solutions, and ecosystem restoration initiatives. Soil organic carbon (SOC), as a major and dynamic component of the carbon cycle, is traditionally quantified through soil sampling and laboratory analyses. Although accurate at local scales, these methods are costly, time-consuming, and spatially sparse, limiting their suitability for large-scale monitoring, underscoring the need for scalable and robust alternatives.

Recent advances in machine learning (ML), and particularly deep learning (DL), offer substantial potential to integrate heterogeneous data streams and reinforce the scientific basis of carbon accounting. However, the application of DL to soil carbon studies remains limited, with most existing work confined to small spatial domains and relatively modest datasets. This limitation reflects the intrinsic complexity of environmental systems, the scarcity of high-quality reference observations, and persistent challenges in multimodal data integration and model interpretability.

Using the pan-European Land Use/Cover Area Frame Survey (LUCAS) soil dataset, this study presents a multimodal deep learning framework for large-scale prediction of SOC stocks. In addition to SOC, the framework estimates texture-related proxies and ancillary soil attributes relevant to carbon stock assessment. The approach integrates a comprehensive suite of data sources, including multispectral Sentinel-2 imagery, climate time series variables, and land-cover information, to jointly exploit spectral and spatio-temporal dependencies.

The proposed architecture integrates modality-specific components tailored to each data type, enabling a coherent spatio-temporal representation of SOC dynamics. Convolutional neural networks (CNNs) are used to extract spatial patterns and vegetation–soil spectral signatures from multispectral imagery, while recurrent architectures, including long short-term memory (LSTM) networks, encode seasonal to interannual variability driven by climatic conditions. Multiple deep learning encoders are systematically compared, ranging from conventional CNN–LSTM architectures to state-of-the-art transformer and vision transformer models, in order to assess their ability to capture long-range dependencies, cross-modal interactions, and complex non-linear relationships underlying SOC distribution.

A comparative analysis further benchmarks the proposed deep learning framework against widely used machine learning methods in soil science, including Random Forest (RF), Extreme Gradient Boosting (XGB), and Multiple Linear Regression (MLR). Model performance is assessed not only in terms of predictive accuracy, but also with respect to implementation complexity and interpretability, highlighting practical trade-offs for operational deployment.

By integrating heterogeneous data sources, this work demonstrates how artificial intelligence can bridge the gap between point-based field measurements and policy-relevant carbon assessments, while supporting state-of-the-art monitoring, reporting, and verification (MRV) frameworks. This analysis contributes to ongoing efforts to develop transparent, scalable, and evidence-based carbon monitoring tools, while explicitly highlighting persistent challenges related to data bias, spatial transferability, and model interpretability.

How to cite: Mercier, M., Marinoni, A., and Selvakumaran, S.: Multimodal Machine and Deep Learning Frameworks for Soil Organic Carbon Monitoring , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13805, https://doi.org/10.5194/egusphere-egu26-13805, 2026.

Quantifying carbon exchanges in natural grasslands is crucial for improving management practices, estimating carbon budgets, and supporting climate mitigation policies. However, direct measurements of net ecosystem CO₂ exchange (NEE) using flux towers are spatially limited, particularly in heterogeneous biomes such as the Brazilian Pampa. This study presents a machine learning framework to upscale carbon exchange observations based on flux towers in natural grasslands used for extensive cattle production in southern Brazil. Continuous CO₂ flux measurements were obtained from multiple flux towers installed across four ecological regions representative of the Brazilian Pampa biome, encompassing different combinations of soil types, vegetation structure, climatic conditions, and grassland management. These long-term observations capture pronounced seasonal and interannual variability in NEE, driven primarily by climate variability and grazing management. Artificial neural networks (ANNs) were trained using eddy covariance flux data, meteorological variables (solar radiation, precipitation, air temperature, and humidity) derived from reanalysis products, and vegetation indicators obtained from satellite remote sensing. The trained models were applied to estimate daily NEE in other regions of the Pampa with different edaphoclimatic and vegetation characteristics where flux towers were installed. Model performance was evaluated using independent subsets of eddy covariance observations, with accuracy assessed using standard statistical metrics for this type of model. The results demonstrate that the machine learning approach successfully reproduces observed seasonal patterns and interannual variability of carbon exchanges, enabling spatially explicit estimation of carbon uptake and emissions in natural grasslands. This framework provides a scalable tool for regional carbon accounting in natural grasslands and for deriving regional emission and uptake factors. The approach contributes to improving monitoring, reporting, and verification (MRV) of nature-based climate solutions and supports policies aimed at low-carbon livestock production and conservation of the Pampa biome.

How to cite: Mergen, A., Sehnem, J., Pinheiro, M., Roberti, D., and Jacques, R.: Integrating eddy covariance and machine learning for the spatial estimation ofcarbon exchanges in natural grasslands of the Pampa biome, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13867, https://doi.org/10.5194/egusphere-egu26-13867, 2026.

Timely detection of climate-driven anomalies in terrestrial CO2 exchange is limited by the latency of current bottom-up and top-down flux products. Dynamic global vegetation model (DGVM) ensembles underpin the annual Global Carbon Budget, yet their reliance on forcing datasets updated on annual cycles delays the assessment of emerging extremes. Here we develop a member-wise machine-learning emulation system that reproduces monthly net biome production (NBP) from DGVM ensembles using near-real-time meteorological reanalysis and atmospheric CO2. The emulators learn each DGVM’s spatiotemporal response on a 0.5° grid, including memory effects from antecedent conditions, and can be run as an ensemble to provide both mean behaviour and spread. In strictly forward evaluation, the emulated ensemble preserves the seasonal cycle and interannual variability of global land–atmosphere CO2 exchange and captures the timing and broad spatial structure of deseasonalized anomalies. Skill is reduced in some tropical forest regions and the strongest positive and negative excursions are damped, indicating a conservative response under extremes. By replacing offline DGVM integrations with lightweight surrogates, this framework reduces product latency to approximately one month and delivers DGVM-consistent near-real-time CO2 flux estimates that can serve as operational priors for integrated carbon-cycle monitoring.

How to cite: Ke, P., Gui, X., Sitch, S., Friedlingstein, P., Liu, Z., and Ciais, P.: Machine-learning emulation of DGVM ensembles enables low-latency terrestrial CO2 flux estimates, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14534, https://doi.org/10.5194/egusphere-egu26-14534, 2026.

Accurate estimates of the spatiotemporal distribution of atmospheric carbon dioxide (CO₂) are essential to evaluate and enforce international climate agreements as well as to infer fluxes of the greenhouse gas. However, current observations are spatially sparse, with satellite and in-situ measurements providing only partial coverage of the Earth’s surface and atmosphere. Atmospheric transport models are often used to infer CO₂ concentrations across unobserved regions by simulating how gases move and mix in the atmosphere. While physically grounded, these models are computationally intensive and notoriously difficult to calibrate with observational data, due to the complexity of atmospheric dynamics and the sparsity of available measurements.

This study investigates the use of generative machine learning for inpainting of CO₂. More specifically, we apply flow matching, an approach that generates samples from an unknown target distribution by iteratively transforming samples from a simple known noise distribution with a deep neural network. In a first step, we train a flow matching model on assimilation data from CarbonTracker (CT2022). This trains the model to respect the physical patterns of atmospheric CO₂ fields, turning it into an effective prior for data assimilation. In a second step, we test the trained flow matching model on conditional generation that is, reconstruction of atmospheric CO₂ from partial observations. For this, we artificially mask parts of the CT2022’s CO₂ in a way that emulates the availability of satellite measurements. In a third step, we infer global CO₂ by conditioning on the total column average CO₂ (XCO₂) measurements from NASA’s Orbiting Carbon Observatory-2 (OCO-2), comparable to other inversions from the OCO-2 v11 MIP, but using a novel approach.

Extensive evaluation against independent and held-out test-sets from in-situ and satellite measurements show physical consistency and decent agreement of the reconstructed global CO₂ fields from OCO-2 measurements. However, challenges remain: specifically, future research needs to alleviate spurious artifacts from the employed posterior conditioning method in both the artificial mask and particularly the conditioning on XCO₂ before the approach can become operational.

Our presented flow matching approach opens up new avenues of research. The prior parameterized by the flow matching model can be investigated itself. For instance, it is possible to perform feature extraction inside the latent space and hence purposefully explore counterfactual scenarios of CO₂ distributions by carefully tracing out paths in the noise distribution and analyzing the corresponding generated CO₂ samples.

How to cite: Groß, J., Benson, V., Lima, M., Winkler, A., and Reimers, C.: Reconstructing Atmospheric CO2 with Flow Matching Models, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15228, https://doi.org/10.5194/egusphere-egu26-15228, 2026.

The precise Measurement, Reporting, and Verification (MRV) of carbon stocks in small-scale afforestation and restoration forests can be served as the  foundation for subsequent carbon sink monitoring and benefit assessment.  Satellite remote sensing method, on the other hand, often  faces insufficient spatial resolution comparing to Unmanned Aerial Vehicle (UAV) imagery. UAV can capture fine details, but often results in "scale mismatch" and systematic estimation bias due to canopy shadows, background soil noise, and spectral saturation effects while applying estimation models directly. To address this technical bottleneck, this study aims to establish an automated carbon stock estimation workflow based on UAV multispectral imagery and to optimize estimation accuracy by identifying the optimal observational resolution through multi-scale analysis. 

The research methodology synchronizes field surveys with remote sensing modeling. First, a comprehensive tree-by-tree biomass inventory was conducted in sample plots. Allometric equations were used to calculate stand biomass, which was then converted into measured carbon stock to serve as Ground Truth for model validation. Subsequently, UAV multispectral images were acquired to calculate vegetation indices (e.g., NDVI) and establish regression models between spectral features and carbon stock. Furthermore, image resampling techniques were adopted to simulate multi-level spatial resolutions ranging from 0.03 to 5 m, systematically analyzing the impact of resolution changes on the Root Mean Square Error (RMSE) and the coefficient of determination (R²). This study clarifies the interference mechanism of spatial scale on canopy spectral signals and identifies the optimal aggregation scale to mitigate background noise. Ultimately, this research provides practical prediction formulas and a Standard Operating Procedure (SOP). In the future, applying this model to UAV-acquired imagery in similar restoration forests will enable rapid, automated carbon estimation without the need for time-consuming field surveys, significantly enhancing the efficiency and economic viability of carbon asset inventories.

Keywords

Aboveground Biomass (AGB), Multispectral UAV, NDVI, Allometric Biomass Model, Scale Effect, Restoration Forest, Carbon Sink Estimation

How to cite: Lee, C.-I. and Ho, H.-C.: Optimizing Aboveground Carbon Stock Estimation in Restoration Forests: A Multi-Scale Analysis of UAV Multispectral Imagery, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16064, https://doi.org/10.5194/egusphere-egu26-16064, 2026.

How to cite: Agrawal, Y., Deshpande, S., Seth Tiwari, P., and Pande, H.: Automated Segmentation of Brick Kilns and Carbon Emission Analysis Using Deep Learning and Life Cycle Assessment , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16105, https://doi.org/10.5194/egusphere-egu26-16105, 2026.

Abstract
More accurate local estimates of biomass and other forest attributes translate
into more accurate national-level estimates, improving forest monitoring and
informing forest policy. Higher-resolution local estimates facilitate more precise
monitoring of forest growth and harvest, allowing for better forest management
planning, and can also be used for verification of forest carbon storage, such as
for tree-based carbon credit programs and afforestation projects.

We present the first time series of high-resolution national maps of tree biomass,
carbon, volume, canopy height, and basal area produced using deep learning
methods applied to 3D point cloud LiDAR data. With hexagonal tiles of a 30
m diameter, the maps enable direct observation of stock change of aboveground
biomass, carbon, and other forest attributes at high resolution, in contrast to
inventory based estimates or coarser resolution remote sensing-based products.
We verify that our approach provides reliable estimates at the national and local
scales by comparing it to additional ground truth plot data from a time series
of local inventories.

The model was trained and validated on ground-truth data from the Danish Na-
tional Forest Inventory (DNFI) by combining field measurements aligned with
more than 20,000 sample plots extracted from two complete national LiDAR
scans. Based on [1], we apply a 3D convolutional neural network (CNN) using
the SENet50 architecture. We extended the approach to perform quantile re-
gression for uncertainty quantification. Our best model achieves an R2 of 0.83
for biomass and carbon, 0.84 for volume, 0.91 for canopy height, and 0.78 for
basal area on validation data.

We find that our model outperforms other state-of-the-art methods, which are
either based on passive 2D imagery or depend on using point cloud data indi-
rectly by extracting summary statistics. By using active LiDAR, we can derive
information from beneath tree canopies, and using the full point cloud enables
the model to learn from detailed information on forest structure, which may be
a key advantage.

The high resolution and accuracy of our method offer unprecedented potential
for time series analysis. The model is sensitive to changes at the individual tree
level, allowing for the detection of individual tree removals or growth. While
large scale forest cover change is easily detected with aerial imagery, thinnings
or partial removals are more difficult to uncover with most methods; however,
our analysis of independent repeated local inventory plots shows that our model
successfully detects smaller scale thinnings and tree growth.

References
[1] Stefan Oehmcke et al. “Deep point cloud regression for above-ground forest
biomass estimation from airborne LiDAR”. In: Remote Sensing of Environ-
ment 302 (2024).

How to cite: Brownell, H., Oehmcke, S., Nord-Larsen, T., and Igel, C.: Time series of national biomass maps from deep learning applied to airborne laser scanning point cloud data, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16682, https://doi.org/10.5194/egusphere-egu26-16682, 2026.

Process-based models (PBMs) and machine learning (ML) offer complementary strengths for representing the coupled carbon-water cycle. PBMs enforce physical principles and provide interpretable diagnostics but rely on incomplete process knowledge, many priors, and very limited use of expanding Earth observations, leading to substantial inter-model spread. ML leverages observations to uncover complex patterns and reduce reliance on assumptions, but can violate physical constraints and extrapolate poorly. Hybrid modeling combines both, uniting ML’s flexibility with PBMs’ interpretability and process consistency.

We present H2CM, a hybrid carbon-water cycle model that merges process‑informed deep learning with direct learning from observations (Baghirov et al., 2025; https://doi.org/10.5194/egusphere-2025-3123). H2CM simulates carbon fluxes—gross primary productivity (GPP), autotrophic respiration, and heterotrophic respiration—and water storages (soil moisture, groundwater, snow) and fluxes (evapotranspiration, runoff). The model is informed by carbon observations—GPP, net ecosystem exchange (NEE) from satellite- and in situ–based inversions, and fAPAR—and by water-cycle observations—evapotranspiration, runoff, terrestrial water storage, and snow. H2CM runs daily at 1° spatial resolution.

H2CM outperforms both purely data-driven approaches and state-of-the-art PBMs in reproducing seasonal NEE, particularly in wet and dry tropics, and it captures the rain‑pulse respiration response in drylands that many models miss. Its estimates of global NEE interannual variability align more closely with satellite- and in situ–based inversion products than do PBM estimates. Finally, we disentangle photosynthetic versus respiratory controls and quantify how different regions (e.g., wet vs. dry tropics) contribute to global variability in land–atmosphere carbon exchange.

How to cite: Baghirov, Z., Reichstein, M., Kraft, B., Ahrens, B., Körner, M., and Jung, M.: Estimating carbon dynamics using H2CM: a hybrid global carbon-water cycle model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19700, https://doi.org/10.5194/egusphere-egu26-19700, 2026.

Carbon farming can deliver climate mitigation and improved soil health, but credible deployment requires scalable MRV that supports additionality assessment and remains operational at farm scale. We present an EO-driven pipeline that integrates heterogeneous Earth-system data with hybrid modelling (machine learning + process-based physics) to estimate crop yield trajectories, soil organic carbon (SOC) evolution, and economic viability under baseline and regenerative management. A case study illustrates how a crop system can transition toward regenerative farming, demonstrating alignment with EU carbon farming policy. Results show how integrated, data-driven approaches can support quantification of both environmental and financial outcomes, enabling credible carbon accounting and guiding targeted investment in sustainable agriculture.

Multi-sensor satellite time series provide indicators of vegetation dynamics, and management proxies relevant to practice adoption (e.g., seasonal soil cover and surface condition). SoilGrids data provide spatially detailed soil information that helps us capture how soil conditions vary across and within fields, and how sensitive each site is. Climate forcing relies on high-resolution CMCC climate projections, enabling stress-testing of productivity and SOC outcomes under plausible future conditions.

A Random Forest model learns non-linear relationships between yield, EO indicators, soil attributes, and climate predictors to generate baseline yield projections. These projections are translated into carbon input assumptions (e.g., residue returns) and coupled to a RothC-class SOC model to simulate SOC evolution under regenerative scenarios such as cover crops.

Farm-level decision metrics integrate transition costs, yield impacts, potential carbon revenues, and land value appreciation to estimate break-even time and NPV, supporting project design and investment appraisal.

How to cite: Galli, G., Zamboni, M., Ricciardelli, A., Quarta, M. L., and Folegani, M.: EO-driven carbon farming MRV: linking crop yield prediction to SOC change, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-20977, https://doi.org/10.5194/egusphere-egu26-20977, 2026.

Accurate estimation of global gross primary productivity (GPP) is fundamental for understanding terrestrial carbon cycling. Eddy covariance (EC) flux observations provide reliable site-scale GPP estimates, but the spatially sparse distribution limits their applicability at large scales. Satellite-based solar-induced chlorophyll fluorescence (SIF) has emerged as a promising proxy for large-scale GPP estimation; however, current satellite SIF observations also suffer from limited spatiotemporal coverage, and uncertainties remain in the SIF–GPP conversion. Moreover, conventional machine learning models trained solely on EC observations often exhibit limited spatial generalization due to the scarcity of spatially representative training samples.

To address these challenges, this study proposes a satellite–ground jointly constrained framework that integrates EC flux measurements and satellite SIF observations using transfer learning and multi-task learning techniques to exploit the complementary strengths of both data sources for global GPP estimation. First, for TROPOMI SIF data that has global spatial coverage but short temporal records, SIF is treated as a source domain to pre-train the model, which is then fine-tuned using long-term EC-derived GPP data as a target domain. This transfer learning-based model (SIFTML) demonstrates improved spatial generalization compared to models trained solely on SIF or EC data, effectively reducing systematic underestimation and overestimation at high and low GPP levels, respectively, while remaining insensitive to the magnitude scaling of source-domain SIF inputs.

Second, for the spatially sparse and track-like distributed OCO-2 SIF observations, a multi-task learning framework based on a mixture-of-experts architecture is developed. A physically constrained loss function derived from the SIF–GPP relationship is introduced to simultaneously achieve seamless SIF reconstruction and high-accuracy GPP estimation by jointly leveraging SIF and EC constraints. Results indicate that the multi-task model outperforms traditional single-task approaches in both GPP estimation and SIF reconstruction.

Overall, this study provides a new paradigm for long-term, high-accuracy global GPP estimation by alleviating limitations associated with the spatiotemporal coverage of ground EC and satellite SIF observations, as well as the uncertainties in SIF–GPP conversion, thereby offering improved support for global carbon cycle research.

How to cite: Guan, X., Ma, Y., Zeng, C., and Lin, L.: Satellite SIF and Ground EC Observation Jointly Constrained Estimation of Global Gross Primary Productivity, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21697, https://doi.org/10.5194/egusphere-egu26-21697, 2026.