Tipping points (TPs) are usually understood as bifurcations due to a slow parameter shift in a bistable system, where the system's state is diffusing around a fixed point in an underlying climate potential. As the shape of the climate potential changes towards the bifurcation, universal early-warning signals (EWS) due to critical slowing down appear.

How reliable is it to transfer this conceptual view to high-level risk assessments of climate change? We set out to test the anatomy and predictability of TPs in the complex system of the global ocean circulation, where changes in ice melt is a control parameter. Several findings highlight the need for more realistic methods to address the risk of TP:

1. The critical threshold of the control parameter can be fuzzy due to sensitive dependence on initial conditions and rate of non-adiabatic forcing changes.

2. In this spatially extended system, there is a high degree of multistability. This leads to a multitude of critical thresholds and abrupt changes in deterministic variability, which can interfer with EWS stemming from noise-driven fluctuations.

3. The universality of EWS in the high-dimensional case is compromised in practise, which may be mitigated by deriving system-specific observables from the properties of an edge state related to the TP.

How to cite: Lohmann, J.: Anatomy and predictability of tipping points in the high-dimensional climate system , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21955, https://doi.org/10.5194/egusphere-egu25-21955, 2025.

Nuclear Magnetic Resonance (NMR) is a crucial logging technique for the unconventional and complex reservoir evaluation. However, the echo spacing is always an issue of borehole NMR measurement, which limits the performance of NMR tools to acquire the short relaxation components.

In this abstract, we proposed a novel Q-Switch technique aiming at breaking through the limitation of dead-time of borehole NMR logging tool, and to achieve much shorter echo spacing. Instead of using resistors of larger resistance in parallel with the radio-frequency (RF) coil to reduce the active dead-time, an inductive coupling circuit was introduced to decrease the ringing-down time significantly after transmitting the RF pulses with high voltage. The Q-Switch circuit consists of inductive coupling coil, capacitors, resistors and active high-voltage MOSFETs. The ringing-down time of RF system was decreased by at least 10 times compared to the system without using proposed Q-switch scheme, leading to echo spacing lower to 0.3 ms under the condition with resonant frequency lower to 500 kHz.

Both simulations and experiments were in great agreements, validating the feasibility and efficiency of proposed Q-switch scheme, and proved to be promising in the borehole NMR applications.

How to cite: Luo, S., Li, X., Yu, H., Wang, Z., Xing, T., Long, Z., Che, C., Liao, G., and Xiao, L.: A Novel Q-Switch Technique for  Borehole NMR Measurement, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-121, https://doi.org/10.5194/egusphere-egu25-121, 2025.

Soil and groundwater contamination at some sites impacts downstream populations when contaminants migrate from groundwater to rivers. Predictive modeling is challenging since it is required to include detailed subsurface structure and groundwater flow models within the site, as well as watershed-scale models for large-scale transport. Now that climate change impacts are major concerns at many sites, it is important to have the capability to represent the water balance change and its impact on contaminant transport both at the site and watershed scale in a consistent manner. This study introduces a new simulation framework to couple a detailed 2D site/hillslope-scale groundwater model to the 3D watershed-scale model to describe contaminant transport from groundwater to river water within the catchment. Within the site, we estimate the contaminant discharges to the river from contaminant sources based on the Richards equation and advection-dispersion equation. The discharges are then applied as the boundary conditions to the watershed-scale model considering the width of the 2D site/hillslope-scale groundwater model and recharge rates for both models.

We demonstrate and validate our framework based on the tritium concentration datasets in surface water and groundwater collected at the Savannah River Site F-Area. Results show that the method can successfully reproduce the contaminant concentration time series in river water.

How to cite: Sakuma, K., Wainwright, H., Xu, Z., Lawrence, A., and Hazenberg, P.: Multiscale model coupling for watershed-scale contaminant transport modeling from point sources in Savannah River Site, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4077, https://doi.org/10.5194/egusphere-egu25-4077, 2025.

As one of crucial factor in atmospheric particulate matter, elemental components exhibit distinct distribution features within different particle size ranges. Crustal elements (such as Al, Si, Fe, Ca, Mg) are primarily concentrated in coarse particulate matter, whereas elements originating from anthropogenic pollution sources (such as heavy metal elements including Pb, Zn, Cd, As, Cr) are more frequently distributed in fine particulate matter. Furthermore, some specific elements may also exhibit peak concentrations in particular particle size, which is closely related to their sources and formation processes. In recent years, there are still some challenges and deficiencies. Further research is needed on the particle size distribution characteristics of complex pollution sources (such as industrial emissions and traffic emissions). Additionally, there is a need to enhance the understanding of the transformation mechanisms and health effects of elemental components within particulate matter. This study selected a typical industrial and mining city to investigate particle size distribution characteristics of elemental components in atmospheric particulate matter. Anderson Eight-Stage Particulate Impactor Sampler was used to collect atmospheric particulate matter in the urban area of Huangshi during winter and summer. Nine particle size range samples were obtained spanning from 0 to 0.4 µm, 0.4 to 0.7 µm, 0.7 to 1.1 µm, 1.1 to 2.1 µm, 2.1 to 3.3 µm, 3.3 to 4.7 µm, 4.7 to 5.8 µm, 5.8 to 9.0 µm, and 9.0 to 10 µm. Energy Dispersive X-Ray Fluorescence Spectrometry (ED-XRF) was employed to determine the concentrations of 17 elemental components, including S, Cl, K, Ca, Ti, V, Cr, Mn, Fe, Co, Ni, Cu, Zn, Rb, Sb, Ba, and Pb. Elements Ca, S, Fe, K, Zn, Ba, and Pb were identified as the primary pollutants during the sampling period. All the elemental concentrations exhibited distinct seasonal variations, demonstrating higher levels in winter compared to summer. Each element demonstrated distinct particle size distribution characteristics with peak concentrations for most elements occurring in the 5.8 to 9.0 µm range and peaks for the remaining elements in the 0.4 to 1.1 µm range. The highest elemental concentrations in both summer and winter were mainly distributed in the 5.8 to 9.0 µm and 0.7 to 1.1 µm size ranges. In summer, most elemental concentrations were negatively correlated with relative humidity. However, in winter, there was no significant correlation with relative humidity. Rainfall had a certain scavenging effect on elements but was also influenced by other meteorological factors. Element S had the highest enrichment factor values in both summer and winter. Element Cl was highly enriched in finer particle size fractions in both seasons. Most elements were slightly enriched across all particle size fractions. Principal component analysis further identified the main sources as soil dust and wind-blown sand, coal combustion, vehicle exhaust emissions, biomass burning, mining and construction activities, and other pollution sources. These findings contribute to the formulation of effective pollution control measures and the protection of public health.

How to cite: Liu, H., Zhang, J., Zhan, C., Liu, S., Liu, T., and Xiao, W.: Size distribution of elemental components in atmospheric particulates from a typical industrial and mining city of Central China, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4641, https://doi.org/10.5194/egusphere-egu25-4641, 2025.

How to cite: Ball, C., Chang, M., Cruise, J., de Valk, C., and Kannan, V.: Industrial High Performance Computing Scalable and FAIR Workflow Opportunities for EO Operations Processing, Operations, and Archiving, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7004, https://doi.org/10.5194/egusphere-egu25-7004, 2025.

Severe convection, including thunderstorms and related phenomena like flash flooding, hail, and strong winds, can have significant socioeconomic impacts. Nowcasting, which provides real-time, short-term predictions, is vital for issuing timely warnings to mitigate these impacts. Satellite imagery is essential for monitoring convection and offering accurate predictions of storm evolution, thereby enhancing early warning systems. Ensemble forecasting, which generates multiple potential scenarios, helps better quantify uncertainties in nowcasting. However, most ensemble forecasting methods are computationally intensive and typically do not incorporate satellite images directly. The Analog Ensemble (AnEn) method, a lower-cost ensemble approach, identifies similar past weather events based on forecast data. For a given time and location, the AnEn method identifies analogs from past model predictions that are similar to current forecast conditions. Then their associated observations are used as ensemble members. Despite its advantages, AnEn struggles with locality and is sensitive to the choice of similarity metrics. This study presents an improved AnEn system that replaces forecast archives with satellite images to identify analogs of convective conditions. The system utilizes pretrained deep learning algorithms (VGG16, Xception, and Inception-ResNet) to assess image similarity. The training dataset consists of daily convection satellite images from EUMETSAT for the period 2020-2023, and the domain covers 40°N to 20°S and -20°W to 4°E. The year 2024 is used for testing, with ERA5 reanalysis of total precipitation as the verification ground-state. For a present convection satellite image this image is encoded and compared to all past encoded images of the training period using different metrics. The most similar images to the current one are then selected and their associated ERA5 total precipitation reanalysis are considered the members or our ensemble. Preliminary results indicate an average maximum precipitation anomaly of 15 mm between the analog ensemble mean and the current reanalysis, showing that the proposed system offers promising improvements in short-term forecasting.

Key words: Convection; Ensemble Forecasting; Deep Learning; VGG; Xception; ResNet; Analog Ensemble; Morocco; Nowcasting; EUMETSAT; ERA5; Morocco; Satellite Images; Remote Sensing;

How to cite: Alaoui, B., Bounoun, C., and Bari, D.: Leveraging Pretrained Deep Learning Models to Extract Similarities for the Analog Ensemble Method Applied to Convection Satellite Imagery, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7258, https://doi.org/10.5194/egusphere-egu25-7258, 2025.

Amid global environmental degradation, understanding the spatiotemporal dynamics and trade-offs of ecosystem services (ESs) under varying land-use scenarios is critical for advancing the sustainable development of social–ecological systems. This study analyzed the Chaohu Lake Basin (CLB), focusing on four scenarios: natural development (ND), economic priority (ED), ecological protection (EP), and sustainable development (SD). Using the PLUS model and multi-objective genetic algorithm (MOGA), land-use changes for 2030 were simulated, and their effects on ESs were assessed quantitatively and qualitatively. The ND scenario led to significant declines in cropland (3.73%) and forest areas (0.18%), primarily due to construction land expansion. The EP scenario curbed construction land growth, promoted ecosystem recovery, and slightly increased cropland by 0.05%. The SD scenario achieved a balance between ecological and economic goals, maintaining relative stability in ES provision. Between 2010 and 2020, construction land expansion, mainly concentrated in central Hefei City, led to a marked decline in habitat quality (HQ) and landscape aesthetics (LA), whereas water yield (WY) and soil retention (SR) improved. K-means clustering analysis identified seven ecosystem service bundles (ESBs), revealing significant spatial heterogeneity. Bundles 4 through 7, concentrated in mountainous and water regions, offered high biodiversity maintenance and ecological regulation. In contrast, critical ES areas in the ND and ED scenarios faced significant encroachment, resulting in diminished ecological functions. The SD scenario effectively mitigated these impacts, maintaining stable ES provision and ESB distribution. This study highlights the profound effects of different land-use scenarios on ESs, offering insights into sustainable planning and ecological restoration strategies in the CLB and comparable regions.

How to cite: Jin, A.: Ecosystem Services Trade-Offs in the Chaohu Lake Basin Based on Land-Use Scenario Simulations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7785, https://doi.org/10.5194/egusphere-egu25-7785, 2025.

Seasonal ecosystems play a crucial role in environmental regulation and biodiversity by hosting complex ecological dynamics that vary with climatic conditions. The Chaschoc-Sejá wetlands are a key example of such systems in southeastern Mexico. The interaction between the lagoon system and the Usumacinta River is highly dynamic; during the rainy season, the lagoons increase in volume, reaching depths of 8 to 10 meters. However, the lagoons completely dry up during the dry season, leaving vegetation at the surface level. 

This project aims to analyze the dynamics of vegetation cover in this environment by comparing high-resolution satellite images (Planet, 3 m) and ortho-mosaics generated with a DJI Mavic 3 Multispectral drone (10 cm). By combining these datasets, we aim to improve our previous vegetation maps and obtain a more accurate and detailed assessment of the Chaschoc-Sejá Lagoon system. Understanding vegetation patterns at a larger scale during specific periods and the variations in plant life within the lagoon and along its shores is a key focus.

Data processing involved classifying vegetation cover and identifying seasonal changes using indices such as NDVI and NDWI. We also generated 3D models to estimate vegetation height. Results show that integrating both techniques significantly improves spatial resolution and temporal accuracy in monitoring these ecosystems. This study provides essential tools for managing seasonal systems and their conservation in the face of climatic and anthropogenic factors. This monitoring will aid in understanding vegetation status, identifying plant species, and contributing to managing and preserving the lagoon system.

How to cite: Nieto, J., Ramírez Serrato, N. L., Romero Herrera, A., Peralta Carreta, C., Herrera Zamarrón, G., Hernández Hernández, M. A., Hernández García, G. D. J., Olea Olea, S., Morales Casique, E., and Cortez Silva, A.: Integrating high-resolution satellite and multispectral drone Imagery for monitoring vegetation in the Chaschoc-Sejá lagoon system, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7921, https://doi.org/10.5194/egusphere-egu25-7921, 2025.

Benevento Province, located in the Campania region of Italy, may experience environmental quality impacts from neighboring developed areas such as Naples and Caserta. Previous studies have suggested that some agricultural chemicals from Naples, such as hexachlorobenzene, may be transported through the air to rural areas of Benevento. Additionally, high concentrations of polycyclic aromatic hydrocarbons (PAHs) have been detected in Naples and Caserta, making Benevento Province a potential PAH "sink." This study systematically investigated the occurrence of PAHs in soil from Benevento Province, southern Italy, and their correlations with environmental factors, soil-air exchange processes, and health risks. Over 95% of sampling sites exhibited ∑₁₆PAHs concentrations at non-polluted levels (9.50-1188 ng/g, mean = 55.0 ± 152 ng/g), and four-ring PAHs were the dominant pollutants contributing to 28.3% of ∑₁₆PAHs. The spatial distribution of PAHs presented significant heterogeneity, with hotspots concentrated near landfills. The results of Positive Matrix Factorization (PMF) model showed that the main sources of PAHs were vehicle emissions, coal/biomass combustion, and petroleum products volatilization/leakage, contributing 42.2%, 40.2%, and 17.6%, respectively. Most of PAHs correlated significantly with total organic carbon in the soil and population density, while only Benzo(b)fluoranthene (BbF) showed a significantly negative correlation with pH. The mass inventory of ∑₁₆PAHs ranged from 0.94 to 29.4 tons, averaging 2.45 tons. The synergistic effects of pollution hotspots and the persistent accumulation of PAHs in the soil suggested that the soil might act as a secondary source of PAHs. Toxicity equivalent and probabilistic risk assessments indicated that health risks from PAHs remained within acceptable limits.

How to cite: Qu, C. and Pu, C.: Investigation of polycyclic aromatic hydrocarbons in the soils of Benevento Province, Italy, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10723, https://doi.org/10.5194/egusphere-egu25-10723, 2025.

The adoption of the FAIR data principles has revolutionised data management in Earth System Sciences (ESS), yet challenges persist in achieving true machine-actionability and comprehensive implementation. 

DeSci Labs introduces two innovative tools—Decentralised Persistent Identifiers (DPIDs) and the CODEX protocol—to address the barriers to FAIR data practice implementation in general; whilst fostering widespread uptake of FAIR principles in the ESS community in particular through involvement in the FAIR2Adapt consortium.

DPIDs are globally unique persistent identifiers based on, and linked directly to, the content of the files it refers to. Each version of every file, regardless of type, is automatically assigned a cryptographic fingerprint, ensuring deterministic resolution and transparent versioning. The DPID has been specifically designed to support FAIR digital research objects. The flexibility to alias DPIDs with existing systems (e.g., DOIs) and programmatic publishing capabilities via NodesLib enhances interoperability while preserving data sovereignty.

The CODEX protocol, an open scholarly infrastructure, further complements this by enabling the storage and retrieval of FAIR digital research objects via a decentralised peer-to-peer network (IPFS). This architecture allows multiple copies of the same content to be stored by different network participants using the same PID. By empowering researchers to collaborate within an open-state repository, the protocol minimises reliance on centralised actors, ensuring long-term accessibility, data integrity, and transparency. Its modular design facilitates diverse gateway applications, maximising participation and reducing barriers to entry.

These tools address core challenges in FAIR adoption by providing robust, scalable, and interoperable solutions tailored to the ESS community. By integrating DPIDs and CODEX into data workflows, researchers can enhance data reusability, improve the provenance of research outputs, and safeguard the collective scientific record. This presentation explores how these technologies can catalyse the next decade of FAIR data practices in ESS, fostering trust, reproducibility, and innovation.

How to cite: Raijmakers, L., Hübinette, E., DeSota, E., Iman, S., and Koellinger, P.: Advancing and Supporting FAIR Principle Adoption through Innovative Social Infrastructure Tools, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13330, https://doi.org/10.5194/egusphere-egu25-13330, 2025.

In mineral exploration and geometallurgical studies, accurately segmenting mineral grains from core scanning datasets may be used to predict metal recovery. This study introduces the application of the Segment Anything Model (SAM), a cutting-edge deep learning tool, to automate the segmentation and extraction of mineral grains from Laser-Induced Breakdown Spectroscopy (LIBS) and hyperspectral core scanning datasets. SAM demonstrates high efficiency and precision in identifying mineral grains, forming the foundation for downstream analyses, including the evaluation of mineral associations, grain size distribution, and other key geometallurgical metrics. Through case studies on pegmatite deposits, this research showcases the potential of SAM to address challenges posed by mineralogically complex ore. By enabling detailed mineralogical characterisation and advancing geometallurgical methods, SAM-based grain extraction presents a transformative tool for supporting sustainable and efficient mining practices.

How to cite: Cai, Y., Manton, R., and Williams, M.: Automated Mineral Grain Extraction for Geometallurgical Studies Using Segment Anything Model (SAM) and Core Scanning Techniques, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13917, https://doi.org/10.5194/egusphere-egu25-13917, 2025.

Synthetic data has become an indispensable tool in climate science, offering extensive spatio-temporal
coverage to address data limitations in both current and future scenarios. Such synthetic data, derived
from climate simulation models, must exhibit statistical consistency with observational datasets to ensure
their utility. Among global climate simulation initiatives, the Coupled Model Intercomparison Project
Phase 6 (CMIP6) represents the latest and most comprehensive suite of General Circulation Models
(GCMs). However, the substantial High Performance Computing (HPC) resources required for these
physics-based models limit their accessibility to a broader research community. In response, genera-
tive machine learning models have emerged as a promising alternative for simulating climate data with
reduced computational demands.
This study introduces an ensemble model based on the Pix2Pix conditional Generative Adversarial
Network (cGAN) to generate high-resolution spatio-temporal maps of monthly global Sea Surface Tem-
perature (SST) with significantly lower computational cost and time. The proposed model comprises two
components: the GAN, which produces simulated SST climatology data , and the Predictor, which is
trained with the variability of the data that forecasts SST anomaly for the subsequent month using the
output data from the previous month. Both components contain the same architecture, but the training
processes are different. The predictor model can be fine-tuned with observed data for some epochs to
adopt its domain.
The ensemble model was calibrated with monthly SST observations from the COBE dataset as in-
put and output. The Empirical Orthogonal Functions (EOF) shows the model’s ability to simulate the
variabilty of the observed data. The model’s performance was evaluated using the temporal Pearson cor-
relation coefficient and mean squared error (MSE). Results demonstrate that the ensemble cGAN model
generates maps with statistical characteristics closely matching those of CMIP6 simulations and obser-
vations, achieving a mean temporal correlation coefficient around 0.5 and an MSE around 1.13 for both
cases.

How to cite: Chakraborty, D. and Mitra, A.: Simulation of Monthly Global Sea Surface Temperature Data using Ensemble GAN Model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14821, https://doi.org/10.5194/egusphere-egu25-14821, 2025.

Deep neural networks have revolutionized various fields due to their remarkable adaptability, enabling them to address related tasks through retraining and transfer learning. These capabilities make them invaluable tools for diverse applications, including climate and hydrological modeling. In an earlier work (Mishra Sharma et al., 2024), we introduced a novel neural network architecture, the Max-Average U-Net (MAUNet), which leverages Max-Average Pooling to downscale gridded precipitation data to higher spatial resolutions. The model demonstrated significant improvements in resolving finer-scale precipitation features, making it well-suited for climate data applications.

In this study, we utilized the MAUNet architecture to tackle the critical task of bias correction in satellite-based precipitation estimates. Bias correction is essential for improving the reliability of precipitation data derived from satellite missions, which often exhibit systematic discrepancies compared to ground-based measurements. Specifically, we focused on correcting biases in precipitation estimates from the Tropical Rainfall Measuring Mission (TRMM) by calibrating them against high-resolution, ground-based gridded datasets from the India Meteorological Department (IMD).

Our experimental results reveal that MAUNet effectively reduces biases in TRMM precipitation estimates, achieving significantly improved agreement with ground truth data. This success is attributed to the model’s robust feature extraction and reconstruction capabilities, which enable it to learn and correct systematic errors in satellite data. The findings also highlight the potential of advanced neural network architectures in addressing bias correction challenges.

This work underscores the utility of deep learning architectures in precipitation modeling, contributing to broader goals of improving the spatial distribution of precipitation estimates. By bridging the gap between satellite observations and ground truth, the MAUNet model offers a comprehensive solution for enhancing precipitation datasets, with significant implications for climate research, hydrological studies, and policy planning.

How to cite: Mishra Sharma, S. C. and Mitra, A.: Leveraging MAUNet for Bias Correction of TRMM Precipitation Estimates, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14839, https://doi.org/10.5194/egusphere-egu25-14839, 2025.

The rising frequency and severity of landslides, exacerbated by the effects of climate change and human development in unstable areas, call for effective risk management strategies. In this context, a systematic collection of all the available data regarding geotechnical aspects, in particular geomaterial parameters, results plays a crucial role, providing a decisive contribution to define strategies for sustainable landslide risk management.

In this work, we present the translation of a geotechnical database to the aims of the project Tech4You Innovation Ecosystem – Goal 1 - Pilot Project 1, useful to identify the typical landslide scenarios, to identify sufficient knowledge for the definition of the geotechnical model and geomaterials typing in similar geo-environmental contexts. The database contains the results of laboratory tests carried out in the past by researchers at CNR IRPI in Rende, relating to 11 sites in Calabria, of which 10 in the Province of Catanzaro and 1 in the Province of Vibo Valentia.  For each site, geotechnical characterisation data of the geomaterials, which represent a key cognitive element, were grouped by type of laboratory test (grain size, indices, Atterberg limits, oedometric, direct shear and triaxial tests). We uploaded these data to validate a tool, named GeoDataTech vers. 2.0, which is an update of a previous version. In particular, we have tested the correct functioning (display, query, extraction data) with a significant sample of data. GeoDataTech vers. 2.0 can manage 2399 laboratory tests to date: 61 oedometric tests, 636 grain size, 537 indices, 78 Atterberg limits, 454 specific gravity, 512 direct shear tests and 121 triaxial tests.

This tool will be available to a wide range of stakeholders (researchers, professionals, territorial administrations, public bodies and citizens) allowing us to acquire, interrogate, export data and to upload their own files to integrate them into the database of the tool, performing advanced analyses with reference to the typification of geomaterials. By enabling the sharing of such data between researchers, practitioners and public institutions, the geotechnical tool will contribute significantly to improving disaster prevention strategies, in particular with regard to the reduction of landslide risks, thereby responding to the growing demand for accessible and interoperable data networks that increase synergic interdisciplinary research on topics such as landslide hazard.

This work was funded by the Next Generation EU—Italian NRRP, Mission 4, Component 2, Investment 1.5, call for the creation and strengthening of ‘Innovation Ecosystems’, building ‘Territorial R&D Leaders’ (Directorial Decree n. 2021/3277)—project Tech4You—Technologies for climate change adaptation and quality of life improvement, n. ECS0000009. This work reflects only the authors' views and opinions, neither the Ministry for University and Research nor the European Commission can be considered responsible for them.

How to cite: Scarcella, G. E., Aceto, L., and Gullà, G.: A relevant accessible and interoperable geotechnical data tool to support the landslide risk management, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15216, https://doi.org/10.5194/egusphere-egu25-15216, 2025.

Changes in land use and cover (LULC) serve as critical indicators of socioeconomic and environmental shifts induced by both natural and man-made factors. This assessment was carried out in the Imphal valley region to forecast changes in land use and land cover. In order to examine the spatiotemporal distributions of LULC, the LULC Classification was analysed using Landsat images from 2007, 2014, and 2017. The CA-Markov Chain model was used to simulate the future LULC for the year 2030 of Imphal valley region based on these the past LULCs. The model result showed that wetland herbaceous will decline by 3.3% and settlement area will expand by 28.71%. The Imphal city area is where the majority of the expanding settlement area is located. As a vital resource for future planning initiatives, this study suggests planners, environmentalists, and decision-makers to prioritise sustainable practices and make appropriate decisions for the sustainability of the region.

How to cite: Nongthouba, M., Oinam, B., and Sachidananda, K.: Prediction of landuse landcover using CA-Markov model for the valley regions of Manipur, India, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15381, https://doi.org/10.5194/egusphere-egu25-15381, 2025.

GNSS (Global Navigation Satellite System) data play a crucial role in both scientific research and practical applications. GNSS datasets are used to monitor atmospheric conditions, tectonic plate movements, and Earth deformation, providing valuable insights for geodetic and geophysical studies. Although widely accessible, GNSS data often lacks the necessary structure and metadata for effective reuse, particularly for data-driven research based on machine learning. To address these challenges, we applied the FAIR (Findable, Accessible, Interoperable, Reusable) data principles to GNSS RINEX observation files hosted by the EUREF Historical Data Centre (EUREF-HDC).

The EU action plan “Turning FAIR into Reality” introduced the concept of FAIR Digital Objects (FDOs), emphasizing the need for Persistent Identifiers (PIDs) and rich, standardized metadata to ensure data can be reliably found, accessed, utilized, and cited. Building on this foundation, we developed a multi-layered FDO structure centered on GNSS RINEX data. Given the established nature of the EUREF-HDC repository, we adapted the FDO concept by prioritizing structured metadata, followed by persistent identifiers and robust (meta)data access procedures.

To implement this approach, we designed the GNSS-DCAT-AP metadata schema, assigned PIDs to both data and metadata, and developed web services enabling humans and machines alike to seamless search, retrieve, and download (meta)data. The effectiveness of our solution was evaluated using the FAIRsFAIR Data Object Assessment Metrics, demonstrating a significant improvement in FAIR compliance.

This work showcases the feasibility of transforming GNSS RINEX data into FAIR Digital Objects and could provide a practical roadmap for other geospatial data repositories seeking alignment with FAIR principles.

How to cite: Bruyninx, C., Miglio, A., Fabian, A., Legrand, J., Pottiaux, E., and Bamahry, F.: Transforming GNSS Data into FAIR Digital Objects, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16092, https://doi.org/10.5194/egusphere-egu25-16092, 2025.

Tropical Cyclones (TCs) are among the most impactful weather phenomena, with climate change intensifying their duration and strength, posing significant risks to ecosystems and human life. Accurate TC detection, encompassing localization and tracking of TC centers, has become a critical focus for the climate science community. 

Traditional methods often rely on subjective threshold tuning and might require several input variables, thus making the tracking computationally expensive. We propose a cost-effective hybrid Machine Learning (ML) approach consisting in splitting the TC detection into two separate sub-tasks: localization and tracking. The TC task localization is fully data-driven: multiple Deep Neural Networks (DNNs) architectures have been explored to localize TC centers using a different set of input fields related to the cyclo-genesis, aiming also at reducing the number of input drivers required for detection. A neighborhood matching algorithm is then applied to join previously localized TC center estimates into potential trajectories over time. 

We train the DNNs on 40 years of ERA5 reanalysis data and International Best Track Archive for Climate Stewardship (IBTrACS) records across the East and West North Pacific basins. The hybrid approach is then compared with four state-of-the-art deterministic trackers (namely OWZ, TRACK, CNRM and UZ), reporting comparable or even better results in terms of Probability of Detection and False Alarm Rate, additionally capturing the interannual variability and spatial distribution of TCs in the target domain. 

The resulting hybrid ML model represents the core component of a Digital Twin (DT) application implemented in the context of the EU-funded interTwin project.

How to cite: Donno, D., Accarino, G., Elia, D., Scoccimarro, E., and Gualdi, S.: Hybrid Machine Learning approach for Tropical Cyclones Detection, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16363, https://doi.org/10.5194/egusphere-egu25-16363, 2025.

The STARS4Water project addresses the critical need to understand the impacts of climate change and anthropogenic activities on freshwater availability and ecosystem resilience at the river basin scale. By developing innovative data services and models tailored to stakeholder needs, the project will improve decision-making processes for sustainable water resource management. A distinctive feature of STARS4Water is its focus on co-creating solutions with local stakeholders using a living lab approach, ensuring that newly developed tools remain relevant and usable beyond the life of the project.

This extension of the original project—funded with a special grant from Unitatea Executivă pentru Finanțarea Învățământului Superior, a Cercetării, Dezvoltării și Inovării (UEFISCDI) from Romania—focuses on a detailed change detection analysis to monitor and quantify land cover transformations in the emblematic Danube Delta region. The objective is to assess how environmental and anthropogenic changes have influenced this ecologically significant wetland over several decades. To achieve this, a comprehensive database of multispectral satellite images from the Landsat archive, spanning from 1985 to 2023, will be constructed. The long-term dataset enables a detailed temporal analysis, important for detecting land cover dynamics over time.

The methodology involves several key phases: (1) data collection and preprocessing of Landsat satellite images to correct errors and align imagery for consistent comparative analysis; (2) sampling and training a deep learning model using convolutional neural network (CNN) architectures, to classify various land cover types; (3) performing land cover classification on the processed images using the trained model, followed by accuracy assessment; and (4) conducting a comprehensive change detection analysis to quantify and interpret the observed transformations in land use and land cover.

The results of this analysis will deliver important knowledge on the long-term dynamics of the Danube Delta landscape, highlighting critical changes with implications for biodiversity, water management and ecosystem services. This approach will support adaptive ecosystem management and contribute to the scientific understanding of climate-related and anthropogenic changes in fragile wetland ecosystems.

Acknowlegments

This work was supported by a grant of the Ministry of Research, Innovation and Digitization, CNCS/CCCDI - UEFISCDI, project number PN-IV-P8-8.1-PRE-HE-ORG-2023-0094, within PNCDI IV.

How to cite: Scrieciu, A. and Toma, A.: Monitoring Long-Term Land Cover Transformations in the Danube Delta using Landsat Satellite Imagery, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16549, https://doi.org/10.5194/egusphere-egu25-16549, 2025.

Abstract: Cloud cover poses a significant obstacle in harnessing multi-spectral satellite imagery for various earth observation applications including disaster response, land use and land cover mapping. To address this issue, this study investigates the potential of Prithvi WxC foundation model (Johannes Schmude et al., 2024), a deep learning architecture designed for weather and climate applications, to perform cloud gap imputation. By leveraging its ability to capture atmospheric dynamics and predict missing data, Prithvi WxC offers a promising solution.

The primary objective is to assess the accuracy and efficiency of Prithvi WxC in reconstructing cloudy pixels in Moderate Resolution Imaging Spectroradiometer (MODIS) surface reflectance; MOD09 (Eric Vermote, 2015). MOD09 data provides valuable information about earth surface, cloud cover and atmospheric conditions, which are instrumental in informing the Prithvi WxC model during the finetuning and imputation process.

This research evaluates the Prithvi WxC foundation model for cloud gap imputation applications and benchmarks its performance against other foundation models, such as Prithvi EO (Jakubik et al., 2023, 2024). The process begins with preprocessing the MOD09 dataset, filtering out missing and cloudy pixels to create clean visible patches, while real-world cloudy patches are used as masks. The preprocessed data is then resampled to align with the temporal and spatial resolution requirements of both the Prithvi WxC and Prithvi EO foundation models. Through rigorous fine-tuning strategies, these models learn to reconstruct the masked regions, effectively filling the gaps caused by cloud cover. Finally, the fine-tuned foundation models are benchmarked using quantitative metrics, such as the Structural Similarity Index Measure (SSIM) and Mean Absolute Error (MAE), complemented by qualitative visual analysis.

This research explores the potential of Prithvi WxC foundation model, pre-trained on extensive weather and climate data, to improve cloud gap imputation in satellite imagery, and subsequently benchmarks it against earth observation foundation models, such as Prithvi EO. Through this evaluation, we aim to enhance scientific understanding via multi-modality and sensor-independent approaches.

 References

Johannes Schmude, Sujit Roy, Will Trojak, Johannes Jakubik, Daniel Salles Civitarese, Shraddha Singh, Julian Kuehnert, Kumar Ankur, Aman Gupta, Christopher E Phillips, Romeo Kienzler, Daniela Szwarcman, Vishal Gaur, Rajat Shinde, Rohit Lal, Arlindo Da Sil: Prithvi WxC: Foundation Model for Weather and Climate." arXiv preprint arXiv:2409.13598, 2024.

C. Roger, E. F. Vermote, J. P. Ray: https://modis-land.gsfc.nasa.gov/pdf/MOD09_UserGuide_v1.4.pdf. NASA, MODIS Surface Reflectance User’s Guide, Collection 6, 2015.

Daniela Szwarcman, Sujit Roy, Paolo Fraccaro, Þorsteinn Elí Gíslason, Benedikt Blumenstiel, Rinki Ghosal, Pedro Henrique de Oliveira, Joao Lucas de Sousa Almeida, Rocco Sedona, Yanghui Kang, Srija Chakraborty, Sizhe Wang, Ankur Kumar, Myscon Truong, Denys: Prithvi-EO-2.0: A Versatile Multi-Temporal Foundation Model for Earth Observation Applications. https://arxiv.org/abs/2412.02732, 2024.

Johannes Jakubik, Sujit Roy, C. E. Phillips, Paolo Fraccaro, Denys Godwin, Bianca Zadrozny, Daniela Szwarcman, Carlos Gomes, Gabby Nyirjesy, Blair Edwards, Daiki Kimura, Naomi Simumba, Linsong Chu, S. Karthik Mukkavilli, Devyani Lambhate, Kamal Das, Ranji: Foundation Models for Generalist Geospatial Artificial Intelligence, 2023.

Eric Vermote: MOD09 MODIS/Terra L2 Surface Reflectance, 5-Min Swath 250m, 500m, and 1km. NASA LP DAAC., NASA GSFC and MODAPS SIPS, NASA, http://doi.org/10.5067/MODIS/MOD09.061, 2015.

How to cite: Medimem, T. B., Padovani, G., Kurihana, T., Kumar, A., Melgani, F., Anantharaj, V. G., and Fiore, S. L.: Finetuning and Benchmarking an AI Foundation Model for Cloud Gap Imputation , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16634, https://doi.org/10.5194/egusphere-egu25-16634, 2025.

The urgency of the climate crisis and the need to accelerate learning and climate action requires that we build on previous knowledge, rather than replicating it. There is an abundance of knowledge on climate change adaptation and mitigation dispersed across websites, projects, platforms, and documents. There is either too much information that is not easily discoverable (sitting in silos) or it is too technical or complex, and not ‘usable’ or fit for purpose in terms (e.g. language or format). Both issues cause redundancy and sometimes replication of work, wasting resources. In the worst case, they can also cause unintended consequences such as maladaptation, or increased vulnerability. However, the issue is not a lack of information, but rather how to organise and connect such knowledge to allow people to discover what already exists and put it to effective use. As such, our goal is to make climate action knowledge findable, accessible, interoperable and reusable (FAIR) and reduce climate change knowledge silos. The recently awarded FAIR2Adapt Project aims to establish a comprehensive FAIR and open data framework for CCA and to demonstrate the impact of FAIR data on CCA strategies. By making CCA data FAIR, FAIR2Adapt will accelerate adaptation actions so that they are visible, understandable, and actionable for various purposes and different types of stakeholders. FAIR taxonomies are one approach to help tackle this issue by making climate change knowledge FAIR and by ensuring, that going forward, platforms have a way to make their knowledge FAIR and thus more reusable by the climate change community. 

The Climate Connectivity Hub and Taxonomy seek to visualize and connect online platform data (e.g. Cordis, Climate-ADAPT, weADAPT, PreventionWeb) to increase discoverability, interoperability and a shared understanding of the research results and their potential application in future policy, research and practice. It builds on past knowledge to scale up and accelerate climate action whilst also identifying key knowledge gaps. This presentation will show that: 1) taxonomies are useful supporting the interoperability of online climate knowledge and can usefully emerge from combined expert and machine learning of project results (e.g. Cordis); 2) shared vocabularies and different interpretations of language and terminology add value to project planning and implementation ; and, 3) the visualization of these elements for decision-makers, planners, researchers, policy makers, etc. can help to enable and scale accelerated climate action. 

How to cite: Bharwani, S., Witton, R., Williamson, K., and Butterfield, R.: Visualizing a climate and disaster resilience taxonomy from research evidence: scaling and accelerating knowledge interoperability, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18822, https://doi.org/10.5194/egusphere-egu25-18822, 2025.

The study explores using advanced deep learning (DL) techniques for spatial-spectral analysis to detect and map peatland degradation at a granular level. Peatlands, vital carbon sinks in global ecosystems, face degradation threats that demand precise and scalable monitoring solutions. Our method combines convolutional neural networks (CNNs), fully convolutional networks (FCNs), and 3D CNNs to examine complex spatial-spectral patterns in SAR, multispectral, and hyperspectral sensor data (e.g., Sentinel-1, Sentinel-2, PRISMA) over the temperate peatland study area of the Monte Bondone region (Latitude: 46°00’48.6” N, Longitude: 11°03'14.6” E), covering an area of 40 hectares as shown in the figures.

CNNs capture spatial relationships between precipitation, temperature, vegetation, soil, and moisture, offering a detailed view of peatland composition. Using multi-dimensional, gridded data from meteorological stations and remote sensing images, CNNs identify patterns affecting peatland health. Fully Convolutional Networks (FCNs) help with spectral unmixing, isolating land cover components at the pixel level, which aids in detecting vegetation degradation and understanding ecosystem changes.

3D CNNs incorporate temporal data to classify Peatland landscapes into different degradation states. The model identifies changes over time, distinguishing between healthy, partially degraded, and fully degraded regions. Deep clustering models also classify peatland areas into degradation states, revealing trends without labeled data.

This deep learning framework supports accurate degradation mapping through spatial-spectral feature extraction, providing precise, pixel-level information to aid ecosystem management and conservation. It helps monitor peatland health and assess environmental changes across diverse landscapes.

How to cite: Kaparthi, H. V. and Vitti, A.: Deep Learning-based Spatial-Spectral Analysis for Peatland Degradation characterization, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18933, https://doi.org/10.5194/egusphere-egu25-18933, 2025.

Integrating heterogeneous hydrological datasets remains a significant challenge in environmental modelling due to variations in feature spaces, data distributions, and temporal and spatial scales across sources. This study introduces a Model-Agnostic Meta-Learning (MAML) approach to address the challenge of integrating heterogeneous hydrological datasets, leveraging a collection of datasets compiled from diverse sources. These datasets, characterized by varying features, distributions, and temporal and spatial scales, provide an ideal basis for evaluating MAML's ability to handle real-world data heterogeneity.

MAML’s unique capability to learn shared representations across datasets with minimal feature overlap and significant variability allows it to effectively transfer knowledge between subsets, offering a flexible and scalable solution for integrating hydrological data with diverse characteristics.

The proposed approach trains a base model on one subset of the data while utilizing MAML's meta-learning capabilities to adapt and transfer knowledge to other subsets with differing feature distributions. To test the model's adaptability, we simulate scenarios with varying degrees of feature overlap. Model performance is assessed using metrics such as mean squared error (MSE), both before and after fine-tuning on unseen data subsets.

Preliminary results demonstrate that MAML effectively learns shared representations across datasets, achieving significant improvements in prediction accuracy. Fine-tuning further enhances the model's adaptability, particularly for datasets with minimal feature overlap. These findings highlight MAML's potential as a powerful and flexible tool for integrating and predicting across heterogeneous hydrological datasets.

This study bridges the gap between advanced meta-learning techniques and hydrological applications, providing new insights into scalable and adaptable data integration methods for environmental sciences.

Keywords: Model-Agnostic Meta-Learning, hydrological datasets, data integration, heterogeneous data, meta-learning, environmental modelling, machine learning. 

How to cite: Slaimi, A. and Scriney, M.: Model-Agnostic Meta-Learning for Data Integration Across Heterogeneous Hydrological Datasets, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19489, https://doi.org/10.5194/egusphere-egu25-19489, 2025.

The semi-arid region of Deraa Oued Noun in Morocco faces significant challenges related to water scarcity, which greatly affects the availability of groundwater resources. With recurring droughts and periods of water shortage, it is imperative to address these challenges and implement effective measures for sustainable groundwater resource management. Artificial groundwater recharge has proven to be a viable solution for alleviating water scarcity issues. By capturing and storing excess water during periods of heavy precipitation or surface water availability, artificial recharge can replenish depleted aquifers and provide a reliable water source during drought periods. However, the success of recharge projects depends on identifying suitable sites that meet specific criteria and maximize the efficiency of the recharge process.

The identification of suitable sites for artificial groundwater recharge in Daraa Oued Noun, through the integration of remote sensing, GIS (Geographic Information System), and MCDM (Multi-Criteria Decision Making) techniques, offers a promising solution to address water scarcity challenges in the context of climate change. The proposed research project aims to provide valuable and spatially explicit information for strategic groundwater resource management.

 This study was conducted in the Deraa Oued Noun district, where water shortages have been observed over the years. The research utilized geology, soil, land use, stream data, and Sentinel-2 and DEM images to develop thematic layers, including lithogeology, soil, slope, lineament density, land use, stream density, and water surface. Additionally, data on the vadose zone thickness were incorporated to enhance the analysis.

By integrating GIS and image processing techniques, these thematic layers were utilized to prepare groundwater recharge maps of the area through a weighted overlay method on a GIS platform. The results revealed that artificial recharge potential was high in the northern and western parts of the study area.

By following a systematic and rigorous methodology, including data collection, remote sensing analysis, MCDM evaluation, and site validation, this project aims to contribute to the successful implementation of artificial recharge projects in the region. By maximizing the efficiency of the recharge method, these projects will help ensure sustainable water supply, mitigate the impacts of drought, and promote long-term water security in Derâa Oued Noun and similar semi-arid regions.

How to cite: Fri, R., Scozzari, A., Haida, S., Kili, M., Erraoui, L., Chaou, J., Mridekh, A., Goumghar, L., and El Mansouri, B.: Using Remote sensing and geographic information system for delineating suitable sites for artificial groundwater recharge: A multi-criteria decision-making approach., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20324, https://doi.org/10.5194/egusphere-egu25-20324, 2025.

We present LUCIE, a data-driven atmospheric emulator that remains stable during autoregressive inference for a thousand of years with minimal drifting climatology. LUCIE was trained using 9.5 years of coarse-resolution ERA5 data, incorporating 5 prognostic variables, 2 forcing variables, and one diagnostic variable (6-hourly total precipitation), all on a single A100 GPU over a two-hour period. LUCIE autoregressively predicts the prognostic variables and outputs the diagnostic variables similar to AllenAI’s ACE climate emulator. Unlike all the other state-of-the-art AI weather models, LUCIE is neither unstable nor does it produce hallucinations that result in unphysical drift of the emulated climate. The low computational requirements of LUCIE allow for rapid experimentation including the development of novel loss functions to reduce spectral bias and improve tails of the distributions. Furthermore, LUCIE does not impose true sea-surface temperature (SST) from a coupled numerical model to enforce the annual cycle in temperature. We demonstrate the long-term climatology obtained from LUCIE as well as subseasonal-to-seasonal scale prediction skills on the prognostic variables. LUCIE is capable of 6000 years of simulation per day on a single GPU, allowing for O(100)-ensemble members for quantifying model uncertainty for climate and ensemble weather prediction.

How to cite: Guan, H., Arcomano, T., Chattopadhyay, A., and Maulik, R.: LUCIE: A Lightweight Uncoupled ClImate Emulator with long-term stability and physical consistency for O(1000)-member ensembles, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20616, https://doi.org/10.5194/egusphere-egu25-20616, 2025.

Analysis of pollen events was conducted using Hirst-type volumetric samplers in Athens and Thessaloniki in combination with CALIPSO vertical aerosol profiles. While Hirst-type ‎[1] volumetric samplers are used to confirm and characterize pollen at ground level, the understanding of pollen vertical distribution and transport is still limited. The utilization of Light Detection and Ranging (LIDAR) for identifying different pollen types is increasingly prevalent, as the depolarization ratio is related to the shape of the pollen particles while other non-spherical particle types are absent ‎[2].
Samplers are situated on the buildings’ rooftops of the Physics and Biology Departments, in Athens and Thessaloniki, respectively. Following ‎[2], intense pollen events are considered when the pollen concentration exceeds 400 grains m-3 for a minimum of two hours each day.
CALIPSO provides unique vertical profile measurements of the Earth’s atmosphere on a global scale ‎[3], with the ability to distinguish between feature types (i.e., clouds vs. aerosol) and subtypes (i.e., marine, dust, clean continental). Only case studies where CALIPSO aerosol layers were classified as marine, dusty marine, dust, or polluted dust were analyzed.
Model simulations were used to exclude the presence of other depolarizing aerosol types. HYSPLIT (Hybrid Single-Particle Lagrangian Integrated Trajectory) was used to trace the origin of the air masses. The atmospheric model RAMS/ICLAMS (Regional Atmospheric Modeling System/Integrated Community Limited Area Modeling System) was selected to describe dust and sea-salt emissions and transport.
Mean values of lidar-derived optical properties inside the detected pollen layers are provided from CALIPSO data analysis. Specifically, there are three observed aerosol layers, one over Athens (12-3-2021) and two over Thessaloniki (2-3-2020, 10-4-2020). Particulate color ratios of 0.652 ± 0.194, 0.638 ± 0.362, and 0.456 ± 0.284, and depolarization ratios of 8.70 ± 6.26%, 28.30 ± 14.16%, and 8.96±6.87 % for 12-3-2021, 2-3-2020 and 10-4-2020, respectively, were misclassified by CALIPSO as marine-dusty marine, dust and polluted dust. The pollen analysis conducted on the 12th of March 2021 indicated that the dominant pollen types were 69% Pinaceae and 24% Cupressaceae. On the 2nd of March 2020, Cupressaceae accounted for 97% of the total pollen, while on the 10th of April 2020, Carpinus represented 76% and Platanus 15%. Consequently, during periods of intense pollen presence, CALIPSO vertical profiles and aerobiological monitoring techniques may be used synergistically to better characterize the atmospheric pollen layers.

Acknowledgements
The research work was supported by the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the “Basic Research Financing (Horizontal support for all Sciences), National Recovery and Resilience Plan (Greece 2.0)” (Project Number: 015144).

[1] J. M. Hirst, Annals of Applied Biology 39, 157-293 (1952).
[2] X. Shang et al., Atmos. Chem. Phys. 20, 15323–15339 (2020).
[3] D. M. Winker et al, BAMS 91, 1211–1229 (2010).

How to cite: Karageorgopoulou, A., Christos, S., Thanasis, G., Χiaoxia, S., Ioanna, P., Vassilis, A., and Elina, G.: Assessment of Atmospheric Pollen Presence in Urban Areas of Greece During CALIPSO Overpasses, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21759, https://doi.org/10.5194/egusphere-egu25-21759, 2025.

The Geostationary Ocean Color Imager-II (GOCI-II), which was launched on February 19, 2020, offers an increased observation times within a day and finer spatial resolution than those of its predecessor, the Geostationary Ocean Color Imager (GOCI), which was launched in 2010. To ensure the reliability of GOCI-II data for practical applications, the accuracy of remote sensing products must be validated. In this study, we employed in situ data from Lake Taihu for validation. We assessed the accuracy of GOCI-II products, including the remote sensing reflectance inverted via two atmospheric correction algorithms (ultraviolet (UV) and near-infrared (NIR) atmospheric correction algorithms), as well as the chlorophyll a (Chl-a) concentration, total suspended matter (TSM) concentration, and phytoplankton absorption coefficient (a_ph). Our results revealed that the UV atmospheric correction algorithm provided a relatively higher accuracy in Lake Taihu, with average absolute percentage deviations (APDs) of the remote sensing reflectance across different bands of 25.17% (412 nm), 29.69% (443 nm), 22.27% (490 nm), 19.38% (555 nm), 36.83% (660 nm), and 33.0% (680 nm). Compared to the products generated using the NIR atmospheric correction algorithm, the derived Chl-a concentration, TSM concentration, and a_ph products from the UV algorithm showed improved accuracy, with APD values reduced by 16.92%, 3.32%, and 10.91%, respectively. When using UV correction, the 412 nm band performed better than the 380 nm band, likely due to the lower signal-to-noise ratio of the 380 nm band and smaller extrapolation errors when assuming a zero signal for the 412 nm band. Considering that the NIR algorithm is suitable for open ocean waters while the UV algorithm demonstrates higher accuracy in highly turbid environments, a combined UV-NIR atmospheric correction algorithm may be more suitable for addressing different types of water environments. Additionally, more suitable retrieval algorithms are needed to improve the accuracy of Chl-a concentration and a_ph in eutrophic waters.

How to cite: Zhao, M., Li, H., Li, H., Zhang, X., Ding, X., and Gong, F.: Assessment of GOCI-II satellite remote sensing products in Lake Taihu, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2237, https://doi.org/10.5194/egusphere-egu25-2237, 2025.

This study explores studies on lakes and paleolakes originating from natural effects. The main objective is to perform a bibliometric analysis of research on naturally occurring lake environments worldwide, covering the period from 2014 to 2024. Data extracted from 1687 documents in the Scopus database were analyzed using VOSviewer software. The results reveal a strict trend towards a focus on geosciences and the environment, underlined by research. This study particularly highlights the relationships between authors, co-authors, keywords, and publishers of specialized journals in this research field, thus providing essential information to guide future research and to value the role of these geological environments, which are rare in the world, based on essentially multidisciplinary geoscience approaches.

How to cite: Abbach, J., El Moussaoui, S., El Talibi, H., and Bouiss, C. E.: bibliometric analysis of natural lakes and paleolakes origin of natural events, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4994, https://doi.org/10.5194/egusphere-egu25-4994, 2025.

Rock masses are characterized by the complex hierarchical structures involving various scale levels. The deformation of rock masses is primarily controlled in weak structural layers between rocks, whereas the rock block can be regarded as a non-deformable block and can move as a whole. In consequence, a new dynamic phenomenon, namely the pendulum-type wave, has emerged, which is a kind of nonlinear displacement wave caused by the overall movement of relatively intact large-scale rock blocks. Aiming at the complex hierarchical structures of rock masses and low-frequency characteristics of pendulum-type waves, the blocky rock masses composed of granite blocks and rubber interlayers are simplified into the block-spring model and wave motion model. Based on Bloch theorem and d’Alembert’s principle, the dispersion relation and equations of motion of 1D blocky rock masses are determined. Research shows that with the increase of the rock size and geomechanical invariant, the initial frequency of the first attenuation zone gradually decreases, and only the low-frequency waves lower than that frequency can propagate in blocky rock masses, which reveals the mechanism of low-frequency characteristics of pendulum-type waves theoretically. The equivalent substitution for the two models and their errors are given, and the results show that the equivalent substitution of the two models is not universal and unconditional. Finally, the influence of hierarchical structures on the dispersion relation and dynamic response is further studied. The larger the stiffness ratio, or the higher the order of hierarchical structures, the smaller is the effect of ignoring the high-order hierarchical structures.

How to cite: Jiang, K. and Qi, C.: Research on dispersion relation and dynamic properties of pendulum‑type waves in 1D blocky rock masses with complex hierarchical structures, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5033, https://doi.org/10.5194/egusphere-egu25-5033, 2025.

Coral reefs in the Mariana Islands serve important roles for the islands’ ecology and economy, contributing to the region’s fisheries, tourism, coastal protection, education, and cultural histories. Despite their immense value, the resilience of these marine ecosystems is threatened by an array of climate-change induced stressors, including ocean acidification and coral bleaching. In response, a team from the University of Guam (UOG) launched a large-scale coral reef mapping campaign to monitor priority reef sites throughout the Mariana Islands using drone technology. The UOG team, consisting of researchers and remote pilots funded by the USGS, Pacific Islands Climate Adaptation Science Center, NASA Guam EPSCoR, and NASA Guam Space Grant, has been conducting drone-based missions to capture high-resolution imagery of priority coral reef sites across Guam and Saipan. Their efforts aim to gather aerial data of coral reefs in Micronesia, providing resource managers with essential information regarding response and recovery. Initially, the campaign used of NASA’s fluid lensing technology developed by Chirayath (2019) for coral reef mapping. This technology combines unmanned aerial systems (UAS), off-the-shelf technology, and machine learning algorithms to create detailed coral reef maps by filtering out distortions caused by light and ocean waves, resulting in clear, high-resolution imagery. In 2024, this process was augmented to employ a new methodology that strategically uses RGB sensors and low tides. This system allows the remote pilots to capture the areas and produce orthomosaic maps at much more efficient rates while maintaining high-resolution quality. By providing these datasets within a shorter turn-around time, local natural resource managers are able to get a timely snapshot of the coral reef sites – providing crucial data of the ecosystem’s health that can help inform conservation decisions. This presentation will outline the collaborative efforts between UOG and regional partners, demonstrating how drones and fluid lensing technology are innovating coral reef monitoring efforts. It will explore how the collected data can help local resource managers make informed decisions regarding coral reef management, showcase coral reefs to the general public, ultimately transforming how local communities can contribute to coral reef resilience.

How to cite: Sayama, J. and Fausto, K.: Innovating Coral Reef Mapping with Drones & NASA Fluid Lensing Technology in the Mariana Islands, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5359, https://doi.org/10.5194/egusphere-egu25-5359, 2025.

In today's changing climate, there is an urgent need to understand the adverse impacts of climate change on natural environments, infrastructures, and industries.Particularly permafrost regions in the Arctic are highly vulnerable to global warming, impacting both the environment and socioeconomic aspects. Thus, systematic monitoring of such environments, is of paramount significance. Advances in Geoinformation technologies, and in particular in Earth Observation (EO), cloud computing, GIS, web cartography create new opportunities and challenges for Arctic research examining the socioeconomic impact of climate change.The rapid advancements in EOin particular have led to an exponential increase in the volume of geospatial data that come from spaceborne EO sensors. This surge, combined with the fast developments in GIS and web cartography present significant challenges for effective management, access, and utilization by researchers, policymakers, and the public. Consequently, there is a growing need for advanced methodologies to organize, process, and deliver geospatial information that comes from EO satellites in an accessible and user-friendly manner.

Recognizing thepromising potential of geoinformation technologies, the European Union (EU) has funded several research projects that leverage advanced technologies such as geospatial databases and WebGIS platforms to streamline EO data handling and dissemination. One such project is EO-PERSIST (http://www.eo-persist.eu), which aims to create a collaborative research and innovation environment focusing on leveraging existing services, datasets, and emerging technologies to achieve a consistently updated ecosystem of EO-based datasets for permafrost applications. To formulate the socioeconomic indicators, the project exploits state of the art cloud processing resources, innovative Remote Sensing (RS) algorithms, Geographic Information Systems (GIS)-based models formulating, exchanging also multidisciplinary knowledge.EO-PERSIST innovative approach is anticipated to contribute to more informed decision-making and broader data accessibility for researchers, policymakers, and other stakeholders.

The present contribution aim is two-fold: at first, to provide an overview of EO-PERSIST Marie Curie Staff Exchanges EU-funded research project; second, to present some of the key project outputs delivered so far relevant to the selected Use Cases of the project and the geospatial database developed for assessing the socioeconomic impacts of climate change in the permafrost Arctic regions.

This study is supported by EO-PERSIST project which has received funding from the European Union's Horizon 2020 research and innovation program under grant agreement No. 101086386.

KEYWORDS:earth observation, cloud platform, Arctic, socioeconomic impact

How to cite: Tselos, G.-N., Detsikas, S. E., Kroszka, B., Grzybowski, P., and Petropoulos, G. P.: Enhancing the use of Geoinformation technologies to assess the socioeconomic impacts of climate change in the Arctic: Insights from the EO-PERSIST Project, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6240, https://doi.org/10.5194/egusphere-egu25-6240, 2025.

The U.S. territory of Guam is threatened annually by high-intensity storms and typhoons due to its location in the western Pacific Ocean. The island’s infrastructure – buildings, roads, and utilities – bear the brunt of typhoon damage, which in turn affects public health, the economy, and natural resources. Traditionally, these impacts have been observed via satellite, radar, and official weather stations.  Damages are assessed in the aftermath of the typhoon with a manual, on-the-ground approach led by the National Weather Service (NWS). This is often exhaustive and time-consuming for the assessment team. Observations from the ground can inadvertently create data gaps on damage assessments due to inaccessible areas caused by vegetative and construction debris, and flooded roads and pathways. This may not capture many impacts eligible for local or federal assistance. To address these data gaps and augment damage assessments, the University of Guam (UOG) Drone Corps program aims to assist local and federal government agencies (e.g., utility companies, public health, emergency services, and natural resource management) by collecting high-resolution aerial imagery to help prioritize and allocate limited resources. This presentation highlights the results of this novel collaboration of UOG, NWS, Guam Homeland Security (GHS), and the Office of the Governor of Guam in the creation of the damage assessment of Typhoon Mawar, which ravaged Guam on 24-25 May 2023. Following the typhoon, UOG worked with NWS to identify and capture imagery of vulnerable sites that were heavily impacted. This presentation will also share how UOG Drone Corps’ data was disseminated among other agencies as supplemental data for natural disaster recovery efforts. The presentation will conclude with a summary of the UOG Drone Corps program model as a resource for developing resiliency strategies for vulnerable island communities using advanced and emerging technologies. 

How to cite: Fausto, K. and Sayama, J.: Deploying UAV technology to assess typhoon impacts in vulnerable communities in Guam , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8063, https://doi.org/10.5194/egusphere-egu25-8063, 2025.

INTRODUCTION. InSAR or Interferometric Synthetic Aperture Radar is a technique for mapping ground deformation using radar images of the Earth's surface collected from orbiting satellites. DInSAR or Differential SAR Interferometry is an active remote sensing technique based on the principle that, due to the very high stability of the satellite orbits, it is possible to exploit the informative contribution carried by the phase difference between two SAR images looking at the same scene from comparable geometries.

AIM. In this setting, the main objective of this study is to evaluate the region near the closed rock salt mine in the south of Albania. Our input for this exercise will be two images of the land near the former rock salt mine in Dhrovjan near the Blue Eye (Saranda, Albania).

RESULTS. By combining the phases of 2 images we produce an interferogram where the phase is correlated to the terrain topography and deformation so if the phase shifts related to the topography are removed from the interferogram, the difference between the resulting products will show surface deformation patterns or cure between the two acquisition dates and this methodology is called differential interferometry Processing, Phase Unwrapping, and at the end creating the displacement map. We use in our study the difference in time with the algorithm which consists of working step by step with these operators: Read the two split products, Applying Orbit files, Back-Geocoding, Enhanced Spectral Diversity, Interferogram, TOPSAR Deburst, and Write. The resulting difference of phases is called an interferogram containing all the information on relative geometry. Removing the topographic and orbital contributions may reveal ground movements along the line of sight between the radar and the target.

The next algorithm we worked with these operators: Read the debursted interferogram, TopoPhaseRemoval, Multilook, Goldstain Filtering, and Write. At the same time from Goldstain Filtering, we add the Snaphu Export operator.

Correct phase unwrapping procedures must be performed to retrieve the absolute phase value by adding multiples of 2π phase values to each pixel to extract accurate information from the signal. In this study, we will use SNAPHU, which is a two-dimensional phase unwrapping algorithm consists of working step by step with these operators: read (the wrapped image) and read (2) the unwrapped image, Snaphu Import, PhaseToDisplacement, and Write. We can display it in Google Earth after saving it as .kmz and also make a profile of the displacements.

DISCUSSION AND CONCLUSIONS

One of the SAR Interferometry applications is deformation mapping and change detection. This work demonstrates the capability of interferometric processing for the observation and analysis of instant relative surface deformations in the radar LOS direction. When two observations are made from the same location in space but at different times, the interferometric phase is proportional to any change in the range of a surface feature directly. All three stages of the work are important and require accurate interpretation knowledge, especially when working with the Snaphu program.

KEY-WORDS

InSAR, DInSAR, Interferogram

How to cite: Belba, P.: The use of InSAR and DInSAR for detecting land subsidence in Albania, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11808, https://doi.org/10.5194/egusphere-egu25-11808, 2025.

Poincaré established a framework for understanding the dependence of a dynamical system's properties on its topology. Topological properties offer detailed insights into the fundamental mechanisms — stretching, squeezing, tearing, folding, and twisting — that govern the shaping of a dynamical system's flow in state space. These mechanisms serve as a conduit between the system's dynamics and its topology [Ghil & Sciamarella, NPG, 2023]. A topological analysis based on the templex approach [Charó et al., Chaos, 2022] involves finding a topological representation of the underlying structure of the flow by the construction of a cell complex that approximates its branched manifold and a directed graph on this complex. A pivotal feature of the cell complex that facilitates the characterization of the flow dynamics is the joining locus, upon which all the fundamental mechanisms that sculpt the flow leave a pronounced signature.

The local dimension d(x) and the inverse persistence θ(x) of the state x of a dynamical system [Lucarini et al., 2016; Faranda et al., Sci. Rep., 2017] provide information on the rarity and predictability of specific states, respectively. We demonstrate herein that these two measures, d and θ  also provide information about the localization of the joining locus.

The present work proposes a new topological method for fingerprinting a system’s nonlinear behavior using the concept of persistent generatexes. This novel approach integrates the strengths of two topological data analysis methods: the templex and persistent homologies. Rather than employing a single cell complex and a digraph to characterize the flow of the system, our approach emphasizes the localization of the joining locus through the calculation of local dimension and the inverse persistence, leading to the construction of a family of nested digraphs. The dynamical paths, namely the nonequivalent ways of travelling through the flow, are found to be the most persistent cycles; here the concept of persistence is used in the sense of the persistent homology approach [Edelsbrunner & Harer, Contemporary mathematics, 2008]. The dynamical paths give us the ‘topological fingerprinting’ of a system’s dynamics.

How to cite: Charó, G. D., Faranda, D., Ghil, M., and Sciamarella, D.: Topological fingerprinting of dynamical systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12947, https://doi.org/10.5194/egusphere-egu25-12947, 2025.

The Er-Rich region is a focal area for understanding the geological evolution of the central-eastern High Atlas, which it covers almost entirely along a north-south transverse line. It is a hinge region between the two major tectonic structures of the High Atlas (the North and South Atlas faults), which reveal a framework of Meso-Cenozoic carbonate, detrital and magmatic rocks.

Previous studies have highlighted the complexity of mapping in this area. To date, no detailed geological map has been produced for this study area, with the exception of the old provisional 1:200,000 map of the Midelt-Rich High Atlas. Remotely-sensed mapping initiatives have also been carried out in the region, except that they do not provide a final interpretation as a geological map, supported by geological maps covering neighboring regions. A detailed geological map of the Er-Rich region, based on the results of remote sensing and field data, is therefore needed in the area. For this purpose, remote sensing geological mapping techniques have been applied to two types of satellite data: 1) Landsat 8 OLI (Optical Land Imager) multispectral optical data, and the Spot 5 panchromatic band acquired by the HRG-2 (High Geometric Resolution) instrument; 2) Sentinel-1 SAR data with dual polarisation (HV-HH).

All the data underwent several pre-processing or correction stages using appropriate software, in particular radiometric and atmospheric correction for Landsat 8 OLI (Optical Land Imager) images using ENVI software. The corrected product of the three Landsat 8 OLI scenes covering the region were then spatially enhanced using the Spot 5 panchromatic band to produce a multispectral image with a high spatial resolution of 5 m using ENVI software. The Sentinel-1 radar data were pre-processed using SNAP toolbox software by applying a series of corrections.

The results obtained by applying the Optimum Index Factor (OIF) method and Principal Component Analysis (PCA), allowing us to select the most significant colored compositions. Moreover, this combination enabled us to delineate with great precision the large outcrops of carbonate rocks (limestones, marl), siliciclastic rocks (conglomerates, sandstones and silts) and magmatic rocks (igneous intrusions).

The lineaments were extracted manually by visual interpretation of Sentinel-1 radar images, after applying directional filtering folowing four general orientations (N0, N45, N90, N135), enabling us to generate a synthetic structural map of the region.

The results obtained were compared with data from geological maps of adjacent areas and approved by field observations, leading to the production of a high-precision geological map, compiled with pre-existing geological literature.

How to cite: Hdoufane, M., Zafaty, O., and Ettaki, M.: Integrated remote sensing data and field investigations for geological mapping and structural analysis in the Er-Rich area (High Atlas, Morocco), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13782, https://doi.org/10.5194/egusphere-egu25-13782, 2025.

How to cite: Mendes, J. and Portier, M.: Uniform Data Access Layer: Advancing Data FAIRness in FAIR-EASE, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15569, https://doi.org/10.5194/egusphere-egu25-15569, 2025.

The impacts of global change, such as extreme heat and water scarcity, are increasingly threatening urban populations. Evapotranspiration (ET) plays a vital role in mitigating urban heat islands and reducing the effects of heat waves. It also serves as a proxy for vegetation water use, making it a critical tool for designing resilient green cities. Despite its importance, high-resolution mapping of urban ET that captures both spatial and temporal dynamics remains limited. This study focuses on the Paris metropolitan area, analyzing ET variability across multiple spatial scales (from 10 m to 10 km) using Sentinel-2 data from the Copernicus system. The Normalized Difference Vegetation Index (NDVI) is calculated with observation scale of 10 m, and then used as a proxy for ET. Universal Multifractal analysis, which have been widely used to characterize and model geophysical fields extremely variable across wide range of space-time scales, are implemented on this new data set. This framework is parsimonious since it basically relies on three parameters only: the mean intermittency codimension C₁, the multifractality index a and the non-conservation parameter H.  Specifically, the multifractality index α (1.3–1.5) and the mean intermittency codimension C₁ (~0.02) were derived to quantify the spatial and temporal heterogeneity of ET. The analysis, spanning 2019–2023, revealed noticeable temporal and spatial variability in ET. The study focuses on a square region of approximately 60 km × 60 km within the area around Paris. This region was further divided into multiple portions of size ranging from 2 to 10 km to assess potential variability over the studied areas. By incorporating both yearly and monthly data, the analysis captured seasonal trends as well as interannual variability, with higher variability observed during the summer months, driven by increased vegetation activity and water demand. Spatially, yearly data was analyzed and ET variability was most pronounced in densely populated areas, such as central Paris, where anthropogenic influences dominate. In contrast, forested areas and urban parks demonstrated significantly more stable ET patterns, underscoring the moderating effect of vegetation cover. These findings highlight the critical role of urban greening in mitigating extreme variability and stress on urban ecosystems.

How to cite: Zhu, S., Gires, A., Schertzer, D., Tchiguirinskaia, I., and Maksimovic, C.: Temporal and Spatial Dynamics of Urban Evapotranspiration in Paris: A Multiscale Perspective, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16203, https://doi.org/10.5194/egusphere-egu25-16203, 2025.

The ESA Directorate of Earth Observation Programmes has been actively leveraging satellite-based environmental information to address fragility contexts, focusing on areas such as environmental crimes, crimes against humanity, cross-border crimes, and onset of crises. Over the past decade, ESA has explored digital intelligence crime analysis by employing advanced data mining and machine learning tools to uncover hidden patterns and relationships in historical crime datasets, enabling better detection, prediction, and prevention of criminal activities.

Despite these advancements, the integration of Earth Observation (EO) capabilities into investigative practices remains limited. This is due to several challenges, including low awareness of EO's potential, a lack of illustrative use cases showcasing its benefits, inconsistencies in satellite data collection compared to investigative needs, high costs of very high-resolution imagery, and restricted access to national intelligence sources. To overcome these barriers, ESA has been investigating strategies to systematically incorporate EO-derived information into investigative frameworks also as legal evidence, aiming to enhance situational awareness and support stakeholders in developing procedures to exploit EO and OSINT for addressing international crimes and assessing fragility contexts, in cooperation with international organizations including Interpol, UNODC and ICC.

Recent developments in EO technology and methodologies have created significant opportunities for more impactful applications. ESA has focused on tailoring EO-based services and OSINT to meet the case-sensitive requirements of security and development end-users, enabling better integration of EO-derived insights into intelligence models. These efforts include developing advanced EO information products that go beyond routine offerings, testing and evaluating these products in collaboration with end-users, and demonstrating their value in operational settings.

The GDA Fragility, Conflict, and Security initiative has been a cornerstone of ESA’s work, involving partnerships with International Financial Institutions (IFIs) to co-design tools that provide precise and timely information. These tools have supported initiatives aimed at reducing inequalities, promoting economic development, and enhancing environmental safety in fragile and conflict or post conflict-affected areas. By combining geospatial data with diverse data sources, ESA has delivered customized analyses and reports to improve emerging threats analysis and decision-making processes.

Several ESA initiatives have demonstrated the benefits of EO services for assessing fragility risk exposure, characterizing dynamic needs in fragile contexts, planning post-conflict reconstruction, and managing natural resources. ESA constantly engages with stakeholders, including the OECD, security organizations, and humanitarian actors, and its community of industries and research centres to promote the adoption of EO in international development, humanitarian aid, and peacebuilding. Through these efforts, ESA continues to advance the role of EO in supporting justice, accountability, and sustainable recovery in fragile settings.

How to cite: Corvino, M. and the Michela Corvino: Leveraging EO for Security and Resilience, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17156, https://doi.org/10.5194/egusphere-egu25-17156, 2025.

The Canary Eddy Corridor is a dynamic region of mesoscale eddy activity, playing a critical role in the transport of physical properties (heat and salt) and biogeochemical properties (nutrients, larvae, plankton) in the eastern North Atlantic. This study investigates the Lagrangian evolution of the trapping capacity of mesoscale eddies according to their lifecycle phases and vertical structure (surface vs. subsurface eddies).

We combine OceanParcels (an open-source Python toolbox) and an eddy identification and tracking algorithm with the GLORYS12V1 reanalysis product and altimetry data from AVISO to simulate particle release and track trajectories within eddies. Applying the eddy tracking algorithm at surface and subsurface levels in GLORYS12V1 reveals that subsurface eddies with a surface signal exhibit subsurface rotational velocities at the eddy core that occasionally exceed those of surface eddy cores. This highlights the potential misrepresentation of eddy transport capacity when relying solely on altimetry data, without accounting for the vertical structure, which can be better resolved through a combination of model outputs and observational data, such as non-standard Argo float configurations. Furthermore, a detailed analysis of the eddy lifecycle phases shows that mature eddies exhibit substantially greater trapping depths compared to their growth and decay stages. These findings align with earlier modeling analyses of dipoles originating south of Madagascar, which also highlight enhanced trapping depths in mature eddies.

The results provide a comprehensive view of the trapping capacity of mesoscale eddies throughout their lifecycle and vertical structure, emphasizing their critical role in biophysical coupling, ecological connectivity, and the transport of biogeochemical properties, as well as microplastics and other pollutants.

Acknowledgments: The first author is grateful for the internship grants ERASMUS +, AMI-MESRI, and TIGER. 

How to cite: Vacca, D., Aguiar-González, B., and Morris, T.: Lagrangian Evolution of the Trapping Capacity of Mesoscale Eddies in the Canary Eddy Corridor: A Numerical Modeling Approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17965, https://doi.org/10.5194/egusphere-egu25-17965, 2025.

The abandonment of agricultural land in India, especially paddy fields, has emerged as a significant challenge for food security and ecosystem sustainability in the country. Although rice production is vital for national food security, research on paddy land abandonment in India remains limited. Some Indian states have reported an alarming decline in paddy cultivation area over the past two decades. The study employs the Udupi district of Karnataka, India, a high-rainfall coastal region where paddy has traditionally been the dominant crop and where paddy land abandonment has been observed, as the study area. This study addresses crucial research gaps by framing these objectives for the study: (1) developing a deep learning framework that utilizes both intensity and phase information from polarimetric Synthetic Aperture Radar (SAR) data for abandoned paddy land detection, (2) leveraging recurrent neural networks (RNNs) to capture temporal patterns in abandonment, and (3) demonstrating an automated, all-weather monitoring approach that overcomes the limitations of traditional optical remote sensing in tropical regions.

Conventional monitoring approaches struggle with persistent cloud cover in tropical regions which limits effective assessment of abandonment patterns. SAR data provides unique capabilities for continuous monitoring under all weather conditions, making it particularly well-suited for tropical regions. However, previous studies have primarily underutilized SAR's potential by concentrating solely on backscattering intensity from ground range detected (GRD) products, overlooking the valuable phase information that could offer deeper insights into land use changes.  In this study, we employ Sentinel-1 Single Look Complex (SLC) data, which offers both intensity and phase information. Considering the temporal nature of paddy land abandonment, we developed a deep learning framework utilizing RNNs viz. LSTM, BiLSTM and BiGRU to effectively capture time-series patterns in the data. This framework analyzes backscattering coefficients (VV and VH polarizations) and polarimetric parameters (entropy, anisotropy and alpha angle) derived from SLC data collected during the Kharif seasons from 2017 to 2024. We carried out extensive ground truth data collection of active and abandoned paddy lands to train and validate our models. The backscattering coefficients were processed through orbit correction, radiometric calibration, TOPSAR deburst, multi-looking, speckle filtering and terrain correction. For deriving the polarimetric parameters, after basic preprocessing steps, the covariance matrix was generated followed by the polarimetric decomposition of the phase-preserved data. Results indicate that our RNN models show promising performance in detecting temporal patterns of paddy land abandonment. The method exhibits a robust ability to produce reliable abandoned land maps in regions prone to cloudy and rainy conditions. Future research should explore polarimetric features across various vegetation types in abandoned lands, expand the methodology to other agricultural systems, and examine the impact of socio-economic and topographical factors on abandonment patterns to support evidence-based land management policies.

How to cite: Kasture, S., Hegde A, A., and Umesh, P.: Deep Learning based Paddy Land Abandonment Detection Using Multitemporal Polarimetric SAR Patterns, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18606, https://doi.org/10.5194/egusphere-egu25-18606, 2025.

The presence of rain in wind farms involves several modeling challenges, as the momentum exchanges between turbulent wakes and the particle phase present subtle phenomena. For instance, rain droplets are typically large enough to exhibit inertia relative to the air carrier phase. Under these conditions, it has been found that the gravitational settling of particles in turbulent flows may be either enhanced or hindered compared to stagnant conditions. While this has significant implications for rainfall transport, ash pollutants, and pollen dispersion, very few studies have been conducted in field conditions. Moreover, the scaling laws and non-dimensional parameters governing this phenomenon have not yet been properly identified, and determining which configurations result in the enhancement or hindrance of settling velocity remains an open question.

We propose a hybrid experimental/numerical approach. Field data from a meteorological mast located at a wind farm in Pays d’Othe, 110 km South-East of Paris, France, were used to characterize the background turbulent flow through a set of sonic anemometers. Additionally, disdrometers were employed to characterize the settling velocity of raindrops, discriminating by particle size. Numerical simulations complement this data analysis. Specifically, 3D space and time vector fields that realistically reproduce the observed spatial and temporal variability of wind fields are generated using multifractal tools. Then, 3D trajectories of non-spherical particles are simulated and their settling velocity derived.

Our findings indicate that the presence of turbulence significantly hinders the settling velocity of raindrops in turbulent environments. Our study covers several distinct rainfall events, allowing us to analyze the influence of turbulent flow properties on this phenomenon.

How to cite: Obligado, M. and Gires, A.: Rainfall Dynamics in Wind Energy Scenarios, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19070, https://doi.org/10.5194/egusphere-egu25-19070, 2025.

Land degradation is a critical challenge to sustainable development, impacting ecosystems, economies, and communities globally. As part of the FAIR-EASE Earth Critical Zone (ECZ) pilot, this study develops a tailored Land Degradation Assessment tool based on the Trends.Earth approach. The tool aims to enhance data accessibility, integration, and usability across environmental domains, supporting decision-making and policy frameworks aligned with the United Nations Sustainable Development Goals (SDGs).
Building upon the robust Trends.Earth implementation, we can integrate customized workflows and datasets to reflect regional variability in degradation indicators, including vegetation productivity, soil health, and land cover changes. Our approach prioritizes FAIR (Findable, Accessible, Interoperable, and Reusable) principles to ensure broad usability and collaboration across scientific and policy communities.
Preliminary results demonstrate the tool's capacity to enhace the detail of the analysis and to identify degradation hotspots. Furthermore, the integration of open-source geospatial tools and standards supports a scalable framework applicable to diverse environmental contexts.
The tool is designed to be embedded within the LandSupport platform, a geospatial decision support system, further enhancing its accessibility and integration into decision-making processes for land management.
This work contributes to advancing interdomain digital services and illustrates the potential of FAIR principles in addressing complex environmental challenges. We invite feedback from the community to refine, expand and customise the tool's application, fostering collaboration for sustainable land management.

How to cite: Mauriello, I. E., Langella, G., Terribile, F., and Miralto, M.: Customizing Trends.Earth for land degradation assessment in the earth critical zone: a FAIR-EASE approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19225, https://doi.org/10.5194/egusphere-egu25-19225, 2025.

Over thirty years ago, Y. Kagan speculated that seismicity could fruitfully be considered as “the turbulence of solids”.  Indeed, fluid turbulence and seismicity have many common features: they are both highly nonlinear with huge numbers of degrees of freedom.  Beyond that, Kagan recognized that they are both riddled with scaling laws in space and in time as well as displaying power law extreme variability and – we could add – multifractal statistics.

Kagan was referring to seismicity as usually conceived, as a sudden rupture process  occurring over very short time periods.  We argue that even at one million year time scales, that the movement of tectonic plates is “quake-like” and is quantitatively close to seismicity, in spite of being caused by relatively smooth mantle convection. 

To demonstrate this, we develop a multifractal model grounded in convection theory and the analysis of the GPlates data-base of 1000 point trajectories over the last 200 Myrs.  We analyzed the statistics of the dynamically important vector velocity differences where Dr is the great circle distance between two points and Dt is the corresponding time lag.  The longitudinal and transverse velocity components were analysed separately.  The longitudinal scaling of the mean longitudinal difference follows the scaling law <Δv(Δr)> ≈ Δr^H with empirical H close to the mantle convection theory value  H = 1.  This high value implies that  mean fluctuations vary smoothly with distance.  Yet at the same time,  the intermittency exponent C₁ is extremely high (C₁ ≈ 0.55) implying that from time to time there are enormous “jumps” in velocity: “Plate quakes”.  For comparison, laminar (nonturbulent) flow has H = 1 but is not intermittent (C₁ = 0), whereas fully developed isotropic fluid turbulence has the (less smooth) value H = 1/3 (Kolmolgorov) but with non-negligible intermittency C₁ ≈ 0.07 and seismicity has very large C₁ ≈ 1.3.  Our study thus quantitatively shows how smooth fluid-like behaviour for the longitudinal velocity component can co-exist with highly intermittent quake-like behaviour.

Whereas the longitudinal component is well modelled by (highly intermittent) convection, the transverse velocity is well modelled by Brownian motion.  In the temporal domain both components (including their strong correlations) display such diffusion behaviour (i.e. with classical exponent H = ½), but are highly intermittent (C_1time = C_1space/2 ≈ 0.27).  Finally, the extreme velocity differences (that appear as occasional spikes in the velocities) have power law probability tails; the “Guttenberg-Richter” exponents in the seismology literature.

The advection - diffusion model is based on an underlying multifractal space-time cascade process.  Using mantle convection theory, we show how the driving multifractal flux (ψ) is related to vertical heat fluxes, expansion coefficients, densities, viscosities and specific heats. Taking typical values predict driving fluxes very close to the observed mean <ψ> ≈ 1/(400 Myrs).  Trace moment analysis shows that the outer space-time scales of the cascade process are ≈17000 km in space and ≈ 50Myrs in time.   Whereas the former corresponds to half the Earth’s circumference, the latter is the typical time required for a plate to randomly “walk” the same distance.

How to cite: Lovejoy, S., Spiridonov, A., and Balakauskas, L.: The turbulence of solids: a multifractal plate tectonic model with Guttenberg-Richter plate “quakes” , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20653, https://doi.org/10.5194/egusphere-egu25-20653, 2025.

Climate change may influence energy demand, with shifts in energy needs not only altering the energy structure but also posing challenges to the sustainability and resilience of energy systems. These impacts could further complicate the feasibility of achieving decarbonization goals. Residential energy sector is a critical component of global energy consumption. As temperature fluctuates and weather variability intensifies, households will adapt energy use to maintain comfortable living conditions. Energy consumption may increase due to climate change, but the magnitude remains uncertain. Considering various income groups around the world, residents may react to climate change heterogeneously.

Traditionally, some models use Heating Degree Days (HDD) and Cooling Degree Days (CDD) to serve as index of temperature change, which are often calculated by formulas below, where i means gridded cell, j means region,  means daily temperature, and represents comfortable temperature,  means population. First, calculate gridded HDD/CDDs as the difference between daily temperature and comfortable temperature. Then aggregate the gridded daily HDD/CDDs to region.

  (1)

   (2)

                                      (3)

However, calculation for HDD/CDDs still have several aspects that could be further improved. First, most temperature data used are predicted on SRES, and HDD/CDDs are assumed to be constant, so HDD/CDDs need to be updated to better reflect future climate change. Second, previous calculation always neglects the impact of crucial factors such as GDP when aggregating gridded temperature difference to regional level, only considering population distributional effects. Third, the difference resulted from income and climate also should be considered, for rich residents can afford more energy consumption, and long-term climate also impact response of people when faced with climate change.

Considering potential shortcomings mentioned above, we update the global HDD/CDDs of 32 regions in Global Change Analysis Model (GCAM). First, we use daily temperature data predicted by four climate models under different Shared Socioeconomic Pathways (SSPs) combining Representative Concentration Pathways (RCPs) scenarios from the Coupled Model Intercomparison Project Phase 6 (CMIP6), thus bridging the gap between climate model and GCAM. For in GCAM climate input module, HDD/CDDs are calculated based on historical climate data and are lack of fine-scale calculation. Second, our calculation adopts two weighting methods considering influence of population and GDP distribution on residential energy demand respectively. Third, beyond global-scale calculation, we refine calculation for China to the provincial level.

Fig. 1 Research Framework

Based on the updated HDD/CDDs, we use GCAM to analyze how climate change impact residential energy demand, aiming to provide scientific support for formulating policies that address the challenges posed by climate change to energy system. Our analysis offers comprehensive insights into residential energy demand change under SSPs and RCPs scenarios, accounting for income heterogeneity. These findings are informative to design effective mitigation policies in the context of climate change.

How to cite: Zhu, M., Zhao, M., Zhu, R., Mei, F., and Ou, Y.: Effects of Climate Change on Residential Energy Structure, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-503, https://doi.org/10.5194/egusphere-egu25-503, 2025.

Digital city contributes to improving resource allocation efficiency and quality of life, with a key component being the division of functional areas. This division directly influences the optimal allocation of urban spatial resources and the efficient operation of various services. However, a mismatch exists between virtual and physical city functions. For instance, many office activities do not occur in physical office spaces. In this study, Guangzhou is taken as the research area to quantify this mismatch in office spaces, utilizing mobile signaling data and POI (Point of Interest) data, and analyzing the factors influencing the mismatch. Mobile signaling data, combined with office software usage records, reveals the precise locations of office activities from a virtual perspective. POI data provides detailed records of physical office locations, while also encompassing other potential locations where office activities may take place. This study reveals the spatial characteristics and influencing factors of this mismatch in Guangzhou, provides scientific support for the development of digital cities.

How to cite: Jiang, H.: Quantifying spatial mismatch between virtual and physical office spaces, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1372, https://doi.org/10.5194/egusphere-egu25-1372, 2025.

The Vulcan Project version 4.0 emissions data product has generated all fossil fuel CO2 emissions across the US landscape, every hour, from 2010-2022 down to the scale of neighborhoods. From this complex landscape, we have extracted FFCO2 emissions for every urban area, following multiple commonly used urban definitions. The information extracted includes both Scope 1 and Scope 2 emissions with a wide array of “functional” attributes such as sector, fuel, vehicle class, building class, road class, and industrial sub-sector. Here, we analyze ~4000 US cities in terms of their size scaling properties. In particular, urban scaling properties provide novel insight into emergent properties such as the relationship between urban metabolism and urban size properties. This relationship varies by region and is indicative of the relationship between urban form and economies of scale including implications for infrastructural development and urban sprawl.

How to cite: Gurney, K., Dass, P., Lobo, J., and Shutters, S.: Scaling Properties of Carbon Emissions in US Cities: Bigger is Better, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3332, https://doi.org/10.5194/egusphere-egu25-3332, 2025.

Urban geo-scenes (UGS) are an abstraction of the basic units of cities. Understanding and functional recognition of UGS is crucial to balancing and optimizing urban spatial layout, rationally allocating urban resources, and enhancing urban resilience and vitality. To construct UGS, urban geo-objects (UGO) e.g., derived from remote sensing must be combined with semantic information, which has seldom be done so far.  Consequently, this study designed a UGS recognition framework based on multimodal deep learning. First, we use very high-resolution satellite data to derive UGOs. Second, the self-built SE-DenseNet branch is used to mine deep physical visual features and social semantics from satellite image data and auxiliary data (POI, building footprints from UGOs). Finally, we build an urban fabric graph model to mine spatial semantics between UGOs.  In addition, a spatial semantic fusion module is introduced for the collaboration and interaction of multi-modal and multi-scale features. We evaluate the effectiveness of the proposed framework in the complex Beijing and Shenzhen regions of China. The overall accuracy is 91.35% and 90.24% respectively, which is higher than the state-of-the-art multimodal methods. In addition, our study also emphasizes the key role of spatial relationships and distribution patterns of UGO in UGS recognition, and the addition of POIs and building heights improves the recognition accuracy. The multimodal UGS recognition framework based on urban fabric can more effectively understand urban functions, thereby achieving urban planning and management.

How to cite: Bao, H., Zhou, L., and Lehnert, L.: Understanding and recognition of geo-scenes based on multimodal spatial semantics to monitor complex urban systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4066, https://doi.org/10.5194/egusphere-egu25-4066, 2025.

The fast-evolving nature of urbanization and its complex patterns require precise and interpretable machine learning models to effectively predict urban growth. To address this challenge, this study introduces a novel framework combining Hybrid Intelligence and Explainable AI (XAI), specifically Shapley Additive Explanations (SHAP) to improve model performance, robustness, and transparency. Using a weighted ensemble technique, the proposed method systemically integrates linear, tree-based, and neural network models to propose a hybrid of Elastic Net, XGBoost, and Wide & Deep Neural Network (EN-XGB-WDN) frameworks for urban growth prediction. The methodology follows a multistep approach and includes the development of the hybrid model, its evaluation for binary classification, integration of SHAP-based feature analysis to identify key drivers of urban growth and improve model interpretability, retraining of the hybrid model to increase accuracy and reduce overfitting, and validation of the proposed framework using standard evaluation metrics including accuracy, precision, recall, F1 score, and AUC. The hybrid model achieves an overall accuracy of 87.34%, a weighted F1-score of 87.18%, and an AUC of 0.9442. The SHAP analysis revealed that Drive Time (DT), Distance from Roads (DfR), and Elevation are the most impactful features to understand the dynamics of urban growth. The findings revealed how variations in specific features, such as higher DT and lower DfR, significantly affect urban growth probabilities. The hybrid model also categorized urban growth probabilities into five classes: very low (40.62%), low (23.27%), moderate (15.38%), high (12.10%), and very high (8.63%), revealing spatial patterns of urban expansion. The framework combines hybrid ensemble methods with SHAP-based explanations to significantly enhance the predictive and explanatory power of urban growth models compared to the limitations of traditional approaches. This study highlights the efficiency of integrating hybrid machine learning and Explainable AI to understand and predict complex urbanization dynamics. The outcomes offer actionable insights for policymakers and urban planners, facilitating data-driven strategies for sustainable urban development. This research demonstrates the effectiveness of hybrid intelligence coupled with Explainable AI, offering a scalable and interpretable framework to better understand and predict urbanization patterns.

How to cite: Khan, D. and Khan, N.: Hybrid Intelligence and Explainable AI for Urban Growth Prediction Modelling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4759, https://doi.org/10.5194/egusphere-egu25-4759, 2025.

Urban geometry plays a key role in determining the urban climate through its complex shading and trapping effects on solar radiation. As a result, urban albedo is typically lower than rural albedo, suggesting a larger solar heat gain in urban areas. As cities grow larger and more heterogeneous in geometry, quantifying the impact of this variation on albedo at the city scale requires computationally efficient models that can also resolve the 3D geometry of real cities. To this end, we developed a simplified 3D urban radiation model and used it to examine the variations in albedo due to heterogeneous geometry in the city of Shanghai. The model reduces computational complexity from O(n²) to O(n) while maintaining an accuracy within 5% compared to traditional 3D models. The case study in Shanghai shows that albedo has a linear relationship with building height but varies nonlinearly with changes in building density. The lowest albedo occurs when the building density (λ_p) is around 0.2 and the building height-to-length (H/L) ratio is 6, while occurs at λ_p > 0.3 with H/L = 1. This suggests that optimizing building geometry could improve the urban climate and potentially being used to increase the utilization of solar energy.

How to cite: Zhou, H. and Wang, K.: Development of simplified 3D urban radiation model to examine the variations of albedo due to heterogenous geometry in the city of Shanghai, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4805, https://doi.org/10.5194/egusphere-egu25-4805, 2025.

Urban heat storage (Qs) is an essential component of urban surface energy balance. Urban with 3D structure has larger surface area than rural and urban would absorb and release more energy than rural. Qs is the main factor for urban heat island (UHI) at nighttime. The quantitative contribution of Qs to UHI is still unclear, due to the lack of a spatio-temporal continuous Qs dataset. In this study, firstly, we developed an urban surface thermal inertia model considering diurnal variation of surface temperature (LST) using hourly LST of Himawari-8. Secondly, the hourly Qs at 2-km resolution in three urban agglomerations in China was simulated by a half-order time derivative method which derived from combining the one-dimensional heat diffusion equation and Fourier’s law for heat conduction, using the urban thermal inertia model and hourly Himawari-8 LST. Thirdly, the relationship between Qs and air temperature (Ta) was studied at different time scales (day and nighttime, four seasons) and different LCZs (local climate zones). The Ta was derived from the interpolation of dense automatic weather stations with more than 10000 sites in China. Finally, some urban heat mitigation measures were provided based on the above analysis. Based on the in-situ observation, the accuracy of urban thermal inertial in this study was higher than other model, RMSE, MAE, R² were improved from 4.65 K, 3.58 K and 0.88 to 1.86 K, 1.53 K and 0.97. In addition, the simulated Qs were validated by the observed Qs (the minus of net radiation, sensible and latent heat flux from in-situ flux tower, and anthropogenic heat flux simulation) in Beijing, Shanghai and Guangzhou, R² could be up to 0.92. The results showed that, Qs was more consistent with Ta at nighttime than daytime, with R² of 0.96 and 0.1, respectively. That showed that Qs is the main factor for nighttime UHI in this study area. During nighttime, the high-rise building has higher Ta than low-rise building, due to higher Qs and release more energy than low-rise. In natural surfaces, water has larger Qs and higher Ta than dense trees. The loop (between hourly Qs and hourly Ta) shape were different at different LCZs, with different loop area and loop slope. Based on the loop area and slope, we found that high-rise building had higher UHI but varied quickly, however, low-rise UHI is lower but would last longer. The water surface in nighttime is also heat source and has a longer time UHI. Therefore, the high-rise building and water surface are not conductive to alleviating the nighttime UHI.

How to cite: Li, N., Guo, F., Dou, J., Ma, Y., and Miao, S.: Remote sensing-driven analysis of hourly urban heat storage and its effects on urban heat islands in China, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5462, https://doi.org/10.5194/egusphere-egu25-5462, 2025.

China has undergone rapid urbanization in terms of population and land use in recent years. However, there are notable lags in "people-oriented" dimensions of urbanization, including urban social services, environmental sustainability, and equity. Here, considering the complex interactions of sub-components of urbanizations, we examined 16 "people-oriented" urbanization indicators across four dimensions - economic, social, environmental, and equity dimensions - from 2005 to 2020. Using methods such as paired t-tests and the evenness measurement, we analyzed and identified the dynamic relationships between these 16 indicators with population/land urbanization at multiple scales, including national, regional, urban agglomeration, and different city sizes. We found that between 2005 and 2020, China's urbanization indicators showed an overall upward trend, with changes ranging from 1.09 to 53.95 times. Among "people-oriented" urbanization indicators, economic and social indicators lagged behind land and population urbanization, while environmental indicators took the lead. The evenness index among indicators showed a "U-shaped" change pattern. Particularly since the implementation of China's New-type Urbanization Plan in 2014, the evenness index among indicators gradually increased from 35.43 to 37.39 in 2020, representing a 6.9% improvement. Looking forward, it is necessary to strengthen investment in social service systems and implement placed-based coordination strategies to promote further development and balanced growth of "people-oriented" urbanization.

How to cite: Gu, T., Huang, Q., and Hou, Y.: Do people-oriented urbanization catch up with land and population urbanization in China？, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6392, https://doi.org/10.5194/egusphere-egu25-6392, 2025.

In recent decades, urban expansion across Europe has accelerated, driving the rapid growth of rural-urban interfaces. The increasingly complex and dynamic nature of territorial transitions calls for the timely development of classification systems designed to systematically organize areas along a spectrum, from distinctly urban to distinctly rural. Current classifications often rely on predefined criteria, such as population size and density, which may not fully capture the nuanced and evolving nature of transitions driven by the complex interplay of socioeconomic processes, demographic shifts, environmental factors, and dynamic geographic forces.

This research addresses existing gaps by employing modern data-driven approaches, including machine learning and clustering techniques, to develop adaptive typologies that integrate diverse demographic, socioeconomic, and environmental variables. Using Switzerland as a case study, the proposed methodology offers a dynamic and scalable framework for territorial classification, supporting the effective management of territorial transitions and landscape conservation in Alpine regions. The analysis leverages a multidimensional dataset derived from the 2020 official census, incorporating 18 variables that encompass demographic profiles, socio-economic, and the physical space characteristics.

We used Self-Organizing Map (SOM) combined with hierarchical clustering. SOM, a type of competitive learning neural network, reduces the complexity of high-dimensional data by mapping it onto a two-dimensional grid of neurons. Visual outputs, such as heatmaps, enhance the interpretation of trends and patterns, providing a clearer understanding of variables distributions and interrelationships. Afterward, the SOM output grid of neurons was aggregated into six distinct clusters, which were mapped onto the geographical space. This produced a visual representation of the spatial organization of territorial typologies along the rural-urban continuum in Switzerland at a detailed municipal level.

The data-driven clustering approach developed in this study proved effective in capturing the complex and diverse nature of Swiss territorial typologies. The key findings reveal a landscape marked by a complex rural-urban interface, extensive intermediate zones, and significant spatial fragmentation. These final six territorial typologies could be characterized as follows: urban centres, representing the main hubs at the highest level of the Swiss urban hierarchy; suburban areas, located near and well-connected to urban centres; two peri-urban areas, distinguished into aging-rural areas and rural-urban edge; rural-forest areas, situated at medium to high elevations, featuring a forested landscape and rural settings; unproductive areas, encompassing high-altitude regions and including critical Alpine glaciers.

How to cite: Tonini, M., Yu, J., and Hagen-Zanker, A.: Exploring Switzerland's Rural-Urban Continuum Through Unsupervised Learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6870, https://doi.org/10.5194/egusphere-egu25-6870, 2025.

Adequate sunlight exposure is crucial for human wellbeing, yet its accessibility in cities is significantly compromised by both could cover and complex three-dimensional (3D) urban structure. Here we adopted an analytical framework that integrated natural day length variations, cloud cover effects, and 3D urban structure to quantify actual sunlight duration in urban areas. By using high-resolution satellite products, fine-scale canopy height data, and detailed 3D building footprints, we mapped the spatiotemporal patterns of sunlight availability and quantified the relative contributions of cloud cover and urban structures on the loss of sunlight for Chinese cities. Our analysis reveals pronounced spatial disparities and trends in urban sunlight resources in China, underscoring the urgent need for evidence-based urban planning strategies that optimize natural light accessibility for sustainable urban development.

How to cite: Wu, X. and Chen, B.: Quantifying urban sunlight accessibility across Chinese cities: Impacts from cloud cover and three-dimensional (3D) urban structure, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7638, https://doi.org/10.5194/egusphere-egu25-7638, 2025.

Over the past few decades, the rapid growth of cities has evolved into a significant social, demographic, and architectural phenomenon, highlighting the vital importance of urban planning in fostering sustainable development. In this context, machine learning has emerged as a game-changing discipline, utilizing advanced algorithms to reshape traditional approaches to urban data management and analysis.This study combines Geographic Information Systems (GIS), Deep Learning techniques, and verified data from the General Directorate of the Spanish Cadastre to perform a comprehensive analysis of the urban environment through façade images in Murcia, one of Spain’s most dynamic metropolitan areas.Leveraging the clustering analysis of the studied variables, an automated binary classification model for façade images was developed using the pretrained EfficientNetB0 architecture in Python. To enhance interpretability, heat maps were generated to visualize the regions the model focuses on during classification. These heat maps reveal the critical features of the facades that guide the model’s decisions, providing valuable insights into the key factor influencing the classification process.The results were integrated into ArcGIS PRO, using the cadastral reference of the properties as a key attribute for a detailed spatial analysis. This approach revealed two significant areas linked to the metropolitan growth of Murcia, laying a strong foundation for future urban studies in the region.

Funding: Twin-ER: Earthquake Risk Pilot Digital Twin. Grant PID2023-149468NB-I00, funded by MCIU/AEI/10.13039/501100011033 and FEDER/EU

How to cite: De La Cruz Luis, M. I., Martinez Cuevas, S., Garcia Aranda, C., Morillo Balsera, M. C., and Poveda Lorente, E.-M.: Advancing Urban Environment Studies in Murcia, Spain through an Automated Façade Image Classification Model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7834, https://doi.org/10.5194/egusphere-egu25-7834, 2025.

Equitable access to daily necessities and services is crucial for enhancing human quality of life and is integral to achieving the United Nations’ Sustainable Development Goals. However, knowledge about global access to these essential resources remains limited and fragmented, primarily due to the absence of a comprehensive infrastructure inventory and scalable measures of accessibility. Here we compiled the most extensive global database of points of interest (POI) to represent six essential infrastructure categories—living, healthcare, education, entertainment, public transit, and work. We used refined 30-meter resolution friction surface data to map travel times to these critical infrastructures as a proxy for accessibility across the urban-rural continuum and assessed disparities across geographic, urbanization, and socio-economic contexts. Our results reveal that access to daily necessities and services is unevenly distributed in terms of total infrastructure, per capita availability, and travel time. Globally, only 38.7% (2.6 billion people) and 50.7 % (3.4 billion people) of the population resides within a 15-minute and 30-minute walking distance of essential daily necessities and services, respectively. These results highlight the urgent need to optimize strategies for planning, allocation, and management of critical infrastructure to promote inclusive and sustainable development.

How to cite: Chen, B., Wu, S., Nelson, A., and Gong, P.: Measuring global human access to essential daily necessities and services, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8416, https://doi.org/10.5194/egusphere-egu25-8416, 2025.

Evaluating human interaction with environmental health determinants in space and time is fundamental to estimate personal environmental exposures. The increasing demand in an exposure assessment of entire populations requires to combine environmental variables at high resolution on large spatial extent, e.g. at nationwide or continental scale, with the space-time activity pattern of each individual in a study population.

Modelling population health and citizens' exposures is a complex process involving multiple procedural steps. One major step is to generate spatio-temporal information on environmental factors, either considered as beneficial for human wellbeing, for example, accessibility to green space or blue space, or considered as having negative health impacts such as the existence of air pollution, noise or heat. To capture the spatial variability these datasets need to be generated at high resolution. To allow for studies comparing cities, regions or countries, a geographical extent of subnational or larger size is required. In addition, data can be temporal to cover diurnal or seasonal variation of an environmental variable. Another major step is to use the environmental factors to as input to models calculating exposures for entire study populations, ranging from a few hundred participants up to millions of citizen. Here, socio-economic variables, mobility, different travel modes, and other daily activities with accompanying location changes need to be considered to mimic the space-time paths of each participant of a study population. These tasks require sufficient flexibility in both constructing environmental models as well as executing those eventually on HPC systems to break computational barriers of common workstations.

We present a computational framework for implementing both procedural steps and show the development of two European scale raster maps on a 25m grid and their subsequent usage to estimate human exposures to greenness visibility and noise. The maps were created with LUE (https://lue.computationalgeography.org/), an open-source modelling framework providing a Python package with currently 115 general-purpose operations for the construction of spatio-temporal simulation models. We implemented two custom focal operations that make use of the LUE framework. The first focal operation calculates for each raster cell the visible green area within a particular buffer size (c.f. Labib 2021, https://doi.org/10.1016/j.scitotenv.2020.143050). The second focal operation aggregates traffic-related noise within a particular buffer size, considering attenuation due to geometric divergence, atmospheric absorption, ground effects and diffraction.

We calculated visible green within a radius of 800m and noise within 1500m radius using 768 CPUs on eight HPC cluster nodes, and then used Campo (https://campo.computationalgeography.org/) for activity-based exposure assessment. The obtained exposure estimates can show considerable differences for different typical human activity patterns, such as homemaker or commuter, as well as a high spatial variability.

How to cite: Schmitz, O., de Jong, K., Shen, Y., and Karssenberg, D.: Assessing human exposures to environmental risk factors at continental-scale: accounting for short range variation in environmental factors and human activity patterns, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8801, https://doi.org/10.5194/egusphere-egu25-8801, 2025.

Anthropogenic heat generated by building energy use contributes to the urban island and climate change. Quantifying high spatiotemporal resolution city scale building energy use (BEU) and anthropogenic heat emission (AHE) is necessary for understanding urban microclimate and sustainable development. However, the current shortage of such data is insufficient to support urban energy management and climate decision-making. We estimated BEU and AHE from buildings in Hong Kong using a GIS-based city-scale building energy model (GIS-CBEM) and investigated their spatiotemporal variations. First, all buildings were categorized into 11 types, and a prototype was developed for each type. These prototypes were then calibrated using annual building energy consumption data from surveys. We studied the energy use profile for each building prototypes under the Typical Meteorological Year (TMY) weather data. Then, we estimated hourly BEU and AHE for all buildings in Hong Kong at the individual building level. The study results unveiled the spatiotemporal variation of buildings in Hong Kong at high resolution and detected divergent structure of building end-use and fuel use for different building prototypes. We found that the total BEU of all buildings in Hong Kong peaked at 5.1 × 10⁹ kWh in August, with 36.7% from HAVC system, while the lowest BEU was found in February at 3.5 ×10⁹kWh, with 14.1% from HAVC system. Total AHE from all buildings reached a maximum of 8.1 × 10⁹ kWh in July and minimum of 4.1 × 10⁹ kWh in February. Our findings have critical significance in enhancing energy efficiency, reducing environmental impact, and promoting sustainable development.

How to cite: Liu, Q. and Zhou, Y.: Unveiling Spatial and Temporal Variations of Building Energy Use and Anthropogenic Heat Emissions in Hong Kong, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9314, https://doi.org/10.5194/egusphere-egu25-9314, 2025.

Traffic congestion continues to challenge urban development, yet most research emphasizes large-scale factors such as road layouts and land use, overlooking localized visual aspects encountered by drivers. This study employs geographically weighted random forest, a non-linear and spatially explicit method, to explore how localized visual features—such as vehicle density, building structures, greenery, and road conditions—impact traffic congestion in Chicago. By integrating transport network dynamics with visual streetscape characteristics, the geographically weighted random forest approach captures spatial heterogeneity and complex interactions more effectively than traditional models. Results demonstrate that incorporating these multi-scale features improves model fit, revealing that greenery mitigates congestion, while dense urban structures and vehicle clusters exacerbate delays. These results highlight the potential of integrating visual characteristics of streetscapes into urban strategies to address congestion more effectively.

How to cite: Xu, M. and Weng, Q.: Modeling the Spatial Dynamics of Traffic Congestion Through Street-Level Visual Features: Evidence from Street View Images in Chicago, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9394, https://doi.org/10.5194/egusphere-egu25-9394, 2025.

Exploring the effect of building morphology on Land Surface Temperature (LST) has received surging attention. In this process, a fundamental precondition is selecting an appropriate spatial statistical unit to calculate building morphological indices and corresponding LST. However, different units lead to divergent results, indicating they inevitably suffer from the Modifiable Areal Unit Problem (MAUP), which brings large uncertainties. This study places special emphasis on proposing a new spatial unit, the Homogenous Unit of Building Morphology (HUBM), to re-describe building morphology and re-analyze its effect on LST with less uncertainty. Results show: (1) building morphology portrayed by HUBM maintains more spatial characteristics and remains relatively stable across scales, which is more consistent with the realistic building environment. (2) The relationship identified by HUBM shows building morphology is not strongly correlated with LST in essence and is regarded as more authentic due to the more objective portrayal of building morphology, while this relationship may be overestimated by previous common units. (3) The effect of building morphology on LST explored by HUBM also remains relatively stable across different scales (R² fluctuation amplitude of 0.08, 0.12, and 0.08 in the spring, summer, and winter, respectively) compared to regular grids (R² fluctuation amplitude of 0.18, 0.2, and 0.2), effectively alleviating the uncertainty associated with the MAUP. These findings provide new insights into re-examining the authentic effect of building morphology on LST, assisting in addressing urban heat island effects and promoting sustainable urban development. Moreover, HUBM can be applicable to other urban issues for mitigating MAUP.

How to cite: Yang, L., Yang, X., and Li, S.:  Is 3D building morphology really related to land surface temperature? Insights from a new homogeneous unit, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10149, https://doi.org/10.5194/egusphere-egu25-10149, 2025.

With rapid urbanization, cities face critical sustainability challenges, including poverty, resource shortages, pollution, and climate impacts. The EU Cities Mission supports 112 cities in developing Climate City Contracts to achieve climate neutrality by 2030 [1], emphasizing strategic, cross-sectoral approaches and stakeholder collaboration. This study introduces a systematic and indicator-based assessment of urban resilience, utilizing EU-sourced environmental, OpenStreetMap, and a few nationally sourced data. The methodology incorporates 12 key indicators, mapped at high resolution for 83 Italian cities using open-source GIS software [3], ensuring full reproducibility and applicability to other European cities. The indicators are categorized into five classes:

(i) nature and biodiversity, including forest canopy coverage, native habitat areas, biodiversity, geodiversity [4], ecological corridors, and heat island effects [5];

(ii) natural hazards, including susceptibility to flooding, earthquakes, wildfires, and landslides [6];

(iii) air pollution, including concentration of PM_2.5 and NO₂;

(iv) transport, including availability of sustainable and affordable transport systems;

(v) social indicators, including population living in close proximity to green spaces or water sources, and public services.

This study evaluates the current state of Italian cities [7], identifies regional differences, and highlights the strengths and weaknesses of each city individually, based on results provided by the urban indicators.

The software developed for this study is flexible, as the input data exists for the whole of Europe and it is easily extensible with modular scripts, to include additional indicators. The scripts processes data to produce spatially distributed results (raster maps) for each indicator in each class listed above and then summarize each indicator with a numerical figure.

Preliminary findings suggest significant regional variation in factors contributing to climate resilience and citizen well-being [8]. Cities in Northern Italy exhibit larger green space coverage but also higher air pollution levels. In contrast, Central Italy stands out for its high species biodiversity and geodiversity. Moreover, results uncover regional spatial patterns, offering actionable insights for policymakers to design locally informed and effective strategies. The findings contribute to advancing sustainability goals, supporting urban transformations toward enhanced resilience and reduced environmental impact. A comprehensive set of urban indicators, including those derived in this study and summarized into a single numerical output for each category, allows ranking of cities and promoting the adoption of data-driven strategies for sustainable development.

References

[1] United Nations (2023) https://sdgs.un.org/goals/goal11

[2] Sarretta et al., Int. Arch. Ph. Rem. Sens. Spat. Inf. Sci. (2021) https://doi.org/10.5194/isprs-archives-XLVI-4-W2-2021-159-2021.

[3] Neteler et al., Env. Mod. Softw. 31 (2012) https://doi.org/10.1016/j.envsoft.2011.11.014

[4] Burnelli et al., Geomorphology 471 (2024) https://doi.org/10.1016/j.geomorph.2024.109532

[5] Morabito et al., Sci. Tot. Env. (2021) https://doi.org/10.1016/j.scitotenv.2020.142334

[6] Loche et al., Earth-Science Reviews 232 (2022) https://doi.org/10.1016/j.earscirev.2022.104125

[7] Alvioli, Land. Urb. Plan. 204 (2020) https://doi.org/10.1016/j.landurbplan.2020.103906

[8] Boeing et al., Lancet Global Health 10 (2022) https://doi.org/10.1016/S2214-109X(22)00072-9

How to cite: Tarasova, E. and Alvioli, M.: Measuring resilience of urban areas using public data and a reproducible approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11618, https://doi.org/10.5194/egusphere-egu25-11618, 2025.

Due to climate change, new challenges arise in drinking water infrastructure planning and in the re-assessment of well-established urban drinking water utilities. We observed temperatures exceeding 25 °C in drinking water supply pipes, which pose a health threat and water quality problem, as these temperatures are favorable for microbial growth.

We set out to predict temperatures in drinking water supply networks. The key step to achieve this goal is to monitor and model soil temperatures and soil moisture derived from meteorological forcing functions. With meteorological observations and soil material properties, we describe the heat transport and water flow from the ground surface into the subsurface and from there into the pipes and with the water in the pipes.

In order to achieve this goal, we solved the heat and water balances jointly at the atmosphere-subsurface interface, using the open-source numerical simulation framework DuMuX. We were able to do this because of the available meteorological observations (e.g., radiation balance, precipitation intensity) next to the newly installed pipes. These balances provide a novel interface condition for heat transport and water flow modelling. We coupled the heat transport through the drinking water pipe walls to the drinking water in the pipes and to the subsurface transport processes.

At a pilot site, we installed typical drinking water pipes (PE and cast iron), backfilled with known material (typical gravelly conditions below roads and naturally existing sandy clay), and applied land-cover (asphalt and natural vegetation). We were able to reproduce the joint measurements of temperatures and soil moisture under various conditions (well-draining gravel vs. less-draining clayey material; vegetation vs. asphalt).

In this presentation, we demonstrate results of the multi-year measurement campaign, the results of 1D and 2D subsurface heat transport models coupled to dynamic hydraulic conditions in the drinking water pipes, and an innovative surrogate-based Bayesian active learning-assisted model calibration methodology.

This work presents an important first step towards predicting temperatures in drinking water supply pipes and will be directly relevant for chemical and biological processes that occur in non-isothermal conditions (e.g., due to climate change), for example, in relation to contaminant remediation. Our results are of relevance for drinking water supply companies, shallow geothermal design, and urban planning.

How to cite: Haslauer, C., Kroeker, I., Nißler, E., Oladyshkin, S., Nowak, W., Class, H., and Osmancevic, E.:   Large Temperatures in Water Distribution Pipes as a Water Quality Threat: Measurements and Modelling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13131, https://doi.org/10.5194/egusphere-egu25-13131, 2025.

Urban climate modelling tools like the Surface Urban Energy and Water balance Scheme (SUEWS) are indispensable for investigating complex surface–atmosphere interactions and guiding urban adaptation strategies. However, these models often present substantial barriers to use: they require extensive technical know-how, involve intricate input datasets, and can be time-consuming to set up and interpret. Recent advancements in Large Language Models (LLMs) hold promise for bridging this gap by transforming complex domain-specific tasks—such as data validation, simulation setup, and error diagnosis—into user-friendly interactive experiences.

In this study, we propose a novel workflow that leverages LLM capabilities—such as generative text, code suggestion, and context-driven troubleshooting—to streamline SUEWS usage and improve accessibility for researchers and practitioners:

• Automated Model Configuration
  We explore the use of LLM-guided prompts to generate properly formatted SUEWS input files, such as specifying hourly meteorological forcing data (e.g., temperature, wind speed, and humidity) or land cover fractions required for accurate simulations. By conversing with the model about location, time range, and data availability, users can rapidly produce consistent and error-checked setup files, reducing manual edits that often lead to inconsistencies.

• Interactive Error Diagnosis
  LLMs can parse error logs and suggest potential solutions in real time. For example, if SUEWS outputs an error related to missing albedo values for a specific land cover type, the LLM can pinpoint the source of the issue and suggest default values or a method for calculation based on site-specific conditions. For example, if a runtime error indicates a mismatch in the date format of meteorological input data, the LLM can identify the exact line causing the error, recommend the correct format, and provide a command or script snippet to rectify the issue. Through iterative dialogue, the model clarifies the root causes of typical setup or runtime issues, explaining how to fix them without requiring the user to trawl through detailed documentation.

• Model Output Interpretation
  Interpreting large volumes of SUEWS output, such as energy balance components (net radiation, latent heat flux, and sensible heat flux) or water budget terms (runoff and evapotranspiration), can be daunting, especially for newcomers. LLMs can summarise key metrics—like energy flux partitioning and surface runoff patterns—and highlight discrepancies in data, thereby assisting in rapid analysis and scenario comparison.

Our findings indicate that an LLM-enabled approach substantially lowers the learning curve and operational overhead associated with SUEWS, while still maintaining scientific rigour. We piloted trial deployments in teaching and professional contexts, reporting improvements in both setup speed and user confidence. Future work includes refining the LLM’s domain-specific training to ensure physically consistent responses—such as maintaining energy balance across flux computations or ensuring water budget closure—and incorporating advanced visualisation plugins for immediate data interpretation.

By harnessing the dialogic strengths of LLMs, we aim to remove barriers to the complexity of urban climate modelling, ultimately broadening participation and fostering more informed decision-making in cities worldwide.

How to cite: Sun, T.: Remove Barriers to Accessible Urban Climate Modelling with Large Language Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13462, https://doi.org/10.5194/egusphere-egu25-13462, 2025.

The frequent occurrences of heat wave events and air pollution episodes have become pressing global concerns. Concurrent heat and ozone pollution events, in particular, have been widely documented across various regions and often result in more severe impacts compared to isolated stressors, leading to increased mortality and morbidity rates. However, our understanding of these compound events in urban environments, particularly their dynamics under different background climates and urban settings, remains limited. In this study, we systematically characterized the frequency, intensity, and duration of compound heat and ozone pollution events during warm seasons across all urban areas in the continental U.S. using long-term, high-resolution daily air pollution and air temperature datasets. Results suggest that urban heat waves, defined by daily maximum temperature, were more frequent, more intense, and longer lasting than their rural counterparts, primarily due to the urban heat island effect. In contrast, over half of the U.S. cities experienced fewer, less intense, and shorter ozone pollution episodes than surrounding rural environments. The spatially heterogeneous disparities in ozone pollution episodes among cities are mainly attributed to whether ozone production is limited by VOC or NOx, as revealed by time series analyses. Despite the overall decreasing trend of surface ozone concentrations during the last two decades, 89% of U.S. cities experienced more frequent compound heat and ozone pollution episodes than rural areas. Additionally, the cumulative heat and ozone intensities were higher in 91% and 88% of U.S. cities, respectively, than in their rural backgrounds. The duration of compound events tends to be shorter in urban areas. These findings highlight the dependence of such compound events on local and background conditions, emphasizing the need for locally tailored mitigation plans to reduce their impacts. This study also calls for detailed regional numerical simulations to elucidate the mechanisms driving these events.

How to cite: Wang, C., Hu, X.-M., and Leffel, J.: Compound heat and ozone pollution in urban areas, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13711, https://doi.org/10.5194/egusphere-egu25-13711, 2025.

Cities are highly interconnected networks of networks (referred to as the Urban Multiplex) that include the power grid and transportation networks, surface water and groundwater, sewerage and drinking water systems, inland navigation and dams – all intertwined with the natural environment and socioeconomic and public health sectors. While the Urban Multiplex is physically and functionally connected, the data produced within its individual sectors are not. This prevents us from fully understanding how the Urban Multiplex is connected, and how failures triggered by external stressors like floods cascade.  

Knowledge Networks are an AI technology that (i) integrates Urban Multiplex data, (ii) produces real-time flood forecasts across the continental U.S., (iii) serves as the foundation to evaluate the total impact of floods on cities, and (iv) supports queries at the nexus of water and energy. This talk will describe the development of the Urban Flooding and Water-Energy Nexus Open Knowledge Networks that aim to provide actionable answers to questions such as:

• Real-time flood mitigation and response: Will my neighborhood flood? Will I have access to water and power? Will this storm disrupt the power grid, drinking water treatment plant, or a bridge?

• Long-term design, planning and research: What is the total socioeconomic impact of this flood? Which critical urban infrastructure will likely fail in a future flood? Which failures will affect the most people or the most vulnerable people? Are there vulnerable communities downstream of this coal mine?

The interdisciplinary team behind this project has brought together academic researchers, industry, federal government, U.S. National labs and local stakeholders. It is funded by the U.S. National Science Foundation’s Convergence Accelerator Program that is structured to enable rapid advancement in highly complex problems of critical societal importance.

How to cite: Yeghiazarian, L. and the Knowledge Networks Team: Knowledge Networks help address urban flooding and water-energy challenges , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13868, https://doi.org/10.5194/egusphere-egu25-13868, 2025.

In recent decades, inequalities in economic, health, and education sectors have intensified spatial clustering of populations and resources, further reinforcing disparities within urban environment. Identifying these geographic boundaries is crucial for developing targeted policies to address inequality effectively. While traditional approaches to studying urban segregation rely primarily on socioeconomic indicators, this research introduces a novel methodology that combines subjective perceptions of the urban environment and objective characteristics of urban areas—such as land use and infrastructure—to identify distinct spatial clusters within Glasgow, a city with a varied socioeconomic landscape. Using MIT Place Pulse dataset of crowd-sourced streetscape perceptions, we developed a deep learning model to predict perception scores for new areas. These perception scores, along with image embeddings and land use information, enabled the geographic clustering of areas based on perceived and functional similarities. Our analysis reveals that perception-based boundaries often diverge from traditional census dissemination areas, suggesting that administrative boundaries may not fully capture the lived experiences of urban space. This research advances our understanding of urban inequality by demonstrating how perceived environmental qualities interact with physical infrastructure to shape distinct urban zones, providing policymakers with new tools for targeted intervention strategies.

How to cite: Pawar, S., Robertson, T., Marino, A., and Anderson, C.: Redefining Urban Clusters: Combining Subjective Perceptions and Objective Data to Map Inequality, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16365, https://doi.org/10.5194/egusphere-egu25-16365, 2025.

The rapid expansion of urban areas has led to significant environmental changes, most notably the Urban Heat Island (UHI) phenomenon, characterized by higher temperatures in urban areas compared to their rural counterparts. Addressing and mitigating UHIs is vital for public health, energy demand management, and enhancing urban livability, especially amidst global climate change. This study focuses on classifying Local Climate Zones (LCZs) in Taipei, Taiwan, using digital building data, satellite imagery, and urban morphological indices. LCZs offer a standardized framework to analyze urban morphology and its influence on local climates. By applying unsupervised clustering methods, we achieved a detailed classification of urban areas, enabling a data-driven exploration of their climatic and morphological characteristics.

To downscale and refine the analysis at the community level, Principal Component Analysis (PCA) was employed to reduce data dimensionality and extract key features such as building coverage, vegetation index, and sky view factor. K-means clustering was then used to categorize urban morphological types, resulting in distinct LCZs across Taipei. Our findings reveal significant differences in environmental variables among clusters. These results highlight how urban morphology, including building density and vegetation cover, impacts local climate conditions. The study also emphasizes the role of thermal comfort, underscoring the complex interplay between urban form and environmental factors.

This research demonstrates the effectiveness of unsupervised classification methods in identifying urban climate zones and provides a practical framework for urban planning and climate adaptation. By enabling targeted interventions, such as greening strategies or ventilation optimization, the study contributes to enhancing urban sustainability and resilience. The findings underscore the importance of interdisciplinary approaches to address the multifaceted challenges of urbanization and climate change.

How to cite: Chen, W.-J., Juang, J.-Y., and Chien, S.-S.: Self-Organizing Local Climate Zones by Using Integrated information in Urban Community – a case study in DaXue Village, Taipei, Taiwan, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17835, https://doi.org/10.5194/egusphere-egu25-17835, 2025.

Urban thermal environment is known to be strongly affected by the composition of urban land cover, with densely built-up areas characterised with distinctly higher temperatures than densely vegetated ones. These observations come from the analysis of relatively coarse land surface temperature (LST) satellite data and conclusions are typically derived for city districts or other variedly defined mapping units. Whilst these analyses provide useful insights to excess heat mitigation at city scales, these do not describe the nuance of the thermal response of the heterogenous urban form at local levels. This study investigated the relationship between LST of variedly configured immediate neighbourhoods of single patches of different land cover types (buildings, paved, grass, trees) extending from 0 to 100m away to determine the shape and type of urban features including water that contribute to the formation of cold and hot urban spaces. The study area comprised three English towns: Milton Keynes, Bedford, and Luton, collectively comprising a wide range of urban forms that are representative for England and other European towns located in the temperate climate zone. The analysis was carried out for two summer days a month apart, capturing the different thermal responses as temperatures rise over summer.  The microscale of the analysis was enabled by downscaled LST obtained from Landsat 8 thermal bands acquired at 100m resolution down to 2m, supported by high resolution spectral indices derived from very high resolution hyperspectral aerial imagery. Patch-level landscape metrics were used to describe the shape of the different patches of urban land cover derived from land cover map at 2m resolution. K-means analysis was used to determine groups of land cover patches of a given type with common thermal and spatial properties. Random forest regression algorithm was used to identify the important descriptors of LST for these groups and ANOVA analysis to determine statistically significant effects for various spatial configuration metrics. The findings suggested that the coldest patches of buildings, grass and paved were associated with highly aggregated patches of trees in the immediate neighbourhood, with PLADJ greater than 73 to 85% and COHESION greater than 93 to 97%, and buildings requiring somewhat lower aggregation levels than grass or paved. Hottest patches of these land cover classes were associated with PLADJ smaller than 63–69% and COHESION smaller than 83–87%, with elevation and distance to water being the most important factors, whose importance increased as the summer progressed. Overall, this study provided further insights into the spatial characteristics of patches of common land cover types in urban areas that contribute to the formation of particularly hot or cold urban spaces, which can facilitate the design of climate resilient cities.

How to cite: Zawadzka, J., Harris, J., and Corstanje, R.: The importance of spatial configuration of urban form in local temperature regulation investigated from very high resolution LST and land cover data and landscape metrics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19759, https://doi.org/10.5194/egusphere-egu25-19759, 2025.

The building stock accounts for 34% of global energy demand and 37% of CO₂ emissions related to energy and industrial processes. Additionally, the current increase in urbanization rates poses significant environmental challenges. Policy makers are becoming increasingly aware of these impacts, developing strategies aimed at improving energy efficiency and obtaining decarbonization of the built environment. Achieving these goals requires modeling actual building stock energy consumption patterns, future energy developing trends as well as the impact of energy retrofitting measures on CO₂ emissions. Urban Building Energy Models (UBEMs) and bottom-up engineering models have proven to be valuable tools. However, these models  require detailed and accurate building attributes related to physical properties (building geometry, height, building type, thermal transmittances, etc.), local climate (air temperature, humidity, solar radiation, etc.) and data related to occupants' energy behavior (occupants’ schedule, heating and cooling energy demand, efficiency of the system etc.). Among others, building construction year is one of the most relevant parameters since it is a key proxy for essential characteristics such as morphology, facade design, building materials, and energy efficiency. However, obtaining building construction year is particularly challenging as it is rarely available in public databases and, when available, the data are often incomplete or inconsistent. In this regard, remote sensing techniques can play a crucial role in the study and monitoring of the building stock. In particular, satellite images represent an excellent tool for the estimation of building age at local or regional scale given their extensive temporal and spatial coverage, as well as and the continuous updates of collections. The study focuses on the city of Parma, for which seven images covering the year range between 1985 and 2011 were selected. After a literature review, five built-up area extraction indices suitable for TM sensor were selected: Normalized Difference Built-up Area Index (NDBI), New Built-up Index (NBI), Band Ratio for Built-up Area (BRBA), Normalized Built-up Area Index (NBAI), and Vegetation Index Built-up Index (VIBI). In addition, Normalized Difference Vegetation Index (NDVI) was also considered, leading to a total of six indices. To improve the ability of these indexes to discriminate urban surfaces from areas with similar spectral signature (bare soil, sand, rock, etc.) annual greenest pixel composite images were generated using Google Earth Engine. Indexes performance was then compared on each image evaluating Receiver Operating Characteristic (ROC) and Precision-Recall (PR) curves, as well as performance metrics such as F1-score and Area Under the Curve (AUC). The results indicate that the NDVI is the best- Finally, temporal series were derived from the classification of images from different years, enabling the assessment of urbanization growth over time and, consequently, the estimation of building ages.

This study was carried out within the RETURN Extended Partnership and received funding from the European Union Next-GenerationEU (National Recovery and Resilience Plan – NRRP, Mission 4, Component 2, Investment 1.3 – D.D. 1243 2/8/2022, PE0000005) – SPOKE TS 1

How to cite: Vigato, T., Dalle Vedove, L., Dalla Vecchia, C., Zandonella Callegher, C., and Zilio, S.: Comparing the effectiveness of Landsat-derived spectral indices for building age prediction in urban energy modelling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20244, https://doi.org/10.5194/egusphere-egu25-20244, 2025.

Despite its potential future collapse and profound impacts, assessing the tipping dynamics of the Atlantic Meridional Overturning Circulation (AMOC) remains a significant challenge. Complex models such as Earth System Models (ESMs) and Earth System Models of Intermediate Complexity (EMICs) introduce substantial uncertainties in identifying tipping points. To address this, recent research has focused on developing conceptual models based on non-linear dynamics to capture the tipping behavior of the system. However, existing conceptual models typically simulate the AMOC response to a single temperature forcing, whereas it is well established that the AMOC is also influenced by freshwater flux.

In this study, we develop and validate an AMOC Tipping Calibration module that incorporates two forcing parameters: global mean temperature and freshwater flux. This module is designed as an emulator for the AMOC response within cGenie, an EMIC. Following validation, the emulator is integrated into SURFER, a simplified climate model that enables rapid and efficient simulations of AMOC trajectories under various scenario-based pathways. Our results show that incorporating both forcing parameters improves the accuracy of AMOC trajectory predictions. The methodology used to develop the two-parameter emulator is generalizable and can be applied to other tipping elements. By facilitating a greater number of simulations than complex models while maintaining calibration to them, this tool represents a significant advancement in exploring and understanding the potential future behaviour of the AMOC and other tipping elements.

How to cite: Laridon, A., Couplet, V., Thiery, W., and Crucifix, M.: Development and Validation of a Tipping Element Emulator Integrated into a Simplified Climate Model to Simulate the AMOC Collapse, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-821, https://doi.org/10.5194/egusphere-egu25-821, 2025.

The Atlantic Meridional Overturning Circulation (AMOC), as recorded in paleoclimate proxies, is one of the climate systems with a potential abrupt transition. Increasing identification of statistical signals—critical slowing down—in observational fingerprints empirically raises concerns that the system may be approaching a tipping point. However, state-of-the-art Earth System Models (ESMs) rarely project an abrupt collapse of AMOC, and its loss of stability has yet to be thoroughly investigated, leaving it unclear whether warning signals of AMOC tipping is overlooked in ESMs or exaggerated in fingerprints. Here, a warning signal over the deep convection site of AMOC is consistently identified in both observations and ESM, and we present that the currently observed signal is reconciled with the modeled one, with warming exceeding the Paris Agreement goal. This warning signal is in accordance with physical stability of the AMOC, the AMOC-induced freshwater convergence into the Atlantic basin, is overestimated in the ESM, so that it projects a delayed tipping point. These results suggest that the observed AMOC is approaching a tipping point akin to the projections of models simulating a much warmer Earth, underscoring potentially overlooked risks in ESMs assessments.

How to cite: Shin, Y., Oh, J.-H., Boers, N., Bathiany, S., Årthun, M., Lee, H., Iwakiri, T., Kim, G.-I., Kim, H., and Kug, J.-S.: Reconciled warning signals in observations and models indicate a nearing AMOC tipping point , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-859, https://doi.org/10.5194/egusphere-egu25-859, 2025.

In this paper, I explore how to assess technologies’ potential to aid a sustainable transformation via societal tipping points. I do this by providing a definition of sustainability that combines justice, well-being, and the value of nature with insights from value-sensitive design and the technology assessment literature. This exploration serves as an additional consideration for the developing societal tipping point scholarship. I argue that research surrounding societal tipping points can be meaningfully bolstered through philosophical reflection on the inherent ethical implications of sustainability, and the value-ladenness of technological development. There is a salient push within various strands of climate adaptation and sustainable transition scholarship towards systems thinking. In order to reorient society within the Anthropocene, and to adapt to a destabilized climate, such scholarship argues that the underlying subsystems society currently relies on need to change. Conceiving societal structures, such as institutional, political, and financial arrangements, the various planetary spheres (bio-, cryo-, hydro-, atmo-, and geosphere), and the techno-scientific infrastructure as interdependent systems has heuristic and practical allure. As a heuristic, it allows researchers and policymakers to account for the numerous interrelated systems that affect climate change and environmental degradation. Practically, this heuristic should enable the identification of impactful and sustainable action. Knowing how the subsystems interoperate, what drives them, and what function they provide, accordingly serves as a baseline to identify possible leverage points to change them. Conceptually mirroring climate tipping points, there is growing interest in societal tipping points as possible catalysts for decisive climate action. This interest is premised on the idea that societal tipping points within a currently unsustainable global societal-ecological-technical system can be identified and operationalized in order to tip the system (or subsystems) into a sustainable direction This premise raises at least two critical issues that have so far received little attention. First, the question arises what tipping towards a more sustainable system would look like. The concept of sustainability is arguably vague, especially when it comes to its aptness in describing climate action. Answering this question requires a reflective and ethically thick conception of sustainability, which in turn, needs to represent a future-oriented conception of justice, well-being, and nature. Second, it is crucial to reflect on the interdependent ways in which technological development and the implementation of new technologies affect the societal values and norms that drive them, since technology plays a central role for achieving societal tipping points. If technology is seen as a an accelerator and facilitator for a sustainable transition, the value-ladenness that technological innovation comes with needs to be addressed. Importantly, some technologies that seem sustainable on the surface, may actually entrench and enforce existing unsustainable modes of behavior and policies. Accordingly, this paper expands on the societal tipping points literature by proposing a concept of sustainability that serves as a means to assess the potential of technologies to facilitate sustainable tipping.

How to cite: Hofbauer, B.: Can We Tip Sustainably? Ethical Considerations on the Role of Sustainable Technology in Societal Tipping Points, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2923, https://doi.org/10.5194/egusphere-egu25-2923, 2025.

The subpolar gyre (SPG) of the North Atlantic plays a pivotal role in the Atlantic Meridional Overturning Circulation (AMOC) and climate through various teleconnections. This study examines the tipping point in thisregion within CMIP6 projections using three models: CESM2-WACCM, MRI-ESM2-0, and NorESM2-LM. These models, selected for exhibiting tipping patterns in at least one emission scenario, reveal distinct yet converging patterns of change, suggesting a destabilization of the subpolar region driven by shifts in salinity, temperature, and density profiles. A consistent feature across the models is pronounced freshening in the upper 150 meters of the water column. This results in a strong stratification, accompanied by cooling in the top 250 meters and warming between 150 and 1500 meters. The resulting enhancement in water column stability leads to a marked reduction in mixed layer depth (MLD). These changes disrupt vertical mixing, weaken nutrient transport, and alter regional circulation dynamics, with cascading effects on marine ecosystems and climate feedback mechanisms. We employed a density-based approach that accounts for the combined effects of temperature and salinity on water density to identify the critical surface salinity leading to the tipping of the SPG. This critical salinity represents a threshold for the salinity level beyond which density-driven stratification results in a stable water column. For stability to break, surface salinity must exceed this critical salinity. All three models consistently identify a critical salinity threshold of approximately 33.8 g·kg⁻¹. When surface salinity drops below this threshold, the subpolar region experiences rapid cooling, reduced convection, and potentially irreversible transitions. The tipping point of the SPG is preceded by an expansion of areas in the SPG where surface salinity falls below this critical threshold, accompanied by a decrease in MLD. To complement our analyses, we used the ISAS dataset to assess how far the system is from an SPG tipping point. Our next step is to establish an observable spatial pattern of early warning. Our findings underscore the vulnerability of the North Atlantic subpolar region to salinity-driven tipping points, which may lead to potentially irreversible transitions. This highlights the critical need for precise monitoring and advanced modeling of salinity dynamics to enhance predictability in future climate scenarios.

How to cite: Almeida, L. and Swingedouw, D.: Critical Salinity as an early warning of Tipping Point in the North Atlantic Subpolar Gyre, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3514, https://doi.org/10.5194/egusphere-egu25-3514, 2025.

Social norms are a key socio-cultural driver of human behaviour and have been identified as a central process in potential social tipping dynamics. They play a central role in governance and thus represent a possible intervention point for collective action problems in the Anthropocene, such as natural resource management. 
A detailed modelling framework for social norm change is needed to capture the dynamics of human societies and their feedback interactions with the natural environment. To date, resource use models often incorporate social norms in an oversimplified manner, as a robust and detailed coupled social-ecological model, scaling from the local to the global World-Earth scale, is lacking. 
Here we present a multi-level network framework with a complex contagion process for modelling the dynamics of descriptive and injunctive social norms. The framework is complemented by social groups and their attitudes, which can significantly influence the adoption of social norms. We integrate the modelling concept of norms together with an additional individual social learning component into a model of coupled social-ecological dynamics with a closed feedback loop, implemented in the copan:CORE framework for World--Earth modelling.
We find that norms generally bifurcate the behaviour space into two extreme states (sustainable vs. unsustainable) divided by regime shifts. Reaching a sustainable (i.e. safe) state becomes more likely with low thresholds of conforming to sustainable norms, as well as lower consideration rates of own resource harvesting success. The success of a generic social norm intervention is also found to be highly dependent on the group topology and exhibit a phase-transition like shape under certain conditions. The regime shifts in thresholds, individual learning and norm intervention hint at exploitable underlying tipping processes.
Our findings suggest that explicitly modelling social norm processes together with social groups enriches the dynamics of social-ecological models and determines safe operating spaces. Consequently, both should be taken into account when representing human behaviour in coupled World--Earth models.

How to cite: Bechthold, M., Barfuss, W., Butz, A., Breier, J., Constantino, S., Heitzig, J., Schwarz, L., Vardag, S., and Donges, J.: Social norms and groups structure safe operating spaces and exhibit regime shifts in renewable resource use in a social-ecological multi-layer network model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3532, https://doi.org/10.5194/egusphere-egu25-3532, 2025.

Forests have considerable potential to influence the stability of the Earth system and mitigate climate change by influencing biogeochemical and biophysical processes. Tree cover, as the primary layer of exchange for carbon, energy and water cycles, play a critical role in such dynamics. However, the persistence and functionality of forests are highly dependent on their resilience to the ongoing rapid changes in natural and anthropogenic pressures. Experimental evidence of a sudden increase in tree mortality across different biomes is rising concerns about the ongoing changes in forest resilience and the associated risks to the climate mitigation potential of forests. Previous global-scale assessments of forest resilience have focused on the use of critical slowing down indicators, such as temporal autocorrelation and variance. These studies have provided important insights, but they can only partially capture the effects of stochastic disturbances and forest management.  

In this study, we explore the potential of spatial statistical indicators (SSI), such as spatial variance and skewness, as early warning signals of regime shifts in global forests. To this aim, we first derive tree cover values for the 2000-2023 period at 0.05-degree spatial resolution for the whole globe by combining multiple satellite observations. We then, develop a machine learning model to disentangle the climate effects on tree cover distributions and elucidate the underlying mechanisms. SSI are ultimately computed on the residuals of the machine learning model and their spatial and temporal variations analysed.

Results show, along with a widespread erosion of tree cover, an increase in both SSI prominently in tropical and boreal forests over the observational period. According to the stability theory, the simultaneous increase in these metrics indicates a rising instability of the system by reflecting an alteration of the shape of the basin of attraction. Such patterns appear largely driven by the increase in stochastic perturbations and human pressures which are not detected using traditional critical slowing down indicators. Overall, this study contributes to better understand the recent dynamics in forest resilience and its underlying mechanisms that can lead to critical transitions. Considering the expected intensification of natural pressures in view of climate change, it is becoming urgent to identify adaptation measures to preserve the long-term stability of global forests and the provision of their ecosystem services.

How to cite: Mura, M., Somasundaram, D., Migliavacca, M., Dakos, V., Cescatti, A., and Forzieri, G.: Spatial variance and spatial skewness as leading indicators of regime shifts in global forests, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6613, https://doi.org/10.5194/egusphere-egu25-6613, 2025.

How to cite: Brunetti, M., Kasparian, J., and Moinat, L.: Detection of global-scale tipping using climate networks , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7369, https://doi.org/10.5194/egusphere-egu25-7369, 2025.

We propose a framework that allows to integrate social (tipping) dynamics into a DICE-style IAM model, in which a policy maker determines abatement effort and savings. The policy maker maximizes the sum of social welfare and a political penalty/reward, depending on whether the majority of the population opposes (penalty) or supports (reward) more ambitious abatement policies. The social process itself depends inter alia on observable climate impacts. We provide numerical simulations that illustrate the impact of the tipping process on policy choices which in turn are built around a total cost of carbon that embraces both the "classical" social cost of carbon and a political cost of carbon. Our initial findings illustrate (i) the considerable scope for political penalties (rewards) to stifle (boost) abatement policies; (ii) an incentive for the policy maker to distort policies in a way that boosts political support; and (iii) a considerable deviation between the total cost of carbon and the social cost of carbon. We argue how the model can be used for the purpose of understanding climate policy making from a "social dynamics" perspective.

How to cite: Kuhn, M., Wagner, G., and Wrzaczek, S.: Optimal climate policies under the shadow of social tipping, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10019, https://doi.org/10.5194/egusphere-egu25-10019, 2025.

The Atlantic Meridional Overturning Circulation (AMOC) is a key tipping element of the climate system due to its influence in regulating the meridional transport of heat and freshwater. Its stability is influenced by the interplay between external forcings (such as greenhouse gasses increase) and internal climate variability. Due to limitations on the deterministic predictability of the AMOC asymptotic state, the concept of a “probabilistic safe operating space” has been proposed. For this purpose, rare-event techniques, specifically the Giardina–Kurchan–Tailleur–Lecomte (GKTL) and Trajectory-Adaptive Multilevel Splitting (TAMS) algorithms, offer promising tools for testing the multistability of the system and assessing this probability at lower computational costs than traditional Monte Carlo methods.  Here, using the intermediate complexity model (PlaSIM-LSG), we estimate the probability of AMOC collapse in sensitivity experiments at different CO₂concentrations and under RCPs scenarios. In particular, TAMS has been applied in order to assess the probability of reaching a low circulation state of the AMOC associated with a 1°C temperature anomaly over central and western Europe. Our findings from sensitivity experiments, consistently with previous studies, indicate that for a wide range of CO₂ concentrations (500-600 ppm), the probability of an AMOC collapse is significantly different from zero (1-10% within 150 years). While such a collapse is unlikely to happen within the 21st century, it becomes likely to happen by 2150 in higher emission scenarios. It is important to note that PlaSIM-LSG does not account for the North Atlantic freshwater flux from Greenland melting which introduces a stabilizing bias for the AMOC-on state. Accounting for this mechanism would likely increase the probability of an AMOC collapse. These results underscore the importance of probabilistic assessments in understanding AMOC stability and highlight the potential for rare-event algorithms to provide insights into the statistical properties of tipping point.

How to cite: Cini, M., Jacques-Dumas, V., and Dijkstra, H. A.: Assessing the Probability of CO2-Driven AMOC Collapses Using Rare Event Algorithms in PlaSIM-LSG, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10648, https://doi.org/10.5194/egusphere-egu25-10648, 2025.

Achieving ambitious climate change targets, such as limiting global warming to 1.5°C, requires both political and social determination. Bottom-up pro-environmentalist behaviours can facilitate crossing social tipping points (STPs), resulting in new social norms with lower impact on global warming. While the passing of  STPs has been described qualitatively, it remains poorly understood how the climate benefits of this phenomenon can be quantified. 
Here, we introduce a stylised system dynamics model that couples socio-ecological contagion with global warming via greenhouse gas emission pathways to estimate the impact of crossing social tipping points on greenhouse gas mitigation and global warming. . This is explored through two examples of bottom-up and top-down mitigation interventions.
Results indicate that a STP could be crossed before 2050. While neither bottom-up nor top-down interventions alone are likely to achieve the 1.5°C target, their combined effect significantly reduces overshoot. This represents a significant step towards understanding how both bottom-up and top-down interventions can be harnessed to mitigate global warming. Our research underscores the importance of bottom-up pro-environmental movements, emphasizing their crucial role in not only reducing personal carbon footprints but also alleviating the burden on technological top-down interventions. This evidence of the benefits of promoting socio-ecological contagion should bolster the determination of individuals and community grassroots groups. Additionally, it should encourage top-down interventions to acknowledge and support the complementary role of collective action.

How to cite: Elliot, T., Donges, J., Pizzol, M., and Otto, I.: Manifesting tipping points in pro-environmental behaviour for climate change mitigation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10788, https://doi.org/10.5194/egusphere-egu25-10788, 2025.

A tipping element is a system that may pass a tipping point, that is, a threshold value of an environmental stressing condition at which a small disturbance can cause an abrupt shift of the tipping element from one state to another, accelerated by positive feedbacks. For example, under rising temperatures and increasing deforestation, the Amazon rainforest could tip from a forest state to a savanna state; one feedback involved is that fewer trees means less evapotranspiration, thus less rainfall and, finally, less trees. Therefore, the fewer trees, the harder it is for the remaining forest to adapt and survive. This phenomenon is called critical slowing down: approaching a bifurcation, a tipping system's resilience decreases, resulting in increasing autocorrelation and variance. The latter indicators are thus often measured to detect bifurcation-induced tipping and are called early-warning signals (EWS).

Let us describe the dynamics of a tipping element Y with the following equation: dY/dt = f(Y, r) + η, with r the environmental stressing condition involved in the tipping behavior and η some noise (e.g., climate variability). Deriving EWS directly from Y's time series relies on the assumption that the noise η is not correlated (white-noise), otherwise any trend in η's autocorrelation would be incorporated in Y's autocorrelation, even if not related to the tipping behavior contained in f(Y, r). In the Amazon rainforest example, increasing deforestation due to human activity is a part of r with a long-term effect, while e.g. ENSO also influences the forest but on short time scales and with sometimes opposite effects depending on its phase (El Niño/La Niña/neutral), and would be part of η.  If for example El-Niño's autocorrelation increases with time, the rainforest autocorrelation might also increase regardless of whether the forest is approaching the bifurcation point or not, therefore the autocorrelation would no longer reflect changes in the forest resilience.

To know how far the system is from the bifurcation point, we want to measure Y's internal autocorrelation (excluding noisy influences η, considering only f(Y, r)), and thus to answer the question: "If we intervene in the system and set the value of Y at time t-1, how does Y evolve at time t?" This defines the direct causal effect of Y_t-1 on Y_t and comes under the heading of causal inference: we look at the influence of setting Y_t-1=y_t-1 on Y_t, whatever the values of the other variables causing Y, which is fundamentally different from a direct measure where the value of Y at time t-1 is, in the general case, dependent on the state of the other variables. To measure how the direct causal effect of Y_t-1 on Y_t  evolves with time (with changing r), we use causal effect estimation, which quantifies the causal effect of hypothetical interventions in a system from observational data --the interventional distribution being rarely available in the majority of systems--- and an assumed causal graphical model that allows us to derive an adjustment expression that controls for confounders. We demonstrate the method on an ideally forced simulated system and discuss potential applications.

How to cite: Lanson, A. and Runge, J.: Causal effect estimation for robust detection of critical slowing down, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10947, https://doi.org/10.5194/egusphere-egu25-10947, 2025.

The Atlantic Meridional Overturning Circulation (AMOC) is one of the most critical tipping elements in Earth’s climate system, with its collapse posing far-reaching implications for weather dynamics and extremes, sea level rise, and Northern Hemisphere cooling. Although it is considered a low-probability but high-impact scenario, recent studies suggest that the AMOC may already be on a trajectory towards collapse. Moreover, current climate models struggle to fully capture the complex interactions between Greenland ice sheet melting and the AMOC slowdown, adding further uncertainty to climate projections. Artificial hosing experiments in the North Atlantic, project the weakening of the AMOC to increase sea surface temperature in the Southern Hemisphere and the tropics, particularly in ocean basins where tropical cyclones form. This warming, combined with rising sea levels and changes in vertical wind shear, could create conditions that promote the development of tropical cyclones and amplify their impacts. Although several studies have explored the relationship between tropical cyclone activity and AMOC weakening, the associated socioeconomic impacts remain uncertain. The goal is to investigate the direct and indirect socioeconomic impacts of tropical cyclones under future climate scenarios characterized by a weakened and fully collapsed AMOC. Will tropical cyclones affect areas that were previously unaffected? Will tropical cyclones' activity intensify, leading to greater societal impacts? To answer these questions, two sets of five-member ensemble simulations were performed for a 2°C stabilization emission scenario using the GFDL ESM2M, with and without induced AMOC collapse. These simulations were then coupled with the MIT coupled statistical-dynamical tropical cyclone model to simulate tropical cyclone activity under these conditions, and the probabilistic climate risk modeling platform CLIMADA was used to analyze the socioeconomic impacts. We anticipate this study to be a stepping stone in a broader ongoing effort to assess the socioeconomic impacts of extreme weather events triggered by tipping points.

How to cite: Colombi, N., Kropf, C. M., Burger, F. A., Meiler, S., Emanuel, K., Frölicher, T. L., and Bresch, D. N.: Tipping the AMOC: Impacts of Tropical Cyclones in a Changing Climate, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11175, https://doi.org/10.5194/egusphere-egu25-11175, 2025.

Plastics represent, in volume, up to 50% of materials present in modern vehicles with most of them being black. Consequently, black plastics are a key material stream to be managed in end-of-life vehicles (ELV) waste. The EU directive on ELV, updated in 2023, introduces new rules covering all aspects of a vehicle life cycle, from its design and market placement until its final treatment as ELV. These new specific criteria now put pressure on car manufacturers and black plastic recyclers to boost circularity in the production and recycling chains, including:

• Improving circular design of vehicles to facilitate removal of materials, parts and components for reuse and recycling;
• Ensuring that at least 25% of the plastics used to build a new vehicle comes from recycling (of which 25% from recycled ELVs).

The first step required to improve the circularity of ELV polymers is to identify the main polymer types present in the stream with optical sensing. Current identification workflows are successfully employed by the plastic-waste recycling industry, based on material-specific signals present in the visible-to-near infrared (VNIR) and short-wave infrared (SWIR) ranges (400–2500 nm). Nevertheless, VNIR/SWIR sensors are unsuitable for identification of black plastics due to the strong signal absorption by dark pigments in this spectral region. In recent years, novel hyperspectral sensors operating in mid-wave infrared (MWIR, 2700–5300 nm) have been successfully employed for identification of black plastics. Yet, the automotive industry requires high-performance materials which led to the development and use of very specific polymer variants, including multi-polymer blends (e.g. ABS/PC), polymer subtypes (e.g. PA6 and PA6.6), and functional additives (e.g. glass fiber, talc, carbon black). Consequently, the identification of the usual polymer classes is not adapted to meet the minimum quality requirements for recycling and, hence, not adequate for future use for car material streams (closed loop). 

Such complexity is justified by the need for high performance and functionality of materials in automotive applications, but impacts recyclability and ultimately leads to downcycling. In order to ensure that high-purity black plastics are obtained at the end of the recycling operation, at the standards needed by the automotive industry, it is necessary to go beyond the identification of main polymer types. 

In this contribution, we address the current challenges and propose solutions to identify the important high-performance polymers used by the automotive industry that could be recycled. Further, we evaluate the suitability of current industrial optical sensing techniques for identification of black plastics originated from ELV waste. We also propose solutions for the identification of plastics with highly-complex composition present in ELVs such as multi-polymer blends (e.g. ABS/PC), polymer subtypes (e.g. PA6 and PA6.6), and functional additives (e.g. glass fiber, talc, carbon black).

How to cite: de Lima Ribeiro, A., Fuchs, M., Madriz, Y., and Gloaguen, R.: Challenges and solutions on identification of high-performance black plastics for closed-loop car recycling  , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12221, https://doi.org/10.5194/egusphere-egu25-12221, 2025.

Resilience is typically defined as the ability of vegetation to recover from external perturbations such as fires or droughts, and it can be quantitatively measured by the rate of recovery following such events. Resilience can also be assessed indirectly, even in the absence of large perturbations. One key metric for this is autocorrelation. A loss of resilience over time, often referred to as "slowing down," can be detected as an increase in autocorrelation. In simple one-dimensional dynamical systems, a reduction in resilience is also associated with increased sensitivity of the system's stable state to external conditions.

Recent studies, using indicators such as the Normalized Difference Vegetation Index (NDVI) and Vegetation Optical Depth (VOD), have found that resilience tends to be higher in wetter regions of tropical forests compared to drier regions, and that resilience has been decreasing across large parts of the Amazon rainforest. Additionally, empirical recovery rates after disturbances have been found to correlate with autocorrelation, supporting the practical relevance of theoretical expectations. However, it remains unclear which specific vegetation properties and processes determine the observed patterns.

Here we use idealized simulations with the state-of-the-art dynamic vegetation model LPJmL and explore how the resilience of natural forests and its indicators depend on (i) climate, (ii) vegetation composition (i.e., the mix of plant functional types), (iii) the vegetation property (variable) being considered, and (iv) the nature of the perturbation(s). We find that autocorrelation qualitatively aligns with the recovery time from large, negative perturbations that affect all tree types similarly.

However, there are exceptions where the factors listed above can influence the relationship in unexpected ways. Specifically, for some tree types and climate regimes, recovery rates and autocorrelation do not align with each other, nor with the forest's sensitivity to climate change. For example, perturbations that alter the relative abundance of tree types can lead to different recovery rates compared to those affecting all tree types uniformly. Moreover, vegetation variables that recover quickly when perturbed in isolation (e.g., fluxes like net primary productivity) may still co-evolve with slower variables they depend on (e.g., carbon stored in trees). We identify key mechanisms behind these features in the model and test their relevance by simulating a more realistic setup, using observed climate data within a geographically realistic domain. We also discuss the relevance of these mechanisms in the real world.

Our findings highlight the need to better understand the nature of disturbances and trends in ecosystems, as well as the mechanisms captured by satellite-derived indicators. This knowledge, along with improved resilience monitoring, will be crucial for making reliable predictions about how ecosystems will respond to human-induced changes.

How to cite: Bathiany, S., Blaschke, L., Morr, A., and Boers, N.: Vegetation resilience and sensitivity in complex dynamic vegetation models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13560, https://doi.org/10.5194/egusphere-egu25-13560, 2025.

The rapid growth of the electronics market, driven by high demand for new technologies, has shortened the lifespan of electronic products, leading to a surge in electronic waste (E-waste). Comprising 25% plastics, E-waste contains unrecovered critical and toxic materials, necessitating advanced recycling strategies. HeliosLab, an infrastructure combining imaging sensors and robotic chemical analyses, was developed at the Helmholtz Institute Freiberg. HeliosLab integrates spectroscopy-based modalities such as RGB and hyperspectral imaging (HSI) across multiple wavelength ranges that can be used to optimize E-waste sorting. The complexity of the hyperspectral data, compounded by multisensory integration, requires sophisticated automated algorithms to efficiently process large volumes of data and extract critical material features. These advancements ensure scalable, fast, and automated detection solutions for industrial-scale E-waste recycling operations.

We are developing smart and novel processing methodologies utilizing state-of-the-art (SOTA) machine learning hyperspectral imaging (HSI) classification models. In this study, we focus on Transformer-based architectures, known for their self-attention mechanisms that effectively capture contextual relationships between their input tokens, which enables unique spatial-spectral feature detection, relevant to remote sensing and HSI applications. Such an approach significantly advances automated polymer identification. 

To test the model’s performance on unseen data and evaluate the generalization performance of those SOTA models in industrial-like environments, multiscene datasets are required. We acquired a new multiscene HSI polymer dataset in the near visible (NIR) to the short-wave infrared (SWIR) (400-2500 nm) using hyperspectral cameras available at HeliosLab. The initial deployment highlighted the challenges related to both, the data quality and quantity, as well as regarding methodological frameworks. This led us to develop a tailored Transformer-based topology capable of detecting polymer fingerprints using novel refined extractions of the spatial and spectral features. Our research and advancements contribute to the automation and optimization of polymer detection in E-waste recycling, paving the way for improved resource recovery and environmental sustainability.

How to cite: Arbash, E., de Lima Ribeiro, A., Fuchs, M., Ghamisi, P., Scheunders, P., and Gloaguen, R.: Optimizing Plastic Identification in E-Waste Recycling through Hyperspectral Imaging and Transformer-Based Machine Learning Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13620, https://doi.org/10.5194/egusphere-egu25-13620, 2025.

The global climate system continues to change under the influence of human activities. Of particular concern is the difficulty of continuing human activities due to irreversible and long-term abrupt changes caused by the global climate system exceeding its tipping point. Climate systems that have the potential to exceed the tipping point are called tipping elements and are being studied. Among them, the Atlantic Meridional Overturning Circulation (AMOC) plays a central role in the movement of materials through the ocean and is connected to many other tipping elements. While there is concern that the AMOC may decrease in strength due to rising temperatures in the Atlantic Ocean, freshwater inflow due to melting ice in the Arctic region has been investigated as a stabilizing factor. Therefore, it is important to comprehensively consider these influences when evaluating the AMOC tipping point.

In the AMOC modelling, Stommel’s two box model describes its nature well. Although, it does not treat freshwater input from multiple estuaries. We have applied three box model (TBM) [1] which divide the Atlantic Ocean into three elements with double estuaries. Freshwater inflows to the Arctic Ocean due to ice sheet thawing in Greenland and permafrost thawing in Siberia were calculated using AWI-ESM [2] and CMIP6 [3] data, respectively. In addition, temperature differences between the southern and northern Atlantic regions were calculated by MRI-ESM2.0 [4].

We also adopted the method of analyzing the time-series behavior of the AMOC as a stochastic process, as in Ditlevsen et al. (2010) [5]. Finally, we estimated the age of AMOC decay based on the analytical AMOC behavior by TBM and by identifying the parameters of the Langevin equation.

[1] E. Lambert, T. Eldevik, P.M. Haugan “How northern freshwater input can stabilise thermohaline circulation”, Tellus A: Dynamic Meteorology and Oceanography, 68 (1) (2016), p. 31051

[2] Ackermann, L., Danek, C., Gierz, P., and Lohmann, G. “AMOC Recovery in a multicentennial scenario using a coupled atmosphere-ocean-ice sheet model”, Geophys. Res. Lett., 47, 2020.

[3] Wang, S., Wang, Q., Wang, M., Lohmann, G., & Qiao, F. (2022). ”Arctic Ocean freshwater in CMIP6 coupled models” Earth’s Future, 10(9)

[4] Yukimoto, Seiji; Koshiro, Tsuyoshi; Kawai, Hideaki; et al. (2019) “MRI-ESM2.0 model output prepared for CMIP6 ScenarioMIP ssp585”

[5] Ditlevsen, P. D. and Johnsen, S. J.: Tipping points: Early warning and wishful thinking, Geophys. Res. Lett., 37, L19703, 2010

How to cite: Kono, K. and Fukuda, T.: Instability analysis of the AMOC with varying freshwater input and sea water temperature in the Atlantic Ocean, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14805, https://doi.org/10.5194/egusphere-egu25-14805, 2025.

In Paris 2015, the global community agreed to keep global warming well below 2.0°C aiming to limit it to 1.5°C above pre-industrial levels. However, recent research has shown that overshooting this temperature guardrail is becoming increasingly likely and several climate data teams across the world recorded 2024 as the first individual year with a global warming level above 1.5°C.

Such temperature levels endanger critical components of the Earth system, the so-called climate tipping elements such as the Greenland and West Antarctic Ice Sheet, The Atlantic Meridional Overturning Circulation, or the Amazon rainforest. In this presentation, we will show the latest evidence on how overshooting temperature targets increases tipping risks. In particular, we will discuss the role of overshooting the 1.5°C and 2.0°C for the stability of critical Earth system components, and also assess the likelihood for climate tipping cascades beyond these global warming levels.

How to cite: Wunderling, N., Högner, A., Möller, T., Ritchie, P., Rockström, J., Steinert, N., and Donges, J. F.: Increased climate tipping risks from temperature overshoots, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15540, https://doi.org/10.5194/egusphere-egu25-15540, 2025.

Populations in mountain ecosystems face the risk of extinction due to climate change. Yet, how these risks will materialise remains unclear because many ecological parameters are unknown. We show that progress can be made by examining how habitats get fragmented and isolated as populations shift to higher elevations. When this shift is slow relative to dispersal, the amount of aggregation and connectivity between habitat fragments determine the warming threshold beyond which populations cannot sustain themselves. If the shift is rapid compared to dispersal, there is also a critical warming rate beyond which populations cannot track their preferred range and go extinct. Through simulations and analyses of stochastic spreading processes on real and artificial landscapes, we investigate how mountain topography, warming rates, and demographic mechanisms affect extinction thresholds. Understanding the link between mountain topography and extinction risks may enable targeted interventions to mitigate these risks, especially in areas with fragmentation bottlenecks.

How to cite: Wuyts, B., Karger, D., Sieber, J., and Boussange, V.: The link between topography and climate extinction risks in mountain ecosystems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16872, https://doi.org/10.5194/egusphere-egu25-16872, 2025.

Critical transition is an increasingly widespread phenomenon in global forest ecosystems, yet predicting its occurrence remains a challenge due to limited understanding of the underlying mechanisms and the variability in empirical relationships between forest health and external perturbations that can lead to critical transitions.

In this study, we demonstrate that the temporal loss of resilience, quantified in terms of critical slowing down (CSD) indicators, can serve as an early warning system (EWS) for critical transitions in forest ecosystems. CSD indicators are analyzed by integrating MODIS satellite data-derived vegetation dynamics and disturbance events from 2000–2021 at 500 m. We applied this approach to 13 living labs distributed across Europe, South America, and China, covering a wide range of environmental gradients and forest management types. Results show that CSD indicators can efficiently capture spatial and temporal variations in critical transitions in near real time. These findings highlight the potential of this EWS for improving forest critical transition predictions as well as providing practical insights for management and adaptation strategies in the context of climate change and at a spatial scale appropriate for decision makers.

How to cite: Somasundaram, D., Elia, A., Mura, M., Pickering, M., and Giovanni, F.: Satellite-based early warning system of critical transitions in forest ecosystems at living lab scale, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16943, https://doi.org/10.5194/egusphere-egu25-16943, 2025.

Tipping points have been identified in several components of the Earth system, raising concerns about abrupt and potentially irreversible changes under climate change. A recent analysis of GRACE satellite data reveals a sudden and unprecedented decline in terrestrial water storage (TWS) during 2015–2016, coinciding with a major El Niño–Southern Oscillation (ENSO) event (Rodell et al. 2024). This decline suggests a recent net drying of the land and raises the hypothesis of a regime shift in the global terrestrial water system. Potential mechanisms include enhanced evapotranspiration, intensifying drought frequency and severity, and land–atmosphere feedbacks. Early warning signals, such as increased autocorrelation and variance observed prior to the decline, support this hypothesis.

To evaluate the significance and rarity of the observed transition, we develop a detection methodology and apply it to both observational estimates and climate model simulations. By analysing fully coupled pre-industrial control simulations, historical simulations, and AMIP-style experiments with prescribed sea surface temperatures, we aim to disentangle the roles of anthropogenic climate change and specific modes of climate variability (e.g., ENSO) in driving this transition. Furthermore, we explore the potential for transitions to alternative states in global TWS. Our work establishes a framework for understanding abrupt changes in TWS and their implications for the terrestrial water cycle in a warming climate.

References

Rodell, M., Barnoud, A., Robertson, F.R. et al. An Abrupt Decline in Global Terrestrial Water Storage and Its Relationship with Sea Level Change. Surv Geophys 45, 1875–1902 (2024). https://doi.org/10.1007/s10712-024-09860-w

How to cite: Fisch, C., Gudmundsson, L., Schumacher, D. L., and Seneviratne, S. I.: Assessment of a potential regime shift in global terrestrial water storage, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17534, https://doi.org/10.5194/egusphere-egu25-17534, 2025.

The Anthropocene epoch is characterized by an excessive use of natural resources and energy that drives the environmental destruction of the planet. However, large inequalities exist among different social groups that benefit to various degrees from the use of resources and energy, as well as among those suffering from the negative impacts of environmental destruction. In this paper, we systematically analyze these differences and discuss a social stratification theory based not only on differences in terms of possessions or social status, but also on differences in how these groups can control and benefit from the planetary material cycles and energy flows or suffer the consequences of environmental degradation. Referring to consumption data, we propose six global socio-metabolic classes and show distinctive patterns in the energy use of these classes. More research is needed to reveal differences in the use of natural resources essential for maintaining the biosphere integrity, such as land, water, nitrogen, and phosphorus. Targeted policy measures that address excessive appropriation of energy and natural resources are needed, as are expansions in infrastructure and institutional change that supports the wellbeing of humankind, and especially of the most marginalized classes.

How to cite: Otto, I. M.: Socio-metabolic class conflicts in the Anthropocene, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17704, https://doi.org/10.5194/egusphere-egu25-17704, 2025.

The Atlantic Meridional Overturning Circulation (AMOC) is considered one of the Earth’s climate tipping elements. Concerns have been raised that global warming could increase the freshwater input into the North Atlantic at high northern latitudes and thereby abruptly interrupt the deep water formation that fuels the AMOC’s lower limb and is necessary to maintain the overturning. To assess the risks such an AMOC tipping scenario poses to societies, it is essential to understand how an AMOC collapse feeds back into the climate system as a whole. It is of particular interest whether an AMOC tipping would have a stabilizing or destabilizing effect on other climate tipping elements. In this context, we studied the impact of an AMOC shutdown on the Amazon Rainforest, which is itself thought to be at risk of undergoing a transition to a savanna state. We forced a km-scale atmosphere-only model with sea surface temperatures from a second lower-resolution coupled climate model simulation that features a collapsed AMOC state. Previous studies indicate that land-atmosphere interactions are different in such convection-resolving models compared to CMIP-type models, possibly affecting the response of precipitation to large-scale perturbations. In general, our simulation confirms the global AMOC-collapse induced precipitation and temperature anomaly patterns also seen in coupled climate model hosing experiments. Most prominently these comprise a cooling and drying of the North Atlantic region and a corresponding southward shift of the tropical rainbelt. However, upon closer examination, we find that over land the signal is attenuated, and in particular precipitation patterns over the Amazon Rainforest appear to be remarkably robust against an AMOC shutdown. This, in turn, means that a tipping of the AMOC would to a first degree neither have a stabilizing nor destabilizing effect on the Amazon Rainforest.

How to cite: Riechers, K., Hohenegger, C., Schmidt, H., Esch, M., and Stevens, B.: Amazon precipitation response to an AMOC shutdown in a km-scale atmospheric model , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17856, https://doi.org/10.5194/egusphere-egu25-17856, 2025.

The relationship between nature and industry has been constantly contested for decades, regardless of the warning on the Earth as transformed by human action (Turner, B. L., et al., 1993) which address the unprecedented changes in the biosphere that have taken place over the last 300 years. Accumulation of the human impacts has also led to the degradation of the atmosphere resulting in anthropogenic warming that have brought tremendous threats to societies. While standing at the tipping point, international societies have made substantive advancement to push industrial and financial sectors taking responsibility in carbon accounting and climate risk assessment (i.e., TCFD). The relationship between climate and industry is complex, but the threat from both of them on biodiversity and natural capital (NC) loss is even devastating. In 2021, the initiative of Taskforce on Nature-related Financial Disclosures (TNFD) was launched and the TNFD Recommendations and Guidance is released in 2023. With a broader focus on the industrial and financial sector dependence and impact on the NC, TNFD contains a big range of ambiguity in developing methodology. Meanwhile, climate risk is regarded as a driving force of ecosystem change in the assessment. How to access relevant data, in suitable spatial and temporal resolution, to quantify the NC related dependency and risk becomes the most fundamental challenge before tipping elements and tipping interactions can be identified to facilitate a social and industrial transformation. 

This study collaborates with a listed high-tech company in Taiwan to assess the relationship between its value chain and NC following the LEAP approach. We have developed a high-spatial-resolution database to identify the dependencies and impacts on NC across different operational locations. Meanwhile, we conduct materiality assessments through internal questionnaires, examining the significance of different types of NC to business operations from the perspectives of Consequences rating and Likelihood rating. Finally, we aim to establish TNFD risk matrices by integrating the assessment results from spatial and materiality assessments, with the hope of helping enterprises to identify NC requiring immediate attention and action.

Overall, the integrated TNFD assessment method combining spatial and materiality analyses serve as tipping elements between enterprises and NC; it may help enterprises systematically quantify their dependencies and impacts on NC, thereby identifying operational locations and types of NC that require priority action. Meanwhile, high-resolution spatial databases can support enterprises in defining the locations, scope, and even causes of NC issues, which in turn can help identify key external stakeholders and initiate new engagements. This integrated assessment approach has the potential to address the methodological gaps in TNFD development and to provide a concrete empirical foundation for business operational transformation, helping enterprises to develop early adaptation and response strategies when facing global ecosystem changes.

How to cite: Chung, M.-K., Lin, K. E., and Tseng, W.-L.: Advancing the pathway towards natural capital based assessment for industrial and financial sectors, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18362, https://doi.org/10.5194/egusphere-egu25-18362, 2025.

The Tipping Point Modelling Intercomparison Project (TIPMIP) is an international intercomparison project that aims to systematically advance our understanding of tipping dynamics in various Earth system components, and assess the associated uncertainties. By connecting and evaluating various models, TIPMIP will fill critical knowledge gaps in Earth system and climate modelling by improving the assessment of overall anthropogenic forcing and long-term commitments (irreversibilities). In this contribution, we report on the status of the project, highlighting recent advances including the finalisation of experimental protocols and first results. Moreover, we provide an overview on the established scientific infrastructure and next steps, inviting the tipping points modelling community for contributions.

How to cite: Loriani, S., Dennis, D., Donges, J. F., and Winkelmann, R.: News from TIPMIP, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18552, https://doi.org/10.5194/egusphere-egu25-18552, 2025.

We are transitioning towards a climate state on Earth featuring rapid changes in response to anthropogenic greenhouse gas emissions and land-use change, with severe observable and projected impacts on the occurrence of extreme weather events and increasing risk of crossing large-scale tipping points. Neither the transition nor the long-term climate state has been observed by (human-made) measurements before, making information on past climatic states increasingly more important to help anticipate future Earth System change. Paleoclimate records have enormously expanded over the past decades, and provide extremely rich information about physical, cryospheric, biological, and ecological processes on many spatial and temporal scales. Yet, it has been difficult so far to directly transform this knowledge on past processes into a more confident evaluation of future projections for the Earth system. In this contribution, I will summarise lessons learned from past climate change on our understanding of climate variability, abrupt changes and climate response to greenhouse gas changes and other forcing. For example, generalizations of classical measures such as equilibrium climate sensitivity can be useful in the palaeoclimate and future context for understanding the response of a climate state to radiative forcing beyond the linear regime, i.e. when (part of) the climate system is close to a tipping point. Finally, this contribution will present the ambition and programme of the starting EU-HORIZON project Past-to-Future (P2F) aiming at developing, expanding and using the wealth of paleoclimate data to improve existing Earth System Models in terms of their ability to describe possibly exotic, out of sample, climate states and the transition pathways towards them from current conditions.

How to cite: von der Heydt, A.: Past to future: Towards fully paleo-informed future climate projections, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19085, https://doi.org/10.5194/egusphere-egu25-19085, 2025.

As part of a project exploring the relation between the Greenland ice sheet stability and the AMOC, we present coupled climate and ice sheet simulations of Greenland with the Earth System Model (ESM) UKESM, a state-of-the-art ESM capable of representing the interactions between ice sheets and the atmosphere and their co-evolution (UKESM-ice; Smith et al., 2021).
Recent large ensemble exercises indicate that there is no sign of non-linear volume change or irreversibility at the scale of the Greenland ice sheet in UKESM-ice, even at high warming levels and despite large ice losses.
We present new simulations exploring Greenland's potential for tipping with modified (snow and ice sheet) parameters and including a recently developed scheme for the marine forcing of outlet glaciers, which was previously omitted from UKESM-ice and prevented the representation of the direct influence of the ocean on the Greenland ice sheet. Results show linear trends of (large) ice volume change at the scale of the ice sheet but local evidence of accelerating melt along the South West margin.
Next steps in the project include providing fresh water from UKESM-ice surface runoff and solid discharge of icebergs to investigate their effect on the strength of the AMOC in NEMO simulations and using a new high resolution NEMO dataset as ocean forcing of the ice sheet in UKESM-ice. 

How to cite: Lang, C., Smith, R., George, S., Marsh, R., and Sinha, B.: ISOTIPIC: Greenland ice sheet potential for tipping with the Earth System Model UKESM-ice, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19501, https://doi.org/10.5194/egusphere-egu25-19501, 2025.

Mounting evidence that human activities are driving Earth’s climate toward dangerous tipping points has raised the question of whether these may be averted by quickly reaching tipping points in human societies. Prior work on social tipping dynamics has focused mainly on defining its key features, identifying and characterising important social tipping elements, and operationalizing interventions for triggering individual elements. However, the success of climate-stabilizing interventions will depend on their timing and coordination across multiple tipping elements that operate on different characteristic time scales, and these coupled dynamics are currently not understood. In this work we explore the challenges of intervention timing and the potential to coordinate subsystem tipping cascades in a multiscale system to achieve a timely whole-system transition. We develop a stylized model in which the changing climate is coupled to a network of social tipping elements such as public support for climate action, political policymaking, financial investment in energy technologies, and energy infrastructure substitution, each with its characteristic dynamics and time scales. We study how intervention timing interacts with tipping cascades between subsystems and derive principles for navigating the system to the desired state. Additionally, we analyze the effects of `windows of opportunity’ - unpredictable system shocks that are likely to become more frequent as climate change intensifies - on our model transformation pathways, and ascertain how these may be exploited to disrupt system-stabilizing feedbacks and synchronize subsystem changes. Our findings emphasise the importance of a complexity-based understanding of human agency and governance of the world-Earth system.

How to cite: Ringsmuth, A., Tilman, A., Everall, J., Campiglio, E., Pieler, M., Constantino, S., and Otto, I.:  Navigating multiscale social tipping dynamics to stabilise Earth’s climate, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19773, https://doi.org/10.5194/egusphere-egu25-19773, 2025.

There is growing concern that the Atlantic Meridional Overturning Circulation (AMOC), a vital Earth system component, could weaken or even collapse under climate change. Despite the severe potential impacts associated with such a transition, it remains extremely challenging to reliably estimate the proximity to a critical threshold and to predict the AMOC's fate under future anthropogenic forcing. We argue that a global viewpoint on the dynamics beyond the detection of early-warning signals is needed for a robust risk assessment. Here we explore the phase space of an intermediate-complexity earth system model, PlaSim-LSG, featuring a multistable AMOC. For two different atmospheric carbon dioxide (CO₂) levels, we explicitly compute the Melancholia (M) state that separates the strong and weak AMOC attractors found in the model. The M state is a chaotic saddle embedded in the basin boundary between the competing states (an edge state). We show that, while being unstable, the M state can govern the transient climate for centuries. The M state exhibits strong AMOC oscillations on centennial timescales driven by sea ice and oceanic convection in the North Atlantic. Combining these insights with simulations under future CO₂ forcing scenarios (SSPs), we demonstrate that in our model the AMOC undergoes a boundary crisis at CO₂ levels projected to be reached in the next decade. Near the crisis, the AMOC behavior becomes highly unpredictable. Founded in dynamical systems theory, our results offer an interpretation of the so-called stochastic bifurcation recently observed in a CMIP6 earth system model under the same time-dependent forcing scenario.

How to cite: Börner, R., Mehling, O., von Hardenberg, J., and Lucarini, V.: Global stability of the AMOC under CO2 forcing: Boundary crisis, long transients and oscillatory edge states, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19783, https://doi.org/10.5194/egusphere-egu25-19783, 2025.

As the world enters a period of accelerated climate change, we require the rapid development of an early warning system (EWS) that identifies whether climatic conditions will result in reaching a tipping point. Tipping points represent critical thresholds where a small disturbance can cause a significant, qualitative shift in a system's state that can have crucial effects on human livelihoods. The impacts of political developments on future emission pathways, highlights the need for warning systems focused on climate risk communication that can be deployed and updated easily by policy teams with data pertaining to representative emission profiles. We are developing an early warning system to detect tipping points using a combination of observational and model data. In this abstract, we introduce the Tipsy-API platform; a dynamics-informed deep learning model to forecast relevant thresholds of the Greenland ice sheet and Atlantic Meridional Ocean Circulation. Following the objective of a “real time” warning system, our framework  iteratively updates forecasts with new observations to adjust the tipping point prediction accordingly. Finally, the framework will be deployed online and be available as an API, which we aim to be interactive and iteratively updated once new information about future warming becomes available. This ongoing work attempts to understand and address the requirements of a UK Government R&D funding agency, with the remit of engaging in high risk and high reward climate research. Thus, our project aims to both reduce uncertainty about tipping points and to allow for necessary open communication with policy makers and other relevant stakeholders.

How to cite: Roesch, C. and Last, C.: Dynamics-informed deep learning for tipping point forecasting, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20396, https://doi.org/10.5194/egusphere-egu25-20396, 2025.

How to cite: Ghasemi, B., Gayen, B., Vreugdenhil, C., and Sohail, T.: Slowing Down of the Atlantic Meridional Overturning Circulation Due to Excess Freshwater: Insights from Turbulence-Resolving Simulations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21133, https://doi.org/10.5194/egusphere-egu25-21133, 2025.

In China's drylands, deserts and areas prone to desertification constitute 44% of the landscape. The desert-oasis transition zone serves as a critical buffer between the desert interior and the oasis, playing an essential role in managing and preventing desertification. Despite its importance, the question of whether ecosystem functions exhibit multistability and experience regime shifts from functional to desertified states remains unresolved, particularly concerning the relationship between changes in vegetation patterns and ecosystem state transitions at the desert edges of arid and hyper-arid regions. In this study, we examined the stability landscapes of ecosystem multifunctionality and vegetation patterns in response to decreasing precipitation at both the inter-desert scale and within individual deserts, as the distance from the oasis to the desert interior increases. We compared the precipitation and distance thresholds for abrupt changes in vegetation pattern indices with those for regime shifts in ecosystem multifunctionality. Our analysis revealed that ecosystem multifunctionality can exist in both functional and desertified states when precipitation ranges between 104.37 mm and 152.56 mm. However, when precipitation drops below 104.37 mm, a complete shift from a functional to a desertified state occurs. The average precipitation threshold for abrupt changes in vegetation pattern indices—such as the size, shape complexity, and connectivity of vegetation patches, flow length, spatial skewness of the landscape, and the power law range, cutoff, and plexpo of the vegetation patch size distribution—is 201.69 ± 34.87 mm, which is higher than the threshold for ecosystem multifunctionality regime shifts. At the scale of individual deserts, changes in vegetation patterns precede regime shifts in ecosystem multifunctionality. These findings suggest that vegetation pattern indices can serve as early warning indicators for desertification in extremely arid desert-oasis transition zones. This study contributes to the enhancement of early-warning systems and supports the monitoring of desertification processes.

How to cite: Li, C. and Zhou, W.: Vegetation patterns as early warning signals for shifts in ecosystem multifunctionality in the desert-oasis transition zone of China’s drylands , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1786, https://doi.org/10.5194/egusphere-egu25-1786, 2025.

Abstract: Climate models often predict that more extreme precipitation events will occur in arid and semiarid regions, where plant phenology is particularly sensitive to precipitation changes. To understand how increases in precipitation affect plant phenology, this study conducted a manipulative field experiment in a desert ecosystem of northwest China. In this study, a long-term in situ water addition experiment was conducted in a temperate desert in northwestern China. The following five treatments were used: natural rain plus an additional 0, 25, 50, 75, and 100% of the local mean annual precipitation. A series of phenological events, including leaf unfolding, fruit setting (onset, summit and end), fruit ripening (onset, summit and end) and leaf coloration of the locally dominant shrub Nitraria tangutorum were observed from 2012 to 2018. The results showed that on average, over the seven-year-study and in all treatments water addition treatments advanced the leaf unfolding date by 1.29–3.00 days, but delayed the leaf coloration date by 1.18–11.82 days. Therefore, the length of the growing season was prolonged by 2.11–13.68 days. However, water addition treatments had no significant effects on all six fruiting events in almost all years, and the occurrence time of almost all fruiting events remained relatively stable compared with leaf phenology. The inter-annual variations of fruiting events were driven by the preceding flowering events rather than temperature or precipitation.

How to cite: Bao, F.: Contrasting responses of fruiting phenology and folia phenology to water additiontreatments in the Desert Shrub Nitraria tangutorum, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2220, https://doi.org/10.5194/egusphere-egu25-2220, 2025.

The study of how topographic and geological complexities drive vegetation dynamics over extended timescales, provides critical insights into the interactions between landscapes and ecosystems. Our study area encompasses the Greater Serengeti-Mara Ecosystem (GSME) in the Kenya-Tanzania transboundary region, renowned for its ecological richness and dynamic environments, most famously as the setting for the world’s largest terrestrial mammal migration. We focus on two case study regions: the Mara River Basin (MRB) and the Ngorongoro Conservation Area (NCA) to investigate localized interactions between geological, topographic, and ecological processes. The ecosystems are supported by a healthy and diverse vegetation cover, impacted by natural as well as anthropogenic factors. MRB is bounded in the north by active normal faulting dominated by Utimbara and Isuria faults whereas NCA is centred on Ngorongoro Crater, a large volcanic caldera. The tectonics of NCA is well-studied but subrecent faulting of Utimbara and Isuria was previously unrecognised and the impacts of these faults on uplift, subsidence and tilting of MRB has been revealed only recently. Previous studies have explored the relationship between precipitation and vegetation dynamics in the region. Limited research has focused on soil properties, primarily examining the effects of volcanic ash on the southeastern sector of GSME. However, the role of tectonics in influencing vegetation and, by extension, the broader ecosystem remains underexplored. We used remote sensing data (Landsat 5, 7, 8 and Sentinel 2) to create a time series analysis from the years 1984 until 2024 to examine the changes in the vegetation cover in the study area. Landsat 7 & 8 and Sentinel 2 data were processed in Google Earth Engine whereas those from Landsat 5 & 7 using Erdas Imagine. The normalised differential vegetation index (NDVI) shows a clear difference in vegetation cover during wet and dry seasons throughout the four decades for both the regions. MRB, which is covered by Quaternary sediments, has a higher vegetation cover throughout the year. NCA is affected by intermittent ash eruptions from Oldoinyo Lengai and has a vegetation cover, which varies at differing altitudes within the region and also shows a considerable seasonal variation at lower altitudes. Additionally, there is a significant difference in precipitation between MRB and NCA. In such a scenario, the vegetation cover in both the regions is likely to be a function of the interaction between the inherent soil properties and precipitation. Interestingly, stable vegetation also persists along active faults. Fault escarpments and fault-bounded wetlands provide seasonally stable vegetation cover, potentially due to localized influences on hydrology and soil properties and may serve as refugia during dry seasons. Our preliminary results highlight the need to integrate geo-tectonic analysis into broader ecosystem studies to better understand their role in sustaining biodiversity and ecosystem resilience.

How to cite: Purohit, C., Ludat, A., Said, A., Machunda, R., Hank, T., Kahle, B., and Kuebler, S.: Exploring the impacts of active faulting & tectonics on the vegetation cover of the dynamic Serengeti-Mara and Ngorongoro ecosystems of East Africa through spectral index analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3466, https://doi.org/10.5194/egusphere-egu25-3466, 2025.

Fairy circles are exceedingly regularly spaced barren circular patches in arid landscapes, typically encircled by a ring of taller grasses. These vegetation patterns occur in Southwestern Africa and Australia and have also been suggested to occur in North Africa, Middle East and Madagascar. The enigmatic origins of fairy circles in arid landscape shave intrigued ecologists and sparked heated debate about the two main competing hypotheses: the termite origin and vegetation self-organization hypotheses.

In the southern part of the Giribes Plains, Kunene region, northwest Namibia, fairy circles form in a distinctive, chain-like arrangement along drainage lines that run from north to south, closely aligned along a slope. These fairy circles are unusual in their extreme elongation, with the most extreme case measuring 32.5 meters long and only 7.7 meters wide. In contrast, the fairy circles in the rest of the Giribes outside the drainage lines are typically circular and exhibit a highly ordered, hexagonal pattern. Based on field work, remote sensing and mathematical modeling we explain the formation of these unique fairy circles.

The soil in the matrix between the circles is covered by physical crust, with some areas featuring a thin biocrust. This is the only place in Namibia where soil crust developed in the matrix. This crust causes the matrix soil to be nearly four times more compact than the soil within the fairy circles. The sand within the fairy circles is coarser (D₅₀ ~600 µm) compared to the matrix soil (D₅₀ ~300 µm), which supports the formation of the crust. Interestingly, sand in fairy circles not aligned with the drainage lines is also coarser (D₅₀ ~450 µm). Hydraulic conductivity, measured using a mini-disk infiltrometer, is three to four times greater within the fairy circles than in the surrounded matrix.

Building on these field observations, we hypothesize that the elongated shape of the fairy circles results from anisotropic soil water diffusion. Water diffuses more readily along the drainage lines than in the surrounding matrix, causing the fairy circles to expand more rapidly along the watercourses than laterally. To test this hypothesis, we used the mathematical model of Zelnik et al. (2015), which simulates biomass and soil water densities under varying water-soil diffusion coefficient ratios, r (r=1outside the drainage lines and r>1 inside the fairy circle) and precipitation rates .

The simulations indicate that, for moderate diffusion ratios and varying precipitation rates, elongated fairy circles form along the drainage lines, while circular fairy circles emerge when the diffusion ratio is lower. The results agree with remote sensing analysis of images take from a drone.  The stability of the pattern to different precipitation rates and r values was also studied. These results support the hypothesis that anisotropic soil water diffusion contributes to the elongated shape of the fairy circles in the Girbies plain, although other factors may also play a role. Indirectly our work supports the self-organization hypothesis for the origin of fairy circles.  The formation of the crust in the matrix remains is still an open question for future research.

How to cite: Yizhaq, H., Getzin, S., katra, I., Kamennaya, N., Peled, Y., and Meron, E.: The origin of the elongated fairy circles in the Giribes Plains, northwest Namibia, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3816, https://doi.org/10.5194/egusphere-egu25-3816, 2025.

Dryland aeolian landscapes are among the most vulnerable ecosystems under accelerating climate shift and land-use changes, where complex interactions between vegetation, soils, and landforms play a crucial role in maintaining ecosystem resilience and services. This study integrates remote sensing, field surveys, and numerical modeling to explore the coevolution of vegetation and aeolian landforms over the past four decades in East Asia’s arid regions, with a particular focus on the feedback mechanisms driving landscape stability in the arid zones under climatic and human forcing.
Analyses of aeolian landforms and climate systems in northern China reveal that declining wind speeds, associated with global terrestrial stilling, have significantly slowed dune migration rates over the past few decades, while widespread vegetation recovery has stabilized dune fields and mitigated desertification. Restoration practices, such as straw checkerboards, have accelerated vegetation recovery, increasing biodiversity and stabilizing soils, though soil fertility remains low compared to natural systems. Dust activity, an integral component of aeolian systems, have been suppressed in these areas, largely due to both climatic shifts and these large-scale restoration projects. Finally, high-resolution satellite images and field observations highlight how vegetation expansion modifies dune morphology through processes such as vegetation anchoring and sand transport alteration, leading to transitions from active to stabilized states. Conceptual models of vegetated dune morphodynamics provide insights into the role of vegetation-soil-landform feedbacks in shaping the arid landscapes.
This study emphasizes the interconnectedness of climate systems, vegetation dynamics, soil properties, and aeolian processes in maintaining ecosystem resilience and restoring ecosystem services. By linking dune morphologies and vegetation dynamics to thresholds of stability and nonlinear responses to climatic and anthropogenic pressures, the findings contribute to a deeper understanding of how dryland ecosystems adapt and evolve. These insights support more effective strategies for soil conservation, landform stabilization, and the restoration of ecosystem functions in the face of ongoing climate and land-use changes.

How to cite: Xu, Z., Wang, L., and Pang, X.: Deciphering Aeolian Landscape Dynamics: Vegetation Recovery and Dune Stabilization under Climatic and Human Influences, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4912, https://doi.org/10.5194/egusphere-egu25-4912, 2025.

Streamflow in several catchments in eastern Australia has decreased considerably (up to 40%) over the past 20 to 30 years, despite stable rainfall levels. This decoupling of streamflow and rainfall undermines the predictive accuracy of rainfall-runoff models used by catchment managers, which typically rely on the assumption of a stable relationship between these variables. Similar non-stationarity in streamflow has been observed in other catchments in the world, and evidence suggests that vegetation processes may be driving this non-stationarity due to increases in temperature and CO₂. Current rainfall-runoff models fail to capture the impact of these vegetation changes on evapotranspiration (ET). While these models account for ET's dependence on soil moisture, they do not consider changes in vegetation biomass and health, which can significantly alter ET and, consequently, runoff.

This contribution presents a methodology for estimating a vegetation-aware ET based on the Penman-Monteith equation and emulators that can account for changes in vegetation biomass and health. The emulators are developed using data from the Australian and New Zealand Flux Research and Monitoring network (TERN OzFlux). This network provides extensive measurements of energy, carbon, and water exchanges across various ecosystems, from which vegetation effects can be estimated under different environmental conditions, and across different vegetation types. Through this research we aim to contribute to understanding evapotranspiration dynamics and offer a reliable and simple tool for estimating vegetation effects, ultimately adding it to more realistic rainfall-runoff simulations.

How to cite: Brieva, C., Jorquera, E., Quijano, J., Kuczera, G., Saco, P., Rodriguez, J., and Kibria, G.: Accounting for Vegetation Feedbacks in Hydrological Models Using a New Vegetation-aware Evapotranspiration Formulation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5142, https://doi.org/10.5194/egusphere-egu25-5142, 2025.

Salt-tolerant Tamarix chinensis roots are crucial in preserving wetland soil and carbon  sequestration, which is essential for wetland ecology. Soil-water-salt conditions influence the growth of these roots in coastal saline areas, but the specific factors and their effects remain unclear. Using principal component and partial least square structural equation modelling (SEM) methods, we studied T. chinensis root features in six Yellow River delta communities. Results showed varied root features across locations, with larger roots further inland. Root growth negatively correlated with soil texture and salinity and positively with groundwater levels. Soil texture and salinity decreased with distance from the coast, while groundwater increased with distance from the Yellow River. This suggests that geographical location influences soil water-salt conditions, impacting root characteristics. The principal component analysis–derived root feature index captured 56.7% of root feature variation. SEM revealed geographical locations indirectly influence root features, with the Yellow River’s proximity primarily affecting them through groundwater and coastal distance influencing via soil sand content and salinity. The study underscores the importance of these findings for wetland conservation and ecology.

How to cite: lizhu, S.: Soil spatial heterogeneity created by river–sea interaction influences Tamarix chinensis root features in the Yellow River Delta, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5353, https://doi.org/10.5194/egusphere-egu25-5353, 2025.

Shrubs, characterized by their multiple dwarf stems, are a dominant plant functional type in arid and semi-arid regions, which cover 40% of Earth's land surface. These ecosystems are fragile and highly susceptible to climate change and human disturbances. The abundance of shrubs serves as an important indicator of ecosystem health, and their projected increase due to CO₂ fertilization and warming climates could significantly alter ecosystem functioning, exacerbate desertification, and impact essential ecosystem services. Monitoring shrub fractional abundance—the proportion of vegetative cover occupied by shrubs—is crucial for understanding these dynamics and guiding sustainable management practices. However, mapping shrub fractional abundance over large areas presents challenges due to their small crowns, sparse distribution, and high density, rendering traditional field surveys and conventional satellite remote sensing techniques inadequate. In this study, we propose an innovative two-step approach that integrates sub-meter resolution Google Earth (GE) imagery with decametric-resolution Sentinel-2 time-series data for accurate and scalable shrub fractional mapping. Our methodology consists of two main steps: (1) a semi-automatic process that uses GE imagery to delineate 1.31 million shrub crowns and generate high-quality training data, and (2) a machine learning model that combines spectral and phenological features from Sentinel-2 data to upscale GE-derived shrub fractional abundance across diverse arid and semi-arid landscapes in Inner Mongolia, China. The model achieved strong predictive accuracy (R² = 0.70), with phenological features—particularly during early May, mid-June, and late September—proving critical for distinguishing shrubs from seasonal vegetation. These periods correspond to key phenophases, including germination, peak growth, and senescence of grasses, which contrast with the perennial phenology of shrubs, highlighting the significance of phenology in differentiating shrubs from dynamic seasonal vegetation. Our results demonstrate the effectiveness of integrating multi-scale remote sensing data with machine learning to address existing limitations in shrub monitoring. This approach provides a scalable and transferable framework for global mapping of shrub fractional abundance, offering valuable insights into shrub encroachment and its implications for ecosystem health in the context of changing climatic and anthropogenic conditions.

How to cite: Liu, Z., Cao, X., Peñuelas, J., Descals, A., Yang, D., Liu, L., Su, Y., Liu, L., Chen, J., and Wu, J.: Mapping Shrub Fractional Abundance: A Multi-Scale Remote Sensing and Machine Learning Framework for Arid Ecosystem Monitoring, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5452, https://doi.org/10.5194/egusphere-egu25-5452, 2025.

Badlands are characterized by highly eroded, rugged landscapes with steep slopes, limited vegetation, and significant soil degradation. In badlands, vegetation plays a key role in erosion mitigation by intercepting runoff and acting as a significant factor in soil stability. However, vegetation dynamics in such an environment are determined by geomorphological factors like slope, erosion, sediment flux, and climatic conditions, characterized by temperature and precipitation patterns.

This study evaluates the significance of such drivers of vegetation transition within the badland systems using a State-and-Transition Model (STM) approach. This model predicts vegetation dynamics as a function of two basic processes: extinction (loss of vegetation) and colonization (vegetation growth over a barren patch of land). It is forced with vegetation states at four different time points (i.e., 1982, 1994, 2015, and 2021), while climate variables (e.g., temperature and precipitation), and sediment fluxes are averaged for the periods between these states. Geomorphological parameters (i.e., topographic elevation and slope) are assumed to be constant throughout the simulation period. It estimates vegetation transition probabilities using logistic regression. The model parameters are optimized through Bayesian methods (i.e., Markov Chain Monte Carlo algorithm) for climate conditions and geomorphology in the Laval catchment in the Draix-Bléone critical zone observatory, southeastern France. Model performance is quantified through repetitive training and testing to ensure the soundness of the predictions.

The results indicate that colonization is negatively impacted by higher slopes and annual sediment fluxes and is supported by increasing mean annual temperatures and summer precipitation. In contrast, vegetation extinction is driven mainly by geomorphic disturbances (e.g., slope and sediment fluxes during extreme events), while climatic factors seem to have little impact on vegetation extinction in this study area. Indeed, the forward prediction model, initiated using the 1982 vegetation state with best-fit parameters as forcing, resulted in a reasonably close match of the predicted states to the conditions observed, i.e., those of 1994, 2015, and 2021, which had an accuracy of ~0.8, with uncertainties of around ~0.35.

The present study integrates both geomorphological and climatic data to develop valid interpretations concerning environmental factors responsible for vegetation dynamics within badland topography, adding to an improved understanding of the ecosystem dynamics of these sensitive environments.

How to cite: Sharma, H., Le Bouteiller, C., and Boulangeat, I.: Geomorphic and climate-driven vegetation dynamics in badlands – A case study from Laval catchment, Draix-Bléone critical zone observatory, SE France, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6012, https://doi.org/10.5194/egusphere-egu25-6012, 2025.

The semi-arid lowlands of Baringo County, Kenya provide numerous ecosystem services to pastoral and agro-pastoral communities. However, these services have been significantly impacted by gradual changes in land cover, climate change, shrub encroachment, and invasion of grasslands by species like Prosopis juliflora. This study aimed to investigate how changes in land cover and roads affect vegetation dynamics between 2000 and 2024. This period was chosen due to the availability of consistent satellite-derived Normalized Difference Vegetation Index (NDVI) data for time series analysis. Using the Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), we analyzed NDVI data for five land cover classes, namely: natural shrubland, artificial grassland, forest, irrigated land, and Prosopis-infested areas. The impact of roads was assessed by calculating the instantaneous energy of high-frequency Intrinsic Mode Functions (IMFs) at buffer distances of 100, 250, 500, 1000, and 1500 meters from the roads in natural ecosystems. The results revealed diverse NDVI trends for different land cover classes. Forest exhibited mixed trends, with some pixels showing positive trends while others remained stable over time. Irrigated agricultural land indicated an increase in trend until 2017, after which it plateaued. Shrubland and artificial grassland maintained steady NDVI values with modest positive trends. Prosopis-infested areas exhibited a positive trend from 2000 to 2017, followed by a decline, likely linked to community-led invasion management efforts. The positive NDVI trends observed in forests and natural shrublands may be attributed to an increased invasion of Prosopis. Seasonal variations were associated with climatic conditions. Statistical analysis indicated that distance from the road had a significant difference on instantaneous energy but with a small effect size. These findings contribute to understanding how infrastructure and land use changes influence vegetation, providing valuable insights for sustainable management of semi-arid rural landscapes.

How to cite: Nyamari, N., Nitschke, S., Kramm, T., Otieno Ochuodho, D., Bareth, G., and Bogner, C.: Impact of Roads on Vegetation Dynamics in the Semi-Arid Baringo County, Kenya, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7446, https://doi.org/10.5194/egusphere-egu25-7446, 2025.

Mangrove wetlands are one of the most significant and vulnerable ecosystems in the world, providing a wide range of services including habitat, flood control and carbon storage, among others. Their vulnerability under climate change scenarios has been well documented, but recent works have shown that coastal wetlands have the capacity to accrete following the trend of SLR under particular circumstances. Suspended sediment concentration (SSC) plays a critical role in the accretion mechanisms that support wetland survival.

Wetlands in the Pacific Islands are among the most vulnerable areas to climate change and they receive considerable sediment from croplands (sugarcane) of their inland catchments. This contribution focuses on mangrove wetlands at the mouth of rivers draining into the Great Sea Reef. The objectives of our research are to evaluate the sediment loads from the catchment upstream of the coastal wetlands and to model the ecogeomorphological feedbacks among catchment, wetland and coastal reef lagoon under current conditions and future climate change scenarios. The methodology simulates the hydro-sedimentological behaviour of the watershed, under current and future scenarios with changes in land use (cropland expansion/management) and extreme events (cyclones). The output of this simulation constutute the input for the eco-geomorphological coastal wetland modelling.

This integrated modelling approach provides a better understanding of the main processes and feedbacks among vegetation, sediments and hydrodynamics within the coastal wetland, considering its interactions with the adjacent terrestrial (catchment) and aquatic (reef lagoon) ecosystems.

How to cite: Jorquera, E., Rodriguez, J., Saco, P., Sandi, S., Quijano, J., and Breda, A.: Effects of climate and human activities on mangrove wetland evolution., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7458, https://doi.org/10.5194/egusphere-egu25-7458, 2025.

The different temporal scales of surface flow and vegetation growth represent a major challenge when simulating the dynamics of ecohydrological systems. The much faster surface flow is therefore commonly simplified by treating it as stationary and linearizing the equations. While the simplified equations can be solved efficiently, they do not resolve individual runoff events. However, the sequence of events can be relevant for the dynamics of dryland vegetation. The nonlinear flow during individual precipitation events can be resolved by employing more sophisticated numerical methods, such as adaptive-time stepping and implicit time-integration. However, this requires the iterative solution of a sequence of discrete linear systems at each time step. This is complicated by asymmetry of the discrete system, originating from the advection of the flow. Here, we explore strategies for the efficient simulation of the nonlinear flow during individual precipitation events when modelling vegetation dynamics over centuries.

How to cite: Kästner, K. and Hinz, C.: Efficient and realistic modelling of individual runoff events in ecohydrological systems , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9937, https://doi.org/10.5194/egusphere-egu25-9937, 2025.

Spatial self-organization is a common response of arid and semi-arid ecosystems to water stress. It may result in periodic patterns such as dots, gaps and labyrinths, or in more irregular arrangements such as scale-free patterns, characterised by a power law distribution of patch sizes. As pattern formation occurs over large spatial domains, in the order of km² , it is often subject to heterogeneous environmental and soil conditions, which may lead to the anisotropic diffusion of resources.

In our project, we study the effects of anisotropic diffusion on pattern formation through the modelling of vegetation dynamics on complex network topologies. 

We employ the well-known reaction-diffusion vegetation model by Gilad et al. [1], in its simplified two-equation version by Zelnik et al. [2]. Two partial differential equations describe the dynamics of soil water and biomass densities.

In our implementation, the diffusive terms refer to network Laplacia, which allows us to the modify the topology on which the model operates.

When the diffusion networks of both water and biomass are regular two-dimensional lattices, we reproduce the observed progression of periodic patterns from gaps to labyrinths to dots for decreasing precipitation. 

To increase the complexity and connectivity of the network we implement the Watts-Strogatz small-world network model [3], in which a controlled number of random shortcuts is drawn over the two-dimensional lattice. Thus the number of shortcuts in the water and biomass diffusion networks become model parameters which may be used as proxies of heterogenous conditions affecting the diffusion of water and biomass respectively.

Our preliminary results show that an increase in anisotropic diffusion (number of shortcuts) has similar effects to an increase in isotropic diffusion in regards to the global variables of the ecosystem, such as average water and biomass densities. However, a small-world network topology induces the formation of steady-state non-periodic patterns, included scale free patterns, in a certain interval of network connectedness. 

Further, these steady-state scale free patterns appear unstable to the expansion of the largest gaps, leading to rapid desertification following a disturbance that may originate from grazing or human intervention. Hence, we uncover the existance of a bistability between two non-periodic patterns with very different ecological value. 

[1] E. Gilad, J. von Hardenberg, A. Provenzale, M. Shachak, and E. Meron. Ecosystem Engineers: From Pattern Formation to Habitat Creation. Physical Review Letters, 93(9):098105, 2004.

[2] Y. R. Zelnik, E. Meron, and G. Bel. Gradual regime shifts in fairy circles. Proceedings of the National Academy of Sciences, 112(40):12327–12331, 2015. 

[3] M. E. J. Newman and D. J. Watts. Scaling and percolation in the small-world network model. Physical Review E, 60(6):7332–7342, 1999.

How to cite: Filippini, S., von Hardenberg, J., and Ridolfi, L.: Vegetation patterning dynamics induced by non-local connections, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11850, https://doi.org/10.5194/egusphere-egu25-11850, 2025.

Dryland vegetation forms spatial patterns as an adaptation to water stress, driven by the uneven distribution of resources. While these patterns aid plant survival, herbivore grazing adds pressure, increasing desertification risks through vegetation loss and soil erosion. We present a novel model integrating vegetation patterning and herbivore grazing dynamics to explore their feedback loops over time. The model accounts for herbivore behaviors, including foraging, movement, and vegetation preferences. Using numerical continuation methods, we analyze solutions such as uniform and patterned vegetation-herbivore dynamics. A key finding is the emergence of traveling waves, where vegetation and herbivores propagate across the landscape. Herbivore distribution within these waves is asymmetric, causing uneven grazing stress. Surprisingly, this dynamic reduces overall grazing impact, enhancing vegetation sustainability compared to uniform grazing. Understanding these dynamics is vital for food security in drylands. By balancing herbivore populations and preserving vegetation, these interactions help mitigate drought and population growth challenges.

How to cite: Singha, J.: Traveling vegetation-herbivore waves can sustain ecosystems threatened by droughts and population growth, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19596, https://doi.org/10.5194/egusphere-egu25-19596, 2025.

Hydrological connectivity at the hillslope scale is a complex, spatially explicit phenomenon where surface and subsurface processes converge and interact, including infiltration, runoff, and lateral flow occurring during a singular rainfall event under specific antecedent soil moisture conditions.

In drylands, where Hortonian runoff generation prevails, such complexity has been conceptually simplified for operational purposes by using connectivity as a proxy for assessing ecosystem "health" or land degradation. Grounded in the current source-sink paradigm, a binary scheme of vegetation (pure sinks) and bare (pure sources) areas is used to distribute potential overland flow according to topography. The connectivity character is then distilled through the concept of Flow Length, with different metrics proposed under this rationale.

Despite this operational simplicity, the quantification of connectivity has yet to reach a standardized status, hindering intercomparison studies and the establishment of assessment baselines for land degradation.

Within the same framework umbrella, we recognize its shortcomings and propose decomposing the connectivity issue into three spatially explicit traits, each representing distinct structural features that emerge at the hillslope scale. This analytical approach aims to separately evaluate the contributions of vegetation patterns and flow routing, without the constraint of the hillslope shape. Facing the challenge of integrating these traits into a unified, synthetic metric for assessing runoff connectivity, we discuss several alternatives. The study is conducted at the experimental site in Benidorm (Alicante, Spain), using UAS-derived orthophotos and DEMs, where lateral variations within a small catchment serve to test the suitability of the proposal. This methodological proposal aims to advance the conceptual discussion toward developing a standardized approach for runoff connectivity evaluation and to inform land degradation assessments in drylands.

How to cite: Arnau-Rosalen, E., Rodriguez-Caballero, E., Marques-Mateu, A., Balaguer-Puig, M., Lopez-Carratala, J., Calvo-Cases, A., Lazaro-Suau, R., and Symeonakis, E.: Towards Standardising Runoff Connectivity Assessment at the Hillslope Scale in Drylands Using Structural Trait (De)composite., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19726, https://doi.org/10.5194/egusphere-egu25-19726, 2025.

Drying trends driven by climate change and water stress pose significant threats to ecosystem functioning and the services they provide to humanity. To better understand ecosystem response to drying trends, we study a mathematical model of plant communities that compete for water and light. We focus on two major responses to water stress: shifts in community composition to stress-tolerant species and spatial self-organization in periodic vegetation patterns. We calculate community bifurcation diagrams of spatially uniform and spatially periodic communities. The bifurcation diagram reveals that as precipitation decreases, spatially uniform community shift from fast-growing to stress-tolerant species. However,  a reverse shift back to fast-growing species occurs when a Turing bifurcation is traversed and patterns form. We further find that the inherent spatial plasticity of vegetation patterns, in terms of patch thinning along any periodic solution branch and patch dilution in transitions to longer-wavelength patterns, buffers further changes in the community composition, despite the drying trend, and yet increases the resilience to droughts. Response trajectories superimposed on community Busse-balloons highlight the roles of the initial pattern wavelength and of the rate of the drying trend in shaping the buffering community dynamics. We discuss the implications of these results for dryland pastures and crop production.

How to cite: Pavithran, I., Ferre, M., Bera, B., Uecker, H., and Meron, E.:  Vegetation pattern formation and community assembly under drying climate trends, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20458, https://doi.org/10.5194/egusphere-egu25-20458, 2025.

Combining satellite data from heterogeneous sources is often needed to produce synthetic indicators for natural resources and ecosystems. For instance, the temporal evolution of the water storage of a lake can be estimated by combining time series of satellite images (Sentinel-1, Sentinel-2) with measurements of the water surface elevation in 1D (nadir altimeters such as Sentinel-3 and 6) or 2D (swatch altimeter such as SWOT) and other available data.

These synthetic indicators are of great value for scientists, stakeholders, and policymakers, but quantifying their uncertainties bring new challenges. 

In the previous example, the estimation of the lake bathymetry is the most important stage as it determines the relationship between the observed water surface extent, the water surface elevation and the estimated relative water storage. 

We propose an optimal transport (OT)-based sensitivity analysis method to determine the probability distribution of the surface corresponding to the lake's observable bathymetry and the relative lake volume variations.
Taking two random variables X (known uncertainty input) and Y (unknown uncertainty output) on a measure probability space, along with their marginals ν and μ, we can define an OT-based global sensitivity measure between a sample Xa of X, and Y, based on the Wasserstein distance of order p between ν and μ.

Most sensitivity analysis methods only cater to the study of a univariate sensitivity measure, whereas using an OT-based sensitivity index allows the study of a multivariate response. Various metrics will be investigated in the presentation (Wasserstein-Bures metric, Wasserstein-Fréchet...) along with their advantages and drawbacks relatively to our multi-dimensional problem.

How to cite: Levillain, J. and Gasnier, N.: Uncertainty propagation on satellite data for synthetic indicators, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1302, https://doi.org/10.5194/egusphere-egu25-1302, 2025.

To understand the mechanisms underlying the magnetic fields of planets, including the Earth and the Sun, previous studies have proposed disk dynamo models as a suitable reduction of the mean-field dynamo equations. One of these is the Rikitake-Hide model, which combines the Rikitake model (a two-disk model for geomagnetic reversal) and the Hide model (a disk model with a motor and mechanical friction). This model reproduces nearly equal intervals of magnetic field reversals with modulated cycles, resembling the fluctuations of sunspots. In this presentation, we discuss the Jacobi stability and aperiodicity of the Rikitake-Hide model using geometrical invariants in the KCC (Kosambi–Cartan–Chern) theory. In the KCC theory, the second and third KCC invariants are related to the Jacobi stability and aperiodicity of trajectories in the system, expressed through variables (electrical currents and angular velocities) and parameters. By calculating the Jacobi stability of the Rikitake-Hide model using the second KCC invariant, we find that the model is Jacobi unstable when fluctuations in magnetic energy (square of the electric current) reach local minima. The instability at local minima manifests as branches in the trajectories of electric currents in the model. Based on the third KCC invariant for the trajectories of electric currents in the Rikitake-Hide model, the aperiodicity of the model may arise from individual electric currents. Although the model is always accompanied by aperiodicity, nearly equal intervals of magnetic field reversals are preserved through the cancellation of aperiodicity by symmetric cyclical fluctuations of electric currents. On the other hand, asymmetric electric currents originating from the motor and mechanical friction in the model alter the periods of magnetic field intervals. Finally, we consider the correspondence between the fluctuations of magnetic energy in the Rikitake-Hide model and sunspots on the Sun and discuss the implications for solar activity.

How to cite: Hirano, M., Nagahama, H., and Yajima, T.: Jacobi stability and aperiodicity of the Rikitake-Hide dynamo model based on KCC-theory, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2556, https://doi.org/10.5194/egusphere-egu25-2556, 2025.

The Maximum Likelihood Ensemble Filter (MLEF) with State Space Localization (MLEF-SSL) was recently developed as a new ensemble data assimilation method that incorporates state space covariance localization in model space. The main motivation for developing this method was to enable global numerical optimization and assimilation of vertically integrated observations in an ensemble data assimilation system with covariance localization. MLEF-SSL uses random projection to compute the localized forecast error covariance and reduce the analysis dimensions to a manageable space. MLEF-SSL is being applied to a global NWP system named Nonhydrostatic Icosahedral Atmospheric Model (NICAM) with the assimilation of atmospheric observations, i.e., the NICAM-MLEF-SSL system, to explore its capability under a realistic high-dimensional dynamical application.

To apply MLEF-SSL in a high-dimensional system such as NICAM, a substantially large number of ensembles of the order of O(10⁴) - O(10⁵) is necessary to represent the localized forecast error covariance, thus, requiring special attention on the algorithmic development for the NICAM-MLEF-SSL system. This presentation will discuss the practical implementation of the NICAM-MLEF-SSL system on the RIKEN supercomputer Fugaku with an emphasis on the use of advanced math libraries for parallel computing.

In addition, the performance of NICAM-MLEF-SSL in assimilation of real atmospheric observations will be evaluated in detail and compared to that of NICAM-LETKF, which is the global data assimilation system that is currently used at RIKEN. 

Future plans related to the use of strong dynamical balance constraints and Artificial Intelligence (AI) techniques in NICAM-MLEF-SSL will also be discussed.

How to cite: Wu, T.-C., Zupanski, M., and Miyoshi, T.: Global Atmospheric Ensemble Data Assimilation using NICAM global icosahedral model and Maximum Likelihood Ensemble Filter with State Space Localization: Real Observation Experiment, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2633, https://doi.org/10.5194/egusphere-egu25-2633, 2025.

Local dimension computed using Extreme Value Theory (EVT) is usually used as a tool infer dynamical properties of a given state ζ of the chaotic attractor of the system. The dimension computed in this way is also known as the pointwise dimension in dynamical systems literature, and is defined using a limit for infinitely small neighborhood in the phase space around ζ. Since it is numerically impossible to achieve such limit, and because dynamical systems theory predicts that this local dimension is almost constant over the attractor, understanding the properties of this tool for a finite scale R is crucial. We show that the dimension can considerably depend on R, and this view differs from the usual one in geophysics literature, where it is often considered that there is one dimension for a given dynamical state or process. We also systematically assess the reliability of the computed dimension given the number of points to compute it.

This interpretation of the R-dependence of the local dimension is illustrated on the Lorenz 63 system for ρ = 28, but also in the intermittent case ρ = 166.5. The latter case shows how the dimension can be used to infer some geometrical properties of the attractor in phase space. The Lorenz 96 system with n = 50 dimensions is also used as a higher dimension example. A dataset of radar images of precipitation (the RADCLIM dataset) is finally considered, with the goal of relating the computed dimension to the (un)stability of a given rain field.

How to cite: Bonte, M. and Vannitsem, S.: Finite-size local dimension as a tool for extracting geometrical properties of attractors of dynamical systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4017, https://doi.org/10.5194/egusphere-egu25-4017, 2025.

How to cite: Brolly, M. T.: Online learning of stochastic closures of quasigeostrophic turbulence, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4165, https://doi.org/10.5194/egusphere-egu25-4165, 2025.

The weak connections between individual blocks in blocky rock mass lead to the possibility of block rotations. Rotation of non-spherical blocks pushes the adjacent blocks away; the phenomenon termed elbowing [1]. Furthermore, rotations in both directions create the lateral displacements in the same direction such that the behaviour becomes non-linear of absolute value type. In order to investigate the resulting pattern of elbowing blocks, we considered a simple one-dimensional structure (chain) of stiff square blocks holding together by springs. Only the end block (driving block) with a fixed centroid position is rotated, while each passive block has two degrees of freedom: translational and rotational. A one-dimensional physical model, analytical model, and numerical model (using the commercial discrete element software UDEC) are developed. It was observed that the passive blocks do not consistently rotate in the same direction as the driving block; instead, when the driving block reaches a certain critical angle, it triggers the change of the direction of rotation and the blocks start inverse rotation. We define the angle of the driving block, when the rotation of the passive blocks returns to zero, as the ‘angle of interior zero’. It was found that this angle is influenced by the magnitude of contact friction and the number of blocks. The result demonstrates the complexity of block motion pattern even in such a simple blocky system.

[1] Pasternak, E., Dyskin, A.V., Estrin, Y. (2006) Deformations in transform Faults with rotating crustal blocks. Pure Appl. Geophys. 163 2011-2030.

Acknowledgement: The authors are grateful to Dr I. Shufrin and the School of Engineering workshop for the help with designing and manufacturing of an initial version of the experimental device.

How to cite: Zhang, M., Dyskin, A., and Pasternak, E.: Triggering inverse rotation in 1D model of blocky rock mass, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4782, https://doi.org/10.5194/egusphere-egu25-4782, 2025.

Real rocks are heterogeneous and contain cracks of various scales. The influence of these cracks on hydraulic fracturing was first studied in the 1980s, when experimental studies showed that the presence of natural fractures can affect the trajectory of the hydraulic fracture. Another type of inhomogeneous material is a homogeneous matrix with fragments of a different material with significantly different mechanical and filtration properties. By varying the composition and size of these fragments, it is possible to create samples with different characteristics. The interface between the matrix and the fragments acts as a weakened area, similar to cracks in a nonuniform fractured material.

Here we present the results of experiments conducted to study the propagation of hydraulic fracture in a heterogeneous material composed of a mixture of gypsum, cement, and fillers. Marbles chips and fine gravel were used as additives in the mixture. The porosity of the original samples without additives was 0.44 and the permeability was 4.15 mD. The porosity and permeability of samples with marble chips were 0.32 and 6.9 mD, respectively, while those with gravel were 0.30 and 2.74 mD. The height of the samples was 60 mm with an outer diameter of 104 mm. Brass inserts with an internal diameter of 10 mm and a length of 18 mm were placed in the center of both sides of each sample during manufacturing. The length of the uncased part of the well was approximately 24 mm.  The prepared sample was placed between two circular aluminum bases, with piezoelectric transducers mounted in the recessed working surfaces. Silicone liquid PMS-5, with a kinematic viscosity of 5 centistokes, was used to saturate the sample. During hydraulic fracturing, silicone liquid PMS-200, with a viscosity of 200 centistokes, was applied. The samples were subjected to radial stresses of 0.5 MPa, along the lateral surface, and axial stresses of 3 MPa were applied along the axis. The main focus of the experiments was on studying the acoustic emission accompanying the propagation of the hydraulic fracturing crack.

The experiments have shown that the presence of inclusions significantly affects the shape and development of hydraulic fractures: under conditions of radially symmetric lateral compression, fractures with three branches form in heterogeneous materials, while fractures with two wings form in homogenous samples. The fracturing pressure in samples with fillers is higher than in homogenous ones, and the highest fracturing pressure values are achieved when there is a more rigid and durable filler present. It has been established that acoustic emission pulses in an inhomogenous material have a wider frequency range than in a homogeneous one. In samples containing fillers, more acoustic emission pulses are recorded, especially when there are rigid fillers present. It is also worth noting that acoustic emission occurs earlier in experiments with homogenous samples relative to the time at which the maximum pressure is reached during hydraulic fracturing. The results of the acoustic emission source location are consistent with the observed patterns of cracks on the surface of the samples.

How to cite: Turuntaev, S., Zenchenko, E., Chumakov, T., and Zenchenko, P.: Laboratory study of hydraulic fracturing in heterogeneous materials, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5113, https://doi.org/10.5194/egusphere-egu25-5113, 2025.

Mechanics of rock failure in uniaxial compression remains a challenge since the wing cracks produced by pre-existing defects/cracks in uniaxial compression are shown to wrap around the initial defect effectively arresting their further growth. As a result, they cannot grow to the extent sufficient for splitting rock samples. This presentation proposes another mechanism of rock failure based on extensive fracture growth caused by zones of tensile stresses formed as parts of self-equilibrating stress field induced by defects distributed in rock (both pre-existing and produced in the process of loading). The fracture driven by localised tensile stresses grows avoiding the zones of compressive stresses thus forming areas of interruption or overlapping. These areas work as distributed bridges constricting the fracture opening. Fractures with constricted opening have the stress intensity factors increasing with fracture growth making the fracture growth unstable and capable to break the rock.

Due to the necessity to avoid the compression zones the growing fracture will deviate form the straight path. If the sample size is not sufficiently large as compared to the size of the compression zones these deviations can be seen as inclined in one direction making the fracture oblique. This can be passed for the conventional shear fracture, which in fact cannot play a role in the process, as the shear fractures do not grow in their own plane forming wings instead. Therefore, the fractures observed in rocks failed in uniaxial compression, in both splitting and oblique failure types are tensile fractures with constricted opening formed and driven by the stress field induced by distributed defects. Whether the resulting failure is splitting or oblique (“shear” failure) depends upon the ratio of the sample size to the compression zones dimensions. We term this dependence “the scale effect in failure type”.

The proposed concept will form a basis for developing models of rock failure in compression necessary for analysing and predicting large scale failures in the rock mass, especially during mining operations.

How to cite: Dyskin, A. and Pasternak, E.: Defect-induced stress field triggering extensive tensile fracture growth in uniaxial compression, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5131, https://doi.org/10.5194/egusphere-egu25-5131, 2025.

The presentation proposes a 2D model of sliding over a rough fault of finite length. Sliding over such a fault involves interaction between asperities of its opposite sides. The interaction proceeds in two stages. At Stage 1 (the initial stage) the interaction involves climbing asperities over the asperities on the opposite side of the fault, causing dilation. As a result, additional resistance to sliding is induced working as friction such that the friction angle becomes the sum of the material friction angle representing friction between the asperity contacts and the average angle of inclination of the tangential line of the asperity contacts. With the increase of shear stress above the critical value of the friction stress shear over the fault is initiated. This corresponds to the conventional interpretation of sliding over rough fault.

This conventional sliding however only proceeds until it reaches the half asperity length, that is when the corresponding local dilation (fault opening) reaches its maximum, which is the asperity height. Under uniform shear stress this is obviously reached at the fault centre. At this point Stage 2 commences. At Stage 2 the asperities which produced maximum dilation (the size of two asperity height) start reducing the dilation creating the effect of local negative friction since the pressure at this stage drives sliding. Subsequently, in the fault zone sliding in Stage 2 the average friction angle drops to that of the material friction angle. As a result, the central zone slides under reduced friction. It was found that there exists a critical magnitude of shear stress, which triggers self-sustained extension of the zone with reduced friction.

Sliding of this type involves creation of loci of negative friction, which can affect the microseismic signals. The presented model provides an additional mechanism of fault sliding and the associated induced seismicity.

How to cite: Pasternak, E. and Dyskin, A.: Rough fault with dilation. Triggering of full sliding, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5137, https://doi.org/10.5194/egusphere-egu25-5137, 2025.

Perhaps the most dynamic component of the Arctic sea ice cover is the marginal ice zone (MIZ), the transitional region between dense pack ice and open ocean. It widens severalfold and moves poleward in a dramatic annual cycle, impacting the climate system, ecological processes, and human accessibility to the Arctic. We’ll discuss multiscale partial differential equation models for MIZ dynamics and the sea ice concentration field. The MIZ is viewed as a liquid-solid phase transition region, or mushy layer, to obtain a model that captures the seasonal cycle. Parameters in the model depend on finer scale structure and are computed using rigorous homogenization methods. We also consider a related floe-scale model with advective forcing to study ice transport processes, jamming, and anomalous diffusion observed in ice floe GPS data.

How to cite: Golden, K.: Multiscale PDE Models for Marginal Ice Zone Dynamics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5286, https://doi.org/10.5194/egusphere-egu25-5286, 2025.

Magnetotelluric (MT) sounding is a vital geophysical exploration method, renowned for its ability to investigate deep geological structures, detect low-resistivity anomalies, and support diverse applications. It is widely used in mineral and geothermal resource exploration, hydrogeological surveys, and deep structural studies, effectively delineating subsurface structures from hundreds of meters to hundreds of kilometers. However, MT inversion, essential for quantitatively analyzing subsurface electrical structures, faces challenges due to the inherent non-uniqueness of MT methods and noise in field data. Traditional deterministic inversion methods, which rely on gradient-based optimization, typically produce a single model or a set of models fitting the data but fail to quantify uncertainties, complicating interpretation and reducing reliability.

To address these challenges, this study adopts a Bayesian inference framework for MT inversion. Unlike deterministic methods, Bayesian inversion treats model parameters as random variables and iteratively updates their prior distributions using observational data, ultimately obtaining posterior probability distributions. This approach enables a quantitative assessment of inversion uncertainties. However, the computational demands of Bayesian inversion, particularly for 2D and 3D MT problems, pose significant challenges, with forward modeling efficiency being a critical bottleneck.

To overcome this, we propose an efficient MT forward modeling method based on the Extended Fourier Deep Operator Network (EFDO), a physics-informed neural operator network. EFDO leverages the principles of Fourier transforms and deep neural operator networks to learn the functional mapping between model conductivity inputs and MT forward responses. By embedding physical laws into the network, EFDO ensures accurate predictions while significantly improving computational efficiency. Once trained, EFDO predicts forward responses in milliseconds, achieving a speedup of 300 times compared to traditional Finite Volume Methods (FVM) while maintaining high accuracy. A multi-GPU distributed parallel training strategy further accelerates EFDO training, drastically reducing preparation time.

Additionally, we integrate Voronoi and Delaunay parameterization techniques with the reversible-jump Markov Chain Monte Carlo (rjMCMC) method to enhance model sampling efficiency. This establishes a robust 2D MT trans-dimensional Bayesian inversion framework. Numerical experiments and tests on the Coprod2 dataset demonstrate the method’s computational efficiency and reliability.

In summary, this study introduces a novel approach combining the physics-informed EFDO network and advanced parameterization techniques to improve efficiency and uncertainty quantification in MT Bayesian inversion, paving the way for rapid, high-dimensional geophysical exploration.

How to cite: Liao, W., Hu, X., and Peng, R.: Efficient 2D MT Forward Modeling and Trans-Dimensional Bayesian Inversion with Physics-Informed Neural Operator Networks, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7723, https://doi.org/10.5194/egusphere-egu25-7723, 2025.

The primitive three-dimensional viscous equations for atmospheric dynamics with the phase transformation of water vapor are studied.
According to the actual physical process, we give the heating rate, mass of water, and precipitation rate, which are related to temperature and
pressure. In fact, in this system, the anelastic approximation is not used, and more complex boundary conditions and dissipation coefficients are given. Providing H2 initial data and boundary conditions with physical significance, we prove the local well-posedness of a unique strong solution to the moist atmospheric equations by the contractive mapping principle and the energy method in the H2 framework.

How to cite: Ma, J., Lian, R., and Zeng, Q.: Local well-posedness of strong solution to a climate dynamic model with phase transformation of water vapor, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8234, https://doi.org/10.5194/egusphere-egu25-8234, 2025.

Understanding and predicting Earth system processes requires more than accurate forecasts—it demands uncovering the underlying relationships among variables and constructing models that are interpretable, physically consistent, and mathematically robust. While machine learning has demonstrated exceptional predictive capabilities, its models often neglect fundamental physical laws, raising concerns about reliability, interpretability, and trust. This work explores the integration of domain knowledge with machine learning through hybrid and causal modeling approaches, aiming to bridge data-driven insights with the principles of the physical sciences. By leveraging these methods, we can enhance our understanding of the data-generating processes and achieve results that are both consistent and explainable. I will present recent advances and strategies in this field, highlighting their potential to revolutionize Earth system research. This effort represents a step toward a long-term AI agenda for developing algorithms that drive knowledge discovery in Earth sciences.

https://arxiv.org/pdf/2010.09031.pdf

https://arxiv.org/abs/2104.05107.pdf

https://doi.org/10.1016/j.physrep.2023.10.005

How to cite: Camps-Valls, G.: Advancing AI for Earth sciences with hybrid and causal models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8388, https://doi.org/10.5194/egusphere-egu25-8388, 2025.

Air pollutant emissions represent key input information for modeling atmospheric chemical composition or evaluating air pollution control policies. Over the last decades, different approaches have been proposed for estimating emissions. These approaches include collecting activity data combined with emission factors to build bottom-up emission inventories, developing appropriate spatial and temporal disaggregation proxies applied to national or regional total estimates to construct top-down emission inventories, and creating complex data assimilation workflows around chemistry-transport models combined with observations to perform air pollution emission inverse modeling.

In the last decade, deep neural networks (DNNs) have demonstrated exceptional ability to model complex spatiotemporal data. Meanwhile, advances in Earth observation systems, such as the Tropospheric Monitoring Instrument (TROPOMI), have enabled the collection of high-resolution atmospheric composition data in near real-time. These developments open up opportunities to integrate the predictive power of DNNs with satellite observations to deliver rapid and accurate estimates of pollutant emissions in near real-time.

Deterministic CTMs offer insights into the forward relationship between emissions and atmospheric composition, and some studies are already suggesting that DNN might be able to estimate with reasonable predictive skills the chemical concentrations obtained from these physics-based models. While the forward mapping is well-defined, the inverse mapping—from atmospheric composition to emissions—is inherently ill-posed. Our objective is ultimately to exploit DNNs for doing air pollutant emission inverse modeling, without using the traditional data assimilation approach. This presents challenges, requiring the application of regularization techniques to address the ambiguity raised by this ill-posed problem.

Here, we will present the preliminary results of our study on training regularized DNNs for inverting the NOx emissions in Spain , utilizing training data derived from the MONARCH air quality model. We will take advantage of the flexibility offered by these models to create different training datasets and assess the performance of our models across different data scenarios.

How to cite: Petticrew, J., Petetin, H., Mas Magre, I., Guevara Vilardell, M., Jorba, O., and Pérez García-Pando, C.: Inverting Spanish NOx emissions using a neural network, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8593, https://doi.org/10.5194/egusphere-egu25-8593, 2025.

We use a quasi-geostrophic model of the wind-driven double-gyre to explore qualitative changes in the system's behavior through a novel topological framework called Templex. This method characterizes and classifies the system's attractors in phase space across various regimes, from low-energy, temporally smooth dynamics to highly energetic and chaotic states. Our study focuses on stationary wind stress forcing to deepen the understanding of the underlying dynamics, starting from this simplified scenario to identify the physical processes involved, while the spatial resolution is gradually increased up to 5 km x 5 km.

The original system of partial differential equations describing the flow is converted into a set of ordinary differential equations using finite-difference methods. We take advantage of the Julia programming language to build the model, apply continuation methods, and perform stability analyses along the branches of a bifurcation tree, subject to pseudo-adiabatic variations in wind intensity. Our findings emphasize the effectiveness of topological methods in revealing the structural aspects of bifurcations, and in examining new pathways to study dynamical systems in geosciences. Moreover, we present novel insights on the existence of an attractor in the infinite-dimensional system, bridging the gap between topological results obtained numerically and the original mathematical model.

Through this presentation, we aim to foster discussion on the potential of topological approaches to advance the understanding of nonlinear systems in geophysical fluid dynamics.

How to cite: Bodnariuk, N., Sciamarella, D., Speich, S., Simonnet, E., and Ghil, M.: The Wind-Driven Double-Gyre Circulation: A topological characterization of attractors across regimes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11928, https://doi.org/10.5194/egusphere-egu25-11928, 2025.

The source term of Chernobyl 2020 wildfires is a tensor consisting of five dimensions: spatial location described by longitude and latitude in a given area with potentially many sources, time profiles, height above ground level, and the size of particles carrying the material. Since the number of concentration measurements is limited, the estimation of this source term is an ill-posed problem. Prior information is thus essential to obtain a reproducible estimate. We show that deep image prior that utilizes the structure of a deep neural network to regularize the inversion is suitable for this problem. The deep network is initialized randomly without the need to train it on any dataset first. The networks is used to represent both the mean and variance of the posterior estimate. The resulting variational Bayes procedure thus introduces smoothness in the spatial estimate of the emissions and reduces the number of unknowns by enforcing a prior covariance structure in the source term. The estimate of the 137Cs emissions during the Chernobyl wildfires in 2020 is compared to the Tikhonov method. The spatial distribution of the proposed method is close to the distribution obtained from satellite observations. 

Acknowledgment: 
This research has been supported by the Czech Science Foundation (grant no. GA24-10400S). 

How to cite: Smidl, V., Brozova, A., Tichy, O., and Evangeliou, N.: Estimation of Spatial-temporal Source Term of Chernobyl Wildfires using Deep Neural Network Prior, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11939, https://doi.org/10.5194/egusphere-egu25-11939, 2025.

An important component of Earth's climate system is the land surface atmosphere interaction. For theoretical and empirical reasons, the strongest constraints to vegetation productivity, and thus drawdown of atmospheric CO2, as well as water use (evapotranspiration) are edaphic in nature. The appropriate treatment of the soil is related only to one of the topics listed in the call for abstract submissions, namely statistical mechanics. A new water balance treatment developed based on combining ecological optimality with distinct percolation scaling results for solute transport in directed as well as random networks as applied to root growth and soil formation delivers observed results for net primary productivity of vegetation, streamflow elasticity, evapotranspiration, plant species richness, and the scale dependence of the water cycle with a single parameter, which can be evaluated in a 2D or a 3D fractal model of plant roots. Most of the results summarized are generated from the 2D (thin soil) model, meaning that most of the above are consistent with predictions that use no adjustable or unknown parameters at all, merely their universal values from percolation theory.

How to cite: Hunt, A. G. and Ghanbarian, B.: With only stochastic and non-linear dynamics methods, the Earth surface component of climate cannot be modeled correctly, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12251, https://doi.org/10.5194/egusphere-egu25-12251, 2025.

We first present briefly recent insights on the effects of time-dependent forcing on systems with intrinsic variability, such as anthropogenic forcing on the climate system. These insights are applied next to the problem of periodic forcing of the wind-driven double-gyre problem. The topological perspective here is provided by applying recent advances in the algebraic topology of autonomously chaotic dynamic systems subject to time-dependent forcing. These advances are applied to the problem at hand.

The application starts by finding a topological representation of the underlying structure of the system’s flow in phase space by the construction of a cell complex that approximates its branched manifold and of a directed graph on this complex. The directed graph corresponds to the way that the flow in phase space moves from one cell of the complex to another.

Fundamental ingredients of the above representation, called generatexes and stripexes, delineate distinct ways of following the dynamical paths on the complex, namely the nonequivalent ways of travelling through the flow in phase space. These mathematically defined pathways will be shown to correspond to physical modes of variability.

How to cite: Ghil, M., Charó, G., Sciamarella, D., Ruiz, J., and Pierini, S.: A topological perspective on the wind-driven ocean circulation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12270, https://doi.org/10.5194/egusphere-egu25-12270, 2025.

By its simplicity and intuitive appeal, the geometric interpretation of analysis provides a complementary understanding of minimum variance estimation.  The geometric interpretation is made possible by using a Hilbert space representation of random variables.  In this presentation we will argue how actually a geometric approach can help to explore/discover new relationships, in identifying assumptions, and provide an alternative pathway of understanding the concept of analysis and estimation of error covariances.  
For example, relationships between analysis increments in cross-validation could be easily derived.  An interpretation of sequential observation processing also follows a simple interpretation.  Important considerations in establishing relationships for an arbitrary number of collocated data sets could also be established.  Then we examine how we can relax the assumption of an optimal analysis.  This will guide us in deriving a new diagnostic of observation statistics with correlated errors.  

How to cite: Ménard, R., Deshaies-Jacques, M., and Vogel, A.: A geometric interpretation of analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13502, https://doi.org/10.5194/egusphere-egu25-13502, 2025.

The Paris Agreement legally commits the international community to keep anthropogenic global warming well below 2.0°C, while major efforts shall be made to hold the 1.5°C-line. This is supported by ample scientific evidence indicating that climate change becomes difficult to manage beyond those lines due to highly nonlinear impacts (like tipping processes). The time range when the Paris guardrails will be transgressed under business as usual is highly relevant for precautionary adaptation measures and for justifying the rapid transformation towards a post-fossil economy. Fully-fledged Earth System Models (ESMs) are usually employed for the pertinent projections, yet they are not only computationally expensive but also lack explicit accounting for natural climate system variability. The latter may significantly distort (and invalidate) the ESM overshoot-timing projections. As an alternative, we present here a purely data-driven stochastic approach based on the persistence properties of the observed global temperatures. We are able to quantify, in a probabilistic way, the natural variability that must be superimposed on the anthropogenic trends in order to retrieve the observed warming behavior. When assuming that the anthropogenic warming continues at the current rate, we actually arrive at comparable overshoot timing estimates as the ESMs and provide an explanation for this finding.

How to cite: Ludescher, J., Yuan, N., Schellnhuber, H. J., and Bunde, A.: Natural variability-focused assessment of climate overshoot timing, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13518, https://doi.org/10.5194/egusphere-egu25-13518, 2025.

Reconstructing spatiotemporal systems from sparse observations remains a long-standing challenge in several domains, including geoscience, air pollution and fluid dynamics. While various data assimilation (DA) and machine learning (ML) methods have shown potential, they still face significant challenges (see [1]): 

• The computational burden of conventional DA algorithms (including error covariance specification), particularly for multivariate, high-dimensional systems.
• Sparse and movable sensor placement, which makes conventional ML models (typically requiring fixed and regularly distributed input data) cumbersome.
• The ill-defined nature of the sparse reconstruction problem, which poses significant risks of overfitting.

We will present our recent works aimed at addressing these challenges. More specifically, we developed latent DA algorithms [2] to reduce the computational burden of variational DA methods. These algorithms demonstrate great potential in efficiently assimilating sparse observations within a reduced-order latent space constructed by neural networks, thanks to the TorchDA library [3]. The latter enables GPU implementation of mainstream data assimilation methods and supports non-explicit state-observation transformation functions, provided they can be learned by a neural network.

We have also employed advanced deep learning techniques, including Voronoi-tessellation CNNs [4] and Vision Transformer-based autoencoders [5], to learn mappings from sparse observations to the complete physical space. These approaches effectively address challenges such as movable sensor placements and varying sensor numbers. Their integration with DA algorithms has also been evaluated.

Finally, our recent work explores [6] the utility of generative AI techniques, particularly denoising diffusion models, for field reconstruction from sparse observations. Generative AI methods offer two main advantages: first, they produce a sample from a probability distribution rather than predicting the mean as a fixed output, which can help mitigate overfitting caused by the illy-defined problem. Second, they inherently function as ensemble predictors by generating several samples, facilitating uncertainty quantification, which is essential in data assimilation. The numerical results tested on cases ranging from fluid dynamics benchmarks to semi-operational air pollution simulations will also be discussed.

[1] Cheng, S., Quilodrán-Casas, C., Ouala, S., Farchi, A., Liu, C., Tandeo, P., Fablet, R., Lucor, D., Iooss, B., Brajard, J., Xiao, D., Janjic, T., Ding, W., Guo, Y., Carrassi, A., Bocquet, M. and Arcucci, R, 2023. Machine learning with data assimilation and uncertainty quantification for dynamical systems: a review. IEEE/CAA Journal of Automatica Sinica

[2] Cheng, S., Chen, J., Anastasiou, C., Angeli, P., Matar, O.K., Guo, Y.K., Pain, C.C. and Arcucci, R., 2023. Generalised latent assimilation in heterogeneous reduced spaces with machine learning surrogate models. Journal of Scientific Computing

[3] Cheng, S., Min, J., Liu, C. and Arcucci, R., 2025. TorchDA: A Python package for performing data assimilation with deep learning forward and transformation functions. Computer Physics Communications

[4] Cheng, S., Liu, C., Guo, Y. and Arcucci, R., 2024. Efficient deep data assimilation with sparse observations and time-varying sensors. Journal of Computational Physics

[5] Fan, H., Cheng, S., de Nazelle, A.J. and Arcucci, R., 2024. ViTAE-SL: a vision transformer-based autoencoder and spatial interpolation learner for field reconstruction. Computer Physics Communications 

[6] Zhuang, Y., Cheng, S. and Duraisamy, K., 2025. Spatially-aware diffusion models with cross-attention for global field reconstruction with sparse observations. Computer Methods in Applied Mechanics and Engineering

How to cite: Cheng, S., Fan, H., Zhuang, Y., Finn, T. S., Lugon, L., Sartelet, K., Duraisamy, K., Arcucci, R., and Bocquet, M.: Hybrid data assimilation and machine learning algorithms for sparse observational data, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13544, https://doi.org/10.5194/egusphere-egu25-13544, 2025.

Bayesian inverse modelling systems are a valuable tool for quantifying sources and sinks of greenhouse gases using atmospheric observations. They are increasingly used to verify national emission inventories submitted to the United Nations Framework Convention on Climate Change (UNFCCC). Recent studies have noted the value of hierarchical Bayesian inverse methods for improved uncertainty quantification in inverse modelling systems, and the importance of including physical constraints such as non-negativity in emissions. However, systems that exhibit these properties can suffer from severe computational bottlenecks, exacerbated by the growing volume of atmospheric observations and the demand for higher spatiotemporal resolution in emission estimates. As a result, spatial dimension reduction and spatiotemporal independence are often placed on emissions estimates, to make the algorithms computationally feasible. This research aims to explore these limits, before introducing novel approaches to address them.

We use a hierarchical Bayesian approach, using Markov chain Monte Carlo (MCMC) sampling, to explore the limits of the spatiotemporal resolution of flux estimates considering different multivariate Gaussian correlations. We demonstrate this by quantifying emissions of methane in the UK. In addition, we demonstrate the utility of a multivariate log-normal emissions distribution, simultaneously maintaining the non-negativity of emissions, as well as the explicit representation of emissions covariance. The results are compared with the emissions calculated using a spatially and temporally independent emissions prior, demonstrating the developments associated with a multivariate approach.

The computational costs associated with MCMC sampling means that the potential for extending the approach to large spatiotemporal parameter spaces is limited. Therefore, alternative inferential methods are explored, with a focus on sequential algorithms - such as Kalman filtering - that are augmented for non-Gaussian, hierarchal inference, in higher dimensional parameter spaces. This work provides the foundation for developing scalable non-Gaussian hierarchical frameworks that combine computational feasibility with improved emissions estimates.

How to cite: Pearson, S., Western, L., Ganesan, A., and Rigby, M.: Scalable Approaches for Hierarchical Non-Gaussian Inverse Modelling for Emissions Estimation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13641, https://doi.org/10.5194/egusphere-egu25-13641, 2025.

The Madden Julian Oscillation (MJO) is the dominant element of the atmospheric variability  in the tropics on intraseasonal zonal timescales. The MJO manifests itself as a slowly eastward propagating envelope coupling large scale circulation and convection. Despite recent progress in the understanding of the MJO, a comprehensive theory of the MJO is still missing, partly due to its complexity associated with moist physics and nonlinearity.
Here, we use a normal mode decomposition of atmospheric reanalysis datasets to show that the MJO can be understood in terms of a superposition of interacting Rossby, Kelvin  and inertio-gravity modes. We discuss the nature of the interaction between these modes, which can be either linear, due to large-scale variations in the moisture, or simply due to the inherent nonlinearity of the equations of the atmosphere.  

How to cite: Raphaldini, B., S. W. Teruya, A., Mayta, V., F.M. Raupp, C., Dias, P., and Peixoto, P.: Normal mode interactions in the Madden Julian Oscillation envelope, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13822, https://doi.org/10.5194/egusphere-egu25-13822, 2025.

Snapback, a post-peak response often observed in brittle rocks, is crucial for accurately characterising fracture properties, particularly regarding size and strain rate effects. This study emphasises the application of the AUSBIT method, which utilises an indirect control approach, to maintain strain rate inside the fracture process zone (FPZ) within the quasi-static range. By selecting different specimen sizes, the AUSBIT technique effectively captures the snapback response while minimising dynamic effects.

The experimental results reveal significant variations in tensile strength measured using the AUSBIT method compared to traditional Brazilian Disc testing. This discrepancy highlights the effectiveness of AUSBIT in accurately reflecting the underlying fracture mechanisms through indirect strain-rate control. These findings offer important insights into rock fracture behaviours and support the potential of AUSBIT as a valuable tool for studying size and rate effects in brittle materials.

How to cite: Dharani Raj, S. V., Nguyen, T. N., Huang, Z., Nguyen, G. D., Karakus, M., Bui, H. H., and Phan, D.: Influence of loading control on tensile strength of rocks , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14623, https://doi.org/10.5194/egusphere-egu25-14623, 2025.

Sampling error is a fundamental limitation of assimilation schemes, such as the EnKF, that employ the sample covariance from an ensemble of forecasts.  Despite the fact that the EnKF is typically applied in situations where the ensemble size is small compared to the system dimension, most of what is known about the effect of sampling error comes from low-dimensional examples or asymptotic results valid when the ensemble size is large. For high-dimensional systems and small ensembles, progress can be made by leveraging (i) the diagonal form of the Kalman-filter update in the optimal coordinates of Snyder and Hakim (2022) and (ii) basic approximations from the theory of random matrices. These yield novel, explicit expressions for the EnKF gain, the (sample) mean and covariance of the EnKF posterior ensemble, and the error covariance of that posterior mean. The expressions show that a single EnKF update will remove almost all variance from the ensemble, unless the observations are very uninformative.  They also identify those directions in the state space for which the EnKF update is effective, improving the state estimate despite sampling errors, and those directions for which sampling errors in the EnKF overwhelm observational information and degrade the state estimate.

How to cite: Snyder, C.: Sampling error in the ensemble Kalman filter for small ensembles and high-dimensional states, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14641, https://doi.org/10.5194/egusphere-egu25-14641, 2025.

Abstract: With the continuous advancement of tunnel construction in cold regions, the shotcrete technique has been widely utilized in tunnel support, forming a concrete-rock composite structure. Influenced by the low-temperature climate in the cold region, the freezing of moisture around the tunnel will exert a significant frost heaving force on this structure, resulting in various degrees of freeze-thaw damage to the structure. Hence, investigating the damage characteristics of the rock-mass concrete material under freeze-thaw cycling conditions is of great significance for the protection of tunnel lining. In this study, an environmental scanning electron microscope (ESEM) test was conducted on the rock-concrete composites specimens, and the micro-damage deterioration mechanism and failure mode of the specimens subjected to uniaxial compression freeze-thaw cycling were analyzed from a microscopic perspective. The test results indicate that: (1) The freeze-thaw cycling causes irreversible damage to the interface area between the rock and concrete, with the number and length of micro-cracks continuously increasing, leading to the fracture and separation of some areas at the interface between the two. It directly affects the macroscopic mechanical performance of the specimens. (2) The freeze-thaw cycling weakens the cementation between particles. It can cause the structure between mineral particles to become loose, resulting in mineral shedding and fracture. Some fibrous minerals are damaged, and the size and quantity of pore structures continuously increase, while the integrity and bonding degree of minerals are damaged. (3) The composite specimens mainly experience splitting failure, sliding failure, and the combined mixed failure under uniaxial compression conditions. The splitting failure belongs to the tensile-shear failure dominated by tensile failure, and the sliding failure is a simple shear failure. Additionally, it is found that the interface failure mode of composite specimens (the inclination angle is 90°) is the tensile-shear failure dominated by tensile failure.

Key words: ESEM; freeze-thaw cycles; rock-concrete specimen; uniaxial compression

How to cite: Zhou, Z., Li, Y., Hu, Y., and Li, K.: Microscopic Deterioration Mechanism and Failure Mode of Rock-Concrete Composites Subjected to Uniaxial Compression and Freeze-Thaw Cycles using ESEM Technology , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15099, https://doi.org/10.5194/egusphere-egu25-15099, 2025.

Extreme rainfall events (EREs) have recently been observed to exhibit teleconnection patterns across long spatial distances. Here we investigate the EREs over the Yellow River basin (YRB) using complex network-based event synchronization analysis. We found that EREs in the YRB are significantly synchronized with multiple regions worldwide, including both local regions and remote regions, and the spatial synchronization patterns exhibit time-scale dependence. Particularly, we found significant synchronization between the EREs in the YRB and those in Europe, with the YRB lags Eastern Europe by 3-5 days lag, while the YRB lags Western Europe by 5–7 days lag. Further analysis reveals that Rossby wave propagation plays a key role in the synchronization of EREs between Europe and the YRB. Wave trains originating in Europe propagate downstream of the Eurasian jet, inducing anomalous circulations over the YRB that enhance vertical upward motion and moisture transport, ultimately triggering EREs. Two distinct wave trains are observed across the Eurasian continent: one associated with Eastern Europe-YRB synchronization, occurring in the mid-latitude region and resembling Rossby wave patterns along the mid-latitude jet stream, with wave energy reaching the YRB after approximately 3 days; and another associated with Western Europe-YRB synchronization, positioned at higher latitudes and relates to Rossby waves along the polar front jet, with wave energy reaching the YRB in about 5 days. Our findings provide valuable insights into the predictability of EREs, offering critical guidance for improving forecasting and early warning capabilities for EREs in the YRB.

How to cite: Cai, L., Yuan, N., Boers, N., and Kurths, J.: Teleconnection of Extreme Rainfall Events between the Yellow River Basin and Europe: A Complex Network Analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15195, https://doi.org/10.5194/egusphere-egu25-15195, 2025.

The Dynamic State Index (DSI) is a scalar diagnostic field that indicates deviations from a steady, adiabatic and inviscid atmospheric basic state. It is defined as Jacobian determinant of the potential voticity, the Bernoulli stream function and the potential temperature. Several works have been demonstrated that the DSI provides a suitable parameter to indicate diabatic processes across all scales. So far, DSI variants for different atmospheric models have been derived and applied to diagnose e.g. the onset and presence of precipitation or the organization of storms. In a novel approach DSI variants that identify specific polytropic states (isobaric, isochoric, isentropic, isothermal) of the atmosphere can be used for an analysis of cloud processes at different climate zones.

How to cite: Rudolph, A., Lisa, S., Peter, N., Storelvmo, T., and Vercauteren, N.: The DSI - a dynamic weather and climate index, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15845, https://doi.org/10.5194/egusphere-egu25-15845, 2025.

Limited-area models (LAMs) often suffer from degradation in their representation of large-scale features compared to that of global models (GMs) due to the restricted domain size and limited observational coverage. To address this, we propose a novel flow-dependent large-scale blending (LSB) method for LAM data assimilation (DA). LSB methods incorporate large-scale information from a GM into the LAM DA system using scale-dependent weights. Our approach, termed as nested EnVar, extends the previously proposed static variational LSB method (nested 3DVar) to an ensemble-based framework. Unlike static LSB methods, nested EnVar simultaneously assimilates both observational and large-scale GM information into LAM forecasts with dynamically adjusting the weight given to GM information based on its estimated flow-dependent uncertainty. 

Through idealized assimilation experiments using a nested system of simplified chaotic models with a single spatial dimension, we demonstrate that nested EnVar effectively reduces large-scale errors in LAM DA as existing LSB methods, and offers better forecasts than GM downscaling. Compared to both traditional DA and other LSB methods, nested EnVar provides more accurate analyses and forecasts when dealing with dense and unevenly distributed observations. By dynamically accounting for GM uncertainty, nested EnVar improves the stability and accuracy of the analysis across scales. 

Our findings suggest that nested EnVar offers a promising alternative to traditional LSB methods for high-resolution simulations of complex, hierarchically structured phenomena. This novel approach has the potential to enhance the effectiveness of high-resolution LAM DA for spatially localized convective-scale observations.

How to cite: Nakashita, S. and Enomoto, T.: Flow-dependent large-scale blending for limited-area ensemble assimilation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16805, https://doi.org/10.5194/egusphere-egu25-16805, 2025.

Traditional abrupt climate change detection method often overlooks the process of the abrupt change. Considering the abrupt change process of climate change is conducive to further understanding the details of climate change. We propose the concept of the transition process of abrupt climate change and develop a climate transition process detection technology based on nonlinear models. Through dynamic segmented fitting, we identify the start time (state) and end time (state) of the abrupt change, as well as the stability parameters of the transition process, which are physical quantities that characterize the change process, and finely depict the characteristics of the transition process. Furthermore, based on the principle of critical slowing down, we derive generalized velocity and generalized force as early warning signals of abrupt climate change. During the change process, we also demonstrate the quantitative relationship between parameters during the system's change process, that is, the product of the degree of change stability and the square of the change amplitude is directly proportional to the change speed. This quantitative relationship has been confirmed in the observational data of global sea surface temperature. On this basis, prediction technology for climate turning points can be developed. This prediction technology has successfully predicted a transition process of the Pacific Decadal Oscillation.

How to cite: Yan, P., Zhao, C., and Li, H.: Transition process detection method of abrupt climate change, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17167, https://doi.org/10.5194/egusphere-egu25-17167, 2025.

Polar vortices are observed in the atmospheres of most solar-system planets, arising as a single cyclone centred on or close to the pole. In contrast, Jupiter’s polar vortices have an unprecedented structure. As revealed by NASA’s Juno spacecraft, they consist of geometric patterns of cyclonic vortices surrounding a central cyclonic vortex at the pole. These crystalline structures were not predicted prior to being observed, and the mechanisms explaining their formation and evolution remain poorly understood. One possible mechanism is that moist convection produces small vortices in the polar regions, with the cyclones then migrating polewards via the ‘beta-drift’ mechanism and merging. Nevertheless, models including these processes do not spontaneously produce polygonal patterns like those on Jupiter. In contrast, this study investigates the stability of an initialized pattern of fully formed polar vortices subjected to these small-scale short-lived processes. This forced-dissipative system is modelled using the shallow-water equations describing a single layer of fluid on a polar gamma-plane. The initialized cyclones are subjected to a stochastic forcing with a short decorrelation time and the factors affecting their stability and time-evolution are studied. These include their degree of shielding (an anticyclonic ring around each cyclone), their depth and the properties of the forcing, in addition to the role of potential vorticity mixing.

How to cite: Cope, L., Thomson, S., Seviour, W., and Shipton, J.: The Dynamics of Jovian Polar Cyclones, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18489, https://doi.org/10.5194/egusphere-egu25-18489, 2025.

Complex chaotic systems exhibit nontrivial internal variability, with a power spectrum typically characterised by resonant broad peaks standing out on a continuous background of frequencies. These resonances correspond to nonlinear excitable modes of the system's evolution that can be generally attributed to long-lasting persistent events, weakly damped instabilities, or critical settings where the chaotic attractor is approaching a crisis [1].

In the context of the climate system, examples of such resonant behaviour are the El Niño Southern Oscillation, caused by a weakly damped instability of the atmosphere-ocean system [2], or the transition between the Warm and Snowball state of the Earth system due to a boundary crisis due to a change of the ice-albedo feedback [3].

On top of describing the relevant features of the internal variability of systems, such resonances also represent the fundamental modes shaping the resilience of the system to external perturbations. In particular, the linear response of the system to general forcing scenarios is solely determined by the Green’s function, for which a decomposition in terms of resonances can be obtained [4].

It is possible to establish a link between the system's nonlinear resonances and the spectral properties of the Koopman operator underlying the evolution in time of the system's observables.

Based on our work in [5], I will show that data-driven techniques developed to investigate the properties of the Koopman operator can be used to extract both resonances and dynamical modes from data. I will provide numerical evidence that the dynamical evolution of the statistical properties of the system can be interpreted as a superposition of such modes. In particular, by employing a projection of generic observables of the system onto the set of nonlinear modes, I will show that it is possible to reconstruct not only correlation functions but also the response of virtually any observable of interest.

Even though so far restricted to low dimensional systems, our results highlight the importance of such nonlinear modes in shaping the variability and response of chaotic systems and provide a way to (a) interpret the relevance of observables as a proxy to investigating dynamical properties of the system and (b) explain the difference between intrinsic variability of observables and their response to perturbations.

References

[1] Chekroun et al., Journal of Statistical Physics (2020) 179:1366–1402, https://doi.org/10.1007/s10955-020-02535-x

[2] Tantet et al., Journal of Statistical Physics (2020) 179:1449–1474, https://doi.org/10.1007/s10955-020-02526-y

[3] Tantet et al., 2018 Nonlinearity 31 2221, https://iopscience.iop.org/article/10.1088/1361-6544/aaaf42

[4] Manuel Santos Gutiérrez and Valerio Lucarini, 2022 J. Phys. A: Math. Theor. 55 425002, https://iopscience.iop.org/article/10.1088/1751-8121/ac90fd

[5] Zagli, Colbrook, Lucarini, Mezić, Moroney, “Bridging the gap between Koopmanism and Response Theory: Using Natural Variability to predict Forced Response”, https://doi.org/10.48550/arXiv.2410.01622

How to cite: Zagli, N., Moroney, J., Lucarini, V., Colbrook, M., and Mezić, I.: Bridging the gap between Koopmanism and Response Theory: Using Natural Variability to predict Forced Response, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18492, https://doi.org/10.5194/egusphere-egu25-18492, 2025.

Intermittency has been initially linked to systems alternating regular and irregular states, while nowadays, it also encompasses systems that switch between two or more regimes. These include geophysical processes such as turbulence, convection, and precipitation patterns, not to mention applications in plasma physics, medicine, neuroscience, and economics. Traditionally, the study of intermittency has focused on global statistical indicators, such as the average frequency of regime changes under fixed conditions, or how these vary as a function of the system’s parameters or forcing, like in the case of climate change. However, these global indicators fail in capturing the local spatio-temporal nature of the regime transitions. 

In this work, we use  global and local perspectives to analyze intermittent systems and uncover reliable indicators of regimes’ changes. In particular, using tools such as the Lyapunov exponents and Covariant Lyapunov Vectors (CLVs), we have been able to characterize both global dynamics and local transitions in five different systems, of various complexities, and for three types of intermittency. We identified some key indicators and precursors of the regime transition that are common, despite the differences in the intermittency mechanism and in the dynamical model properties. At the same time, intermittency-type related mechanisms have been unveiled. For instance, we discover very peculiar behaviors in the Lorenz 96 (L96; Lorenz, 1996) and in the Kuramoto-Shivanshinki models (KS;  Kuramoto, Y. and Tsuzuki, T., 1976; G. Sivashinsky, 1977) that, despite their notoriety, have been so far unseen. In the L96 we identified crisis-induced intermittency with pseudo-periodic intermissions. In the KS equations we detected a spatially global intermittency which follows the scaling of type-I intermittency. 

In our local analysis, we uncover the relation between the CLVs mutual alignment and the regimes’ change, a connection that is present in all of the systems and for all types of intermittency considered. In the case of the on-off intermittency, the angle between CLVs is an effective precursor of the jump to the “on” regime. Furthermore, in all systems, the last CLV (the most stable), was found to carry important information about the dynamical features of the intermittency. In the case of type-I intermittency it allowed us to reconstruct the limit cycle around which the intermittency develops, in merging-crisis type of intermittency the last CLV successfully detected the pseudo-periodic intermissions, while in the case of on-off intermittency it allowed us to indicate the “off” state. 

The identification of these general and fundamental mechanisms driving intermittent behaviours, and in particular the detection of indicators spotting the regimes’ change, have the potential to be impactful in the study of turbulent geophysical fluids, rainfall patterns or atmospheric deep convection. In particular, they could be used to define a latent space of reduced dimension in which a neural network can be trained to automatically predict regime changes.

How to cite: Barone, A., Savary, T., Demaeyer, J., Vannitsem, S., and Carrassi, A.: Characterization of dynamical properties and indicators of regime change in intermittent chaotic systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18862, https://doi.org/10.5194/egusphere-egu25-18862, 2025.

It is well-established that large eddies significantly influence the turbulent transport of heat and scalars in the atmospheric surface layer. However, the mechanistic understanding of how large eddies originating from both the ground (updrafts) and aloft (downdrafts) regulate flux convergence (FC) and divergence (FD) remains relatively unexplored. Based on turbulence data measured at 12 levels, spanning from 1.2 m to 60.5 m above the ground, we observe a notable increase in the variability of sensible heat flux magnitudes with height. Our results show that FC and FD of sensible heat are primarily linked to variations in the respective transport efficiencies (RwT) at different heights. Using the cross-wavelet transform, we find that in FC cases, the regions with high wavelet coherence expand with height, resulting in higher RwT at higher levels compared to low ones. Conversely, in FD cases, the regions with high wavelet coherence decrease with height, leading to lower RwT at higher levels. Large eddies with length scales of approximately 120 to 500 m have a significant impact on amplifying or attenuating RwT at higher levels compared to lower levels. Using conditional sampling to extract the updrafts and downdrafts of large eddies, distinct patterns are observed in the characteristics of updrafts and downdrafts between FC and FD groups, especially in their flux contribution and transport efficiencies. This work emphasizes the significant contribution of asymmetric turbulent transport by updrafts and downdrafts to the discrepancy between the observed turbulent fluxes and those predicted by the Monin-Obukhov similarity theory.

How to cite: Gao, Z., Li, L., Liu, H., and Yang, B.: Flux Convergence and Divergence Linked to Asymmetric Transport by Large Turbulent Eddies , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20158, https://doi.org/10.5194/egusphere-egu25-20158, 2025.

In the spirit of Hasselmann’s program, Climate Tipping Points are often studied in systems described by stochastic differential equations where a combination of noise and a parameter approaching or crossing a bifurcation threshold leads to tipping. However, in some cases, a multistable system forced by another potentially chaotic system is a more appropriate description and might give rise to unexpected effects.

We show how tipping in such chaotically forced systems is affected by varying the relative timescale between the chaotic forcing and the forced multistable system. Further, we explain how periodic orbits of the forcing system can help to understand this effect and how they can be used to characterise the chaotic tipping window.

How to cite: Roemer, R. and Ashwin, P.: The chaotic tipping window and its dependence on relative timescales, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20233, https://doi.org/10.5194/egusphere-egu25-20233, 2025.

Detecting subtle, spatially localized changes in climate datasets is essential for understanding evolving regional dynamics and extreme events—key topics in Earth and planetary sciences. We present EagleEye, a novel distribution-free approach for identifying local density anomalies between two multivariate datasets. By leveraging a nearest-neighbor strategy and a binomial-based test, EagleEye pinpoints regions of significant deviations without relying on complex model assumptions.

We illustrate its potential using temperature reanalysis products, where EagleEye reveals localized shifts in historical climate patterns—including notable positive anomalies around Greenland—that may indicate emerging changes in temperature fields. Owing to its scalability and interpretability, EagleEye can handle large, high-dimensional datasets and integrate multiple climatic variables within a single region. This framework offers an innovative avenue for probing evolving climate dynamics, highlighting areas undergoing rapid change, and supporting an enhanced understanding of climate variability—making it a valuable tool for geoscience applications and beyond.

How to cite: Springer, S.: EagleEye: Unsupervised Detection and Quantification of Local Density Anomalies in Climate Data, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21320, https://doi.org/10.5194/egusphere-egu25-21320, 2025.

Recent studies found that the concept of typicality of extreme events is relevant in case of several weather and climate extremes related to atmospheric circulation anomalies, such as heatwaves or cold spells. This concept refers to the property of extreme events of being similar among each other while being at the same time anomalous with respect to usual weather conditions, referring to both spatial patterns and temporal evolutions. The aim of typicality analysis is to clarify whether this property applies to the extreme events of interest and to quantify the strength of typicality. In this presentation, I will discuss different types of typicality and give a practical guidance on how to perform a typicality analysis using climate datasets. I will present the most important steps of the analysis, discussing also several tools we can use to quantify typicality, such as standard statistical measures, Taylor diagrams, dynamical systems-based indicators.

How to cite: Galfi, V. M.: Typicality analysis for extreme climate events, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21616, https://doi.org/10.5194/egusphere-egu25-21616, 2025.

We study the hemispheric to continental scale regimes that lead to summertime heatwaves in the Northern Hemisphere. By using a powerful data mining methodology -archetype analysis - we identify characteristic spatial patterns consisting of a blocking high pressure systems embedded within a meandering upper atmosphere circulation that is longitudinally modulated by coherent Rossby Wave Packets. Periods when these atmospheric regimes are strongly expressed correspond to large increases in the likelihood of extreme surface temperature. Most strikingly, these regimes are shown to be typical of surface extremes and frequently reoccur. Three well publicised heatwaves are studied in detail - the June-July 2003 western European heatwave, the August 2010 "Russian" heatwave, and the June 2021 "Heatdome" event across western North America. We discuss the implications of our work for long-range prediction or early warning, climate model assessment and post-event diagnosis.

How to cite: Chapman, C., Monselesan, D., Risbey, J., Hannachi, A., Lucarini, V., and Matear, R.: The Typicality of Regimes Associated with Northern Hemisphere Heatwaves, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3877, https://doi.org/10.5194/egusphere-egu25-3877, 2025.

Turbulence remains a pressing challenge for aviation safety and efficiency, as highlighted by recent incidents involving Singapore Airlines, Qatar Airways, and Scandinavian Airlines. Among the various types, Clear Air Turbulence (CAT) poses the greatest hazard due to its occurrence in clear skies, rendering it difficult to detect and predict. Furthermore, the unprecedented changes in Earth's climate are reshaping atmospheric dynamics on a global scale, with profound implications on aviation. As a companion of ClimaMeter, a platform designed to assess and contextualize extreme weather phenomena in relation to climate change, we introduce here TurboMeter. It is designed to use ERA5 reanalysis data to investigate the meteorological drivers of turbulence events by comparing them with historical analogues under similar atmospheric conditions. Turbulence diagnostics, including Ellrod’s indices, are used to evaluate the roles of jet streams, wind shear, and convective activity at typical cruising altitudes.

To illustrate TurboMeter, we present some recent aviation turbulence events occurred during 2024. Our findings reveal that they are closely linked to intensified jet streams and enhanced convective activity, influenced by the growing impacts of anthropogenic climate change. These results highlight a concerning trend: changing climatic patterns are altering the atmospheric drivers of turbulence, particularly CAT, with significant implications for flight safety and operational planning. Our study evidences the urgent need for improved weather forecasting and turbulence prediction models to mitigate aviation risks in a rapidly warming climate.

How to cite: Alberti, T., Rapella, L., Coppola, E., and Faranda, D.: TurboMeter: attributing aviation turbulence events to climate change, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5631, https://doi.org/10.5194/egusphere-egu25-5631, 2025.

How to cite: Faranda, D. and the The ClimaMeter team: ClimaMeter: a rapid attribution framework for weather extreme events, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5780, https://doi.org/10.5194/egusphere-egu25-5780, 2025.

Since Lorenz's pioneering work, dynamical systems theory has provided a powerful framework for studying complex systems. Among these, the study of their instantaneous properties is particularly significant for understanding short-lived yet impactful extreme events. Here, we propose an analogues-based index to measure the instantaneous predictability of dynamical systems over different forecasting horizons. We demonstrate its application in both classical dynamical systems and the Euro-Atlantic sector atmospheric circulation. Furthermore, recognizing that the onset of extreme events often involves processes operating across different scales, we introduce a novel framework that enables the exploration of scale-dependent dynamical properties. Given the flexible and generalizable nature of these methods, we believe they open new research avenues for studying extreme events from a dynamical systems perspective and will serve as valuable tools for deepening our understanding of extreme events.

How to cite: Dong, C., Gualandi, A., Lucarini, V., and Mengaldo, G.: Advancing the understanding of extreme events through the lens of dynamical system theory, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7645, https://doi.org/10.5194/egusphere-egu25-7645, 2025.

Extreme weather events, including heatwaves, extreme precipitation, tropical cyclones, and other hazards, pose significant risks to society and ecosystems. Recent advancements in observational techniques, numerical modeling, theoretical frameworks, and AI methods have greatly improved our understanding and prediction of extreme weather events. However, despite significant progress, key challenges remain unresolved, particularly in achieving a thorough understanding of the physical drivers of extreme events, improving the transparency of AI-based prediction methods, and evaluating the vulnerability and resilience of cities to their impacts. To address these challenges, we present various approaches drawn from different fields, including dynamical systems theory, explainable AI, and NLP-based methods. Given the flexible and generalizable nature of these methods, we believe they may pave the way toward more robust solutions for addressing the challenges posed by extreme weather events.

How to cite: Mengaldo, G.: Progress and Challenges in the Study of Extreme Weather, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7685, https://doi.org/10.5194/egusphere-egu25-7685, 2025.

Compound climate and weather extremes have received significant attention in recent years due to the increased risks that they pose to the environment, human societies, and the economy. While prior studies have identified associations between various hazards in disaster databases, investigations focussing on droughts and floods remain rare. In this study, we analyze the impacts of concurrent or sequential drought-flood extremes from two widely used disaster databases: the Emergency Events Database (EM-DAT) and its geocoded version (GDIS), as well as the DesInventar database. The analysis focuses on the period from 1960 to 2018, aligning with GDIS temporal coverage. We define concurrent or sequential hazards as instances where a flood occurs during a drought period or within four months following a drought.  

Our findings for the global extratropics reveal that the economic losses and the number of affected people resulting from the identified drought-flood events are two to eight times higher than those ascribed to isolated droughts or floods, with a confidence interval ranging from two to twelve. Specifically, in DesInventar, the impact ratio (the mean impact of concurrent or sequential events divided by the mean impact of isolated events) for indirectly affected individuals and financial losses is approximately three. In EM-DAT, the impact ratio reaches three for economic damages and eight for affected individuals. Furthermore, the impact ratios are notably higher in the last 30 years of the study period compared to earlier decades, emphasizing the increasing severity of the drought-flood compound events.

These results highlight the amplified negative impacts when droughts and floods occur concomitantly or sequentially, highlighting the need for more robust policies to address their socio-economic risks, particularly under changing climatic conditions.

How to cite: Worou, K. and Messori, G.: Amplified Socio-Economic Impacts of Concurrent or Sequential Drought-Flood Events: Insights from Disaster Databases (1960–2018), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8719, https://doi.org/10.5194/egusphere-egu25-8719, 2025.

Climate change is an ongoing process that is modifying weather patterns and influencing weather phenomena and extreme events such as heatwaves, droughts, and floods. In this study, we investigate whether climate change can also play a role in enhancing wildfires by focusing on a set of three recent wildfires in Europe (i.e., events occurred in Central Sweden in July 2018, France in July 2022, and in Sicily and Greece in July 2023). We employ the concept of analogues to assess the influence of climate change on the atmospheric conditions underlying wildfire development monitored through the fire weather index, by comparing past and present atmospheric patterns similar to those that occurred during the wildfire. Our analysis focuses on both reanalysis data and high-resolution regional climate models to attribute the observed changes and provide future projections. Our findings show that climate change has altered critical factors supporting wildfire development, such as temperature, humidity, and wind patterns. The results from our sample of three events point out that climate change has increased wildfire hazards in Europe, which is projected to further increase for similar fire weather conditions in the future.

How to cite: Lu, C., Nogherotto, R., Alberti, T., Messori, G., Coppola, E., and Faranda, D.: Assessing the impact of climate change on wildfire development: insights from analogues and regional climate models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10235, https://doi.org/10.5194/egusphere-egu25-10235, 2025.

Extreme weather and climate events, marked by unexpected and severe conditions at the edges of historical distributions, significantly impact human health, society, and ecosystems. With global warming driving an increase in the frequency and intensity of these extremes, there is an urgent need to enhance weather prediction beyond the typical 7–10-day range. Among the atmospheric and oceanic components studied for improving predictability, the stratosphere stands out due to its slower and more predictable changes, which can have persistent impacts on surface weather patterns.

Research has highlighted the stratosphere's role in driving weather and climate extremes, particularly in the extratropical Northern Hemisphere. Events involving a weak or strong stratospheric polar vortex can precede the occurrence of surface extremes, making the polar vortex a key link between stratospheric variability and surface climate predictability. While various studies have previously identified this teleconnection, the processes connecting anomalous vortex states to extreme surface events are not yet fully understood.

In VORTEX project we employ a methodology based on advancements in dynamical systems theory to explore the relationship between anomalous polar vortex states and extreme precipitation and temperature events. This approach characterizes each vortex-extreme event's recurrence, persistence, and predictability, providing dynamic insights that traditional methods cannot. By identifying the intrinsic predictability of stratospheric patterns tied to extremes, this methodology offers a pathway to improve sub-seasonal to seasonal climate models, focusing future efforts on better representing critical patterns that influence extreme weather.

How to cite: Alvarez-Castro, C., Peña-Ortiz, C., Gallego, D., and Faranda, D.: VORTEX project: The role of the polar vortex on the predictabIlity of extreme events in the Northern Hemisphere, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10570, https://doi.org/10.5194/egusphere-egu25-10570, 2025.

Anthropogenic climate change (ACC) is one of the most demanding challenges facing our society. The intensification and increased frequency of many extreme events due to ACC are among its most impactful consequences, threatening human health, infrastructure, and ecosystems. In this context, raising the awareness of the general public of the relationship between ACC, extremes, and associated impacts becomes a crucial task.

This work is grounded in attribution science and focuses on quantifying and understanding the influence of internal climate variability on extreme events. Among the many tools available for attribution, we use ClimaMeter [Faranda et al. 2023], a rapid framework designed to provide context for extreme events in relation to ACC. ClimaMeter’s approach emphasizes the dynamics associated with extreme events and identifies weather conditions similar to those characterizing the event of interest, leveraging the analogues methodology for conditional attribution [Yiou, 2014]. The analysis provided by such a framework enables the evaluation of significant changes over time of the event’s dynamics and associated meteorological hazards and links them to ACC.

An essential part of ClimaMeter’s methodology is quantifying the influence of natural variability relative to ACC in explaining the changes associated with the event. Specifically, three modes of Sea Surface Temperature variability are taken into account: the El Niño-Southern Oscillation, the Atlantic Multidecadal Oscillation and the Pacific Decadal Oscillation. These three modes are considered with equal weight and changes not explained by them are assumed to be due to ACC [Faranda et al., 2023]. While the methodology is rapid and easy to communicate, it also has some limitations. In this work, we investigate the implications of this approach. First, we test it on a pre-industrial simulation of the IPSL climate model to evaluate its performance under stationary climate conditions. Additionally, we explore a generalization of the current methodology, aiming to refine the quantification of natural variability by weighing the three modes based on the event region and associated hazard. This generalized approach has the potential to expand ClimaMeter’s methodology and provide new insights into the complex mechanisms linking natural variability and extremes.

How to cite: Naldesi, C., Vrac, M., Bertrand, N., and Faranda, D.: Exploring a new methodology to quantify natural variability in conditional extreme event attribution, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10822, https://doi.org/10.5194/egusphere-egu25-10822, 2025.

Hemispheric symmetries, including those in zonal-mean eddy kinetic energy and in hemispheric-mean planetary albedo, are a characteristic feature of Earth’s climate. Whether such a symmetry also holds for extreme surface windspeeds driven by midlatitude cyclones is currently unclear. We address this question by focusing on the regions with the peak of storm tracks over the North Atlantic, North Pacific and Southern Ocean. We analyse reanalysis and satellite datasets and employ objectively calculated storm tracks to associate cyclones with surface winds they produce. Additionally, we check for existence of trends in extreme windspeeds of each basin. Results show a statistically distinguishable hemispheric asymmetry in extreme surface windspeeds, with the North Hemisphere having stronger extremes, driven primarily by extreme windspeeds occurring during winter and in proximity to cyclones. This implies that cyclones in the North Hemisphere drive stronger surface windspeed extremes than in the South Hemisphere. The North Hemisphere also has higher extreme windspeeds above the boundary layer (700 hPa), pointing to the role of large-scale processes in driving these differences. Lastly, trends in the extreme surface windspeeds across all basins are positive in the reanalysis dataset, and statistically significant in the North Pacific and Southern Ocean.

How to cite: Stanković, A. and Caballero, R.: Winter cyclones drive stronger surface wind extremes in the North Atlantic than in the Southern Ocean, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10966, https://doi.org/10.5194/egusphere-egu25-10966, 2025.

North Pacific blocking patterns, defined by persistent high-pressure systems that disrupt atmospheric circulation, are pivotal elements of mid-latitude weather dynamics. These blocking events play a significant role in shaping regional weather extremes, such as prolonged cold spells or heatwaves, and can redirect storm tracks across the Pacific. For instance, the 2021 Pacific Northwest heatwave demonstrated the profound impact of blocking on terrestrial temperatures, where an upstream cyclone acted as a diabatic source of wave activity, intensifying the blocking system. This led to heat-trapping stable stratification, which elevated surface temperatures to unprecedented levels (Neal et al., 2022). Similarly, marine heatwaves in the Northeast Pacific have been linked to high-latitude blocking events, which weaken westerly winds, suppress southward Ekman transport, and enhance ocean stratification, thereby increasing sea surface temperatures (Niu et al., 2023). The predictability of North Pacific blocking events is governed by the intricate interplay of large-scale atmospheric dynamics, ocean-atmosphere interactions, and internal variability (Smith et al., 2020).

This study investigates the differences in predictability between eastern and western North Pacific blocking events, using a modified version of the Davini et al. (2012) blocking index to distinguish their geographical locations. Identified blocking events were tracked using a block-tracking algorithm until they dissipated. Predictability was assessed by identifying an analogue pair for each blocking event. Specifically, after classifying blocks as eastern or western, geopotential height maps for each event were compared to all other days in the dataset. The analogue pair for an event was defined as the day with the smallest root mean square (RMS) distance. Predictability was then evaluated by averaging the error evolution of the tracks between events in each analogue pair.

Using CMIP6 model simulations and ERA5 reanalysis data, the study revealed that eastern blocks are significantly more persistent and stable than their western counterparts. Eastern blocks exhibited longer durations and greater resistance to atmospheric variability, resulting in improved forecast accuracy. In contrast, western blocks were found to be more transient and challenging to predict due to their susceptibility to dynamic instabilities.

References

Davini, P., Cagnazzo, C., Gualdi, S. and Navarra, A., 2012. Bidimensional diagnostics, variability, and trends of Northern Hemisphere blocking. Journal of Climate, 25(19), pp.6496-6509.

Neal, E., Huang, C.S. and Nakamura, N., 2022. The 2021 Pacific Northwest heat wave and associated blocking: Meteorology and the role of an upstream cyclone as a diabatic source of wave activity. Geophysical Research Letters, 49(8), p.e2021GL097699.

Niu, X., Chen, Y. and Le, C., 2023. Northeast Pacific marine heatwaves associated with high-latitude atmospheric blocking. Environmental Research Letters, 19(1), p.014025.

How to cite: K Xavier, A., Hamilton, O., Faranda, D., and Vannitsem, S.: Comparative predictability of eastern and western north pacific blocking events, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12284, https://doi.org/10.5194/egusphere-egu25-12284, 2025.

The physical mechanisms underlying climate-induced extreme events are inherently complex, arising from the compounding nature of multiple drivers and/or hazards. Leveraging the chaotic nature of the atmosphere, a novel approach, based on results from dynamical system theory, has recently been adopted to reveal the drivers of both individual and compound extremes. Central to this approach is the co-recurrence ratio, which quantifies the instantaneous dynamical coupling between multiple variables in terms of joint recurrences of atmospheric configurations to similar ones in the past.

While the co-recurrence ratio has demonstrated potential in revealing the atmospheric drivers of certain extremes, its performance may depend heavily on factors such as the choice of geographical domain(s), selection of variables, and the thresholds used to define extremes. These sensitivities remain underexplored, limiting the broader applicability of this approach.

In this study, we aim to address these gaps by assessing the sensitivity of the co-recurrence ratio in a European setting, focusing on daily winter extremes in temperature, wind, and precipitation. For this analysis, we adopt a bivariate focus, diagnosing the coupling between large-scale circulation patterns and single hazard variables.

By exploring these sensitivities, this work seeks to enhance the understanding of the robustness of the co-recurrence ratio and its effectiveness in diagnosing the atmospheric drivers of various types of extremes.

How to cite: Reiter, A. C., Drews, M., Messori, G., Faranda, D., and Dahl Larsen, M. A.: Sensitivity of Dynamical Coupling to Large-Scale Circulation in European Winter Extremes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13213, https://doi.org/10.5194/egusphere-egu25-13213, 2025.

The co-occurrence of wintertime cold spells in North America and wet, windy extremes in Europe, known as the Pan-Atlantic compound extremes, is linked to distinct dynamical pathways. One of those dynamical pathways involves the presence of a persistent high-pressure system west of Greenland. This high-pressure anomaly tends to simultaneously induce a southward displacement of a trough over the eastern United States and sustain an upper-level trough over southwestern Europe, creating conditions that induce both cold spells in North America and extreme precipitation in Europe. The co-occurrence of the Pan-Atlantic compound extremes has been investigated in previous studies. However, the causal association between extremes on both sides of the Atlantic has yet to be verified. In this study, we aim to assess the relationship between these compound extremes and to uncover the causal mechanisms driving their co-occurrence. Preliminary findings indicate that high-pressure anomalies over Greenland are a main driver of both phenomena, providing a coherent dynamical link that bridges these geographically distinct extreme events. The study further seeks to clarify the underlying dynamics and improve predictability for such interconnected extreme weather events, which can help to better manage and mitigate their impacts.

How to cite: Krouma, M. and Messori, G.: Causality and predictability of the Pan Atlantic compound extremes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13374, https://doi.org/10.5194/egusphere-egu25-13374, 2025.

Extreme weather events such as heatwaves, droughts, thunderstorms, and cyclones threaten human lives, ecosystems, and economic stability. Tracking and characterizing the spatiotemporal dynamics of such events is essential for understanding their cascading impacts on socioeconomic and environmental systems. When the detection and characterization of extremes are done in real-time, they can provide critical information that benefits many sectors, including agriculture, emergency management, and regulatory authorities.

To offer a tool for operational monitoring of weather-related hazards across Europe, we developed RHITA (Real-time Hazards Identification and Tracking Algorithm), an online framework designed for the rapid, automated, and objective spatiotemporal detection of hazards driven by extreme weather events. RHITA is intended for a wide range of users, including scientists, policymakers, authorities, and the general public. It leverages the ERA5 dataset for real-time detection, and the algorithm is calibrated using the EM-DAT dataset, which documents global disaster occurrences and impacts.

RHITA currently offers two main features: (1) real-time tracking and spatiotemporal characterization of extreme weather events such as heatwaves, droughts, cold spells, cyclones, and storms, focusing on associated hazards like extreme temperatures, water deficits, heavy precipitation, and strong winds; and (2) publicly available, up-to-date, transboundary historical spatiotemporal hazard catalogs for Europe.

How to cite: Cazzaniga, G., Burq, A., Vrac, M., and Faranda, D.: RHITA: a framework for real-time detection and characterization of weather extremes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13784, https://doi.org/10.5194/egusphere-egu25-13784, 2025.

Even though tropical cyclones (TCs) are well documented from the moment they reach a certain intensity to the moment they start to evanesce, many physical and statistical properties governing them are not well captured by gridded reanalysis or simulated by earth system models. Thus, the tracking of TCs remain a matter of interest for the investigation of observed and simulated tropical. Many cyclone tracking schemes are available. On the one hand, there are trackers that rely on physical and dynamical properties of the TCs and users prescribed thresholds, which make them rigid, and need numerous variables that are not always available in the models. On the other hand, there are trackers leaning on deep learning which, by nature, need large amounts of data and computing power. Besides, given the number of physical variables needed for the tracking, they can be prone to overfitting, which hinders their transferability to climate models. In this study, the ability of a Random Forest (RF) approach to track TCs with a limited number of aggregated variables is explored. Hence, it becomes a binary supervised classification problem of TC-free (zero) and TC (one) situations. Our analysis focuses on the Eastern North Pacific and North Atlantic basins, for which respectively 514 and 431 observed tropical cyclones tracks record are available from the IBTrACS database over the 1980-2021 period. For each 6-hourly time step, RF associates TC occurrence or absence (1 or 0) to atmospheric situations described by predictors extracted from the ERA5 reanalysis. Then situations with TC occurrences are joined for reconstructing TC trajectories. Results show good ability of the method for tracking of tropical cyclones over both basins and good ability for spatial and temporal generalization as well. It also shows similar TC detection rate as trackers based on TCs' properties and significantly lower false alarm rate. RF allows us to detect TC situations for a range of predictor combinations, which brings more flexibility than threshold based trackers. Last but not least, this study shed light on the most relevant variables allowing to detect tropical cyclone.

How to cite: Vaittinada Ayar, P., Bourdin, S., Faranda, D., and Vrac, M.: Ensemble Random Forest for Tropical Cyclone Tracking, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13891, https://doi.org/10.5194/egusphere-egu25-13891, 2025.

Europe has been identified as a heatwave hotspot, where heatwave intensities have outpaced other mid-latitude regions in the Northern Hemisphere (Rousi et al. Nat. Comms. 2022). Accelerated European heatwave trends have been found to be associated with the increased persistence of Eurasian double jets, a specific set-up of the large-scale circulation in which the Northern hemisphere polar and subtropical jets occur as two clearly separated branches. However, if observed trends are projected to continue with anthropogenic warming and to what degree the present generation of climate models constitute useful tools to assess changes in the atmospheric circulation has not yet been ascertained.

In this study, we benchmark 11 CMIP6 climate models to evaluate their ability to reproduce the main characteristics of double jets and their relationship to heat extremes, aiming to identify the best-performing models for future projections. Our findings show that, on average, the models tend to underestimate the frequency of double jets by 80%. Moreover, half of the climate models underestimate the intensity of double-jet-associated heatwaves over Western Europe, with the remaining models even showing a negative anomaly in heatwave intensity during double jet events in the region. Furthermore, climate models fail to capture the growth rate of double jet persistence, with the model mean trend at -0.4 days per decade, while the observed rate is approximately 1.5 days per decade. The bias in the persistence trend of double jet in models is strongly correlated with the underestimation of the western European heat extreme trend, with an R² value of 0.42.

Despite this, some models show reasonable agreement with the observations, and these models are further analyzed to project circulation-driven changes in extreme heat. Using EC-Earth3-Veg-LR, we observe an increase in double jet frequency from 2020 to 2060, at a rate of 0.2 days per decade. Our work highlights the need for better representation of double jet characteristics and their relationship with heat extremes in climate models to enhance preparedness for future heat risks.

How to cite: Liu, S., Tian, Y., and Kornhuber, K.: Large-scale atmospheric circulation as a source of uncertainty in western European heat extreme projections , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14873, https://doi.org/10.5194/egusphere-egu25-14873, 2025.

Extreme precipitation events (EPEs) are meteorological phenomena that are likely to intensify as a result of climate change. They are a major concern for our society, especially in densely populated areas, as they can have significant economic and environmental impacts. Therefore, identifying large-scale atmospheric circulation that lead to EPEs is crucial for detecting geographical areas at risk and mitigating their adverse impacts.

To achieve this objective, we study the circulation patterns associated with EPEs in Italy. Initially, we focus on North-Central Italy and we identify the precipitation extremes in three datasets: ARCIS 3.0, MSWEP, and CERRA LAND. Circulation types associated with the EPEs are obtained by applying Self Organizing Maps (SOMs), an unsupervised artificial neural network widely used in synoptic climatology, to anomalies of geopotential height at 500 hPa and mean sea level pressure. Since ArCIS, the reference dataset, is limited to North-Central Italy, we extend the analysis to the whole of Italy using CERRA-Land. Such choice is based on the fact that it produced the most similar results to ArCIS in North-Central Italy compared to MSWEP.

We then generate composites of various variables (all retrieved from ERA5) for each SOM pattern to better understand the circulation patterns and characterize the atmospheric dynamics associated with extreme events. Additionally, we analyze the probability of exceeding the 99th percentile of wet-days to identify the areas impacted by each pattern. Composites for the different circulation types show variations in the synoptic pattern's position within the Mediterranean basin, as well as differences in the direction and intensity of moisture flux. These patterns influence distinct regions and display varying frequencies across seasons.

In future works the classification obtained by this study will be applied to climate model simulations, aiming to investigate the role of anthropogenic climate change in the dynamics leading to EPEs in Italy. 

How to cite: Iacomino, C., Pascale, S., Zappa, G., Iotti, M., Grazzini, F., Portal, A., and Ghinassi, P.: High-risk atmospheric circulation patterns for Italian precipitation extremes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15668, https://doi.org/10.5194/egusphere-egu25-15668, 2025.

Regional climate emulators provide computationally efficient tools for generating high-resolution climate projections, bridging the gap between coarse-scale models and the detailed resolution required for local-scale hazard assessments. Climate hazards from extreme precipitation events are projected to increase in frequency and intensity under global warming, emphasizing the need for accurate modeling of convective processes. However, traditional numerical methods are constrained by low resolution or the high computational costs of kilometer-scale simulations.

To overcome these limitations, we introduce GNN4CD, a novel deep learning emulator that estimates kilometer-scale (3 km) hourly precipitation from coarse atmospheric data (~25 km). The model leverages graph neural networks and a hybrid imperfect approach (HIA) for downscaling, initially trained on ERA5 reanalysis and observational data, and applied using regional climate model (RegCM) data for present-day and future projections.

GNN4CD demonstrates exceptional performance in reproducing precipitation distributions, seasonal diurnal cycles, and extreme percentiles across Italy, even when trained on northern Italy alone. The model captures shifts in precipitation distributions, particularly for extremes, across historical, mid-century, and end-of-century scenarios. Additionally, evaluations using an ensemble of convection-permitting regional models confirm GNN4CD's ability to replicate ensemble spreads for both present-day and future projections essential for estimating the uncertainty in the future climate change signal..

How to cite: Coppola, E., Blasone, V., Di Gioia, S., Sanguinetti, G., Arora, V., and Bortolussi, L.: Graph neural networks based climate emulator for kilometer scale hourly precipitation : a novel hybrid imperfect approach, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17645, https://doi.org/10.5194/egusphere-egu25-17645, 2025.

The North Atlantic Oscillation (NAO) is the dominant pattern of atmospheric variability over the North Atlantic region, having its greatest influence on Europe during the winter months. In winter, positive NAO index values are linked to warmer temperatures and increased precipitation in western and northern Europe, whereas southern Europe tends to experience colder and drier conditions. These drier conditions can pose significant challenges for agriculture and livelihoods. An overall positive trend in the NAO index has been observed in winter in recent decades. However, how precipitation dynamics in the Mediterranean region respond to the shift towards a higher NAO index are largely unknown, partly due to the poor capture of NAO’s upward shift in climate models. 

Here we examine the impact of the shift towards a higher NAO index on precipitation dynamics in the Mediterranean region in winter. We employ a novel statistical model to analyse next-day precipitation conditional on past observations. The analysis focuses on conditioning drought persistence on different NAO states to assess their influence on the distributional characteristics of drought durations across the Mediterranean region. We present preliminary analyses that contribute to the growing body of evidence that long-term positive trends in the NAO index have an impact on rainfall patterns and drought occurrence in Europe. Understanding the role of teleconnections in regional climate variability and long-term trends is essential for robust regional climate projections for improved risk assessment and policy planning.

How to cite: Schultz, E., Spanjers, B., and Coumou, D.: The impact of the upward trend in the NAO index on precipitation dynamics in the Mediterranean region, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17852, https://doi.org/10.5194/egusphere-egu25-17852, 2025.

People are increasingly turning to courts to combat climate crisis. In the early 2000s, fewer than 10 climate change litigation cases had been filed globally. By 2024, this number has grown to over 2,500, with more than half originating in the United States. Some of these cases rely on extreme weather attribution science to link damages to anthropogenic climate change. Developing rigorous, legally useful assessments of damage attributable to climate change is an increasingly pressing need.

We present a framework for forecast-based impact attribution which can link physically consistent hazards to impacts, providing evidence for legal cases and climate cost recovery laws. As a case study, we analyze the severe impacts of Storm Irene in August 2011 when it was undergoing extratropical transition in the north-eastern USA. In the state of Vermont, Irene caused rainfall of up to 180 mm within a few hours, leading to fluvial and pluvial flooding with catastrophic consequences that caused $850 million in economic damages. By integrating an operational weather forecast model (ECMWF’s IFS) and hydrological models with economic impact assessments, we assess the extent to which these damages can be attributed to anthropogenic climate change.

This research underscores the potential of interdisciplinary attribution methodologies to enhance the scientific basis for judicial adjudication on climate change and climate law-making.

How to cite: Ginesta, M., Ermis, S., Stuart-Smith, R., and Franta, B.: Impact Attribution for Climate Law: The Case of Storm Irene, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17937, https://doi.org/10.5194/egusphere-egu25-17937, 2025.

The framework of weather analogues is a powerful methodology for the detection of the climate change fingerprint on weather extremes, that has been widely used in several contexts. The procedure has several advantages compared to standard model-based attribution exercises, being fast and not computationally expensive. Here we address whether the detection of analogs based on impacts (e.g., environmental, socio-economic) of a severe weather event can provide added value on the attribution of the event intensity or likelihood to climate change.

As a case study, we analyse the twin Emilia-Romagna flood event of May 2023. It caused a sizable amount of casualties, widespread destruction and substantial economic damage. We detect analogues of the river runoff as an impact-based observable of interest, addressing it in an univariate context, but also jointly with other observables (i.e., in a multivariate framework), such as mean sea-level pressure, total precipitation, and 850 hPa vorticity. We therefore detect the optimal set of variables for performing multivariate analysis and the appropriate analysis domain. We suggest that by combining river runoff with other observables by carefully selecting the spatial domain, we obtain a clearer view of the role played by anthropogenic climate change for this event, also including the additional vulnerability linked to the environmental impact of human activities, such as land-use change and freshwater diversion.

How to cite: Lembo, V., Ginesta, M., Alberti, T., D'Agostino, R., and Faranda, D.: Towards an impact-based approach to the detection of analogues: the case study of Emilia-Romagna floods in May 2023, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17947, https://doi.org/10.5194/egusphere-egu25-17947, 2025.

Many natural systems show emergent phenomena at different scales, leading to scaling regimes with signatures of chaos at large scales and an apparently random behavior at small scales. These features are usually investigated quantitatively by studying the properties of the underlying attractor. This multi-scale nature of natural systems makes it practically impossible to get a clear picture of the attracting set as it spans over a wide range of spatial scales and may even change in time due to non-stationary forcing.

Here we present a review of some recent advancements in characterizing the number of degrees of freedom and the predictability horizon of geophysical and complex systems showing non-hyperbolic chaos, randomness, state-dependent persistence and predictability. We compare classical approaches, based on Lyapunov exponents and correlation dimension, with novel approaches based on combining adaptive decomposition methods with concepts from extreme value theory. We demonstrate that the properties of the invariant set depend on the scale we are focusing on and that the proposed formalism can be generally helpful to investigate the role of multi-scale fluctuations within complex systems, allowing us to deal with the problem of characterizing the role of stochastic fluctuations across a wide range of physical systems as well as the role of different dynamical components in determining the predictability of rare events in complex systems.

How to cite: Alberti, T., Faranda, D., and Lucarini, V.: The predictable chaos of rare events in geophysical and complex systems, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18710, https://doi.org/10.5194/egusphere-egu25-18710, 2025.

The Diurnal Temperature Range (DTR) serves as a crucial meteorological indicator, reflecting the difference between daily maximum and minimum temperatures and the magnitude of diurnal extremes. The anomalous values of DTR are often linked to the occurrence of various climatic extremes such as droughts, heatwaves, and wet spells, which make it necessary to understand the evolution of DTR both historically and for the future. This study focuses on analyzing the evolution of DTR globally by employing the non-stationary Multidimensional Ensemble Empirical Mode Decomposition (MEEMD) method. To accomplish this, historical temperature data spanning 69 years (1951-2019) and CMIP6 Bias corrected data covering 150 years (1951-2100) were utilized. The non-linear trend characteristics in temperature are computed using CRU 0.50 x 0.50 gridded temperature data for historical trends and five different bias-corrected climate projection datasets of NASA Earth Exchange Global Daily Downscaled Projections (NEX-GDDP-CMIP6) for the assessment of trends in future DTR by considering two SSP scenarios, i.e., SSP 245 and SSP 585, each corresponding to intermediate and high emissions scenarios. The CMIP6 models that are considered are CanESM5, GFDL CM4, MIROC6, NorESM2-MM, and MPI-ESM1-2-HR. The results from the analysis reveal the decrease in global DTR, with a faster rate of increase in minimum temperature than in maximum temperature. However, the southern regions of Australia and Africa showed an increase in DTR. The CMIP6 simulations showed that CanESM5 and MPI-ESM1-2-HR showed a decreasing trend in global DTR for both scenarios of ssp, with an increase in DTR for South America and the southern part of Africa for CanESM5, while GFDL CM4, MIROC6, and NorESM2-MM showed a decrease in global DTR. The findings underscore the importance of understanding regional climatic variations when assessing global temperature trends. The observed contrasting regional patterns in DTR highlight the influence of localized hydroclimatic factors, including land-use changes, aerosols, and atmospheric feedback mechanisms. These insights are crucial for refining climate models and improving future climate projections under different emission pathways. Overall, the study emphasizes the necessity of incorporating non-linear approaches like MEEMD to capture complex climatic trends and underscores the role of DTR as a key indicator of climate change and its impacts at both global and regional scales.

How to cite: Reddy, M. S. B., Rajendran, V., and Tewari, M.: Analyzing the Historical and Projected Evolution of the Global Diurnal Temperature Range (DTR), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18740, https://doi.org/10.5194/egusphere-egu25-18740, 2025.

Heatwaves (HWs) are extreme climate events increasingly magnified under climate change, posing significant risks to both human and environmental systems. The Mediterranean region, recognized as a climate change hotspot, is experiencing a worrying amplification of both atmospheric and marine heatwaves. In this presentation we will discuss the evolution and interplay of these phenomena emphasizing their compounding effects when occurring simultaneously.

Our findings reveal a clear increase in HW frequency, intensity, and duration, with the concurrence of atmospheric and marine heatwaves resulting in a significant local amplification of marine heatwave intensity. While atmospheric heatwaves remain largely unaffected by this interaction. This interaction has become more prominent in recent years, highlighting the increasing complexity of extreme climate phenomena in this region.

The results underscore the urgent need for regionally tailored strategies to mitigate the cascading impacts of compounding heatwaves, as their intensification under climate change exacerbates threats to Mediterranean ecosystems and communities.

How to cite: Khodayar Pardo, S., Pastor, P., and Paredes-Fortuny, L.: Unraveling the Rising Threat of Atmospheric and Marine Heatwaves in the Mediterranean Region, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19577, https://doi.org/10.5194/egusphere-egu25-19577, 2025.

Starting January 7, 2025, devastating wildfires have swept through the Los Angeles metropolitan area and nearby regions. By January 10, the fires had caused ten deaths, destroyed thousands of structures, displaced nearly 180,000 residents, and scorched approximately 30,000 acres. This study employs the extended ClimaMeter (climameter.org <http://climameter.org/>) protocol to explore the potential role of climate change in exacerbating the severity of this event. Specifically, we examine whether climate change has modified the atmospheric conditions, represented by the mean sea level pressure, that contribute to wildfire occurrence, represented by the fire weather index, by analyzing historical and current weather patterns similar to those observed during the fires. Our methodology integrates both reanalysis datasets and high-resolution regional climate models to assess observed changes and project future fire risk scenarios. The results indicate a significant increase in the fire weather index across much of California and surrounding regions, which suggests that this event can be ascribed to human-driven climate change. The models show a similar signal in the present climate and project increases in fire weather hazard in the future.

How to cite: Nogherotto, R., Lu, C., Cazzaniga, G., Erika, C., and Faranda, D.:  January 2025 Wildfires in Southern California are attributable to Anthropogenic Global Warming, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19698, https://doi.org/10.5194/egusphere-egu25-19698, 2025.

The El Niño-Southern Oscillation (ENSO) represents one of the most significant drivers of global climate variability. This study investigates the chaotic parameter regimes of the Jin-Timmermann model, particularly focusing on the dynamics identified by Guckenheimer et al. where chaotic attractors emerge. We analyze the reduced three-dimensional system with specific attention to the critical parameters δ = 0.225423, ρ = 0.3224, which govern the time-scale interactions between oceanic and atmospheric processes. Using topological data analysis (TDA), we characterize the structural transitions between periodic and chaotic behaviors in the model's parameter space. Our methodology combines persistent homology with dynamical systems theory to identify distinct topological signatures associated with strong El Niño events. We validate these theoretical findings against observational data from the ERA5 reanalysis and NOAA/ERSSTv5 Niño 3.4 index, focusing particularly on the relationship between topological features and prolonged dry conditions in Southeast Asia. This approach provides new insights into the non-systematic relationship between strong El Niño events and regional climate impacts, while establishing a novel framework for comparing theoretical models with observational data. Our results demonstrate the utility of topological methods in understanding complex climate phenomena and suggest new possibilities for improving ENSO prediction capabilities.

How to cite: Sanchez Muniz, M., Brown, M., and Paranamana, P.: Characterizing ENSO Through Topological Analysis of Jin-Timmermann Model's Chaotic Regimes, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20545, https://doi.org/10.5194/egusphere-egu25-20545, 2025.

Extreme rainfall events during the Indian monsoon season pose significant challenges due to their socioeconomic and environmental impacts. Understanding the spatial and temporal dynamics of these events requires robust analytical and statistical methods capable of capturing complex relationships within rainfall generating systems. Complex network approaches have emerged as powerful tools for analyzing spatiotemporal patterns in climate data, offering new insights into extreme weather phenomena.

This study compares two methodologies for constructing and analyzing climate networks to study the spatiotemporal structure and dynamics of heavy precipitation events in India during the monsoon season across multiple time scales. Specifically, we introduce a novel combination of Discrete Wavelet Decomposition with Event Coincidence Analysis (ECA), referred to as Multi-Scale Event Coincidence Analysis (MSECA) and compare the results with the existing Multi-Scale Event Synchronisation (MSES). From a conceptual perspective, MSECA appears to be a more reasonable method compared to MSES, as it mitigates certain undesired effects of temporal clustering of rainfall extremes across various timescales.

Our results reveal distinct differences in network properties depending on the methodology used, highlighting the sensitivity of network-based analyses to the choice of construction technique. These differences affect the identification of dominant heavy rainfall patterns and their underlying drivers, such as large-scale atmospheric circulation and/or local feedback mechanisms at daily to monthly temporal scales.

Our work underscores the importance of methodological rigor and the potential of complex network approaches in advancing the understanding of extreme rainfall events in monsoon-dominated regions. This comparison provides a foundation for developing standardized practices for network-based climate studies, enabling more robust assessments of extreme weather phenomena.

How to cite: Bishnoi, G., Dhanya, C. T., and Donner, R. V.: A Comparison of Methodologies for Studying Heavy Precipitation Events during the Summer Monsoon Season in India Using Complex Network Approaches, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21222, https://doi.org/10.5194/egusphere-egu25-21222, 2025.

While precipitation is the primary driver of streamflow variability, temperature also plays a significant role. Temperature influences streamflow by modifying precipitation, evapotranspiration, and soil moisture. While this relationship is often studied using hydrological or black-box models, the causal effect of temperature dynamics on streamflow at the catchment scale is not fully understood. This study investigates the causal relationship between precipitation, temperature, and streamflow time series using the PCMCI+ causal discovery method. Having the causal structure, the total causal effect of temperature on stream flow is estimated. The analysis is conducted on CAMELS-GB (671 catchments) and LamaH (859 catchments) datasets to study the causal effects of temperature on streamflow across a wide range of catchments with different climate and physiographic characteristics. Preliminary results indicate that temperature significantly influences streamflow within a specific range, which changes over time for most catchments. The changes in the range within which the temperature has high causal effects on the temperature might be due to the shift in catchment storage and precipitation patterns, leading to a change in catchment response to temperature. These findings highlight the importance of identifying a relationship between temperature streamflow variability from a cause-and-effect perspective. This suggests that incorporating causal information can improve the modelling of the hydrological systems under changing climate. 

How to cite: Abbasizadeh, H., Maca, P., and Hanel, M.: Influence of Temperature on Streamflow Dynamics: A Multi-Catchment Analysis Using the PCMCI+ Causal Discovery Algorithm, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-292, https://doi.org/10.5194/egusphere-egu25-292, 2025.

River discharge prediction is critical for water resource management, yet equifinality—where multiple model configurations achieve similar accuracy—complicates process understanding. We explored this phenomenon using Long Short-Term Memory (LSTM) models trained on UK river basins, incorporating geomorphic descriptors derived from Digital Terrain Models and other environmental features from the CAMELS-GB dataset, including land cover, soil, and climate variables. Explainable AI techniques revealed that the models rely on different, yet equally effective, combinations of correlated features to achieve comparable performance. This variability underscores the complexity of hydrological systems and highlights the importance of integrating explainability and domain knowledge in machine learning to enhance model interpretability and robustness.

How to cite: Chen, Q., Mudd, S., and Moulds, S.: Equifinality in River Discharge Prediction Revealed Through Explainable AI , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-482, https://doi.org/10.5194/egusphere-egu25-482, 2025.

Accurate modeling of soil water movement in the unsaturated zone is essential for effective soil and water resources management. Physics-informed neural networks (PINNs) offer promising potential for this purpose, but necessitate retraining upon changes in initial or boundary conditions, posing a challenge when adapting to variable natural conditions. To address this issue, inspired by the operator learning with more universal applicability than function learning, we develop a physics-informed deep operator network (PI-DeepONet), integrating physical principles and observed data, to simulate soil water movement under variable boundary conditions. In the numerical case, PI-DeepONet achieves the best performance among three modeling strategies when predicting soil moisture dynamics across different testing areas, especially for the extrapolation one. Guided by both data and physical mechanisms, PI-DeepONet demonstrates greater accuracy than HYDRUS in capturing spatio-temporal moisture variations in real-world scenario. Furthermore, PI-DeepONet successfully infers constitutive relationships and reconstructs missing boundary flux condition from limited data by incorporating known prior physical information, providing a unified solution for both forward and inverse problems. This study is the first to develop a PI-DeepONet specifically for modeling real-world soil water movement, highlighting its potential to improve predictive accuracy and reliability in vadose zone modeling by combining data-driven approaches with physical principles.

How to cite: Ye, Q., Huang, Z., Zheng, Q., and Zeng, L.: Predicting Water Movement in Unsaturated Soil Using Physics-Informed Deep Operator Networks, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1863, https://doi.org/10.5194/egusphere-egu25-1863, 2025.

Recent applications have demonstrated the strength of deep learning (DL) in information extraction and prediction. However, its limitations in interpretability have delayed its popularity for use in facilitating advancement of hydrologic understanding. Here we present a framework using explainable artificial intelligence (XAI) as a diagnostic tool to investigate distributed soil moisture dynamics within a watershed. Soil moisture and its movement generated by physically based hydrologic model were used to train a long short-term memory (LSTM) network, whose feature attribution was then evaluated by XAI methods. The aggregated feature importance presents abrupt rise in the model’s nodes located in riparian area, indicating threshold behavior in runoff generation and development of hydrologic connectivity at the watershed scale, which helps explain the rapid increase in streamflow. This work represents a demonstration of the potential of XAI to uncover underlying physical mechanisms and to help develop new theories from observed data.

How to cite: Ye, S., Li, J., Chai, Y., Liu, L., Sivapalan, M., and Ran, Q.: Using explainable artificial intelligence as a diagnostic tool , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2661, https://doi.org/10.5194/egusphere-egu25-2661, 2025.

How to cite: Huynh, N. N. T., Garambois, P.-A., Renard, B., and Monnier, J.: A General Framework for Integrating Neural Networks into Numerical Resolution Methods for Spatially Distributed Hydrological Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2740, https://doi.org/10.5194/egusphere-egu25-2740, 2025.

Flood susceptibility mapping (FSM) plays a crucial role in proactive flood risk management, particularly in light of increasing fluvial flooding events. Traditional FSM methods, such as physics-based and qualitative approaches, are hindered by either high computational demands or inherent uncertainty. To address this, machine learning (ML) models have become an increasingly popular FSM approach, though commonly cited as black-box approaches due to the difficulty associated with understanding their underlying mechanisms. In order to better understand the ML approaches used for FSM, this study uses the gradient-weighted class activation mapping (Grad-CAM) to interpret flood susceptibility predictions of a convolutional neural network (CNN) for the Don River watershed in Ontario, Canada. Grad-CAM is an explainable algorithm highlighting input regions that are influential to the output, aiding the user in understanding and visualizing model selected important features used to arrive at the prediction. Grad-CAM results are compared to the commonly used shapley additive explanation (SHAP) algorithm. SHAP is used to calculate the relative contribution of each input onto the output, and provides a benchmark for comparisons due to its popularity.

A two dimensional CNN with an architecture of two convolutional layers, two pooling layers and a fully connected layer is used to predict flood susceptibility. The inputs to the CNN include topographical and climactic variables across the entire watershed, with a 60-40% training and testing split respectively. The results of the CNN were compared against the floodplain map of the Don River. Using the area under curve- receiver operating characteristics (AUC-ROC) as a performance metric, the CNN exhibits high performance with an AUC-ROC of 0.96.

The study highlights the potential of CNNs for flood susceptibility mapping, as well as compares two explainable machine learning algorithms, helping to further their application within FSM. Explainable algorithms are essential to decision makers in flood risk management for proactive planning and resource allocation. Future work should explore expanding the scope to predict flood susceptibility at a nationwide level.

How to cite: Khalid, R. and Khan, U. T.: Explainable convolutional neural network for flood susceptibility mapping in Southern Ontario  , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3859, https://doi.org/10.5194/egusphere-egu25-3859, 2025.

Floods are the costliest hazard in Canada in terms of direct infrastructure damage. Flood susceptibility modelling (FSM) identifies flood hazard areas; input features are dependent on the study area and modelling methods, which affect the reliability and accuracy of FS maps. Typical features in FSM are static topographical inputs (digital elevation model, land use, wetness index, height above nearest drainage, etc.). Though meteorological variables have been included in FSM, they are often low temporal resolution (e.g. annual); seasonal meteorological variables are often not included. The 2023 Canadian National FS map was developed using machine learning (ML) ensembles, with features that include historical flood events and 30 years of climate data. This research initiates the update to the existing Canadian FS map by expanding the suite of input features used and comparing the impact of three feature selection methods (partial correlation, partial mutual information, combined neural pathway strength) on three types of ML algorithms: random forest, artificial neural network (ANN) and convoluted neural network (CNN). The expanded set of features includes geospatial indices and flood-specific meteorological data such as spring temperature, precipitation, and vapour pressure. Data from preceding seasons to specific flood events is also included. Preliminary findings from the feature selection methods show that including seasonal flood-specific meteorological data provides important information leading to better model performance. Model performances of the three algorithms were comparable. Random forest with extreme gradient boosting led to the highest model performance (AUC = 0.98, F1 = 0.94), followed by CNN (AUC = 0.0.96, F1 = 0.90). ANN ensemble with leave-one-out-cross-validation resulted in the lowest model performance (AUC = 0.91, F1 = 0.85). Results contribute to the development of an improved national FS map for Canada.

How to cite: Dunbar, K. E., McGrath, H., and Khan, U. T.: Enhancing flood susceptibility modelling in Canada: Integrating seasonal meteorological data, feature selection and machine learning approaches, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3871, https://doi.org/10.5194/egusphere-egu25-3871, 2025.

Effective flood forecasting is critical for informed decision-making and timely emergency response. Traditional physical models, which rely on fixed-resolution spatial grids and input parameters, often incur substantial computational costs, limiting their capacity to accurately predict flood peaks and provide prompt hazard warnings.  This paper introduces methods to ensure physical consistency in machine learning models, aiming to develop a fast, stable, accurate, cross-regional, and downscaled neural flood forecasting foundation model. Specifically, we present a Physics-embedded Neural Network, which integrates the momentum and mass conservations of flood dynamics into a neural network. Additionally, we combine this Physics-embedded Neural Network with a diffusion-based generative model, enhancing physical process consistency for long-term, large-scale flood forecasting. We also briefly introduce other models that integrate physics and machine learning, such as the FloodCast model by incorporating hydrodynamic equations into its loss function to maintain physical consistency, and the UrbanFloodCast model by learning physical consistency from urban flood dynamic data. The performance of these models will be analyzed using our proposed FloodCastBench dataset, a comprehensive collection of low-fidelity and high-fidelity flood forecasting dataset and benchmark. Results from the dataset demonstrate that incorporating physical consistency significantly enhances flood forecasting accuracy, demystifies the black-box nature of machine learning frameworks, and increases confidence in addressing dynamical systems. Finally, we propose a Spatiotemporal Foundation Model capable of forecasting floods across a variety of scales and regions.

How to cite: Xu, Q. and Zhu, X. X.: Towards Physics-consistent Foundation Models for Flood Forecasting, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4428, https://doi.org/10.5194/egusphere-egu25-4428, 2025.

Catchments are the fundamental units of hydrological analysis and integrate a vast number of physical, biological, and anthropogenic processes. Traditional hydrological modelling approaches, however, adopt a bottom-up perspective, aggregating small-scale physical principles to predict large-scale catchment behaviour. While effective for prediction, this approach can fall short in advancing our understanding of emergent processes and their interactions given the strong dependence on a priori assumptions. To address this gap, causal discovery algorithms offer a promising alternative by moving beyond simple correlation to directly identifying the dynamic causal structures emerging at the catchment scale. In this study, we applied the PCMCI+ algorithm to the CAMELS-US dataset in combination with a subsequent causal effect estimation. We explored how and to what extent dynamic causal structures can be learned from hydro-meteorological data alone, and which catchment properties and conditions influence their expression. We find that causal discovery in hydrology faces challenges due to non-stationarity, unsuitable conditional independence tests, and unmet methodological assumptions. Despite these limitations, our approach reconstructed physically plausible relationships controlled by meaningful catchment properties. These results highlight the potential of causal discovery in hydrology, where it could serve as a complementary framework for model evaluation studies or as an integral part of the model development process.

How to cite: Strahl, D., Gnann, S., Wiesner, K., and Wagener, T.: Using Causal Discovery to Identify Drivers and Controls of Streamflow in Large Sample Hydrology, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4845, https://doi.org/10.5194/egusphere-egu25-4845, 2025.

Water resources management depends heavily on hydrological modeling for reservoir operation and risk mitigation, especially in data-scarce regions. Hybrid approaches that combine artificial intelligence and conceptual models offer great potential for accurate streamflow prediction. However, their implementation can be time-consuming and applied in different configurations. This study comprehensively compares two promising hybrid frameworks: the Conceptual-Data-Driven Approach (CDDA) and the Ensemble Approach. The analysis was conducted in the Upper Blue Nile Basin in Ethiopia over the period from 2002 to 2019. Six baseline models were developed, including CNN-LSTM (data-driven), NAM and HBV-Light (lumped), and SWAT+, WEAP, and HEC-HMS (semi-distributed). All models achieved NSE ≥ 0.85 during the validation period, with CNN-LSTM performing best (NSE = 0.94). Each model was integrated into the two hybrid frameworks using Random Forest (RF) or Artificial Neural Networks (ANN). Results showed that the Ensemble Approach outperformed CDDA by combining two conceptual models. ANN performed better than RF across both frameworks. Hybrid modeling significantly improved semi-distributed models, while lumped and data-driven models showed minimal benefits. In the Ensemble Approach, normal and extreme flows simulated using semi-distributed models performed best when supported by CNN-LSTM or lumped models. Our analysis also demonstrated the robustness of the Ensemble Approach for selecting the supporting model. These findings emphasize the value and feasibility of the Ensemble Approach for improving streamflow prediction and better supporting decision-making in data-scarce regions. Nevertheless, a thorough understanding of the opportunities in hybrid modeling requires further research with a specific focus on operational forecasting.

How to cite: Mohamed, A., M. Ali, A., Ali, A., Hassan, O., E. Elbasheer, M., and Abdelaziz, M.: Can CNN-LSTM and lumped models improve (extreme) streamflow prediction of semi-distributed models? A comparative analysis of two hybrid frameworks, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5052, https://doi.org/10.5194/egusphere-egu25-5052, 2025.

Model Intercomparison Projects in the Earth Sciences have shown, that the outputs of
Earth System Models often show large variations and can therefore give quite different results,
with no single model consistently outperforming others. Examples include Global Water
Models (GWMs), as well as Global Climate Models (GCMs). The high computational costs
of running such models make comprehensive statistical analyses challenging, a common issue
with many complex models today. Machine learning models have become popular surrogates
of slow process-based models, due to their computational speed, at least once trained. This
speed makes it possible to use techniques from Explainable AI (XAI) to analyze the behavior
of the surrogate model.
Here, we analyze long-term averages of the GWM ’Community Water Model’ (CWatM)
for different parts of the global domain for actual evapotranspiration E_a, total runoff Q and
groundwater recharge R. We train an artificial neural network on the model’s input and output
data and use three different strategies to assess the importance of input data: LassoNet for sub-
set selection and feature ranking, along with Sobol’ indices and DeepSHAP for interpretability.
Our results show that subset selection can effectively reduce model complexity before XAI
analysis. For some hydrological domains the number of relevant input
variables for a chosen output reduces to less then 15 variables out of 98 model inputs, while
others remain more complex, requiring many variables for performances with R² > 0, 8.

How to cite: Faber, L., Wiesner, K., Tang, T., Wada, Y., and Wagener, T.: Using Explainable Artifical Intelligence (XAI) to Analyze the Behavior of Global Water Models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5727, https://doi.org/10.5194/egusphere-egu25-5727, 2025.

Physics-based hydrological modelling provides great opportunities for risk assessment and water resources management. However, diagnostic model evaluation and quantitative uncertainty assessment remain a challenge: (1) Model choices, boundary conditions, and prior assumptions about input, parameter or data uncertainty might be hard to formulate or justify; (2) rigorous propagation of uncertainties struggles when the analysed model structure is not “true”, and (3) a full propagation of uncertainties is often computationally prohibitive for complex models.

Alternative approaches promote the extraction of information directly from data, thereby avoiding overly strict physics-based constraints and the pitfalls of uncertainty quantification. Challenges of these data-driven approaches include the lack (or difficulty of) explainability, transparency, and transferability to unseen scenarios.

To explore the frontier of where those two perspectives (should) converge, we investigate the potential of surrogate models (computationally cheaper, data-driven representations of complex models) as a binding link with several potential benefits: (1) they alleviate the computational burden and thereby allow for a fully Bayesian uncertainty analysis; (2) they are flexible enough to overcome structural deficits of the original complex model, thereby enabling a better predictive performance, and (3) being data-driven, we can elegantly fuse the information from available data into their training process.

Methodologically, we propose a weighted data-integrated training of surrogates via two competing approaches that differ technically, but also philosophically, and reveal complementing insights about the strengths and weaknesses of the physics-based model and about the additional information in the available data, thereby facilitating deeper system understanding and improved (hybrid) modelling. We demonstrate the proposed workflow on didactic examples and a real-world case study. We expect this approach to be generally useful for modelling dynamic systems, as it contributes to more realistic uncertainty assessment and opens up ways for model development.  

How to cite: Guthke, A., Reiser, P. L., and Bürkner, P.: Training Surrogates with Knowledge and Data: A Bayesian Hybrid Modelling Strategy, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6278, https://doi.org/10.5194/egusphere-egu25-6278, 2025.

Neural networks are efficient and effective in predicting system states in hydrology. However, most current approaches lack hydrological flow partitioning, do not allow for training on measurements of multiple variables, or lack capability to tightly integrate physically-based components. To address these shortcomings I propose and evaluate an approach referred to below as Dynamical System Neural Network (DSNN). DSNN is a feedforward neural network with an architecture that resembles the organisation in components of the real-world system it represents. In hydrology, the DSNN represents each water flow (e.g. seepage, snow melt) by a collection of input, hidden, and output neural layers, where each input is the state of a hydrological storage (e.g. groundwater storage influencing seepage) or other variable (e.g. air temperature influencing snow melt). These components are interconnected to form a single neural network of the complete dynamical system considered, where all storages and flows are explicitly quantified. If physical understanding of a flow and its parameterization is available, a known formulation can be used as a replacement of a neural network component. The DSNN is applied forward in time, backpropagating gradients over all timesteps. It can be run in spatially lumped or semi-distributed mode. To demonstrate the approach, a DSNN is presented of the Austrian Dorfertal (Kals) Alpine catchment containing snow and subsurface water storages and associated flows including streamflow. The DSNN is trained, validated, and tested on daily streamflow over ~40 years. To explore the capability of the DSNN in estimating the magnitude and dynamics of internal system storages (snow water equivalent, subsurface water storage) and flows (evapotranspiration, sublimation, snowmelt, seepage), the DSNN is first trained and tested with streamflow data generated by a conceptual model. The DSNN turns out to be capable of reproducing - with a satisfactory level of precision - the system states and fluxes calculated by the conceptual model, with decreasing performance when measurement error is added to the artificially generated streamflow data before training. To explore its predictive performance, the DSNN is applied on measured streamflow data for the Dorfertal, comparing multiple DSNN setups that represent all flows as neural network components or only a subset of flows where remaining flows are represented with a standard conceptual model (e.g. linear reservoir). Preliminary results indicate that in predictive performance, in most setups, the DSNN outperforms a standard conceptual model trained on the same streamflow data, with NSE values for testing of 0.74 and 0.71, respectively. This preliminary result indicates DSNN to be a promising approach for blending process-based and neural network based modelling as well as for training (i.e. calibration) of neural network models on measurements of multiple hydrological variables as these are all explicitly represented by the DSNN and can thus be incorporated in the loss function (e.g. streamflow, snow depth, groundwater, evapotranspiration).

How to cite: Karssenberg, D.: Dynamical system neural network for hydrological modelling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7382, https://doi.org/10.5194/egusphere-egu25-7382, 2025.

How to cite: Zheng, Y., Wang, C., and Jiang, S.: Advancing distributed hydrological modeling with hybrid machinelearning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7770, https://doi.org/10.5194/egusphere-egu25-7770, 2025.

The integration of remote sensing observations into hydrological modeling frameworks presents a significant opportunity for improving spatial and temporal predictive capabilities across continental domains. This research introduces a novel hybrid distributed hydrological model that addresses key challenges in computational efficiency, by using a GPU-enabled computational infrastructure, and in predictive accuracy by assimilating multi-source remote sensing datasets, specifically satellite-based soil moisture and evapotranspiration, at a high spatial resolution (1km×1km) and temporal scale (hourly). The model addresses critical challenges in regional hydrological forecasting by leveraging advanced data assimilation techniques and machine learning methodologies.

The proposed hybrid modeling framework synthesizes physically-based distributed hydrologic modeling principles with data-driven machine learning approaches, facilitating a more comprehensive representation of land surface hydrological processes. A key innovation is the GPU-enabled cell-to-cell routing algorithm, which enables fast and efficient computational processing of complex hydrological connectivity and water movement across large spatial domains. By integrating remote sensing observations, the methodology enables enhanced initial condition specification and improved parameter estimation, particularly in regions characterized by sparse ground-based measurement networks.

Preliminary analytical results demonstrate significant improvements in model performance, particularly in capturing spatial and temporal variability of hydrological states and fluxes. The approach substantively advances current methodological capabilities in hydrological forecasting, offering a promising framework for developping enriched tensorial numerical solvers, addressing complex hydroclimatic prediction challenges in data-limited environments.

How to cite: Ettalbi, M., Garambois, P.-A., Huynh, N.-N.-T., Ferreira, E., and Baghdadi, N.: GPU-Enabled Cell-to-Cell Routing in a High Resolution Hybrid Distributed Hydrological Model with Multi-Source Remote Sensing Data Assimilation: A Continental-Scale Computational Approach , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9176, https://doi.org/10.5194/egusphere-egu25-9176, 2025.

The formation of floods, as a complex physical process, exhibits dynamic changes in its driving factors over time and space under climate change. Due to the black-box nature of deep learning, its use alone does not enhance understanding of hydrological processes. The challenge lies in employing deep learning to uncover new knowledge on flood formation mechanism. This study proposes an interpretable framework for deep learning flood modeling that employs interpretability techniques to elucidate the inner workings of a peak-sensitive Informer, revealing the dynamic response of floods to driving factors in 482 watersheds across the United States. Accurate simulation is a prerequisite for interpretability techniques to provide reliable information. The study reveals that comparing the Informer with Transformer and LSTM, the former showed superior performance in peak flood simulation (NSE over 0.6 in 70% of watersheds). By interpreting Informer’s decision-making process, three primary flood-inducing patterns were identified: precipitation, excess soil water, and snowmelt. The controlling effect of dominant factors is regional, and their impact on floods in time steps shows significant differences, challenging the traditional understanding that variables closer to the timing of flood event occurrence have a greater impact. Over 40% of watersheds exhibited shifts in dominant driving factors between 1981-2020, with precipitation-dominated watersheds undergoing more significant changes, corroborating climate change responses. Additionally, the study unveils the interplay and dynamic shifts among variables. These findings suggest that interpretable deep learning, through reverse deduction, transforms data-driven models from merely fitting nonlinear relationships to effective tools for enhancing understanding of hydrological characteristics.

How to cite: Xu, Y. and Lin, K.: Uncovering the Dynamic Drivers of Floods through Interpretable Deep Learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10256, https://doi.org/10.5194/egusphere-egu25-10256, 2025.

Earth system data, measured by satellites and terrestrial stations and simulated by increasingly complex models, provide valuable information for identifying functional relationships within the Earth system. These relationships are essential for understanding complex interactions and predicting changes, for example, in climatic or ecological processes, but often only occur in certain spatiotemporal sections or within certain threshold values. With the increasing spatiotemporal resolution of remote sensing products and models, a manual analysis is impractical, and hypothesis-driven approaches can lead to undiscovered hidden relationships. Previous work proposed the SONAR (automated diScovery Of fuNctionAl Relationships) decision-tree algorithm to automatically search for functional relationships in earth system data without a-priori assumptions. We analyzed the proposed algorithm using artificially generated data to evaluate SONAR's functionality.  We tested if the choice of statistical indicator (Pearson’s r, Spearman’s ρ, Kendall’s τ, and Mutual Information) influences the functionality of the SONAR algorithm and which factors are important for the identification of functional relationships. Using 1512 synthetic data sets and the developed SAMPI (Similarity of A Manifested and Prototypical decision tree Indicator) coefficient, we demonstrate how the performance of the algorithm changes under different variations of the data sets - including the number of designated splits, the presence of interfering variables and the strength and nature of the underlying functional relationships. In particular, we show which statistical indicator provides the best results under these conditions. The results demonstrate that the SONAR algorithm is very versatile, especially when employing the most reliable statistical indicator. The SONAR algorithm could, therefore, have far-reaching applications, for example, in analyzing climatic patterns or investigating dependencies between environmental factors.

How to cite: Thöne, C., Bäthge, A., and Reinecke, R.: The effects of different statistical indicators in the new decision-tree-based SONAR algorithm for automated detection of functional relationships in Big Earth Data, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10444, https://doi.org/10.5194/egusphere-egu25-10444, 2025.

Terrestrial evaporation (E) is an essential climate variable, linking water, energy and carbon cycles. As E is influenced by the state of the atmospheric boundary layer, vegetation and soil, its modelling is a complex task, resulting in a myriad of simulation approaches. To combine the strong predictive skills of data-driven models with the interpretability and physical consistency of process-based models (PBMs), a new research field of differentiable modelling has emerged¹. 

Here, we present a differentiable framework for E estimation, facilitating online training of NNs as intermediate PBM components. It is inspired by the GLEAM framework for estimating E, which applies offline training (i.e., outside the PBM) of neural networks (NNs) predicting evaporative stress^2,3. Building upon the Julia SciML ecosystem’s implementation of universal differential equations⁴, a wide array of numerical methods are available for solving the PBM’s ordinary differential equations (ODEs) and calculating the parameter sensitivities⁵. In this way, the effect of the numerical methods on the obtained hybrid model can be investigated, moving beyond the direct automatic differentiation through explicit Euler solutions of ODEs as often applied in other hydrological hybrid modelling approaches. 

References

¹Shen, C., Appling, A.P., Gentine, P. et al., Differentiable modelling to unify machine learning and physical models for geosciences, Nat. Rev. Earth. Environ., 4, 552–567, 2023, https://doi.org/10.1038/s43017-023-00450-9

²Koppa, A., Rains, D., Hulsman, P. et al., A deep learning-based hybrid model of global terrestrial evaporation, Nat. Commun., 13, 1912, 2022, https://doi.org/10.1038/s41467-022-29543-7

³Miralles, D. G., Bonte, O., Koppa, A. et al., GLEAM4: global land evaporation dataset at 0.1° resolution from 1980 to near present, preprint, 2024, https://doi.org/10.21203/rs.3.rs-5488631/v1 

⁴Rackauckas, C., Ma, Y.,  Martensen, J. et al., Universal differential equations for scientific machine learning, ArXiv, 2020, https://doi.org/10.48550/arXiv.2001.04385 

⁵Sapienza, F., Bolibar, J., Schäfer, F. et al., Differentiable Programming for Differential Equations: A Review, ArXiv, 2024, https://doi.org/10.48550/arXiv.2406.09699 

How to cite: Bonte, O., Miralles, D. G., Koppa, A., and Verhoest, N. E. C.: Universal differential equations for estimating terrestrial evaporation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11693, https://doi.org/10.5194/egusphere-egu25-11693, 2025.

How to cite: Chang, A. Y.-Y., Leonarduzzi, E., Grams, C. M., and Humphrey, V. W.: A Physics-Constrained Emulator for High-Resolution Soil Moisture , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12068, https://doi.org/10.5194/egusphere-egu25-12068, 2025.

Terrestrial evaporation (E) is a critical component of the water cycle, returning nearly 60% of continental precipitation to the atmosphere and dissipating approximately 50% of surface net radiation. A prevalent approach for estimating E involves computing a theoretical maximum, known as potential evaporation (E_p), and scaling it based on a multiplicative stress factor, often referred to as “evaporative stress” (S) or “transpiration stress” (S_t) when specifically applied to plant transpiration. Like stomatal or surface conductance, S_t is governed by a complex nonlinear interplay of environmental drivers such as soil moisture, air temperature, radiation, and atmospheric vapor pressure deficit. This complexity is not yet fully understood, which further hampers its accurate physical modelling and limits our ability to comprehend transpiration’s sensitivity to the changing environment.

The fourth generation of the Global Land Evaporation Amsterdam Model (GLEAM4) has yielded a global dataset of transpiration by integrating multi-source remote sensing data following a hybrid approach, in which E_p is computed based on a process-based model and S_t is calculated by employing deep neural networks. These neural networks are trained on global eddy covariance and sap flow measurements for both tall and short vegetation, and are informed by a set of environmental controls or biotic factors. These factors include soil moisture, vapor pressure deficit, atmospheric CO₂ concentration, wind velocity, air temperature, downwelling shortwave radiation, LAI, and vegetation optical depth. Beyond the predictive capabilities of these deep neural networks, the relationships between environmental controls and S_t within these neural networks remain under exploration, leaving uncertainty as to whether GLEAM4 accurately represents real-world processes. To explore the relationships, we employ the SHapley Additive exPlanation (SHAP) method, which quantifies the marginal contributions of predictors to model predictions, offering insights into the relative importance of environmental drivers in determining S_t.

Our findings highlight dominant S_t drivers across various climatic regimes and ecosystems, revealing their contributions' temporal evolution. Additionally, we investigate how S_t responds to shifts in environmental conditions, including climate and vegetation changes, water stress, atmospheric aridity, and rising CO₂ levels. Our study enhances global understanding of transpiration dynamics and provides critical insights into the impacts of diverse hydroclimatic drivers, thereby supporting broader applications within the hydrology and climate communities.

How to cite: Ruan, F., M. Baez-Villanueva, O., Bonte, O., Koppa, A., Li, W., Camps Valls, G., Yang, Y., and G. Miralles, D.: Global Vegetation Stress Drivers based on Hybrid Modelling and Explainable AI, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12928, https://doi.org/10.5194/egusphere-egu25-12928, 2025.

Vegetation plays an important but complicated role in modulating land-atmosphere interactions and the water cycle. Under global change, increasing vegetation greenness trends have been observed, which further complicate the control of vegetation in the earth system. Despite growing interest in the role of vegetation in the hydrological processes, large uncertainties still exist, particularly when it comes to the underexplored response of streamflow to vegetation greening. In this study, we explore the watershed-relevant biophysical controls of vegetation greening on streamflow. In order to do so, we develop a hybrid ecohydrological model. This model adheres to the water balance principles, while it simultaneously has a flexible structure that enables integrating physical insights from observational data. The multi-task learning optimization ensures physical consistency across a range of processes and temporal frequencies, which allows us to investigate the cascading impacts of vegetation changes across the water cycle, leading up to the streamflow as an end-process. Ecohydrological insights are directly derived from observational data, while physically meaningful model parameters reflect how ecosystem functions and hydrological processes respond to vegetation changes. We find that the marked change in streamflow can be attributed to vegetation change controls on diverse biophysical processes. Our research highlights the potential of hybrid models to capture complex earth system processes by exploiting multiple observational data streams, machine learning and physical constraints.

How to cite: Blougouras, G., Brenning, A., Migliavacca, M., and Reichstein, M.: Hybrid hydrological modelling of the biophysical impacts of earth’s greening on streamflow, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13311, https://doi.org/10.5194/egusphere-egu25-13311, 2025.

Cyanobacterial blooms have become more frequent and intense in Lake Superior since 2012, primarily due to increased nutrient loads, with phosphorus being the main limiting factor. To protect water quality, extensive monitoring of lakes and streams is crucial, but it is not cost-effective or practical to measure nutrients frequently across all ecosystems. This study presents a cost-effective, transferable solution using machine learning (ML) models to predict phosphorus concentrations and loads based on conventional water quality parameters like streamflow, dissolved oxygen, conductivity, turbidity, transparency, and total suspended solids. The research introduces an explainable hybrid ML framework combining probabilistic principal component analysis (P2CA) with several ML models, including Bagging Ensemble Learning, Boosting Ensemble Learning, Gaussian Process Regression, and Support Vector Regression, to enhance prediction accuracy. Results demonstrate that the P2CA-Boosting Ensemble Learning model consistently outperforms other approaches. To confirm its effectiveness, the developed model was tested with the same input data from a different river catchment, proving it works well in different environments. This study highlights the potential of combining P2CA with Boosting Ensemble Learning as a powerful tool for water quality management in streams and rivers.

How to cite: Kumar, A. and Zhang, K.: Development of a hybrid machine learning model to predict total phosphorus in streams over the north shore of Lake Superior, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13799, https://doi.org/10.5194/egusphere-egu25-13799, 2025.

The field of hydrology has undergone significant transformation over the past decade, driven by advancements in machine learning and data-driven techniques. A key breakthrough came from the work of Kratzert et al. (2018), who demonstrated that purely data-driven LSTM models could outperform traditional hydrological models in over 600 catchments across North America. However, while these models significantly improve predictive performance, they often sacrifice interpretability and explainability.

To address this trade-off, researchers have explored new approaches that merge physical principles with data-driven methods. One promising innovation is the concept of differentiable modeling, introduced by Chen et al. in 2022. This approach transforms physical models into differentiable functions, allowing neural networks to represent and learn model parameters. By doing so, differentiable modeling enhances flexibility while maintaining a foundation in physical principles.

This research presents a novel differentiable hydrological model called the Differentiable Distributed Regression Model (dDRM). The dDRM builds on the principles of differentiable modeling with the structure of a conceptually lumped model using a simplified representation of physics ("smooth" HBV model). Inspired by the simplicity of the LSTM model, which aggregates data at the catchment level rather than relying on a grid-based representation, we introduce four equally sized elevation zones instead of grid cells in the dDRM. These zones inherently reflect differences in hydrological processes, such as precipitation, temperature, and snowmelt dynamics, enabling the model to account for spatial heterogeneity while maintaining computational efficiency.

By leveraging the principles of differentiable modeling, the dDRM achieves a balance between explainability and predictive performance. To evaluate model performance, we tested the dDRM across sixty-three catchments in southern Norway, in a gauged setting. Only precipitation and temperature were used as input data. For benchmarking purposes, we also trained an LSTM model to the same catchments. 

Our results demonstrate that the dDRM outperforms the fine-tuned LSTM model in both daily predictions and cumulative runoff volumes. These findings underscore the potential of differentiable hydrological models to bridge the gap between performance and interpretability. By combining physical principles with data-driven techniques, the dDRM provides a pathway toward more effective and understandable forecasting tools in hydrology.

How to cite: Beil-Myhre, B., Shrestha, R., and Matheussen, B. V.: The Differentiable Distributed Regression Model (dDRM) Balancing Explainability and Predictive Performance, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15436, https://doi.org/10.5194/egusphere-egu25-15436, 2025.

Groundwater is a critical resource for drinking water supply, agriculture, and ecosystems in general. In regions facing water scarcity, such as Brandenburg (Germany), effective groundwater management is essential. This requires accurate assessments of groundwater dynamics, which data-driven models can deliver through efficient and reliable groundwater level (GWL) predictions. To effectively develop and apply data-driven models for groundwater level prediction, a deeper understanding of which and how the input features influence the groundwater level prediction is crucial.

Our primary objective is to assess the impact of the input features of a Deep learning (DL) model that predicts GWLs using feature attribution methods. Specifically, the influence of climatic features as well as different land use patterns is examined. This study employs a global DL model based on the Long Short-Term Memory (LSTM) architecture to predict seasonal GWLs for 16 weeks ahead. We utilize a comprehensive set of features, including dynamic features such as climatic variables (e.g., temperature, precipitation, relative humidity) and static features such as Corine land cover. By incorporating these, we aim to capture the complex interactions between climate, landuse and groundwater levels.

For the feature attribution itself, we apply the Shapley value sampling method. It analyses the effect of an alternation of an input feature to the respective chosen objective. The choice of that function is essential for the obtained results. We alternate the corresponding objective function in three distinct ways: first, by using the total change of the predicted GWL for the whole period of interest; second, per prediction horizon, i.e. per predicted week of the 16 week prediction; and third, through a decomposition into partial scale-respective signals of the period of interest using the discrete wavelet transform. Besides understanding which input features are most important for the predictive performance of the LSTM model, the results enable us to identify further aspects of the dynamics learned by the model. For example, if and when the model switches from extrapolation to prediction, and at which temporal scales different factors play a role; e.g. if forest vegetation is more important for seasonal or weekly effects on groundwater levels. This multi-faceted approach allows us to gain a deeper understanding of the factors influencing GWLs and their temporal dynamics, both for static and dynamic input features. Ultimately, feature attribution methods can enhance the awareness for reasonable land-use, hence, groundwater management and lead to better predictive models.

How to cite: Engel, M., Kunz, S., Wetzel, M., and Körner, M.: Multitemporal and Multiscale Feature Attribution Methods to Understand the Impact of Climatic and Land Use Features on the Prediction of Groundwater Levels, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15651, https://doi.org/10.5194/egusphere-egu25-15651, 2025.

The Integrated Multi-satellite Retrievals for GPM (IMERG) is a global satellite-based precipitation dataset that provides near real-time precipitation estimates by combining multiple satellite measurements. IMERG integrates microwave (MW) observations from low-orbit satellites with precipitation estimates inferred from the brightness temperature of geostationary infrared (IR) imagery. MW measurements provide accurate precipitation estimates due to their direct interaction with precipitation particles, while IR measurements offer broader spatial and temporal coverage by inferring precipitation from cloud-top brightness temperatures. Together, these complementary techniques balance precision and coverage to improve global precipitation monitoring. However, IR-based precipitation estimates are inherently less reliable due to the weak direct correlation between brightness temperature and precipitation. Conversely, MW-derived estimates are more accurate but spatially constrained by the limited footprint of low-orbit satellites. To investigate the contributing factors in IR precipitation error calibration, we leveraged ERA5 Land, a high-resolution reanalysis dataset that includes surface variables across nine domains, such as temperature, soil moisture, radiation, and vegetation indices. These variables offer a comprehensive lens for understanding the impact of the land surface on precipitation dynamics. We employed the XGBoost machine learning model to predict the errors in IR precipitation estimates relative to MW-derived benchmarks. Additionally, SHapley Additive exPlanations (SHAP) values were used to interpret the model’s predictions, uncovering how individual input features contribute to error correction.

Our findings indicate that the explainable machine learning model can correct the infrared (IR) precipitation estimates to resemble microwave (MW) products, achieving notable improvements across statistical metrics. In the preliminary analysis of 165 countries and territories, the XGBoost model’s calibration improved the RMSE in all validation datasets, with a median reduction of 19.89% and an average reduction of 22.5%. Similarly, the correlation coefficient improved, with a median increase of 18.43% and an average increase of 54.49%. Moreover, the spatial and temporal distributions of the variables' SHAP values show various patterns. The clustered spatial distribution may represent the local climate attributes in specific geographic regions, providing insights into how regional environmental factors influence precipitation estimates. Meanwhile, the temporal distribution may imply seasonal variation, which can help identify patterns in precipitation dynamics and refine IR-based calibration by accounting for temporal variability in precipitation processes. This study provides a robust framework for leveraging land surface variables to refine IR-based precipitation products. By integrating reanalysis data with machine learning models, we present a scalable solution for improving precipitation monitoring in data-sparse regions, particularly where MW observations are unavailable.

How to cite: Hung, H. T. and Wang, L.-P.: IRMerg: Enhancing Global Infrared Precipitation Estimates with Land Surface Variables and Contributing Factors Analysis Using Explainable Machine Learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16863, https://doi.org/10.5194/egusphere-egu25-16863, 2025.

Long Short-Term Memory (LSTM) networks have shown strong performance in rainfall–runoff modelling, often surpassing conventional hydrological models in benchmark studies. However, recent studies raise questions about their ability to extrapolate, particularly under extreme conditions that exceed the range of their training data. This study examines the performance of a stand-alone LSTM trained on 196 catchments in Switzerland when subjected to synthetic design precipitation events of increasing intensity and varying duration. The model’s response is compared to that of a hybrid model and evaluated against hydrological process understanding. Our study reiterates that the stand-alone LSTM is characterised by a theoretical prediction limit, and we show that this limit is below the range of the data the model was trained on. We show that saturation of the LSTM cell states alone does not fully account for this characteristic behaviour, as the LSTM does not reach full saturation, particularly for the 1-day events. Instead, its gating mechanisms prevent new information about the current extreme precipitation from being incorporated into the cell states. Adjusting the LSTM architecture, for instance, by increasing the number of hidden states, and/or using a larger, more diverse training dataset can help mitigate the problem. However, these adjustments do not guarantee improved extrapolation performance, and the LSTM continues to predict values below the range of the training data or show hydrologically unfeasible runoff responses during the 1-day design experiments. Despite these shortcomings, our findings highlight the inherent potential of stand-alone LSTMs to capture complex hydro-meteorological relationships. We argue that more robust training strategies and model configurations could address the observed limitations, ensuring the promise of stand-alone LSTMs for rainfall–runoff modelling.

How to cite: Baste, S., Klotz, D., Espinoza, E., Bardossy, A., and Loritz, R.: The Extrapolation Dilemma in Hydrology: Unveiling the extrapolation properties of data-driven models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16971, https://doi.org/10.5194/egusphere-egu25-16971, 2025.

Precise streamflow forecasting in river systems is crucial for water resources management and flood risk assessment. This study focuses on the Tagus Headwaters River Basin (THRB) in Spain, a key hydrological basin providing essential water for urban, industrial, and irrigation purposes. Additionally, a significant portion of its water resources is transferred to the Segura River Basin through the Tagus-Segura water transfer, Spain’s most extensive hydraulic infrastructure. Given that nearly all available water in the THRB is allocated for these demands, precise streamflow forecasting is vital. For streamflow estimation in this basin, we evaluated the Soil and Water Assessment Tool (SWAT+), a physically-based model, and three AI-based models: support vector regression (SVR), feed-forward neural network (FFNN), and long short-term memory (LSTM) models, across four gauging stations within the THRB. For the AI-based models, rainfall and time-lagged runoff data were used as input data. Additionally, an ensemble machine learning technique was evaluated, using the outputs of both physically-based and AI-based individual models as inputs for the ensemble model. The results show that the AI-based models and the ensemble machine learning technique significantly outperformed the SWAT+ model. While the precision of the AI-based models was considerably higher than that of the SWAT+ model, the application of the ensemble technique enhanced the precision of the AI-based models by 18 to 26% during the calibration period and 4.1 to 9.2% during the validation period. Furthermore, the Shapley Additive Explanations (SHAP) methodology was used to explore how each model contributes to the predictions in the ensemble technique. This work was supported by the Spanish Ministry of Science and Innovation, under grants PID2021-128126OA-I00.

How to cite: Asadi, S., Jimeno-Sáez, P., López-Ballesteros, A., and Senent-Aparicio, J.: In the application of physically-based and interpretable AI-based models for streamflow simulation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18102, https://doi.org/10.5194/egusphere-egu25-18102, 2025.

 Soil moisture, one of the essential climate variables, forms a fundamental bridge between hydro-meteorological processes and influence climate dynamics. It is extremely variable and is driven by numerous hydrological, agricultural and ecological factors. Soil moisture subsequently impacts soil forming processes, root zone water availability, infiltration rates, runoff, groundwater storage and vegetation-soil interaction. Despite its significant contribution in hydro-ecological interaction, its variability at subsurface is not yet explored adequately. Precise estimation of soil moisture at various depths is crucial because it affects water retention characteristics and modulates the vertical and lateral movement of water within the soil profile. This subsurface information is integral to understanding recharge rates, groundwater interactions, and the overall water balance within a catchment. In this study, we present an automated machine learning framework designed to predict soil moisture at multiple depths of 10 cm, 20 cm, 30 cm, and 40 cm leveraging Bayesian optimization. We collected data from our hydrological observatory set up constituting an automatic weather station, a pan evaporimeter and a soil moisture recorder. To evaluate model performance, we categorized the dataset into four scenarios (S1, S2, S3, and S4), with each subsequent scenario incorporating a greater number of observations and rainfall events. We used 11 input features to train this AutoML model by integrating several hydrological and meteorological variables with in-situ soil moisture data. Among the predictor variables, humidity, dew point, and rainfall emerged as the most influential factors driving soil moisture variability. The model was trained to calculate the performance metrices for the entire dataset and for subsets containing only rainfall instances. Our optimized model demonstrated superior performance, with an R² of 0.88–0.99 and RMSE < 0.022 for the overall dataset, and R² of 0.76–1.00 with RMSE < 0.06 for rainfall-specific data across all soil moisture depths.

How to cite: Singh, V., Singh, A., and Gaurav, K.: An automated machine learning framework for multi-depth soil moisture prediction using hydro-meteorological datasets, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-18468, https://doi.org/10.5194/egusphere-egu25-18468, 2025.

The delineation of catchment areas from elevation is a fundamental step in lumped process-based models (PBMs). Most machine learning (ML) approaches for rainfall-runoff modeling spatially aggregate inputs to represent basin-wide processes. Elevation-based lumping, however, disregards both human interventions such as drainages and underground hydrological flows, which can lead to significant model inaccuracies. In this work, we employ DRRAiNN (Distributed Rainfall-Runoff Artificial Neural Network) – a fully distributed neural network architecture – to infer catchment areas directly from observed precipitation and discharge dynamics without prior delineations.

As a first evaluation of the potential to infer actual catchment areas with DRRAiNN, we trained the model on relatively sparse data from 2006 until 2015: Radolan-based hourly precipitation data as input with a spatial resolution of 4x4 km and only daily discharge measurements from 17 stations in the Neckar river basin as target output. Elevation and solar radition were given as additional parameterization input. As DRRAiNN is fully differentiable, we were then able to infer station-specific attribution maps via backpropagation through space and time. To evaluate the alignment between the inferred attribution maps and elevation-based catchment areas, we compute the Wasserstein distance between attributions inside and outside the catchment boundaries. A higher distance indicates better agreement. The results show that DRRAiNN learns to propagate water in a physically plausible manner. Further, we reveal deviations that indicate additional water flows that are undetectable from elevation data alone. Our findings thus suggest that DRRAiNN captures key rainfall-runoff dynamics while avoiding the limitations of lumped models.

The quantitative evaluations alongside qualitative comparisons underscore the model’s potential for uncovering hidden hydrological processes. We show that catchment area estimates can be inferred from relatively little discharge data, which may, in the future, potentially be substituted by satellite data. As a result, DRRAiNN may be applicable in ungauged catchments. Given actual discharge measurements or discharge estimations, DRRAiNN can be used to analyze the hydrological dynamics of surface and subsurface runoff as well as baseflow esimations and has the potential to uncover unexpected and unknown runoff dynamics that would not be detectable otherwise.

How to cite: Scholz, F., Zarfl, C., Scholten, T., and Butz, M. V.: Inference of catchment areas from modeled discharge dynamics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19057, https://doi.org/10.5194/egusphere-egu25-19057, 2025.

Continental and global water models have long been trapped in slow growth and inadequate predictive power, as they are not able to effectively assimilate information from big data. While Artificial Intelligence (AI) models greatly improve performance, purely data-driven approaches do not provide strong enough interpretability and generalization. One promising avenue is “differentiable” modeling that seamlessly connects neural networks with physical modules and trains them together to deliver real-world benefits in operational systems. Differentiable modeling (DM) can efficiently learn from big data to reach state-of-the-art accuracy while preserving interpretability and physical constraints, promising superior generalization ability, predictions of untrained intermediate variables, and the potential for knowledge discovery. Here we demonstrate the practical relevance of a high-resolution, multiscale water model for operational continental-scale and global-scale water resources assessment. (https://bit.ly/3NnqDNB). Not only does it achieve significant improvements in streamflow simulation compared to the established national- and global water models, but it also produces much more reliable depictions of interannual changes in large river streamflow, freshwater inputs to estuaries, and groundwater recharge. 

How to cite: Song, Y., Shen, C., Ji, H., and Rahmani, F.: High-Resolution Differentiable Models for Operational National and Global Water Modeling and Assessment, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20878, https://doi.org/10.5194/egusphere-egu25-20878, 2025.

Long Short-Term Memory (LSTM) networks have become popular for streamflow prediction in hydrological systems due to their ability to model sequential data. However, their reliance on lumped catchment representation and climate summaries often limits their capacity to capture spatial heterogeneity in rainfall patterns and evolving rainfall trends, both of which are critical for hydrological consistency. This study explores the limitations of LSTM-based streamflow predictions by employing a distributed conceptual hydrological model, SIMHYD, coupled with Muskingum-Cunge routing, to generate synthetic datasets representing diverse hydroclimatic conditions. These datasets are designed to replicate rainfall-runoff dynamics across selected catchments from all 18 ecoregions in CAMELS-US and key Indian river basins, providing a robust testbed for evaluating model performance under controlled conditions. The pre-trained LSTM model is tested against synthetic discharge data, enabling direct comparisons to assess its ability to simulate realistic hydrological responses. Performance is evaluated using multiple metrics, including Nash-Sutcliffe Efficiency (NSE), Kling-Gupta Efficiency (KGE), Percent Bias (PBIAS), and mean peak timing errors, to identify systematic deviations. Results reveal that LSTM models struggle with spatially variable and temporally shifting rainfall patterns, leading to inaccuracies in peak flow timing, magnitude, and overall discharge volumes. These issues highlight vulnerabilities in current LSTM-based flood forecasting systems, particularly in their ability to generalize across diverse climatic conditions and regions. This study also characterizes specific failure pathways, such as underestimation of extreme events and poor temporal coherence in hydrographs, which are critical for operational forecasting. By diagnosing these limitations, the study provides a framework for integrating process-based hydrological knowledge with data-driven techniques to improve model robustness. The findings underscore the importance of using synthetic datasets and diverse diagnostic tools to evaluate and enhance the reliability of LSTM-based models, paving the way for hybrid approaches capable of addressing the complexities of real-world hydrological systems.

How to cite: Dubey, S., Bhasme, P., and Bhatia, U.: Characterizing possible failure modes: Insights from LSTM-Based Streamflow Predictions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-935, https://doi.org/10.5194/egusphere-egu25-935, 2025.

Measuring river surface velocity enables river discharge estimation, a fundamental task for hydrologists, environmental scientists, and water resource managers. While traditional image-based velocimetry methods are often effective, they struggle to produce complete velocity fields under complex environmental conditions. Poor lighting, reflective glare, lack of visible surface features, or excessive turbulence can all result in regions where feature tracking fails, leading to gaps in the resolved velocity field. Addressing these gaps through the reconstruction of missing velocity measurements is an important research challenge. Recently, researchers have employed deep learning to address various hydrology problems, demonstrating promising improvements. In this work, we propose a neural operator-based model to address the challenge of missing velocities in river surface velocimetry. Specifically, our model is based on the Fourier neural operator with a graph-enhanced lifting layer. It is trained on the river surface velocimetry reconstruction task using a self-supervised paradigm. Once trained, it can be used to infer missing velocities in unseen samples. Experiments conducted on a dataset collected from a river in the Netherlands demonstrate our approach’s ability to accurately infill missing surface velocities, even when faced with large amounts of missing data. We attribute this robustness to the neural operator’s ability to learn continuous functions, which enhances our model’s capacity for high-level feature representation and extraction. Our findings suggest that the reconstructed velocity fields produced by our model can act as reliable ground truth data for deep learning-based methods. In the future, we aim to improve our model’s performance and generalization by incorporating additional data collected from a wider range of rivers and under varying environmental conditions.

How to cite: Chen, X., Winsemius, H., and Taormina, R.: Graph-enhanced Neural Operator for Missing Velocities Infilling in River Surface Velocimetry, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1644, https://doi.org/10.5194/egusphere-egu25-1644, 2025.

Historical records observed at hydrological stations are scarce in many regions, bringing significant challenges to the hydrological predictions for these regions. Transfer learning (TL), increasingly applied in hydrology, leverages knowledge from data-rich catchments (sources) to enhance predictions in data-scarce catchments (targets), providing new insight into data-scarce region predictions. Most existing TL approaches pre-train models using large meteoro-hydrological datasets to improve overall generalizability to target catchments. However, the predictive performance for specific catchments would be constrained due to irrelevant source data inputs and the lack of effective source fusion strategies. To address these challenges, this study proposes the Dual-Source Adaptive Fusion TL Network (DSAF-Net), which utilizes a pre-trained dual-branch feature extraction module (DBFE) to extract knowledge from two carefully selected source catchments, minimizing noise and redundancy associated with larger datasets. A cross-attention fusion module is then incorporated to dynamically identify key knowledge of the target catchment and adaptively fuse complementary information. This fusion module is embedded after each layer in the DBFE to enhance multi-level feature integration. Results demonstrate that DSAF-Net achieves superior prediction accuracy to single-source TL and large dataset TL strategies. These findings highlight the potential of DSAF-Net to advance hydrological forecasting and support water resource management in data-scarce regions.

How to cite: Gao, Y. and Liang, D.: Dual-Source Adaptive-Fusion Transfer Learning for Hydrological Forecasting in Data-scarce Catchments, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2616, https://doi.org/10.5194/egusphere-egu25-2616, 2025.

The increasing frequency of extreme hydrological events, such as floods and droughts, poses significant challenges for operators of drinking water reservoirs in maintaining a balance between water supply and demand. While the security of supply typically requires high water levels to meet consumer demands throughout the year, ensuring flood protection, on the contrary, necessitates that reservoir storage is kept partially free to accommodate high inflows. Accurate inflow forecasting is essential for making risk-based operational decisions, including the timely release of water from drinking water reservoirs to mitigate flood risks. While deep learning approaches, particularly Long Short-Term Memory (LSTM) networks, have become prevalent in rainfall-runoff modeling, most existing studies focus on small, homogeneous datasets limited to single hydrological basins. This study leverages the newly published CAMELS-DE dataset to develop a regionally trained and finetuned LSTM model encompassing 1,582 catchments across Germany. We apply this regional model to five small catchments upstream of drinking water reservoirs and compare its performance against basin-specific LSTM models. Our findings demonstrate that the regionally trained LSTM model significantly improves the accuracy of inflow estimates, especially when finetuned to our target catchments. This is highlighting its potential for enhancing reservoir management strategies in the face of climate change.

How to cite: Johnen, G., Niemann, A., Nistahl, P., Dolich, A., and Hutwalker, A.: Forecasting Reservoir Inflows Using Regionally Trained and Finetuned LSTM Models: A Case Study with CAMELS-DE, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2761, https://doi.org/10.5194/egusphere-egu25-2761, 2025.

Deep learning models such as the Long Short Term Memory Network (LSTM) are capable of representing rainfall-runoff relationships and outperform classical hydrological models in gauged and ungauged settings (Kratzert et al., 2018). Previous studies have shown that combining multiple precipitation data in a single LSTM significantly improves the accuracy of simulated runoff, as the neural network learns to combine temporal and spatial patterns of inputs (Kratzert et al., 2021). However, every operational runoff forecasting setting requires meteorological forecasts. Nearing et al. (2024) have developed a global runoff forecast model based on an LSTM, with an ECMWF forecasting product as additional input over the forecast horizon. Compared with observed or reanalysis meteorological input data, forecasting products generally have a lower accuracy, with different reliabilities between various forecasting products. This is where the synergies of several meteorological forecasts combined with historical observational and reanalysis data can be used in a single deep learning model.

This study investigates how well LSTMs can predict runoff when trained on (1) multiple archived meteorological forecasts and (2) a combination of multiple archived meteorological forecasts and reanalysis data. All meteorological input data are aggregated to the catchments of the LamaH-CE dataset (Klingler, Schulz and Herrnegger, 2021). Runoff predictions are evaluated for a 24 hours forecasting horizon.  Preliminary analyses indicate that the coupling of reanalysis data and forecasting products from different sources improves the accuracy of operational runoff forecasting, suggesting that the model is able to learn and adjust real-time biases in forecasting data.

Klingler, C., Schulz, K. and Herrnegger, M. (2021) ‘LamaH-CE: LArge-SaMple DAta for Hydrology and Environmental Sciences for Central Europe’, Earth System Science Data, 13(9), pp. 4529–4565. DOI: 10.5194/essd-13-4529-2021.

Kratzert, F., Klotz, D., Brenner, C., Schulz, K. and Herrnegger, M. (2018) ‘Rainfall–runoff modelling using Long Short-Term Memory (LSTM) networks’, Hydrology and Earth System Sciences, 22(11), pp. 6005–6022. DOI: 10.5194/hess-22-6005-2018.

Kratzert, F., Klotz, D., Hochreiter, S. and Nearing, G.S. (2021) ‘A note on leveraging synergy in multiple meteorological data sets with deep learning for rainfall–runoff modeling’, Hydrology and Earth System Sciences, 25(5), pp. 2685–2703. DOI: 10.5194/hess-25-2685-2021.

Nearing, G., Cohen, D., Dube, V., Gauch, M., Gilon, O., Harrigan, S., Hassidim, A., Klotz, D., Kratzert, F., Metzger, A., Nevo, S., Pappenberger, F., Prudhomme, C., Shalev, G., Shenzis, S., Tekalign, T.Y., Weitzner, D. and Matias, Y. (2024) ‘Global prediction of extreme floods in ungauged watersheds’, Nature, 627(8004), pp. 559–563. DOI: 10.1038/s41586-024-07145-1.

How to cite: Konold, O., Feigl, M., Klingler, C., and Schulz, K.: From multiple meteorological forecasts to river runoff: Learning and adjusting real-time biases to enhance predictions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4247, https://doi.org/10.5194/egusphere-egu25-4247, 2025.

Runoff forecasting is a long-standing challenge in hydrology, particularly in unmeasured catchments and rapid flash flood prediction. For unmeasured catchment forecasting, we introduce the encoder-decoder-based dual-layer long short-term memory (ED-DLSTM) model^[1]. This model fuses static spatial granularity attributes with temporal dynamic variables to achieve streamflow forecasting at a global scale. ED-DLSTM reaches an average Nash efficiency coefficient (NSE) of 0.75 across more than 2000 catchments from historical datasets in the United States, Canada, Central Europe, and the United Kingdom. Additionally, ED-DLSTM is applied to 150 fully ungauged catchments in Chile, achieving a high NSE of 0.65. The interpretability of the transfer capacities of ED-DLSTM is effectively tracked through the cell state induced by adding a spatial attribute encoding module, which can spontaneously form hydrological regionalization effects after performing spatial coding for different catchments.

Moreover, rapid flood prediction with daily resolution is challenged to capture changes in runoff over short periods. To address this, we also propose a benchmark evaluation for runoff and flood forecasting based on deep learning (RF-Bench) at an hourly scale. We introduce the Mamba model to hydrology for the first time. The benchmark also includes Dlinear, LSTM, Transformer, and its improved versions (Informer, Autoformer, Patch Transformer). Results indicate that the Patch Transformer exhibits optimal predictive capability across multiple lead times, while the traditional LSTM model demonstrates stable performance, and the Mamba model strikes a good balance between performance and stability. We reveal the attention patterns of Transformer models in hydrological modeling, finding that attention is time-sensitive and that the attention scores for dynamic variables are higher than those for static attributes.

Our work ^[2,3] provides the hydrological community with an open-source, scalable platform, contributing to the advancement of deep learning in the field of hydrology.

[1] Zhang, B., Ouyang, C., Cui, P., Xu, Q., Wang, D., Zhang, F., Li, Z., Fan, L., Lovati, M., Liu, Y., Zhang, Q., 2024. Deep learning for cross-region streamflow and flood forecasting at a global scale. The Innovation 5, 100617. https://doi.org/10.1016/j.xinn.2024.100617

[2] Zhang, B., Ouyang, C., Wang, D., Wang, F., Xu, Q., 2023. A PANN-Based Grid Downscaling Technology and Its Application in Landslide and Flood Modeling. Remote Sensing 15, 5075. https://doi.org/10.3390/rs15205075

[3] Xu, Q., Shi, Y., Bamber, J.L., Ouyang, C., Zhu, X.X., 2024. Large-scale flood modeling and forecasting with FloodCast. Water Research 264, 122162. https://doi.org/10.1016/j.watres.2024.122162

How to cite: zhang, B., xu, Q., and ouyang, C.: Runoff Forecasting in Unmeasured Catchments and Rapid Flash Flood Prediction Based on Deep Learning., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4387, https://doi.org/10.5194/egusphere-egu25-4387, 2025.

Accurate water level prediction is essential for flood risk management, water resources management, inland water transportation and climate resilience. Traditional statistical methods, such as autoregressive models, and physically-based hydrological simulations, have been widely used in water level forecasting. However, these approaches often struggle to capture complex, dynamic, and nonlinear interactions in a hydrological system, particularly those affected by climate change. In recent years, machine learning models have emerged as a promising alternative, offering improved predictive accuracy and adaptability across varying environmental conditions. A special type of such models is the Graph Neural Network (GNN), which focuses especially on the reproduction of spatial dependencies, and hence it can be employed to capture the spatial dynamic of the hydrologic/hydraulic system, by treating hydrological networks as graph structures (e.g. nodes as gauges). Going one step further, GNN models can be combined with sequence-based machine learning techniques, such as the Long short-term memory (LSTM) neural network, to capture simultaneously the spatial and temporal dynamics of the system. In this work, we develop and assess a series of advanced hybrid-graph structured machine learning models (such as GNN-LSTM) to make hydrometric predictions across a long river channel. The developed models will be assessed on the basis of alternative performance metrics and against a series of traditional benchmark statistical and machine learning models such as ARIMA and LSTM respectively. As a test case, we exploit data from 19 water level gauges in the Red River of the North, which spans 885 km, serving the natural boundary between North Dakota and Minnesota and has experienced several severe historical flood events.

How to cite: Koutsos, G., Kossieris, P., Thomopoulou, V., and Makropoulos, C.: Leveraging Graph Neural Networks for water level prediction, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4500, https://doi.org/10.5194/egusphere-egu25-4500, 2025.

River level predicting underpins the management of water resource projects, steers navigational activities in rivers, and protects the lives and properties of riverside communities, etc. Traditionally, hydrological-hydraulic coupled models have been at the forefront of simulating and predicting river levels, achieving notable success. Despite their utility, these models encounter limitations due to the exhaustive demand for various data types—often difficult to obtain—and the ambiguity in determining downstream boundary conditions for the hydraulic model. Responding to these limitations, this study utilizes Long Short-Term Memory (LSTM) model, a deep learning technique, to predict river levels using upstream discharges. Three approaches were used to further enhance the accuracy and reliability of our model. Firstly, we incorporated historical water level data at or downsteam of the predicted station as input, secondly, we classified the datasets based on physical principles, and thirdly, we employed data augmentation techniques. These methods were evaluated within the Jingjiang-Dongting river-lake system in China. It achieves high prediction accuracy of water level and can mitigate the impact of input inaccuracies. The incorporation of water level data as input and the Classification-Enhanced LSTM model that segregates the input data according to rising and recession trends of water level,significantly improve prediction accuracy under extreme water level conditions compared with other deep learning approaches. The proposed model uses easily accessible data to predict water levels, offering enhanced robustness and new strategies for improving prediction accuracy under extreme conditions. It is applicable for predicting water levels at any hydrological station along a river and can enhance the prediction accuracy of hydraulic models by proving a robust downstream boundary condition.

How to cite: Luo, J.: Classification-Enhanced LSTM Model for Predicting River Water Levels, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4683, https://doi.org/10.5194/egusphere-egu25-4683, 2025.

Accurate soil moisture prediction is increasingly important due to its critical role in water resource management and agricultural sustainability under global climate change. While machine learning models have achieved high accuracy in soil moisture prediction, their ability to generalize to different environmental and meteorological conditions remains a significant challenge. Existing models often perform poorly when applied to conditions that differ from their training data, highlighting the need for approaches that improve generalization while effectively capturing underlying soil moisture dynamics.

In this study, we propose a novel soil moisture prediction model that combines self-supervised learning with a Transformer architecture. The performance of the model was compared with the widely used Long Short-Term Memory (LSTM)-based approach to evaluate its ability to generalize. The proposed model outperformed the baseline in tasks such as capturing extreme soil dryness, adapting to unobserved meteorological humidity conditions, and forecasting soil moisture dynamics at untrained depths. Further analysis revealed that the model’s success stems from its capability to learn comprehensive representations of underlying soil moisture processes. These results highlight the potential of advanced deep learning methods to improve prediction and our process understanding of soil hydrology in a changing climate.

How to cite: Wang, L., Shi, L., and Jiang, S.: Improving generalization of soil moisture prediction using self-supervised learning, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6294, https://doi.org/10.5194/egusphere-egu25-6294, 2025.

This study evaluates a new approach to improving streamflow forecasting with deep learning: it focuses on the novel application of a dynamic multimodal feature fusion mechanism that adapts fusion operations based on the data's characteristics. Two baseline Long Short-Term Memory (LSTM) architectures are used, applying two dynamic fusion methods: dynamic operation-level fusion and attention-based fusion, to combine heterogeneous and multisource hydrometeorological data. The models are used for univariate (single flow gauge) and multivariate (multi-gauge) streamflow forecasting approaches. Applying these four approaches to the Severn Basin in the UK, known for long medium- to high-flow periods and shorter low-flow intervals, shows that the dynamic operation-level fusion consistently improved over the attention-based fusion in key performance metrics. In the multivariate case, Nash-Sutcliffe Efficiency (NSE) improved by 1.43%, Mean Absolute Error (MAE) decreased by 1.73%, Mean Absolute Scaled Error (MASE) dropped by 1.82%, and high-peak MAE decreased by 3.36%. For the univariate case, NSE improved by 1.44%, MAE decreased by 4.02%, MASE dropped by 3.89%, and high-peak MAE improved by 2.8%. In addition, multivariate models were considerably faster than univariate models, with training and inference times reduced by 74.57% and 73.81%, respectively. The multivariate models showed a 2.75% increase in NSE and a 72.04% decrease in MASE, indicating they captured better the hydrologic variability than the univariate models. Conversely, univariate models had a 20.59% lower MAE, a 21.17% lower high-peak MAE, and greater stability as indicated by tighter interquartile ranges, suggesting better error minimisation and more reliable predictions. Notably, in two river stations all models underperformed due to rapid flow variability and flashy hydrological responses in smaller catchment areas, suggesting in the future the use of higher-resolution climatic data. Overall, the study shows the potential of new dynamic multimodal fusion techniques, navigating the operational trade-offs between speed, stability, and accuracy across multi and uni-variate training strategies in streamflow forecasting. Nonetheless, the need for an optimal operational balance remains, suggesting further refinement of fusion techniques and focusing on minimising uncertainty.

How to cite: Theodosiadou, K., Rodding Kjeldsen, T., and Barnes, A.: Improving high-flow forecasting using dynamic multimodal feature fusion, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7020, https://doi.org/10.5194/egusphere-egu25-7020, 2025.

Accurate nutrient predictions are crucial for river water quality management. While deep learning (DL) has shown promise in various Earth science applications, challenges such as data scarcity and limited interpretability hinder its use in river nutrient predictions. Building on insights into the physical dynamics of nutrients, this research investigates how incorporating extreme weather indices as additional input data, which are often overlooked in current DL-based nutrient prediction, could affect model performance. Additionally, we aim to improve model interpretability by developing hybrid DL-physical structures and identify the optimal structure for predicting nutrient indicators. 
 
The study proposes an assessment workflow and demonstrates its application by predicting dissolved inorganic nitrogen (DIN) and soluble reactive phosphorus (SRP) concentrations at the outlet of the Salmons Brook catchment, UK, where nutrient observations are scarce. The workflow includes two key decisions: selecting the input dataset and defining the DL-physical hybrid structure, each with two options. Comparing multiple predictions generated from all decision combinations enables the evaluation of the impacts of extreme weather events and different hybrid structures. 
 
The simulations demonstrate that incorporating extreme weather indices as additional inputs enhanced performance for both nutrient indicators, particularly in capturing extreme values. Overall, the choice of input dataset had a greater impact on the simulations than the hybrid structure, highlighting the importance of careful input selection and preprocessing in DL model development. Integrating results from a physical model into a DL model can improve simulation interpretability by introducing nutrient-related physical processes. In addition to the hybrid structure, incorporating insights into the physical behaviour of nutrients further enhances the interpretability of DL-based predictions, which is crucial for gaining the trust of domain experts, especially when validating results. 

How to cite: Tang, J., Liu, L., Chun, K., and Mijic, A.: Enhancing River Nutrient Predictions with Extreme Weather Indices and DL-Physical Hybrid Structures for Improved Interpretability, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7094, https://doi.org/10.5194/egusphere-egu25-7094, 2025.

Deep learning methods in hydrology have traditionally focused on deterministic models, limiting their ability to quantify prediction uncertainty. Recent advances in generative modeling have opened new possibilities for probabilistic modelling in various applied fields, including hydrological forecasting (Jahangir & Quilty, 2024). These models learn to represent underlying probability distributions using neural networks, enabling uncertainty quantification through sampling in a very flexible framework.

In this submission we introduce vLSTM, a variational extension of the traditional long short-term memory (LSTM) architecture that quantifies predictive uncertainty by adding noise sampled from a learned multivariate Gaussian distribution to perturb the model’s hidden state. The vLSTM preserves the traditional LSTM’s state-space dynamics while introducing a probabilistic component that enables uncertainty quantification through sampling. Unlike mixed-density networks (MDNs) which directly model the distribution of the target variable, vLSTM’s uncertainty is obtained by perturbations to the hidden state, providing a novel approach to probabilistic prediction. In rainfall-runoff modeling, vLSTM offers a different mechanism for uncertainty quantification to the well established MDN models (Klotz et al., 2022). This approach enriches the existing toolkit of uncertainty methods in deep learning while maintaining the simplicity of sampling for probabilistic predictions.

To rigorously evaluate probabilistic predictions across different model architectures, we develop new information-theoretic metrics that capture key aspects of how uncertainty is handled by a particular model. These include the average prediction entropy H(X), which quantifies model confidence, and average relative entropy D_KL(pq), which measures the average alignment between the predicted distribution of a model and a target, among others. The proposed metrics take advantage of non-parametric estimators for Information Theory which have been implemented in the easy to use UNITE toolbox (https://github.com/manuel-alvarez-chaves/unite_toolbox). By expressing these metrics in compatible units of bits (or nats), we enable direct comparisons between different uncertainty measures. We apply these metrics to our newly introduced vLSTM and the existing MDN models to show strengths and weaknesses of each approach. This information-theoretic framework provides a unified language for analyzing and understanding predictive uncertainty in probabilistic models.

References

• Jahangir, M. S., & Quilty, J. (2024). Generative deep learning for probabilistic streamflow forecasting: Conditional variational auto-encoder. Journal of Hydrology, 629, 130498. https://doi.org/10.1016/j.jhydrol.2023.130498
• Klotz, D., Kratzert, F., Gauch, M., Keefe Sampson, A., Brandstetter, J., Klambauer, G., Hochreiter, S., & Nearing, G. (2022). Uncertainty estimation with deep learning for rainfall–runoff modeling. Hydrology and Earth System Sciences, 26(6), 1673–1693. https://doi.org/10.5194/hess-26-1673-2022

How to cite: Alvarez Chaves, M., Gupta, H., Ehret, U., and Guthke, A.: Evaluating uncertainty in probabilistic deep learning models using Information Theory, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7205, https://doi.org/10.5194/egusphere-egu25-7205, 2025.

Deep learning (DL)-based hydrological models, particularly those using Long Short-Term Memory (LSTM) networks, typically require large datasets for effective training. In the context of large-scale rainfall-runoff modeling, dataset size can refer to either the number of watersheds or the length of the training period. While it is well established that training a regional model across more watersheds improves performance (Kratzert et al., 2024), the benefits of extending the training period are less clear.

Empirical evidence from studies such as Boulmaiz et al. (2020) and Gauch et al. (2021) suggests that longer training periods enhance LSTM performance in rainfall-runoff modeling. This improvement is attributed to the need for extensive datasets to ensure proper model convergence and the ability to capture a wide range of hydrological conditions and events. However, these studies neglected the influence of data recency (or data recentness), which is critical for operational applications that forecast current and future hydrological conditions. In the context of climate change and anthropogenic interventions, the assumption of stationarity (i.e., that historical patterns reliably represent future conditions) may no longer hold for hydrological systems (Shen et al., 2022). Consequently, the selection of training periods should account for potential non-stationarity, as more recent data may better reflect current rainfall-runoff dynamics. Intriguingly, Shen et al. (2022) found that calibrating hydrologic models to the latest data is a superior approach compared to using old data, and completely discarding the oldest data can even improve the performance in streamflow prediction.

This study aims to address two research questions: (1) As the number of watersheds increases, is it still necessary to train LSTM models on decades of historical observations? (2) Can LSTM models achieve comparable performance using shorter training periods focused on more recent data? Specifically, we examine whether models trained on recent data outperform those trained on older data and explore how different temporal partitions of historical records affect predictive skill.

This study leverages a comprehensive dataset comprising streamflow records from over 1,300 watersheds across North America, representing diverse climatic and hydrological regimes, with streamflow data spanning 1950 to 2023. Training periods are designed to isolate the effects of temporal data recency while keeping period lengths consistent. This approach enables a systematic comparison of model performance using exclusively older (e.g., pre-1980) versus exclusively recent data (e.g., post-1980). This research provides evidence-based recommendations for selecting training data while balancing computational costs, data availability, and prediction accuracy.

References

Boulmaiz, T., Guermoui, M., and Boutaghane, H.: Impact of training data size on the LSTM performances for rainfall–runoff modeling, Model Earth Syst Environ, 6, 2153–2164, https://doi.org/10.1007/S40808-020-00830-W/FIGURES/9, 2020.

Gauch, M., Mai, J., and Lin, J.: The proper care and feeding of CAMELS: How limited training data affects streamflow prediction, Environmental Modelling and Software, 135, https://doi.org/10.1016/j.envsoft.2020.104926, 2021.

Kratzert, F., Gauch, M., Klotz, D., and Nearing, G.: HESS Opinions: Never train an LSTM on a single basin, Hydrology and Earth System Science, https://doi.org/10.5194/hess-2023-275, 2024.

Shen, H., Tolson, B. A., and Mai, J.: Time to Update the Split-Sample Approach in Hydrological Model Calibration, Water Resour Res, 58, e2021WR031523, https://doi.org/10.1029/2021WR031523, 2022.

How to cite: Yu, Q. and Tolson, B.: Empirical Evidence of the Importance of Data Recency in LSTM-Based Rainfall-Runoff Modeling , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7468, https://doi.org/10.5194/egusphere-egu25-7468, 2025.

The increasing complexity of water pollution and its impact on aquatic ecosystems necessitates the accurate prediction of water pollutant loads for effective river management. Total Organic Carbon (TOC), a key indicator of organic pollution levels, is central to assessing ecosystem health and informing water treatment strategies. However, conventional process-based modeling methods, while capable of providing precise water quality predictions, require extensive input data and significant computational resources, limiting their practical application. Consequently, alternative modeling approaches, particularly those leveraging artificial intelligence, have been explored. Recent advancements in deep learning have improved predictive modeling in environmental sciences. These approaches have showed effectiveness in hydrological applications, such as streamflow forecasting, by capturing complex nonlinear relationships within environmental systems. Despite these advancements, a notable limitation of these models is their difficulty in maintaining physical consistency, specifically in adhering to the principle of mass balance—a fundamental concept in both hydrology and water quality modeling. In this study, we evaluate a Mass-Conserving Long Short-Term Memory network integrated with QUAL2E kinetics (MC-LSTM-QUAL2E) to predict TOC loads in river systems. By incorporating representations of decay and reaeration processes within a mass-conserving neural network framework, this model combines data-driven prediction capabilities with the requirements of physical consistency. A key component of this framework is the trash cell, designed to simulate TOC transformations based on QUAL2E dynamics. Within the trash cell, TOC decay and reaeration are modeled using parameters k_decay​ and k_reaeration​, which are determined by environmental variables such as temperature, pH, dissolved oxygen, total nitrogen, and total phosphorus. The QUAL2E module updates the trash state at each timestep to account for TOC losses due to decay and gains from reaeration, ensuring mass conservation.  The MC-LSTM-QUAL2E model was compared to a conventional LSTM model using environmental variables, including temperature, pH, dissolved oxygen, and nutrient levels, as inputs. The analysis used data from 2012 to 2020, with the period from 2012 to 2017 designated for training and 2018 to 2020 for tests. Model performance was assessed using metrics such as Nash-Sutcliffe Efficiency (NSE), Kling-Gupta Efficiency (KGE), Root Mean Square Error-observations standard deviation Ratio (RSR), and Percent Bias (PBIAS). By maintaining mass balance and incorporating QUAL2E dynamics, the model provides reliable predictions of TOC loads in river systems and offers insights into associated biochemical and hydrological processes.

How to cite: Jeong, H., Lee, B., Lee, Y., and Lee, S.: Predicting total organic carbon loads in river using a mass-conserving LSTM integrated with QUAL2E kinetics, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7531, https://doi.org/10.5194/egusphere-egu25-7531, 2025.

Integrating deep learning techniques into hydrology has opened a new way to improve urban flood modeling, with various solutions being developed to address urban flood problems driven by climate change and urbanization. However, predicting urban inundation in near real-time for large urban areas remains challenging due to computational demands and limited data availability. This work proposes a deep learning-based super-resolution framework that enhances the spatial resolution of process-based urban flood modeling outputs using convolutional neural networks (CNNs) while improving computational efficiency. This study investigates the interaction between deep learning model architecture and the underlying physical processes to improve prediction accuracy and robustness in urban pluvial flood mapping. The methodology will be applied to various urban flood scenarios, including extreme rainfall events and hurricane-induced flooding, and its performance will be evaluated through quantitative indicators and sensitivity analyses. The applicability and scalability of this model will also be discussed. In particular, strategies to enhance model reliability and integrate additional hydrological information under extreme conditions will be explored. The study will further address uncertainty estimation in deep learning-based super-resolution models and scalability challenges associated with super-resolution approaches for large-scale flood simulations. The findings aim to demonstrate the potential of deep learning as an innovative tool in hydrological modeling and to enable more effective flood risk management strategies.

How to cite: Choi, H., Woo, H., Kim, M., Ryu, H., Lee, J., Lee, S., and Noh, S. J.: Developing a Deep Learning-Based Super-Resolution Urban Flood Model: Towards Scalable and Reliable Hydrological Predictions, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8045, https://doi.org/10.5194/egusphere-egu25-8045, 2025.

Sea level rise and increasing coastal flood risks demand the development of accurate and efficient coastal risk models capable of generating large ensembles of projections to support robust adaptation strategies. The latest IPCC report emphasizes the importance of projecting storm surge changes and their associated uncertainties, alongside mean sea level rise. However, the high computational cost of storm surge simulations continues to limit the feasibility of generating large ensembles.
Artificial Intelligence (AI) is emerging as a promising alternative to simulate storm surge scenarios with significantly reduced computational costs. Despite recent advancements, key challenges remain in accurately representing extreme events and ensuring robust model extrapolation under changing climate conditions. While AI-based surrogate models have been proposed in the literature, gaps persist in understanding their performance limits for extreme events in future scenarios, hindering their application in climate adaptation planning.
To address these challenges, we developed a deep learning (DL) surrogate model of the physics-based Global Tide and Surge Model (GTSM). The DL model is trained using reanalysis data (ERA5) and historical scenarios from the Coupled Model Intercomparison Project Phase 6 (CMIP6) High Resolution Model Intercomparison Project (HighResMIP). Our analysis focuses on the DL model's performance in simulating extreme storm surge events, validated against GTSM outputs for both historical reanalysis and future projections, with a case study along the New York coastline.
To enhance the surrogate model’s performance for extreme events, we explore various loss functions, including a customized quantile loss function, and test alternative DL architectures with different input configurations. Results demonstrate that the quantile loss improves the model's accuracy for extremes compared to standard loss functions such as mean square error. Additionally, fine-tuning DL models with specific Global Climate Model forcing fields improves the alignment of AI-predicted storm surge trajectories with GTSM outputs, even under diverse spatiotemporal resolutions and model setups.
These findings highlight the critical importance of selecting appropriate loss functions and training datasets to ensure robust performance over extreme events and projected future scenarios. Our globally applicable framework, relying solely on open-source data, offers a promising pathway to scalable and efficient storm surge projections, with implications for robust coastal adaptation planning.

How to cite: Longo, E., Ficchì, A., Muis, S., Verlaan, M., and Castelletti, A.: Projecting storm surge extremes with a deep learning surrogate model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-9088, https://doi.org/10.5194/egusphere-egu25-9088, 2025.

Machine learning is increasingly important for rainfall–runoff modelling. In particular, the community started to widely adopt the Long Short-Term Memory (LSTM) network. One of the most important established best practices  in this context is to train the LSTMs on a large number of diverse basins  (Kratzert et al., 2019; 2024). Intuitively, the reason for adopting this practice is that training deep learning models on small and homogeneous data sets (e.g., data from only a single hydrological basin) leads to poor generalization behavior — especially for high-flows. 

To examine this behavior, Kratzert et al. (2024) use a theoretical maximum prediction limit for LSTMs. This theoretical limit is computed as the L1 norm (i.e., the sum of the absolute values of each vector component) of the learned weight vector that relates the hidden states to the estimated streamflow. Hence, for random vectors we could simply obtain larger theoretical limits by increasing the size of the network (i.e., the  number of parameters). However, since LSTMs are trained using gradient descent, this relationship is more intricate. 

This contribution explores the relationship between the theoretical limit and the network size. In particular, we will look at how increasing the network size in untrained models increases the prediction limit and contrast it to the scaling behavior of trained models.

How to cite: Klotz, D., Baste, S., Loritz, R., Gauch, M., and Kratzert, F.: The relationship between theoretical maximum prediction limits of the LSTM and network size, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-10650, https://doi.org/10.5194/egusphere-egu25-10650, 2025.

Accurate streamflow prediction is critical for flood forecasting and water resource management, particularly in data-scarce regions like Central Asia (CA), where traditional hydrological models struggle due to insufficient discharge data. Deep learning models, such as Long Short-Term Memory (LSTM), have demonstrated the potential for global hydrologic regionalization by leveraging both climate data and catchment characteristics. We used a transfer learning (TL) approach to improve streamflow predictions by first pretraining LSTM models on catchments from data-rich regions like Switzerland, Scotland, and British Columbia (source regions). These deep learning models were then fine-tuned on the data scarce target region (CA basins). This approach leverages the knowledge gained from the source regions to adapt the model to the target region, enhancing prediction accuracy despite the data scarcity in CA. Incorporating lagged streamflow alongside ERA-5 climate data boosted prediction accuracy, particularly in snowmelt and glaciers influenced basins like Switzerland (median NSE=0.707 to 0.837), British Columbia (median NSE= 0.775 to 0.923) and CA (median NSE=0.693 to 0.798). K-Means algorithm was applied to categorize catchments from four global locations into five clusters (labeled 0–4) based on their specific attributes. The predictive performance of fine-tuned LSTM model has significantly enhanced when leveraging a pre-trained model with cluster 2, as demonstrated by higher median metrics (NSE=0.958, KGE=0.905, RMSE=10.723, MSE=115.055) compared to both the locally trained model (NSE=0.851, KGE=0.792, RMSE=20.377, MSE=415.579) and individual basin-based training approaches (NSE=0.69, KGE=0.692, RMSE=25.563, MSE=676.110). These results highlight the effectiveness of pretraining the LSTM model on diverse clusters (0, 1, 2, and 4) before fine-tuning on the target region (cluster 3). Moreover, pretraining the LSTM model with clusters 0 and 4 resulted in enhanced performance by increasing the number of basins, whereas the impact was minimal or even declined when using clusters 1 and 2, as well as when all basins from the four clusters were included. These findings demonstrate the feasibility of transfer learning in addressing data scarcity challenges and underscore the importance of diverse and high-quality training data in developing robust, regionalized hydrological models. This approach bridges the gap between data-rich and data-scarce regions, offering a pathway to improved flood prediction and water resource management.

How to cite: Hassan, J., Rowan, J., and Mukherjee, N.: Exploring the Transferability of Knowledge In Deep Learning-Based Streamflow Models Across Global Catchments, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11240, https://doi.org/10.5194/egusphere-egu25-11240, 2025.

Effective discharge forecasting is critical in operational hydrology. This study explores novel methods to improve forecast accuracy by combining data assimilation techniques and hydrograph decomposition. Traditional rainfall-runoff modeling, including AI-based approaches, typically simulates the entire discharge signal using a single model. However, runoff is generated by multiple processes with contrasting kinetics, which a single-model approach may fail to capture adequately. This study proposes using hydrograph decomposition to separate baseflow and quickflow components, training specific forecasting models for each component individually, and then merging their outputs to reconstruct the total discharge signal. This approach is expected to enhance forecast accuracy for both floods and droughts, identifying long-term dependencies governing baseflow to improve seasonal low-flow forecasts. Experiments will be conducted using a subset of the CAMELS dataset.

How to cite: Saint Fleur, B. E., Gaume, E., and Akil, N.: Improving AI-based discharge forecasting through hydrograph decomposition and data assimilation, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11682, https://doi.org/10.5194/egusphere-egu25-11682, 2025.

Accurately forecasting streamflow is essential for effectively managing water resources. High-quality operational forecasts allow us to prepare for extreme weather events, optimize hydropower generation, and minimize the impact of human development on the natural environment. However, streamflow forecasts are inherently limited by the quality and availability of upstream weather sources. The weather forecasts that drive hydrological modeling vary in their temporal resolutions and are prone to outages, such as the ECMWF data outage in November of 2023. 

Here, we present HydroForecast Short Term 3 (ST-3), a state-of-the-art probabilistic deep learning model for medium-term (10-day) streamflow forecasts. ST-3 combines long short-term memory architecture with Boolean tensors representing data availability and dense embeddings for processing of the information in these tensors. This architecture allows for a training routine that implements data augmentation to synthesize varying amounts of availability of weather inputs. The result is a model that 1) makes accurate forecasts even in the case of an upstream data outage, 2) achieves higher accuracy by leveraging data of varying temporal resolutions including regional weather inputs with shorter lead times than the most common medium term weather inputs, and 3) generates individual forecast traces for each individual weather source, facilitating inference across regions where weather data availability is limited. 

Initial results across CAMELS sites in North America indicate that the incorporation of near-term high resolution weather data increases early horizon forecast KGE by nearly 0.25 with meaningful improvements in metrics seen across our customers’ operational sites. Validation metrics across individual weather sources, as well as model interrogation through integrated gradients highlights a high level of fidelity in the model’s learned physical relationships across forecast scenarios.

How to cite: Lambl, D., Topp, S., Butcher, P., Elkurdy, M., Reed, L., and Sampson, A. K.: Increasing the Accuracy and Resilience of Streamflow Forecasts through Data Augmentation and High Resolution Weather Inputs, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-12110, https://doi.org/10.5194/egusphere-egu25-12110, 2025.

Machine learning (ML) models have transformed our ability to perform reasonably-accurate, large-scale river discharge modeling, opening new opportunities for global prediction in ungauged basins. These ML models are data-hungry, and results have conclusively shown that ML techniques do best when a single ML model is trained on all basins in the dataset. This is contrary to inuitions from the hydrological sciences, where individual basin calibration traditionally provides the best forecasts. 

We bridge this gap between intuitions from traditional ML and hydrology by pre-training a single global model on basins in the worldwide Caravan dataset (~6000 basins), and then fine-tune that model on individual basins. This is a well-known practice within ML, and for us serves the purpose of producing models aimed at high-quality local prediction problems while still capturing the advantages of large-sample training. We show that this leads to a significant skill improvement. 

We have also conducted analysis of geophysical and hydrological regimes that benefit most from fine-tuning. These results point to how flood forecasting and water management agencies and operators can expect to fine-tune large, pretrained models on their own local data, which may be proprietary and not part of large, global training datasets.

This work illustrates how local agencies like national hydromet agencies or flood forecasting agencies might be able to leverage machine learning based hydrological forecast models while also maximizing the value and information of local data by tailoring large, pretrained models to their own local context. This is an important step in allowing local agencies to take ownership of these global models, and directly incorporate local hydrological understanding to improve performance.

How to cite: Ryd, E. and Nearing, G.: Fine Flood Forecasts: Calibrating global machine learning flood forecast models at the basin level through fine-tuning., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-13027, https://doi.org/10.5194/egusphere-egu25-13027, 2025.