This session focuses on the role of scalable, adaptive, and AI-enhanced computing approaches in supporting the entire natural hazard management cycle: from early detection and warning to modelling, impact forecasting and decision support. We invite contributions that explore but not limited to innovative methods and real-world applications across the areas such as: (i) Early detection and rapid warning systems, leveraging sensor networks, remote sensing, and predictive analytics, (ii)Time-critical simulations and forecasting models, (iii) AI applications in natural hazard contexts, including real-time/near real-time earthquake signal analysis, landslide and wildfire risk mapping, flood extent detection, and uncertainty-aware forecasting using ML-based ensemble models, (iv) Operational platforms and decision-support tools, integrating real-time data streams with adaptive modeling, (v) Climate change impacts on hydro-geological hazards, with a focus on floods, landslides, and droughts, (vi) Physics-informed learning and the integration of climate scenarios, (vii) AI-driven coupled hazard modeling using multi-source data, (viii) Representation of hydrological interactions among atmosphere, vegetation, and soil, and, (ix) Case studies demonstrating the application of such methods etc.
We invite contributions that showcase novel approaches in computational science, AI / machine learning, modeling systems, or hybrid workflows that improve readiness and responsiveness during natural disasters. We particularly encourage interdisciplinary submissions that highlight collaborative work across geoscience, computer science, and emergency management. This session aims to bring together researchers, practitioners, and system developers working at the intersection of geoscience and urgent computing to advance the state of natural hazard mitigation and civil protection.
The south of the Iberian Peninsula, particularly the Baetic System, is one of the most seismically active regions of the Iberian Peninsula. Its complex seismotectonic configuration causes recurrent moderate to strong earthquakes, posing a significant hazard to society and the built environment, requiring rapid and accurate post-event assessment of ground-motion intensity. These high-risk areas coincide with densely populated areas of Murcia, such as Lorca, or the province of Almeria. In addition, population dynamics vary significantly between summer and winter, due to seasonal tourism and residential tourism, which increases vulnerability and the need for rapid and accurate assessments following an earthquake. To address this need, the Machine Learning Estimator for Ground Shaking Maps (MLESmap) was developed as a rapid-response framework that combines high-quality physics-based simulations with Machine Learning techniques to infer spatially distributed ground-motion intensity measures within seconds after earthquake initiation. Trained on a large ensemble of synthetic seismic scenarios, MLESmap provides near real-time predictions of ground-motion intensity fields, such as acceleration levels and shaking patterns.
Our methodology incorporates both offline and online phases in a comprehensive workflow. It begins with the generation of a synthetic training data set generated by the CyberShake platform. Then predictor characteristics are extracted before the validation and learning stages. The result is a model that can be used for fast inference validated with start-of-art methodologies and available real data .
To evaluate the influence of surface representation on model performance, synthetic simulations are carried out using both 1D and 3D seismic velocity models, allowing for a systematic comparison of their impact on training and prediction accuracy. In addition, different learning strategies are explored, as for example, multi-objective approaches that allow for the simultaneous estimation of multiple measures of ground motion intensity. These analyses quantify the influence of velocity model dimensionality and training strategy on the performance of MLESmap predictions
Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Spain, Italy, Iceland, Germany, Norway, France, Finland and Croatia under grant agreement No 101093038, ChEESE-2P, project PCI2022-134980-2 funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR)
How to cite: Blanco-Prieto, R., Zamora, N., Monterrubio-Velasco, M., and de la Puente, J.: From physics-based simulation to ground motion models using Machine-Learning Estimator for Ground Shaking Map, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-1865, https://doi.org/10.5194/egusphere-egu26-1865, 2026.
Mining-induced geological hazards in the mountainous regions are often characterized by wide impact zones and complex subsurface structures, which pose significant challenges for the precise identification of landslides and the analysis of their formation mechanisms. To address this issue, we takes the Leji landslide in southwestern China as a case study and integrates SBAS-InSAR technology, two-dimensional decomposition modeling, UAV photogrammetry, field geological investigation, and AMT sounding to establish a multidimensional “Space-Air-Ground-Subsurface” detection strategy. This multi-dimensional framework enables the systematic acquisition of both surface deformation and subsurface structural information of the Leji landslide, thereby elucidating its controlling factors and causative mechanisms. The results reveal that the central parts of Landslide I and Landslide II exhibit the most significant deformation, with surface displacement dominated by downslope subsidence. The maximum annual average subsidence rates range between –60 mm/y and –80 mm/y. The cumulative deformation zones retrieved by SBAS-InSAR closely coincide with the mining areas detected by AMT. Through data fusion, the boundary angles of the mining areas were determined as 77° in the upslope direction and 48° in the downslope direction along the dip, and 77° and 55° in the strike direction. Comprehensive analysis indicates that the Leji landslide is a Quaternary soil creep landslide formed under the combined influence of fault–fold structures, frequent heavy rainfall, and both open-pit and underground mining activities, and it remains in an active state. This study demonstrates that the “Space-Air-Ground-Subsurface” collaborative observation system effectively overcomes the limitations of single techniques in landslide mechanism research, providing a reliable technical pathway and scientific basis for understanding the development mechanisms and disaster risk mitigation of mining-induced landslides.
How to cite: Zhang, Y. and Zhu, Y.: Using SBAS-InSAR and Audio-Magnetotelluric Sounding to Characterize Two-Dimensional Deformation and Failure Mechanisms in Mining-Induced Landslides, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3666, https://doi.org/10.5194/egusphere-egu26-3666, 2026.
In recent years, the Yizhong River basin in Deqin County, Yunnan Province, has experienced frequent debris flow events, posing a significant threat to surrounding residential areas and infrastructure. This study aims to investigate the hydrodynamic characteristics and hazard risk of debris flows in this basin under extreme rainfall conditions, providing a scientific basis for disaster risk reduction and prevention. The research employed UAV aerial photogrammetry, field investigations, and numerical simulation techniques to construct a high-resolution 3D terrain model of the Yizhong River basin. Using the continuum mechanics method based on deep integration and embedded by Physics-Informed Neural Networks (PINNs), the movement processes of flood-type and landslide-type debris flows were simulated under two extreme rainfall frequencies: 1% and 0.5%. Simulation results reveal that the frequent initiation of debris flows in the Yizhong River basin is influenced by multiple factors, including topography, material source conditions, and rainfall intensity. Under the 1% rainfall frequency, both types of debris flows trigger slope instability along the channel, leading to the entrainment of additional source material and enlarging the affected area. At the 0.5% rainfall frequency, the drainage channels within Deqin County were completely overwhelmed, with major transport arteries largely blocked. Substantial volumes of debris flow material entered the Zhiqu River, with overflow even burying Deqin No. 1 Middle School. Risk assessments for single-channel debris flow in the Yizhong River basin revealed that at the 0.5% rainfall frequency, debris flows within the Yizhong River channel reached an extremely high risk level, highlighting the inadequate capacity of the existing protection measures. Consequently, urgent attention must be given to stabilising unstable slopes along both banks of the Yizhong River basin and constructing drainage and diversion facilities within Deqin County. Future regional disaster prevention and mitigation efforts should prioritise curbing the frequent occurrence of debris flow disasters in the Yizhong River basin at their source.
How to cite: Wang, W., Zhu, S., Zhang, Y., and Chen, C.: Dynamic characteristics and risk assessment of debris flow under extreme rainfall in Yizhong River Basin of Deqin, Yunnan Province, from numerical simulation to PINN model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4409, https://doi.org/10.5194/egusphere-egu26-4409, 2026.
The EuroHPC Center of Excellence for Exascale in Solid Earth (ChEESE CoE, 2018–2026, DOI: 10.3030/101093038) is preparing 11 community-driven, open-source codes to run optimally on large accelerated supercomputing infrastructures (Leonardo, LUMI, MareNostrum-5). The CoE works with flagship codes in different areas of geophysics (earthquakes, tsunamis, volcanoes, magnetohydrodynamics, geodynamics, and glacier modelling), focusing on performance, scalability, CI/CD on EuroHPC systems, and portability across current and emerging hardware architectures. During 2025, the CoE has been awarded more than 1 million node-hours on GPU-accelerated systems such as Leonardo (4 NVIDIA A100 per node), LUMI-G (8 AMD MI250X per node), and MareNostrum-5 (4 NVIDIA H100 per node). The resulting simulations and use cases are being stored in data lakes together with their metadata for use by the scientific community, for example to train AI models or to be accessed through the European Plate Observing System (EPOS). All codes and applications under the ChEESE umbrella are available in open GitLab/GitHub repositories and undergo an SQAaaS process to obtain software quality badges. In addition, the project aims to enable urgent supercomputing services for emergencies during high-impact events (earthquakes, tsunamis, and volcanoes), including the associated technical challenges and recommendations on access policies. This is done in collaboration with end users such as civil protection agencies in various European countries. For example, ChEESE researchers tested an urgent supercomputing service for earthquakes during the September 19th 2025 drill, in collaboration with the Mexican Seismological Service (SSM).
Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Spain, Italy, Iceland, Germany, Norway, France, Finland and Croatia under grant agreement No 101093038, ChEESE-2P, project PCI2022-134973-2 funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR.
How to cite: Folch, A.: ChEESE: the European Center of Excellence for supercomputing in geosciences, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-5866, https://doi.org/10.5194/egusphere-egu26-5866, 2026.
Droughts are hydroclimatic anomalies driven by precipitation deficits and increased evapotranspiration, posing an escalating threat under a warming Mediterranean climate. Assessing drought risk remains challenging due to the complex interactions between biophysical conditions and human systems, as well as limitations in impact reporting. Moreover, drought impacts are highly heterogeneous across sectors, as different types of drought affect socio-environmental systems differently. In this context, we develop a spatio-temporal deep learning framework to predict sector-specific drought impacts and identify the environmental and climatic drivers of these impacts.
We combine two primary data sources: the European Drought Impact Database (EDID), which contains above 13,000 georeferenced drought impact reports spanning 1970 to 2023 and aggregated into four sectors (agriculture, ecosystem, energy, and socio-economic); and a set of physical drivers, including precipitation, temperature, drought indices, vegetation indices, and population density, derived from various sources for the period 2001–2021.
The prediction task is formulated as a spatio-temporal segmentation problem using a 3D U-Net architecture to capture dependencies in climate and environmental conditions over a one-year period. The preprocessing workflow harmonizes all variables to a spatial resolution of 0.25° and an 8-day time step. Seasonally varying predictors are transformed into anomalies, and all variables are normalized. Input samples are arranged as tensors with shape 36×48×16×16 (C×T×H×W), representing one year of conditions, while the target consists of a binary impact map (1×1×16×16) corresponding to the subsequent 8-day period. The training dataset is balanced through equal sampling of impact and no-impact cases. Consequently, the model learns to use one year of spatio-temporal context to predict drought-affected areas at the next time step.
Initial results for the agricultural sector indicate that traditional drought indices have limited predictive skill for drought impacts. A baseline evaluation of the Standardized Precipitation-Evapotranspiration Index (SPEI) across multiple thresholds shows that the 1-month SPEI achieves a PR-AUC of 0.13 and an ROC-AUC of 0.32 for the impact class over the 2018-2020 test period, with similarly low performance for the 3-, 6-, and 12-month variants. In contrast, preliminary model experiments demonstrate a substantial improvement over the baseline, achieving an F1 score of 0.43, a PR-AUC of 0.41, and a ROC-AUC of 0.71, despite remaining limitations in predictive performance.
These limitations are primarily attributed to noise and spatial uncertainty in the ground-truth labels, as EDID impacts are reported at coarse administrative units (NUTS3) and uniformly assigned to all grid cells within each region, constraining pixel-level learning. In addition, drought impacts are influenced by large-scale atmospheric circulation patterns and remote climate teleconnections (e.g., ENSO and NAO) that are not explicitly represented in the current feature set. Future work will address these limitations by incorporating large-scale circulation and teleconnection indicators, developing strategies to mitigate label noise, and extending the modelling framework to additional sectors. Once predictive performance is optimized, explainable AI methods based on Integrated Gradients will be applied to identify the most influential climatic and environmental drivers, enabling sector-specific interpretation of drought impact mechanisms and their temporal dynamics.
How to cite: Sapena, M., Papadopoulos, N., Athanasiou, G., Papoutsis, I., and Camps-Valls, G.: Predicting multi-sectoral drought impacts in the Mediterranean with spatio-temporal deep learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7822, https://doi.org/10.5194/egusphere-egu26-7822, 2026.
Heavy precipitation is a major hazard associated with tropical cyclones, often causing substantial economic losses and casualties through secondary disasters such as floods, landslides, and debris flows. The southeastern coast of China is one of the region most severely impacted by tropical cyclones. Under the context of global warming, the risks posed by tropical cyclone precipitation are expected to increase further. Accurate simulation of tropical cyclone rainfall is crucial for assessing flood hazards and provides a scientific basis for regional disaster risk mitigation policies. In this study, based on MSWEP precipitation data and tropical cyclone track data, we developed a China-focused tropical cyclone precipitation simulation model using the XGBoost algorithm reconstructed the precipitation field of TCs from 2000~2020. First, based on the tropical cyclone best-track data provided by the China Meteorological Administration, a rainfall field was constructed as a collection of 100 km × 100 km grid cells, forming an approximately circular domain with a radius of about 1000 km centered on the tropical cyclone. Mean precipitation for each grid cell was then extracted from the MSWEP dataset. Fifteen predictor variables were selected, including cyclone center latitude and longitude, grid center latitude and longitude, distance and azimuth between grid center and cyclone center, elevation, slope, aspect, wind speed and direction, cyclone forward direction, distance to land, season, and whether the cyclone center was over land. Based in MSWEP data from 2000 to 2020, a model was trained to predict precipitation in each grid using XGBoost algorithm. Based on this model, a reconstructed dataset of tropical cyclone rainfall for 2000–2020 was generated and evaluated. The main results indicate that, for a 70:30 train-test split, the model achieved RMSE=173.768mm, MAE= 85.504mm, and R²=0.674, demonstrating good performance. The simulated data effectively reproduce the spatial distribution of total tropical cyclone precipitation. Comparison of precipitation distribution maps based on MSWEP and simulated data further confirms that the model captures the spatial characteristics of total tropical cyclone rainfall with reasonable accuracy.
How to cite: Tao, K. and Xu, W.: A Machine Learning-Based Tropical Cyclone Precipitation Simulation in China, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8510, https://doi.org/10.5194/egusphere-egu26-8510, 2026.
Extreme climate change intensifies the spatiotemporal variability of soil moisture and temperature fields, thereby increasing the frequency and uncertainty of hydrogeological hazards such as floods, landslides, and droughts. These processes are governed by highly nonlinear water–heat coupling in unsaturated soil, where state variables and constitutive parameters are strongly interdependent. This complexity poses significant challenges for conventional physics-based numerical models due to difficulties in parameterization and uncertainty in boundary conditions, while purely data-driven models often lack physical consistency and interpretability. To address these limitations, this study proposes a hybrid modeling framework that integrates physical mechanisms with deep learning by embedding constitutive relationships and physical constraints derived from water–heat transport equations in unsaturated soil into a deep neural network. The proposed approach enables accurate prediction of the spatiotemporal evolution of soil moisture and temperature while preserving physical consistency. Numerical experiments were conducted for multiple soil types and boundary conditions, and the effects of data sparsity and noise on model performance were systematically evaluated. The results demonstrate that the hybrid model significantly outperforms purely data-driven approaches in terms of prediction accuracy and generalization capability, particularly in capturing localized moisture transport fronts and nonlinear dynamic behaviors. Further validation using bench-scale laboratory water–heat coupling experiments demonstrates that the proposed framework not only reconstructs key hydrothermal constitutive parameters but also successfully reproduces the temporal evolution of volumetric water content and temperature in unsaturated soil. Overall, this study provides a robust hybrid modeling strategy for simulating coupled water–vapor–heat processes in unsaturated soil. The proposed framework highlights the potential of physics-constrained deep learning for complex hydrological processes and supports its application in hydrogeological hazard analysis and risk assessment.
How to cite: Tian, S., Wang, Q., Lu, Y., Su, W., and Liu, Y.: Physics-Constrained Artificial Intelligence for Modeling Water–Heat Processes in Unsaturated Soil under Climate Change, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15418, https://doi.org/10.5194/egusphere-egu26-15418, 2026.
Groundwater plays an important role in flood formation yet, flood forecasting in coastal basins is often limited by inadequate representation of surface and groundwater interactions. In this study, we use a Graph Neural Network (GNN) to evaluate the added value of incorporating hourly groundwater information for short-term flood forecasting. Harris County, Texas is considered as a case study. The region is monitored by an extensive network of rainfall and channel-level sensors, supplemented by United States Geological Survey (USGS) wells providing hourly groundwater level data. Within the GNN framework, the sensor network is represented as a graph, where nodes correspond to monitoring areas and edges represent learned hydrological influence paths. Node inputs include recent precipitation, recent streamflow level changes, and normalized groundwater hydraulic load anomalies derived from Harvey Hurricane (2017) and post-Harvey flood events from 2018 to 2023. Results show that including a single groundwater-based prediction variable improves prediction ability by approximately 20% compared to precipitation and level-based reference models. This gain is strongest in areas with continuous groundwater withdrawal and accelerated recharge, where enhanced hydraulic gradients can intensify coastal storage exchange and enhance hydrogeological memory. The learned graph also provides an interpretable, directed interaction structure that supports data-driven causal hypotheses about network connectivity. Furthermore, we estimated the time delay dependency associated with the lag between two stations in our study area, which form a head-to-tail pair. The learned delay between these two stations is sub-daily, with a magnitude of ~0.5 to 0.7 days, corresponding to roughly 12 to 17 hours. This information can guide the parameterization of lag in rainfall-runoff modeling workflows. The results indicate that shallow groundwater dynamics can act as an important regulator of short-term urban flood response in coastal basins. When designing next-generation warning systems for Harris County and similar regions, groundwater levels and rainfall effects should be considered together.
How to cite: N Geykli, A., Gul, E., and Hassanzadeh, E.: Enhancing short-lead flood forecasting by integrated modeling of surface and groundwater , EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15179, https://doi.org/10.5194/egusphere-egu26-15179, 2026.
We present a Digital Twin that tracks the evolution of unrest caused by dike intrusion at volcanoes, leveraging HPC computational models and Artificial Intelligence algorithms to combine real-time monitoring data and physics-based predictions.
The Digital Twin includes three main components. A preliminary, offline scenario database is produced by simulating ground deformation due to dike
intrusion using the finite element HPC software GALES. The model calculates the three-dimensional elastostatic response induced by overpressurized dikes within a spectrum of geometries, positions and orien. The computational domain can include DEM topography and heterogeneous rock properties, e.g. from seismic tomography surveys. Scenarios are used to train a machine learning module that reconstructs the source of observed deformation patterns. The source is identified in terms of dike size, position, orientation and intensity (dike opening). An auto-encoder, trained on multi-parametric observational time series, detects unrest by identifying variations from the long-term trends at multiple stations. As unrest is detected, inversion of the observed deformation is performed by the trained ML module, providing the location and size of dike intrusion. The geodetic dataset is updated in near-real-time, providing the ability to model dike evolution as it rises towards the surface.
The Digital Twin has been applied restrospectively to the December 2018 dike intrusion at Mount Etna, tailoring ground deformation simulations to the specifics of the volcano, including observed distribution of dike properties. Results show the ability of the Digital Twin to identify unrest and track the evolution of the dike towards the surface to the eruptive vent.
The Digital Twin is available through a dedicated GitLab repository for the EU-funded DT-GEO project, including the case study application.
How to cite: Montagna, C. P., Bruni, R., De Paolo, E., Allegra, M., Garg, D., Cannavò, F., and Papale, P.: Near-real-time detection of dike-intrusion-indiced unrest: a Digital Twinfor Mount Etna volcano, Italy, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18997, https://doi.org/10.5194/egusphere-egu26-18997, 2026.
How to cite: Gatti, F., Perrone, N., Lehmann, F., and Fresca, S.: Shake Anywhere: a simulation-free AI-based earthquake ground motion generator for any source/any geology., EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3994, https://doi.org/10.5194/egusphere-egu26-3994, 2026.
Soil moisture dynamics play a critical role in slope stability, especially for rainfall-induced group-occurring landslides. With the growing availability of remote sensing–derived soil moisture products, there is increasing potential to improve landslide susceptibility assessment. However, few studies have explicitly incorporated both the spatial and temporal dynamics of soil moisture into susceptibility modeling. This study introduces a novel framework that integrates a Residual-Sparse Autoencoder (ResSAE) with Random Forest (RF), Support Vector Machine (SVM), and Artificial Neural Network (ANN) algorithms to enhance landslide susceptibility prediction using remotely sensed soil moisture data. Spatio-temporal soil moisture information for the study area in Nanping, China, is obtained from three open-access datasets: SMCI1.0, ERA5-Land, and SMAP-L4. Results show that antecedent soil moisture features extracted by ResSAE substantially improve prediction accuracy. The influence of rainfall, antecedent period length, and dataset source is further evaluated. Further analysis reveals that antecedent soil moisture over the prior seven days captures most of the hydrological memory relevant for slope failure, while additional rainfall data contribute only marginal gains. Optimal performance is achieved with ERA5-Land for RF, SMAP-L4 for SVM, and SMCI1.0 for ANN.Overall, the study highlights the importance of incorporating spatio-temporal soil moisture into susceptibility assessment. The proposed approach enables efficient and cost-effective predictions, supporting near-real-time applications and offering potential to strengthen regional to global rainfall-induced landslide prevention and mitigation strategies.
How to cite: An, N. and Xie, E.: Enhancing landslide hazard assessment by considering spatio-temporal soil moisture dynamics, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3068, https://doi.org/10.5194/egusphere-egu26-3068, 2026.
Urban flooding is emerging as an increasingly severe global challenge due to climate change and urbanization. Although machine learning offers numerous solutions for urban flood forecasting, its application remains constrained. Existing research remains constrained by the scarcity of traditional hydrological monitoring data, and the absence of systematic comparisons across multiple models creates uncertainty when selecting the most suitable algorithms and features, making the decision-making mechanisms for selecting the most suitable algorithms and features remains unclear. To address these challenges, social media data was adopted as the sole basis in this study to evaluate and compare the performance of seven typical machine learning algorithms in urban flood forecasting. The Shapley Additive exPlanations (SHAP) framework was established, investigating the adaptability of the selected algorithms based on a multidimensional feature system while elucidating the decision-making mechanisms for selecting the most suitable algorithms and features. The results suggest that: (1) Social media data can serve as the sole source for precise urban flood identification, overcoming the real-time and spatial coverage limitations of traditional methods. (2) Different machine learning models show significant performance heterogeneity; reliance on a single model risks systematic bias, whereas ensemble tree models demonstrate superior predictive performance. (3) Feature importance is highly model-dependent, exhibiting contextual sensitivity and interactive influence mechanisms. Therefore, feature engineering should be based on multi-model consensus, prioritizing features with significant differences such as natural characteristics and risk exposure.
How to cite: Miao, R., Huang, R., and Zheng, J.: Adaptability of Multiple Social Media Data Integrated Machine Learning Algorithms in Urban Flood Forecasting using the SHAP Framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3659, https://doi.org/10.5194/egusphere-egu26-3659, 2026.
Driven by global climate change, extreme weather events leading to short-duration heavy rainfall have emerged as a primary challenge for urban disaster prevention and resilience. Frequent and intense rainfall not only significantly increases the risk of urban pluvial flooding but also disrupts the stable operation of public infrastructure. Traditional drainage system designs often rely on static solutions that are inadequate for coping with the rapid intensity changes and high uncertainty of extreme rainfall, further exacerbating disaster risks in urban areas.
This study integrates advanced data analytics with machine learning to propose a rainfall and flood risk prediction system based on Self-Organizing Maps (SOM) and Long Short-Term Memory (LSTM). Leveraging Internet of Things (IoT) technology, the study incorporates high-resolution data (10-minute intervals) from flood-prone communities in Taipei City between 2015 and 2021. The multi-source dataset includes radar reflectivity, meteorological observations, sewer water level monitoring, and historical flood records to build a hydro-meteorological model with strong spatial and temporal representation. Preliminary results indicate that incorporating wind speed and direction data significantly enhances prediction accuracy and reduces uncertainty. Through SOM technology, the system performs refined classification of high-dimensional meteorological data, excelling in identifying extreme rainfall patterns. Combined with LSTM’s capability to capture temporal sequence characteristics, the system predicts rainfall and water level fluctuations. Furthermore, through a monitoring mechanism for sewer water level rise rates, integrating terrain and sewer spatial characteristics to provide localized, dynamic notifications and tailored response recommendations.
By combining AI-driven uncertainty analysis with real-time hydrological monitoring, this research strengthens flood forecasting capabilities under diverse wind field conditions, providing a science-based decision-support framework. The application of this model not only enhances the precision of community-scale flood prevention planning but also offers an adaptive regional warning strategy for urban climate adaptation. Ultimately, this system will effectively bolster urban disaster resilience and provide local governments with robust decision-support tools to achieve long-term sustainable development goals.
How to cite: Chen, L.-Y., Pan, T.-Y., Lai, J.-S., and Chang, M.-J.: Development of a Urban Flood Prediction Model Using SOM-LSTM: Integrating Environmental IoT and Sewer Water Level Rising Rates, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7239, https://doi.org/10.5194/egusphere-egu26-7239, 2026.
High-performance computing (HPC) plays a central role in advancing AI-based approaches for time-critical natural hazard applications, especially in regions where observational data are limited. In seismology, the scarcity of strong-motion records for large earthquakes poses a major challenge for the development of purely data-driven ground motion models. Here, we highlight the use of HPC to generate large, high-fidelity synthetic earthquake datasets specifically tailored for training machine-learning (ML) models for rapid ground motion forecasting in Southern Iceland.
Using the CyberShake workflow on HPC systems, we compute an unprecedented ensemble of approximately 100,000 physics-based earthquake scenarios, spanning magnitudes Mw 5.0–7.4, at 350 synthetic stations across the Southern Iceland Seismic Zone and the Reykjanes Peninsula Oblique Rift. Seismic wave propagation is simulated deterministically up to 2 Hz using three alternative Earth velocity models, allowing us to systematically investigate how subsurface velocity heterogeneity influences ground motion. By exploiting seismic reciprocity, the computational cost scales with the number of virtual recording sites rather than with the number of earthquakes, making it feasible to explore tens of thousands of rupture scenarios on Tier-0 HPC systems. The resulting simulations combine multiple velocity models, dense site coverage, and designed magnitude distributions, forming a comprehensive and carefully curated training dataset.
This large HPC-generated database is then used to train machine-learning surrogate models within the Machine Learning Estimator for Ground Shaking Maps (MLESmap) framework, including both tree-based ensembles and deep neural networks. Although these ML models provide near-instantaneous predictions of ground motion intensity measures during post-event response, their reliability ultimately depends on the quality, diversity, and physical realism of the underlying training data.
Our results show that HPC-driven simulation workflows can effectively close the data gap in regions with limited observations, delivering physically grounded datasets that support robust AI models for time-critical seismic hazard assessment. More broadly, this work underscores the role of HPC not only as a computational tool for modeling extreme events, but as a cornerstone of next-generation AI-driven systems for hazard forecasting and emergency response.
Funded by the European Union. This work has received funding from the European High Performance Computing Joint Undertaking (JU) and Spain, Italy, Iceland, Germany, Norway, France, Finland and Croatia under grant agreement No 101093038, ChEESE-2P, project PCI2022-134980-2 funded by MCIN/AEI/10.13039/501100011033 and by the European Union NextGenerationEU/PRTR).
How to cite: Monterrubio-Velasco, M., Zamora, N., Blanco-Prieto, R., Riaño, A. C., Vázquez, F., Chapagain, B., and de la Puente, J.: HPC-enabled large-scale physics-based seismic simulations as training data for AI-driven ground motion forecasting in Southern Iceland, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7756, https://doi.org/10.5194/egusphere-egu26-7756, 2026.
The GANANA project is an EU–India initiative that builds on three pillars: geohazards, weather and climate and life sciences, each linked to a EuroHPC Center of Excellence (CoE). In particular, the ChEESE-2P CoE pillar advances the use of High-Performance Computing (HPC) for geophysical hazard assessment and risk mitigation. It harnesses flagship HPC codes to deliver integrated, physics-based and data-driven solutions for earthquakes, tsunamis, smoke dispersion, and cascading hazards, with a strong focus on urgent computing,operational readiness and rapid response. We present GANANA’s high-level framework and first results across three core geophysical hazard domains. For earthquakes, urgent computing workflows enable near-real-time ground-shaking simulations using physics-based solvers, supporting rapid impact assessment for civil protection. These workflows are complemented by Artificial Intelligence / ML techniques for seismic data monitoring, where deep-learning pipelines automate event detection, phase picking, and magnitude estimation, and are tightly integrated with physics-based simulations to enhance robustness in data-scarce and tectonically complex regions. For tsunamis, GANANA extends established HPC workflows for rapid forecasting and high-resolution inundation mapping, triggered by seismic events, with particular emphasis on operational applicability and transferability to new coastal regions.
The workflow focused on wildfire spread and smoke dispersion, aims to develop an integrated forecasting system for urgent computing applications built upon expertise on the development of HPC codes for Numerical Weather Prediction (NWP) and atmospheric dispersion models. A defining feature of GANANA is its structured, bidirectional exchange of codes, expertise, and operational practices between Europe and India, enabling the adaptation, validation, and deployment of advanced HPC technologies in diverse geographical and institutional contexts.
A key aspect of the project is also the cascading hazard - framework. Preliminary demonstrators show that this exchange significantly improves model performance, interoperability, and time-to-solution, while simultaneously fostering capacity building and shared ownership of advanced HPC tools. GANANA thus illustrates how sustained international collaboration can transform mature exascale-ready codes into scalable, user-oriented systems for geophysical hazard forecasting and early warning.
How to cite: Zamora, N., Srivastava, N., Sánchez, C., Mingari, L., Folch, A., Macías, J., Monterrubio-Velasco, M., Diez-Ventura, G., Esquivel-Mendiola, L. I., Vázquez-Novoa, F., Badia, R. M., and de la Puente, J.: Physics-Based and AI-Driven HPC Workflows for Geophysical Hazards in GANANA project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-7839, https://doi.org/10.5194/egusphere-egu26-7839, 2026.
Hernandez, E.¹, Folch, A¹, Mingari, L.¹, Stramondo, S.², Trasatti, E.², Ganci, G.², Corradini, S.², Gonçalves, P.³, Brenot, H.⁴, Pacini, F. ³
- Geociencias Barcelona (GEO3BCN), CSIC, Barcelona, Spain
- Instituto Nazionale di Geofisica e Vulcanologia (INGV), Sezione di Bologna, Bologna, Italy
- Terradue, Roma, Italy
- Royal Belgian Institute for Space Aeronomy (BIRA), Brussels, Belgium
The ESA Geohazards Early Digital Twin Component (GET-it) project aims to deliver interactive, scenario-based tools for decision-making during geohazard crises. Within this framework, we present recent advancements in volcanic ash and gas dispersion modeling through the integration of satellite data assimilation into the FALL3D model. The main innovation consists of assimilating SEVIRI-derived SO₂ mass loading during the 2021 Cumbre Vieja eruption (La Palma) and volcanic ash during the 2018 Mount Etna eruption. These enhancements significantly improve the accuracy of quantitative forecasts of volcanic clouds, which are critical for aviation safety and public health.
The assimilation system implemented in FALL3D is based on the Local Ensemble Transform Kalman Filter (LETKF), an ensemble-based technique with localization designed to run efficiently on parallel computing platforms. The observation operator maps the model state to satellite retrievals, enabling sequential assimilation cycles. After each cycle, the corrected 3D concentration field initializes a new forecast, reducing uncertainty in cloud position and concentration. For La Palma, three assimilation steps were performed at 3-hour intervals using SEVIRI SO₂ retrievals, improving consistency with independent observations of cloud height.
To enable operational use, these simulations have been deployed on the Geohazards Thematic Exploitation Platform (TEP) by Terradue. The implementation leverages Common Workflow Language (CWL) workflows and Docker containers, ensuring reproducibility and scalability. The platform provides interactive visualization of eruption scenarios, including maps and time series, and allows users to modify key eruption source parameters (e.g., column height, intensity) through predefined scenarios (low, medium, high).
This work demonstrates the potential of combining Earth observation data with advanced numerical modeling in a cloud-based environment to deliver actionable information for crisis management. Future developments will focus on extending these capabilities to other geohazards and enhancing real-time operational readiness.
How to cite: Hernandez Plaza, E.: Advancing Volcanic Crisis Management through Satellite Data Assimilation in FALL3D within the ESA GET-it Digital Twin Framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-9859, https://doi.org/10.5194/egusphere-egu26-9859, 2026.
Tsunami Early Warning Systems (TEWS) in the NEAM region (North-East Atlantic, the Mediterranean, and connected seas) operate under strict time constraints, particularly for near-field events where coastal impact may occur within a few minutes. In the NEAM region, operational chains typically use decision matrices and precomputed scenario databases. In Spain, the TEWS is operated by the Instituto Geográfico Nacional (IGN), and this work is carried out jointly with IGN to support operational decision-making. These established tools can be reinforced with rapid products that provide early indicators of coastal impact within some minutes or even seconds of the first source estimate. One example is the use of Faster-Than-Real-Time (FTRT) simulations, already implemented in the current system.
Here we present a workflow in which neural-network surrogates are trained on large sets of physics-based tsunami scenarios, enabling fast inference of coastal impact metrics. The Tsunami-HySEA code is used to generate large-scale simulation sets, providing the data required by models designed for near-instant inference on standard CPUs. The surrogates models learn to map solid Earth earthquake source descriptors (capturing some uncertainty in fault parameters) to warning-relevant coastal metrics, focusing on maximum wave height and first-arrival time at multiple sites. Once trained, the models deliver predictions within seconds, facilitating rapid updates as source estimates evolve. Model interpretability is assessed using SHAP values, confirming how each input influences the predictions. The results confirm that the patterns follow the physical principles of tsunami generation and propagation. In an operational workflow, model results are fed into an automated reporting layer that produces tables, maps and graphics for Civil Protection within seconds, enabling rapid situational updates as source estimates evolve.
We first report initial results for Atlantic sources affecting SW Spain. Approximately 250,000 HySEA simulations covering multiple Atlantic fault segments, focal mechanisms and magnitudes were used to train models. The results for forecast points along the Huelva–Cádiz coast show good agreement with observed patterns of maximum wave height and meet operational speed requirements, with errors remaining within the acceptable range for TEWS procedures. We then describe the extension of the methodology to the Western Mediterranean, covering the Spanish Mediterranean coast and the Balearic Islands. This extension involves defining and parameterising multiple tsunamigenic fault systems, assembling and controlling the quality of high-resolution topo-bathymetric datasets, and designing robust training and validation strategies.
A practical limitation is that, despite comprehensive coverage of the targeted fault systems, rare source realisations or parameter combinations may fall outside the effective support of the training distribution, which can reduce reliability of point predictions. To handle such cases in operations, we complement deterministic estimates with threshold exceedance probabilities, enabling risk-aware decisions while preserving consistency with established TEWS procedures.
How to cite: Rodríguez Gálvez, J. F., Macías Sánchez, J., Gaite Castrillo, B., Sánchez Linares, C., González del Pino, A., Castro Díaz, M. J., Cantavella Nadal, J. V., and Puertas González, L. C.: AI- and HPC–Driven Tsunami Decision Support for the Spanish TEWS: Atlantic Results and Western Mediterranean Extension, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-10252, https://doi.org/10.5194/egusphere-egu26-10252, 2026.
The Lower Mekong River Basin (LMRB) is a flood-prone region experiencing increasing flood risk due to climate change and human activities. This growing challenge underscores the need for robust hydrological models capable of accurate flood prediction. Although purely deep learning approaches have demonstrated strong predictive performance, their data-driven nature does not explicitly represent the underlying physical mechanisms, which limits their interpretability.
In this study, we develop an interpretable deep learning framework based on a Long Short-Term Memory (LSTM) model to predict river discharge across multiple subbasins in the LMRB, with post-hoc interpretation provided by SHapley Additive exPlanations (SHAP). Feature contributions and dominant flood drivers are analysed using SHAP, enabling transparent interpretation of the model’s predictions. The LSTM model demonstrates high predictive performance, achieving Nash–Sutcliffe Efficiency values above 0.9 across all subbasins, although the largest flood peaks are slightly underestimated in midstream subbasins during extreme rainfall events. SHAP analysis indicates that soil-related variables are predominant contributors to discharge prediction, and their influence is partially mediated through interactions with precipitation and runoff. Furthermore, the relative importance of contributing variables changes over time: soil and vegetation-related variables dominate in earlier periods from 2013 to 2017, whereas hydrometeorological variables are more influential after 2017.
Overall, this study highlights the potential of post-hoc interpretable techniques applied to deep learning models for identifying the main contributing variables for discharge prediction and the drivers of flood events across the subbasins of the LMRB, providing valuable insights to support improved flood risk management.
How to cite: Qiu, Y., Shi, X., and He, X.: An interpretable deep learning framework for flood prediction in the Lower Mekong River Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14089, https://doi.org/10.5194/egusphere-egu26-14089, 2026.
Timely detection of landslide precursors is essential for life-saving early warnings, yet remains challenging due to the subtle, non-linear nature of pre-failure ground motion and the computational intensity of processing SAR time series. To address this, we present an operational automated InSAR framework, co-developed under India’s National Supercomputing Mission and the India-EU GANANA HPC collaboration, that processes multi-temporal SAR data optimized on HPC infrastructure (AIRAWAT) to enable long-term satellite-based monitoring large areas for landslide hazard assessment and early warning.
The system ingests Sentinel-1 SLC, IW data and ancillary geospatial layers (DEM and historical landslide inventories). Using GMTSAR-automated workflows, it generates displacement time series and LOS velocity maps across large, landslide-prone regions. These outputs are analysed to identify accelerated displacement trends for known landslides. Threshold values are identified based on the movement signatures, key precursors to slope failure, days to weeks before catastrophic events.
Critically, the entire pipeline, from SAR data ingestion to risk classification, is optimized for low-latency execution on HPC, enabling updates within 24–48 hours of new satellite acquisitions. Outputs are translated into a dynamic risk alert system (Green–Red) and delivered via an interactive dashboard with API access, designed for integration into national disaster response workflows.
Currently piloted in the Himalayas and Western Ghats, this framework demonstrates a scalable, HPC-driven paradigm for time-critical geo-hazard monitoring directly supporting rapid situational awareness and proactive evacuation decisions. The architecture is extensible to other InSAR-monitored hazards (e.g., subsidence, volcanic unrest).
The framework was tested with the well-documented Nepal Earthquake (7.8 M) on 25 April 2015, which triggered more than 47,200 co-seismic landslides. The displacement and coherence time-series were plotted at the crown points and centroids of the landslide polygon. The time-series plots show prominent trends towards the event date. Significant peak was observed in the displacement derived from 08 February 2015 and 21 April 2015 (Sentinel-1 Ascending) interferogram, which may be used as an early warning precursor.
How to cite: Singh, Y. K., Prabhu, T. S. M., Sidar, V. K., and Khare, M. K.: Towards an Operational InSAR Framework on HPC for Time-Critical Landslide Precursor Detection and Early Warning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16040, https://doi.org/10.5194/egusphere-egu26-16040, 2026.
Our study investigates the dynamics of the Gross Glienicker Lake, a groundwater fed lake in the Berlin-Brandenburg region of Germany. This lake (similarly to many others in the region) is experiencing a significant water decline mainly driven by the climate, loosing more than 2 meters of its water levels since the 1970s. To understand the hydrogeological system better, and to identify potential mitigation measures we applied a coupled groundwater-surface water model using HydroGeoSphere (HGS). This 3-D model simulates the hydrological processes of the catchment with high spatial and temporal resolution, incorporating all available geological and hydrological data from the area.
The model was mainly created to evaluate the impacts of different future climate projections on the water levels. We have investigated 3 different RCP scenarios using 43 different climate projection simulations. We have employed machine learning tools to fill in any future data gaps, for example future levels of a river boundary condition and future groundwater extraction rates given population growth trends. To access the uncertainties originating from the HGS model, we have used a meta modeling framework. Meta modeling uses a machine learning based surrogate model (an LSTM in this case), to emulate the input-output numerical relationship of the HGS model in a computationally efficient way. Once trained, the meta model can emulate an HGS model run accurately in a couple of seconds. We fed the meta model with thousands of perturbed climate inputs, showing that the model output is robust even under extreme climatic conditions.
Our results showed that the lake is highly sensitive to precipitation variability, therefore future projections diverge significantly given the scenarios. Except for the wet scenario, all predictions show further water level decrease and they also reveal a strong shift in the seasonal dynamics.
How to cite: Somogyvári, M., Mahmoodi, N., Ölmez, C., Tügel, F., Schneider, M., and Merz, C.: Meta modeling: using machine learning to assess the model uncertainty of a high-resolution groundwater model, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-16429, https://doi.org/10.5194/egusphere-egu26-16429, 2026.
Severe droughts in the Mekong Delta have exerted profound social and economic impacts in recent decades, underscoring the need for advanced predictive tools to enhance drought mitigation and preparedness. This study presents an AI-based framework that integrates precipitation moisture diagnostics with deep learning to significantly improve drought prediction in the Vietnamese Mekong Delta (VMD). First, moisture source contributions were quantified by using the Water Accounting Model-2layers (WAM-2layers), a moisture tracking tool with the ERA5 reanalysis data as inputs, revealing that over 60% of VMD precipitation originates from upwind source regions, with humidity and wind speed identified as dominant causal drivers of drought-period deficits. Building on this physical insight, a Convolutional Gated Recurrent Unit (ConvGRU) model was employed and explicitly trained with these external atmospheric variables. The model demonstrated robust multi-type drought forecasting skill at a 3-month lead, accurately detecting ~90% of meteorological and ~80% of agricultural droughts with low false-alarm rates (<10%), and reliably reconstructing major historical drought events. This work establishes a synergistic methodology, in which process-based diagnostics inform and validate an AI-driven prediction system, directly contributing to more reliable, physically interpretable early warning and supporting agricultural resilience and economic stability in this climate-sensitive delta.
How to cite: Shi, J. X. and Zhou, K.: AI-Enhanced Drought Forecasting: Fusing Moisture Source Diagnostics and Deep Learning in the Mekong Delta, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-17624, https://doi.org/10.5194/egusphere-egu26-17624, 2026.
Effective crisis management requires timely and accurate decision support, leveraging advanced computational methods and geoscientific insights. This study focuses on enhancing decision support for flood and landslide scenarios by integrating geoscientific concepts into prognosis modeling within immersive situation representation frameworks. Building upon the experiences and outcomes of the oKat-SIM project (optimized disaster response through simulation), we demonstrate how coupling high-performance computing with Geographic Information Systems (GIS) can improve real-time response capabilities in civil protection. The project aligns with the foundational goal stated in the Leopoldina report, emphasizing the significance of geoscientific process understanding in decision-making to prepare for, mitigate, and manage natural disasters effectively.
Our approach transcends traditional mapping by utilizing immersive and dynamic 3D representations through synchronized augmented reality (AR) glasses, allowing crisis management teams to maintain interpersonal communication while interacting with floating 3D scenario displays. This integration augments situational awareness and facilitates decision-making in high-pressure environments, such as crisis management centers. The involvement of end-users - both, operational and administrative personnel from municipalities and regional authorities - is crucial throughout the process of application development, allowing iterative improvements driven by real-world feedback.
Technical building aspects include: real-time landslide susceptibility and run-out modelling tightly coupled with GIS-based preprocessing and executed inside a Unity-based immersive runtime, enabling near-real-time scenario updates driven by HPC- and AI-assisted workflows. Advanced rendering techniques such as Gaussian Splatting, multi-resolution terrain streaming, and federated data fusion are leveraged to efficiently integrate remote sensing data, simulation outputs, and uncertainty layers into synchronized AR/3D views, providing scalable, low-latency situational awareness and decision support for time-critical crisis management.
Our case studies demonstrate the effective visualization of historical and potential disaster scenarios, fostering deeper understanding of complex interdependencies and enabling faster, informed decision-making. This interdisciplinary effort bridges geoscience and computational technologies, advancing operational platforms for flood and landslide preparedness and response, and fostering collaborative advancements for modern crisis management.
How to cite: Zeilinger, G., Kauling, S., Oswald, O., Prasannan, A., and Conrad, F.: Integrating Geoscientific Concepts in Prognostic Modeling for Immersive Situation Representation: Enhancements from the oKat-SIM Project, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19582, https://doi.org/10.5194/egusphere-egu26-19582, 2026.
Rapid and accurate estimates of ground-motion intensity measures are critical for seismic hazard assessment and disaster response. Empirical ground-motion models provide fast predictions, but suffer from large uncertainties, especially in regions with sparse observations. Physics-based simulations offer physically consistent shaking intensity estimates but remain computationally prohibitive for real-time applications and large-scale scenario analyses. We present a machine-learning framework that predicts high-resolution ground-motion intensity maps conditioned on earthquake source parameters, combining physics-consistent predictions with near-instantaneous inference. The framework predicts a suite of intensity measures widely used in seismic hazard and earthquake-engineering studies -- including peak ground velocity (PGV), peak ground acceleration (PGA), and response spectra -- for arbitrary double-couple sources embedded in a realistic 3D medium, inherently capturing complex geological and topographic effects.
Our approach leverages two complementary training datasets obtained from waveform simulations: spatially sparse shaking intensity maps generated via reciprocity methods and spatially dense intensity maps from forward simulations. A conditioned U-Net is first pretrained on abundant spatially sparse maps to learn global spatial features, subsequently fine-tuned using a limited set of spatially dense maps. This staged training strategy significantly reduces training data requirements while maintaining high predictive accuracy. Applied to the San Francisco Bay Area and Wellington, New Zealand, the framework produces physics-consistent intensity maps with speedups of 6–7 orders of magnitude compared to traditional wave-propagation simulations. This enables scalable, near-instantaneous hazard assessment for both rapid disaster response and comprehensive scenario-based analyses.
How to cite: Ramadan, F., Nissen-Meyer, T., Koelemeijer, P., and Fry, B.: Near-Instantaneous Physics-Based Ground-Motion Maps Using Sparse-to-Dense Deep Learning, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-734, https://doi.org/10.5194/egusphere-egu26-734, 2026.
Thermokarst activity is intensifying under a warming climate, and retrogressive thaw slumps (RTSs) in the Beiluhe region of the Qinghai–Tibet Plateau represent one of the most active examples. To produce regional, multi-year RTS inventories, we applied a domain-adaptation AI approach to improve model transferability across optical remote-sensing imagery acquired under diverse illumination conditions. From 2019 to 2022, the number of mapped slumps increased from 803 to 885, and the total affected area expanded from 1,727 ha to 2,329 ha. Despite these rapid changes, how hydroclimatic forcing, especially precipitation and land surface temperature (LST), jointly influences slump-related ground deformation remains unclear. Here, we analyze InSAR-derived surface deformation in relation to precipitation across different LST regimes. RTSs exhibiting larger seasonal deformation amplitudes also show higher subsidence rates. When LST is below ~0 °C, greater annual subsidence is associated with drier years; when LST is above 0 °C, greater subsidence more often occurs in wetter years. These results highlight precipitation and temperature controls on RTS deformation and emphasize the need to consider combined hydroclimatic conditions when interpreting remote-sensing deformation signals in permafrost terrain.
How to cite: Hu, X., Zhou, Y., Lin, Y., and Song, Y.: Hydroclimatic Controls on Thaw Slump Deformation on the Qinghai–Tibet Plateau, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-21782, https://doi.org/10.5194/egusphere-egu26-21782, 2026.
In recent years, technological advances in the use of geospatial data (such as satellite images, anthropogenic and/or environmental raster and vector open data, etc.) for hydrogeological risk assessment, combined with advanced analysis techniques (e.g., machine learning), have become increasingly valuable. These technologies can be utilized by local and national authorities for land planning and emergency management to better understand the dynamics associated with climate change. This understanding can help guide actions aimed at safeguarding not only environmental resources but also socio-economic assets and citizens’ lives.
In pursuit of this goal, a partnership has been established between the Po River Basin District Authority (AdBPo), the Italian Space Agency (ASI), and academic and research institutions such as the University of Bologna (UNIBO), the University of Modena and Reggio Emilia (UNIMORE), the University of Padova (UNIPD), and the Institute of Environmental Geology and Geoengineering of the National Research Council of Italy (CNR-IGAG). The aim is to implement a downstream service for monitoring landscape evolution related to fluvial systems (geomorphological classification), and slope dynamics (including landslides and rock glaciers) and to quantitatively evaluate the exposed assets.
The PARACELSO project (Predictive Analysis, MonitoRing, and mAnagement of Climate change Effects Leveraging Satellite Observations) aims to develop a modular and interoperable GIS cloud-based platform that supports the analysis of natural phenomena (such as fluvial hydrodynamics, landslides, and rock glaciers) using multi-sensor satellite data imagery provided by:
- DIAS platforms deployed by the Copernicus Programme (e.g., Sentinel 1-2),
- ASI missions such as CosmoSkyMed, PRISMA, and SAOCOM.
Furthermore, a methodology integrating Earth Observation and geospatial data analysis, to evaluate the exposed assets, has been implemented using open-source libraries.
To facilitate this, within the Big Data HPC MarghERita infrastructure— a supercomputing system named in honor of the scientist Margherita Hack and provided by the Emilia-Romagna Region — computational resources are employed for the high-performance processing, analysis, and storage of large volumes of acquired satellite imagery, as well as additional geospatial datasets. The platform executes the project-developed algorithms to investigate the temporal evolution of fluvial and slope systems. Furthermore, the infrastructure supports the access, visualization, and sharing of the processed and analyzed data.
The project has received funding from ASI through the “I4DP_PA (Innovation for Downstream Preparation for Public Administrations)” Call for Ideas.
How to cite: Hammouti, M., Zazzeri, M., Sterlacchini, S., Correa Da Mota, T., Mazzanti, M., Pancaldi, M., Agostini, M., Bizzi, S., Cecchetto, M., Berti, M., Brardinoni, F., Corsini, A., Tondo, M., Critelli, V., Mulas, M., Candela, L., D'Amato, L., and Simonelli, T.: Cloud-Based GIS Platform for the Management of Hydrogeological Risks in the Po Basin, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19618, https://doi.org/10.5194/egusphere-egu26-19618, 2026.
The processing and analysis of the large volumes of data generated by Interferometric Synthetic Aperture Radar (InSAR) require a significant investment of time, particularly in regions with complex geodynamic behavior. While InSAR presents notable advantages in terms of spatial coverage, precision, or data acquisition speed, traditional analytical methods can be insufficient to fully capture the complexity of deformation patterns or to efficiently manage the increasing amount of available data.
Integrating machine learning techniques into the InSAR computations and interpretation workflow enhances efficiency and automation. These methods enable automated detection of deformation patterns, improved separation of geophysical signals from atmospheric or orbital noise, and the identification of subtle or non‑linear ground motion that may be overlooked by conventional approaches. Such capabilities provide a more robust, reproducible, and sensitive framework for deformation analysis, which is essential for subsequent inversion procedures.
We describe in this presentation first results obtained in the Guadalentin Basin (SE Spain) using all these combined methodologies, as well as the comparison with previous studies for the area.
How to cite: Carrillo, R., Núñez, D., Pardo, E., and Fernández, J.: A New Machine Learning Method for Advanced Treatment of InSAR Deformation Data: Preliminary Results from the Guadalentín Basin (Spain), EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19335, https://doi.org/10.5194/egusphere-egu26-19335, 2026.
