The seismic risk assessment has gained significant popularity in recent years due to the increasing development of infrastructure and urbanization in seismically active locations across the globe. Earthquakes pose serious issues as natural events because of their unpredictability and the extensive harm they may do to infrastructure, buildings, and people's lives. Ground motion at the time of the earthquake can depend on several local sites and event characteristics, such as the size of the seismic event, the depth of the earthquake focus, the distance from the epicentre, the local geology and soil conditions. However, traditional probabilistic seismic hazards using ergodic ground motion models do not consider these variations, leading to a further less accurate damage or risk assessment. Hence, the present work aims to perform a comprehensive seismic risk assessment by incorporating three-dimensional physics-based numerical modelling, which explicitly incorporates the path and site-specific characteristics that cater for non-ergodicity. Here, physics-based ground motion has been simulated for controlling events corresponding to typical sites present in Shimla city, Himachal Pradesh, India. Furthermore, to assess the associated risk for the region exposure, data of the building inventory of Shimla has been gathered using Google Street View (GSV) images, and for the classification of the building inventory to different building typologies, deep machine learning-based Convolution neural network (CNN) models are trained. The developed CNN model has shown great precision in identifying the building class for the region. After classification, suitable well-known fragility functions are mapped to each class, and subsequent risk is calculated. Finally, the results developed using physics-based hazard are compared with the conventional empirical approach. The study results will provide the respective stakeholders with the technical knowledge for the region's hazard and subsequent risk.

How to cite: Shukla, S. S., Choudhary, R., and Jaya, D.: Seismic risk assessment using 3D physics-based seismic hazard: A case study for Shimla city, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-408, https://doi.org/10.5194/egusphere-egu25-408, 2025.

This study presents the site characterization of 133 selected stations in the National Accelerometer Network of Greece. All available earthquake recordings for various distances, magnitudes and azimuths around the station are compiled and processed to estimate a stable and reliable average Horizontal -to-Vertical Spectral Ratio (eHVSR). The earthquake records used have magnitude range 4 ≤ M < 6, with focal depth ranging from 0 to 40km and hypocentral distances 12 km ≤ R_hyp ≤ 300 km. Using the Diffuse Filed Concept for earthquakes (DFCe), and incorporating limited a priori geological information, available for the area around the station, the estimated eHVSRs were utilized in an inversion framework to estimate the best misfit velocity profiles down to seismological bedrock (where Vs>=3km/sec). Comparisons of these estimated velocity profiles with existing ones for the selected stations based on other geophysical or/and geotechnical methods, revealed good agreement, encouraging broader application of this methodology for the rest of stations.

In accordance with recommendations from the SERA project, seven key indicators were calculated for each of the 133 stations and are presented as follows: (1) Resonance frequency (f₀), (2) Shear wave velocity of the uppermost 30 meters (Vs₃₀), (3) Surface geology description, (4) Current EC8 site class, (5) Depth of the seismological bedrock (H_3km/s), (6) Depth of the engineering bedrock (H_0.8km/s) and (7) V_SZ full profiles where available. Such comprehensive site characterization of accelerometer stations enhances regional and international strong-motion databases (e.g. ESM db) and contributes to exploiting earthquake recordings to their full potential for engineering and seismological applications.

How to cite: Chatzianagnostou, E. and Theodoulidis, N.: 1D Site characterization of National Accelerometer stations in Greece based on earthquake recordings and the Diffuse Filed Concept (DFCe), EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-8382, https://doi.org/10.5194/egusphere-egu25-8382, 2025.

Although the European Forest Fire Information System (EFFIS), provided by the Copernicus Emergency Management Service, offers three different methods for determining forest fire danger, the Canadian method is usually used and accepted in Croatia. The Canadian Fire Weather Index (FWI) estimates the forest fire danger level based on meteorological parameters (air temperature, humidity, wind speed and precipitation amount) related to 12 UTC for the given day at the meteorological station or to a grid point of a numerical weather prediction model.

Thanks to the EFFIS statistics portal, it is possible to see the extent to which Croatia has been at risk from forest fires in recent years based on the areas burned and the number of fires. The Copernicus Climate Change Service (C3S) provides a much more detailed overview of the burned areas. The combination of data from the Climate Change Service and the Emergency Management Service can provide a better overview of forest fires in Croatia. The forest fire danger levels are analyzed spatially between different regions such as the continental, mountainous and Adriatic parts of Croatia. In order to find an appropriate duration of the fire season, the forest fires within and outside the fire season are listed. The aim of the spatio-temporal analysis is to show the most endangered areas and the seasonal trend of forest fires in Croatia.

How to cite: Škurić Kuraži, D. and Herceg Bulić, I.: Spatio-temporal analysis of forest fires in Croatia, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15531, https://doi.org/10.5194/egusphere-egu25-15531, 2025.

This communication presents a unified modeling framework for human-caused wildfire ignitions across representative European regions (pilot sites, PS), aiming to enhance understanding of ignition drivers and support wildfire risk management. Our approach models ignition probability at a fine spatial resolution (100 m), identifies key influencing factors, and enables cross-regional comparisons.

We calibrated Random Forest models using historical fire records and geospatial datasets, including land cover, accessibility, population density, and dead fine-fuel moisture content (DFMC). Models were developed individually for each PS and compared to a comprehensive model integrating all PS. Spatial autocorrelation effects on model performance were also evaluated.

Model performance was robust, with AUC values ranging from 0.70 to 0.89. DFMC anomaly emerged as the most influential variable across all PS. Among human-related factors, proximity to the Wildland-Urban Interface was most significant, followed by distance to roads, population density, and wildland coverage. The full model achieved an AUC of 0.81, highlighting mean DFMC and anomaly as dominant ignition drivers modulated by accessibility and population density. Local model performance, however, dropped by 0.10 AUC in regions such as Southern Sweden and Attica, Greece.

These findings underscore the importance of integrating fine-scale spatial and environmental data for wildfire ignition modeling. The developed models provide valuable insights into wildfire ignition hazards and support the implementation of targeted mitigation policies in fire-prone European landscapes.

How to cite: Gelabert Vadillo, P. J., Jiménez-Ruano, A., Ochoa, C., Alcasena, F., Sjöström, J., Marrs, C., Ribeiro, L. M., Palaiologou, P., Bentué-Martínez, C., Chuvieco, E., Vega-García, C., and Rodrigues, M.: Modeling Human-Caused Wildfire Ignition Probability Across Europe, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-11004, https://doi.org/10.5194/egusphere-egu25-11004, 2025.

Wildfires are an intensifying global challenge, driven by climate change, which increases their frequency, severity, and spatial extent. Accurate wildfire risk assessment and forecasting are essential for effective mitigation, resource allocation, and long-term planning. The Canadian Fire Weather Index (FWI) is a widely used fire danger rating system that integrates four primary daily meteorological variables—24-hour accumulated precipitation, wind speed, relative humidity, and temperature—into six components representing fuel moisture, ignition probability, and fire spread potential. Its temporal "memory" feature, which tracks moisture changes over time, makes it particularly valuable for capturing wildfire dynamics.

However, the FWI reliance on specific daily input data at noon poses challenges for its application in regions or scenarios lacking such precise temporal measurements. To address this limitation, FWI proxies computed using daily mean data offer a practical alternative. Yet, these proxies often lack the fidelity required to fully replicate the FWI values.

This study focuses on enhancing the emulation of the original FWI using daily mean data and other proxy variables by leveraging advanced deep learning techniques. We explore a spectrum of architectures, ranging from conventional machine learning models to state-of-the-art approaches like convolutional neural networks (CNNs) and Convolutional Long Short-Term Memory (ConvLSTM) networks. These models are tailored to capture the spatial and temporal complexities of wildfire behavior while maintaining robustness in the face of variable data availability.

Our research centers on the Iberian Peninsula, a Mediterranean region highly vulnerable to extreme wildfire events. By utilizing high-resolution, geo-referenced datasets, we validate the ability of these models to emulate the original FWI with high accuracy. To enhance model interpretability, we integrate eXplainable Artificial Intelligence (XAI) techniques, providing actionable insights into the decision-making processes and addressing concerns about the "black box" nature of deep learning.

This work demonstrates how daily data, combined with cutting-edge deep learning methods, can effectively emulate the FWI, offering a scalable and reliable solution for wildfire risk prediction in regions where traditional inputs are unavailable. The proposed models bridge the gap between limited data availability and the growing need for precise fire danger indices, enabling improved assessment and planning for wildfire-prone regions.

By advancing the science of wildfire modeling through daily data-driven approaches, this study contributes to a deeper understanding of spatial and temporal wildfire dynamics. It highlights the potential of integrating geoscience, climatology, and artificial intelligence to develop practical tools for wildfire risk mitigation, resilience, and decision-making in a rapidly changing climate.

Acknowledgments: This research work is part of R+D+i project CORDyS (PID2020-116595RB-I00) with funding from the Spanish Ministry of Science, Innovation and Universities MCIN/AEI/10.13039/501100011033. O.M. has received the research grant PRE2021-100292 funded by MCIN/AEI /10.13039/501100011033.

How to cite: Mirones, Ó., Baño-Media, J., and Bedia, J.: Daily Data-Driven Emulation of the Fire Weather Index: Deep Learning Solutions for Wildfire Risk Prediction, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16116, https://doi.org/10.5194/egusphere-egu25-16116, 2025.

Urgent Computing (UC) refers to the use of High-Performance Computing (HPC) and High-Performance Data Analytics (HPDA) and Artificial Intelligence (AI) modules during or immediately following emergencies. It typically integrates complex end-to-end workflows with scalable computing resources, where multiple model realizations are necessary to account for input and model uncertainties, all under strict time-to-solution constraints. Enabling urgent HPC in unpredictable events such as earthquakes can significantly enhance resilience and response efforts. The temporal horizon for UC usually spans from minutes to a few hours, providing decision-makers with rapid estimates of the potential outcomes of emergency scenarios. In particular, high-resolution synthetic ground motions for earthquakes can complement the tools used by seismological services for impact analysis. Here, the Urgent Computing Integrated Services for Earthquakes (UCIS4EQ) is proposed as an innovative UC seismic workflow designed to rapidly generate synthetic estimates of the consequences (such as synthetic time histories, shakemaps, PGA/PGV, among other proxies) of moderate to large earthquakes (M > 6). Over the last six years, UCIS4EQ has been developed from scratch and received contributions within the framework of three European projects (DT-GEO, eFlows4HPC, and ChEESE CoE). In this work, we demonstrate the technological maturity of UCIS4EQ and its operational readiness in collaboration with the Mexican Seismological Service (SSN). Furthermore, this work addresses the challenges we face to reach operational maturity addressing the specific requirements of a seismological service for an urgent computing framework providing reliable outcomes for decision making with global coverage.

How to cite: Monterrubio-Velasco, M., Boehm, C., Iglesias, A., Diez, G., Bhihe, C., Esquivel, L., Zamora, N., Tuinstra, K., and de la Puente, J.: Earthquake Shaking Simulation Workflow for Urgent Computing Services: Challenges and Advances, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17470, https://doi.org/10.5194/egusphere-egu25-17470, 2025.

The Chamoli Glacial flood happened in the Indian state of Uttarakhand on the 7^th of February 2021. This disaster was triggered by a rockslide-induced glacial burst near the Ronti peak. The event unleashed a massive debris flow that devastated the area’s critical infrastructure, including the Rishiganga and Tapovan Vishnugad hydropower projects. The event underscored the vulnerability of the fragile Himalayan geology, challenges in development, disaster preparedness and early warning systems.

PARATUS project's forensic approach is based on the combination of three specific forensic methodologies: Investigation of Disasters (FORIN), Post Event Review Capability (PERC), and Detecting Disaster Root Causes (DKKV). The forensic analysis investigates the disaster’s causes, multi-dimensional impacts and responses, highlighting the key vulnerabilities across physical, socio-cultural, economic and institutional dimensions. The study identifies poor infrastructure resilience, environmental degradation and limited emergency response capacity as major contributors to the severity of the disaster. Cascading effects such as sedimentation and artificial lake formation further exacerbated the risks. The immediate aftermath saw significant disruptions in transportation and communication networks, hindering rescue operations despite the swift deployment of ground and aerial relief to the affected population.

In the recovery phase, coordinated efforts under India’s National Disaster Management Plan facilitated relief and reconstruction. However, challenges associated with the long-term rehabilitation of the people affected by the disaster still persist. The governmental institutions are currently focusing on building resilience through slope stabilization, improved early warning systems and sustainable infrastructure development. Addressing systemic vulnerabilities, including governance gaps and socio-economic inequities remains a critical step toward mitigating future risks. This forensic analysis builds on existing scientific literature and institutional reports revealed by the Government of India to assess and emphasize the necessity of integrating multi-hazard approaches and localized strategies for disaster risk reduction in vulnerable mountainous regions like the central Himalayas.

How to cite: Ghosh, P., den Bout, B. V., Westen, C. V., and Atun, F.: Chamoli Glacial Burst: Investigating the vulnerability of the Himalayan geology with the support of Forensic Analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-20476, https://doi.org/10.5194/egusphere-egu25-20476, 2025.

Subsidence is a geological phenomenon that continuously affects Mexico City. Over time, the impact of this phenomenon has been extensively studied using various methodologies, primarily at a regional scale. In recent years, efforts have shifted toward mapping subsidence at a local scale using technologies such as photogrammetry and LiDAR. These studies aim to establish a reference database to validate or complement regional-scale initiatives.

Field-based studies on subsidence often involve identifying problematic areas and analyzing topographical changes and structural damage over time. However, it is crucial to quantify and understand the limitations and capabilities of these techniques to establish a reference framework and ensure the reliability of the obtained data. Currently, precision methodologies are within everyone's reach thanks to technologies like photogrammetry and LiDAR from smartphones.

To achieve this, two controlled experiments (one conducted in the field and one in a laboratory setting) were carried out, in which 3D reconstructions of a box with known dimensions were made. Ten photogrammetry and ten LiDAR surveys were performed to compare the measurements obtained from the digital model with those taken from the physical object.

In the laboratory experiments, the average percentage error using photogrammetry was 1.03% (0.20 cm). Specifically, the error for a 16-cm-tall box was 1.44% (0.27 cm), while for a 20-cm-tall box, it was 0.61% (0.12 cm). For LiDAR, the average percentage error was 1.51% (0.27 cm), with errors of 1.50% (0.26 cm) for the 16-cm box and 1.52% (0.27 cm) for the 20-cm box. In field experiments, photogrammetry yielded an average percentage error of 0.88% (0.3 cm), whereas LiDAR showed an average error percentage of 2.17% (0.62 cm).

These findings confirm LiDAR and photogrammetry's potential for high-precision subsidence monitoring, providing a robust and accessible validation method. Utilizing mobile devices such as the iPhone 13 Pro Max extends the reach of these methodologies, enabling more accessible and practical research in urban contexts where subsidence poses significant challenges to infrastructure and quality of life.

How to cite: Vidal, G., Ramírez, N. L., Jácome, M. P., López, N., Reyes, T. A., and Yépez, F. D.: Accuracy Analysis of Photogrammetry and LiDAR Point Clouds Using an iPhone 13 Pro Max, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7731, https://doi.org/10.5194/egusphere-egu25-7731, 2025.

To address the challenge of inconsistent line-of-sight (LOS) deformation datum derived from interferometric measurements of different Synthetic Aperture Radar (SAR) images—and the significant variation in LOS direction between near-range and far-range within the same image—this contribution proposes an InSAR deformation datum  connection method with a fixed LOS direction. The method combines Bayesian inference and the Markov Random Field (MRF) model, integrating InSAR and GNSS deformation data to achieve unified deformation datum for adjacent and even different-orbit SAR interferometric results.

A simulation experiments, using Sentinel-1 imaging parameters and GNSS velocity field data, and a real-world validation with InSAR data of the 2023 Southern Turkey earthquake are conducted. In the simulation, the root-mean-square error  of LOS displacement rate difference in the overlapping regions of adjacent-track SAR images decreased 99%. In the real-world experiment, the root-mean-square error of displacement difference reduced from 20 mm to 8 mm, demonstrating the effectiveness of the proposed method.

We have three key contributions：(1) Unified Deformation datum: Successfully realize an InSAR deformation datum connection with fixed LOS direction in SAR images; (2) Adjacent-Track Stitching: Achieve seamless stitching of adjacent-track SAR deformation results from a single data source; (3) Real-Data Validation: Reduce the mean displacement difference in overlapping regions of adjacent-track SAR images of the 2023 Southern Turkey earthquake from 20 mm to 8 mm.

How to cite: Bian, W., Motagh, M., and Wu, J.: InSAR Deformation Datum Connection with A Fixed Line-of-Sight Direction: A Bayesian inference and the Markov Random Field (MRF) model integration, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19686, https://doi.org/10.5194/egusphere-egu25-19686, 2025.

In 2023, Otis strengthened from a slight tropical storm into a major hurricane (Category 5) within only about 12 hours before it made landfall. The storm slammed into Mexico's coast with maximum sustained winds of over 165 mph and hurricane-force winds extending up to 30 miles from its center. The SICT (Secretariat of Infrastructure, Communications and Transportation) warned of a total closure of the Mexico-Acapulco highway in the Chilpancingo-Acapulco section. Faced with reports of hundreds of landslides through the lines, the SICT deployed more than 1000 workers, 100 vehicles and 300 pieces of heavy machinery in the hope of “restoring traffic as soon as possible and providing safety to users.” Unfortunately, predictions could not anticipate close enough the Otis destructive force.

Ensuring the proper functioning of road infrastructure is a fundamental aspect in risk management. Landslides have the potential to impair critical transportation infrastructure, particularly road networks in the hilly regions in Mexico. Recognizing the extremely changing climate conditions in the Mexican Pacific coasts are becoming increasingly difficult to predict, in this research advanced technologies are integrated into an intelligent digital scenario to simulate and control this linear infrastructure before, during and after extreme rainfalls occur.

The strategic roads digital twin comprises i. dynamic susceptibility maps, ii. satellite radar information of control points (the landslides pathologies are easily detected through them), iii. an artificial intelligence slope stability calculator (in near real-time) for pointing incipient instability, and iv. a semi-immersive scenario for analyzing future states based on the information of pluvial stations and control points, once this information is analyzed with the intelligent calculator. For communicate the input conditions, the aggravating factors and the future responses, a digital twin of potentially affected road sections (detected on the dynamic maps) is developed. Simulate scenarios before rainfall increases, help to make informed maintenance and risk prevention decisions in road infrastructure in areas with high geotechnical complexity and strong seasonal rainfall patterns. Exploiting precalculated extremely dangerous conditions, this digital twin can serve as an early warning system because it is programmed for immediate communication of graduated alarms that announce the proximity to dangerous states.

How to cite: García, S., Trejo, P., and Ángeles, B.: A digital twin for management of landslides and slope incidents on strategic road infrastructure, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14209, https://doi.org/10.5194/egusphere-egu25-14209, 2025.

Gravitational mass movements are recurrent events in Brazil, usually triggered by intense rainfall. When such rainfall events occur in urban areas, particularly on slopes, they often result in disasters, causing loss of human lives, social impacts, and economic damage. Thus, mapping and monitoring landslide susceptible areas are extremely important, as well as the implementation of a system capable of predicting their occurrence in advance. In this context, this study aims to assess the efficiency of the TRIGRS numerical model as a component of a prediction system for landslides on urban slopes. As a first step, the influence of the drainage network, which is altered due to urbanization on slopes, will be analyzed in relation to the safety factor, moisture profile, and pore pressure. The drainage network was calculated using a digital terrain model derived from LIDAR data. The TRIGRS model was applied to a small watershed located in the municipality of Campos do Jordão, São Paulo, Brazil. During the 72 hours analyzed period, two heavy rainfall events stroke the area and landslides were registered. The registered landslides show the model efficiency on the identification of the most susceptible areas, because they happened in areas identified by TRIGRS as extremely susceptible to landslides. The combined geotechnical and geophysical methodology for soil characterization and the use of more realistic drainage network feeding the TRIGRS has shown to be useful urban planning and early warning systems. This study is part of Brazilian Council for Scientific and Technological Development (CNPq) Project coordinated by GEGEP/UFPE, with the participation of Cemaden, and in collaboration under development with the National Research Council of Italy (CNR). It aims to implement a methodological procedure.

How to cite: Ernesto de Moraes, M. A., M. Mendes, R., Bortolozo, C. A., Metodiev, D., das Dores S. Medeiros, M., R. M. Andrade, M., S. G. Mendes, T., and Q. Coutinho, R.: Effects of Drainage Network on the Identification of Landslide-Susceptible Areas Using the TRIGRS Model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-21040, https://doi.org/10.5194/egusphere-egu25-21040, 2025.

Quantifying the vulnerability of roads caused by debris flows is crucial for regional hazard mitigation in remote areas. However, the changing climate has increased the uncertainties in providing reliable vulnerability assessment due to the altered pattern of rainfall. Such change has induced the increased frequency and magnitude of debris flows, impacting the safe operation of highways. In this case, a reliable method was developed to help on the improvement of vulnerability quantification with the involvement of AI and Flo-2D simulation techniques before applying this proposed framework to a case study in the Gyirong Zangbo Basin, Tibet, China. In detail, a deep learning model was developed to estimate the physical vulnerability of roads in the event of a future debris flow with the consideration of a series of factors, including spatial locations of roads to the debris-flow channel, debris-flow catchment area (A_c), length of main channel (L), topographic relief (R), mean slope of main channel (J), and rainfall (P). After that, debris-flow simulations were performed to validate the physical vulnerability assessment results, which can further benefit the accurate quantification of economic loss on a regional scale. Here, in addition to the direct loss of the damaged roads, the indirect loss caused by the damaged roads was also estimated using a complex network theoretical approach that account for regional socioeconomic development and the time needed for road restoration. Overall, this study can form part of an early warning system to assist on the effective management of debris flows on a regional scale in mountainous areas.

How to cite: Qiu, C.: Vulnerability quantification of roads caused by future debris flows in mountainous areas, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3505, https://doi.org/10.5194/egusphere-egu25-3505, 2025.

Geohazard chains in watersheds often involve a series of interconnected events, such as landslides that propagate along slopes, intrude into river channels, form landslide dams, and result in dam breaches and outburst flooding. Because the sub-processes within a geohazard chain are coupled, one or more of these events can trigger subsequent ones, leading to larger spatial and temporal scales than isolated disasters. This results in more destructive power and a wider impact area. In this study, a numerical case study focuses on the most recent geohazard chain event: the 2018 Baige landslide in Sichuan Province, China. This event can be divided into several sub-processes based on the coupling order within the chain. The first landslide formed a landslide dam, followed by another landslide at the same location, which overlapped with the first, creating a higher dam. This ultimately led to a larger-scale dam breach and outflow.
To simulate this sequence, a series of validated depth-averaged models for geohazard chains was employed, along with a standard LxF central differencing scheme to retain high resolution and avoid Riemann characteristic decomposition. The landslide propagation was modeled using a visco-inertial friction law. The numerical predictions were verified against field measurements from the literature, demonstrating the feasibility of using μ(K) visco-inertial rheology to simulate large-scale landslides and landslide dam formations. The overtopping failure of the two overlapping landslide dams and the subsequent outburst flooding were successfully simulated using the proposed model. Maximum discharge results indicate the model's capability to capture the interaction between dam breaches and outburst floods. The numerical findings, validated by existing literature, provide a reliable assessment for emergency relief and hazard mitigation. This modeling framework is expected to contribute to improved mitigation strategies for geohazard chains.

How to cite: Xie, Y., Zhou, G., Cui, K. F. E., Lu, X., and Li, N.: Numerical study of  2018 Baige landslides induced geohazards chain and dynamic proesses, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15456, https://doi.org/10.5194/egusphere-egu25-15456, 2025.

Climate change has intensified droughts in many parts of the world, severely impacting different sectors. In particular, the agricultural sector is highly sensitive to precipitation deficits and the resulting soil moisture deficit, leading to a drastic reduction in crop productivity. There is an urgent need to ensure access to food for a growing population in future, making it essential to address agricultural drought induced crop yield losses. Multimodal satellite and reanalysis climate data archives, coupled with advancements in machine learning, offer a promising avenue to address this issue, but studies are often limited to the calculation of drought indices. In order to produce actionable insights and allow for time to prepare for drought-related food production deficits, specific information on crop losses is needed. Therefore, this study demonstrates the potential of the machine learning algorithm Random Forest (RF) for annual crop yield forecasting using multimodal datasets, for two agriculturally important drought-prone regions in India and Germany. Using 11 climate variables from ERA5 data and PKU GIMMS NDVI (version 1.2) from 1990 to 2021, an RF model was trained to predict crop yields for two common crops across the study sites. The model was evaluated at different spatial scales and the spatial transferability of the model was also tested, using Root Mean Square Error (RMSE; absolute error metric) and Mean Absolute Percentage Error (MAPE; relative error metric). Feature importance was also assessed across scales and across different study sites, using the mean decrease in impurity as a post-hoc explainability tool. Results show that different features are important for accurate crop yield predictions in different regions, for different crops, and across different space-time scales. Spatial transferability requires retraining the model with local data, due to the strong influence of local climatic and agricultural conditions as well as data availability. Findings pave the way for long lead time predictions of drought impacts on agricultural productivity purely open source data, contributing directly to improving global food security equitably, as the methods are equally applicable in data-rich and data-poor contexts. 

How to cite: Sahu, H., Garg, P. K., Vijay, S., and Dasgupta, A.: Estimating drought impacts on crop yield using AI and EO, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2768, https://doi.org/10.5194/egusphere-egu25-2768, 2025.

The Amazon Basin (AB), the largest hydrographic basin in the world, spans across seven countries in South America. It constitutes a highly intricate system, rich in natural resources, and is marked by substantial biological heterogeneity. The AB plays a pivotal role in the regulation of environmental processes, serving a key component of the global hydrological cycle and climate systems. Understanding the increasing frequency, intensity and spatial extent of extreme drought events in this region is vital for safeguarding the regional ecosystem. This study aims to classify extreme drought events in the AB using the Standardized Precipitation-Evapotranspiration Index (SPEI), derived from ERA5 reanalysis data, covering the period from 1980 to 2024. To assess both agricultural and hydrological droughts, this research incorporates the accumulation periods of 6 and 12 months (SPEI-6 and SPEI-12). The ranking methodology accounts for various SPEI time scales, the extent of the affected area, and the average SPEI intensity within that area. The results highlight that the 2023/24 drought episode was the most intense ever recorded in the AB, with over 90% (80%) of the region affected for the month of January for SPEI-6 (SPEI-12), surpassing known past mega-events, such as the 2005, 2010 and 2015/16 episodes. These extreme conditions were observed across all timespans. Specifically, for January 2024 under the SPEI-6 and for September 2024 under the SPEI-12, more than half of the AB was categorized as experiencing exceptional drought, as established by the 1st percentile of the SPEI distribution. Furthermore, the results underscore the persistence of consecutive periods of drought, especially since the beginning of 2020. With the climate projections indicating continued warming in the region, increased evapotranspiration and lower rates of rainfall are expected, potentially leading to even drier periods. This marks the significance of studies focused on understanding the development and impacts of droughts, as they play a critical role in the mitigation of future environmental risks.

How to cite: Albuquerque, R., Monteiro dos Santos, D., Miranda, V., Gouveia, C., Liberato, M., Trigo, R., Peres, L., and Libonati, R.: Ranking of extreme drought events in the Amazon Basin between 1980 and 2024, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14311, https://doi.org/10.5194/egusphere-egu25-14311, 2025.

In territories with complex orography such as the Macaronesian archipelagos of Madeira, Azores, Canary Islands, and Cape Verde, the spatial resolution of the Coupled Model Intercomparison Project Phase 6 (CMIP6) is not sufficient to account for all the atmospheric phenomena that occur in these archipelagos with such a complex microclimatic structure. This research presents a climatic dataset at a spatial resolution of 3x3 km² in all the Macaronesian archipelagos derived from high-resolution regional climate simulations performed with Weather Research and Forecasting (WRF) model, applying the pseudo-global warming (PGW) method. The dataset is focused on the following parameters: temperature, precipitation, solar radiation, wind, and cloud coverage. Meteorological stations (ECAD and METAR) and reanalysis ERA5 data were used for the validation of the model results in the recent past period (1982–2019). We worked with two periods for future projections (2030–2059 and 2070–2099) under two representative scenarios (SSP2.6 and SSP8.5). These indicators include annual and seasonal statistics and variability for each parameter. The dataset aims to support regional climate adaptation strategies, contributing to the broader scientific understanding of climate in insular environments.

How to cite: Barrancos, J., Carrillo, J., Tondreau, P. S., Expósito, F. J., Pérez, J. C., González, A., and Díaz, J. P.: Pseudo–Global Warming climate projections at convection-permitting resolution in the Macaronesia region, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7204, https://doi.org/10.5194/egusphere-egu25-7204, 2025.

The Nearest Neighbour declustering technique is utilized to differentiate dependent events, such as aftershocks and foreshocks from independent events, such as isolated and mainshocks events . The estimated background field could either be stationary or non-stationary over time and may exhibit patterns that depend on both space and time. Any residual deviations from a time-stationary and spatiotemporally-independent Poisson point field could offer insights into regional loading processes and merit further investigation (Zaliapin and Ben-Zion, 2020). We apply the adopted technique on the Southern California region, an area that includes four significant events with magnitudes greater than 7, over the years 1981- 2021 and the catalog's completeness ranges between magnitudes 2 to 3 (Zaliapin and Benzion, 2020). For generating the complete background set, both outdegree and closeness centrality yielded nearly identical mainshock node counts for background detection in our study region, highlighting the robustness of these centrality measures.   In a tree network, hierarchy identification might not be straightforward, but utilizing centrality can aid in placing elements accurately. Higher centrality values indicate a simpler structure compared to lower centrality values. Although the traditional highest magnitude method produces results almost similar to those of the centrality measure from network analysis, the network-based approach offers new possibilities for future research in the study of earthquake sequences and their evolution. In a spatially inhomogeneous, temporally homogeneous Poisson process (SITHP), there is a strictly positive probability that two events may occur arbitrarily close to each other  and NN method works better for declustering with this condition (Luen and Stark, 2012). In this study, three temporal statistical tests have been conducted: the Conditional Chi square(CC) test, the Brown-Zhao(BZ) test, and the Kolmogorov smirnov (KS) test on the complete background set. It was found that the KS test, which assumes the time series follows a uniform distribution and does not require any adjusting parameters, is more reliable than the other two tests(requires more tuning constants). For almost all magnitude cut-offs, the temporal tests fail the null hypothesis; however, for a magnitude of 3.4, the temporal test is satisfied, but the space time test ( Luen and Stark test) fails the null hypothesis. For the nearest neighbour (NN) method, the null hypothesis is rejected for all magnitude ranges in our study region. Consequently, it can be concluded that NN declustering is not effective for this dataset or the number of data points is low. Notably, the Luen and Stark space time test yielded a value of 0 for most magnitudes, except for magnitudes 4 and 4.2. This suggests two potential scenarios: either the earthquakes are inadequately declustered, leading to some background events being overlooked or there is another possibility that this model is not fit for the Poisson process and suggesting a need for an alternate conditional model.

References:

Luen, B., & Stark, P. B. (2012). Poisson tests of declustered catalogues. Geophysical journal international, 189(1), 691-700. https://doi.org/10.1111/j.1365-246X.2012.05400.x.

Zaliapin, I., & Ben‐Zion, Y. (2020). Earthquake declustering using the nearest‐neighbour approach in space‐time‐magnitude domain. Journal of Geophysical Research: Solid Earth, 125(4), e2018JB017120. https://doi.org/10.1029/2018JB017120.

How to cite: seal, A. and Jana, N.:  Statistical Analysis on Background Seismicity of Southern California Region: Application of Nearest Neighbour Declustering and Network  Analysis, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-693, https://doi.org/10.5194/egusphere-egu25-693, 2025.