Winter windstorms are the most damaging natural hazard in Europe in terms of insured losses, with impacts often arising from storm clusters rather than isolated events. In reinsurance practice, losses are aggregated over sequences of storms affecting the same region within a limited time window. Yet, attributing individual damages to specific storm events remains challenging. The distribution of costs between insurance and reinsurance companies critically depends on this attribution, making robust and transparent criteria essential for a fair allocation of losses. This study introduces a method to systematically link individual insurance claims to extra-tropical cyclones, enabling event-based attribution of damages. The method is applied to the Generali France loss portfolio to build a catalogue linking individual claims to storm events. The resulting catalogue provides a foundation for risk assessment, loss modelling, and reinsurance applications. We focus on storm clusters, defined as successive storms affecting the same region within a 96 h period. We show that damaging storms within clusters are more intense than isolated events, with lower minimum sea-level pressure and higher vorticity. Losses within clusters are dominated by a single storm, accounting on average for about 70 %%
Forest ecosystems are increasingly threatened by disturbances such as fires, droughts, storms, and insect and pathogen outbreaks. Accurate and timely disturbance mapping is essential for understanding their dynamics and informing mitigation strategies to combat widespread forest decline. Traditional inventories, such as the U.S. Forest Service’s Insect and Disease Survey (IDS), provide detailed information on biotic and abiotic disturbances; however, they have varying coverage and inherent uncertainties in the location, extent, and timing of disturbances due to data-collection constraints. Other approaches, such as satellite remote sensing, can, in principle, overcome some of these challenges by providing large-scale coverage and continuous spatio-temporal observations. However, robust disturbance classification algorithms need to be developed, which in turn require good-quality labels.We present a novel approach for refining disturbance classification labels by combining IDS with Sentinel-1 radar backscatter change detection to produce a new reference dataset, Sentinel-1 Disturbance Mapping (S1DM). The disturbed patches identified by Sentinel-1 are typically located within 200–330 m of IDS locations and generally agree on disturbance timing. Sentinel-1 tends to detect bark beetle disturbances up to 2 years earlier than IDS, and some defoliator events are also detected 1 year prior to IDS. When statistically examined against manual labels from high-resolution images from PlanetScope, we found that S1DM performed better than IDS for wind and bark beetle disturbances, but not for defoliators. For bark beetle disturbances, S1DM improves the median Intersection over Union from 0.007 to 0.076, an absolute gain of 6.9 % over IDS.By integrating spatial and temporal information on disturbance occurrence from Sentinel-1 change detection with information on the corresponding disturbance agent from IDS, S1DM provides a high-quality forest disturbance reference dataset for developing remote sensing forest classification models.Our approach highlights the benefits of combining satellite-based remote sensing with traditional aerial survey data, reducing aerial survey costs while providing a scalable method adaptable to various regions.
In this study, we extend the application of the Maximum Entropy model (MaxEnt), traditionally applied to ecological research and less explored in natural hazard studies, to a novel context by characterising a multi-hazard scenario (i.e., flood-triggered sinkholes) in the Orléans karst region (Val d'Orléans) of France. Many regions of the world exhibit complex hazard landscapes where networks of multi-hazard interrelationships (cascades) pose challenges due to the potential interactions between hazards and the different temporal and spatial scales of hazard events. While mountainous, coastal and volcanic regions have been recognised as multi-hazard forming zones, karst terrains have received little attention despite being prone to multi-hazard events due to their distinct geology, geomorphology, hydrogeology and other environmental characteristics. Incorporating karst-specific multi-hazard scenarios into resilience planning processes supports disaster risk reduction efforts by raising the awareness of citizens, protecting elements at risk and facilitating decisions on disaster prevention. To support this aim, we developed a multi-hazard susceptibility map for the karst region of Val d'Orléans that characterises flood-triggered sinkholes. We applied MaxEnt, a machine learning method, to forecast the spatial probability distribution of flood-triggered sinkholes. Model inputs included the location of past sinkhole occurrences and geo-environmental factors contributing to sinkhole formation (e.g., topography, local geology, hydrology and flood hazard). We validated the performance of the model by initially using 70 % of the sinkhole inventory data and keeping the remaining 30 % for testing. This validation process assessed the model's performance using the Area Under the Curve of the Receiver Operating Characteristic (AUC-ROC). The resulting map reveals areas located up to 1 km south of the Loire River and areas with lowest elevation with highest susceptibility to flood-triggered sinkholes. We conclude that our approach to producing this type of multi-hazard scenario and map is useful for identifying flood-triggered sinkholes in Val d'Orléans and other karst areas around the globe, supporting effective land use planning.
This study presents a high-resolution framework for assessing climate-related risk at the building scale by implementing the IPCC risk concept, defining risk as a function of vulnerability, exposure and hazard. The framework focuses on pluvial flood risk related to people's well-being and mobility. Hazard is driven by a 100-year rainfall event (36 mm h−1), modelled with a hydrodynamic flood simulation incorporating topography, drainage capacity, and land use. Exposure is differentiated by impact type, considering residents on ground floors for well-being and building proximity to flooded streets for mobility and accessibility. Social vulnerability is quantified using socioeconomic indicators such as age, income, and education. The framework is demonstrated using empirical data from Hamburg, Germany, identifying risk hotspots where high social vulnerability coincides with elevated flood exposure. To support practical implementation, we introduce a Python-based ArcGIS pluvial flood risk toolbox that enables automated, building-level risk mapping. The transparent and flexible design makes the framework transferable to other cities, supporting climate adaptation planning and risk-informed decision-making.
Debris flows are frequently triggered by rainstorms after wildfires and pose severe threats to downstream residents and buildings in mountainous regions. However, there has been limited focus on developing a comprehensive framework to assess the physical vulnerability of buildings to postfire debris flows. This study presents a quantitative approach for establishing a physical vulnerability model based on observed building damage and simulated debris flow intensities. Detailed field surveys established a building damage database in Kule village, Yajiang County. Numerical simulations using the FLO-2D model were performed to reproduce the debris flow process and quantify the debris flow intensity, including the flow depth, flow velocity, impact pressure, momentum flux, overturning moment, and relative burial height. Physical vulnerability curves were developed for brick-concrete buildings and compared with those obtained in previous studies, and the differences in vulnerability curves, intensity indicators, and functional models were examined. The results revealed that the lognormal cumulative distribution function (LNCDF) model achieved the best performance, with relative error less than 10 % and prediction accuracy exceeding 85 %. Critical thresholds for complete building damage were identified as a flow depth of 2.5 m and impact pressure of 25 kPa. The momentum flux demonstrated greater sensitivity in distinguishing different damage categories, whereas the impact pressure provided more precise vulnerability index predictions. The proposed physical vulnerability model can evaluate the building structural resistance to debris flows in wildfire-affected areas, providing a systematic foundation for risk management and mitigation strategies.
Pluvial (flash) floods frequently cause damage in rural and urban watersheds as a result of short-term, intense local precipitation events that cause infiltration excess runoff and overland flow. Unlike fluvial floods, pluvial floods are primarily characterized by surface runoff and flow in small ditches and creeks, making them unsuitable for evaluation using common extreme value statistics based on long-term river discharge data. Precipitation statistics alone are insufficient for predicting pluvial floods because these floods are also influenced by hydrological and hydrodynamic processes. We propose a new regional-scale pluvial flood index (PFI) that considers precipitation as well as hydrological and hydrodynamic processes to assess the hazard of surface flooding. The PFI is based on local pluvial flood hazard areas (PFHA), which are defined as areas where water depth, flow velocity, or both exceed thresholds that endanger pedestrians and vehicles. We defined four PFI classes based on historical and design events, ranging from no hazard to very large flood hazard. The PFI serves as a simple, dimensionless measure and information tool to support regional to local scale pluvial flood management.PFHA and PFI were calculated for various events using radar-based precipitation input, dynamic simulations of infiltration and saturation excess, and hydrodynamic simulations of surface runoff. PFI forecasting requires quantitative precipitation data as well as appropriate processed-based distributed hydrodynamic and hydrological models at large temporal and spatial scales. We demonstrate the PFI's applicability and utility by creating large-scale flash flood hazard maps and hindcasting an extreme historical event. Furthermore, the PFI can link to detailed local flash flood hazard information, assisting municipal decision-making. It can also be a key component in operational pluvial flood warning systems, providing information on the occurrence and severity of floods on a scale of several hectares to square kilometres. This educates stakeholders and the community, improving real-time warning systems, preparedness, and planning decisions.
The assessment of flood-related losses to cultural heritage (CH) remains one of the most underexplored areas in flood risk management, largely due to the complexity of CH assets and their multiple, often intangible, values. In this study, extensive field data collection and archival research on artwork restoration costs were undertaken to support a synthetic approach for developing vulnerability models for both the CH buildings and artworks across three primary CH asset types: places of worship, museums, and libraries/archives. The methodology was applied to the historic city of Florence (Italy), enabling the derivation of mean and percentile vulnerability curves from a sample of 48 inspected CH buildings. For a 500-year flood scenario, estimated average losses amount to approximately EUR 2.5 million for the CH buildings and EUR 3 million for artworks per asset, with total damages to CH in the city reaching approximately EUR 527 million. While direct monetary loss estimates, i.e., restoration costs are subject to considerable uncertainty, the model results align well with available ex-post data, particularly for places of worship. These findings demonstrate that flood-related monetary losses to CH assets are far from negligible when compared to other damage categories, such as residential buildings, and therefore warrant increasing attention from the scientific and policy-making communities.
Estimating seismic hazard is crucial for enhancing societal resilience and risk mitigation strategies. Probabilistic Seismic Hazard Analysis (PSHA) is the current standard framework, traditionally relying on empirical earthquake rupture forecasts (ERFs) and ground-motion models. In this framework, physics-based earthquake cycle simulators are emerging as powerful tools in PSHA, capable of replicating observed seismicity and seismic hazard statistics. Here, we present a quantitative consistency evaluation of physics-based PSHA models at the Eastern Betic Shear Zone in south-eastern Spain against both historical macroseismic intensity data and instrumental ground shaking records. We use synthetic catalogues from RSQSim earthquake-cycle simulations to construct two physics-based ERFs that we pipeline into a PSHA calculation. Results indicate that the physics-based ERFs derived from the best-performing simulation model, previously ranked against empirical benchmarks, achieve the best overall agreement with observed macroseismic intensities and acceleration records at 10 sites, outperforming both the lower-performing simulation and a traditional area-source model. Our findings highlight that the incorporation of physics-based models into PSHA is appropriate, inherently enabling the inclusion of fault-system rupture complexity and interactions, all key challenges in PSHA. We also advocate for the hybridization of physics-based models with traditional approaches in PSHA to better capture epistemic uncertainties in the hazard representation.
In 2022, southwestern France experienced an exceptional wildfire season, recording a burned area 14 times higher than the 2006–2023 average. Here, we assess the rarity (return period) of the fire weather conditions observed in 2022 and how anthropogenic climate change (ACC) has already altered and will continue to alter the probability of fire weather conditions associated with the three largest wildfires (Landiras-1: 12 552 ha; Landiras-2: 7124 ha; La Teste-de-Buch: 5709 ha). Drawing from the daily Fire Weather Index (FWI) computed from two reanalysis datasets (1959–2023) and a nationwide wildfire record dataset (2006–2023), we first sought to quantify the rarity of those conditions across a range of spatial (local vs. regional) and temporal (fire duration vs. 30 d window) scales. Our results demonstrate that the rarity of FWI conditions is generally the highest at local and fire duration scales with the associated return periods increasing from 6 to 34 years, from 22 to 38 years, and from 6 to 101 years
Large rock slope instabilities develop over long periods and creep slowly over hundreds or thousands of years, until they undergo a “slow to fast” evolution towards catastrophic collapse. Capturing this transition is key to manage related risks, especially considering ongoing climate change scenarios and human activities, that are expected to strongly influence geohazards. However, this is a challenging task due to the complexity of the underlying processes. Long-term, area-wide monitoring of slope movements is essential to understand landslide dynamics and evolution. Despite being widely used for landslide investigations, C-band SAR interferometry datasets suffer from decorrelation in vegetated areas and fast movements, limiting displacement retrieval in alpine regions. Emerging L-band systems, with reduced temporal decorrelation, can complement higher-frequency data by enabling measurements also in vegetated areas and capturing larger displacements. This work aims at analysing the potential benefits and limitations of L-band SAR interferometry applied to alpine landslide monitoring and at understanding if these data can help in mitigating current shortcomings of C-band SAR interferometry. We explore three different scenarios of large alpine slope instabilities in the European Alps, that threaten important economic and societal assets. We perform site-specific analysis, validation and interpretation of L-band SAR interferometry products derived from ALOS-2 PALSAR-2 and SAOCOM-1 satellite imagery, as well as of terrestrial data acquired by the GAMMA L-band SAR (GLSAR) instrument. Our results highlight the contributions of L-band InSAR products to the practical characterisation and interpretation of large rock slope instabilities and provide important recommendations for the recently launched L-band satellite SAR missions ALOS-4 PALSAR-3 and NISAR, as well as for the future L-band satellite SAR mission ROSE-L.
Forecasting future vent opening position is fundamental for managing volcanic hazards, and is usually based on the spatial density of past vents or other crust weakness indicators. Here, a novel empirical approach inspired by dyke propagation models is applied to the Campi Flegrei caldera. Results show that dyke azimuthal direction and propagation length are statistically independent, that azimuth correlates with topographic peaks within 7 km from the caldera centre, and that propagation length exhibits two main peaks at 2 and 4 km. Based on these results, we develop two new vent opening probability maps with maxima well correlating with caldera's structure and recent seismicity.
Understanding the relationships between flood risk perception and flood behaviour is crucial for effective risk management and risk communication strategies, but quantitative research in this area remains challenging. Based on a survey of 1007 residents in four different localities of Chile exposed to river floods, this study builds and applies a framework for assessing flood risk perception and flood behaviour at the individual, household, neighbourhood, and municipality levels. Results show that almost all respondents were aware of flood risk. Economic and personal resources strongly influence worry and preparedness: households in better economic situations were less worried about floods, lower economic resources at the municipal and neighbourhood levels prompted households to adopt cautionary measures. Experiences where the flood passed outside the household increased worry and preparedness. Worry decreased as trust in the neighbours increased. Overall, worry and preparedness in the study area were intermediate, with an increasing dispersion from the municipality to the individual level. Increasing worry did not necessarily translate into higher preparedness. Municipalities exhibited different flood behaviours, and some neighbourhoods exhibited flood behaviours different to those of their municipalities, evidencing important differences across the analysed levels. The results suggest that risk communication and risk management strategies should be tailored to the needs of specific neighbourhoods exposed to flooding.
The risk posed by natural hazards has gained growing attention in recent decades, largely due to the intensification and recurrence of extreme events, with the climate crisis identified as the primary driver. Landslide risk is no exception, although its impacts are generally less evident than those of floods or, particularly, severe droughts. In both cases, urban expansion has further exacerbated the problem, especially since the mid-twentieth century in more developed regions. This residential growth often took place in poorly regulated settings, particularly during its early stages, leading to the occupation of areas that were environmentally, culturally, or from a landscape perspective unsuitable, and frequently exposed to natural hazards. In fact, the risk of landslides affecting buildings located on susceptible terrain can largely be attributed to ineffective land management, often resulting from the absence of specific regulations. This study introduces a set of risk indices that serve as objective tools for the dynamic assessment of landslide risk in extensive and spatially fragmented territories divided into local entities. Based on these indices, criteria are proposed to evaluate the degree of risk and the adequacy of its management within each local entity, considering the evolution of urban development. Finally, a classification system is presented that organizes all cases according to their severity, offering a decision-support tool for public authorities tasked with ensuring effective land management.
In mountainous regions, intense rainfall can trigger thousands of landslides within hours. The drivers that control the occurrence of such landslides, and the methods for predicting the zones susceptible to their initiation have been extensively studied. Yet, for many of the most severe disasters associated with these landslide events, the main impacts on local communities occurred far from the source areas where most modelling efforts are focused. Sediments mobilized high on slopes by rainfall-triggered landslides can be transported many kilometres downstream, causing significant impacts along their path, while also feeding river systems with large amounts of sediments and consequently increasing flood risks. Such chain of cascading hazards significantly increases the destructive potential of landslides as well as their impact zone. Effective risk mitigation must therefore address not just susceptibility to initiation but also landslide mobility and hazard interactions – yet such studies remain rare.With this work, we emphasize the importance of capturing what we refer to as the landslide–debris-rich flood continuum (landslide source, runout and related debris-rich floods) for accurate inventory, susceptibility and exposure mapping when landslide mobility is high – as it is often the case for extreme rainfall events. We apply this approach in two districts of eastern Zimbabwe (> 8000 km2), severely impacted by Cyclone Idai in March 2019. Using simple, replicable methods, we mapped over 14 000 (mostly) shallow landslides and 94 km2
Barrier islands are vulnerable to extreme storm events which cause erosion and deposition of sediment. These morphological changes pose risks to both the built environment and natural habitats but are also affected by them. This study investigates the morphological impacts of Hurricane Ian (2022) on two barrier islands along Florida's Gulf Coast: the urbanized Fort Myers Beach, and Lovers Key, a naturally vegetated island and State Park. Using high-resolution pre- and post-storm topo-bathymetric datasets, we quantify patterns of erosion, sediment deposition and dune crest change. In addition, we investigated the morphological response of the developed and natural barrier islands by integrating spatially varying land cover data into the numerical model XBeach. Results show that the built environment on Fort Myers Beach significantly affects sediment transport pathways, causing localized erosion and deposition patterns distinct from those observed on the vegetated Lovers Key Island where dune crest lowering, landward migration, and storm-induced breach were prominent. Model simulations that incorporated detailed spatial variability of vegetation and built environment, replicated observed morphological changes with reasonable Brier Skill Scores, including the location of breach formation. Sensitivity analyses demonstrated that relatively small changes in roughness coefficient, wave skewness and asymmetry factor, morphological acceleration factor, and boundary water levels influence erosion intensity and sediment deposition patterns. Additionally, introducing supplemental vegetation patches in the model showed less dune erosion on vegetated barrier island, indicating that revegetation of islands may be beneficial. The findings provide insights into the complex interplay between storm forcing, land cover variability, and barrier island morphodynamics, and emphasize the importance of incorporating detailed land use and vegetation data in morphodynamic models to better assess barrier island responses to future storms under evolving climatic conditions, ultimately aiding efforts to enhance coastal resilience and adaptive management.
Flash floods and landslides have caused severe economic damages and loss of life, especially in mountainous regions. To support effective risk management there is a growing interest in multi-hazard assessment. In this study a globally applicable Machine Learning (ML) Framework for landslide and flash flood susceptibility mapping was applied and evaluated in the Italian region Liguria that is frequently and severely impacted by both hazards. A relatively dense inventory of past events was constructed to facilitate the training of the ML Framework. The analysis revealed substantial similarities in the causative factors for the two hazards. There is a considerable area of Liguria susceptible to both hazards, although flash floods most often occur in river valleys whereas landslide susceptibility is also high in the upper courses of river catchments. We found a very high susceptibility along the coastline where many villages and cities are located. The unified framework allows for the integration of different hazard types under a consistent modelling structure. This enhances the comparability of results and supports the development of integrated mitigation strategies for any region of interest.
Tree falls along linear infrastructures and in particular powerlines pose a significant economic, safety and environmental challenge for the companies and institutions managing these infrastructures. The quick progression and affordability of remote sensing technologies such as drone-based inventories offers the opportunity to quickly and efficiently map individual trees along these infrastructures, enabling precise vegetation management to reduce risks. Here, we show how the hybrid empirical and mechanistic wind risk model ForestGALES can be applied to assess the vulnerability of individual trees to windfalls along selected powerlines in southern Norway. The validation dataset contained 180 recorded individual tree falls along powerlines from the winter 2020–2021. There was no major wind event recorded that winter. However, still, the ForestGALES model performed adequately, with an AUC (area under the curve) of 0.67. Combining the vulnerability index from ForestGALES with all other available tree and environmental variables in a machine learning model (extreme gradient boost algorithm) did however significantly improve the prediction performance. These results highlight how a combination of high-quality remote sensing data at the individual tree level can be utilized with ForestGALES and machine learning to provide managers with high-resolution vulnerability information for vegetation management.
Tsunamis are high-impact natural disasters capable of causing significant social, economic, and environmental losses. Despite advances in tsunami warning systems, accurately predicting tsunami occurrence remains a challenge due to the uncertainty associated with seismic rupture characteristics. This study develops a methodology that integrates historical earthquake records, numerical modelling and statistical analysis to derive a tsunami-occurrence criterion, expressed as a binary labelling threshold for identifying whether an earthquake generates a tsunami. As part of this methodology, a global simulation-based database (TSUSY Database) was constructed using earthquake focal mechanism data from the USGS database and validated against tsunami records from the NOAA catalogue, covering events from 1976 to 2023. Through numerical simulations, maximum wave heights were estimated for each event as the maximum value within the entire simulation domain, and used to define thresholds that label earthquakes as tsunamigenic or non-tsunamigenic, with the aim of balancing missed events and unnecessary alerts. By providing a simulation-based criterion for tsunami occurrence, the methodology supports the development of decision tools for real-time tsunami assessment and has been incorporated into an operational tsunami decision-support system that can assist Tsunami Warning Centres in their warning procedures.
Flood mapping and water depth estimation from Synthetic Aperture Radar (SAR) imagery are crucial for calibrating and validating hydraulic models. This study uses SAR imagery to evaluate various preprocessing (especially speckle noise reduction), flood mapping, and water depth estimation methods. The impact of the choice of method at different steps and its hyperparameters is studied by considering an ensemble of preprocessed images, flood maps, and water depth fields.The evaluation is conducted for two flood events on the Garonne River (France) in 2019 and 2021, using hydrodynamic simulations and in-situ observations as reference data. Results show that the speckle filtering method choice can significantly alter flood extent estimations with variations of several square kilometers. Additionally, the selection and tuning of flood mapping methods significantly affect performance. While supervised methods outperformed unsupervised ones, well-tuned unsupervised approaches (such as local thresholding or change detection) can achieve comparable results. The compounded uncertainty from preprocessing and flood mapping steps also introduces substantial variability in the water depth field estimates.This study highlights the importance of considering the entire processing pipeline, encompassing preprocessing, flood mapping, and water depth estimation methods and their associated hyperparameters. Rather than relying on a single configuration, adopting an ensemble approach and accounting for methodological uncertainty should be privileged. For flood mapping, the method choice has the most influence. For water depth estimation, the most influential processing step was the flood map input resulting from the flood mapping step and the hyperparameters of the methods.