Votre recherche

Abstract Background Relating the geographical distribution of intermediate freshwater snail hosts (viz. vectors of schistosomes) to local environmental attributes offers value for understanding the epidemiological landscape of schistosomiasis transmission in a changing aquatic environment. Schistosomiasis—both urogenital and intestinal—causes significant human suffering, affecting approximately 240 million people globally and grouped within the neglected tropical disease (NTD) umbrella. This study addresses the following questions: 1. Where are the most suitable habitats for intermediate host snails in the Lower Shire Valley (LSV) in Malawi? 2. Which environmental factors are strongly associated with the geographical distribution of such snails in the LSV? Methods This paper presents the first species distribution models (SDMs) for intermediate snail hosts for urogenital and intestinal schistosomiasis in Chikwawa and Nsanje Districts, which together form the LSV). The SDMs developed for this study are ensemble machine learning approaches based on Random Forest (RF), Support Vector Machines (SVM), and multilayer perceptron (MLP) and are specific to the Bulinus africanus group and Biomphalaria pfeifferi . The former transmits urogenital schistosomiasis ( Schistosoma haematobium ), while the latter transmits intestinal schistosomiasis ( Schistosoma mansoni ). Results The SDMs reveal the following: 1) currently, Bu. africanus group not only has a wide distribution across central Chikwawa and eastern Nsanje but is also concentrated in floodplains, and the LSV has few habitats that can support Bi. pfeifferi , and 2) vegetation cover is the most important predictor of Bu. africanus group distribution, whereas precipitation variables are most important for Bi. pfeifferi in the LSV. Thus, Bu. africanus group habitat is the most dominant and abundant, while Bi. pfeifferi suitable habitat is patchy and scarce. Conclusion The distribution of suitable habitats for potential urogenital and intestinal schistosomiasis transmission across LSV is not uniform and typically non-overlapping. Understanding the spatial and temporal distributions of these snails is important for controlling and eliminating schistosomiasis. Graphical Abstract

Study region: This study aims at the Kunhar River Basin, Pakistan, that has been facing repeated flood occurrences on a recurring basis. As the flood susceptibility of this area is high, its topographic complexity demands correct predictive modeling for strategic flood planning. Study focus: We developed a system of flood susceptibility mapping based on Geographic Information Systems (GIS), Principal Component Analysis (PCA), and Support Vector Machine (SVM) classification. Four kernel functions were applied, and the highest-performing was the Radial Basis Function (SVM-RBF). The model was validated and trained using historical flood inventories, morphometric parameters, and hydrologic variables, and feature dimensionality was reduced via PCA for increased efficiency. New hydrological insights: The SVM-RBF model recorded an AUC of 0.8341, 88.02% success, 84.97% predictability, 0.89 Kappa value, and F1-score of 0.86, all of which indicated high predictability. Error analysis yielded a PBIAS of +2.14%, indicating negligible overestimation bias but within limits acceptable in hydrological modeling. The results support the superiority of the SVM-RBF approach compared to conventional bivariate methods in modeling flood susceptibility over the complex terrain of mountains. The results can be applied in guiding evidence-based flood mitigation, land-use planning, and adaptive management in the Kunhar River Basin. © 2025 The Author(s)

Urban flooding threatens Indian cities and is made worse by rapid urbanization, climate change and poor infrastructure. Severe flooding occurred in cities such as Mumbai, Chennai and Ahmedabad. This has caused huge economic losses and displacement. This study addresses the limitations of traditional flood forecasting methods. It has to contend with the complex dynamics of urban flooding. We offer a deep learning approach which uses the network Convolutional Neural Networks (CNN) and Long Short-Term Memory (LSTM) networks to improve flood risk prediction. Our CNN-LSTM model combines spatial data (water table, topography) and temporal data (historical model) to classify flood risk as low or high. This method includes collecting data pre-processing (MinMaxScaler, LabelEncoder) Modeling, Training and Evaluation. The results demonstrate the accuracy of flood risk predictions and provide insights into flexible strategies for urban flood management. This research highlights the role of data-driven approaches in improving urban planning to reduce flood risk in high-risk areas. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2026.

Flood risk assessment is an effective tool for disaster prevention and mitigation. As land use is a key factor influencing flood disasters, studying the impact of different land use patterns on flood risk is crucial. This study evaluates flood risk in the Chang-Zhu-Tan (CZT) urban agglomeration by selecting 17 socioeconomic and natural environmental factors within a risk assessment framework encompassing hazard, exposure, vulnerability, and resilience. Additionally, the Patch-Generating Land Use Simulation (PLUS) and multilayer perceptron (MLP)/Bayesian network (BN) models were coupled to predict flood risks under three future land use scenarios: natural development, urban construction, and ecological protection. This integrated modeling framework combines MLP’s high-precision nonlinear fitting with BN’s probabilistic inference, effectively mitigating prediction uncertainty in traditional single-model approaches while preserving predictive accuracy and enhancing causal interpretability. The results indicate that high-risk flood zones are predominantly concentrated along the Xiang River, while medium-high- and medium-risk areas are mainly distributed on the periphery of high-risk zones, exhibiting a gradient decline. Low-risk areas are scattered in mountainous regions far from socioeconomic activities. Simulating future land use using the PLUS model with a Kappa coefficient of 0.78 and an overall accuracy of 0.87. Under all future scenarios, cropland decreases while construction land increases. Forestland decreases in all scenarios except for ecological protection, where it expands. In future risk predictions, the MLP model achieved a high accuracy of 97.83%, while the BN model reached 87.14%. Both models consistently indicated that the flood risk was minimized under the ecological protection scenario and maximized under the urban construction scenario. Therefore, adopting ecological protection measures can effectively mitigate flood risks, offering valuable guidance for future disaster prevention and mitigation strategies. © 2025 by the authors.

This study presents a novel bivariate-output deep learning framework based on LeNet-5 for the simultaneous prediction of urban flood susceptibility and inundation depth in Seoul, South Korea. Unlike previous studies that relied on single-output models, the proposed approach jointly learns classification and regression targets through a shared feature extraction structure, enhancing consistency and generalization. Among six tested architectures, the Le5SD_CBAM model—integrating a Convolutional Block Attention Module (CBAM)—achieved the best performance, with 83% accuracy, an Area Under the ROC Curve (AUC) of 0.91 for flood susceptibility classification, and a mean absolute error (MAE) of 0.12 m and root mean squared error (RMSE) of 0.18 m for depth estimation. The model’s spatial predictions aligned well with hydrological principles and past flood records, accurately identifying low-lying flood-prone zones and capturing localized inundation patterns influenced by infrastructure and micro-topography. Importantly, it detected spatial mismatches between susceptibility and depth, demonstrating the benefit of joint modeling. Variable importance analysis highlighted elevation as the dominant predictor, while distances to roads, rivers, and drainage systems were also key contributors. In contrast, secondary terrain attributes had limited influence, indicating that urban infrastructure has significantly altered natural flood flow dynamics. Although the model lacks dynamic forcings such as rainfall and upstream inflows, it remains a valuable tool for flood risk mapping in data-scarce settings. The bivariate-output framework improves computational efficiency and internal coherence compared to separate single-task models, supporting its integration into urban flood management and planning systems. © 2025 by the authors.

Study region: Shanghai, China Study focus: This paper proposes a comprehensive framework for quantifying storm surge floods in coastal cities by incorporating the influences of both climate change and urbanization. The framework achieves a physically process-based numerical simulation of storm surge-induced flood hazards due to tropical cyclones in coastal cities by coupling the fast flood inundation model (SFINCS) and the land use change model (GeoSOS-FLUS), along with the numerical nested model for storm surges (Delft 3D Flow & Wave). Using a 1000-year tropical cyclone simulated by the STORM model as an example, this study analyzes and maps coastal flood impacts under the moderate climate scenario (SSPs245) and high emission scenario (SSPs585), and also evaluates the impact of land use changes on these scenarios. New hydrological insights for the region: Taking Shanghai, China as an example, the results show that by 2100, urban land use changes will lead to an increase in the extent of 1000-year TC flooding areas by 4.91–34.00 %, underestimating the inundation area of storm surges if future urban land use changes are not considered. Additionally, our predictions indicate the vulnerability of Chongming island and Changxing island to the impacts of climate change, despite the protective role of coastal embankments considered in the tropical cyclone storm surge simulation. The results of this study represent an important contribution to a better understanding of how future urban land use changes will affect storm surge flooding risks in and around Shanghai. The proposed methodology can be applied to coastal areas worldwide that are vulnerable to tropical cyclones, aiding in the formulation of hazard mitigation policies to alleviate flood impacts in these regions. © 2025 The Authors

Abstract. Developing predictions of coastal flooding risk on subseasonal timescales (2–6 weeks in advance) is an emerging priority for the National Oceanic and Atmospheric Administration (NOAA). In this study, we assess the ability of two current operational forecast systems, the European Centre for Medium-Range Weather Forecasts Integrated Forecasting System (IFS) and the Centre National de Recherches Météorologiques climate model (CNRM), to make subseasonal ensemble predictions of the non-tidal residual component of coastal water levels at United States coastal gauge stations for the period 2000–2019. These models were chosen because they assimilate satellite altimetry at forecast initialization and attempt to predict the mean sea level, including a global mean component whose absence in other forecast systems complicates assessment of tide gauge reforecast skill. Both forecast systems have skill that exceeds damped persistence for forecast leads through 2–3 weeks, with IFS skill exceeding damped persistence for leads up to 6 weeks. Post-processing forecasts to include the inverse barometer effect, derived from mean sea level pressure forecasts, improves skill for relatively short forecast leads (1–3 weeks). Accounting for vertical land motion of each gauge primarily improves skill for longer leads (3–6 weeks), especially for the Alaskan and Gulf coasts; sea-level trends contribute to reforecast skill for both model and persistence forecasts, primarily for the East and Gulf coasts. Overall, we find that current forecast systems have sufficiently high levels of deterministic and probabilistic skill to be used in support of operational coastal flood guidance on subseasonal timescales.

This is the second part of a two-part study that investigates an extreme rainfall event that occurred from 8 to 12 December 2018 over central Vietnam (referred to as the D18 event). In this part, the study aims to evaluate the practical predictability of the D18 event using the quantitative precipitation forecasts (QPFs) from a time-lagged cloud-resolving ensemble system. To do this, 29 time-lagged (8 d in forecast range) high-resolution (2.5 km) members were run, with the first member initialized at 12:00 UTC on 3 December and the last one at 12:00 UTC on 10 December 2018. Between the first and the last members are multiple members that were executed every 6 h. The evaluation results reveal that the cloud-resolving model (CReSS) predicted the rainfall fields in the short range (less than 3 d) well for 10 December (the rainiest day). Particularly, the CReSS model shows high skill in heavy-rainfall QPFs for this date with a similarity skill score (SSS) greater than 0.5 for both the last five members and the last nine members. The good results are due to the model having good predictions of relevant meteorological variables such as surface winds. However, the predictive skill is reduced at lead times longer than 3 d, and it is challenging to achieve good QPFs for rainfall thresholds greater than 100 mm at lead times longer than 6 d. These results also confirmed our scientific hypothesis that the cloud-resolving time-lagged ensemble system (using the CReSS model) improved the QPFs of this event in the short range. Furthermore, the results also demonstrated that a decent QPF can be made at a longer lead time (by a member initialized at 18:00 UTC on 4 December). In addition, the ensemble-based sensitivity analysis (ESA) of 24 h rainfall in central Vietnam shows that it is highly sensitive to initial conditions, not only at lower levels but also at upper levels. The rainfall is sensitive to both kinematics and moisture convergence at low levels, and such sensitivities decrease with increasing lead time. The ESA also facilitates a better understanding of the mechanisms in the D18 event, implying that it is meaningful to apply ESA to control initial conditions in the future. © Author(s) 2025.

Flooding is an escalating hazard in arid and rapidly urbanizing environments such as Jeddah, Saudi Arabia, where the lack of historical flood records and sparse monitoring systems challenge effective risk prediction. To address this gap, this study aims to develop an accurate and interpretable flood susceptibility-mapping framework tailored to data-scarce urban settings. The research integrates a stacked ensemble model—comprising machine learning: XGBoost, CatBoost, and Histogram-based Gradient Boosting (HGB)—with SHapley Additive exPlanations (SHAP) to enhance prediction accuracy and model transparency. Random Forest was excluded from the final model stack due to inferior classification performance. A diverse set of geospatial inputs, including digital elevation model, slope, flow direction, Curve Number, topographic indices, and LULC (from ESRI Sentinel-2) were used as predictors. Furthermore, 92 and 198 flooded and non-flooded points were used for model validation. The model achieved strong predictive performance (AUC = 0.92, Accuracy = 0.82) on the validation set. In the absence of official flood records, model outputs were intersected with road network data to identify 395 road points in highly susceptible zones. Although these points do not represent a formal validation dataset—due to the general lack of detailed flood event records in the region, particularly in relation to infrastructure—they provide a valuable proxy for identifying flood-prone road segments. SHAP explainability analysis revealed that TRI, TPI, and distance to rivers were the most globally influential features, while Curve Number and LULC were key drivers of high-risk predictions. The model mapped 139 km2 (8.7 %) of the area as very high flood susceptibility and 325 km2 (20.3 %) as high susceptibility, outperforming individual learners. These results confirm that stacked ensemble learning, paired with explainable AI and creative validation strategies, can produce reliable flood susceptibility maps even in data-constrained contexts. This framework offers a transferable and scalable solution for flood risk assessment in similar arid and urbanizing environments. © 2025 Elsevier Ltd

With global warming, the hydrological cycle is intensifying with more frequent and severe droughts and floods, placing water resources and their dependent communities under increasing stress. Guidance and insights into the projection of future water conditions are, therefore, increasingly needed to inform climate change adaptation. Hydrological projections can provide such insights when suitably designed for user needs, produced from the best available climate knowledge, and leverage appropriate hydrological models. However, producing such hydrological projections is a complex process that requires skills and knowledge spanning from the often-siloed disciplines of climate, hydrology, communication, and decision-making. Groundwater projections are still underrepresented compared to surface water projections, despite the importance of groundwater to sustain society and the environment. Accordingly, this paper bridges these silos and fills a gap by providing detailed guidance on the important steps and best practices to develop groundwater-inclusive hydrological projections that can effectively support decision-making. Using an extensive literature review and our practical experience as climate scientists, hydro(geo)logists, numerical modelers, uncertainty experts and decision-makers, here we provide: (a) an overview of climate change hydrological impacts as background knowledge; (b) a step-by-step guide to produce groundwater-inclusive hydrological projections under climate change, targeted to both scientists and water practitioners; (c) a summary of important considerations related to hydrological projection uncertainty; and (d) insights to use hydrological projections and their associated uncertainty for impactful communication and decision-making. By providing this practical guide, our paper addresses a critical interdisciplinary knowledge gap and supports enhanced decision-making and resilience to climate change threats. © 2025 Commonwealth of Australia. Earth Science New Zealand. Acclimatised Pty Ltd and The Author(s). Earth's Future published by Wiley Periodicals LLC on behalf of American Geophysical Union.

Named Data Networking (NDN) represents a promising Information-Centric Networking architecture that addresses limitations of traditional host-centric Internet protocols by emphasizing content names rather than host addresses for communication. While NDN offers advantages in content distribution, mobility support, and built-in security features, its stateful forwarding plane introduces significant vulnerabilities, particularly Interest Flooding Attacks (IFAs). These IFA attacks exploit the Pending Interest Table (PIT) by injecting malicious interest packets for non-existent or unsatisfiable content, leading to resource exhaustion and denial-of-service attacks against legitimate users. This survey examines research advances in IFA detection and mitigation from 2013 to 2024, analyzing seven relevant published detection and mitigation strategies to provide current insights into this evolving security challenge. We establish a taxonomy of attack variants, including Fake Interest, Unsatisfiable Interest, Interest Loop, and Collusive models, while examining their operational characteristics and network performance impacts. Our analysis categorizes defense mechanisms into five primary approaches: rate-limiting strategies, PIT management techniques, machine learning and artificial intelligence methods, reputation-based systems, and blockchain-enabled solutions. These approaches are evaluated for their effectiveness, computational requirements, and deployment feasibility. The survey extends to domain-specific implementations in resource-constrained environments, examining adaptations for Internet of Things deployments, wireless sensor networks, and high-mobility vehicular scenarios. Five critical research directions are proposed: adaptive defense mechanisms against sophisticated attackers, privacy-preserving detection techniques, real-time optimization for edge computing environments, standardized evaluation frameworks, and hybrid approaches combining multiple mitigation strategies. © 2025 by the authors.

A systems-oriented review of machine learning (ML) over the entire flood management spectrum, encompassing fluvial flood control, pluvial flood management, and resiliency-risk characterization was undertaken. Deep learners like long short-term memory (LSTM) networks perform well in predicting reservoir inflows and outflows. Convolution neural networks (CNNs) and other object identification algorithms are being explored in assessing levee and flood wall failures. The use of ML methods in pump station operations is limited due to lack of public-domain datasets. Reinforcement learning (RL) has shown promise in controlling low-impact development (LID) systems for pluvial flood management. Resiliency is defined in terms of the vulnerability of a community to floods. Multi-criteria decision making (MCDM) and unsupervised ML methods are used to capture vulnerability. Supervised learning is used to model flooding hazards. Conventional approaches perform better than deep learners and ensemble methods for modeling flood hazards due to paucity of data and large inter-model predictive variability. Advances in satellite-based, drone-facilitated data collection and Internet of Things (IoT)-based low-cost sensors offer new research avenues to explore. Transfer learning at ungauged basins holds promise but is largely unexplored. Explainable artificial intelligence (XAI) is seeing increased use and helps the transition of ML models from black-box forecasters to knowledge-enhancing predictors. © 2025 by the authors.

Water risk management has been adversely affected by climate variations, including recent climate change. Climate variations have highly impacted the hydrological cycles in the atmosphere and biosphere, and their impact can be defined with the teleconnection between climate signals and hydrological variables. Water managers should practice future risk management to mitigate risks, including the impact of teleconnection, and stochastically simulated scenarios can be employed as an effective tool to take advantage of water management preparation. A stochastic simulation model for hydrological variables teleconnected with climate signals is very useful for water managers. Therefore, the objective of the current study was to develop a novel stochastic simulation model for the simulation of synthetic series teleconnected with climate signals. By jointly decomposing the hydrological variables and a climate signal with bivariate empirical mode decomposition (BEMD), the bivariate nonstationary oscillation resampling (B-NSOR) model was applied to the significant components. The remaining components were simulated with the newly developed method of climate signal-led K-nearest neighbor-based local linear regression (CKLR). This entire approach is referred to as the climate signal-led hydrologic stochastic simulation (CSHS) model. The key statistics were estimated from the 200 simulated series and compared with the observed data, and the results showed that the CSHS model could reproduce the key statistics including extremes while the SML model showed slight underestimation in the skewness and maximum values. Additionally, the observed long-term variability of hydrological variables was reproduced well with the CSHS model by analyzing drought statistics. Moreover, the Hurst coefficient with slightly higher than 0.8 was fairly preserved by the CSHS model while the SML model is underestimated as 0.75. The overall results demonstrate that the proposed CSHS model outperformed the existing shifting mean level (SML) model, which has been used to simulate hydroclimatological variables. Future projections until 2100 were obtained with the CSHS model. The overall results indicated that the proposed CSHS model could represent a reasonable alternative to teleconnect climate signals with hydrological variables.

Atmospheric methane (CH4) concentrations have increased to 2.5 times their pre-industrial levels, with a marked acceleration in recent decades. CH4 is responsible for approximately 30% of the global temperature rise since the Industrial Revolution. This growing concentration contributes to environmental degradation, including ocean acidification, accelerated climate change, and a rise in natural disasters. The column-averaged dry-air mole fraction of methane (XCH4) is a crucial indicator for assessing atmospheric CH4 levels. In this study, the Sentinel-5P TROPOMI instrument was employed to monitor, map, and estimate CH4 concentrations on both regional and global scales. However, TROPOMI data exhibits limitations such as spatial gaps and relatively coarse resolution, particularly at regional scales or over small areas. To mitigate these limitations, a novel Convolutional Neural Network Autoencoder (CNN-AE) model was developed. Validation was performed using the Total Carbon Column Observing Network (TCCON), providing a benchmark for evaluating the accuracy of various interpolation and prediction models. The CNN-AE model demonstrated the highest accuracy in regional-scale analysis, achieving a Mean Absolute Error (MAE) of 28.48 ppb and a Root Mean Square Error (RMSE) of 30.07 ppb. This was followed by the Random Forest (RF) regressor (MAE: 29.07 ppb; RMSE: 36.89 ppb), GridData Nearest Neighbor Interpolator (NNI) (MAE: 30.06 ppb; RMSE: 32.14 ppb), and the Radial Basis Function (RBF) Interpolator (MAE: 80.23 ppb; RMSE: 90.54 ppb). On a global scale, the CNN-AE again outperformed other methods, yielding the lowest MAE and RMSE (19.78 and 24.7 ppb, respectively), followed by RF (21.46 and 27.23 ppb), GridData NNI (25.3 and 32.62 ppb), and RBF (43.08 and 54.93 ppb).

Floods have major impacts on the Mediterranean region, but little is currently known about their potential evolution in the context of climate change. This is due in particular to the limited ability of climate models to reproduce extreme meteorological events such as heavy rains that lead to flash floods, especially at the local scale over smaller basins. This study is the first to explore future flood scenarios over 12 Mediterranean basins using an ensemble of 12 high-resolution convection-permitting climate models and the GR5H hourly rainfall-runoff model. The results indicate an overall increase in flood intensity across all basins, particularly for the most severe events, but also a strong spatial variability in the change signal depending on the geographic location. There is good agreement among the convection-permitting climate models on an increase in hourly and daily rainfall extremes in the Mediterranean, but these changes are not strongly correlated with changes in flood-peak intensity, indicating that change in rainfall intensity alone is a poor predictor of future flood hazards. At present, this type of analysis is hampered by the short duration of the available high-resolution climate simulations. Longer timeseries would be required to better assess the robustness of the projected changes against climate variability.

The increasing threats of global flood risk mandate rapid and accurate high-resolution flood modeling strategies over large scales. In the United States, the National Oceanic and Atmospheric Administration (NOAA) Office of Water Prediction (OWP) has operationalised a Flood Inundation Mapping (FIM) framework utilising the Height Above Nearest Drainage (HAND)-Synthetic Rating Curve (SRC) approach. It translates streamflow into stage and subsequently maps the inundation over the floodplain. It is a low-fidelity FIM framework, suitable for large-scale applications with much less computational effort. The SRCs are calculated for each river segment using Manning's equation; however, uncertainty in Manning's parameters and missing bathymetry impart bias in SRC calculation, and thus in FIM. An SRC adjustment factor (λsrc), introduced by OWP, calibrates SRCs against USGS rating curves, HEC-RAS 1D rating curves, and National Weather Service (NWS)-Categorical Flood Inundation Mapping (CatFIM) locations. Adjusted SRCs improve the FIM predictions but are limited to locations with the above data sources. In this paper, we develop machine learning models to predict the λsrc over the entire United States river network. Results show that the eXtreme Gradient Boosting model yielded the strongest predictability, with an R2 of 0.70. The impact of λsrc on FIM predictions is evaluated for Hurricane Matthew in North Carolina and synthetic flood events in 15 watersheds. For Hurricane Matthew flooding, the mean percentage improvements in Critical Success Index (CSI), Probability of Detection (POD), and F1 Score are 17.5%, 20% and 12.5%, while for synthetic events, the improvements are 2.59%, 4.93%, and 3.03%, respectively. © 2025 The Author(s)

The preparation of accurate multi-hazard susceptibility maps is essential to effective disaster risk management. Past studies have relied mainly on traditional machine learning models, but these models do not perform well for complex spatial patterns. To address this gap, this study uses two meta-heuristic algorithms (Genetic Algorithm (GA) and Particle Swarm Optimization (PSO)) to provide an optimized Random Forest (RF) model with better predictive ability. We focus on four significant hazards—landslides, land subsidence, wildfires, and floods—in Kurdistan Province, Iran, using Sentinel-1 and Sentinel-2 satellite imagery collected between 2015 and 2022. Furthermore, two models of RF-GA and RF-PSO were utilized to create multi-hazard susceptibility, which were evaluated using receiver operating characteristic (ROC) curves and area under the curve (AUC). The RF-GA algorithm achieved 91.1% accuracy for flood hazards, 83.8% for wildfires, and 99.1% for landslide hazards. In contrast, utilizing RF-PSO resulted in a 95.9% accuracy for land subsidence hazards. The combined RF-GA algorithm demonstrated superior accuracy to individual RF modeling techniques. Furthermore, eastern regions are more prone to floods and land subsidence, whereas western areas face more significant risks from landslides and wildfires. Additionally, floods and land subsidence exhibit a considerable correlation, impacting each other’s occurrence, while wildfires and landslides demonstrate interacting dynamics, influencing each other’s likelihood of occurrence. © The Author(s) 2025.

Floods can cause varying degrees of damage to village houses, and rapid assessment of houses damage is crucial for post-disaster safety evaluations. To address this issue, the paper proposes a DeepLabv3-Dual Attention (DDA) model incorporating improved ResNet-50, Atrous Spatial Pyramid Pooling (ASPP), Selective Kernel(SK)-Attention, and Self-Attention mechanisms. In the DDA model, the ASPP structure is enhanced with Self-Attention, while the backbone network employs SK-Attention improved ResNet-50, enabling more effective feature extraction across both spatial and channel dimensions. Validated on the “7.20” heavy rain disaster dataset from Zhengzhou, DDA achieves optimal performance (learning rate: 0.01, batch_size: 16, epoch: 88) with 87.5% accuracy, 98.5% global accuracy, 77.8% MIoU, and 86.5% F1-score, outperforming DeepLabv3+, Swin-Unet, U-Net, YOLO baseline models. Ablation studies confirm dual-attention synergy improves MIoU by 2.5%, precision by 2.5% over single mechanisms. Further, a damage quantification framework is developed using OpenCV and Canny edge detection to extract contours, coupled with a maximum tangent circle algorithm for pixel-level width measurement and skeletonization-based pixel-level length calculation. The results of damage quantification model show that the proportion of images with damage width quantification error ⩽ 20% are 85.29%. This study serves as a reference for intelligent assessment and risk classification of structural damage to buildings affected by flooding. © 2025 Elsevier Ltd