Enhancing cross-regional transferability of super-resolution-based flood surrogate models for data-scarce catchments.

Flood maps are essential for disaster preparedness, insurance and climate adaptation, yet US official maps cover only one-third of river channels. Here we present a deep-learning framework that learns from existing records to produce a spatially complete 30-m flood hazard map for the contiguous US, filling unmapped regions and updating legacy extents. We estimate that the 2023 national database omits 11 million people and 4.1 million buildings in flood zones, adding 69% and 81% beyond the official baseline respectively. Crucially, our socioeconomic analysis of 917 cities reveals that entirely unmapped areas disproportionately harbor highly vulnerable demographics, particularly the elderly and children, exposing them to unrecognized risks and compounding existing inequalities. By solving the trilemma between accuracy, scale, and cost in national scale flood mapping, our open-source method broadens public access to risk information, paving the way for targeted and equitable climate adaptation.

The World Health Organisation (WHO) aims to eliminate malaria by 2030; yet, the disease remains endemic in Sub-Saharan Africa. Stagnant floodwaters provide ideal breeding grounds for mosquitoes. Previous estimates of potential malaria risk in flood zones have been limited due to insufficient large-scale geospatial data. Here, we integrate high-resolution flood maps (2000–2018) from the Global Flood Database, malaria incidence data from the Malaria Atlas Project, and geospatial population data across 492 flood-prone zones in 38 countries. We used a geospatial statistical models to assess malaria relative risk and drivers. We found that in East and West Africa, malaria relative risk is elevated in flood-prone regions compared to national baselines. We estimate that sim12 million individuals diagnosed with Plasmodium falciparum (Pf )were exposed to flooding events, representing one-third of the population affected by floods. Our analyses find that flood exposure is one of the main drivers of the malaria burden in flood zones. These findings identify critical malaria hotspots and key drivers in flood-prone zones, and can help inform WHO’s malaria eradication strategies by guiding policymakers on the geographic distribution of vulnerable areas.

Slovakia: Kysuca, Torysa, Topľa, and Gidra rivers Efficient and accurate prediction of fluvial floods is a necessary task for flood preparedness and risk reduction. This study focuses on predicting the extent of fluvial floods under three flood scenarios (Q 10 , Q 100 , Q 1000 ) using three machine learning models: random forest (RF), extreme gradient boosting (XGBoost), neural networks (NN) and one deep learning model (U-Net). The models were trained using official flood maps, created under the second cycle of EU Flood Directive (2007) implementation, and seven high-resolution physical-geographic and land cover predictors. Model performance was assessed using three metrics and training time. Overall, we studied four river sections in Slovakia (Kysuca, Gidra, Torysa, and Topľa) for model development, with three river sections being used for training and the remaining one for testing and calculating performance measures. Results suggest that transferability of the modeled fluvial flood extent on similarly long and large river sections provided the highest performance. This finding is demonstrated by lower and balanced numbers of FP and FN pixels, specifically, for the Torysa/Kysuca, Topľa/Kysuca or Kysuca/Torysa training/testing river sections. The RF and XGBoost models required the least time for training. The optimized U-Net model resulted in less training time than the NN model. The results have high potential for near real-time flood mapping or operational early warning. • Machine and deep learning were used to transfer fluvial flood extent among rivers. • Models were innovatively trained/tested on official flood maps and seven predictors. • RF, XGBoost, NN, U-Net models and Q 10 , Q 100 , Q 1000 flood scenarios were applied. • The highest performance was achieved on similarly long and large river sections.

Accurate, real-time flood mapping from synthetic aperture radar image is vital for disaster response but is hindered by a persistent trade-off between model accuracy and computational efficiency. This paper introduces FM-Mamba, a novel network that leverages a state space model to break this bottleneck. Its core innovation is an encoder built with non-causal Mamba blocks, which captures essential long-range spatial context with linear complexity, paired with a parameter-efficient decoder designed for precise boundary recovery. Evaluated on the Sen1Floods11 and S1GFloods benchmarks, FM-Mamba achieves leading segmentation accuracy, matching or exceeding state-of-the-art methods in F1-score and IoU. Crucially, it accomplishes this with only 3.93 million parameters and a drastically reduced computational footprint, demonstrating a superior balance of performance and efficiency that is ideal for operational, real-time flood mapping.

Emergent capabilities for remote coastline and flood mapping with ultra-high-resolution SAR: a case study in American Samoa

We present the Global Water and Flood Mapping System (GWFMS), an experimental web-based portal supported by NASA, designed to facilitate access to high-resolution surface water and related products derived from commercial satellite data. The system architecture is modular, enabling integration of additional commercial satellite data sources; its current configuration is optimized for PlanetScope optical imagery, which offers near-daily global coverage at high spatial resolution. This temporal frequency enables the detection and monitoring of ephemeral or small-scale flood events that are typically missed by lower-resolution or less frequently updated datasets. Products generated by GWFMS are disseminated through an on-demand online service, which can be activated for example, during emergency scenarios, such as developing flood events, to support rapid response. The current system configuration utilizes the PlanetScope Ortho Analytic 4B Surface Reflectance (SR) product from Planet Labs. Surface water is detected using the Normalized Difference Water Index (NDWI), in conjunction with a multi-Otsu thresholding algorithm to delineate water bodies. Validation of the water detection output against the Global Surface Water dataset demonstrates strong overall accuracy exceeding 90%, with a false positive rate of approximately 3.5%. Despite its effectiveness, the system adopts a conservative detection approach in areas characterized by complex terrain, cloud shadow contamination, or high turbidity, which may reduce sensitivity in these regions. Nevertheless, GWFMS significantly enhances the capacity to monitor surface water dynamics globally, offering a valuable complement to existing global water datasets-particularly for disaster response agencies and other stakeholders requiring timely, high-resolution hydrological information.

Pluvial floods can cause severe socio-economic impacts on coastal urban areas like the Metropolitan Area of Barcelona. This study combined the development of high-resolution flood maps, based on a large-scale coupled 1D/2D model and empirical functions, to quantify direct economic damage to buildings and determine risk to pedestrians and vehicles. Importantly, the flood model included a network of 36 municipalities and covered 636 km2. Three scenarios were considered: single-hazard (extreme precipitation), multi-hazard (coincident extreme precipitation and storm surge), and adaptation (implementation of resilience measures). In total, 20 rain events were applied for each scenario: 5 were historic design storms, while 15 considered the effect of climate change (60 simulations in total). By the end of the century, results show potential increases in expected annual damage of up to 36%, from €139.8 M to €190.3 M. Risk for pedestrians could increase by 25% (494 ha to 620 ha) and for vehicles by 26% (59 km to 75 km) in the T10 single-hazard scenario. In the multi-hazard case, the socio-economic impacts are approximately 5% higher, while the adaptation simulations considering sustainable urban drainage systems show reductions between 6 and 18%. The metropolitan results were compared and validated with a previous assessment done in the City of Barcelona. Based on these results, urban planners, emergency responders, and public administrations can develop effective adaptation measures based on cost–benefit analyses for current and future climate scenarios. Compared to previous studies, this approach adapts existing urban-scale methodologies to regional-scale flood risk assessment.

This study evaluates flood vulnerability in the Chennai City Region, Tamil Nadu, using remote sensing and GIS techniques to guide urban development planning. With rapid urbanization and recurrent flooding, Chennai faces heightened risks from heavy monsoon rains, inadequate drainage, and encroachment on natural floodplains. Sentinel-2 and Landsat satellite imagery, combined with GIS data such as digital elevation models (DEM) and land-use maps, were used to classify land cover, map flood extents, and assess flood vulnerability. A multi-criteria evaluation using Analytical Hierarchy Process (AHP) identified key vulnerability factors, including population density, elevation, land use, and proximity to water bodies and drainage infrastructure. The study also conducted sensitivity analyses, including map-removal sensitivity analyses, to quantify the impact of individual parameters on flood vulnerability mapping. The findings reveal significant urban expansion (85% of the area) and widespread impermeable surfaces contributing to high surface runoff and limited infiltration. Topographic Wetness Index (TWI), drainage density, slope, and distance from streams were used to assess flood-prone zones further. The Normalized Difference Vegetation Index (NDVI) was calculated to evaluate the extent and health of vegetation affected by flooding. At the same time, DEMs and terrain analysis provided insights into low-lying areas with higher flood vulnerability. The research identified flood-prone zones classified into low, medium, and high-risk areas, covering 24.4%, 50.2%, and 25.4% of the study region, respectively. These results underscore the need for sustainable land-use management, improved drainage infrastructure, and climate-resilient urban development strategies to mitigate flood vulnerability in Chennai. The comprehensive assessment aims to support flood vulnerability management efforts and urban resilience planning in the region.

The HAND of flood mapping: multi-dimensional evaluation across data-rich and data-poor basins using existing maps, ground observations, and remote sensing data

Predicting coastal subsidence and sea-level scenarios in the Sundarbans Delta using InSAR and artificial intelligence for sustainable coastal management.

A transferable deep learning framework for flood mapping: Spatial generalization across hydro-climatic regimes using satellite imagery

Mapping flood memory: How risk perception and social vulnerability drive flood insurance patterns in the U.S

Impacts of climate change and urbanization on soil moisture dynamics have reduced regional flood resilience

Abstract. Coastal flooding is among the most damaging natural hazards in Europe, yet large-scale assessments have typically relied on simplified static “bathtub” models and coarse elevation data. Here, we present a novel pan-European methodology that applies a dynamic flood model at 25 m resolution, forced by location-specific total water level hydrographs. These hydrographs integrate mean sea level, tides, storm surge, and wave setup with spatially varying foreshore slopes, allowing storm type, duration, and shape to be explicitly represented. More than 51 000 coastal target points were used to reconstruct events, and the methodology was validated against 12 local-scale historical floods across diverse coastlines. The validation results confirmed the robustness of the large-scale methodology while highlighting the strong dependence of the results on the resolution and vertical accuracy of the underlying digital elevation model. At continental-scale, sensitivity analyses quantified uncertainty from model selection, hydrograph shape, and storm type. Results show that static flood models systematically overestimate inundation, with errors exceeding 25 % in low-lying coastal floodplains such as Belgium and the United Kingdom. At the continental scale, storm type variability explains 41 % of flood map uncertainty, while hydrograph shape has a smaller but measurable effect. Including coastal protection standards reduces the estimated exposed floodplain by more than half, underscoring the critical role of defenses. By bridging the gap between global static assessments and local dynamic models, this study establishes a methodological benchmark for continental-scale flood hazard mapping.

Floods are among the most widespread and destructive natural hazards, posing persistent threats to human life, infrastructure, and socioeconomic systems worldwide. Over recent decades, flood mapping and modeling have evolved significantly due to advances in remote sensing, Geographic Information Systems (GIS), and data-intensive analytical methods. This review synthesizes recent progress and remaining challenges in flood susceptibility, inundation, hazard, and risk assessment, with particular emphasis on the growing role of machine learning, deep learning, and hybrid modeling frameworks for spatial prediction and decision support. Based on a comprehensive review of more than 130 peerreviewed studies, the paper examines methodological developments spanning conventional hydrological and hydraulic models, data-driven approaches, and integrated hybrid frameworks. These methods increasingly leverage multi-source geospatial data, including satellite imagery, digital elevation models, rainfall products, and socio-environmental indicators. Emerging research trends reveal a shift toward intelligent data fusion, ensemble modelling, hybrid architectures, and Generative Pre-trained Transformer (GPT) architectures that combine physical process understanding with learning-based algorithms to enhance predictive accuracy, robustness, and scalability across diverse climatic and urban settings. The review also highlights thematic advances in urban flood analysis, flash flood susceptibility mapping, real-time forecasting, and model performance evaluation under data uncertainty. Despite substantial progress, key challenges persist related to model generalization, interpretability, data quality, and operational implementation. By critically assessing current methodologies and research gaps, this study outlines future directions for next-generation Geospatial Artificial Intelligence (GeoAI) in flood mapping and modeling.

Abstract In April 2024, northern Oman experienced an extreme flash flood triggered by rainfall totals exceeding one to two years of the regional average within 24 h. This study evaluates the performance of three remote sensing approaches for mapping flood-activated channels: Sentinel-2 Tasseled Cap Transformation (TCT) Brightness, Sentinel-1 Amplitude Change Detection (ACD), and Sentinel-1 InSAR Coherent Change Detection (CCD). Multi-temporal optical, SAR amplitude, and SAR coherence datasets were processed and compared with hydrological terrain indices derived from TanDEM-X elevation data. Results show that CCD provided the clearest and most spatially consistent delineation of flood channels, unaffected by cloud cover and less prone to noise than ACD, while integrating changes over time into a single product. Inside flow channels, coherence difference was shown to drop by up to 0.6 and being considerably lower than during stable conditions. TCT effectively highlighted bleaching of alluvial deposits under clear-sky conditions, and ACD proved most useful where flooding persisted at the time of acquisition. The combined analysis demonstrates that CCD, supported by optical and terrain data, offers a robust and transferable method for post-event flood mapping in arid regions. Its compatibility with Sentinel-1’s acquisition strategies makes it a practical tool for preliminary flood mapping and post-event assessment, especially in the context of increasingly frequent extreme rainfall events.

ABSTRACT In recent times, the development of algorithms to delineate water surface maps has significantly boosted flood monitoring and mitigation efforts by utilizing dual polarization, multi‐temporal Sentinel‐1 synthetic aperture radar (SAR) data. The Sentinel‐1 mission, with its global land monitoring capability, has been widely employed for SAR‐based flood mapping. Compared to single‐image flood algorithms, change‐detection methods offer superior results by deriving flood extent from classified changes, requiring data‐based parameterization. This study critically evaluates the effectiveness of three cutting‐edge thresholding algorithms—Edge Otsu, Bmax Otsu, and Kittler–Illingworth (KI)—for automated flood water detection using dual polarization, multi‐temporal Sentinel‐1 SAR data, focusing on the September 2019 flood event in North‐eastern Thailand. Utilizing Google Earth Engine for preprocessing and image correction, the study examines three Sentinel‐1 change detection models—Difference Image, Normalized Difference Flood Index (NDFI), and Normalized Difference Sigma‐naught Index (NDSI). Among 27 combinations of inputs, change detection methods, and thresholding algorithms, the “Harmonic data‐S1GBM (2016–2017)” input paired with the KI thresholding algorithm and the NDSI change detection method achieved the highest overall accuracy of 86.29% (calculated using user accuracy, producer accuracy, and overall accuracy metrics against 2000 validation samples from GISTDA flood maps and Sentinel‐2 NDWI data). This combination proved most effective in distinguishing flooded from non‐flooded areas, underscoring the importance of selecting optimal data inputs and algorithms for accurate flood inundation mapping. The results highlight the superiority of the KI thresholding algorithm, particularly when used with harmonic data inputs, and establish a robust framework for future flood monitoring applications using Sentinel‐1 SAR data. Furthermore, the study emphasizes that for global and automatic flood services, algorithms should not depend on locally optimized parameters, as these cannot be automatically estimated and vary spatially, significantly affecting mapping accuracy.

Abstract Typhoons lead to meteorological disasters along the coastal region of China. Typhoons induce secondary hazards in cities such as flooding, landslides, vibration of high-rise buildings etc. This paper introduces an improved multi-criteria decision-making (MCDM) method, called the ECC model, for assessing the distribution of vulnerability to typhoon-induced flooding. The novelty of this method is its integration of weights calculated initially using the traditional entropy weight method (EWM), coefficient of variation method (CVM), and criteria importance through intercriteria correlation (CRITIC) method through a matrix-based weight combination approach; the vulnerability level is reflected through comprehensive scores. By applying historical data related to typhoon-induced flood disasters to this method, the vulnerability distribution in the Hainan Island region was estimated. The collected statistical data demonstrated that the proposed improved assessment method achieved results that corresponded highly with actual disaster situations, indicating the reliability of this method. Moreover, the results showed that the proposed ECC model predicted the disaster site due to the Yagi event accurately and reflected the field situation. This study serves as a valuable reference for disaster prevention/mitigation measurement.

Examining the mismatch between regulatory flood boundaries and actual flood extent: Evidence from the 2019 Nebraska floods