Silva et al. (2025) A methodological approach to flood dynamics based on satellite-derived spectral indices and altimetric forecast models: a case study in Southern Brazil

Identification

• Journal: The Science of The Total Environment
• Year: 2025
• Date: 2025-10-13
• Authors: Ubiratan Joaquim da Silva, Admilson da Penha Pachêco, Antonio Miguel Ruiz-Armenteros, Juarez Antonio da Silva, Julia Wickenhäuser, Tácito Richarles Ferreira da Silva, Luciana Barros Pinto, Anderson Luiz Ribeiro de Paiva, Leidjane Maria Maciel de Oliveira, Sylvana Melo dos Santos
• DOI: 10.1016/j.scitotenv.2025.180686

Research Groups

• Department of Civil and Environmental Engineering, Federal University of Pernambuco, Recife, Brazil
• Department of Cartographic and Surveying Engineering, Center for Technology and Geosciences, Federal University of Pernambuco, Recife, Brazil
• Department of Cartographic, Geodetic and Photogrammetry Engineering, University of Ja´en, Ja´en, Spain
• Microgeodesia Ja´en Research Group (PAIDI RNM-282), University of Ja´en, Ja´en, Spain
• Center for Advanced Studies on Earth Sciences, Energy and Environment (CEACTEMA), University of Ja´en, Ja´en, Spain
• Hydraulic Research Institute of the Federal University of Rio Grande do Sul (IPH/UFRGS), Porto Alegre, Brazil

Short Summary

This study developed an integrated remote sensing methodology to monitor flood-prone areas in the Porto Alegre Metropolitan Region, Brazil, combining topographic modeling (HAND) with spectral indices (NDWI, WNDWI) and validating results against hydrodynamic simulations (HEC-RAS). The approach demonstrated high accuracy, with WNDWI outperforming NDWI in detecting flood extent, particularly in turbid urban environments.

Objective

• To develop a methodology to assess the potential of combining HAND-derived flood susceptibility mapping with multispectral satellite indices (Landsat 8 - OLI and Landsat 5 - TM) to support flood monitoring in urban areas of the Porto Alegre Metropolitan Region (MRPA), Rio Grande do Sul, Brazil.
• To evaluate the capacity of the HAND model to detect vulnerable zones in a densely urbanized environment.
• To compare the sensitivity of NDWI and WNDWI indices in identifying water pixels under complex urban and turbid conditions using time series imagery.
• To validate the proposed approach through an integrated analysis of fuzzy classification techniques and Hydrologic Engineering Center–River Analysis System (HEC-RAS) hydraulic modeling results.

Study Configuration

• Spatial Scale: Porto Alegre Metropolitan Region (MRPA), Rio Grande do Sul state, Brazil, covering 10,916.60 square kilometers. The study focused on the Jacuí River basin and surrounding areas.
• Temporal Scale: Analysis of flood events in 2015 and 2024, with a non-flood reference scene from 2006. Landsat imagery was acquired for June 10, 2015, June 5, 2024, and August 5, 2006. Land Use and Land Cover (LULC) maps were analyzed for 2006, 2015, 2024, and 2025.

Methodology and Data

• Models used:
  • Height Above the Nearest Drainage (HAND) model: For flood susceptibility mapping.
  • Normalized Difference Water Index (NDWI): For water detection.
  • Weighted Normalized Difference Water Index (WNDWI): For water detection, particularly in turbid conditions.
  • Fuzzy logic: Applied for binarization of spectral index outputs to classify water pixels.
  • Otsu method: Used to automatically determine thresholds for fuzzy logic binarization.
  • Hydrologic Engineering Center–River Analysis System (HEC-RAS) 6.3.1: Used for 2D unsteady flow hydrodynamic simulations and as reference data for validation.
  • Large Basin Hydrological Model (MGB): Used to generate upstream hydrographs for HEC-RAS simulations.
• Data sources:
  • Shuttle Radar Topography Mission (SRTM) Digital Elevation Model (DEM): 30 meter spatial resolution, used for HAND model.
  • Landsat 8 Operational Land Imager (OLI) and Landsat 5 Thematic Mapper (TM) imagery: 30 meter spatial resolution, Collection 2 surface reflectance data, used for NDWI and WNDWI calculation.
  • LiDAR DEM (1 meter resolution) and detailed bathymetric surveys: Combined with SRTM for HEC-RAS topographic input.
  • Brazilian National Water and Sanitation Agency (ANA) stations: Provided rainfall inputs and streamflow series for HEC-RAS.
  • Hydropower records: Provided streamflow series for HEC-RAS.
  • MapBiomas data: Used for Land Use and Land Cover (LULC) analysis.

Main Results

• The HAND model effectively identified low-lying, high-risk areas near watercourses but showed limitations in densely urbanized zones due to the 30 meter resolution of the SRTM DEM.
• WNDWI consistently mapped a larger spatial extent of flooding compared to NDWI across all analyzed scenes. In 2015, WNDWI mapped 676.290 square kilometers (6.20% of the study area) as flooded, while NDWI mapped 516.285 square kilometers (4.73%). In 2024, WNDWI mapped 2423.789 square kilometers (22.21%), significantly more than NDWI's 1206.391 square kilometers (11.06%).
• WNDWI demonstrated superior thematic accuracy when validated against HEC-RAS hydrodynamic simulations. For the 2015 flood, WNDWI achieved an overall accuracy of 98.45% with a Kappa coefficient of 0.75 (substantial agreement), an AUC of 0.55, and an F1-score of 0.219. NDWI showed an overall accuracy of 96.20% with a Kappa of 0.52 (moderate agreement), an AUC of 0.36, and an F1-score of 0.112.
• For the more severe 2024 flood, WNDWI achieved an overall accuracy of 99.10% with a Kappa of 0.82 (almost perfect agreement), an AUC of 0.74, and an F1-score of 0.730. NDWI recorded an overall accuracy of 96.13% with a Kappa of 0.61 (substantial agreement), an AUC of 0.44, and an F1-score of 0.130.
• WNDWI consistently outperformed NDWI in detecting floodwaters in highly turbid and heterogeneous urban areas, attributed to its weighted combination of NIR and SWIR bands.
• Both indices showed reduced sensitivity in transitional zones characterized by mixed land cover types (e.g., urban structures, forested areas, cultivated lands, bare soils, and varying moisture levels).

Contributions

• This study presents a novel integrated methodology combining predictive topographic modeling (HAND), dynamic flood extent analysis using multispectral satellite indices (NDWI and WNDWI with fuzzy logic), and validation against hydrodynamic simulations (HEC-RAS).
• It demonstrates the superior performance and robustness of the Weighted Normalized Difference Water Index (WNDWI) over the traditional Normalized Difference Water Index (NDWI) for accurate flood mapping in complex, turbid, and heterogeneous urban environments.
• The research enhances the reliability and accuracy of remote sensing-based flood mapping, providing crucial spatial data and predictive tools for urban planning, flood risk management, and the implementation of resilient strategies in climate-vulnerable metropolitan areas.
• The findings support the development of scalable early warning systems and disaster response planning by providing timely and accurate information on flood susceptibility and inundation extent.

Funding

• Microgeodesia Ja´en Research Group (PAIDI RNM-282) from the University of Ja´en.
• Federal University of Pernambuco (UFPE) research project No. 031162 “Geophysical and Geodetic Studies for Monitoring Hydrographic Basins and Dams (EGMBHB)”.
• Federal University of Pernambuco (UFPE) research project No. PVK1141-2024 “Vegetation and Water Cover of River Basins Using Orbital Images in the State of Pernambuco”.
• National Council for Scientific and Technological Development (CNPq) project No. 406919/2022-411 “National Observatory for Water Security and Adaptive Management (ONSEADAdapta)”.
• National Council for Scientific and Technological Development (CNPq) project No. 408269/2023-5 “Use of Socio-Inclusive Green Technologies as Urban Drainage Control Measures”.
• Universidad de Ja´en/CBUA (for open access charge).

Citation

@article{Silva2025methodological,
  author = {Silva, Ubiratan Joaquim da and Pachêco, Admilson da Penha and Ruiz-Armenteros, Antonio Miguel and Silva, Juarez Antonio da and Wickenhäuser, Julia and Silva, Tácito Richarles Ferreira da and Pinto, Luciana Barros and Paiva, Anderson Luiz Ribeiro de and Oliveira, Leidjane Maria Maciel de and Santos, Sylvana Melo dos},
  title = {A methodological approach to flood dynamics based on satellite-derived spectral indices and altimetric forecast models: a case study in Southern Brazil},
  journal = {The Science of The Total Environment},
  year = {2025},
  doi = {10.1016/j.scitotenv.2025.180686},
  url = {https://doi.org/10.1016/j.scitotenv.2025.180686}
}