Ahmed et al. (2025) Flood Mapping from Satellite Imagery: A Response-Based Framework for Quantifying Flood Hazard and Uncertainty

Identification

• Journal: Earth Systems and Environment
• Year: 2025
• Date: 2025-10-07
• Authors: Rabia Ahmed, Md Mamunur Rashid, Nishan Kumar Biswas, Shubhra Sharma
• DOI: 10.1007/s41748-025-00862-1

Research Groups

• School of Ocean Science and Engineering, The University of Southern Mississippi, Ocean Springs, MS, USA
• Hydrological Sciences Laboratory, NASA Goddard Space Flight Center, Greenbelt, MD, USA
• University of Maryland Baltimore County, Baltimore, MD, USA
• Department of Disaster Science and Climate Resilience, University of Dhaka, Dhaka, Bangladesh

Short Summary

This study develops a response-based framework for probabilistic flood hazard mapping using time series of satellite-derived flood maps and multiple digital elevation models (DEMs). The framework provides a rapid and computationally inexpensive alternative to numerical hydrodynamic models, particularly for data-scarce regions, while quantifying flood hazard and associated uncertainties.

Objective

• To formalize a response-based framework for deriving probabilistic flood hazard maps using half-monthly time series of remotely sensed water extent (from Sentinel-1 images) and multiple DEM datasets, incorporating watershed responses and quantifying associated uncertainty.

Study Configuration

• Spatial Scale: Northeastern region of Bangladesh, approximately 21,000 km².
• Temporal Scale: Half-monthly scale for monsoon months (May to October) from 2015 to 2023 (9 years).

Methodology and Data

• Models used:
  • Flood Detection: Threshold-based method using Sentinel-1 backscatter coefficients, calibrated with monthly water extent climatology derived via the Edge Otsu Algorithm.
  • Inundation Depth Estimation: Two static methods:
    1. Grid method (600 m x 600 m grids).
    2. Polygon method (hydrologically connected water pixels).
  • Uncertainty Analysis: Square Root Error Variance (SREV) to quantify total uncertainty and contributions from different sources.
  • Validation Model: Hydrologic-hydrodynamic modeling framework (HHMF) coupling SWAT (hydrologic model) and HEC-RAS (hydrodynamic model).
• Data sources:
  • Satellite: Sentinel-1 satellite images (acquired and processed via Google Earth Engine - GEE).
  • Digital Elevation Models (DEMs):
    • SRTM Global 1 Arc Second (30 m) Version 3
    • ALOS World 3D − 30 m (AW3D30) Digital Surface Model (DSM)
    • NASADEM
    • Copernicus DEM – Global-30 m Digital Elevation Model (COPDEM)
    • Forest And Buildings removed Copernicus 30 m DEM (FABDEM)
  • Validation Data: HHMF simulations for the June 2022 flood, using Climate Change Initiative-Land Cover Maps of 2015, modified SRTM DEM, and GPM IMERG precipitation data.

Main Results

• The framework generated 1070 inundation depth maps, combining 107 half-monthly maps with 5 DEMs and 2 depth estimation methods.
• Validation against HHMF simulations showed high accuracy for satellite-derived surface water extent (88% accuracy, F1 score of 0.77, Kappa coefficient of 0.81).
• Inundation depth distributions from the proposed static methods were comparable to HHMF simulations, with most pixels showing depths between 1 m and 5 m.
• Probabilistic flood hazard maps were generated, including the probability of inundation exceeding 1 m and 20-year return period flood depths.
• Low-lying areas, particularly Haors and river corridors, showed high flood probabilities (approaching 1) and significant inundation depths (up to 4 m, with some areas exceeding 4-8 m for a 20-year flood).
• Significant uncertainty was observed in flood depth estimates, with the total uncertainty in the probability of inundation depth ≥ 1 m ranging from 0 to 0.5.
• Digital Elevation Models (DEMs) were identified as the primary source of uncertainty, contributing at least 50% of the total uncertainty in flood probability for approximately 73% of pixels, and for 20-year flood depth for approximately 79% of pixels, compared to depth estimation methods.

Contributions

• Proposes a novel response-based framework for probabilistic flood hazard mapping that integrates watershed responses by utilizing time series of satellite-derived flood maps.
• Offers a rapid and computationally inexpensive alternative to traditional, computationally demanding numerical hydrodynamic models, particularly valuable for data-scarce and resource-limited regions.
• Systematically quantifies uncertainty in flood hazard estimates and identifies the relative contributions of different sources (DEMs and depth estimation methods).
• Leverages publicly available, open-source Sentinel-1 imagery and multiple global DEMs, enhancing accessibility and applicability.
• Provides high-resolution (30 m) flood hazard maps and uncertainty estimates, supporting informed decision-making for infrastructure development, land-use planning, and enhancing community resilience.

Funding

The authors declare that no funds, grants, or other support were received during the preparation of this manuscript.

Citation

@article{Ahmed2025Flood,
  author = {Ahmed, Rabia and Rashid, Md Mamunur and Biswas, Nishan Kumar and Sharma, Shubhra},
  title = {Flood Mapping from Satellite Imagery: A Response-Based Framework for Quantifying Flood Hazard and Uncertainty},
  journal = {Earth Systems and Environment},
  year = {2025},
  doi = {10.1007/s41748-025-00862-1},
  url = {https://doi.org/10.1007/s41748-025-00862-1}
}