Scholarlink

<p>Flood hazard mapping is crucial for planning flood protection and mitigation measures, minimizing flood losses, and developing effective flood response plans. Typically, flood hazard mapping adopts an event-based deterministic approach to compute flood hazard for a specific design event (e.g., a 100-year flood), assuming that the probability of flood drivers (e.g., discharge) approximates the probability of flood extent and inundation. In contrast, this study develops a framework that utilizes satellite images for response-based probabilistic flood hazard mapping, where flood hazard maps are generated from the time series of flood maps. Employing a calibrated flood detection algorithm and Sentinel-1 imagery, we first derive surface water extent maps for each half of the monsoon month (May to October) from 2015 to 2023 across the northeastern region of Bangladesh (approximately 21,000&#xa0;km²). We derive flood depths using two static depth estimation techniques and five digital elevation models (DEMs). Flood depths are calculated as the difference in the elevations of a selected pixel and the highest elevation of the watered (flooded) pixels within a defined boundary. Two approaches are adopted to define the boundary: (1) dividing the study area into 600&#xa0;m x 600&#xa0;m grids, and (2) dividing the area into polygons of connected watered pixels. Finally, flood hazard maps and their associated uncertainties are computed at each pixel for a 20-year return period (i.e., a 20-year flood). Additionally, the probability of flood occurrence exceeding an inundation depth (e.g., 1&#xa0;m) is also investigated. The study reveals significant differences (i.e., uncertainty) in the flood depth estimates originating from the DEMs and the depth estimation methods. The proposed approach provides a rapid and computationally inexpensive alternative to numerical hydrodynamic models for response-based probabilistic flood mapping, particularly for data-scarce and resource-limited regions.</p> Graphical Abstract <p>Based on the graphical snapshot, the study offers a response-based flood hazard mapping framework that leverages satellite images. Northeastern Bangladesh is considered a case study area. Inundation depth is estimated from Sentinel-1-derived water extent maps using two static methods and five digital elevation models (DEMs). In the first method, the study area is divided into small 600&#xa0;m x 600&#xa0;m grids, while in the second method, flooded polygons are identified using interconnected pixels. Within each grid or polygon, the highest elevation point is identified, and the flood depth for each pixel is calculated by subtracting the pixel’s elevation in the DEM from the highest flooded point’s elevation. This procedure computes flood depth at each pixel and provides a flood depth map. The study produces flood depth maps for each half of the monsoon months (May to October) from 2015 to 2023 using the two flood depth estimation methods and five DEMs, resulting in 1070 inundation depth maps. These maps are then used to create probabilistic inundation maps showing the flood depths and corresponding uncertainty at each pixel for a 20-year return period. The areas surrounding the Haors (unique wetlands characterized by seasonal freshwater depressions) are more prone to flooding. The contribution of DEMs to the uncertainty in the 20-year flood depth is higher compared to that from the depth estimation methods. The proposed framework offers a rapid and cost-effective alternative to computationally demanding numerical hydrodynamic models in response-based probabilistic flood mapping, especially in areas with limited data and resources. </p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Flood Mapping from Satellite Imagery: A Response-Based Framework for Quantifying Flood Hazard and Uncertainty

  • 期刊论文

  • 发表日期：

    2025年10月7日

查看全文
  • Raihan Uddin Ahmed,
  • Md Mamunur Rashid,
  • Nishan Kumar Biswas,
  • Smita Sharma

摘要

Flood hazard mapping is crucial for planning flood protection and mitigation measures, minimizing flood losses, and developing effective flood response plans. Typically, flood hazard mapping adopts an event-based deterministic approach to compute flood hazard for a specific design event (e.g., a 100-year flood), assuming that the probability of flood drivers (e.g., discharge) approximates the probability of flood extent and inundation. In contrast, this study develops a framework that utilizes satellite images for response-based probabilistic flood hazard mapping, where flood hazard maps are generated from the time series of flood maps. Employing a calibrated flood detection algorithm and Sentinel-1 imagery, we first derive surface water extent maps for each half of the monsoon month (May to October) from 2015 to 2023 across the northeastern region of Bangladesh (approximately 21,000 km²). We derive flood depths using two static depth estimation techniques and five digital elevation models (DEMs). Flood depths are calculated as the difference in the elevations of a selected pixel and the highest elevation of the watered (flooded) pixels within a defined boundary. Two approaches are adopted to define the boundary: (1) dividing the study area into 600 m x 600 m grids, and (2) dividing the area into polygons of connected watered pixels. Finally, flood hazard maps and their associated uncertainties are computed at each pixel for a 20-year return period (i.e., a 20-year flood). Additionally, the probability of flood occurrence exceeding an inundation depth (e.g., 1 m) is also investigated. The study reveals significant differences (i.e., uncertainty) in the flood depth estimates originating from the DEMs and the depth estimation methods. The proposed approach provides a rapid and computationally inexpensive alternative to numerical hydrodynamic models for response-based probabilistic flood mapping, particularly for data-scarce and resource-limited regions.

Graphical Abstract

Based on the graphical snapshot, the study offers a response-based flood hazard mapping framework that leverages satellite images. Northeastern Bangladesh is considered a case study area. Inundation depth is estimated from Sentinel-1-derived water extent maps using two static methods and five digital elevation models (DEMs). In the first method, the study area is divided into small 600 m x 600 m grids, while in the second method, flooded polygons are identified using interconnected pixels. Within each grid or polygon, the highest elevation point is identified, and the flood depth for each pixel is calculated by subtracting the pixel’s elevation in the DEM from the highest flooded point’s elevation. This procedure computes flood depth at each pixel and provides a flood depth map. The study produces flood depth maps for each half of the monsoon months (May to October) from 2015 to 2023 using the two flood depth estimation methods and five DEMs, resulting in 1070 inundation depth maps. These maps are then used to create probabilistic inundation maps showing the flood depths and corresponding uncertainty at each pixel for a 20-year return period. The areas surrounding the Haors (unique wetlands characterized by seasonal freshwater depressions) are more prone to flooding. The contribution of DEMs to the uncertainty in the 20-year flood depth is higher compared to that from the depth estimation methods. The proposed framework offers a rapid and cost-effective alternative to computationally demanding numerical hydrodynamic models in response-based probabilistic flood mapping, especially in areas with limited data and resources.