Unlocking the full potential of Sentinel-1 for flood detection in arid regions

Research output: Contribution to journalArticleResearchpeer review

Authors

  • Shagun Garg
  • Antara Dasgupta
  • Mahdi Motagh
  • Sandro Martinis
  • Sivasakthy Selvakumaran

External Research Organisations

  • University of Cambridge
  • Helmholtz Centre Potsdam - German Research Centre for Geosciences (GFZ)
  • RWTH Aachen University
  • German Aerospace Center (DLR)
View graph of relations

Details

Original languageEnglish
Article number114417
Number of pages23
JournalRemote sensing of environment
Volume315
Early online date9 Oct 2024
Publication statusPublished - 15 Dec 2024

Abstract

Climate change has intensified flooding in arid and semi-arid regions, presenting a major challenge for flood monitoring and mapping. While satellites, particularly Synthetic Aperture Radar (SAR), allow synoptically observing flood extents, accurately differentiating between sandy terrains and water for arid region flooding remains an open challenge. Current global flood mapping products exclude arid areas from their analyses due to the sand and water confusion, resulting in a critical lack of observations which impedes response and recovery in these vulnerable regions. This paper explores the full potential of Sentinel-1 SAR to improve near-real-time flood mapping in arid and semi-arid regions. By investigating the impact of various parameters such as polarization, temporal information, and interferometric coherence, the most important information sources for detecting arid floods were identified. Using three distinct arid flood events in Iran, Pakistan, and Turkmenistan, different scenarios were constructed and tested using RF to evaluate the effectiveness of each feature. Permutation feature importance analysis was additionally conducted to identify key elements that reduce computational costs and enable a faster response during emergencies. Fusing VV coherence and amplitude information in pre-flood and post-flood imagery proved to be the most suitable approach. Results also show that leveraging crucial features reduces computational time by ∼35% as well as improves flood mapping accuracy by ∼50%. With advancements in cloud processing capabilities, the computational challenges associated with interferometric SAR computations are no longer a barrier. The demonstrated adaptability of the proposed approach across different arid areas, offers a step forward towards improved global flood mapping.

Keywords

    Coherence, Disaster management, Flood mapping, Machine learning, Remote sensing, SAR

ASJC Scopus subject areas

Sustainable Development Goals

Cite this

Unlocking the full potential of Sentinel-1 for flood detection in arid regions. / Garg, Shagun; Dasgupta, Antara; Motagh, Mahdi et al.
In: Remote sensing of environment, Vol. 315, 114417, 15.12.2024.

Research output: Contribution to journalArticleResearchpeer review

Garg, S., Dasgupta, A., Motagh, M., Martinis, S., & Selvakumaran, S. (2024). Unlocking the full potential of Sentinel-1 for flood detection in arid regions. Remote sensing of environment, 315, Article 114417. https://doi.org/10.1016/j.rse.2024.114417
Garg S, Dasgupta A, Motagh M, Martinis S, Selvakumaran S. Unlocking the full potential of Sentinel-1 for flood detection in arid regions. Remote sensing of environment. 2024 Dec 15;315:114417. Epub 2024 Oct 9. doi: 10.1016/j.rse.2024.114417
Garg, Shagun ; Dasgupta, Antara ; Motagh, Mahdi et al. / Unlocking the full potential of Sentinel-1 for flood detection in arid regions. In: Remote sensing of environment. 2024 ; Vol. 315.
Download
@article{4beac415bb3343f9926597b334529d20,
title = "Unlocking the full potential of Sentinel-1 for flood detection in arid regions",
abstract = "Climate change has intensified flooding in arid and semi-arid regions, presenting a major challenge for flood monitoring and mapping. While satellites, particularly Synthetic Aperture Radar (SAR), allow synoptically observing flood extents, accurately differentiating between sandy terrains and water for arid region flooding remains an open challenge. Current global flood mapping products exclude arid areas from their analyses due to the sand and water confusion, resulting in a critical lack of observations which impedes response and recovery in these vulnerable regions. This paper explores the full potential of Sentinel-1 SAR to improve near-real-time flood mapping in arid and semi-arid regions. By investigating the impact of various parameters such as polarization, temporal information, and interferometric coherence, the most important information sources for detecting arid floods were identified. Using three distinct arid flood events in Iran, Pakistan, and Turkmenistan, different scenarios were constructed and tested using RF to evaluate the effectiveness of each feature. Permutation feature importance analysis was additionally conducted to identify key elements that reduce computational costs and enable a faster response during emergencies. Fusing VV coherence and amplitude information in pre-flood and post-flood imagery proved to be the most suitable approach. Results also show that leveraging crucial features reduces computational time by ∼35% as well as improves flood mapping accuracy by ∼50%. With advancements in cloud processing capabilities, the computational challenges associated with interferometric SAR computations are no longer a barrier. The demonstrated adaptability of the proposed approach across different arid areas, offers a step forward towards improved global flood mapping.",
keywords = "Coherence, Disaster management, Flood mapping, Machine learning, Remote sensing, SAR",
author = "Shagun Garg and Antara Dasgupta and Mahdi Motagh and Sandro Martinis and Sivasakthy Selvakumaran",
note = "Publisher Copyright: {\textcopyright} 2024 The Authors",
year = "2024",
month = dec,
day = "15",
doi = "10.1016/j.rse.2024.114417",
language = "English",
volume = "315",
journal = "Remote sensing of environment",
issn = "0034-4257",
publisher = "Elsevier Inc.",

}

Download

TY - JOUR

T1 - Unlocking the full potential of Sentinel-1 for flood detection in arid regions

AU - Garg, Shagun

AU - Dasgupta, Antara

AU - Motagh, Mahdi

AU - Martinis, Sandro

AU - Selvakumaran, Sivasakthy

N1 - Publisher Copyright: © 2024 The Authors

PY - 2024/12/15

Y1 - 2024/12/15

N2 - Climate change has intensified flooding in arid and semi-arid regions, presenting a major challenge for flood monitoring and mapping. While satellites, particularly Synthetic Aperture Radar (SAR), allow synoptically observing flood extents, accurately differentiating between sandy terrains and water for arid region flooding remains an open challenge. Current global flood mapping products exclude arid areas from their analyses due to the sand and water confusion, resulting in a critical lack of observations which impedes response and recovery in these vulnerable regions. This paper explores the full potential of Sentinel-1 SAR to improve near-real-time flood mapping in arid and semi-arid regions. By investigating the impact of various parameters such as polarization, temporal information, and interferometric coherence, the most important information sources for detecting arid floods were identified. Using three distinct arid flood events in Iran, Pakistan, and Turkmenistan, different scenarios were constructed and tested using RF to evaluate the effectiveness of each feature. Permutation feature importance analysis was additionally conducted to identify key elements that reduce computational costs and enable a faster response during emergencies. Fusing VV coherence and amplitude information in pre-flood and post-flood imagery proved to be the most suitable approach. Results also show that leveraging crucial features reduces computational time by ∼35% as well as improves flood mapping accuracy by ∼50%. With advancements in cloud processing capabilities, the computational challenges associated with interferometric SAR computations are no longer a barrier. The demonstrated adaptability of the proposed approach across different arid areas, offers a step forward towards improved global flood mapping.

AB - Climate change has intensified flooding in arid and semi-arid regions, presenting a major challenge for flood monitoring and mapping. While satellites, particularly Synthetic Aperture Radar (SAR), allow synoptically observing flood extents, accurately differentiating between sandy terrains and water for arid region flooding remains an open challenge. Current global flood mapping products exclude arid areas from their analyses due to the sand and water confusion, resulting in a critical lack of observations which impedes response and recovery in these vulnerable regions. This paper explores the full potential of Sentinel-1 SAR to improve near-real-time flood mapping in arid and semi-arid regions. By investigating the impact of various parameters such as polarization, temporal information, and interferometric coherence, the most important information sources for detecting arid floods were identified. Using three distinct arid flood events in Iran, Pakistan, and Turkmenistan, different scenarios were constructed and tested using RF to evaluate the effectiveness of each feature. Permutation feature importance analysis was additionally conducted to identify key elements that reduce computational costs and enable a faster response during emergencies. Fusing VV coherence and amplitude information in pre-flood and post-flood imagery proved to be the most suitable approach. Results also show that leveraging crucial features reduces computational time by ∼35% as well as improves flood mapping accuracy by ∼50%. With advancements in cloud processing capabilities, the computational challenges associated with interferometric SAR computations are no longer a barrier. The demonstrated adaptability of the proposed approach across different arid areas, offers a step forward towards improved global flood mapping.

KW - Coherence

KW - Disaster management

KW - Flood mapping

KW - Machine learning

KW - Remote sensing

KW - SAR

UR - http://www.scopus.com/inward/record.url?scp=85205672329&partnerID=8YFLogxK

U2 - 10.1016/j.rse.2024.114417

DO - 10.1016/j.rse.2024.114417

M3 - Article

AN - SCOPUS:85205672329

VL - 315

JO - Remote sensing of environment

JF - Remote sensing of environment

SN - 0034-4257

M1 - 114417

ER -