Multi‐temporal synthetic aperture radar flood mapping using change detection

Miles Adam Clement, Chris Kilsby, P. Moore

Journal of Flood Risk Management · 2017 · 395 citations · 34 references

DOIFull text

Open access

Concepts

TL;DR

A change‑detection and thresholding approach, adapted from prior work, was applied to 13 Sentinel‑1 SAR images (VH and VV polarizations) to delineate flooding extent during the 2015‑2016 Yorkshire winter floods. The method identified peak flood extents on 29 Dec 2015, achieving 97 % accuracy against a Sentinel‑2 optical image, with VV polarization yielding fewer misclassifications, and showed good agreement with Environment Agency Flood Maps while also detecting pluvial flooding beyond the floodplains.

Abstract

A change detection and thresholding methodology has been adapted from previous studies to determine the extent of flooding for 13 Sentinel‐1 synthetic aperture radar images captured during the floods of winter 2015–2016 in Yorkshire, UK . Both available polarisations, VH and VV , have been processed to allow for a comparison of their respective accuracy for delineating surface water. Peak flood extents are found on 29 December 2015 during the aftermath of storms Eva and Frank. Results have been validated against a Sentinel‐2 optical image, with both polarisations producing a total accuracy of 97%. Of the two polarisations, VV produces fewer misclassifications, mirroring the similar results reported in previous research. Mapped results are compared to the Environment Agency Flood Maps for Planning ( EA FMP ), with good correlation observed for inundation on the floodplains. Differences occur away from the floodplains, with the satellite data identifying pluvial flooding not highlighted by the EA FMP .

References

34