Sentinel-1 Synthetic Aperture Radar Burned Area Detection Using Expectation Maximization in a Multiscale Approach
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
This research explores wildfire burn mapping using Sentinel-1 Synthetic Aperture Radar (SAR) imagery for the 2021 Woods Creek Fire in the Helena-Lewis and Clark National Forest in Montana and the 2021 French Fire near Lake Isabella in Kern County California. Sentinel-1 SAR imagery is used since it can be collected during most weather conditions as well as in heavy smoke and is useful in the upper latitudes where wildfires often occur. Both the ascending and descending orbits as well as the co-polarity (VV) and cross-polarity (VH) are evaluated. The increase of wildfire occurrence is the result of lower precipitation and fuel moisture content as a result of climate change. The methodology by which SAR imagery detects wildfire burns is adapted to use SAR imagery from Google Earth Engine (GEE) and a method provided by the Alaska Satellite Facility (ASF). This method utilizes a stationary wavelet transform and math morphology to process imagery at various scales and expectation maximization in order to generate change classes. The resulting burn area is compared to Sentinel-2 differenced Normalized Burn Area (dNBR) and MODIS Burned Area. The ascending orbits of Sentinel-1 burned areas provided the best results compared to those of the descending orbits likely due to the limited ability of GEE to radiometrically terrain correct SAR imagery.