Emergency flood mapping in Australia with Sentinel-1 and Sentinel-2 satellite imagery

Abstract

Timely inputs for spatial planning are essential to support decisions about preventive or damage controlling measures, including flood. Climate change predictions suggest more frequent floods in the future, implying a need for flood mapping. The objectives of the study were to evaluate the suitability of Sentinel-1 SAR data to map the extent of flood and to explore how land cover classification through different machine learning techniques and optical Sentinel-2 imagery can be applied as an emergency mapping tool. The Australian floods in March 2021 were used as a case study. Google Earth Engine was used to process and classify the flood extent and affected land cover types. Our study revealed the great suitability of Sentinel-1 SAR data for emergency mapping of flooded areas. Furthermore, land cover maps were produced using random forest (RD) and support vector machines (SVM) on optical Sentinel-2 Imagery. The presented workflow can be implemented in other parts of the world for the rapid assessment of flooded areas.