Global near real-time flood mapping using a fully automated Sentinel-1-based processing chain

Abstract

An automated Sentinel-1-based processing chain for global near real-time flood detection and monitoring is presented. In contrast to optical satellites Sentinel-1, as a C-band radar satellite mission, operates independently from weather and illumination conditions. Within the scope of crisis management the processing chain provides time-critical information about the flood extent about one hour after data availability. The processing chain consists of pre-processing the radar data, preparing global auxiliary data, unsupervised classification as well as post-processing with a fuzzy logic-based approach. The final results are disseminated via a web-service to the users. Deployed in Python 3 the overall run time including data retrieval, processing and dissemination is less than 40 minutes. Future development foresees the integration of Cython in order to reduce calculation time.