Flooded Rice Paddy Detection and its Accuracy Assessment Using Sentinel-1 and Planetscope Images: A Case Study of 2018 Spring Flood in West Java Indonesia

This research evaluates damaged rice paddy fields by using remotely sensed data for the agricultural insurance recently launched in Indonesia. The relatively extensive flooding had occurred in the Tegalluar area in Bojongsoang in the 2018 spring. The Sentinel-1 and the PlanetScope satellites observed the Tegalluar area during the flooding period. We propose an automatic thresholding method to detect flood areas in rice paddy fields using two Sentinel-1 C-band SAR data acquisitions pre- and post-flooding. We confirmed the flood detection accuracy using the visible and near-infrared images acquired by the PlanetScope satellites. Our proposed method showed that the VV data outperformed the VH data in correlation ratio and discriminant accuracy. The overall classification accuracy of non-flood and flood areas reached 84.7% with VV data and 80.6% with VH data, including the error caused by the time difference between Sentinel-1 and PlanetScope data acquisitions. Utilizing the speckle reducing filters with SAR data improved the overall classification accuracy by 5 %.