A Simple and Straightforward Approach in Mapping Rice Area Using Multitemporal Sentinel-1 SAR Data

Abstract

Sentinel-1 sensor capable to provide temporally dense and open access data is a candidate platform for rice monitoring. The objective of this study was to exploit Sentinel-1 C-band Synthetic Aperture Radar (SAR) multitemporal imagery to detect and to map rice cropping areas of Cianjur, Indonesia, through (1) learning backscattering behavior over rice fields; and (2) generating rice area maps via several thresholding combinations. Layer manipulation (r.series aggregate operation) and RGB false-color composite images were constructed to better visualize the separability of rice areas among other types of land utilization. The result shows an averaged accuracy of about 82% with the highest accuracy of 85.6% using maximum and minimum backscatter coefficient thresholds on VH polarization. It is, therefore, concluded that a simple, straightforward approach could be beneficial for a specific task while minimizing the computational burden for further broad-scale implementation.