Rice crop mapping with RADARSAT-2 data in the middle of Taiwan

Rice is an important crop in Taiwan because it provides not only livelihoods but also employment for a majority of rural populations. The size of rice fields in Taiwan is small and fragmented. The optical satellite data, such as Landsat and MODIS data, are easily contaminated by cloud cover that potentially affect the mapping results. The microwave data, which are able to penetrate though cloud cover, can partly overcome such issues of optical data. This study aims to map rice fields using RADARSTA-2 data for Changhua County in the middle of Taiwan. The data were processed for the 2014 first rice cropping season through four main steps: (1) data pre-processing, including radiometric and terrain corrections and speckle noise filtering of the backscattering coefficient of cross-polarization (VH) data, (2) Normalized Difference Sigma-0 Index (NDSI) calculation using 2 SAR images of sowing and heading dates selected based cropping calendar, (3) image classification using a threshold of NDSI, and (4) accuracy assessment of the mapping results. The classification map compared with the government's rice crop map indicated an overall accuracy of 80% and kappa coefficient of 0.60, respectively. The comparisons between the rice areas derived from the Radarsat-2 data classification with the government's rice area statistics showed the consistency with the government's rice area statistics. This study demonstrates the validity of the mapping approach for delineating rice fields in the middle of Taiwan using RADARSAT-2 data.