Optimum number of RADARSAT datasets for estimation of rice cropping areas in Chi watershed

Abstract

This study was carried out to determine optimum number of radar image datasets for mapping areas occupied by rice crop in Chi watershed. Four sets of RADARSAT-1 SAR fine mode data were taken into consideration. These included the datasets acquired on 29th Jun., 16th Aug., 28th Sep., and 27th Oct. 2007. The optimum number of data sets was determined by comparing performances of the classifications using different number of datasets. These included (1) the classification using all 4 sets of data described above, and (2) the classification using 3 sets of selected data, i.e., data acquired on 29th Jun., 16th Aug., and 28th Sep. The selection of datasets was made by using Feature Selection Algorithm. The results showed that the optimum number of datasets was 3. This is because the use of all the 4 sets and the 3 selected sets of data both yielded satisfactory results, i.e., Overall Accuracies > 84 %, Producer’s Accuracies and User’s Accuracies for rice cropping areas > 83 %. In addition, analysis of KHAT and Z-statistic showed that there was no significant difference between the map generated using 4 sets of data and that using 3 sets of data