Use of RADARSAT-1 SAR data (standard mode) to map cultivated areas of rice crop in the northeast region

Abstract

This research was conducted to investigate performance of using RADARSAT-1 SAR (standard mode) to map areas of the in-season rice cropping in the northeast region, in each particular year. Three sets of the radar data acquired on different dates, i.e., 4th July, 14th September, and 25th October 2006, were used. The Maximum Likelihood classifier was applied to classify these data based on 2 classification schemes. The scheme #1 included 4 classes. Rice cropping areas were divided into 2 classes (Classes 1 and 2) representing 2 major cropping patterns. Class 1 was the areas of rice crop grown around June and harvested after last week of October. Class 2 was the areas of rice crop grown around June and harvested before last week of October. Other classes were field crops and forest. The mapping result was moderately accurate for the rice crop in Class 1 (Producer’s Accuracy = 75.00 % and User’s Accuracy = 78.95 %). However, for the rice crop in Class 2, the accuracy was somewhat poor (Producer’s Accuracy = 54.55 % and User’s Accuracy = 60.00 %). The classification scheme # 2 consisted of 3 classes. This scheme was designed for use when distinguishing the rice cropping patterns was not concerned. The rice crop of both cropping patterns were, therefore, merged into 1 class (Class 1+2). Others classes were still remained. Mapping accuracy was improved slightly for the rice cultivated areas. (Producer’s Accuracy = 74.19 % and User’s Accuracy = 82.14 %). In conclusion, the RADAR remote sensing technique used in this study yielded moderately accurate results. The technique can be applied to map the rice cultivated areas when all patterns of the rice cropping were considered as one class. However, distinguishing different rice cropping patterns was limiting.