Detection of Abandoned Rice Paddies with Satellite Remote Sensing

Abstract

Increase in abandoned agricultural lands is a big problem in Japan. Satellite remote sensing is one of promising approaches for periodic monitoring of the abandoned agricultural lands with low cost. We developed a method of detection of abandoned rice paddies with satellite data and evaluated it with ground observation data. We found that the difference of NDVI between a rice paddy and an abandoned rice paddy was particularly large in one month of just after planting as well as two months of just after harvesting. By taking advantage of this feature, we produced the map of abandoned rice paddies by using ALOS AVNIR-2 data taken in the post-harvest period. Error matrices were produced to evaluate some accuracies : Producer's accuracy for abandoned rice paddies was 71.8%, and user's accuracy for abandoned rice paddies was 86.1%. Although most of the previous studies used only satellite data taken in the post-planting period, this study shows utility of satellite data in the post-harvest period.