Estimation of paddy field area using the area ratio of categories in each mixel of Landsat TM

Abstract

It is important to estimate the area of a paddy field accurately in order to evaluate the production of rice and the methane flux from paddy fields, which contributes to the greenhouse effect. Paddy fields are irrigated and a single crop is cultivated in them. Therefore, paddy fields are easy to classify compared to upland fields and forests. However, mixels decrease the accuracy of classification and increase the error in area estimation. To overcome these problems, we developed a model to estimate the area of paddy fields in pixels from Landsat TM data. We applied this model to verification sites in Hokkaido and Miyagi Prefecture. In these verifications of our model, we were able to estimate paddy field areas with 1-4 per cent and 3-3 per cent overall error, respectively. These results show that we can apply our model to practical use.