The paddy rice damage caused by a flood in the Kanto district in Japan has been analysed using multitemporal Landsat-5 Thematic Mapper images. The image acquired immediately after the flood showed the spread of turbid water over the paddy fields and the image acquired 1 month after the flood showed the paddy rice damage in the inundated areas. The flood damage assessment was undertaken using several multivariate analyses. (1) The correlation analysis revealed that the turbidity level of flood water was related to the rice damage. (2) The decrease of the rice yield was estimated from both images by multiple regression models. (3) The flood damage relationship was determined by a cluster analysis using the higher-order principal components of the multitemporal image. Based on the results of these analyses, distribution maps of the rice yield levels were estimated at each date and a classification of the flood damage pattern was obtained.
展开▼
