Abstract
We analyze paddy fields in Nara basin in order to show that the pattern expand method is reasonable for quantitative analysis of an area which is a mixture of water, vegetation and soil. The pattern expand method has been developed to analyze remote sensing multispectral data such as Landsat TM data.
First, the pattern expand method is applied to three suitable seasons' Landsat TM data. Next, these data are coregistered using affine transformation with several sets of control points, and the characteristics of paddy fields are studied through three seasons. Using three pattern expand coefficients, paddy fields can be recognized from other land cover types, and classified to three paddy field types. The coefficients of these paddy fields are compared through three seasons on the same parameter space. Lastly, paddy fields are extracted using three pattern expand coefficients, and the paddy field areas of six cities and towns in Nara basin are estimated. The paddy field areas from this analysis are in good agreement with those from statistical data.
