2007 年 27 巻 2 号 p. 129-140
In order to grasp the balance between supply and demands of vegetables such as cabbage, lettuce and Chinese cabbage which are produced on a large scale, forecast of their harvest time and area is a problem to be awaiting solution in Japan. This study proposed an approach of forecasting the harvest time and area of cabbage by using SPOT images in Tsumagoi country and discussed practical application to provide forecasting information. Firstly, cabbages field-based were extracted by the proposed post-classification method using the field polygon data. The field polygon data is a section of cultivated field which was interpreted from the digital aerial photography. Next, the pixel-based cabbages harvest time was estimated for extracted cabbage fields from the profiles of normalized difference vegetation index (NDVI) corresponding to cabbage growing and relationship between NDVI and harvest date. Results of extracted cabbage field and estimated harvesting period respectively derived from SPOT5 images of May 5 and July 28, 2005, were verified by ground investigation. According to the profile of NDVI, cabbage pixels with multiple growing stages in the image can be divided into two groups of leaves number increase term and leaves weight increase term. Results for the cabbages during the early stage in term of leaves number increase, presented a low accuracy because their spectral reflectance in SPOT is similar to bareland. Results for the cabbages in term of leaves weight increase showed strong coincidence with ground investigation. Furthermore, it was found that results of forecasted harvest area respectively from multi-years SPOT image can give a trend analysis of cabbage possible supplying field comparing with the historical information of the past harvested field and shipped amount to market. As a result, this study suggested an operational application of satellite remote sensing in forecast of vegetable harvest time and area, aimed to provide the services for the coordination between supply and market demand of vegetables. Moreover the past information contained in ancillary data such as remote sensing data, shipped field data and so on should be incorporated into the forecasting process.