Abstract
This study used IKONOS "Geo" imagery to detect damage to pine stands caused by pine wilt disease. The study was conducted in pine stands on Mt. Tsukuba in Ibaraki Prefecture, Japan. 66 plots (10m × 10m each) were established in the study area, taking damage ratio into consideration ; all the trees were tallied and examined for damage. Ratios of damage, healthy pine trees, and hardwood trees were then calculated in terms of basal area. Pan-sharpened colour imagery and multi-spectral imagery were analysed, and normalised difference vegetation index (NDVI) values were calculated from the images. The relationship between the mean NDVI for pixels in the plots and the field-observed conditions in the plots was then examined. The damage ratio and the ratio of hardwood trees tended to decrease and the ratio of healthy pine trees tended to increase as the NDVI increased for the two imagery types. Moreover, to validate a prediction method, regression equations were derived from half of the plots and applied to the other half of the plots. The root mean square errors (RMSEs) of predicting the damage ratio and ratio of healthy pine trees from the NDVI were approximately 0.15 and 0.12, respectively, for the pan-sharpened imagery ; the RMSEs for the same two ratios were approximately 0.17 for the multi-spectral imagery.
