MONITORING LEAF-OUT OF DECIDUOUS FOREST USING NDVI OF TERRA/MODIS BETWEEN 2000 AND 2011 OVER GIFU PREFECTURE, JAPAN

摘要

Daily Normalized Difference Vegetation Index (NDVI) images of Terra/MODIS were analysed to detect leaf-out date of deciduous forest between 2000 and 2011 over forests in Gifu prefecture, Japan to know annual variations. Change of NDVI in spring was modeled by a linear regression analysis at each pixel, and leaf-out date was estimated. Advantages of this method were as follows. 1) Leaf-out date was able to be determine daily basis in each pixel. 2) Since no smoothing or curve fitting was applied, the method was able to analyze small NDVI change. 3) The method was not suffered by noises by clouds which appeared as NDVI decrease. The method estimated leaf-out date quite accurately in deciduous forest where NDVI changed great in spring in snowy areas. Leaf-out started in early April and late June in lowland and high mountains exceeding 2500m ASL, respectively. No clear inter-annual change trend appeared in the leaf-out maps. Ranges of the earliest and latest leaf-out during the 12 years were between about 10 and 20 days and were longer in lower areas than higher areas. Leaf-out synchronized well at different elevation in a small area, however, synchronization was less clear at separated pixels.