時系列NOAA/AVHRRデータを用いた土地被覆変化傾向抽出

抄録

The goal of this study is to use NDVI and LST (land surface temperature) time series NOAA/AVHRR data to estimate global tendencies in land cover changes from 1982 to 1999. As a preprocessing step, Temporal Window Operation-correction (TWO Method) was performed to reduce the effects of clouds. The mean (AVG_NDVI and AVG_LST) and standard deviation (STD_NDVI and STD_LST) values were calculated for each NDVI and LST data over the 18 period. The results showed that NDVI varied according to rate of vegetation coverage ; and LST varied by latitude, altitude and whether or not the area was vegetated. In order to identify any abnormal trends, AVG_NDVI and STD_NDVI were used to calculate the monthly deviation. These calculations revealed that Africa suffered a period of drought from July 1984 to January 1985. In addition, in order to reveal changes in land cover by season, profile matching, taking into consideration phonological shifts, was implemented, and the difference between these results and average (DIF_NDVI and DIF_LST) values were determined. The results of these comparisons allowed identification of regional changes in NDVI under the influence of abnormal weather patterns. Furthermore, the gradient was extracted for 18 years of NDVI and LST data, using the linear regression analysis from DIF_NDVI and DIF_LST. The two gradient values plotted for NDVI and LST gradient indicated several important trends, such as an increase in NDVI in many grassland areas, including farmland ; and a increase in LST in desert areas.