Developing NDVI estimation model of forest area in Japan, using temperature data
2006
Noguchi, I.(Hokkaido. Inst. of Environmental Sciences, Sapporo (Japan)) | Buhe, A. | Takada, M. | Hamahara, K. | Takahashi, H. | Tamada, K.
In this study, estimation model of the Normalized Difference Vegetation Index (NDVI) using temperature and location data was developed. Data sets for analysis were the NDVI images in 2003 from SPOT4/VEGETATION (Maximum NDVI values in 10-days synthesis), and temperature data from AMeDAS 173 stations, located forest area in 2003. Ten days averages of temperature (before 10days time-lag) were selected from the ten days moving averages of temperature as input data of models by statistics analysis. Unusual lower NDVI values were excluded as outlier values for the effects of cloud jamming. Furthermore, site and area models were developed considering vegetation in each five class divided by the warm index (WI). It was considered that there were the lower and higher limited NDVI values because NDVI values shows stable values in summer and winter seasons. Thus, models were used a linear-regressions model between lower and higher limited temperature values. The calculated limited temperature value was used when input-data was lower and higher than the limited temperature values in each warm index class. As per the results, it was found the good correlation with the site model results and the observation values of NDVI (r=0.968, mean difference between model results and observation values: 0.087). Furthermore, the 6 area models in all Japan were developed based on the site models using following factor; latitude, longitude, altitude, mean temperature and WI. Finally, it was found that the rather good correlation coefficient and mean difference between the area model results and observation values was 0.923 and 0.131, respectively.
اظهر المزيد [+] اقل [-]