Papers

HOME > Research > Papers
Department Disaster Information Research Division
Date 2017
Name of Journal Environmental Monitoring and Assessment
Title A comparative study on generating simulated Landsat NDVI images using data fusion and regression method—the case of the Korean Peninsula
Author Lee Mi Hee
Type ±¹³»Àú³Î
Abstract Landsat optical images have enough spatial and spectral resolution to analyze vegetation growth characteristics. But, the clouds and water vapor degrade the image quality quite often, which limits the availability of usable images for the time series vegetation vitality measurement. To overcome this shortcoming, simulated images are used as an alternative. In this study,
weighted average method, spatial and temporal adaptive reflectance fusion model (STARFM) method, and multilinear regression analysis method have been tested to produce simulated Landsat normalized difference vegetation index (NDVI) images of the Korean Peninsula. The test results showed that the weighted average method produced the images most similar to the actual images, provided that the images were available within 1 month before and after the target date. The STARFM method gives good results when the input image date is close to the target date. Careful regional and seasonal consideration is required in selecting input images. During summer season, due to clouds, it is very difficult to get the images close enough to the target date. Multilinear regression analysis gives meaningful results even when the input image date is not so close to
the target date. Average R2 values for weighted average method, STARFM, and multilinear regression analysis were 0.741, 0.70, and 0.61, respectively.
Keyword Landsat, MODIS, Simulated image, STARFM, NDVI, Multilinear regression analysis
View 2224