Abstract
We have studied an algorithm and a new vegetation index for analyses of ADEOS-II/GLI data, based on the pattern decomposition method.1) To simulate spectral response patterns of the GLI sensor, reflectances of about 450 samples were measured in the field with a spectrometer covering the spectral range of the GLI sensor. About 96% of the information of the nineteen-dimensional GLI data was successfully transformed into three pattern decomposition coefficients.2) It was confirmed.by the experiment using the spectrometer that land cover ratios in a pixel are estimated from the pattern decomposition coefficients correctly. Furthermore, GLI data with 250m and 1km spatial resolutions were simulated from LANDSAT/TM data with 30m spatial resolution. Using the data, it was shown that land cover ratios in the GLI pixel estimated from the pattern decomposition coefficients are nearly equal to those of TM pixels in corresponding areas of the GLI pixel.3) A new vegetation index, VIPD (Vegetation Index based on Pattern Decomposition) was developed. VIPD utilizes all the nineteen-dimensional GLI data, and reflects the amount of vegetation and the degree of vegetation vigor. The index is more sensitive for vegetation cover ratio, for the vertical thickness of vegetation, and for vegetation type, such as broad leaves and needle leaves, than is NDVI.
From these results, it became evident that the algorithm based on the pattern decomposition method is sufficiently able for analyzing hyper-multidimensional GLI data, and the new vegetation index is useful in the study of vegetation.
