Endmember classes determination using spectral similarity analysis for MODIS reflectance channels

Endmember variability is one of reasons causing errors of model fits in spectral mixture analysis (SMA). The endmembers used in endmember matrix should be higher intra-class similarity and lower inter-class similarity, which depends on spectral wavelengths. The quantification of spectral similarity of intra- and inter-class helps to determine the appropriate endmember classes and optimize the endmember selection. In this study, five spectral matching algorithms were used to quantify the spectral similarity, and three spectral class-matching metrics based on the matching algorithms were used to analyze the spectral similarity of intra- and inter-class. The spectral class-matching metrics of a measured spectral library (ENVI spectral libraries) and MODIS image spectra were quantified and assessed to determine the appropriate the endmember classes for MODIS reflectance channels.