Estimation of fractional vegetation cover by unmixing HJ-1 satellite hyperspectral data

Abstract

Remote sensing provides the possibility for large-scale or even global monitoring of the fractional vegetation cover (FVC). In this paper, multiple endmember spectral mixture analysis (MESMA) method was used to extract vegetation information of Xinjiang's Shihezi area using the hyperspectral data acquired by Chinese HJ-1/HSI small satellite in the arid area. The Endmember average root mean square error (EAR) and pure pixel index (PPI) indices were combined to select the endmember spectra. The retrieved FVC from the HJ-1/HSI image data was verified with the in-situ measurements, and compared with the linear spectral mixture model (LSMM) result. The comparison shows that the MESMA method enables the use of different endmember combinations for different image pixels, thus can perform much better than the simple linear spectral unmixing analysis in the estimation of regional fractional vegetation cover information.