Development of In Situ Experiments for Evaluation of Anisotropic Reflectance Effect on Spectral Mixture Analysis for Vegetation Cover

Abstract

Owing to its flexibility, spectral mixture analysis (SMA) based on the linear spectral mixing model (LSMM) is a useful tool for subpixel vegetation cover estimation. Multiple scattering and endmember spectral variability are the two reasons that produce errors in the LSMM-retrieved cover fraction estimates. Although the anisotropic reflectance properties are well documented, their effect has barely been investigated in the studies of subpixel vegetation cover estimation. This letter developed a series of controlled in situ experiments (using a checkerboard mixture design) to evaluate the anisotropic reflectance effect (ARE) on fractional vegetation cover estimation using SMA based on the LSMM. The results illustrate that ARE has a large impact on SMA for vegetation cover estimation, and the developed approach allowing for ARE produces more accurate estimates, as the value of root-mean-square error drops more than 50%. This letter may open a new perspective for using SMA to estimate vegetation cover by emphasizing the importance of integrating ARE and characterizing anisotropic reflectance properties of an endmember class as another source of intraclass variability that is likely to be ignored.