Comparison of scaling effects in fraction of vegetation cover between algorithms based on linear mixture model using VI

Area-averaged values of fraction of vegetation cover (FVC) derived from spectral data of different spatial resolution are perturbed by systematic biases, called scaling effect. The scaling effect degrades the compatibilities in FVC from multiple sensors. Objectives of this study is to compare scaling effects in FVC by two algorithms based on linear mixture model with vegetation index (VI); Isoline-based LMM and VI-based LMM. Previous study indicated that Isoline-based LMM has degree of freedom, which is the endmember spectra in the model that enables to mitigate the scaling effect under the two-endmember LMM. Results of numerical simulations using satellite data indicate that Isoline-based LMM with an optimum set of endmember spectra outperformed VI-based LMM with regard to the scaling effect, as expected from the scaling theory. It implies that the degree of the scaling effect may be controlled in some extent by choosing an appropriate set of endmember spectra.