Using negative soil adjustment factor in soil-adjusted vegetation index (SAVI) for aboveground living biomass estimation in arid grasslands

Abstract

All values of soil adjustment factor (L) from 0 to 1 in the soil-adjusted vegetation index (SAVI) were found to be undesirable in arid areas with sparse vegetation cover. We hypothesized that the negative soil adjustment factor in the SAVI would be satisfactory in arid areas. In the study, we explored the potential of negative soil adjustment factor in the SAVI against a set of field measurements during the growing season of 2009 in the desert steppe of Inner Mongolia, China. As hypothesized, negative soil adjustment factor was found to give satisfactory reduction of soil background noise. SAVI (L = −0.2) (R2 = 0.64, RMSE = 18.4 g m−2, rRMSE = 25.5%, n = 120) provided dramatic improvements over optimal NDVI (normalized difference vegetation index) (that is, SAVI (L = 0)) (R2 = 0.44, RMSE = 22.7 g m−2, rRMSE = 31.4%, n = 120) for aboveground living biomass (AGB) estimation among tested vegetation indices. A physical basis for the negative soil adjustment factor was also explored. Results showed that, the intersected points between approximate vegetation isolines and soil line were situated in the domain of positive values of NIR (near-infrared) and red in the first quadrant of NIR-red plane, and the slopes of vegetation isolines increased with increasing AGB in the desert steppe, thus resulting in negative factor by shifting the origin toward positive intersected points. This runs counter to reported vegetation isolines behavior for SAVI, in which the intersected points were situated in the region of negative values of NIR and red in the third quadrant. The good performance of negative factor was also observed for percent green cover (PGC) estimation in the desert steppe. We further proposed a conceptual model to describe vegetation isolines behavior in arid grasslands. Our findings will complement widely used soil adjustment factor. Further study is needed to validate the negative soil factor and test the conceptual model in other arid grasslands.