Remotely sensed estimation and mapping of soil moisture by eliminating the effect of vegetation cover

Soil moisture (SM), which plays a crucial role in studies of the climate, ecology, agriculture and the environment, can be estimated and mapped by remote sensing technology over a wide region. However, remotely sensed SM is constrained by its estimation accuracy, which mainly stems from the influence of vegetation cover on soil spectra information in mixed pixels. To overcome the low-accuracy defects of existing surface albedo method for estimating SM, in this paper, Qinghai Lake Basin, an important animal husbandry production area in Qinghai Province, China, was chosen as an empirical research area. Using the surface albedo computed from moderate resolution imaging spectroradiometer (MODIS) reflectance products and the actual measured SM data, an albedo/vegetation coverage trapezoid feature space was constructed. Bare soil albedo was extracted from the surface albedo mainly containing information of soil, vegetation, and both albedo models for estimating SM were constructed separately. The accuracy of the bare soil albedo model (root mean square error=4.20, mean absolute percent error=22.75%, and theil inequality coefficient=0.67) was higher than that of the existing surface albedo model (root mean square error=4.66, mean absolute percent error=25.46% and theil inequality coefficient=0.74). This result indicated that the bare soil albedo greatly improved the accuracy of SM estimation and mapping. As this method eliminated the effect of vegetation cover and restored the inherent soil spectra, it not only quantitatively estimates and maps SM at regional scales with high accuracy, but also provides a new way of improving the accuracy of soil organic matter estimation and mapping.