Sub-canopy Soil Moisture Modeling in n-Dimensional Spectral Feature Space

This paper attempts to quantify soil moisture in various canopy cover conditions using n-dimensional spectral signatures, including land surface temperature, vegetation index, albedo, and others. First, the feature vector between the pixels and various moisture contents was indentified. Normalization of the varying distance from a user-defined initial state to any pixel location, and coefficients related with n-dimensional spectral feature space were calculated, assigning weights to each parameter. Then, a soil moisture index was developed using a linear combination of the first order polynomials. The Extended Fourier Amplitude Sensitivity Test (eFAsT) was used to calculate the relative variance contribution of model input parameters to the variance of soil moisture predictions. Results derived from satellite data including Enhanced Thematic Mapper Plus (ETM+) and the Moderate Resolution Imaging Spectroradiometer (worms) imagery demonstrated significant correlations between the index and soil moisture obtained for different ecosystems and vegetation cover. The best agreement, the coefficient of determination (R-2), between the index and soil moisture were 0.58 and 0.65 for ETM+ and MODIS data, respectively. eFAST sensitivity analysis indicates that land surface temperature might be the most determinant factor in soil moisture estimation, then albedo, followed by NDVI.