Toward a Calibration-Free Model for Optical Remote Sensing of Soil Moisture

Abstract:

A recently developed physically-based model to retrieve soil moisture from optical images was evaluated in this study. The model was derived based on a simple two-flux radiative transfer model describing diffuse reflectance from a uniform, optically thick, absorbing and scattering medium. The model exhibited an unprecedented linear relationship between a novel transformed reflectance and the surface soil moisture in the shortwave infrared bands such as bands 6 and 7 of Landsat 8. Accuracy of the model was tested based on laboratory-measured spectral reflectance data of a broad range of Arizona soils in the optical domain (400 – 2500 nm). Additionally, the original model was further simplified by combining bands 6 and 7 data which reduced the number of model parameters from two to one. The remaining physically-significant parameter was directly measured for the Arizona soils, exhibiting little variability among those varied soil textures. New findings in this study significantly advance this new method toward its application without the need for ground-based model calibration. Further study of potentials and limitations of this model for large-scale application using optical satellite data (e.g. Landsat, MODIS) remains a goal of future research.