Footprint-scale soil moisture spatial-temporal variability and implications for satellite validation

Soil moisture is an important boundary variable that controls land surface processes, including interactions with climate, weather, biogeochemical and ecological phenomena. Remote sensing will ultimately provide appropriate spatial coverage of moisture content fields for use in Earth system monitoring and research. However, considerable spatial variability in moisture content exists within remote sensing footprints, and errors in ground-based estimates of footprint means are largely unquantified. This research characterizes soil moisture variability within a number of aircraft- and satellite-footprint-scale sites and quantifies the related uncertainty in ground-based estimates of the footprint mean. The footprint sites are located in Iowa and Oklahoma, and are associated with the SGP97, SGP99, SMEX02, and SMEX03 experiments, which aimed to validate airborne (ESTAR, PSR) and space-borne (AMSR-E) microwave radiometers. In chapter 2, soil moisture spatial variability is characterized with varying surface wetness conditions across scales from 2.5 m to about 50 km. Ground-based soil moisture measurements from SGP and SMEX experiments are employed for the analysis. Results revealed a preferential pattern of soil moisture variability with mean moisture content and scale of the soil moisture field. In chapter 3, the scaling behavior of soil moisture variability with support scale was examined. The semi-variogram function of soil moisture was deployed to reproduce log variance versus log support. Results indicated the existence of a multi-scale nested correlation structure of soil moisture, which resulted in a piece-wise decaying pattern of log variance with increasing log support. In chapter 4, high-resolution soil moisture images from airborne radiometers were utilized to characterize more detailed features of footprint-scale soil moisture distributions. A mixture of Gaussian distributions was proposed to fit the observed footprint-scale soil moisture distributions and the benefits of the mixture model were discussed. In chapter 5, ground-based measurements of soil moisture were combined with the antecedent precipitation index (API) to improve the accuracy of validation data. Ground-based validation data from a limited number of samples were improved by utilizing the API with a Gaussian mixture model. However, direct upscaling of point measurements by correcting bias of the API did not improve the accuracy of the footprint-scale mean estimate. Implications of the results of this study and suggestions for future research are presented in chapter 6.