Satellite and aircraft based microwave remote sensing of soil moisture

This dissertation includes research in three broad areas: (i) nature and behavior of the microwave signal under the conditions of sloping terrain and strong surface heterogeneity, (ii) algorithm validation and accuracy assessment of soil moisture products derived using passive and active techniques, and (iii) sensitivity analysis of radar derived products to soil moisture variability for potential use in synergistic passive-active temporal change detection disaggregation methodologies. The outlined research issues were addressed using several different soil moisture data sets (incl. station, aircraft and satellite) over three study areas located in SW US and NE Mexico [the North American Monsoon Experiment (NAME)], and SE Australia [the Goulburn (NAFE'05 domain) and the Murrumbidgee (NAFE'06) catchments, where NAFE stands for National Airborne Field Experiment].Remote sensing offers numerous advantages for soil moisture mapping over the traditional in situ or modeling approaches including high temporal frequency, global spatial coverage and frequent repeat. However, the coarse spatial resolution of the radiometer estimates and the lack of operational active algorithm for retrieval of absolute soil moisture are limiting factors in the development of fine resolution global soil moisture data sets. Furthermore, the accuracy of the available satellite-derived estimates is typically assessed in terms of agreement with point-collected station data. The strong surface heterogeneity combined with the dynamic nature of the soil moisture make the 'point' accuracy approach prompt to errors since it is largely controlled by the ground and moisture conditions observed at a single point. Therefore, this dissertation (i) evaluates the capabilities of temporal change detection methodologies for retrieval of relative soil moisture estimates using SAR observations and their potential for usage in disaggregation techniques, as well as (ii) examines the 'optimum' spatial resolution when down-scaling radiometer retrievals, and (iii) analyzes the effect of the spatial resolution of the evaluation data sets.The first part of this dissertation presents a topography -- vegetation correction algorithm developed based on a polynomial fitting procedure. The methodology was applied to the QuikSCAT observations at 2.225 km over the NAME (mountainous terrain and semi-arid types of vegetation). Comparison analysis between the un-corrected and the corrected backscatter data indicated that the sigma° sensitivity to soil moisture can be significantly improved if the topographic induced variability is adequately accounted for.Currently the only operational soil moisture product is provided by the Advanced Microwave Scanning Radiometer (AMSR-E). AMSR-E calibration and retrieval accuracy have been extensively studied since its launch on board NASA's Aqua platform in 2002. This dissertation builds on the previous work in the following ways: (i) algorithm validation and accuracy assessment of a C-band derived product (ii) assessment of AMSR-E sensitivity to changes in soil wetness under the conditions of extensive irrigation and, (iii) evaluation of the impact of standing water on AMSR-E retrieval accuracy.Soil moisture retrieval from active microwave observations is more challenging as compared to the passive case due to the more complex sensor -- ground surface interaction. Consequently accurate physical modeling of this interaction requires a large amount of ancillary data. The active retrieval methodology evaluated here was developed using a temporal change detection approach, and sigma° observations acquired from the Advanced Synthetic Aperture Radar Global (ASAR) Monitoring (GM) mode (1 km). The resulting ASAR GM relative soil moisture product and QuikSCAT backscatter data were further evaluated for their sensitivity to soil wetness at several spatial scales ranging from point to 25 km using three different data sets: station (point), aircraft [Polarimetric L-band Multi-beam Radiometer (PLMR), 1 km], and satellite (AMSR-E, 25 km). Despite the good temporal agreement achieved over SE Australia, spatial correlations exhibited strong spatial-scale dependence for both active systems.