|Bryant, R. - UNIVERSITY OF ARIZONA|
|Rahman, M. - UNIVERISTY OF ARIZONA|
|HOLIFIELD COLLINS, CHANDRA|
|Sano, E. - EMPRAPA, BRAZIL|
|Slocum, K - ARMY TEC|
Submitted to: Water Resources Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: October 1, 2005
Publication Date: November 1, 2005
Citation: Thoma, D., Moran, M.S., Bryant, R., Rahman, M.M., Holifield Collins, C.D., Skirvin, S.M., Sano, E.E., Slocum, K. 2005. Comparison of four models for determining surface soil moisture from c-band radar imagery. Water Resources Research 42(1): 1-12. Interpretive Summary: Accurate estimates of soil moisture across broad landscapes would be useful for predicting rangeland health, developing cropping strategy, determining cross country vehicle mobility, and could aid decision making in a broad range of natural resource management programs. Ground based measurements of soil moisture though accurate are time consuming and difficult to obtain over large areas and automated sensors are expensive. Currently orbiting radar satellites offer an alternative based on their sensitivity to changes in soil moisture. This study accomplished an evaluation of four mathematical models based on their ability to predict surface soil moisture from radar satellite imagery in semi-arid Arizona rangelands. Evaluation of the models was made by comparing model-estimated soil moisture from satellite images to soil moisture measured in the field at the time satellite images were acquired. The most significant contribution of this study was demonstration of a new model that had an accuracy of (91%), can be used globally in any sparsely vegetated landscape, and is relatively simple to implement.
Technical Abstract: Four models for deriving estimates of near-surface soil moisture from radar imagery in a semi-arid rangeland were evaluated against in situ measurements of soil moisture. The models were based on empirical, physical, semi-empirical and image-difference approaches. The Integral Equation Method (IEM) model was used in both the physical and semi-empirical approaches. In all cases spatial averaging to the watershed scale improved agreement between radar backscatter or model output and observed soil moisture. Variation in radar backscatter explained 85% of the variation in observed soil moisture at the watershed scale. The rmse between modeled and observed soil moisture was 0.13 and 0.04 for the physical and best semi-empirical adjustment to the physical model, respectively. The newly proposed delta index model explained 91% of the variability in soil moisture with rmse = 0.04. Results show that the delta index is scaled to the range in observed soil moisture and indicate it may work as a purely image-based model in any sparsely vegetated environment that has time invariant roughness. Because it is image-based, the delta index implicitly accounts for surface roughness, topography, and vegetation. These are land surface variables required by most physical models that are often difficult to obtain over large geographic areas.