A quasi-global evaluation system for satellite-based surface soil moisture retrievals

Authors: Crow, Wade; MIRRALES, D; Cosh, Michael

Submitted to: IEEE Transactions on Geoscience and Remote Sensing
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 12/28/2009
Publication Date: 6/10/2010
Citation: Crow, W.T., Mirrales, D.G., Cosh, M.H. 2010. A quasi-global evaluation system for satellite-based surface soil moisture retrievals. IEEE Transactions on Geoscience and Remote Sensing. 48(6):2516-2527.

Interpretive Summary: Estimates of surface soil moisture derived from satellite observations have value for a range of agricultural applications including: drought monitoring, crop yield forecasting, and long-term precipitation forecasting using numerical weather prediction models. Unfortunately, a severe lack of accurate ground-based soil moisture observations has hindered the thorough testing of surface soil moisture retrieval algorithms. Without such testing, it is very difficult to objectively compare two competing retrieval approaches and learn what approach works best. This uncertainty has - in turn - limited the development of key agricultural applications for this data. Here, we apply a novel evaluate approach (based on a data assimilation mathematical framework and ancillary rainfall observations) that allows for the objective inter-comparison of competing soil moisture retrieval strategies over a much wider spatial and temporal domain than was previously possible. Global application of the approach will make it much easier to objectively test remotely-sensed soil moisture products and therefore aid in the development of key water-resource applications for this type of remote sensing data.

Technical Abstract: To date, limitations in the availability of in situ surface soil moisture observations have restricted the validation of remotely-sensed soil moisture products to a small number of heavily-instrumented watershed sites. A recently developed data assimilation technique offers the potential to greatly expand the geographic domain over which retrievals can be evaluated by effectively substituting (relatively plentiful) rain gauge observations for (less commonly available) ground-based soil moisture measurements. The technique is based on calculating the Pearson correlation coefficient (Rvalue) between rainfall errors and Kalman filter analysis increments, realized during the assimilation of a remotely-sensed soil moisture product into the Antecedent Precipitation Index (API). Here the existing Rvalue approach is modified by 1) reformulating it within an anomaly space where long-term seasonal trends are explicitly removed and 2) basing the calculation of API analysis increments on a Rauch-Tung-Striebel smoother instead of a Kalman filter. This reformulated approach is then applied to a number of Advanced Microwave Scanning Radiometer soil moisture products acquired within three heavily-instrumented watershed sites in the Southern United States. Rvalue -based evaluations of soil moisture products within these sites are verified based on comparisons with available ground-based soil moisture measurements. Results demonstrate that, without access to ground-based soil moisture measurements, the Rvalue methodology can accurately mimic anomaly correlation coefficients calculated between remotely-sensed soil moisture products and soil moisture observations obtained from dense ground-based networks. Sensitivity results indicate that the predictive skill of the Rvalue metric is enhanced by both proposed modifications to its methodology. Finally, using this newly verified approach, Rvalue calculations are expanded to a quasi-global (50 degree S to 50 degree N) scale using only rainfall products derived from the Tropical Rainfall Measurement Mission Precipitation Analysis. Results demonstrate a generally positive trend of retrieval value with decreasing microwave brightness temperature frequency and refine expectations concerning the global distribution of land areas in which remotely-sensed surface soil moisture retrievals can contribute to climate forecasting applications.