Submitted to: Journal of Hydrometeorology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/23/2010
Publication Date: 12/1/2010
Publication URL: http://handle.nal.usda.gov/10113/56317
Citation: Mirrales, D., Crow, W.T., Cosh, M.H. 2010. A technique for estimating spatial sampling errors in coarse-scale soil moisture estimates derived from point-scale observations. Journal of Hydrometeorology. 11:1423-1429. Interpretive Summary: Satellite-based surface soil moisture estimates offer the potential to aid in a variety of key agricultural water resource applications including drought monitoring, irrigation scheduling and accurate rainfall prediction. However, difficulties associated with evaluating spaceborne soil moisture estimates form a major technical barrier to their successful implementation in these areas. Soil moisture estimates obtained from ground-based instrumentation can be used for verification of corresponding satellite estimates - however the spatial scales between ground-based observations (essential 1 to 5 cm) is several orders of magnitude smaller than the spatial resolution of the satellites (typically > 10 km). Consequently, attempts to validate (and therefore improve) satellite measurements are frequently thwarted by large sampling errors encountered when trying to use point-scale ground observations to characterize satellite-scale soil moisture estimates. This paper describes a statistical technique to estimate the magnitude of these sampling errors. Once accurately estimated, these sampling errors can be properly accounted for in evaluation strategies. Consequently, implementation of this technique will improve our ability to optimize the utility of satellite-based soil moisture retrievals for critical agricultural and water resource applications.
Technical Abstract: The validation of satellite surface soil moisture retrievals requires the spatial aggregation of point-scale ground soil moisture measurements up to coarse resolution satellite footprint scales (>10 km). In regions containing a limited number of ground measurements per satellite footprint, a large component of the total observed difference between satellite retrievals and ground based soil moisture estimates is attributable to spatial sampling error in aggregated ground values. Such error hampers the validation of satellite soil moisture estimates by obscuring the direct estimation of their intrinsic error. So-called “Triple Collocation" (TC) procedures offer a viable solution to this up-scaling problem by accurately approximating root-mean-square error in three independently-acquired estimates of soil moisture. Here, we apply TC to coarse-scale soil moisture data products extracted from passive microwave remote sensing, land surface modeling and low density ground-based observations with the goal of explicitly estimating the spatial sampling uncertainty of coarse-scale soil moisture estimates derived from ground observations. Based on comparisons with very dense ground-based soil moisture networks, the TC approach is able to estimate the magnitude of point-to-footprint spatial sampling errors to within 0.0064m3m-3 and substantially improve our ability to validate remotely-sensed soil moisture retrievals using sparse ground-based observations.