Despite the obvious physical connection between surface soil moisture conditions and antecedent rainfall, relatively little attention has been paid to integrating surface water balance information obtained from both spaceborne surface soil moisture and precipitation retrievals. This paper will develop two Kalman filter-based strategies for enhancing the accuracy of satellite-based rainfall retrievals over data-poor land areas using AMSR-E surface soil moisture retrievals. As a first step, both approaches will be evaluated over areas of the continental United States (CONUS) in which extensive ground-based precipitation resources are available as a benchmark.

The first approach is based on the application of an adaptive Kalman filtering strategy to the assimilation of AMSR-E soil moisture retrievals into a simple linear water balance model forced by a range of satellite-based precipitation products. Modeling uncertainties derived via tuning of Kalman filter error parameters - based on the statistical analysis of filtering innovations - will be compared with actual errors in precipitation forcing products to evaluate whether the approach is capable of reliably estimating the relative accuracy of various global precipitation products in the absence of ground-based rainfall observations. Results will be evaluated based on their potential for enhancing the calibration of rainfall retrieval algorithms over data-poor land areas

A second approach will be aimed at evaluating the degree to which discrete soil moisture corrections - dictated by the Kalman filter during the assimilation of AMSR-E soil moisture products into a satellite-rainfall-driven water balance model - can be used to accurately filter errors in TRMM rainfall products. The CONUS-wide correlation between such corrections (commonly referred to as “analysis increments”) and actual satellite rainfall errors will be evaluated, and a simple boot-strapping procedure for partitioning Kalman filter analysis increments into model and rainfall error contributions will be presented. The approach will be evaluated based on its potential for improving the relative accuracy of near-daily, real-time TRMM rainfall products.

Adaptation of the second approach over data-poor land areas has the potential to enhance hydrologic forecasting and monitoring applications by improving the accuracy of precipitation forcing data feed into hydrologic models. However, current hydrologic data assimilation systems are constrained in that remotely-sensed soil moisture is used only to correct internal model states and not model inputs. Prospects for the development of an alternative hydrologic data assimilation approach, capable of simultaneously correcting both rainfall inputs and internal soil moisture states, will be discussed.