Remote sensing observatory validation of surface soil moisture using Advanced Microwave Scanning Radiometer E, Common Land Model, and ground based data: Case study in SMEX03 Little River Region, Georgia, US

Optimal soil moisture estimation may be characterized by intercomparisons among remotely sensed measurements, ground-based measurements, and land surface models. In this study, we compared soil moisture from Advanced Microwave Scanning Radiometer E (AMSR-E), ground-based measurements, and a Soil-Vegetation-Atmosphere Transfer (SVAT) model for the Soil Moisture Experiments in 2003 (SMEX03) Little River region, Georgia. The Common Land Model (CLM) reasonably replicated soil moisture patterns in dry down and wetting after rainfall though it had modest wet biases (0.001-0.054 m(3)/m(3)) as compared to AMSR-E and ground data. While the AMSR-E average soil moisture agreed well with the other data sources, it had extremely low temporal variability, especially during the growing season from May to October. The comparison results showed that highest mean absolute error (MAE) and root mean squared error (RMSE) were 0.054 and 0.059 m(3)/m(3) for short and long periods, respectively. Even if CLM and AMSR-E had complementary strengths, low MAE (0.018-0.054 m(3)/m(3)) and RMSE (0.023-0.059 m(3)/m(3)) soil moisture errors for CLM and soil moisture low biases ( 0.003-0.031 m(3)/m(3)) for AMSR-E, care should be taken prior to employing AMSR-E retrieved soil moisture products directly for hydrological application due to its failure to replicate temporal variability. AMSR-E error characteristics identified in this study should be used to guide enhancement of retrieval algorithms and improve satellite observations for hydrological sciences.