Validation and Assessment of Model-Simulated Soil Moisture in the North American Land Data Assimilation System Phase 2

Accurate soil moisture is a critical initial condition in numerical weather and climate prediction systems because it partitions the surface energy and water fluxes into different parts, and further affects atmospheric boundary layer circulation and local convection activities. Since in-situ soil moisture observations are limited both in time and space, the soil moisture products from offline land surface models driven by the observed precipitation and/or radiation fluxes often serve as alternatives for initial conditions of weather and climate models. Therefore it is very important to make a comprehensive evaluation and assessment on such products. In this study, the 31-year (1 January 1979 to 31 December 2009) soil moisture simulated from four state-of-the-artland models (NCEP/Noah, NASA/Mosaic, OHD/SAC, Princeton and Washington/VIC) during the North American Land Data Assimilation System (NLDAS) phase 2 experiment has been evaluated using two observed datasets (Illinois soil moisture databank and Oklahoma Mesonet soil moisture dataset). Illinois databank is used to evaluate simulated soil moisture at monthly and annual time scales, and the Oklohma dataset is used to assess simulated soil moisture at daily time scales using anomaly correlation, root mean square error and bias. The results show that although the performance of four models varies, all of them have demonstrated the ability to simulate soil moisture well for different time scales.