Soil moisture background error covariance and data assimilation in a coupled land-atmosphere model

This study characterizes the space-time structure of soil moisture background error covariance and paves the way for the development of a soil moisture variational data assimilation system for the Noah land surface model coupled to the Weather Research and Forecasting (WRF) model. The soil moisture background error covariance over the contiguous United States exhibits strong seasonal and regional variability with the largest values occurring in the uppermost soil layer during the summer. Large background error biases were identified, particularly over the southeastern United States, caused mainly by the discrepancy between the WRF-Noah simulations and the initial conditions derived from the used operational global analysis data set. The assimilation of the Soil Moisture and Ocean Salinity (SMOS) soil moisture data notably reduces the error of soil moisture simulations. On average, data assimilation with space-time varying background error covariance results in 33% and 35% reduction in the root-mean-square error and the mean absolute error, respectively, in the simulation of hourly top 10 cm soil moisture, mainly due to implicit reductions in soil moisture biases. In terms of correlation, the improvement in soil moisture simulations is also observed but less notable, indicating the limitation of coarse-scale soil moisture data assimilation in capturing fine-scale soil moisture variation. In addition, soil moisture data assimilation improves the simulations of latent heat fluxes but shows a marginal impact on the simulations of sensible latent heat fluxes and precipitation.