Evaluating the Impact of Assimilating Soil Moisture Variability Data on Latent Heat Flux Estimation in a Land Surface Model

Accurate specification of the soil moisture in land-surface models has the potential to improve the evapotranspiration estimates from these models. However, soil moisture is highly variable in space and time due to variability in climatic, topographic, vegetative, and soil properties. It is anticipated that including information on soil moisture variability into a land-surface model will improve model estimates of evapotranspiration. In this experiment, the spatial variability of soil moisture was measured over an agricultural field in southern Ontario over a 70 m × 70 m area ten times during the growing season. These data were used to update the Canadian Land Surface Scheme (CLASS) using three assimilation techniques. The techniques evaluated included two versions of ensemble Kalman filter (EnKF), and direct insertion of soil moisture data into the model. The results showed that assimilating observed soil moisture variability into CLASS improves model latent heat flux estimates by up to 14%. The amount of improvement depends on the method and timing of assimilation. The effect was largest at the beginning of the growing season, while it was smallest at peak growth. Application of EnKF, considering both instrumental error and field variability, resulted in greater improvement in latent heat flux estimates compared to the other two methods. This study showed that assimilation of soil moisture variability into CLASS can result in greater improvement in modelled ET comparing to assimilating of the mean of the sampling area.