Prediction of Turbulent Heat Fluxes by Assimilation of Remotely Sensed Land Surface Temperature and Soil Moisture Data into an Ensemble-Based Data Assimilation Framework

Accurate estimation of turbulent heat fluxes is important for water resources planning and management, irrigation scheduling, and weather forecast. Land surface models (LSMs) can be used to simulate turbulent heat fluxes over large-scale domains. However, the application of LSMs is hindered due to the high uncertainty in model parameters and state variables. In this study, a dual-pass ensemble-based data assimilation (DA) approach is developed to estimate turbulent heat fluxes. Initially, the common land model (CoLM) is used as the LSM (open-loop), and thereafter the ensemble Kalman filter is employed to optimize the CoLM parameters and variables. The first pass of the DA scheme optimizes vegetation parameters of CoLM (which are related to the leaf stomatal conductance) on a weekly-basis by assimilating the MODIS land surface temperature (LST) data. The second pass optimizes the soil moisture state of CoLM on a daily-basis by assimilating soil moisture observations from Cosmic-ray instrument. The ultimate goal is to improve turbulent heat fluxes estimates from CoLM by optimizing its vegetation parameters and soil moisture state via assimilation of LST and soil moisture data into the proposed DA system. The DA approach is tested over a wet and densely vegetated site, called Daman in northwest of China. Results indicate that the CoLM (open-loop) model typically underestimates latent heat flux and overestimates sensible heat flux. By assimilation of LST in the first pass, the turbulent heat fluxes are improved compared to those of the open-loop. These fluxes become even more accurate by assimilation of soil moisture in the second pass of the DA approach. These findings illustrate that the introduced DA approach can successfully extract information in LST and soil moisture data to optimize the CoLM parameters and states and improve the turbulent heat fluxes estimates.