Comparison of different model error treatments and assimilation schemes in land surface temperature assimilation

In this study, land surface temperature (LST) is assimilated by using Common Land Model (CoLM) and Ensemble Kalman Filter (EnKF). To found the most reasonable method on model error treatment and remote sensed land surface temperature assimilation, some methods on the model ensemble generation and the construction of observe operators are compared. Though experiments show the two methods have similar result, forcing and parameter perturbation can generate ensemble more reasonable compared with perturbing only state variables, and is more easily to achieve in realistic assimilation. In observe operator comparison, by the component temperature decomposition method, the land surface temperature of remote sensing can update ground surface temperature in the CoLM model directly, which has more obvious physical meaning than other observe operator. A synthetic experiment also shows this method have the best result in the comparison.