The sensitivity and optimization of the model parameters for the simulation of latent heat flux

The land surface model (LSM) is very complex, and its parameters are one of the key sources of model prediction error. In order to improve the ability of the simulation, the study addressed the parameter sensitivity analysis, optimization and its effect on the simulated latent heat fluxes from the Noah LSM. The LH-OAT sensitivity analysis method was conducted for the Noah LSM to assess the sensitivity of model prediction of the latent heat flux to various model input parameters, the dominant vegetation and soil parameters were determined. Then, the model parameters are estimated from Noah LSM with a Bayesian framework by applying the Shuffled Complex evolution Metropolis Algorithm(SCEM-UA), an efficient Markov Chain Carlo sampler. The Noah model prediction using the optimal parameters shows that the LE simulated latent heat fluxes matched measurements fairly well with an R² value of 0.9394, Root Mean Squared Error (RMSE) of 36.77W/m2, and mean bias error(MBE) of -1.17W/m². Results demonstrate the ability of the combination LH-OAT method and the SCEM-UA algorithm for sensitivity analysis and parameter optimization in the Noah land surface model.