Quantifying contributions of uncertainties in physical parameterization schemes and model parameters to overall errors in Noah-MP land modeling using observations from eddy flux sites

Quantifying contributions of errors in model structure and model parameters to biases in a land surface model (LSM) is critical for model improvement, but has not been done systematically for many global land surface models. This paper investigates the uncertainties in the Noah with multiparameterization (Noah-MP) LSM with dynamic vegetation by examining the interactions between imperfect parameterization schemes (PSs) and improper parameter values (PVs). A number of Noah-MP physical ensemble simulations were conducted at 92 eddy flux sites to quantitatively assess the impacts of the PS uncertainties on model performance, and then the key parameters in the two combinations of schemes with significant differences were calibrated. The results show that five subprocesses—the surface exchange coefficient (SFC), soil moisture threshold, radiation transfer (RAD), runoff and groundwater, and surface resistance to evaporation—have the most significant influence on the performances of simulated sensible heat flux, latent heat flux, net absorbed radiation and gross primary productivity in the Noah-MP LSM with dynamic vegetation, and that the interaction between SFC and RAD contributed up to 80% of the variation in the model performance at some sites. It is also shown that tuning the PSs and optimizing the PVs should be jointly applied to reduce the errors in the Noah-MP LSM, although compared to tuning PSs, parameter optimization happens to make less robust model improvement. Finally, this study emphasizes that reducing the significant uncertainties in soil parameters and exploring the errors caused by missing physical features are crucial to improving LSMs with dynamic vegetation.