A better simulation of water and carbon fluxes in a typical desert grassland ecosystem through the Common Land Model

How to improve the accuracy and fit-for-purpose of Land surface models (LSMs) in simulating carbon and water fluxes in arid regions, is still a great challenge, since the latent heat flux and terrestrial respiration are underestimated. In this study, we first modified the root water uptake, soil respiration, and stomatal conductance processes and introduced a Bayesian parameter optimization algorithm to optimize the newly introduced fitting parameters. Then, the performance of the calibrated model was evaluated by using eddy covariance observations at a typical desert grassland station. The results showed that the modified soil respiration model better captures the seasonal pattern of terrestrial ecosystem respiration (TER), with the distance between simulated and observed indices (DISO) value decreased from 1.45 to 0.95. Additionally, modifying the root water uptake process can noticeably enhance the simulation of water flux (i.e., latent heat flux) with DISO value decreased from 1.43 to 1.28, but uncertainties remain in simulating carbon flux (i.e., gross primary productivity). Through evaluating three stomatal conductance models under different levels of water stress, we found that both BallWoodrow-Berry (BWB) and Ball-Berry-Leuning (BBL) models could improve the CoLM performance in simulating carbon and water fluxes due to their effective response to the water stress with DISO values decreased from 1.96 to 1.68 and 1.67, respectively, while the Unified Stomatal Optimization (USO) model has a relatively poor simulation performance, which fails to capture the impact of water stress with DISO value increased to 2.13. Our study provides valuable insights for improving the simulations of carbon and water fluxes in arid regions by using CoLM.