Evaluating the vegetation-atmosphere coupling strength of ORCHIDEE land surface model (v7266)

Plant transpiration dominates terrestrial latent heat fluxes (LE) and plays a central role in regulating the water cycle and land surface energy budget. However, Earth system models (ESMs) currently disagree strongly on the amount of transpiration, and thus LE, leading to large uncertainties in simulating future climate. Therefore, it is crucial to correctly represent the mechanisms controlling the transpiration in models. At the leaf scale, transpiration is controlled by stomatal regulation, and at the canopy scale, through turbulence, which is a function of canopy structure and wind. The coupling of vegetation to the atmosphere can be characterized by the coefficient omega. A value of omega & RARR;0 implies a strong coupling of vegetation and the atmosphere, leaving a dominant role to stomatal conductance in regulating water (H2O) and carbon dioxide (CO2) fluxes, while omega -> 1 implies a complete decoupling of leaves from the atmosphere, i.e., the transfer of H2O and CO2 is limited by aerodynamic transport. In this study, we investigated how well the land surface model (LSM) Organising Carbon and Hydrology In Dynamic Ecosystems (ORCHIDEE) (v7266) simulates the coupling of vegetation to the atmosphere by using empirical daily estimates of omega derived from flux measurements from 90 FLUXNET sites. Our results show that ORCHIDEE generally captures the omega in forest vegetation types (0.27 +/- 0.12) compared with observation (0.26 +/- 0.09) but underestimates omega in grasslands (GRA) and croplands (CRO) (0.25 +/- 0.15 for model, 0.33 +/- 0.17 for observation). The good model performance in forests is due to compensation of biases in surface conductance (Gs) and aerodynamic conductance (Ga). Calibration of key parameters controlling the dependence of the stomatal conductance to the water vapor deficit (VPD) improves the simulated Gs and omega estimates in grasslands and croplands (0.28 +/- 0.20). To assess the underlying controls of omega, we applied random forest (RF) models to both simulated and observation-based omega. We found that large observed omega are associated with periods of low wind speed, high temperature and low VPD; it is also related to sites with large leaf area index (LAI) and/or short vegetation. The RF models applied to ORCHIDEE output generally agree with this pattern. However, we found that the ORCHIDEE underestimated the sensitivity of omega to VPD when the VPD is high, overestimated the impact of the LAI on omega, and did not correctly simulate the temperature dependence of omega when temperature is high. Our results highlight the importance of observational constraints on simulating the vegetation-atmosphere coupling strength, which can help to improve predictive accuracy of water fluxes in Earth system models.