Résumé
Land surface models (LSMs) typically represent soil moisture control on stomatal conductance through an empirical sensitivity function, without considering plant hydrology. This study proposes integrating water transfer representation within the soil-plant-atmosphere continuum in the ORCHIDEE land surface model. This new configuration includes vegetation hydraulic architecture and a stomatal control based on leaf water potential ψ leaf ) , along with a mechanistic representation of water absorption by roots via radial diffusion around the roots. An adaptive numerical scheme is implemented to prevent numerical instabilities during hydric stress, reducing hourly instabilities by a factor of 2. The implementation and the standard configuration of ORCHIDEE are calibrated and evaluated at FLUXNET sites with eddy-covariance flux measurements. A detailed assessment is carried out at two well-documented forest sites (FR-Hes and FR-Pue), where both configurations perform similarly regarding the seasonal dynamics of latent heat flux (RMSEs of 16.0 W/m 2 for the potential-based configuration and 15.8 W/m 2 for the standard configuration at FR-Hes). An evaluation of leaf water potential at FR-Pue shows correlations of 0.87 and 0.72 for predawn and midday ψ leaf respectively. A second evaluation across 135 sites from the FLUXNET2015 database highlights similar performances for both configurations. Finally, a global assessment of the differences between the two schemes emphasizes the good performance of the hydraulic architecture model. Overall, the new hydraulic architecture provides a more mechanistic description of stomatal conductance response to soil water stress and paves the way for incorporating physiological processes controlling tree mortality and using in situ observations to calibrate plant responses to water stress.<p>Plain Language Summary Land surfaces models (LSMs), which aim at representing the exchanges between the land surfaces and the atmosphere, usually represent the rate of water exchanges between the vegetation and the atmosphere according to soil water stress defined as an empirical function and without considering the plant water status. Here, we propose to represent the water transfers from the soil-root interface toward the leaves via a hydraulic architecture model, accounting for potential water storage in the plant. This model provides a more mechanistic description of the response of stomatal conductance to soil water stress and opens for the representation of important physiological processes such as cavitation. In order to evaluate the model, a detailed analysis has been performed at two forest sites. It enabled to understand the behavior of the new configuration and to verify its capacity to reproduce observations of water exchanges with the atmosphere. An extended analysis over 135 FLUXNET sites illustrated the overall good performances of the model. Finally, a study at global scale highlighted significant differences between the new hydraulic architecture and the standard model with respect to energy, water and carbon fluxes