Evaluating CO₂ effects on semi-empirical and empirical stomatal conductance simulation in land surface models

Ongoing changes in climate and carbon dioxide (C_s) in the atmosphere have profound effects on plant transpiration and, consequently, on the water balance. Land surface models (LSMs) reflect plant response to these changes by simulation of stomatal conductance (g_s). However, the plant response is not well understood and varies with climate. In this study, the simulation of g_s within different LSMs is reviewed and a new approach, a Mixed Generalized Additive Model (MGAM) for g_s simulation, is developed. The alternative g_s estimation is proposed as a solution for the high parameterisation uncertainty in semi-empirical g_s simulation models, and high dependency on mathematical functions for environmental stress factors in empirical g_s simulation models. MGAM has high Pearson and Spearman correlations (87% and 85%, respectively) and efficiency coefficients (71%), with low error values (0.07 mol/m²s) in g_s simulation. The global sensitivity analysis of the MGAM approach shows the necessity of considering the interaction between C_s and other key climate variables in g_s simulation. The high accuracy and low uncertainty to the first-order key climate factors in g_s simulation highlight the MGAM model's importance in future studies.