Enhanced runoff simulation with improved evapotranspiration accounting for vegetation response to climate variability

Rainfall-runoff simulation plays a crucial role in predicting high runoff events. Hydrological models that use potential evapotranspiration (PET) equations are biased in runoff simulation because they neglect the role of vegetation responses to environmental variables such as ambient CO₂ concentration, air temperature (T_a), net radiation (R_n), and vapor pressure deficit (VPD). The modification of Penman-Monteith PET (PET_PM) by incorporating vegetation response to environmental variables through stomatal conductance (g_s) leads to complexity and uncertainty. This study used a mixed generalised additive model (MGAM) to simulate g_s as a nonlinear function of environmental variables. By integrating MGAM derived g_s into PET_PM, the modified model, PET_MGAM was developed. Using data from three eddy covariance flux tower sites with different vegetation types, PET_MGAM outperformed PET_PM, showing higher Nash-Sutcliffe Efficiency (NSE) and Kling–Gupta efficiency (KGE) values for runoff simulations of the catchments associated with the flux towers. The results showed that PET_MGAM moderated the runoff underestimation simulated by PET_PM, under both wet and dry conditions. shapley additive explanations (SHAP) analysis highlighted the contribution of key environmental variables to PET estimation under wet and dry climates. PET_MGAM accounts for the interactive effects of CO₂, T_a, and VPD in a modified PET equation, leading to more accurate estimates of water balance components under wet and dry climate conditions.