An improved representative of stomatal models for predicting diurnal stomatal conductance at low irradiance and vapor pressure deficit in tropical rainforest trees

The predictions of stomatal conductance (g sw ) by the unified stomata optimization (USO) series models in tropical rainforest trees exhibited pronounced biases. However, little attention has been devoted to the structural issues within the USO series models themselves. This study introduced a novel approach by integrating the Farquhar photosynthesis model with the random forest (RF) algorithm to investigate diurnal variations in leaf stomatal responses among six tropical tree species in South China. The results revealed that the USO model and its derivative significantly overestimated g sw when vapor pressure deficit (VPD) and irradiance were low. The overestimation was primarily attributed to the assumption of a linear relationship between g sw and net assimilation rate (A n ) and the issue of unbounded g sw when VPD was low. The relationship between g sw and A n was indeed non-linear due to a negative correlation between the intercellular: atmospheric CO 2 concentration ratio (C i /C a ) and irradiance. The relationship between C i /C a and irradiance indicated that C i /C a was higher at low irradiance, declined and tended to gradually stabilize at high irradiance. An empirical coefficient was determined by using the monthly mean of daytime minimum VPD to represent g sw as finite values at low VPD. The revision achieved a substantial improvement in predictive accuracy of g sw at low VPD while preserving g sw responsiveness under high VPD.