Estimation of mesophyll conductance in <i>Ginkgo biloba</i> from the PSII redox state using a machine learning approach

Mesophyll conductance (gm) has been proved to be one of the important factors limiting photosynthesis and thus affects the estimation of plant productivity and terrestrial carbon balance. However, beyond the leaf scale, gm is usually assumed to be infinite because of the unavailability of the estimating technology. In this study, we first verified the important role of gm on photosynthesis by utilizing a wide range of ginkgo (Ginkgo biloba L.) families. Then, the dataset was adopted to establish a random forest-based gm estimation approach with the drivers being selected under the guidance of several mechanistic models (e.g. Farquhar, von Caemmerer, Berry model, the mechanistic light reaction model of photosynthesis). This model exhibited high predictive accuracy, utilizing both the measured fraction of open reaction centers in PSII (qL) (R2 = 0.71, RMSE = 0.008) and the estimated qL (R2 = 0.70, RMSE = 0.008) as inputs. Since qL, a key physiological driver in the model, can be obtained from chlorophyll fluorescence of PSII (SIFPSII) using the open-closed (OC) redox model of photosynthetic electron transport, this leaf-scale model could potentially be applied beyond the leaf scale, provided that environmental data are available. Direct measurements also confirmed the close relationship between qL and gm under ambient CO2 concentration and saturated light conditions. Our findings pave the way for additional attempts to estimate gm across a variety of scales.