Incorporating dynamic schemes of canopy light extinction coefficient improves transpiration model performance for fruit plantations

The light extinction coefficient (C) describes the efficiency of light interception by a plant canopy, and is a key parameter commonly used to partition radiant energy between a vegetation canopy and the soil surface in several important transpiration models. The C has been observed to vary considerably in the field, but is often parametrized as a fixed constant value, resulting in large potential uncertainty in transpiration estimation. We hypothesized that C changes with leaf area index (LAI) and environments, and that accurate characterization of the C variation pattern would improve transpiration modelling. To test these hypotheses, the relationships between measured C and potential influencing factors were explored for three different fruit plantations in different climate zones in China. Based on these relationships, a new method was proposed to capture C dynamics that considered both LAI and environmental factors. Additionally, 16 C schemes (15 dynamic C schemes and one fixed C scheme) were individually embedded in the Penman-Monteith model, and comprehensive comparisons were made to determine the most effective C scheme for simulating transpiration based on a Bayesian framework. Results showed that C displayed clear variations across all sites, and that there were significant relationships of C with LAI and environmental factors. Compared with the model using a fixed C scheme, the model using the best dynamic C scheme significantly improved model performance by increasing estimation accuracy and decreasing model uncertainty. Specifically, the average coefficients of determination (R2) increased from 0.35 to 0.76, 0.82 to 0.85, and 0.90 to 0.93 for the orange orchard, the jujube orchard, and the grape greenhouse, respectively. Additionally, the average mean relative errors (MRE) decreased by 54.99%, 26.62%, and 13.95%, respectively, and the average uncertainty band widths (B) decreased by 46.81%, 21.74%, and 11.43%, respectively. Model performance improvement was more significant at sites with more serious environmental stresses. The results of our study emphasize the necessity for considering dynamic C patterns in water consumption estimation, particularly for regions facing severe abiotic stresses that are likely to occur in many regions under future climate conditions.