A hybrid transpiration model for water-limited conditions

Abstract

Soil water deficit is considered the dominant environmental constraint for plant transpiration (and photosynthesis) under water-limited conditions. Proper representation of soil water stress is thus critical for sound performance of both process-based and empirical models for transpiration simulation (e.g., process-based BTA-ψ and empirical modified Jarvis-Stewart model, hereafter MJS). However, very often process-based models suffer from a lack of data (such as water potential for the BTA-ψ model); empirical models are difficult to transfer the optimal form of environmental stress functions across sites. In this study, a hybrid model is proposed to address these two limitations. The model is a combination of the BTA model and an empirical function of volumetric soil water content (θ), and referred to as BTA-θ. The BTA-θ model is compared against the BTA and MJS models regarding their capability in simulating transpiration under subtropical humid climate zone and Mediterranean climate conditions at both daily and hourly resolutions. Three different water stress functions were adopted for BTA-θ and MJS to test the transferability of optimal response function across species and climatic zones. Overall, BTA-θ estimated transpiration reliably with the Nash-Sutcliffe coefficient of efficiency being larger than 0.5 at two sites for both wet and dry periods and outperformed BTA significantly under various soil moisture conditions. BTA-θ with different water stress functions performed comparably for each of the two climatic zones. The various forms of water stress function in BTA-θ had negligible effect on parameterization of the BTA equation. However, the MJS model constructed with three water stress functions performed variably and had a big influence on parameterization of other stress functions. These results suggest that the hybrid BTA-θ model is superior to the process-based BTA model in simulating transpiration under different levels of soil water deficit and is more robust than the purely empirical MJS model in selecting appropriate stress functions. BTA-θ provides a structure to incorporate other plant water stress data for transpiration modelling.