Species with larger vessel area have higher bias for the original Granier equation in calculating sap flux density

Abstract

Accurate estimate of transpiration is crucial to understanding global water and carbon cycles. The thermal dissipation (TD) method with Granier’s equation, which is the most popular approach to measure whole-tree water use, requires species-specific calibration parameters to calculate transpiration. Calibrations require direct measurement of water uptake, often through gravimetric or potometric methods; therefore, it is often disregarded because of limited time and resources. It remains unclear which factor influences the variation of calibration parameters. In this study, we compiled published calibration experiments to elucidate the factors that could influence the variation in coefficients and attempted to predict the correction parameter. The results showed that: (1) only 14 % of the experiments provided accurate sap flux density estimates using the default parameter values in the Granier’s equation, and sap flux was underestimated to varying degrees in 72% of the experiments; (2) the regression slope of TD-measured and the actual sap flow with the y-axis intercept set to zero (hereinafter referred to as “Slope”) was significantly correlated to the mean vessel area (VA), i.e., species with larger VA consistently had higher Slope and greater bias. Stem diameter also had a positive influence on the Slope after accounting for covariation with the VA. Finally, a multiple linear regression model based on these two characteristics (VA and stem diameter) explained 89% of the variation of Slope. This study provides a simple calibration procedure for the users of TD-sensors with a linear correction model based on easily measurable vessel area and stem diameter. It may also provide a new development avenue for sensor manufacturers or others to improve the TD method to reduce its sensitivity to wood characteristics – or provide some “out-of-the-box” correction factors across a range of wood characteristics.