An evaluation of necessary model complexity for accounting for radial variability in the upscaling of whole-tree transpiration.

Abstract

Accurate scaling of outer-xylem sap flux observations to whole-tree transpiration is highly dependent on the proper characterization of sap flow with radial depth. While radial variability of sap flow has been identified as a substantial source of error in estimating whole-tree transpiration, there has yet to be a universal adoption of a single scaling method. This study evaluated the temporal dependency of the radial profile of sap flux on seasonality and meteorological drivers and assessed the accuracy of six radial profile estimation methods of varying complexity in comparison with a robust multi-point estimation of sap flow. Of the six scaling approaches examined in this study, the use of a linear regression to compare outer flow rates with whole tree flow generated the least error when compared with the multi-point estimation (0.6% error). A cross validation demonstrated that statistical models remain accurate when applied to individuals not included in the training of the statistical models (mean adjusted R2 = 0.961). These findings suggest that future sap flux studies seeking to scale to whole-tree or stand level could robustly instrument a subset of trees with multiple radial depth measurement points to train one of the statistical models described in this study. These models can then be applied to additional trees instrumented with a single measurement point in the outer xylem, reducing the cost of deployment and potential sources of error.