A comparison of sap flux-based evapotranspiration estimates with catchment-scale water balance

Many researchers are using sap flux to estimate tree-level transpiration, and to scale to stand- and catchment-level transpiration; yet studies evaluating the comparability of sap flux-based estimates of transpiration (Et) with alternative methods for estimating Et at this spatial scale are rare. Our ability to accurately scale from the probe to the tree to the watershed has not yet been demonstrated, nor do we know the relative impact of the main sources of variability on our scaled estimates. Accounting for the variability in the radial distribution of sap flux within the sapwood, the variability of transpiration among trees and between plots within the catchment, and the variability in stand density, sapwood area, and leaf area are critical for making landscape inferences about transpiration. During 2004 and 2005, we continuously monitored 40 trees in three plots within a 13.5-ha gauged watershed comprising a 50-year-old eastern white pine plantation within the Coweeta Basin in western North Carolina, USA. We scaled sap flux-based estimates of stand transpiration (Et) and surface area-based estimates of stand interception (Ei) to the catchment and compared these with water balance estimates of evapotranspiration (E, precipitation minus runoff, P − Ro). For both years, the sum of sap flux scaled Et and Ei were 14 and 7% lower than evapotranspiration estimated from P − Ro. Our results show that a considerable amount of variation exists at each scaling step encountered; however, a simple scaling exercise revealed that omitting among plot variation affected the sap flux scaled Et estimate by 48%. Thus, the largest source of variability in scaling to the landscape was landscape variation in stand density and sapwood area.