Contribution of sapwood traits to uncertainty in conifer sap flow as estimated with the heat-ratio method

• We measured sapwood water content, density, and depth in five conifer species.
• We determined the effect of trait variability on whole-tree sap flow estimates.
• Use of a fixed sapwood thermal diffusivity may overestimate conifer sap flow.
• Sapwood trait data that are not tree-specific impart errors sap flow estimates.
• Uncertainty of sap flow estimates should reflect error in sapwood traits.

Inferring whole-tree sap flow rates (Q) with thermometric sap flow sensors requires specification of physiological and structural attributes of trees. Using sap temperature measurements to estimate Q with the heat-ratio method (HRM) requires quantification of the water content (mc), basic density (ρb), and depth (Rs) of sapwood. Values of mc and ρb serve to estimate sapwood thermal diffusivity (k), a necessary variable in the calculation of heat-pulse velocity (Vh) that is often set to a nominal value (k_nom); mc and ρb are also used to convert Vh to sap velocity (Vs). The sapwood area across which Vs is integrated is often estimated on the basis of Rs. Because mc and ρb are correlated and influence Q through estimation of k and the conversion of Vh to Vs, we sought to quantify the potential error introduced when Q is calculated with k_nom rather than with k as estimated from measurements of mc and ρb in five coniferous species. We also examined how variation in mc, ρb, and Rs across sampling scales may contribute to uncertainty in estimated Q. Across the observed range of mc and ρb, the two traits contributed to a net decline in the process of estimating Vs from Vh (with k_nom). This suggests that the use of k_nom rather than a calculated k may result in overestimation of Q when mc and ρb co-vary as they did in our study. Variability in mc and ρb across sampling scales could induce errors greater than 10% in Vs (and hence Q), while within- and among-tree variability in Rs could impart even greater errors (up to 130%). We propose that this uncertainty be represented in the expression of error in whole-tree sap flow estimates and in statistical analyses involving those estimates.