Characterizing the heterogeneity of eastern hemlock xylem water isotopic compositions: Implications for the design of plant water uptake studies

Abstract

AbstractXylem water isotopic compositions (2H, 18O; δXYLEM) can be used to estimate plant water uptake depths; however, environmental heterogeneity in these measurements may prevent reaching robust conclusions. Bayesian mixing models used to estimate plant water uptake depths often assume that measurements of δXYLEM and candidate water uptake sources are normally and identically distributed. We tested if δXYLEM measured across 30 Eastern hemlock (Tsuga canadensis (L.) Carrière) trees met these assumptions. Bootstrap simulations suggested that the distributions of hemlock δXYLEM data were non‐normal in March, April, June, and July and that between 15 and 26 hemlock δXYLEM samples were required to reject the assumption of normality. In June, July, and August, δXYLEM was significantly predicted by a multivariate linear regression with tree sapwood depth or elevation, rejecting the assumption of independently distributed observations. A comparison of dry season hemlock water uptake depth estimates between a Bayesian mixing model and a process‐based ecohydrological model calibration showed differences, with the Bayesian model estimating a substantially greater proportion of shallow water uptake. These results highlight the need for standardized field sampling protocols for δXYLEM and analytical methods that will lead to more robust estimates of plant water uptake depths. These findings also suggest that water uptake functions conditioned on landscape and tree structural variables could substantially advance the representation of plants in ecohydrological models.