Effects of sample size on sap flux-based stand-scale transpiration estimates.

In this study, we aimed to assess how sample sizes affect confidence of stand-scale transpiration (E) estimates calculated from sap flux (F(d)) and sapwood area (A(S_tree)) measurements of individual trees. In a Japanese cypress plantation, we measured F(d) and A(S_tree) in all trees (n = 58) within a 20 x 20 m study plot, which was divided into four 10 x 10 subplots. We calculated E from stand A(S_tree) (A(S_stand)) and mean stand F(d) (J(S)) values. Using Monte Carlo analyses, we examined the potential errors associated with sample sizes in E, A(S_stand) and J(S) using the original A(S_tree) and F(d) data sets. Consequently, we defined the optimal sample sizes of 10 and 15 for A(S_stand) and J(S) estimates, respectively, in the 20 x 20 m plot. Sample sizes larger than the optimal sample sizes did not decrease potential errors. The optimal sample sizes for J(S) changed according to plot size (e.g., 10 x 10 and 10 x 20 m), whereas the optimal sample sizes for A(S_stand) did not. As well, the optimal sample sizes for J(S) did not change in different vapor pressure deficit conditions. In terms of E estimates, these results suggest that the tree-to-tree variations in F(d) vary among different plots, and that plot size to capture tree-to-tree variations in F(d) is an important factor. The sample sizes determined in this study will be helpful for planning the balanced sampling designs to extrapolate stand-scale estimates to catchment-scale estimates.