Figure 3. Distribution of mean δD biases per tree species. The error bars correspond to minima and maxima. Letters indicate significant differences between species based on a Dunn-Sidák test (p < 0.05).

Deuterium biases and offsets against environmental variables

Both pre-dawn LWPs (–1.97 to –0.18 MPa), RSWC (0.60 to 1.64 g/g) and xylem water δD (–56.5 to –22.5\($\textperthousand$\)) varied over broad ranges, making it possible to test the ability of these variables to predict variations in δD biases and offsets (Figure 4). We found that neither δD biases nor δD offsets were correlated with isotopic composition (Spearman; p=0.17 and p=0.78, respectively) (Figure 4a, 4d), pre-dawn LWP (Spearman; p=0.73 and p=0.22, respectively) (Figure 4b, 4e), or RSWC (Spearman; p=0.38 and p=0.42, respectively) (Figure 4c, 4f). The linear mixed-effect modelling confirmed that xylem water δD, pre-dawn LWP and RSWC where not significantly related to the δD bias. The model also found significant differences between C. bella and all other species (Table 2). The ‘site’ random effect had very low (<10-5) variance, implying that sites had a minimal effect on the δD bias. These modelling results suggest that while CVD-induced δD biases were independent of water content and water stress levels, inter-specific differences cannot be ignored.