Performance and functional traits variation in relation to cold temperature
To answer our first research question, we evaluate the difference between performance (apical growth rate) between the study sites and cold (mean minimum temperature) as covariable in an analysis of covariance test (ANCOVA). To respond our second research question, we first evaluate the difference between functional traits (hydraulic transport, wood anatomy and root pressure) between the study sites and cold (mean minimum temperature) as covariable in an analysis of covariance test (ANCOVA). To perform these analyses, we build general linear models with a dependent response variable, “study site” as a categorical factor and “mean minimum temperature” as continuous variable as covariable, we use anova function in the car package in R Core Team (2020). We verify the assumptions of the ANCOVA analysis such as the independence between the covariate and the independent variable, the homogeneity of variance (Levene test in car library), and the linear relation between the covariate and the outcome. Then we then conducted a Tukey post-hoc analysis to identify the difference between sites, using the function glht in the package multcomp in R Core Team (2020). After this, we evaluate the relation of the trade-off between efficiency and safety of water transport and the performance of liana species, by conducting a principal component analysis (PCA). We use the function prcomp, considering mean cero and standard deviation one. To reduce the factors and reflect the combination of functional traits that are best associated with the growth rate, a second step was perform linear regressions between the PCA score obtained for each trait in each site and the mean apical growth rate (AGR) for the species, this regression reflect which specific trait combination is associated with performance. Finally, in relation to the efficiency-safety tradeoff, we evaluate the selection of narrow xylem vessels as a response to freezing temperature towards in higher latitudes (Fig. 2), through the evaluation of the frequency distribution of the xylem vessel diameter (µm) and their relative contribution to total hydraulic conductivity (%) with a Kolmogorov-Smirnov test, using the ks.test function in R Core Team (2020).