Effects of environmental covariates on growth sensitivity

Equation 4: We examined the effects of species, mean climatologies and elevation at each site, and tree ontogeny, on <i>t</i>_FFD, <i>t</i>_SMD(_<em>T</em>_-1), <i>t</i>_VPD(_<em>T</em>_-1), <i>t</i>_SMD(_<em>T</em>₎, and <i>t</i>_VPD(_<em>T</em>₎, using a Generalized Additive Model (GAM). Variables characterizing tree ontogeny included tree age (years) and size (logBA) at time of sampling. To determine whether the splines vary with species, we run an interaction model of covariates. In this interaction model, the fitting process optimizes the performance where some species may behave differently from others, in interactions with environmental covariates.Equation 5: We were interested in knowing how climate sensitivity <i>t</i> depends smoothly on covariates, the hypothesis being that the effects of MAT and MAP are felt toward the cold and warm edges (or wet and dry edges) of the tree species' sample distributions. This hypothesis is tested by examining the confidence intervals (CIs) of the covariate’s predicted fits, which here are zero-centered. The null hypothesis ‘no edge effect’ is rejected whenever the CIs at the extremes or limits of the range of the species’ data exclude zero (see Figure S2; Simpson, 2018). All tests were two-sided, i.e., we made no hypothesis about the directionality of responses.The strength and significance of covariates was determined using the associated F-ratio and <i>P</i>-value. We used the ‘bam’ function, <i>k</i> =5, and the fREML method to fit these species’ models. The data contained in this folder are raw outputs from these models.A subset of the tree-ring data extracted from the CFS-TRenD repository is also included. <br>