Statistical analyses
Differences in fragment survivorship between source-recipient treatments
were analysed with a chi-squared test at the end of the heat stress
period (T2) and end of the recovery period (T4). Differences in 1) Shoot
mortality per fragment, 2) herbivory impact, 3) C:N of leaves, 4)
growth, 5) new leaf production and 6) total leaf length were each
compared between treatments (n = 6) within discrete time periods using
Analysis of variance (ANOVA). Normality was tested using Shapiro-Wilks
test and Q-Q plots, and homogeneity of variance was tested with Levene’s
test and residuals plot. No transformations were applied. Tukey-HSD
analysis was used to examine differences between groups for significant
main effects. To assess the relationship between thermal performance,
source location and 1) maximum absolute realised temperature and 2)
maximum thermal stress anomalies, linear mixed effects models were used
using the lme4 package in R (Bates et al. 2015). Source location
and temperature metric were treated as fixed effects and recipient
location’ and ‘time point’ were treated as random effects, to account
for differences in performance across locations and seasonality.
Separate models were conducted for each temperature metric (i.e. max
temperature or heat stress anomaly) and performance metric (i.e. growth,
new leaf production and total leaf length) combination. We started with
a full model based on a priori hypotheses about the inclusion of
terms and higher order interactions. We identified the minimum adequate
model by a stepwise removal of non-significant terms using likelihood
ratio tests of the model with the effect in question against the model
without the effect in question. Visual inspection of residual plots did
not reveal any obvious deviations from homoscedasticity or normality and
therefore no transformations were conducted. P-values were obtained by
likelihood ratio tests of the full model with the effect in question
against the model without the effect in question. All analyses were
conducted in R (version 4.0.3, 10/10/2020).