2.6 Data analysis
For each species, we analysed three performance metrics that have
fitness consequences: overall survival, height increase after
DRO, and Thiaki respectively), and the inherent growth rate. Analysis
was limited to just 17 months given a) partial canopy closure by some
trees at the DRO was beginning to lead to neighbour shading b) this
allowed us to exclude the impacts of ex-Tropical Cyclone Jasper, which
made landfall close to the DRO and Cow Bay sites on the
13th Dec 2023 causing extensive regional flooding.
Probability of survival was assessed using a generalised linear model
with a binomial response with a ‘log-link’ family. Survival data (1 =
alive, 0 = dead) was the response variable. To identify the impact of
provenance and site on growth for each species, we fitted a generalised
linear model with a Gaussian distribution. We ran parallel analyses of
absolute fitness (growth increment over ~17 months) as
well as the inherent growth rate (Stanton & Thiede 2005, Lamontet al. 2023), \(IGR=ln\mathrm{\Delta}H/lnH\), where ∆H is the
change in height and H is the initial height. For each model, we looked
at the interaction effects of provenance (factor, two levels) and site
(factor, three levels), and included covariates for initial height
(numeric, continuous) and the underlying spatial autocorrelation as the
autologistic term (numeric, continuous) (Dormann 2007).
To determine if species-level wood density could explain the strength
and direction of local adaptation, we performed ordinary least squares
regression with wood density as the independent variable and local
adaptation as the dependent variable. Local adaptation was calculated as
the difference between fitness in sympatry and in allopatry averaged
across the three sites (Blanquart et al. 2013), with local
adaptation present if fitness in sympatry is higher than in allopatry.
Fitness of each population at each site was derived from the estimated
marginal means of either survival probability, height increment, or
inherent growth rate from the analyses described above. All analyses
were performed using R Statistical Software (R Core Team 2022).
Two species – Alstonia scholaris and Toona ciliata – were found to have a complex interaction between initial growth and
insect herbivory in that once individuals (irrespective of provenance)
obtained a certain size they were subject to devastating insect damage
from either an unidentified caterpillar or native Cedar tip moth
(Hypsipyla robusta ) respectively. This damage caused loss of the
apical meristem and a reduction in overall height in these species at
sites where they first gained a larger height - as such these species
were excluded from growth analyses (but retained in survival analyses).