Data analysis
All analyses were carried out in the R environment version 3.6.1 (R Core Team 2019) and graphics were obtained through the packagesggplot2 , jtools and interactions (Wickham, 2016; Long, 2019b, 2019a). Species climatic niches were estimated from the weighted average values of mean annual temperature (bio 1) and annual precipitation (bio 12) of all sites where each species was recorded. Weights were based on the average AGWB of the species in each site across all census intervals. We opted to weigh the species’ niches with their AGWB rather than abundance because by better synthesizing the species’ responses to local conditions. We ran the analysis for each forest type separately because climate is strongly correlated with forest type. This approach prevented us from including climate and forest type in the same models, avoiding high variance inflation factors (VIF > 4).
Balance between recruitment and extinction
We tested if the number of extinct and recruited species differed significantly from each other, and estimated the temporal trend of each group (i.e., the interaction between extinct/recruited and year). For this, we used generalized linear mixed effects models (GLMM) using Poisson family. Because observations were nested within sites and within census intervals, we accounted for the random effects of site and census interval. We used the natural logarithm of the multiplication between site sampled area and census interval length as an offset variable to control for their effects on the species count.
count of species ~ status*year (eqn 1)
Equation 1 describes the global model of extinction/recruitment balance with all variables included in the R notation. Interaction effects are represented by “*”.