Results from candidate models using AICc for the 7 study sites.

Only interpretable predictor variables in at least one site are provided (see Table S2 for all models analyzed). The first value represents the Akaike's weight (wi) given to each model in the candidate set followed by the R2 statistic and the direction of the relationship. Bold values represent the best model in each candidate set. Abbreviations are as follows: rpop = per capita growth rate (r = ln(nt/nt−1) for the total brook trout population, radult = r for adults, ryoy = r for young-of-the-year, dtrout = density of all brook trout, dadult = density of adult brook trout, dyoy = density of young-of-the-year brook trout, sptT = mean April-June maximum temperature, sut-1T = mean July maximum temperature, and sptQ = mean March-June discharge. Missing values represent predictor variables that were correlated with another predictor variable in the candidate set and therefore removed. No models were constructed for ryoy at main stem sites because few YOY were found in those sites. Models with an * were considered interpretable models using criteria from Grossman et al. [14].