Estimating size‐specific brook trout abundance in continuously sampled headwater streams using Bayesian mixed models with zero inflation and overdispersion

Abstract – We examined habitat factors related to reach‐scale brook trout Salvelinus fontinalis counts of four size classes in two headwater stream networks within two contrasting summers in Connecticut, USA. Two study stream networks (7.7 and 4.4 km) were surveyed in a spatially continuous manner in their entirety, and a set of Bayesian generalised linear mixed models was compared. Trout abundance was best described by a zero‐inflated overdispersed Poisson model. The effect of habitat covariates was not always consistent among size classes and years. There were nonlinear relationships between trout counts and stream temperature in both years. Colder reaches harboured higher trout counts in the warmer summer of 2008, but this pattern was not observed in the cooler and very wet summer of 2009. Amount of pool habitat was nearly consistently important across size classes and years, and counts of the largest size class were correlated positively with maximum depth and negatively with stream gradient. Spatial mapping of trout distributions showed that reaches with high trout counts may differ among size classes, particularly between the smallest and largest size classes, suggesting that movement may allow the largest trout to exploit spatially patchy habitats in these small headwaters.