Food acquisition is an important modulator of animal behavior and habitat selection that can affect fitness. Optimal foraging theory predicts that predators should select habitat patches to maximize their foraging success and net energy gain, likely achieved by targeting areas with high prey availability. However, it is debated whether prey availability drives fine-scale habitat selection for predators. We assessed whether an ambush predator, the timber rattlesnake (Crotalus horridus), exhibits optimal foraging site selection based on the spatial distribution and availability of prey. We used passive infrared camera trap detections of potential small mammal prey (Peromyscus spp., Tamias striatus, and Sciurus spp.) to generate variables of prey availability across the study area and used whether a snake was observed in a foraging location or not to model optimal foraging in timber rattlesnakes. Our models of small mammal spatial distributions broadly predicted that prey availability was greatest in mature deciduous forests, but T. striatus and Sciurus spp. exhibited greater spatial heterogeneity compared to Peromyscus spp. We found the spatial distribution of cumulative small mammal encounters (i.e. overall prey availability), rather than the distribution of any one species, to be highly predictive of snake foraging. Timber rattlesnakes appear to forage where the probability of encountering prey is greatest. Our study provides evidence for fine-scale optimal foraging in a low-energy, ambush predator and offers new insights into drivers of snake foraging and habitat selection. 

(1) We quantified small mammal relative availability with widely-distributed camera traps, (2) projected small mammal encounters across the study area with landscape predictors, and (3) used radio-telemetry-derived timber rattlesnake foraging locations to model optimal foraging in timber rattlesnakes. We deployed game cameras to detect small mammals from 2017–2018 at 242 randomly-chosen, unique sites. We positioned each game camera overlooking a log or downed woody debris to simulate a stereotyped ambush position of timber rattlesnakes (Crotalus horridus). We modeled counts of small mammal observations with remotely-sensed landscape characteristics to predict small mammal encounters across the study area. We modeled snake foraging using mixed-effects Bernoulli GLMs with foraging behavior as a binomial function of the spatially-explicit small mammal encounter rates.