Estimating the Effects of Environmental Variables and Gear Type on the Detection and Occupancy of Large-River Fishes in a Standardized Sampling Program Using Multiseason Bayesian Mixture Models

Sampling in non-wadeable rivers presents methodological challenges for monitoring fish species. Changing environmental conditions may affect the ability to accurately capture species (i.e., detection) and consequently may lead to inappropriate inferences on occupancy rates. We used hierarchical Bayesian multiseason mixture models to estimate occupancy and detection of 41 of 52 fish species in the Kankakee River, Illinois, by using data from a standardized monitoring program. Fish were sampled with AC boat electrofishing and shoreline seining over 7 years. Some centrarchids (e.g., Smallmouth Bass Micropterus dolomieu) were efficiently sampled by boat electrofishing, whereas most other species had low detection probabilities. Moderate changes in environmental conditions, such as water velocity and temperature, produced moderate changes in detection and occupancy. Generally, when species had high detection probabilities, changes in environmental conditions produced relatively small changes in the estimated detection probabilities. Our results also suggested that some sport fishes collected from rivers with only moderate environmental fluctuations are unlikely to produce strongly biased estimates of detection and occupancy among years. However, many species had detection probabilities that were low, imprecisely estimated, or both. Overall, we demonstrate that long-term fisheries monitoring can effectively detect some species at levels that are often relevant for management, but assessments of species with lower and more uncertain detection probabilities may not provide adequate information for management decisions. We recommend the use of sampling designs that allow the estimation of both detection and occupancy.