Using Single-Pass Surveys to Assess Spatial and Temporal Patterns in Brook Trout Abundance: Correcting for Imperfect Detection

Spatial and temporal variability in detection probabilities can make it challenging to infer population trends from count data. Over the past decade, hierarchical models have been developed to distinguish ecological patterns (i.e. abundance) from observation processes (i.e. detection) using repeated surveys (e.g. repeated visits, multi-pass, double-observer). However, most stream fish survey data collected by state and federal agencies use single-pass electrofishing and do not have the repeated sampling necessary for these models. This is a problem for understanding population trends in responses to climate change over broad spatial extents. A new model has been developed to account for detection probability using conditional likelihood from single-pass surveys. This model has performed well in simulations but remains to be tested with field data. We used 14 years of brook trout data from Shenandoah National Park to compare abundance and detection estimates from 3-pass sampling with estimates from just the first pass using this new model. There was good correlation in abundance estimates using the two methods (adults: 0.66, YOY: 0.95). The single-pass method slightly overestimated detection and therefore underestimated abundance compared with the 3-pass Dail-Madsen model. This new single-pass model provides a viable option for using historic data to evaluate population trends.