Optimal Stream Survey Design for Detection of Fish Population Trends: Insights Gained from Long-Term Monitoring

Use of surveys to detect trends in population abundance over time can be very difficult due to high natural variability. Trout populations are no exception, with reported coefficients of variation (CV) ranging from 15%-108%. Given the finite amount of resources available to managers, knowledge of how survey design affects the ability of monitoring to detect population trends would be highly beneficial. To assess these effects, we applied varying spatial and temporal survey designs to a brook trout (Salvelinus fontinalis) population dataset that contained nineteen years of fine-scale electrofishing survey data, and had a positive linear trend over that interval. Spatial designs revealed that annual sampling of four 50 M sections (10% of available habitat) that were initially randomly selected but remained fixed over the monitoring period resulted in CV (51.4%) and statistical power (0.88) values comparable to those obtained from sampling the entire stream (46.7%, 0.99). Temporally, sampling the entire stream at a fixed annual frequency ranging from 2 to 5 years resulted in significant trend detection in only 29% (4/14) of all monitoring possibilities. The low detection rate was caused by high abundance variability, evidenced by the fact that annual monitoring required a minimum of seventeen years before reaching adequate power (0.81).