Temporal Variation in Trout Populations: Implications for Monitoring and Trend Detection

We summarized the temporal variation in populations of inland trout Salmo, Salvelinus, and Oncorhynchus spp. from streams in North America and determined the statistical power to detect trends over time. The coefficients of variation in abundance averaged 49% (SD = 27% range = 15-108%) over time for all ages of trout. Temporal variation was lower when more age-classes were monitored, but whether abundance or biomass was more variable differed among Populations. Detecting population trends was difficult when using the traditional alpha = 0.05 criterion. For example, detecting it 5% annual decline with good power (1 - beta >= 0.80) would require about 20 years if only one site were monitored. Even when alpha was relaxed to 0.20, 15 years were required to detect it 5% annual decline when the variation was average. Using a network of sites improved the ability to detect changes: a 5% annual decline at alpha = 0.05 could be detected in 10 years when 30 sites were monitored. For high-value populations, it may require relaxing alpha to ensure that declines are detected, even if this increases the risk of claiming change when none has occurred and thus undertaking unnecessary management action. For example, a 5% annual decline could be detected with good power (>= 0.80) in 8 years when a network of 30 sites is monitored at alpha = 0.20. Thus, biologists should monitor the least-variable component of a population, monitor a network of sites, and increase alpha for species of concern to ensure that real population trends are detected. Estimates of trend parameters (and their uncertainty) should be considered in addition to whether or not a statistical test for trend is significant. A pilot study or existing data can help estimate the variation that is typical of the population(s) to be monitored, determine whether trends can be reliably detected, and identify how Much risk needs to be incurred to detect trends.