Reliably diagnosing Fraser River sockeye salmon declines in a variable world

There are numerous indicators that can be used to categorize populations as being a conservation concern, but these alternatives may vary in their reliability. To determine which of 18 quantitative decline indicators most reliably categorize the conservation status of a population, we used a stochastic model to simulate time series representing populations of sockeye salmon (Oncorhynchus nerka), allowing us to examine the effects of different levels of process variation and observation error. We examined whether each indicator’s assessed status accurately predicted the subsequent trend of the population using a Receiver Operating Characteristic analysis as an integrated measure of reliability. Indicators that measured decline from a historical baseline were the most reliable. Indicators that measure decline over the most recent 3 generations, which are widely used, were sensitive to process variation and observation error. Our results suggest that, when available, longer time-series of abundance should be used to evaluate population decline.