Effects of sampling duration and intervals on confidence in population trend estimates

Abstract

Species population trends are fundamental to conservation, underpinning lUCN red-list classifications, many national lists of threatened species and are also used globally to convey to policy makers the state of nature. Clearly, its crucial to quantify how much we can trust population trend data. Yet many studies analyzing large numbers of population time series lack a straightforward way to estimate confidence in each trend. Here we artificially degrade 27,930 waterbird population time series to see how often subsets of the data correctly estimate the direction and magnitude of each populations true trend. We find you need to sample many years to be confident that there is no significant trend in a population. Conversely, if a significant trend is detected, even from only a small subset of years, this is likely to be representative of the populations true trend. This means that if a significant decline is detected in a population, it is likely to be correct and conservation action should be taken immediately, but if the trend is insignificant, confidence in this can only be high with many samples. Our full results provide a clear and quantitative way to assign confidence to species trends, and lays the foundation for similar studies of other taxa that can help to add rigor to large-scale population analyses.