Premium
When can we trust population trends? A method for quantifying the effects of sampling interval and duration
Author(s) -
Wauchope Hannah S.,
Amano Tatsuya,
Sutherland William J.,
Johnston Alison
Publication year - 2019
Publication title -
methods in ecology and evolution
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 3.425
H-Index - 105
ISSN - 2041-210X
DOI - 10.1111/2041-210x.13302
Subject(s) - population , statistics , sampling (signal processing) , population size , sign (mathematics) , scale (ratio) , magnitude (astronomy) , trend analysis , iucn red list , reliability (semiconductor) , demography , geography , econometrics , ecology , mathematics , biology , computer science , cartography , power (physics) , mathematical analysis , physics , filter (signal processing) , quantum mechanics , astronomy , sociology , computer vision
Species’ population trends are fundamental to conservation. They are used to determine the state of nature, and to prioritize species for conservation action, for example through the IUCN red list. It is crucial to be able to quantify the degree to which population trend data can be trusted, yet there is not currently a straightforward way to do so. We present a method that compares trends derived from various samples of ‘complete’ population time series, to see how often these samples correctly estimate the sign (i.e. direction) and magnitude of the complete trend. We apply our method to a dataset of 29,226 waterbird population time series from across North America. Our analysis shows that, for waterbirds, if a statistically significant ( p < .05) trend is detected, even from only a few years, it is likely to reliably describe the sign (positive or negative) of the complete trend, but is unlikely to accurately match the percentage change in population per year. If no significant trend is detected, a many‐years long sample is required to be confident that the population is truly stable. Furthermore, an insignificant trend is more likely to be missing a decline rather than an increase in the population. Sampling infrequently, but regularly, was surprising reliable in determining trend sign, but poor at determining percentage change per year. By providing percentage estimates of reliability for combinations of sampling regimes and lengths, we have a means to determine the reliability of species population trends. This will increase the rigour of large‐scale population analyses by allowing users to remove time series that do not meet a reliability cut‐off, or weighting time series by reliability, and could also facilitate planning of future monitoring schemes. While the specific values estimated by our analysis might not be applicable to other taxa or systems, the methods are easily transferable, and we provide the tools to do so.