Uncertainty in population estimates: a meta-analysis for petrels

Abstract

Population estimates are commonly generated and used in conservation science. All estimates carry inherent uncertainty, but little attention has been given to when and how this uncertainty limits their use. This requires an understanding of the specific purposes for which population estimates are intended, an assessment of the level of uncertainty each purpose can tolerate, and information on current uncertainty. We conducted a review and meta-analysis for a widespread group of seabirds, the petrels, to better understand how and why population estimates are being used. Globally petrels are highly threatened, and aspects of their ecology make them difficult to survey, introducing high levels of uncertainty into population estimates. We found that by far the most common intended use of population estimates was to inform status and trend assessments, while less common uses were trialling methods to improve estimates, and assessing threat impacts and conservation outcomes. The mean coefficient of variation for published estimates was 0.17 (SD = 0.14), with no evidence that uncertainty has been reduced through time. As a consequence of this high uncertainty, when we simulated declines equivalent to thresholds commonly used to trigger management, only 5% of studies could detect significant differences between population estimates collected 10 years apart for populations declining at a rate of 30% over three generations. Reporting of uncertainty was variable with no dispersion statistics reported with 38% of population estimates and most not reporting key underlying parameters: nest numbers/density and nest occupancy. We also found no correlation between population estimates and either island size, body size or species threat status - potential predictors of uncertainty. Synthesis and applications: Key recommendations for managers are to be mindful of uncertainty in past population estimates if aiming to collect contemporary estimates for comparison, to report uncertainty clearly for new estimates, and to give careful consideration to whether a proposed estimate is likely to achieve the requisite level of certainty for the investment in its generation to be warranted. We recommend a practitioner-based Value of Information assessment to confirm where there is value in reducing uncertainty.

Generating unbiased estimates of burrowing seabird populations

“…To evaluate the strength of our unbiased estimators of burrow numbers we simulated declines of 30, 50 and 80% over three generations for Antarctic prions and white‐headed petrels following Bird et al (2021) and assessed their power to detect significant differences between repeat estimates (Supporting information).…”

“…Typically the number of burrows and the proportion that are occupied are measured separately and combined to obtain an estimate for a given area, often a whole island (Sutherland and Dann 2012). In most cases large uncertainty in these estimates prevents the detection of significant changes in seabird population sizes over time (Bird et al 2021). Uncertainty is usually greater for burrow detection than burrow occupancy, since burrows are often distributed discontinuously across challenging terrain, and their entrances may be obscured by vegetation (Rayner et al 2007, Rodríguez et al 2019).…”

Trialling camera traps to determine occupancy and breeding in burrowing seabirds

“…Remote Sensing in Ecology and Conservation published by John Wiley & Sons Ltd on behalf of Zoological Society of London This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes. (Bird et al, 2021;Rodríguez et al, 2019). The most important metrics for tracking conservation status and population-level responses to impacts are the number of breeding burrows and their success rate, both of which rely upon one-off or repeat measures of burrow occupancy.…”

“…Collectively there are strong ecological conservation and fisheries management incentives for understanding petrel population trends, but they are difficult to monitor (Rodríguez et al., 2019). Seasonal occurrence, nocturnal activity at colonies and challenges in finding, counting and ascertaining the contents of burrows all contribute to high levels of uncertainty when populations are assessed (Bird et al., 2021; Rodríguez et al., 2019). The most important metrics for tracking conservation status and population‐level responses to impacts are the number of breeding burrows and their success rate, both of which rely upon one‐off or repeat measures of burrow occupancy.…”

“…In recent years, audio playback for small species and manual burrow inspections for larger species have emerged as the most widely used occupancy detection methods in burrowing seabird studies (Bird et al., 2021). Playback requires calibration to account for non‐responsive birds (Soanes et al., 2012).…”