Using opportunistic citizen science data to estimate avian population trends

Abstract

Determining population trends is critical for conservation. For most bird species, trends are based on count data gathered by institutions with formalized survey protocols. However, limited resources may prevent these types of surveys, especially in developing countries. Ecotourism growth and subsequent increases in opportunistic data from birdwatching can provide a source of population trend information if analyses control for inter-observer variation. List length analysis (LLA) controls for such variation by using the number of species recorded as a proxy for observer skill and effort. Here, we use LLA on opportunistic data gathered by eBird to estimate population trends for 574 North American bird species (48% of species declining) and compare these estimates to population trends based on 1) formal breeding bird surveys (54% of species declining) and 2) population estimates from eBird data controlled using more rigorous correction (46% of species declining). Our analyses show that eBird data produce population trends that differ on average by only 0.4%/year from formal surveys and do not differ significantly from estimates using more control metrics. We find that estimates do not improve appreciably beyond 10,000 checklists, suggesting this as the minimum threshold of opportunistic data required for population trend estimation. Lastly, we show that characteristics affecting a species' ubiquity, such as geographic and elevational range, can affect its population trend estimate. Our results suggest that opportunistic data can be used to approximate species population trends, especially for widespread species. Because our protocol uses information present in all checklists, it can be applied to a diversity of data sources including eBird, trip reports, and bird atlases.