Robustness of simple avian population trend models for semi-structured citizen science data is species-dependent

Citizen scientists are collecting opportunistic biodiversity records on unprecedented temporal and spatial scales, vastly outnumbering the records achievable from structured surveys. For example, BBS is one of the world’s most comprehensive national monitoring schemes, covering over 1.5% of the UK's land surface. To achieve this remarkable coverage, BBS volunteers conducted approximately 7,600 visits to just under 4,000 sites in 2017. Yet, this impressive number is dwarfed by the almost 80,000 complete lists that were collected by BirdTrack participants at over 16,000 locations during the same time.

However, unlike BBS volunteers who visit randomly selected sites, and systematically survey them, BirdTrack participants are free to go when and where they choose. The bounty of opportunistic records, such as BirdTrack lists, therefore hides potential biases in the selection of sites and visit times, as well as variation in observer effort and skill. Such quantity-quality trade-offs between surveys, like BBS, and less structured schemes, such as BirdTrack, remain poorly understood.

Recent work has advocated the use of simple statistical trend models as a quick way to leverage opportunistic citizen science data to produce trends, and as a means to fill knowledge gaps on species or regions that are poorly covered by structured surveys. In this study, BTO scientists examined the robustness of population trends of common UK birds derived from BirdTrack and BBS results. They derived trends in BirdTrack reporting rates from BirdTrack lists, using simple statistical models which accounted for variation in observer effort, such as the time spent compiling each BirdTrack list, but not for the non-random nature of the ways sites in BirdTrack are selected. They compared these trends to population trends derived from BBS survey data. For 90 of 141 species,  year-to-year changes in BirdTrack reporting rates were positively correlated with BBS trends. However, such correlations were higher for widespread species and those exhibiting marked population change. Less agreement among trends was found for rarer species and those with small or uncertain population trajectories.

This study's findings suggest that simple statistical models for estimating population trends from complete BirdTrack lists are robust only for widespread and common species, and do not provide a silver bullet for the monitoring of rarer species, even in a scheme with many observers and extensive coverage. Notwithstanding, all BirdTrack records are valuable. They help to map changing distributions and assess variation in migrant arrival and departure dates and BTO scientists are actively developing novel statistical methods to integrate BirdTrack data into improved distribution and trend estimates.

Notes

Thank you to the many thousands of citizen scientists who contribute bird records to BirdTrack or the Breeding Bird Survey and the past and present organizers and staff of both schemes. BirdTrack is operated by the BTO, and supported by the Royal Society for the Protection of Birds, BirdWatch Ireland, Scottish Ornithologists’ Club, the Welsh Ornithological Society and BirdLife International. The BTO/JNCC/RSPB Breeding Bird Survey is a partnership jointly funded by the BTO, RSPB, and JNCC, with fieldwork conducted by volunteers. We thank the donors to BTO’s BirdTrack Research Appeal and The Newstead Charity for their financial support. Computations for this study used JASMIN, the UK’s collaborative data analysis environment (http://jasmin.ac.uk).