In five countries, 2006–2023 trends in GBIF opportunistic bird record shares agreed weakly with breeding-bird survey trends (r = 0.28–0.68) and held the information of 10–282 survey sites

Question: Does the change in the share of GBIF opportunistic bird records of a species agree with its population trend from a structured national breeding-bird survey? Data: April–July GBIF HUMAN_OBSERVATION bird records (GBIF API facets, retrieved 2026-10-24) for the UK, Sweden, Norway, Czechia and the USA, 2006–2023, with copies of the surveys, other standardised count schemes and ringing data removed. Reference trends came from the UK Breeding Bird Survey, Swedish fixed routes, Norwegian TOV-E, Czech JPSP and the North American BBS (US routes). Method: Survey trends are Poisson log-linear slopes with site effects. GBIF trends are Poisson slopes of species records with the log of all bird records as offset (record share). We compared 116–445 species per scheme by correlation and by an 'equivalent number of survey sites': the number of random survey sites whose trends agree as well with trends from an independent half of the sites. Results: Pearson r was 0.68 (UK; 0.46 without the eradicated ruddy duck), 0.31 (Sweden), 0.28 (Norway), 0.43 (Czechia) and 0.44 (USA); Spearman 0.24–0.47. Two random halves of each survey agreed better (r = 0.42–0.81). The GBIF shares held the trend information of about 10, 55, 11, 127 and 282 survey sites (Spearman criterion; 1–13% of each scheme). GBIF trends were compressed: OLS slope 0.12–0.55 of the survey trend. Greenfinch fell 8.9%/yr (UK) and 8.2%/yr (Sweden) on surveys, but its GBIF share changed by −1.0 and +0.6%/yr. A positive control, the GBIF copies of the Swedish and Norwegian surveys analysed by the same method, gave r = 0.81 and 0.69 (0.96 and 0.79 against survey presence trends). eBird and national portals did not differ consistently, and a 1.5- to 20-fold increase in records from 2006–2014 to 2015–2023 did not consistently improve agreement. Raw record counts rose 4.5–32%/yr. Limitation: we tested the simple national record-share index, not checklist or occupancy models; GBIF content changes, so we include the exact API responses. Conclusion: in these countries, GBIF record shares contain a weak trend signal and do not replace structured surveys.

GBIF · citizen science · opportunistic data · eBird · breeding bird survey · population trends · reporting rate · monitoring · data validation · birds