Year-to-year variance of local bird counts did not increase from early to recent decades in North American, UK, Swedish and Czech surveys at equal observer experience

We tested if local bird populations became more variable from year to year. We used within-observer annual growth rates of 43,632 site x species series in four schemes: North American Breeding Bird Survey (1970-1994 vs 2000-2024), UK Breeding Bird Survey (1994-2008 vs 2010-2024), Swedish fixed routes (1996-2010 vs 2011-2025) and Czech JPSP (1982-2002 vs 2004-2024). A Poisson pseudo-likelihood model with series fixed effects gave the late/early ratio of squared growth deviations. The model controls observer tenure at the site, observer experience in the scheme, survey-date gap and abundance. Intervals come from a bootstrap of sites and years. No scheme showed an increase. The ratios were 0.951 (95% interval 0.912-0.991) in North America, 0.971 (0.878-1.062) in the UK, 0.909 (0.790-1.034) in Sweden and 1.096 (0.742-1.759) in Czechia (19 sites). The combined ratio was 0.951 (0.917-0.987). Variance decreased strongly with observer experience (up to 30% less for observers with more than 40 earlier surveys), and mean tenure was 2 to 3.3 times larger late. Without observer terms the apparent decrease was larger (0.926, 0.910, 0.874, 0.709). The North American decrease is not robust: 1980-1999 vs 2000-2019 gave 0.999 (0.965-1.028). When series were not paired across periods, North American raw excess variance was 16% higher late, which shows that series composition must be held fixed. Sedentary North American species decreased most (0.884); a cold-winter term explained little of this. Main limitation: the measure includes counting error. Simulations on the real survey structure show no bias at the null, but a 20% change in process variance appears as only 3-12%. Thus moderate process-variance increases cannot be excluded in the UK, Sweden or Czechia.

population variability · breeding bird survey · observer effects · observer experience · growth rate variance · climate change · North American BBS · UK BBS · Swedish Bird Survey · Czech JPSP · Poisson pseudo-likelihood · long-term monitoring