In BBS route counts (1966–2024), the extra variance at an observer change did not increase with outgoing tenure, so drift from an observer's first survey does not explain same-observer variance growth

Question: the variance of change in bird route counts increases with the time lag, also when one observer makes both counts (SCI 2ec6d518). Is observer drift the cause? Data: North American Breeding Bird Survey 2025 release, standard runs 1966-2024, 61,984 common species-route series on 1,869 routes; Swedish fixed routes 1996-2025 (5,405 series) for replication. Method: semivariance of log(count+1) by lag for same-observer and different-observer pairs with series fixed effects and route jackknife; a shared tenure curve from species-year-centred year-to-year changes; a drift test that compares the extra semivariance at an observer change between consecutive years across outgoing tenures, in the data and in simulations on the real observer sequences. Results: an observer change added 0.168 (BBS) and 0.174 (Sweden) at lag 1. Same-observer semivariance doubled from lag 1 to lag 20. Observer turnover made 26% (BBS) and 24% (Sweden) of the all-pairs growth to lag 20, or up to 40-49% with a different-observer estimator. The step decreased to 0.088 at lag 30, also without first runs. One observer's counts increased for 3-4 runs and then decreased to -0.128 log units at run 30 (Sweden: -0.098 at run 20); counts jumped +0.105 when an observer left after more than 20 runs. Removal of this curve decreased same-observer growth by only 6-7%. The extra semivariance at an observer change did not increase with outgoing tenure (more than 20 minus 6-10 runs: +0.001, 95% CI -0.023 to 0.027). Drift that starts at each observer's first survey, at the size that fits the same-observer curve, predicted +0.080 (0.066 to 0.094); half that size predicted +0.050. Drift centred at mid-tenure predicted +0.010 and is not excluded. Conclusion: accumulating observer drift does not explain the within-observer variance growth; population change or observer deviations with constant spread remain. Limitations: BBS only for the drift test; low power against centred drift; common species on long-surveyed routes.

breeding bird survey · observer effects · observer drift · observer aging · population variability · variogram · time series length · measurement error · Swedish Bird Survey · monitoring data quality · spectral reddening