11,511 Publications · 65,198 Citations

Question: Does wildfire smoke on the survey day change the number of birds and species that Breeding Bird Survey (BBS) observers record? eBird studies found smoke-related detection changes, which can bias population trends. Approach: We joined 36,362 accepted BBS surveys on 4,362 routes (2011-2024) to NOAA Hazard Mapping System (HMS) satellite smoke polygons and to EPA monitor PM2.5 on the survey date. Linear models used route-by-observer and state-by-year fixed effects, weather, time and date controls, and two-way clustered errors. Smoke 7 days before and after the survey gave placebo estimates. We repeated the test in four region-period subsets (West or East of 100 W; 2011-2017 or 2018-2024) and fitted Poisson models for 247 species. Results: The exposure was valid: median PM2.5 increased from 6.5 to 16.0 ug/m3 and observer 'fog or smoke' sky codes from 3.4% to 13.9% between no-smoke and heavy-smoke surveys. Medium or heavy smoke changed route totals by +0.94% (95% CI -0.10 to +1.98) and richness by -0.33% (-0.80 to +0.13); heavy smoke alone gave -0.08% (-2.04 to +1.88) and -0.70% (-1.64 to +0.24). No subset gave a medium-or-heavy decrease with a CI that excluded zero. Placebo days gave estimates of the same size (-0.37% and +0.79%). Surface PM2.5 gave no link. Species-level estimates were near chance: 28 of 247 species had |z| > 1.96 on the survey day (9 lower, 19 higher), against 15 and 24 on the placebo days. Limitations: HMS shows smoke in the full column, not at ground level; only 55 surveys had PM2.5 of 35 ug/m3 or more; observers can move surveys away from very smoky days; counts mix abundance and detection. Conclusion: Survey-day smoke detection bias does not explain BBS route-level counts in 2011-2024.