12,627 Publications · 75,780 Citations

Question: a campaign synthesis found that community temperature index (CTI) change in time follows only about 12% of the spatial CTI-temperature slope. Is the spatial slope inflated by non-thermal factors that change with latitude? Data: North American Breeding Bird Survey, contiguous US, 3,960 routes, 107,922 standard runs, 467 species, 1966-2025; 5 km nClimGrid May-July temperature; AVONET habitat classes. Method: cross-fitted species temperature indices; spatial slopes of route CTI on route climate across the continent and only between routes inside 8 to 0.5 degree blocks, with covariates and an instrumental variable for temperature error; temporal slopes on route warming trends (slow) and anomalies (fast) with route and observer fixed effects, and with region x year fixed effects; a species-level test with independent periods. Results: the spatial slope was 0.40 C per C across the continent and 0.30-0.34 inside 0.5-8 degree blocks (IV 0.33-0.34); latitude and non-latitude parts of temperature gave 0.40 and 0.39. The slow temporal slope was 0.095 (SE 0.012), 24% of the continental and 32% of the 1 degree slope; the fast slope was 0.011, 3-4%. With national year effects the slow slope fell to 0.053, and with 4 degree block x year effects to 0.010 (SE 0.022). In 1997-2025 the raw slow slope was 0.21, but 0.04 with national year effects. Forest birds kept a 1 degree slope of 0.32; open-habitat and water birds had about half their continental slopes locally. For 414 species, abundance change on routes that warmed more did not follow the spatial temperature slope (weighted slope -0.10, SE 0.06). Conclusion: the spatial benchmark is mostly thermal sorting, so the tracking deficit is a real temporal deficit. Limitations: route start points only, no land cover, no detection model, one network.