Chicago window-killed birds (52 species, 1978–2016) have abrupt tarsus and bill shifts that adults and juveniles share, and tarsus does not follow the June temperature of the hatch year

Question: Weeks et al. (2020) reported that 52 migratory bird species killed at Chicago buildings became smaller in 1978–2016, and that warm summers caused part of this change. If development in warm summers makes birds smaller, a change must first occur in juveniles, and adults that hatched before the change must not change. Tarsus is a bone length that does not change after fledging. Approach: we used the public table of 70,716 birds. We estimated year effects separately for spring adults, fall adults and fall juveniles (cell fixed effects for species, sex, age and month), scanned for level shifts (sup-Wald test, season-clustered errors), compared juveniles with same-season adults and with next-spring adults, and tested regional June temperature with several time models. Results: the largest tarsus shift, −0.92% (SE 0.13), occurred between fall 1987 and spring 1988. It had the same size in spring adults (−0.87%), fall adults (−0.83%) and juveniles (−1.02%), and it was negative in all 30 species with data. The spring-1988 adults had hatched by 1987, so development cannot cause this shift. It is about 38% of the total tarsus decrease; a gradual decrease of −0.39% per decade remains. Bill length decreased by 3.3% per decade and shifted by +7.6% in 2015 in all 38 species. For tarsus, juveniles correlated with fall adults of the same season (r = 0.70 for first differences) but not with next-spring adults that contain their own cohort (r = −0.01). Wing, mass and bill agreed. With the original design, we reproduced a −0.13% tarsus association per SD of June temperature (p = 0.003). With the 1987/88 shift in the model, it was −0.04% (p = 0.34). Juveniles did not respond to the temperature of their hatch summer more than same-season adults (difference −0.01% per SD, 95% CI −0.17 to +0.16). Fall adults, which did not develop in that summer, gave the same −0.12% as juveniles. Cross-correlations of smoothed series, which the original study used, have null ranges of ±0.47 to ±0.84 and changed sign with the smoothing. Conclusion: in this series, a part of the size decrease and most year-to-year variation come from effects of the collection season that birds of all ages share (measurement, carcass handling or sample origin), not from development. Limitations: regional temperature in place of species-specific ranges; the contrast cannot exclude an effect of the originally reported size; the cause of the shifts and of the remaining gradual trends is unknown.

body size · temperature-size rule · museum specimens · window collisions · measurement artifact · structural break · developmental plasticity · migratory birds · tarsus · wing length · bill length · reanalysis