Data from: Timing and duration of molt in temperate-zone passerines is shaped by frequency of double brooding, migration distance, and wing length

AbstractPrevious studies have shown that duration of the post-breeding molt in temperate-zone passerines is inversely related to migration distance. However, it is unclear whether this relationship arises directly—because long-distance migration selects for more rapid molt, or indirectly—because long-distance migration precludes double brooding and allows molt to begin earlier under more favorable environmental conditions. We investigated this issue by examining the effects of migration distance, frequency of double brooding, and wing length on the timing and duration of the post-breeding molt for 33 passerine species captured during banding operations at Powdermill Avian Research Center in southwestern Pennsylvania, USA. Consistent with predictions of the indirect pathway hypothesis, we found that long-distance migration is strongly associated with single brooding, and single-brooded species initiate flight-feather molt about one month earlier than double-brooded species. However, contrary to the predictions of the indirect pathway hypothesis, frequency of double brooding has no effect on molt duration; instead, molt duration increases with increasing wing length and decreases with increasing migration distance. The strong negative relationship between migration distance and molt duration is consistent with the direct pathway hypothesis and suggests that the time constraints imposed by long-distance migration lead directly to natural selection for rapid post-breeding molt. Our results also indicate that the rapid molt of long-distance migrants is achieved by an increase in the number of remiges being replaced simultaneously (molt intensity), and apparently also by an increase in the rate at which remiges grow. Unexpectedly, our estimates of molt duration were bimodally distributed, with 25 species showing rapid molt (40–65 d) and 8 showing slow molt (82–103 d). Overall, our findings clarify the effects of migration distance, double brooding, and wing length on phenology of passerine post-breeding molt, and highlight opportunities to expand avian life-history theory to incorporate interspecific variation in molt strategies.MethodsCollection of Molt DataData on passerine post-breeding molt were collected 1986–2004 in association with year-round netting and banding operations at Powdermill Avian Research Center (40.164°N, 79.267°W) in southwestern Pennsylvania, USA (Mumme et al. 2025a,b). Mist nets were located in a 10-ha area of shrubby second growth adjacent to deciduous forest, ponds, and agricultural land. Weather permitting, netting and banding occurred 5–6 days per week, beginning 1 h before sunrise and ending mid-afternoon. Each banded bird was assigned a total remex molt score (0–90), primarily by one individual (RSM), based on the molt status of the 18 remiges (9 primaries and 9 secondaries) on the right wing, using the 0–5 molt scoring system of Ginn and Melville (1983): 0 = old feather, 1 = old feather missing or new pin feather, 2 = new feather emerging from sheath up to one-third grown, 3 = new feather between one-third and two-thirds grown, 4 = new feather more than two-thirds grown but still sheathed at base, and 5 = fully grown new feather with no trace of the sheath. Total remex score was recorded for all individuals, but in a subset of captured birds separate scores were also recorded for each of the 18 remiges, allowing calculation of molt intensity, which for this analysis we defined as the number of remiges being actively replaced (i.e., with molt score 1–4) during mid-molt, when molt was 20–80% completed (total remex molt score 18–72).Our analysis focused on species that undergo a complete post-breeding molt at our study site, and we excluded species or individuals having partial or suspended molts. Additional criteria for species inclusion were either: (1) at least 20 records of birds in active molt, i.e., with remex molt score 1–89, or (2) at least 15 records of birds in active molt with some data from individuals recaptured during the same molt cycle, as recapture data substantially improve estimation of molt parameters. A total of 33 species from 11 passerine families, with a combined 3,324 records of birds in active molt, including 2,068 records of molt intensity, are included in the final dataset.All but 1 of the 33 species examined breed locally in the Powdermill area; the exception is the molt-migrant Catharus ustulatus (Swainson’s Thrush), which we included because individuals of this species regularly initiate and complete molt during their migratory stopover at Powdermill before continuing south to their nonbreeding range in Central and South America. Of the remaining 32 species, 24 are completely migratory at Powdermill and 5 are partial migrants, i.e., migratory species whose nonbreeding range includes the Powdermill region; for these species, the local breeding population may comprise a mix of individuals that migrate after completing molt and others that remain resident through the winter (Table 1). We classified 3 species—Poecile atricapillus (Black-capped Chickadee), Haemorhous mexicanus (House Finch), and Cardinalis cardinalis (Northern Cardinal)—as nonmigratory permanent residents; although some individuals in northeastern H. mexicanus populations may migrate short distances, the great majority appear to be nonmigratory.Estimation of Molt ParametersWe estimated molt start date and molt duration for the 33 species using methods developed and described previously (Mumme et al. 2025a,b). In brief, individual molt scores of birds in active molt were expressed as the proportion of remex molt completed and then arcsine-transformed to linearize the molt process and account for its slow progression at the earliest and latest stages of molt, when few remiges are being actively replaced. Transformed molt scores were then rescaled as a proportion, allowing mean ± SE molt start date and molt duration to be taken directly from estimates of the intercept and slope of regression equations obtained via either Pimm mixed-model regression (Mumme et al. 2025a), for 29 species where data were available for individuals recaptured during the same molt cycle, or standard Pimm regression (Pimm 1976), for 4 species where recapture data were lacking. These methods produce accurate and unbiased estimates of molt phenology whenever molt data are collected across a broad range of dates, and are unaffected by molt-dependent sampling bias, which is widespread in the Powdermill data (Mumme et al. 2025a).Ecological and Morphological Explanatory VariablesWe examined how interspecific variation in molt phenology is influenced by two ecological and one morphological explanatory variables: estimated migration distance, frequency of double brooding, and wing length. To estimate migration distance for the 25 completely migratory species, we used the range maps available at Birds of the World (Billerman et al. 2025) and the polygon tool of the image analysis program Fiji (Schindelin et al. 2012) to estimate the centroid of the stationary nonbreeding range; we then calculated the distance between the centroid and Powdermill. For the 5 species of partial migrants (Table 1), we used the centroid of the portion of the nonbreeding range located to the south of Powdermill. The 3 species considered to be permanent residents were assigned a migration distance of 0. For frequency of double brooding, we used the 33 species accounts at Birds of the World (Billerman et al. 2025) to qualitatively determine if double brooding was described as occasional, regular, or frequent (scored as 1 or “Yes”; 19 species) vs. absent or rare (scored as 0 or “No”; 14 species) for breeding populations in the Powdermill region. Data on wing length, a strong indicator of overall body size and one that is particularly relevant to remex molt, were obtained from birds measured during banding at Powdermill (Mulvihill et al. 2004).Statistical AnalysisIn most analyses we controlled for the non-independence resulting from shared evolutionary histories via Phylogenetic Generalized Least Squares (PGLS) regression. PGLS regressions were run with the R package caper version 1.04 (Orme et al. 2025) under R version 4.5.1 (R Core Team 2025) based on a phylogeny (Supplementary Material Figure S2) of the 33 species in our dataset drawn from McTavish et al. (2025) using the R package clootl version 0.1.2 (Cornell Lab of Ornithology Open Tree of Life, Aves tree 1.5, 2023 taxonomy; Miller et al. 2025).Our general approach was to build PGLS regression models for molt start date and molt duration that included all 3 explanatory variables: estimated migration distance (continuous), wing length (continuous), and frequency of double brooding (categorical). Initial models included interaction terms, but non-significant interactions were excluded from final models. To determine if similar statistical patterns were evident within the single family best represented in the dataset, we also built comparable PGLS models restricted to the 13 species of New World warblers (Parulidae).Because of unexpected strong bimodality in the molt duration data, our initial PGLS model for all 33 species (duration model 1) explained a relatively small proportion of the interspecific variation in molt duration (adjusted R² = 0.29). We therefore built two additional PGLS models of molt duration to account for and incorporate this unanticipated bimodality; for these analyses we excluded frequency of double brooding as an explanatory variable, as duration model 1 showed that it has virtually no effect on molt duration. In molt duration model 2, we included the slow/fast molt duration dichotomy as a categorical explanatory variable along with migration distance and wing length. In duration model 3, we included molt intensity as a continuous explanatory variable along with migration distance and wing length. Duration models 2 and 3 explained a substantially higher proportion of the total interspecific variation in molt duration, with adjusted R2 values of 0.84 and 0.79, respectively.Across all PGLS models, predictor variables showed weak collinearity (correlation coefficients < 0.6 and variance inflation factors < 2.0), and all residuals approximated a normal distribution. With the exception of molt duration model 1 (discussed above), PGLS model performance was generally good, with adjusted R² values of 0.55–0.92.Literature CitedBillerman, S. M., B. K. Keeney, G. M. Kirwan, F. Medrano, N. D. Sly, and M. G. Smith, Editors (2025). Birds of the World. Cornell Laboratory of Ornithology, Ithaca, NY, USA. https://doi.org/10.2173/bow.Ginn, H. B., and D. S. Melville (1983). Moult in Birds. British Trust for Ornithology, Tring, UK.McTavish, E. J., J. A. Gerbracht, M. T. Holder, M. J. Iliff, D. Lepage, P. C. Rasmussen, B. D. Redelings, L. L. Sánchez, and E. T. Miller (2025). A complete and dynamic tree of birds. Proceedings of the National Academy of Sciences 122(18):e2409658122.Miller, E., E. J. McTavish, and L. L. Sanchez-Reyes (2025). clootl: Fetch and explore the Cornell Lab of Ornithology Open Tree of Life avian phylogeny. https://github.com/eliotmiller/clootl.Mulvihill, R. S., R. C. Leberman, and A. J. Leppold (2004). Relationships among body mass, fat, wing length, age, and sex for 170 species of birds banded at Powdermill Nature Reserve. Eastern Bird Banding Association Monograph, no. 1.Mumme, R.L., R. S. Mulvihill, and D. Norman (2025a). Estimation of molt phenology under molt- and date-dependent sampling bias: A comparison of conventional and recapture-based methods. The Wilson Journal of Ornithology 137:532–549.Mumme, R. L., R. S. Mulvihill, and D. Norman (2025b). Sex- and age-related differences in post-breeding molt phenology are phylogenetically and ecologically widespread in passerines. Ornithology 143:ukaf039.Orme, D., R. Freckleton, G. Thomas, T. Petzoldt, S. Fritz, N., Isaac, and W. Pearse (2025). caper: Comparative analyses of phylogenetics and evolution in R, R package version 1.0.4. https://doi.org/10.32614/CRAN.package.caperPimm S. (1976). Estimation of the duration of bird molt. The Condor 78:550–553.Schindelin, J., I. Arganda-Carreras, E. Frise, V. Kaynig, M. Longair, T. Pietzsch, S. Preibisch, C. Rueden, S. Saalfeld, B. Schmid, J.-Y. Tinevez, D. J. White, V. Hartenstein, K. Eliceiri, P. Tomancak, and A. Cardona (2012). Fiji: an open-source platform for biological-image analysis. Nature Methods 9:676–682.