Many species have shifted their breeding phenology in response to climate change. Identifying the magnitude of phenological shifts and whether climate-mediated selection drives these shifts is key for determining species’ resilience to climate change. Birds are a strong model for studying phenological shifts due to numerous long-term research studies; however, generalities pertaining to drivers of phenological shifts will emerge only as we add study species that differ in life history and geography. We investigated 32 years of reproductive timing in a non-migratory population of dark-eyed juncos (Junco hyemalis). We predicted that plasticity in reproductive timing would allow females to breed earlier in warmer springs. We also predicted that selection would favour earlier breeding, and we asked whether temperatures throughout the breeding season would predict the strength of selection. To test these predictions, we examined temporal changes in the annual median date for reproductive onset (i.e., first egg date), and we used a sliding window analysis to identify monthly spring temperatures driving these patterns. Next, we explored plasticity in reproductive timing and asked whether selection favoured earlier breeding. Lastly, we used a sliding window analysis to identify the time during the breeding season that temperature was most associated with selection favouring earlier breeding. First egg dates occurred earlier over time and strongly covaried with April temperatures. Further, for individual females that bred in more than one year, they typically bred earlier in warmer Aprils, exhibiting plastic responses to April temperature. We also found significant overall selection favouring earlier breeding (i.e., higher relative fitness with earlier first egg dates) and variation in selection for earlier breeding over time. However, temperature across diverse climatic windows did not predict the strength of selection. Our findings provide further evidence for the role of phenotypic plasticity in shifting phenology in response to earlier springs. We provide evidence for the role of selection favouring earlier breeding, regardless of temperature, thus setting the stage for adaptive changes in female breeding phenology. We suggest for multi-brooded birds that advancing first egg dates likely increases the length of the breeding season, and therefore, reproductive success.

Study system and breeding data

Since 1983, a breeding population of Dark-eyed Juncos has been monitored at Mountain Lake Biological Station (MLBS) and the surrounding Jefferson National Forest (37°22’N, 80°32’W) (Chandler et al. 1994). All monitoring protocols were approved by Indiana University’s Institutional Animal Care and Use Protocol (#12-050). At the beginning of each breeding season (April-May), birds on the study site were caught using mist nets or Potter traps and banded with a unique USFWS metal band and distinctive combinations of colour bands. Researchers searched for nests every year, identifying parents and tracking the progress of the nest. First egg date, expressed as ordinal date, was observed directly, or for nests found after the start of egg-laying, was calculated based on the day nestlings hatched or left the nest (Nolan et al. 2002). Breeding data from 1983–2015 were used for this study except for 2013 due to limited research effort. Records where female ID or first egg date were unknown were removed. Female subjects that were implanted with exogenous testosterone during a separate five-year study in the population were also removed (Clotfelter et al. 2004; Ketterson et al. 2005).

To calculate true first egg dates, we excluded any known re-nests. Also, knowing that the first nest found for a female might not be her true first nest, we eliminated nests whose first egg dates came later than each year’s median first egg date from known re-nests. Our data filtering resulted in 1,244 first nests of 935 female juncos between 1983 and 2015; females had one to five years of data (x̃ = 1). Annual differences in research effort (number of nests found) did not explain variation in first egg dates (see Supplementary Materials; Fig. S1). Because the distributions of first egg dates were not normal in some years, we calculated median annual first egg dates from first nests. Using both first nests and re-nests for each year, we calculated the annual total number of eggs and total number of fledglings produced by each female. Females were grouped into two age classes based on plumage (Pyle 1997) or records from previous breeding seasons: second years (SY; first breeding season) and after second years (ASY; second or later breeding season).  

Temperature data

Between November 16, 1971 and January 31, 1998, temperature data (daily minimum; T_min and maximum temperature; T_max) were collected from MLBS via a National Oceanic and Atmospheric Administration (NOAA) weather station (Network ID GHCND: USC00445828, hereafter, “Logger A”). On June 24, 1994, a second data logger (Campbell CR10) was established at MLBS that records temperature every half hour. To permit comparing data across devices, we calculated daily T_max and T_min from this MLBS data logger (hereafter, “Logger B”). From T_min and T_max, we calculated a daily midpoint (median) temperature (T_mid) for both loggers. Since the two weather stations overlapped from 1994-1997, we confirmed that Datasets A and B were strongly correlated and then combined the datasets (see Supplementary Materials).

Monthly average T_min, T_max, and T_mid were calculated for March–August for each year. Data were available for all years (1983-2015), except for missing March and April data for 1991 and 2002 and missing March data for 2004.

Files have been analysed with R.