Study species and study site
Savannah Sparrows are small migratory songbirds that live in open grasslands throughout North America (Wheelwright and Rising 2008). In the spring, males migrate from wintering grounds in the southern United States to breeding grounds across the United States and Canada (Bédard and LaPointe 1984; Tufts 1986; Woodworth et al. 2016). We conducted our research at the Bowdoin Scientific Station on Kent Island, New Brunswick, Canada (44°35′N, 66°46′W). Our island-living, colour-marked population of Savannah Sparrows has been studied for more than three decades (Williams et al. 2013) and is known to exhibit strong site philopatry (Wheelwright and Mauck 1998). Males in this population arrive from migration in mid-April, approximately one month before female arrival, and begin to defend breeding territories from rival males (Woodworth et al. 2016). Males regularly produce songs which they learn by emulating vocal tutors heard during their natal year and at the outset of their first breeding season (Mennill et al. 2018). Songs are made up of a complex series of notes produced in a consistent and stereotyped order (Fig. 1). Each male produces a single song type that is individually distinctive (Williams et al. 2013). Song persists between the arrival from spring migration and departure for fall migration, but the pattern of this variation has not been studied previously in quantitative detail.
Recording technique
In 2014, we recorded male Savannah Sparrows using automated digital recorders (Wildlife Acoustics Song Meters, model: SM2; details in Mennill et al. 2012). We used eight different recorders, each mounted at a height of 1 m on a wooden stake, and moved the eight recorders around among 13 locations at the study site, allowing us to monitor the 34 males. Each recorder location was separated by 100‒150 m. Given the small territories of Savannah Sparrows (territory diameter is less than 50 m, and often as small as 25 m), each autonomous recorder allowed us to record between one and seven males simultaneously. We are confident that our recorders sampled all songs produced by males on their territories, because territories are small and songs are routinely heard from at least two territories away. In-person observation sessions confirmed that males only sing from within the boundaries of their territories, and that song posts were within range of the recorders. In total we recorded n = 34 males for ten 24-h long recording sessions between mid-April and mid-September, with approximately 14-day intervals between subsequent recording sessions at each of the 13 microphone locations. In all of the recordings where multiple males were present, we distinguished individuals on the basis of their individually distinctive song. The song of each individual was known from in-person focal recording sessions, which allowed us to connect each song to a specific colour-banded individual (as in Williams et al. 2013; Mennill et al. 2018).
Analysis of pairing and breeding status
To determine the breeding activities of the 34 males, we collected behavioural observations and monitored nests every 2 days. We divided the breeding season into five different stages (as in Foote and Barber 2009): (1) the “pre-pairing stage”, a time-period when males had arrived on the breeding ground but had not yet attracted a mate (in our study population this stage begins in mid-April or late-April and persists for approximately 15 days, until females arrive and pair with males); (2) the “fertile stage”, a time-period when males had paired with a fertile female (8 days prior to the laying of the penultimate egg; in our study population this stage occurs between early and late May); (3) the “incubation stage”, a time-period when males were paired with a female incubating eggs (a period of 12 days; in our study population this stage typically occurs in early June); (4) the “hatchling stage”, a time-period when males were paired with a female who was provisioning nestlings (a period of 9 days; in our study population this stage typically occurs in late June); and (5) the “fledgling stage”, a time-period when males were paired with a female provisioning fledglings (starting with the end of the nestling stage to the start of the next brood, or, if no consecutive brood is attempted, a period of 7 days after the end of the nestling stage). We estimated the length of the fertile period, relative to egg-laying dates, based on the following logic: female Savannah Sparrows are thought to be fertile during the 1‒3 days when nests are built and during the following period of 3‒5 days when eggs are laid (1 egg per day). In cases where nests were found after laying was complete, we estimated the first day of incubation by back-dating 12 days from known hatch dates (Dixon 1978; Wheelwright and Rising 2008). The first nesting attempts of birds in our study population are quite synchronous. For re-nesting attempts, stages 2 through 5 were repeated later in the summer. Thus the timing of each breeding stage varied individually, especially as the breeding season progressed and the breeding activities of the study animals became asynchronous.
Analysis of diel and seasonal variation
We analyzed field recordings by visually scanning sound spectrograms using Syrinx-PC Sound Analysis Software (J. Burt, Seattle, WA). This software allowed us to visualize 5 min of recording at a time, to compare the field recordings to an on-screen reference recording of each individual, and annotate the songs with a time-stamped annotation. From these annotations we tallied the vocal output of each male.
We defined the “dawn chorus” as songs that occurred during the period 30 min before sunrise until sunrise (as in Liu 2004; Naguib et al. 2016). We defined “daytime song” as songs that occurred between sunrise and sunset. We defined the “dusk chorus” as songs that occurred following sunset. Sunrise and sunset times for each day were obtained from the National Research Council, Herzberg Institute of Astrophysics sunrise/sunset calculator (www.nrc-cnrc.gc.ca) for the nearby city of Saint John, New Brunswick. In addition to tallying the songs produced by birds during the three periods of sunrise, daytime, and sunset, we also calculated hourly values of song output during 1-h bins. We excluded days where the recordings showed a heavy influence of weather (i.e. very rainy or windy days) and we could not be confident that we had sampled all songs. The number of songs recorded in September was typically zero, and we did not include zero-song recordings in our analysis because we could not be certain if the bird remained in the area of the recorder at that time (birds may have already begun their southward migration, or territory boundaries may have eroded after breeding had concluded). On average, we included 7 ± 3 days of recording spread across the season for each individual.
Statistical analysis
We analyzed diel and seasonal variation in song output using linear mixed models (LMM). Our fixed effects were time of day (subdivided into 1-h periods; or subdivided into bins of dawn, daytime, and dusk), breeding stage (pre-pairing, fertile, incubation, hatchling, or fledgling), and interaction between time of day and breeding stage. We included male identity as a random effect to account for the fact that the same males were sampled repeatedly. For any analysis that showed statistical significance for the fixed effects, we conducted a Tukey’s post hoc test of honestly significant differences. We used JMP (v14 SAS Institute Inc. 2019) for all statistical analyses.