Light pollution alters the phenology of dawn and dusk singing in common European songbirds

(a) Study sites and data collection

Each day, between 6 January and 17 April 2012, we recorded all bird vocalizations around dawn and dusk in 12 forested sites (0.6–1.8 ha) in Southern Germany. Sites were chosen such that (i) half were affected by artificial night lighting (street lamps), (ii) within each light ‘treatment’, half were affected by traffic noise (vehicles driving on a busy road adjacent to the site), (iii) they were in non-urban areas and as similar as possible in parameters other than noise and light. Data from different sites can be considered to be independent, because sites were between 0.5 and 28 km apart and chosen to avoid clustering of the same light conditions. Two pairs of one lighted and one unlighted site were selected along the same road, allowing direct comparison under similar noise conditions. For a detailed description of the sites (one site had to be excluded from most analyses) and their locations, see [17].

At each site, we placed two Song Meter SM2+ (Wildlife Acoustics, Concord, MA; http://www.wildlifeacoustics.com/products/song-meter-sm2-birds) recorders on the ground, 70–130 m apart, in order to maximize song detection. We programmed each device to record sounds (stereo, sampling rate 22 050 per second) between 1.5 h before local sunrise until 1.5 h after local sunset (times based on the coordinates of each plot). Sound files were stored as wav files onto Secure Digital High Capacity cards (Laxer, Fremont, CA).

Each recording device also contained a temperature sensor (inside the box), which we programmed to log air temperature every 5 min. Temperature data were stored as text files onto the same digital data cards.

(b) Data extraction

We analysed each recording using Song Scope v. 4.1.1 (Wildlife Acoustics, Concord, MA; http://www.wildlifeacoustics.com/products/song-scope-overview), as explained in detail in [17]. We noted the song of the six most common species at the study sites: European robin Erithacus rubecula, common blackbird, song thrush Turdus philomelos, great tit Parus major, blue tit and common chaffinch Fringilla coelebs.

On each day between 6 January and 17 April 2012, we noted for each species and at each recorder, whether dawn or dusk song was produced (yes/no). This was the case whenever we detected at least three song repetitions (strophes) within less than 5 min during the relevant period. The dawn chorus was broadly defined as singing that occurred in the period between 1.5 h before sunrise until 1.5 h after sunrise. The dawn chorus usually started before sunrise. Mean onset of singing (±s.d.) in min from sunrise: robin: −57.7 ± 19.1 (n = 686 recording days), blackbird: −53.2 ± 17.3 (n = 869), song thrush: −49.1 ± 14.8 (n = 522), great tit: −31.4 ± 25.6 (n = 1018), blue tit: −16.2 ± 25.3 (n = 874), chaffinch: −9.1 ± 18.0 (n = 766). The dusk chorus was broadly defined as singing that occurred in the period between 1.5 h before sunset until 1.5 h after sunset. The dusk chorus usually stopped before sunset for the great tit, the blue tit and the chaffinch, and after sunset for the robin, the song thrush and the blackbird. Mean cessation of singing (±s.d.) in minutes from sunset: robin: 29.8 ± 20.6 (n = 515), song thrush: 27.5 ± 11.3 (n = 509), blackbird: 19.3 ± 14.1 (n = 817), great tit: −16.0 ± 18.8 (n = 931), blue tit: −20.4 ± 23.7 (n = 737), chaffinch: −26.0 ± 22.0 (n = 462). In total, we analysed 1579 recorder days for dawn singing (i.e. the sum of the number of days analysed for each recorder, equivalent to 4737 h) and 1444 recorder days (4332 h) for dusk singing.

During each dawn and dusk period (as defined above) and for each recorder, we noted the presence of rain (yes/no; recognizable on the sonogram as broad-frequency, low-amplitude, continuous sound), and extracted temperature at sunrise/sunset from the text files. We excluded recordings when heavy rainfall made song detection unreliable (n = 6 days at dawn, n = 4 days at dusk).

(c) Statistical analysis

All statistical analyses were performed with R v. 3.1.0 [26] and the R-package lme4 v. 1.1.7 [27]. We used generalized linear mixed models with binomial error distribution (fitted by the Laplace transformation), with ‘site’ and ‘recorder nested within site’ as random effects to control for variation owing to site and for similarity between the two recorders at each site. For each species and each period (dawn/dusk) separately, we tested whether the probability of singing was determined by the presence of artificial light (factor ‘light’: yes/no) in interaction with ‘date’, and by ‘rain’ (yes/no) and ‘relative temperature’. Because temperature and date were strongly correlated (Pearson correlations: dawn: r = 0.50, n = 87, p < 0.001; dusk: r = 0.66, n = 85, p < 0.001), we computed relative temperature as residuals of a mixed effect model with temperature (°C) as dependent variable, ‘date’ as fixed effect with random slope and ‘site’ as random intercept. For the song analyses, the intercept for the date was set to 1 January for the two tit species. For the other species, that started singing from mid-February onwards, the intercept was set at the median date for which they were singing in half of the sites (robin: 1 March, blackbird: 21 February, song thrush: 3 March, chaffinch: 23 February). We did not include traffic noise as a variable in the models, because (i) previous work suggested that traffic noise did not affect the daily timing of dawn and dusk singing [17] and (ii) there is no clear prediction about how traffic noise would affect the seasonal timing of dawn or dusk singing. We controlled for between-site variation in the presence or density of each species by removing those recorders where the focal species was singing during less than one quarter of the entire recording period from 15 February onwards (number of recorders excluded at dawn/dusk: robin 0/1, song thrush: 2/1 (absent from one lighted site), blue tit: 2/2, chaffinch: 1/5). We checked whether the song of the same individual was picked up by both recorders at a site. If this was the case (only one song thrush at each of two lighted sites), we removed the data from one recorder. In winter, malfunctioning of the recorders owing to battery problems led to missing data (January: 9 days at dawn, 12 at dusk; February: 6 days at dawn, 7 at dusk).

All tests are two-tailed, and p-values lower than 0.05 are considered significant. We report means and their standard errors.

(a) Natural variation in the phenology of dawn and dusk singing

As expected, for all species, the probability of singing at dawn and at dusk increased from winter to breeding (main effect of date; figure 1 and tables 1 and 2). The two tit species had already started dawn and dusk singing when the recordings started, but the other species only started producing dawn or dusk song much later. Apart from isolated instances in early winter (five for robins, three for blackbirds, seven of them in the same lighted site), the earliest dawn chorus was recorded on 16 February (blackbird). For all species except the song thrush the dawn chorus seems to develop earlier in the season than the dusk chorus (figure 1).

• Download figure
• Open in new tab
• Download PowerPoint

(b) Effect of artificial night lighting on the phenology of dawn and dusk singing

Overall, the probability of dawn or dusk singing was affected by the presence of street lighting in five out of six species (interaction between light and date, or a main light effect; tables 1 and 2). The main effect of light reflects a difference in the probability of singing at dawn or dusk during the period when this probability is increasing, because all species are singing almost every day and at every site later in the recording period (figure 1).

Robins, blackbirds and great tits were more likely to produce a dawn chorus earlier in the season in the lighted sites compared with the dark sites (figure 1 and table 1). The blue tit showed a similar, but non-significant trend (figure 1 and table 1). The effect was in the opposite direction for song thrushes; they produced dawn song somewhat later in the season in the sites affected by artificial night lighting (figure 1 and table 1). In the chaffinch, the probability of dawn singing did not differ between lighted and dark sites across the entire recording period (figure 1 and table 1).

Similar to the dawn chorus, robins, blackbirds, great tits and blue tits were more likely to produce a dusk chorus earlier in the season in the lighted sites compared with the dark sites (figure 1 and table 2). In the song thrush and the chaffinch, the probability of dusk singing did not differ between lighted and dark sites across the season (figure 1 and table 2).

(c) Effect of weather on the phenology of dawn and dusk singing

For all species, the probability of singing at dawn or dusk was reduced when it was raining, with the song thrush (not significant for dawn song) and the blackbird (not significant for dusk song) being the least affected (tables 1 and 2). Except for the song thrush and the blue tit, all species were more likely to produce a dawn chorus on days that were warm relative to the time of year (table 1), and a similar effect was detected for the dusk chorus (table 2).

(a) The phenology of dawn and dusk singing

The six species under investigation showed a marked difference in their seasonal timing of singing (figure 1). Blue tits and great tits started dawn and dusk song earliest. The great tit was the most consistent in producing dawn and dusk singing over the entire period. In this species—and in the closely related blue tit—winter singing is more or less common, depending on the year [23]. Pairs can form or stay stable over winter, and males may use song to keep in touch with the partner [23]. Resident males may also use song to announce territory ownership inside winter flocks [28]. Resident chaffinches and blackbirds commenced morning and evening singing in mid-February (figure 1), on days of warm weather, and when food became more available [29]. They were followed later in February by the robin, which may be a partial migrant [30] in the area. Song thrushes started dawn and dusk singing the latest in the season (figure 1); this species rarely winters in Bavaria and most birds arrived in the study area in early March.

As the season progressed and breeding approached, the likelihood of dawn and dusk singing clearly increased (figure 1). This effect of date is mimicked by the daily timing of singing: all species started singing earlier relative to sunrise and (all except the blue and great tit) later relative to sunset closer to the start of breeding [17]. In the blue tit, this seasonal effect was also observed in sleep duration, even when controlling for differences in day length: males (and females) slept less as the season progressed [31].

Interestingly, dawn singing seems more dominant in early spring, with the dusk chorus becoming more prevalent later in the season, especially in the tit species, in line with Slagsvold [29]. The song thrush appears to invest more than the other species in the evening song peak, for reasons that remain to be studied.

In line with other work on song activity [17,29,32–35], we found that dawn and dusk singing was less likely on relatively cold or rainy days. Rain and colder temperatures may negatively impact song production because of the associated costs of singing in adverse conditions, or because other behaviours (e.g. foraging) become more important [33].

(b) Effects of artificial night lighting on the phenology of dawn and dusk singing

At sites with artificial night lighting all species except the song thrush and the chaffinch were more likely to produce dawn and dusk song earlier in the season (figure 1). Thus, our study suggests that light pollution can lead to a faster seasonal development of the dawn and the dusk chorus. Interestingly, the effect was most pronounced in the robin and the blackbird (figure 1), which are the two species that naturally sing earliest at dawn and that are most affected by light in terms of earlier daily singing, whereas absent in the chaffinch, a species whose daily timing of singing was not affected by light [14,17]. The only exception is the song thrush, where artificial night lighting was associated with a (minor) delay in the development of dawn singing (figure 1), despite a similar effect on the daily timing of dawn song as in the robin and blackbird [17]. We discuss three non-mutually exclusive mechanisms that might explain the overall earlier annual initiation of dawn and dusk song peaks in sites with artificial light at night.

(c) Conclusions and outlook

Our study suggests that artificial light at night advances the seasonal occurrence of singing in those songbird species that are also affected by light with respect to their daily timing of dawn and dusk singing (with the exception of the song thrush). Although we controlled for potentially confounding factors such as temperature, traffic noise and bird density (for more detail, see Da Silva et al. [17]), differences between lighted and non-lighted sites in, for example, quality of the males or arrival time could still have influenced our results. Experimental studies are now needed to confirm that the observed effects are indeed causally linked to the presence of artificial night lighting.

Our results add to accumulating evidence that light pollution has the potential to alter natural seasonal rhythms [18,20]. This and previous work on effects of light pollution suggest that birds breeding in lighted environments become territorial earlier and breed earlier than those in naturally dark habitats. However, the evolutionary consequences of the observed effects remain unknown. Singing earlier in the year could have positive fitness consequences for males, for example increasing the likelihood of attracting a (high-quality) social mate, or of siring extra-pair offspring [14]. Pairs could also compensate for smaller clutch sizes or lower productivity per nesting attempt in urban habitats by producing multiple broods owing to longer breeding seasons [64]. On the other hand, earlier singing or singing over a longer period may also come at a survival cost, owing to an increased risk of predation or because of exhaustion or elevated stress levels. Long-term individual-based studies are needed to address these issues, and to obtain a better understanding of the evolutionary consequences of artificial night lighting.