Introduction

The nocturnal behaviours of diurnal birds have been little studied aside from captive bird species such as pigeons and chickens. In those studies, two distinct sleep phases have been identified, namely slow-wave sleep (SWS) and rapid-eye-movement (REM) sleep (Rattenborg et al. 2009; Aulsebrook et al. 2021). Even though the neurons of birds are organised in a nuclear manner rather than laminar, there are characteristic changes in brain activity patterns during sleep of which some are remarkably similar to those features in mammals (Beckers and Rattenborg 2015). Yet, not all sleep activities are the same such as the amount of REM sleep (Roth et al. 2006) and pupillary behaviour (Ungurean et al. 2021).

Some bird studies address different aspects of sleep rather than the neuronal basis and processes. For example, data are available on how the age and sex of individuals influences sleep (Steinmeyer et al. 2010; Stuber et al. 2015) as well as temperature and artificial light (Raap et al. 2016; Stuber et al. 2017; Ren et al. 2021). Moreover, there is emerging evidence that sleep and biological rhythms are further influenced and modified by natural as well as non-natural environmental factors such as light, noise, the presence/absence of predators, and pollution. These parameters strongly differ in captivity versus natural conditions (Voirin et al. 2014; Gravett et al. 2017), with little know with respect to species-specific behaviour and its modulations under natural conditions.

The Eurasian Blue Tit (Cyanistes caeruleus, Linneus, 1758)—in the following just named Blue Tit—is a common and regular sedentary breeding bird in Germany (Barthel and Helbig 2005), having a population size of about 3.3 million breeding pairs (Südbeck et al. 2007). Blue Tits are known to use artificial nest boxes with a preference for small-diameter entries (Kempenaers and Dhondt 1991). Along with using nest boxes or secondary cavities for breeding (Newton 1998), Blue Tits roost mostly solitarily in nest boxes at least in winter time (Winkel and Hudde 1987; Mainwaring 2011). Here, we made use of these nesting behaviours combined with relatively new, easy-to-use and inexpensive imaging methods to gain insights into the nocturnal behaviours of free-living birds.

Methods

Imaging, behaviour calculations and definitions

Timestamped image data were collected in nest boxes each equipped with a motion-detecting Internet Protocol (IP) camera (Edimax IC-3110W) with a 5 mm/f2.8 lens and ten infra-red (IR) light emitting diodes (LEDs; 850 nm peak emission a wavelength range not visible to Blue Tits (Hart et al. 2000). The first and the last image triggered by bird motion in a given box were defined as ‘entry’ and ‘departure’ times. The sunrise and sunset times were calculated according to longitude (9.849° E) and latitude (51.996° N). Timing of the first and last image was used to calculate the total residence time spent in the nest box. The onset of resting was defined by the first minute after entry in which no image was taken (evening latency) (Table 1). According to this, the waking period was defined as the time between the last minute before departure in which no image was taken and last image at all (morning latency) (Table 1). Furthermore, the midpoint of sleep was calculated by subtracting the morning latency and the evening latency from total residence time and adding half residence time to the morning latency time point, resulting in a clock time referred to as midpoint of sleep (Table 1).

Table 1 Parameter list

Nocturnal activities

By analysing the number of images taken during a night, an activity pattern was created based on movements of the bird. Since our method uses the timestamp of the taken images, we cannot distinguish sleep, resting, grooming or different habits, but still we will stick to the wording ‘sleep’ if no activity was detected for a comprehensible reading. In line, we will coin the term ‘awake’ for any activity. The images taken by the camera were summarised minute-stepwise to gain the activity profile during the night (actigraph) (Supplementary Figure S1). Since the sensitivity of cameras differed within the various nest boxes, we deduced the nocturnal activity from Common House Martin box 01 only. This camera was most sensitive to motion. To evaluate a rhythmic pattern of nocturnal activity (Supplementary Figure S1), minutes in which at least three images were taken are referred as activity. Furthermore, resting periods alias sleep periods were quantified and are defined as minutes with less than three images per minute. Summarising these periods led to the portion spent awake and the portion spent asleep. In addition, we analysed the frequency (φ) and periodicity (τ) of arousals. These parameters represent the mean number of active minutes per hour and how many minutes pass between arousal events. To avoid the summary of intermitting activity, 5-min gaps were introduced in consecutive activity, resulting in so called awakenings. These peak activity periods were further analysed by summarising the minutes in which more than 3 images per minute were consecutively triggered by movement. These bouts represent the time spent awake during maximum activity (awakening duration). Finally, we quantified the time without any image in a row representing so-called sleep bouts. All parameters are summarised in a schematic (Supplementary Figure S1) and in a table including the calculations (Table 1).

Weather data

Weather data were obtained from a weather station 10 km away from the study site, which is within the range used in other studies (Schlicht and Kempenaers 2020). Data (rain, temperatures, sunlight) were provided by meteostat.org (Lamprecht 2021) (Supplementary Figure S2), and daylength was provided by Time and Date AS (Steffen Thorsen 2022).

Data analysis

Linear regression analysis and Pearson correlation were used to examine correlations between certain predictor variables such as temperature (minimal and average) and rain and the response variables such as \(\varphi\) arousal, evening and morning latencies as well as active hours. All parameters were analysed using a self-written Matlab script. Time calculations were performed with Microsoft Excel 2016. Statistics were carried out using GraphPad Prism 8 software. All data points are depicted within the graphs as grey dots (scatter plot) underlying the colour-coded mean values of each recording session and the corresponding inter-sessional standard error.

Results

This study was conducted from November 2014 through July 2020 at the ‘Entdeckerturm’ Langenholzen, which is a bird observation hide near Alfeld/Leine, Germany (52,0°N 9.85°E) designed to pique people’s interest in nature and wildlife (Supplementary Figure S3 A and B). The hide is equipped with six nest boxes for Common House Martins (Schwegler box 9B, east side), ten nest boxes for Common Swifts (north and west side), a Common Kestrel box (south side) and ten bat boxes (north side). Observations made around the study site and of the nest boxes were reported to the citizen science platform naturgucker.de and can be followed at http://naturgucker.de/?gebiet=-1543032447&tab=2.

To monitor birds’ nesting activities, 12 nest boxes were manually equipped with motion-detecting IP cameras. Each camera was set up to take an image upon the detection of movement, with the images timestamped and automatically emailed for evaluation.

Before starting to roost in a given nest box, different species of birds were observed exploring the nest boxes several times during the day. Due to a smaller entry diameter, the nest boxes designed for Common House Martins were nearly exclusively explored then occupied by Blue Tits, whereas the other nest boxes, especially the Common Swifts boxes, were also visited by Great Tits (Parus major, Linneus, 1758), Eurasian Tree Sparrows (Passer montanus, Linnaeus, 1758), House Sparrows (Passer domesticus, Linneus, 1758), Eurasian Wrens (Troglodytes troglodytes, Linneus, 1758), and Eurasian Nuthatch (Sitta europaea Linneus, 1758).

In total, four nest boxes (1–4) for Common House Martins and two bat boxes (1 + 2) were regularly occupied by Blue Tits (Supplementary Figure S3 B). As seen in Table 2, Common House Martin box 1 was most favoured, followed by bat box 2. The nest box visits during the night varied among different years (Table 2). Very rarely did the data suggest a bird changed nest boxes during the night (from 03.12. to 04.12.2018, Session 5, from bat box 2 to bat box 1). Of note, once out of five New Year’s Eve nights a bird left the nest box, potentially due to fireworks (2015/2016 00:11 a.m.) (Bosch and Lurz 2019).

Table 2 Summary of recorded nights

Seasonal dependence on sleep timing and duration

The average times spent each month in the nest boxes showed a strong seasonal rhythm, whereas in the dark winter months, the Blue Tits stayed for around 14 to 16 h (October through February) within the nest box, they only roosted on average 8 h in May (Fig. 1A). Interestingly, the midpoint of sleep—a circadian key parameter—varied during October through July by only half an hour (Fig. 1B). Even though the length of the stay in the nest box continuously increased as the nights got longer, the portion spent resting and active during the night remained constant. From November through April, the time spent awake varied from 4 to 5.5% and in turn the time spent asleep was from 94.6 to 96%. Only starting from May onwards, the portions slightly change: The time awake increased from 11 to 20%, whereas time spent asleep decreased from 80 to 88%. Overall, a bird spent on average 94.0 ± 4.3% resting and 6.0 ± 4.3% active (Fig. 1C).

Fig. 1
figure 1

Seasonal variation of nest box stay and sleep duration: A Total time spent by Blue Tits in a nest box. The mean value of different recording sessions are colour-coded (see legend) and grey shows all values. B Midpoint of sleep for all sessions (as Central European Time; CET). C Fraction of sleep and awake/arousal state in different seasons

The timing of going to rest and starting the activity was strongly dependent on the daylength in the experimental seasonal periods. The entry and the departure times showed a clear correlation to sunset and sunrise, respectively. From November through February, Blue Tits birds entered the nest boxes usually at least 10 min after sunset (Fig. 2A). Additionally, the birds usually left the nest boxes 30 min prior to sunrise, independent of the session and season (Fig. 2B).

Fig. 2
figure 2

Seasonal variation of entry, departure and resting. A Entry of the Blue Tits relative to sunset (dashed line) in minutes across the sessions. B Departure of the bird relative to sunrise (dashed line) in minutes. C The time the birds need to calm down and roost. D The time the birds became active before departing the nest box

On average, the Blue Tits birds spent 3–5 min each night coming to rest, termed ‘evening latency’ (Fig. 2C). In the morning, the birds spent on average 4–7 min preparing for the day before departing the nest box (morning latency) (Fig. 2D).

Resting–activity cycles during the night

The Blue Tits activity during night-time hours appeared as a repetitive pattern of high activity and resting periods (Supplementary Figure S1). Our analyses revealed regular arousals (φ arousal) and a repetition of consecutive activity periods (τ) ranging from 66 min in November to 130 min in February (Fig. 3A, B). This rhythmic pattern led to several awakenings per night ranging from 1 up to 18 activity maxima during each night (Fig. 3C) lasting between 1 and 6 min (Fig. 3D). Extreme restless nights of up to 21 activity maxima were also recorded. Of note, not every night showed this regular pattern of activity. Sometimes a given bird was more active with shorter resting periods and more active bouts. Birds were only active for a minor fraction of the time, so we also quantified sleep durations (Supplementary Figure S1). The mean sleep bout duration during a night varied from 114 ± 94 min in May up to 302 ± 194 min in December. Except in July, the Blue Tits birds slept on average the longest time in December and January. The sleep periods were shortest in May and June with a minimum of only 42- and 50-min sleep in a row, respectively (Fig. 3E).

Fig. 3
figure 3

Sleep parameter: A the frequency of arousals, whereas B depicts period (τ) of awakenings (extreme values are graphically omitted in grey). C The number of awakening events per night (extreme values are graphically omitted in grey). D The median duration of awakening events in minutes for different sessions. E The mean duration of a sleep bout

Environmental factors influencing sleep and activity parameters

The correlation of the arousal frequency against the ambient temperature (which is usually 2 °C higher in the nest boxes, at least during the wintertime) showed at least for some seasons a significant correlation. Our data indicate that warmer temperatures during the night (T min) resulted in increased arousal frequencies (Fig. 4A, Table 3). In contrast, our data show no influence of the temperature on the evening or morning latency periods and thus does not impact falling asleep or morning awakening, even though the Common House Martin nest boxes face east (Fig. 4B, C, Table 3). We also found no connection between the temperature (here daily average temperature) and the activity hours of the birds for the four coldest months (November through February). Except for season 4, the data show no correlation between cold temperatures and prolonged daily activity (Fig. 4D, Table 3).

Fig. 4
figure 4

Parameter correlations: A The dependency of arousal events to the minimum temperature across the sessions. B, C Comparisons of the first (A) and the last (B) resting periods versus the minimum temperature. D Comparison of active hours versus daily average temperature for the four coldest months (November till February). E Comparison of active hours and total daily precipitation

Table 3 Pearson correlation of predictors and response variables

Finally, we addressed whether rainfall modulated the daily activity of the Blue Tits. When comparing total rainfall against the daytime activity of the birds, we found a correlation for at least some sessions (session 2, 4 and 5), where precipitation shortened the time, the birds were active outside the nest boxes. In three of 6 sessions, rainfall shortened the daytime activity by around 7 ± 4 min per 1 mm additional rainfall (Fig. 4E). All these findings are corroborated by a Pearson correlation of the above-mentioned variables (Table 3).

Discussion

This study presents data on the nocturnal behaviours of free-living Blue Tits in a rural setting. We found a clear seasonal influence on sleep behaviour, mainly determined by the presence or absence of daylight. Furthermore, we found ultradian sleep–wake rhythms during the night and could demonstrate that environmental cues such as ambient temperature and rain events can modulate the behaviour of the birds.

The Blue Tits birds visited different nest boxes, not only within one recording session, but also during different sessions. We know that several birds were sleeping in given nest boxes, since very occasionally we could observe two Blue Tits sleeping at the same time in different nest boxes. Unfortunately, we cannot pin down the total number nor the gender of the birds that slept in the nest boxes. Thus, this could result in some higher variation of the evaluated parameters, given that individual sleep behaviour have been shown to differ (Steinmeyer et al. 2010; Mueller et al. 2012).

When looking at the coldest months, we could not clearly see a prolongation of activity time by colder temperatures which might be a consequence of harder foraging. Our experimental setup included a bird feeder close to Entdeckerturm; it is possible that cold weather might have increased the number of bird feeder visits (Bonter et al. 2013), which is a parameter not monitored in this study.

Our data show that the Blue Tits’ roost timing was mainly determined by light. During winter the birds roosted within the nest box 14.5 h in total in November, and this declined to 9 h in May. These data perfectly match published data about the sleep of Blue Tits (Steinmeyer et al. 2010). Even though the sleep duration varied between 8 and more than 15 h, the midpoint of sleep was observed to be little changed. The birds spent around 2 to 6 percent of the night awake, which is in line with published data (Steinmeyer et al. 2010). The birds entered the nest box to roost slightly after sunset. This timing changed slightly during the seasons and the time difference towards sunset decreased with seasonal progression, resulting in entry around sunset in March and April. In contrast, in summer (June and July), birds entered the nest boxes to roost well before sunset. The departure of the birds in the morning did not show any seasonal effect. The birds usually left the roosting site around 30 min prior to sunrise. Again, exceptions were found in June and July in which the birds left around sunrise.

Evening and morning latency showed no seasonal variation, and both parameters were not influenced by the environmental temperature. Evening latencies were shorter and less variable than morning latencies. All these findings are in line with Blue Tits’ sleep evaluated by (Steinmeyer et al. 2010). The frequency of arousals (φ arousal) during night changed slightly during the season and decrease towards February and March and increase again in April. Starting with the breeding time, the arousal frequency rapidly increased in May through July, whereas the mean duration of the awakenings stayed constant throughout the whole session. All parameters point to a seasonal dependent sleep quality in Blue Tits and are in line with published data (Steinmeyer et al. 2010). Most likely, slight and minor differences can be best explained by our technique, since it was not possible for us to discriminate between awake (eyes open, eyes closed), grooming, preening or stretching. Our data are based on consecutive motion detection of the cameras and not behavioural assessment of the bird by images as conducted in most studies.

Compared to the above-discussed sleep behaviour of Blue Tits, we were also able—by chance—to directly compare some parameters with Great Tits (Session 6) and Common Starlings (session4) using the same strategy. In contrast to Blue Tits, both Great Tits and the Common Starling chose a Common Swift nest box to roost in. The total time the birds spent within the nest boxes did not differ among species (Supplementary Figure S4 A). It might be that Great Tits as well as Common Starlings are later chronotypes as indicated by the midpoint of sleep, which is slightly delayed compared to Blue Tits (Supplementary figure S4 B). This later phenotype of Great Tits and Common Starlings is mainly determined by a later departure in the morning. Blue Tits left the nest boxes 20 to 30 min prior to sunrise, whereas Great Tits and Common Starlings left around sunrise. There is no huge difference in the entry time of these different species (Supplementary Figure S4 D, E). It is known that both species (Great Tits and Common Starlings) show seasonal variation in sleep parameters (Stuber et al. 2015; van Hasselt et al. 2020).

In addition, we found a dependency of the frequency of awakening on the temperature. Our measurements of the temperature showed that an increase of the environmental temperature of 1 °C led to an increase within the nest box of 1.8 °C; in front of the camera, it might be even warmer. Taking this into account, we found that warmer nights led to more awakenings of the birds. This dependency was also significant in a Pearson correlation and is also in line with published data (Stuber et al. 2017). However, our data also show that this particular phenomenon is more pronounced in cold roosting times, since the session 6 (only May, June, July), does not show this pattern due to overall high awakening frequencies. This finding is particularly important, since winters are getting milder in times of climate change, and this could lead to less effective sleep in songbirds, contributing to lower fitness of the birds. Even though sleep quality seems not to be linked to reproductive success (Steinmeyer et al. 2013), detrimental effects on the individual cannot be ruled out and needs to be analysed in the long term. In addition, we found that rain, which is another environmental parameter, shortened daytime activity of the Blue Tits, which is in line with published data (Schlicht and Kempenaers 2020). Differences between the sessions might again be explained by inter-individual differences such as sex and age (Steinmeyer et al. 2010; Schlicht and Kempenaers 2020).

The method—IP motion-detecting cameras, sending emails and/or saving images with a time stamp—used in this study can—in contrast to high-performance laboratory use-only devices such as ONEIROS (Massot et al. 2019)—represent a cheap, efficient and powerful tool to investigate wild animals and their behaviours without any manipulation. The data can be acquired in a remote fashion and analysed offline. Motion-detecting cameras are already used as traps, e.g. to study wildlife in remote areas, mainly to survey species inventory of a certain location and detect the presence or absence of hidden and hard-to-observe species (Norouzzadeh et al. 2018; Jansen et al. 2020). There are already efforts to extend camera traps towards data collectors to study the activity pattern of frogs, for example, (Barata et al. 2018) or different behaviours such as roosting of white-bellied herons (Khandu et al. 2020) or sleep of black rhinos (Santymire et al. 2012). This manuscript further corroborates the usefulness of IP motion-detecting cameras to investigate and, more importantly, quantify behaviour.