Introduction

Singing is the most complex form of communication in songbirds (Oscines), primary serving to establish territory and attract a mate for reproduction1. Most bird species are active during the day, which is also reflected in their vocal activity. The intensity of singing varies throughout the day and is higher in the mornings and evenings due to better sound propagation conditions2. However, typical diurnal birds have also been known to sing at night3,4. This behaviour is assumed to provide additional benefits, such as an increased chance of producing offspring and enhanced territory defence. This is supported by research on the willie wagtail (Rhipidura leucophrys), proving the role of males’ nocturnal and diurnal songs in mate attraction and territory defence5. On the other hand, singing at night is costly both energetically (e.g., decreased resting time, increased energy expenditure due to activity in lower nighttime temperatures6) and ecologically (increased risk of predation).

Nocturnal singing by diurnal birds is geographically variable; however, the current state of knowledge limits the ability to draw more general conclusions. For example, in North America, nocturnal singing by diurnal birds has been documented in at least 51 of 82 analysed families, and over 70% of the species that vocalise nocturnally are considered diurnal4. In Central Europe, this behaviour has been observed in 15 of 32 families, and over 85% of nocturnal singing birds are typically diurnal species3. These studies have shown that nocturnal singing by diurnal birds is quite common in temperate regions, allowing bird species to be grouped into regular, occasional, and rare nocturnal singers. On the other hand, a study on bird species from the Afrotropical highlands found that only three of 50 typically diurnal species sang during astronomical night, producing just 10 songs7. Most likely, tropical birds have been reported to undertake less risky behaviours than temperate birds8, either due to different life histories of species from various regions or to avoid competition for acoustic space with other animals (e.g., insects or anurans) vocalising at night.

The nocturnal singing of diurnal birds has been confirmed to be stimulated by additional light in the environment9,10. Light is transformed into a nerve impulse, causing neurological and hormonal changes in the brain responsible for singing11. Both anthropogenic and natural sources of light can stimulate nocturnal singing. In light-polluted areas, some birds modify their singing time12. For example, the american robin (Turdus migratorius) sings intensely at night in areas with a large amount of artificial light, but does not vocalise in dark areas during the night13. In contrast, the willie wagtail (Rhipidura leucophrys) sings less at night in light sources, such as streetlights, compared to dark urban and rural areas14. Similarly, the vocal activity of diurnal birds at night depends on the phase of the moon, the environment’s primary source of natural light. In the case of diurnal species – the undulated tinamou (Crypturellus undulatus), the percentage of the moon illumination correlated with its phase is positively associated with nocturnal vocal activity15. However, the opposite pattern occurs in the white-throated sparrow (Zonotrichia albicollis) and the ovenbird (Seiurus aurocapilla), which sing less intensively at night when the moon is full16. Vocal activity during bright nights (illuminated by artificial and natural light sources) is strongest in naturally early-singing species, with large pupils and sensitive retinas that allow more light to pass through, stimulating the organism to sing6. Examples of such species include the common blackbird (Turdus merula)17, the great tit (Parus major)18, the european robin (Erithacus rubecola)19, and the american robin13. To conclude, light may have different effects on different species, while environmental conditions, anatomical structure of the eye, and phenotypic plasticity can modify its role20. However, the number of studies examining the entire community rather than individual species is limited.

Predation pressure is another significant factor that can regulate nocturnal singing by diurnal birds. One hypothesis suggests that (1) birds sing more intensely at night when there is strong pressure from predators during the day, and (2) birds sing less intensely at night when there is strong pressure from nocturnal predators21. Therefore, we expect the intensity of night singing in the same bird species to vary across different locations (with birds singing more intensely at night in some areas than in others) due to varying levels of predation risk. The type of environment may also play a role in the intensity of nocturnal singing. Kułaga and Budka3 showed that more species sang at night in open areas than in forests, where predation pressure may be higher due to owls and small mammalian predators. However, in this study, predation pressure was not directly measured. In addition, open areas are brighter than dense forests, which may stimulate birds to produce more songs. All these complex associations suggest that, to fully understand the function and evolution of nocturnal singing by diurnal birds, it is necessary to consider this behaviour in biologically and ecologically diverse species at different stages of the reproductive cycle, reflecting various determinations for singing in various habitats, each subject to unique ecological pressures.

This study aims to determine the prevalence and intensity of nocturnal singing by diurnal birds in temperate regions of Central Europe, along with factors that influence it. Specifically, it aims to answer more specific questions: Which diurnal bird species sing at night? What is the prevalence and intensity of night singing in particular species? Does the intensity of nocturnal singing change across the night and breeding season? Does the intensity of nocturnal singing depend on the type of environment and species-specific factors, including phylogeny?

Additionally, the influence of natural light and predation pressure on the nocturnal vocalisations of diurnal bird species is to be investigated. It is anticipated that the intensity of nocturnal singing by diurnal birds in temperate regions will exhibit species-specific, seasonal, and ecological variation. A comprehensive sampling approach and multidimensional analysis will allow us to identify the biological and environmental factors that predispose entire bird communities, rather than individual species, to engage in nocturnal singing behaviour.

Results

Prevalence and intensity of nocturnal singing

Our study recorded 128 species belonging to 16 orders and 44 families across all four locations. Three species—bearded reedling, barn owl, and little owl—were found singing only during the night, whereas 82 species were recorded only during the day. In total, 46 bird species were recorded at night (Table 1; Fig. 1). These species belonged to 9 orders and 20 families, classified into four categories of activity type. The first category, non-territorial species (NT), included the common buzzard, mallard, goose, eurasian wigeon, black-headed gull and grey heron. The second group comprised typical nocturnal species (TN) and included only five species of owls. The third category consisted of typical diurnal species (TD), with 21 species recorded at night. The fourth category contained territorial species that are both diurnal and nocturnal (TDN; 14 species recorded at night). Two categories were taken into account in all analyses: (1) typically diurnal species and (2) species that are both diurnal and nocturnal. These two categories included 35 species belonging to 6 orders and 15 families.

Table 1 List of bird species recorded vocalising at night.
Full size table
Fig. 1
figure 1

Phylogenetic tree of species detected in the study. Species marked in red were recorded during the day and at night, species marked in black were recorded only during the day, species marked in blue were recorded only at night. The graph was generated using the phyloT v2 generator.

Full size image

Species that sing at night differ in the number of songs they produce. The highest number of songs was recorded for the corncrake, while the lowest for the green sandpiper (Table 1).

The phylogenetic tree of species detected in the study is presented in Fig. 1. We observed that in two families every detected species sang at night (Locustellidae and Alaudidae). There are several families in which some detected species were vocally active at night while others were not (e.g., Acrocephalidae, Charadriidae and Turdidae). Additionally, families in which no species sang at night can be identified (i.e. Picidae, Regulidae and Corvidae).

The prevalence of nocturnal singing among orders is presented in Fig. 2. The order with the highest number of species singing at night was Passeriformes – 21 species. Two of them (thrush nightingale and bluethroat) sang at more recording points at night than during the day. The order of Charadriiformes included six species singing at night. One of them – eurasian woodcock – was present at a higher number of points at night than during the day. The order of Gruiformes consisted of 5 species that sang at night, two of which were present at more recording points at night than during the day. The order of Galliformes included two species; Cuculiformes – one species; and Pelecaniformes – one species singing at night.

Fig. 2
figure 2

The prevalence and intensity of singing among orders. The analysis included (1) typically diurnal and (2) both diurnal and nocturnal bird species. Nocturnal species were excluded. The prevalence (Y-axis) is ratio of the number of sampling points where birds were recorded at night to the number of points where these birds were recorded for a given sampling point during the day.

Full size image

Patterns of Temporal and Spatial night song activity

The largest number of diurnal species singing at night was recorded in the Upper Nurzec River Valley (20 species) and the Warta Landscape Park (20 species). In the Biebrza National Park, 17 species were recorded, and in the Białowieża Primeval Forest, 14 species of diurnal birds singing at night were identified.

Patterns of vocal activity during the night differed between species (Fig. 3). In many species (e.g., sedge warbler, whinchat, bluethroat, eurasian skylark, common nightingale), an increase in song production before sunrise was observed. In the case of common crane, the opposite pattern was followed. In some species, the pattern was difficult to define (e.g., corncrake). It was also found that in some species the intensity of nocturnal songs differed between populations (e.g., northern lapwing, common snipe, common crane), whereas in other species the pattern remained similar across populations (whinchat, Savi’s warbler, eurasian skylark).

Fig. 3
figure 3

Temporal pattern of nocturnal vocal activity across different species and populations (represented here as locations) of the same species.

Full size image

The temporal pattern of birds’ vocal activity during the night appears to be different for species belonging to different categories of activity type (Fig. 4a). Typically diurnal species exhibit two distinct peaks of singing intensity: in the middle of the night and at the end of the night. Similarly, species categorised as “both diurnal and nocturnal birds” sing more as the night progresses. However, in the case of typically nocturnal birds, the pattern seems to be opposite—the highest activity occurs in the first part of the night and gradually decreases over time. When the data from all locations were aggregated and the night was divided into three equal parts (beginning, middle, and end), the number of songs produced by species during the night (TD and TDN) was found to be higher at the end of the night than at the beginning (Fig. 4b; χ2 = 15.54, df = 2, p < 0.001). A similar pattern in nocturnal vocal activity was observed across the four locations (Supplement Fig. S1 online).

Fig. 4
figure 4

Patterns of temporal nocturnal song activity. (a) Pattern of vocal activity during the night according to activity type and locations. (b) Pattern of vocal activity between different parts of the night. Dots represent partial residuals, while whiskers indicate 95%-CI. (c) Pattern of vocal activity between different weeks of monitoring. Dots represent partial residuals, while whiskers indicate 95%-CI.

Full size image

The relative abundance of songs varies between locations (Supplement Fig. S2 online), while the diversity of species observed singing at night appears to remain constant over time. However, a higher diversity in the last week of recordings compared to the first week was observed (Fig. 4c; Z = 3.78, p < 0.001).

The intensity of nocturnal singing—level of the Moon illumination

The results show that the moon illumination is an effective stimulus for the probability of bird singing, with the strongest response observed in both diurnal and nocturnal species compared to typically diurnal ones (Fig. 5a; interaction: χ2 = 34.14, df = 1, p < 0.001). However, the abundance of songs increases only for both diurnal and nocturnal species, not for the typically diurnal community (Fig. 5b; interaction: χ2 = 54.23, df = 1, p < 0.001). Moreover, due to the increased probability of singing at the species level, increased birds’ diversity with increasing moon illumination was observed (χ2 = 26.26, df = 1, p < 0.001).

Fig. 5
figure 5

The influence of natural light on the vocal activity of birds (typically diurnal and both diurnal and nocturnal species). (a) The influence of scaled moon illumination on the probability of bird vocal activity and (b) the abundance of songs. Dots represent partial residuals, solid lines are estimates, while shaded polygons indicate 95%-CI of the estimates.

Full size image

Intensity of nocturnal singing—level of predation pressure

Both in forests and open areas, when nocturnal predators are absent, the number of species singing at night was noticed as higher (Fig. 6b), but this result is not statistically significant (χ2 = 1.95, df = 1, p = 0.16). However, the number of vocally active species during the night was observed as higher when diurnal predators are present at the recording points (Fig. 6a, χ2 = 3.99, df = 1, p = 0.046). This model also accounted for the habitat type (open vs. forest), which was the strongest predictor (Fig. 6b, χ2 = 27.63, df = 1, p < 0.001). Significantly more songs by diurnal birds were recorded during the night at points located in open areas (meadows, arable fields, wasteland, forest clearings) than in forested ones (Fig. 6b).

Fig. 6
figure 6

The influence of predation pressure on the vocal activity of birds (a) Difference in the number of species singing at night between habitat types and presence or absence of nocturnal predators (b) Number of species singing at night in the case of presence or absence of day predators. Dots represent partial residuals, while whiskers show 95%-CI.

Full size image

Intensity of nocturnal singing—species-specific factors

No significant relationship between the frequency of occurrence at recording points and (1) a species’ migration type (χ2 = 0.74, df = 2, p = 0.69), (2) minimum bird length (χ2 = 0.01, df = 1, p = 0.92) was found. However, the frequency of occurrence at recording points was higher in species whose preferred habitat (reported in the literature as typical) is mixed (edge of a forest and open area) or open area (χ2 = 9.62, df = 2, p = 0.008) compared to a forest.

Discussion

Our study has shown that nocturnal singing by diurnal bird species is quite common in temperate regions. Of the 128 recorded temperate bird species, 46 sang at night in Central Europe (36%). 76% of these were classified as diurnal species (typically diurnal and both diurnal and nocturnal species). La4 reported that nocturnal vocalisations were confirmed in at least 30% of North American birds. Over 70% of nocturnally vocalising birds are considered diurnal. The percentage of nocturnal singers in temperate regions on different continents is similar, which may indicate that there exists a single pattern of nocturnal singing prevalence in these regions.

The presence of nocturnal singing by diurnal birds varies between taxa. We found families where every detected species sang at night (Locustellidae and Alaudidae) but also families where no species sang at night (i.e., Picidae, Regulidae and Corvidae). The most common pattern was families in which some species produced vocalisations at night while others did not (i.e., Acrocephalidae, Charadriidae, and Turdidae). This disproportion in the occurrence of nocturnal singing suggests that this behaviour has evolved independently multiple times. Likewise, La4 showed that the phylogenetic signal in nocturnal vocalisations is insignificant, indicating that nocturnal vocalisations are not restricted by phylogeny. The order with the highest number of species singing at night was Passeriformes. The thrush nightingale and bluethroat sang at more recording points at night than during the day. These two species belong to the genus Luscinia, which is known for its nocturnal activity22. A proposed explanation of why these birds sing intensely at night is that they sing hidden inside the bush, so they feel less at risk of predator attack. Additionally, this energetically costly behaviour6 might be used by females to assess the quality of males23.

Even within a single family, the intensity of nocturnal vocalisation varies among species. These species can be divided into three categories: those singing regularly (i.e., corncrake, thrush nightingale, eurasian coot), those singing occasionally (i.e. common grasshopper warbler, wood lark, whinchat), and those singing rarely (i.e. spotted flycatcher, great bittern, redwing). A similar variation in nocturnal singing intensity among different species was reported in North America4.

The nocturnal vocal activity of a single species can vary across different locations. This may indicate that night singing is not only species-specific but also population-specific. In the case of woodlark in the Upper Nurzec River, we observed a higher intensity of nocturnal singing before sunrise. When recording this species in the Białowieża Primeval Forest, a broad peak of the highest nocturnal vocalisations occurred in the middle of the night, but in the Warta Landscape Park, the intensity of singing at night was relatively consistent, with a small peak before sunrise (Fig. 3). This species is probably affected by many location-dependent factors reflected in its vocal activity.

On the other hand, in the case of the whinchat, the pattern of nocturnal singing intensity appears to be similar across locations – vocalisations begin before sunrise when the amount of light is slightly higher. Another interesting example is when birds are present during the day at all locations but sing at night only in one. The cuckoo represents such a pattern, as this species sang at night only in the Biebrza National Park. We can hypothesise that in this location a higher amount of natural light is present at night because the large open landscape is covered mainly by marshes. A consistent pattern of nocturnal vocalisations has not been found between locations for all the detected species. This may be due to species composition differences, light availability variation across different areas or reducing acoustic competition occurring during the day24.

Considering vocal intensity at night in species with different activity-type categories, a distinct pattern has been observed. Typically, diurnal species have two peaks of singing intensity: in the middle and at the end of the night. Both diurnal and nocturnal species sing more songs as the night progresses. After summing up the diurnal species (typically diurnal and both diurnal and nocturnal) across all locations (Fig. 4b), we found that birds sang more songs at the end of the night than at the beginning of it. It can be generalised that birds in our study sing less at the start of the night when likely resting after the day It is the time after dusk chorus - the second diurnal peak of intensive singing, when many birds consume a lot of energy while vocalising25, which is why they need to regenerate later. They then wake up and vocalise after a brief period of regeneration. Since sleeping is crucial for birds’ regeneration, its deprivation has vast consequences for their physiology (affecting body mass, temperature, hormone levels) and behaviour (impairing attentiveness, motivation, reaction times, coordination, emotional stability and increasing stress behaviour26. Thus, it is inevitable that birds spend some part of the night sleeping, while our study shows that this occurs primarily at the beginning of the night (Fig. 4b). Since our study focused exclusively on nights, we cannot exclude the possibility that birds also begin resting before sunset.

The relative number of observed species that sing at night was higher on the last day of recordings (middle of May) when accounting for daytime diversity. This may be related to individuals establishing territories and intensifying their search for mates. We have not identified a peak in nocturnal singing intensity at the beginning of the breeding season, which is typical during the day when males seeking partners and defining territories27.

Significantly more songs of diurnal birds were recorded during the night at points located in open areas (meadows, arable fields, wasteland, forest clearings) than in forest areas. It was also found that the frequency of occurrence at recording points at night is higher for species whose typical habitat is mixed (edge of the forest and open area) or an open area. This may be related to the greater amount of incoming light reflected by the moon at night in open areas. In the forest, the extensive and leafy tree crowns limit the amount of light. Additionally, in forest areas, there is higher pressure from nocturnal predators such as owls. Through singing, birds can be more easily located, therefore, to reduce risky behaviours diurnal birds may vocalise less often or not at all during the night21.

Changes in moonlight intensity are complex and depend on multiple factors. The moon phase, although convenient and generally well understood, is a poor proxy for moonlight intensity because it does not take into account many factors influencing illumination28. Our study used scaled moon illumination as the time of the moon in the sky during the night scaled to night length and then multiplied by the visible percentage of the illuminated moon surface. It is a stricter approach, because it considers more factors influencing the level of illumination than only the moon phase. We found that when scaled moon illumination is higher, both diurnal and nocturnal birds sing more songs, but this dependency is not present in typically diurnal birds. Diurnal species are adapted for bright environments and, therefore, may be susceptible to changes in the amount of light at night. Surprisingly, only species in the category: “both diurnal and nocturnal birds” react to higher illumination at night (Fig. 5). These species are likely to be more sensitive to detecting small changes in the light conditions at night. Many studies confirm the stimulating role of the moon in singing. Dickerson et al. (2020) found that the nocturnal song of the willie wagtail (Rhipidura leucophrys), which is typically a diurnal bird, increased with lunar illumination. Males of the vermilion flycatcher (Pyrocephalus obscurus) increased the song rate and sang songs with more elements during full moon nights than new moon nights10. The white-browed sparrow weaver (Plocepasser Mahali) also increased its total song output during the full moon compared with the new moon30.

High predation pressure at night can limit nocturnal singing in diurnal birds31.The number of species that are vocally active at night was found to be statistically significantly higher when diurnal predators were present at the recording points. To minimise the threat, birds can sing more at night when daytime predators are inactive. This study also found that when nocturnal predators are absent, the number of species singing at night is higher, but this result is not statistically significant when accounting for habitat type and the presence of diurnal predators. A study of male veery (Catharus fuscecens) birds showed that birds sang less intensely after dusk when owl vocalisations were played back to them21.

In conclusion, the nocturnal singing of diurnal birds is a relatively common phenomenon in Central Europe. The intensity of this behaviour varies between orders, families, and even between individual species. Birds generally sing less at the beginning of the night and more at the end. More diurnal bird songs were recorded in open areas than in forests, probably due to greater access to light in non-forested areas. Higher moon illumination is reflected in a higher number of songs produced by both diurnal and nocturnal birds, which confirms the significant role of light in stimulating night singing. Predation pressure may affect song timing in diurnal birds. This study is the first to provide a detailed assessment of the nocturnal activity of diurnal Central European bird species in a wide range of environments, indicating the factors that influence the prevalence and intensity of this behaviour.

Methods

Study area

The study was conducted in four locations: three in Eastern Poland (Upper Nurzec River Valley, Biebrza National Park, and Białowieża Primeval Forest) in 2020 and one in Central Poland (Warta Landscape Park) in 2021. The Upper Nurzec River Valley has a typical agricultural landscape consisting of two main types of habitats. The first is a large peat meadow stretching along the Nurzec River. The second habitat type is extensively cultivated fields located on the valley’s edge. Agriculture is extensive; therefore, over 140 species of breeding birds have been observed there, including typical farmland species. Another study location is Biebrza National Park, the largest national ark in Poland. It is covered mainly by marshes with a unique variety of plant communities and rare wetland birds. The rest of the area consists of fields, meadows, and forests, home to approximately 180 breeding bird species. The last study location in Eastern Poland is Białowieża Primeval Forest, which contains the best-preserved natural forest on the East European Plain. Many different forest types and some open glades are habitats for about 160 breeding bird species.

The single study location in Central Poland - the Warta Landscape Park- is covered by forests, valuable meadows, grasslands, and mostly wetlands – the latter closely related to Warta’s periodic flooding. This area is a breeding site for about 150 bird species.

The habitat diversity of the study areas has enabled the detection of a wide range of typical diurnal bird species singing at night. In this way, reliable data on the prevalence of such behaviour in birds have been obtained. Additionally, covering several locations in the study has allowed us to isolate areas with different predation pressures. The close proximity of the first three locations to each other (Eastern Poland) has minimised the effect of phenological differences. However, data on the night singing of diurnal birds from Central Poland have been used to determine possible differences in the intensity of this behaviour in more distant areas.

Selection of recording points

76 recording points, based on satellite images, were selected − 20 points for each location; only the area of Biebrza National Park had 16 points due to harsh field conditions. In each location, half of the points were located in open habitats, such as marshlands, meadows, or arable fields, while the other points were in forests (with the distance from the recording point to the forest edge exceeding 100 m). The distance between neighbouring recording points was greater than 1 km, minimising the probability of recording the same individuals at different points.

Fieldwork

At each of the 76 recording points, an automatic sound recorder (AudioMoth 1.1.0; one channel, 48 kHz/16 bit sampling rate, medium gain) recorded for one day (24 h) every 7 days from late March to mid-May on different days at each location (in total 7 × 24 h of recordings per point). Depending on the habitat conditions, the recorders were placed on trees or shrubs at 3–4 m above ground level using plastic bands. The audio recorders were protected against moisture using waterproof baggies and hygroscopic bags. Monitoring of birds of prey and predatory mammals occurred three to four times per season, approximately every 1.5 months (from March to September). During the inspection, all tracks of predators observed and vocalisations heard were noticed along designated transects (about 680 m per recording point; Fig. 7). To identify a particular species that left tracks or traces, photos were taken to determine the predator species. In the next step, the predators were divided into those active during the day and those active at night. The predators were also categorised as posing a threat to singing birds or not. Only the predators that threatened vocalising birds were used in the statistical analysis. Based on the literature and the Birds of the World database32we selected only predator species that include adult birds in their diet and whose hunting method indicates that they feed on adult birds. The predation pressure of owls was determined based on nocturnal soundscape recordings.

Fig. 7
figure 7

Shape of the predator monitoring transect. The monitored trail ran from the centre of the transect, heading north, then east, and then west through the centre, then south, and ending at the centre.

Full size image

Analyses of acoustic and predator monitoring

Firstly, we defined night as the period from nautical dusk to nautical dawn. Nautical dusk occurs when the Sun’s geometric centre falls 12 degrees below the horizon (in our study area and study period, this happened between 8:18 and 10:15 PM). Nautical dawn is when the Sun’s geometric centre rises to 12 degrees below the horizon, and the sky is still completely dark (in our study area and study period, between 2:33 and 4:40 AM). The most restrictive definition of night (astronomical night) was not applicable since it did not occur in May. Thus, at each recording point, between 4 h 20 min to 8 h 39 min of nocturnal soundscape recordings were analysed. This significant difference in night length is due to its progressive shortening as the season advances. At each recording point, we used light intensity recorder (HOBO UA-002-64) to ensure that we did not include dusk and dawn in our definition of night. According to the recorders, the light intensity was zero throughout the entire nautical night. To estimate nocturnal singing activity in diurnal birds, the analyses did not cover the entire set of soundscape data, instead, one-minute sound samples from the nocturnal recordings at 20-minute intervals were selected.

Additionally, daytime samples were analysed to assess the number of bird species singing at each recording point. One-minute sound samples at 10-minute intervals from one hour before sunrise to two hours after sunrise were selected. In total, approximately 10,000 one-minute sound samples were analysed.

Bioacoustic data were analysed using Raven Pro 1.6 software. To generate spectrograms, the following parameters were applied: 60% brightness, 60% contrast, a 1024 points window size, and a Hamming window type. During manual spectrogram scanning and listening to the recordings, all the recorded songs and calls were classified by species. Our analyses included the songs of songbirds and the calls of non-songbirds that serve a similar function to those of songbird songs — territory defence and mate attraction. Thus, the songs of blackbirds and the calls of corncrakes were considered nighttime vocalisations, whereas the contact calls of great tits were not.

Environmental and species-specific factors

To understand the effect of natural light on bird singing activity, the illumination of the moon at night was measured. This was done by calculating the relative illumination of the moon on a specific night, which was represented by the time the moon was above the horizon during the night, scaled to the night length, and then multiplied by the percentage of the illuminated moon. This calculation provided us with the relative intensity of light that illuminated the night.

To understand which species-specific factors could affect the prevalence of singing during the night, a database of all the detected species assigned to a particular genus, family, order, migration type, minimum and maximum body length, minimum and maximum weight, and preferred habitat type was created, with all the data drawn from the online database Birds of the World from Cornell Lab of Ornithology32.

All the detected bird species were classified into four different categories based on activity type (Table 1): (1) non-territorial species (NT), (2) territorial and typically nocturnal species (TN), (3) territorial and typically diurnal species (TD), (4) territorial species active both diurnally and nocturnally (TDN). The first category of non-territorial species includes birds that did not breed in the study locations, and whose vocalisations were unrelated to mating. The second group typically comprises nocturnal species and consists only of owls. The third category consists of species that mainly sing during the day. The fourth category includes territorial species that commonly sing both during the day and at night. Based on a literature search, this classification differentiates four groups with different circadian rhythms and territory usage.

Statistical analysis

Temporal patterns in singing intensity at night were based on a scaled time, which accounted for the absolute time each species sang relative to the length of the night. This approach allowed us to compare singing intensity within the same night and across different nights throughout the season and across various localities. Additionally, temporal patterns at the seasonal level were analysed, both within and between localities, using diversity data and singing intensity. The effects of natural light, predation pressure, and species-specific factors on the prevalence and intensity of bird songs were assessed using different linear models, as described below. Data in all models included species classified as TD and TDN (for details, see the section on environmental and species-specific factors in the methods).

To test in which part of the night (beginning, middle, and end) birds sing the most, we built a negative binomial generalised linear mixed model (GLMM) with the number of songs as the response variable, the part of the night (a three-level factor) as the fixed effect, and species ID as the random intercept effect. In this model, the random effect of a particular night was not accounted for since it was over-parametrized and did not converge.

Temporal changes in bird activity, measured as diversity, probability of singing, and singing intensity, were analysed with moon illumination data, and the resolution of these data (sampling point/week) was the same. To achieve this, we built a set of three GLMMs that tested responses in the diversity of birds vocally active during the night and the probability of singing and its intensity. The first model was a Poisson GLMM, which tested the change in the diversity of birds that sang during the night in response to moon illumination intensity and week ID. Moreover, this model also included Point ID as a random intercept effect and the diversity of species detected during the day as an offset (log-scaled). The second GLMM (binomial) included the presence/absence of birds during the night as the response variable, the intensity of moon illumination at night, species activity type (two categories: TD and TDN) and week ID as fixed effects, with point ID as a random intercept effect. Since we aimed to identify differences in response to the moonlight between the two species activity types, we also included an interaction between moon illumination intensity and species activity type. The last model was a negative binomial GLMM with the same set of predictors as the second GLMM but with count data as a response variable.

The effect of predation pressure on nighttime singing was analysed based on data at the resolution of sampling points, where all information about the diversity of birds at night and during the day, as well as the presence of predators, was summed up over the observation period. This effect was assumed not to vary weekly but to be more constant over time for a specific locality. Moreover, in our approach, the presence of particular predators was treated as a binary variable because passive recordings of owls could lead to pseudoreplication. Additionally, burrows, prey remains, and excrements could originate from the same individuals. To test the effect of predators on the diversity of night-singing birds, we used a negative binomial GLM with the number of observed species singing at night at a sampling point as the response variable, type of habitat (open vs. forest), presence of nocturnal predators, and presence of diurnal predators as fixed effects, and the number of observed species during the day as an offset (log-scaled). Also, a model that included the interaction between the habitat type and the presence of predators during the day and night was tested; however, this model did not show a better fit.

Finally, to understand which species-specific factors could affect the prevalence of singing at night, we built a negative binomial GLMM with the count of sampling points where a species sang at night as the response variable, typical habitat type, bird size, and migration type as fixed effects. Moreover, this model also included location ID as a random intercept effect and the count of points where the species was detected during the day as an offset (log-scaled). It is important to note that in this analysis, we used the typical habitat type from the literature (Birds of the World database32. In contrast, in other analyses, the habitat type specified in the field for every recording point was used.

All analyses were performed in R software (v. 4.3.233). Models were estimated using the Restricted Maximum Likelihood method with the glmmTMB package (1.1.834). Negative binomial GLMMs refer to the models where variance is estimated as a quadratic function of the mean35. The significance of variables was tested using Type-II ANOVA when our focus was on the main effects, or Type-III ANOVA when we focused on the interaction terms. Model quality was evaluated using the performance (v. 0.12.236 and DHARMa v. 0.4.737) packages. We assessed over/under dispersion, zero-inflation, collinearity of predictors, linearity of relationships depending on the link-function, and homogeneity of variance. Models were visualised using visreg (v. 2.7.038) and ggplot (v. 3.5.14) packages.