Lunar synchronization of daily activity patterns in a crepuscular avian insectivore

Abstract Biological rhythms of nearly all animals on earth are synchronized with natural light and are aligned to day‐and‐night transitions. Here, we test the hypothesis that the lunar cycle affects the nocturnal flight activity of European Nightjars (Caprimulgus europaeus). We describe daily activity patterns of individuals from three different countries across a wide geographic area, during two discrete periods in the annual cycle. Although the sample size for two of our study sites is small, the results are clear in that on average individual flight activity was strongly correlated with both local variation in day length and with the lunar cycle. We highlight the species’ sensitivity to changes in ambient light and its flexibility to respond to such changes in different parts of the world.

To assess the impact of light pollution on animal behavior, it is essential to first understand the behavioral responses of animals to periodic changes in natural light conditions (Parker & Smith, 1990;Stephens & Krebs, 1985).
In this study, we present an analysis of the lunar-associated behavior of European Nightjars (Caprimulgus europaeus; hereafter referred to as nightjar; Figure 1) by testing the hypotheses that the lunar cycle affects their nocturnal flight activity. Within the order Caprimulgiformes (Nightjars and their allies), many species are aerial insectivores adapted to a crepuscular/nocturnal lifestyle (Mayr, 2010;White, Barrowclough, Groth, & Braun, 2016).
All members of the Caprimulgidae are visual hunters and predominantly detect flying prey against the sky (Cramp et al., 1985, Evens et al., 2018. Optimal foraging conditions for this hunting method are restricted to periods of twilight, unless nocturnal light allows prolonged foraging activities (Jetz, Steffen, & Linsenmair, 2003). Earlier observation-based studies already suggested that several species of Caprimulgiformes synchronize their activities with the lunar cycle, showing low activity during periods of nocturnal darkness (Brigham, Gutsell, Wiacek, & Geiser, 1999) and high singing, reproductive and foraging activity around full moon (Brigham & Barclay, 1992;Holyoak, 2001;Jackson, 1985;Mills, 1986;Perrins & Crick, 1996).
Despite these indications of population-level responses to the lunar cycle, variation in activity patterns at the individual level in relation to nocturnal light conditions have not yet been quantified, except in the context of thermoregulation (e.g., Smit, Boyles, Brigham, & McKechnie, 2011) and migration (Norevik, Akesson, Andersson, Backman, & Hedenstrom, 2019). Individual nightjars became increasingly heterothermic in response to lower foraging opportunities associated with new moon periods, irrespective of relatively constant food resources (Smit et al., 2011). Similarly, better foraging opportunities during full moon periods most likely drive light-dependent fuelling opportunities and influence the departure timing at stopover sites of migrating nightjars (Norevik et al., 2019). From the above, we expected that moonlight allows nightjars to be active longer and outside the period around dusk and dawn. As such, our study aims at testing the hypothesis that the probability that an individual is active depends on local variation in light levels. We investigated this by analyzing daily activity patterns of individuals from three distinct parts of the species' breeding range (Mongolia, Belgium and Sweden; Figure 2a), during two discrete periods of the annual cycle (the breeding and the nonbreeding season) in relation to the lunar cycle, while controlling for variation in local day length. www.light pollu tionm ap.info). We captured nightjars in breeding areas using ultra-fine mist nets (Ecotone, 12 × 3m) and tape lures (Evens, Beenaerts, Witters, & Artois, 2017). We marked each individual with a unique alphanumeric ring and fitted a data logger dorsally between the wings (Evens, Conway, et al., 2017;Norevik, Akesson, & Hedenström, 2017). In total, we tagged 90 adult individuals, 20 in Mongolia, 10 in Belgium, and 60 in Sweden with a 1.2 g SOI-GDL3pam data logger (Mongolia and Belgium (Liechti et al., 2018) or a 2.1g Multidata logger (MDL; Sweden (Norevik et al., 2019). Each logger contained sensors that recorded air pressure, ambient light intensity, and acceleration in 5-min intervals. The SOI-GDL3pam data loggers contained additional sensors to record air temperature in 5-min intervals and magnetic field in 4-hr intervals. Activity is measured as the sum of the absolute differences in acceleration on the z-axis (SOI-GDL3pam: a summary variable stored for each 5-min interval (Liechti et al., 2013); MDL: a summary variable stored for each one-hour interval (Norevik et al., 2019). In F I G U R E 1 The European Nightjar (Caprimulgus europaeus) is a crepuscular insectivore that mainly breeds in semi-natural, dry landscapes total, we recovered eleven loggers (two in Mongolia, one in Belgium and eight in Sweden; all from males; Appendix S1: Supplementary Materials T1). These low recovery rates did not allow a formal comparison of flight activity between study sites.

| Activity data
Activity data from SOI-GDL3pam data loggers were transformed to hourly estimates to fit the data of the MDL loggers. We subdivided daily activity data from the breeding and the nonbreeding season into three groups: daytime (from sunrise until sunset), twilight (from sunset until evening nautical twilight; from morning nautical twilight until sunrise), and night (from evening nautical twilight until morning nautical twilight). We categorized activity data into two classes: active (e.g., flight and foraging) and inactive (e.g., resting and roosting). We distinguished between these categories based on an activity threshold computed as the 97.5th percentile of daytime activity using a Linear Quantile Mixed Model (Geraci, 2014) with individual identity as random intercept, and period (winter versus. summer), time of the day and date as covariates. This threshold was chosen because nightjars remain inactive during daytime, with resting and preening as their main activities. During twilight they spent much of their time flying to forage and to commute between local sites (Evens et al., 2018;Evens, Beenaerts, et al., 2017;Mills, 1986;Wynne-Edwards, 1930), and at night additional foraging activities can occasionally take place (Brigham et al., 1999;Evens et al., 2018).
Data on the timing of day, night, and twilight (i.e., sunset and sunrise) and the lunar cycle (i.e., altitude of the moon above the horizon and fraction of illuminated moon) were extracted for the known breeding sites and estimated nonbreeding sites and for each interval using the R-package "suncalc." We did not take into account variation in local weather conditions which may lead to additional noise, and hence may weaken our results rather than create systematic biases (Penteriani et al., 2013).
Of all hourly intervals, 7,434 (32%) were categorized as active, while the remaining 15,358 intervals were categorized as inactive.
We modeled nocturnal and twilight activity using Generalized Linear Mixed Models (GLMMs) with maximum likelihood using the R package glmmTMB version 0.2.3 (Brooks et al., 2017;Geraci, 2014;Team, 2019). To asses patterns of nocturnal activity, we constructed two models: 1) a model for the conditional mean containing the fraction of illuminated, visible moon (i.e., fraction of illuminated moon when above the horizon, continuous variable: percentages) and altitude of the moon above the horizon (continuous variable: radians) as the main predictors (Table 1) and 2) a zero-inflated model which allows modeling the probability of excess zeros in the conditional part of the model. To assess patterns of crepuscular activity, we constructed a model with twilight period (categorical variable: dusk or dawn) as the main predictor (Table 1).
The conditional model (nocturnal activity model) and the crepuscular activity model use a negative binomial distribution with log-link function, whose variance was set to increase linearly with its mean.  Although we cannot exclude that some individuals were affected by the devices, direct observations in the field and assessment of the recaptured individuals showed no signs of negative effects.
The recovery rate (11%) is lower than expected (Evens, Conway, et al., 2017;Norevik et al., 2019) and is probably caused by i) the late deployment of loggers on nonresident individuals with no fidelity to the study site in Belgium (late August 2018) and ii) a low recovery success due to bad weather conditions during a two-week trapping session in Mongolia (July 2019).

| RE SULTS
The nightjars' sensitivity to changing light conditions can be seen in the actograms (Figures 2 and 3 During twilight, activity was higher at dusk than at dawn, and higher during the breeding season than during the nonbreeding season (Table 1; Figure 3g). Twilight activity was independent of the moon phase (Table 1).

| D ISCUSS I ON
Although the sample size for two of our study sites is small, the re-    Table 1. (g) Differences in the probability of activity between dusk and dawn during the breeding season (green) and during the nonbreeding season (orange). Shown are model estimates and their 95% confidence intervals based on the model in Table 1 Activity Hour Hour  Nocturnal activity data display a noticeable diagonal band ( Figure 2) indicating a progressive shift in the nightjars' daily activity throughout subsequent moon phases (see also Norevik et al., 2019).
This general pattern is consistent between all individuals in our study, even though they resided in different parts of the world. Flight activity of the tracked individuals corresponds to daily changes of the moon's trajectory (i.e., the moon's altitude above the horizon) and the fraction of illuminated moon (Figure 3a-f, Appendix S1: Supplementary Materials 2). Around new moon, nightjars were largely inactive during the night (Figure 3a), whereas around full moon they seemed to fully exploit the increased ambient light by being active all night ( Figure 3c). The close relationship between nocturnal activity and night light is also suggested by relatively high before-midnight activity during a first-quarter moon and high after-midnight activity during a last quarter moon phase (Figure 3b and d).
The flight activity of nightjars is presumably organized by en- foraging conditions during moonlit nights (Phalan et al., 2007, Mackley et al., 2011, Pinet et al., 2011, Penteriani et al., 2014, Rubolini et al., 2014, Dias et al., 2016, Roeleke et al., 2018; but see Cruz et al., 2013). Our study shows that the nocturnal flight activity of European nightjars correlates with both the altitude of the moon and the fraction of illuminated moon. In line with previous findings (Mills, 1986), this suggests that low ambient light levels, that is, less than the light intensity of a quarter moon (0.01-0.03 lux (Kyba, Mohar, & Posch, 2017)), limit movement and/or foraging opportunities. Flying in a local area (Brigham & Barclay, 1992, Zwart et al., 2014 or commuting between breeding and foraging sites might be safer during lighter conditions, because of a reduced risk of colliding with dark objects (Cresswelll & Alexander, 1992;Evens et al., 2018). Moonlight may also increase prey visibility and hence foraging success, which is probably why nightjars invest most energy in territorial display and reproduction during periods with the greatest moonlight levels (Jackson, 1985;Mills, 1986;Perrins & Crick, 1996). Our study shows that nightjars are usually inactive during moonless parts of the night. Similarly, low light levels during moonless nights affected thermoregulation in several species of nightjars, whereby the birds entered torpor following reduced foraging opportunities (Brigham et al., 1999;Doucette, Brigham, Pavey, & Geiser, 2012;Smit et al., 2011). However, moonlit nights do not necessarily imply high foraging activity. For example, Afrotropical nightjars reduced their nocturnal activity F I G U R E 4 Nightjar activity in relation to light, moon altitude, and time. (a) Individual activity in relation to the fraction of visible moon. Activity data were collected at 5-min intervals. Shown is the probability that a nightjar was active at night in relation to the fraction of illuminated moon. Each color corresponds to one individual (same color as in Figure 2: black = Belgian, blue and red = Mongolian). (b) The probability of nightjar activity at night in relation to the altitude of the moon above the horizon (expressed in radians). Shown is the estimate and 95% confidence intervals based on the model in Table 1 during full moon, presumably in response to high predation risk (Brigham et al., 1999;Jetz et al., 2003).
Foraging success may be higher during moonlit nights if prey activity is higher than during dark nights (Jetz et al., 2003). Nightjars typically perch at the edge of open fields while foraging (Evens, Beenaerts, et al., 2017). The likelihood to detect and hawk flying insects silhouetted against the sky (Camacho, Palacios, Sáez, Sánchez, & Potti, 2014;Evens et al., 2018;Jackson, 2003) should be higher when the sky is illuminated (Ashdown & McKechnie, 2008;Jetz et al., 2003;Mills, 1986). Alternatively, prey might be harder to catch in moonlight, either because insects can better detect the predator (Penteriani, Kuparinen, del Delgado, & M., Lourenço, R. and Campioni, L., 2011) and make evasive manoeuvres or because insects fly higher and therefore nightjars would have to work harder to achieve the same net energy intake. Although the emergence of insects peaks around sunset and sunrise (Malmqvist et al., 2018), it has also been suggested that nocturnal insect activity is associated with near full moon (Jetz et al., 2003;Nowinszky, Petrányi, & Puskás, 2010). One study suggested that the nightly flight activity of Lepidopterans-the nightjars' main food source-decreases during full moon nights (Raimondo, Strazanac, & Butler, 2004), whereas another study showed that the activity of species associated with open habitats increased during moonlit nights (Nowinszky, Kiss, Szentkirályi, Puskás, & Ladányi, 2012).
If nightjars are sensitive to relatively subtle changes in ambient light conditions, as our study suggests, we predict that artificial night lighting, especially skyglow during overcast nights (Jechow et al., 2017), will influence their behavior. Artificial night light can be perceived far from its source, even in uninhabited areas (Falchi et al., 2016), and is known to alter the behavior of many  , Da Silva et al., 2015, and how this in turn affects population dynamics. www.vr.se/) and Lund University (https://lundu niver sity.lu.se/).

CO N FLI C T O F I NTE R E S T S
We declare we have no competing interests.