The diel activity pattern of mountain hare (Lepus timidus) on managed heather moorland in Scotland

Abstract The research presented in this paper provides an insight into the behavioral ecology of mountain hares on heather moorland in the Lammermuir Hills of southeast Scotland. We examine the seasonal and diel activity patterns using camera traps over a period of 12 months. The rate of camera detections was calculated for the different divisions of the 24‐hr cycle (daylight, dusk, night, and dawn). During autumn and winter (October–February), the activity pattern was crepuscular with greater activity at dusk than at dawn. Daylight activity was relatively low, and there was a regular pattern of small peaks of activity during the night. In spring and summer (March–September), peaks of crepuscular activity remained evident but daylight activity was much more prevalent than during autumn and winter, and night activity was lower. We discuss the problematic definition of twilight and present an explanation for seasonal changes in the pattern of diel activity that is linked to the reproductive cycle of the mountain hare.

and is molted in spring (Flux, 2009;Hewson, 2009). In winter, the hare adapts its diet toward vegetation that may still be accessible above snow cover or can be reached by digging (Hewson, 1962;Hulbert et al., 2001).
Diel (24-hr cycle) activity studies are of importance in understanding the rhythms of mammalian life and the factors that influence and control these rhythms. Many studies have shown that mountain hares are predominantly crepuscular/nocturnal animals (reviewed in Angerbjörn & Flux, 1995) but the details of their circadian rhythms and seasonal variations are not well defined. Camera traps are a useful tool in recording profiles of animal activity (reviewed in Rowcliffe et al., 2014), particularly for animals such as mountain hares that are predominantly nocturnal and that live in remote habitats. Ikeda et al. (2016) and Ogurtsov et al. (2018) used camera trapping to study diel activity and seasonal changes in their studies of mountain hares in the boreal forests of Japan and Russia, respectively.
Counting mountain hare is much easier when they are active.
Using part of the same dataset presented here, one of us has explored how activity studies can inform the best time of year and the best time in the 24-hr cycle to count (Pettigrew, 2020): this at a time when there is controversy surrounding uncontrolled culling (Brooker et al., 2018).
This study examined the activity of mountain hares in the Lammermuir Hills of southeast Scotland. We used camera trapping to explore in detail the patterns of circadian rhythm and its seasonal variations. The data presented inform our knowledge of the natural history of this native mammal species.

| Study area
The study was undertaken within a 1 km square area of heather moorland managed for driven grouse shooting on the eastern slopes of Meikle Says Law, Lammermuir Hills, Scotland (SW corner of 1 km square study area: Latitude 55.813758, Longitude −2.607933).
The general topography of the study area, the pattern of muirburn (Yallop et al., 2006), and the location of camera traps are shown in Figure 1b. During mountain hare surveys (Pettigrew, 2020), up to 20 individual hares have been routinely counted by the Lammermuir Hare Group in the camera study area.

| Camera traps
Five camera traps (AUCEE 12 MP H9 Hunting Cameras) were set at 0.7 m above ground level on posts with a 5-10 degree downward tilt at the locations shown in Figure 1b between October 2018 and September 2019 (8,760h). We selected nonrandom locations to maximize total detection rates, by avoiding placing cameras in deep heather where they would be unlikely to record hare movement. Three of the locations were regenerating heather areas with associated hare runs and two were dry gullies with a mixture of heather, grasses, and rushes. A pilot study found 184 hare detections for 5 nonrandom cameras in a single month but only 32 hare detections for 5 random cameras (Pettigrew & Di Vita, unpublished data). Maximizing detections contributes to the smoothness of activity profiles and the confidence that can be placed in their key features (Lashley et al., 2018). The cameras were separated by at least 50 m (exceeding the detection limit of the infrared cameras (Brown & Gehrt, 2009) and they were orientated pointing away from nearest neighbors in order to reduce the likelihood of double counting.
The cameras were set on Greenwich Mean Time throughout the year and took 3 consecutive still images after a motion trigger, followed by a delay of 10 s. Images were 8 MP. Infrared detection distance was set to 'far', and motion sensitivity was set to 'medium'.
Camera trap images were downloaded every 2 weeks and camera times were reset if they had incurred time errors (which were less than 6 min). For each image that contained a mountain hare, we evaluated whether or not it represented an independent event. It is common for camera trap studies to apply an arbitrary threshold duration between photos to assign independence of events. Our general rule was that camera triggers separated by less than 2 min were not counted as independent events and those separated by more than 2 min were counted as independent events. However, we have The image data and applications of exceptions to the 2-min general rule are available in the supporting files S1a and S1b, respectively, linked to from the Data Accessibility Statement. [Correction added on 24 May 2021 after first publication: the location of the links to the supporting files has been clarified in the preceding sentence.] For analysis of the independent events (subsequently called 'detections'), we define a 2-hr postsunset (dusk) period and a 2-hr presunrise period (dawn). We defined daylight as between sunrise and sunset, and night as between the end of the dusk period and the beginning of the dawn period. The times of sunset and sunrise and the hours of daylight and night were calculated for each day using the data tables at https://www.theti meand place.info/uk/ for the nearby village of Gifford ( Figure 1a). The time stamps for detections were tabulated in Excel, categorized as morning (before 12:00) and afternoon (12:00 and after) and then re-calculated as decimal time relative to either sunset (afternoon group) or sunrise (morning group). Vazquez et al. (2019) found that this 'average anchoring method' was the best means for reliable extraction of crepuscular activity at higher latitudes.
The detections from the five cameras for each day of a month were assigned to the divisions: daylight, dusk, night, or dawn.
Daily detection rates for these divisions were calculated and then summed for the month. The monthly detection rates for each of the divisions of the day were expressed as a percent- In order to provide a rational basis for grouping the months of the year, we plotted an activity index for the daylight and night periods akin to that of Manly et al. (2002). For this specific analysis (in contrast to the definitions above), night is defined as the period between sunset and sunrise (i.e., it includes both the dusk and the dawn period). This is done to simplify the graph and focus on the activity changes in the daylight period. We tested for significant differences in hare detection rate using an ANOVA with monthly detection rate as the response variable and the divisions of the 24-hr cycle (daylight, dusk, night, dawn)

| RE SULTS
The analysis of the results from camera trapping follows a step-bystep sequence which corresponds to the sequence of Figures 2-5.
We present the raw data across the months of the year in Figure 2.
We analyze the rise in daylight activity in spring and summer  (Figures 2 and 3).
We found a significant difference in detection rate between the divisions of the 24-hr cycle for October-February (F 3,16 = 78.78, p < .001) and March-September (F 3,24 = 32.08, p < .001). The Tukey post hoc test showed that, for October to February (Figure 5a), the mean detection rate was greatest for dusk, followed by dawn, night, and daylight with all differences significant (p < .01). For the March-September results (Figure 5b), the mean detection rate was highest for dusk which was significantly greater than that for dawn (p < .001). Dusk (p < .001) and dawn (p < .01) frequencies were significantly greater than those for night and daylight, but night and daylight frequencies showed no significant difference (p = .99).
In a separate ANOVA (F 7,40 = 39.6, p < .001), the detection rate was also compared for the different divisions of the 24-hr cycle between the two divisions of the year. Only the daylight division showed a significant difference (p = .002) between October-February and March-September.
On the basis of Figure 4 and the statistical analysis of Figure 5, we conclude that the activity that we see throughout the year is crepuscular. If we redefine our Activity index (described in Materials and Methods) as: we find Activity Indices (mean ± SD) over the whole year of 0.55 ± 0.38 for daylight, 2.35 ± 0.53 for dusk, 0.99 ± 0.29 for night, and 1.60 ± 0.42 for dawn. Activity index differs between 24-hr divisions (ANOVA

| D ISCUSS I ON
We suggest that there is a need for a consensus to be reached in the animal ecology community concerning the definitions of twilight, dusk, and dawn. Everyday usage defines twilight as the half-light that follows sunset or precedes sunrise and 'crepuscular' as behavior that inhabits these time periods. However, the scientific literature is inconsistent on the definition of these terms in the context of diel activity. Many authors use a 2-hr span that straddles sunset or sunrise (Caravaggi et al., 2018;Ikeda et al., 2016;Ogurtsov et al., 2018). Other authors (Campera et al., 2019;Narendra et al., 2010) define dusk and dawn in terms of astronomical twilight which ends or begins when the sun is 18° below the horizon (Thorsen, 2020). Still other authors (such as Bennie et al., 2014) use nautical twilight that ends or begins when the sun is 12° below the horizon (Thorsen, 2020). These variations in definition across studies means that it is not always possible to compare like with like.
We propose that the triggers of activity in species should be considered when defining twilight for biological studies. During postsunset and presunrise periods, the fall or rise in light intensity is exponential (Kishida, 1989;Thorsen, 2020). A wide range of animal species have Activity index = Proportion of detections in daylight, dusk, night, or dawn period∕Proportion of time in daylight, dusk, night, or dawn period F I G U R E 3 Levels of daylight and night activity of the mountain hare in the different months of the year. Activity index = Proportion of detections for the daylight or night periods (open circles and closed circles, respectively)/Proportion of time for the daylight or night periods. For this specific purpose, 'night' is defined as the period between sunset and sunrise (ie it includes both the dusk and the dawn period). The activity index was calculated for each month of the year. A value of one for that index (broken line) indicates that the proportion of detections and the proportion of time for that division of the 24-hr cycle are equal. The months between March and September indicated by the grayshaded area are associated with values for the daylight activity index (open circles) that approach a value of 1. This is in contrast to the low values for the daylight activity index found for the months of October to February been shown to alter behavior in response to these steep changes in light intensity (Leopold & Enyon, 1961;Narendra et al., 2010;Pepin & Cargnelutti, 1994). On that basis, we find no justification for defining The spring-summer period includes the months of March-September. The total number of detections in this period was 1,082. In both (a) and (b), light shading represents daylight, mid-gray shading represents dawn or dusk, and dark gray shading represents night. The left vertical line represents sunset, and the right vertical line represents sunrise F I G U R E 5 Mountain hare detection rates during daylight, dusk, night, and dawn over the 5 months of October-February (a) and the 7 months of March-September (b). The closed circles represent the means and the vertical bars the standard deviation show postsunset and presunrise activity peaks that are offset because of their use of this definition of twilight.
In our analysis of activity patterns, we define the twilight periods as 2 hr after sunset and 2 hr before sunrise. Our preference would have been to use astronomical twilight which, at the latitude of the present study, is close to 2 hr for much of the year. However, in mid-summer, the sun never reaches 18° below the horizon and the limit of astronomical twilight does not exist. Our 2 hr twilight periods are therefore a pragmatic solution and the 2 hr spans do match the width of the postsunset and presunrise activity peaks that we see (e.g., in Figure 4a).
There are two published studies on mountain hare activity using camera traps. They were conducted in forested areas in Hokkaido, Japan (Ikeda et al., 2016), and in the Valdai upland of Russia (Ogurtsov et al., 2018). Both use a definition of twilight as straddling sunset and sunrise but, taking that into account, their seasonal data show a bimodal crepuscular pattern with increased daylight activity in summer, a pattern that resembles our own results. However, the Ogurtsov et al. (2018) activity profiles for winter and spring followed a nocturnal activity pattern with a single peak in the middle of the night. This pattern differs from our own results and may be a consequence of the different habitats under study.
We have argued above that the triggers of activity in species should be considered when defining twilight for biological studies. The low light intensity of twilight stimulates secretion of the hormone melatonin from the pineal gland (Goldman, 2001). Melatonin determines activity levels and timings in crepuscular and nocturnal animals, and laboratory studies have demonstrated that this basic mechanism is subject to seasonal variations coupled to the reproductive cycle (Bartness & Goldman, 1989;Goldman, 2001). The Syrian hamster (Mesocricetus auratus) is a 'long-day breeder' and Goldman proposed that, in January and February, the gonads of the Syrian hamster become fully active. Under the influence of the reproductive hormones, the animal enters a photorefractory phase.
A similar timing of scrotal development in the male occurs in the mountain hare (also a 'long-day breeder') and all female adults are pregnant between March and June (Angerbjörn & Flux, 1995). Thus, in mountain hare, we propose that the onset of daylight activity that we see in Spring may be associated with a partial suppression of the melatonin-mediated circadian rhythm by the reproductive hormones. In the Syrian hamster (and we propose with mountain hare), this photorefractory phase ends with the shorter days as autumn approaches and that leads to regression of the reproductive organs, the restoration of full photosensitivity and the return to the autumn-winter pattern of crepuscular/nocturnal activity. This simple model based on the interaction of the reproductive hormones with the underlying circadian rhythm is likely to be subject to influence by other environmental factors, such as temperature, and physiological factors, such as nutrition.
We have argued above that our results can be interpreted as showing a basic crepuscular pattern that is modified during the breeding season leading to increased daylight activity. Although such activity represents a risk in terms of predation, it may be essential for efficient reproduction. In Scotland, the only raptors capable of taking an adult mountain hare are golden eagle (Aquila chrysaetos) or white-tailed eagle (Haliaeetus albicilla) but neither are resident in the study area (this may change in the future with the recent reintroduction of the golden eagle to southern Scotland (https://www. golde neagl essou thofs cotla nd.co.uk/). Therefore, the risks to a prey animal of daylight activity in this specific area may be low. Would the diel activity pattern be different for animals living in unmanaged areas with greater predation pressure? It would be informative to carry out a diel behavioral study of mountain hares in the alpine areas of the Cairngorm Mountains of Scotland that are not managed for grouse and where there are aerial predators.

| CON CLUS ION
The data presented are the first of their type on the Lammermuir mountain hare population, indeed the first for southern Scotland.
Despite the limited size of the camera trap study area (0.25 km 2 ) and small number of camera trap locations (n = 5), the density and activity of mountain hares at the camera trap locations provided a large number of detections. This allows us to perform an informative analysis of patterns of circadian and seasonal activity. We find a greatly increased activity in daylight in spring and summer that is superimposed on the underlying crepuscular pattern present through the year. We suggest that the driver for seasonal change in activity is the reproductive cycle and the mechanism is hormonal change.

ACK N OWLED G M ENTS
We thank the Mayshiel Estate and head gamekeeper, Phil Savage, for allowing us access to the study area and the Peoples Trust for Endangered Species for funding purchase of the AUCEE camera traps.

E TH I C A L A PPROVA L
The study has no ethical implications.

DATA AVA I L A B I L I T Y S TAT E M E N T
Camera trap photographs of mountain hare Lepus timidus per month Scotland (S1a), https://doi.org/10.5061/dryad.m0cfx pp3p.
Calculation of times relative to sunset and sunrise for mountain hare Lepus timidus camera trap records Scotland (S1b), https://doi. org/10.5061/dryad.2bvq8 3bpx. Calculation of detection rate for