Early View e01251
Research article
Open Access

Human avoidance, selection for darkness and prey activity explain wolf diel activity in a highly cultivated landscape

Peter Sunde

Corresponding Author

Peter Sunde

Department of Ecoscience, Aarhus University, Aarhus, Denmark

Contribution: Conceptualization (lead), Data curation (supporting), Formal analysis (lead), Funding acquisition (lead), ​Investigation (lead), Methodology (lead), Project administration (equal), Resources (lead), Validation (equal), Visualization (lead), Writing - original draft (lead), Writing - review & editing (lead)

Search for more papers by this author
Sofie Amund Kjeldgaard

Sofie Amund Kjeldgaard

Department of Research & Collections, Natural History Museum Aarhus, Aarhus, Denmark

Contribution: Conceptualization (supporting), Data curation (equal), Formal analysis (supporting), ​Investigation (equal), Methodology (equal), Project administration (supporting), Resources (supporting), Software (supporting), Supervision (supporting), Validation (lead), Visualization (supporting), Writing - original draft (supporting), Writing - review & editing (supporting)

Search for more papers by this author
Rasmus Mohr Mortensen

Rasmus Mohr Mortensen

Department of Ecoscience, Aarhus University, Aarhus, Denmark

Contribution: Formal analysis (equal), Methodology (supporting), Visualization (equal), Writing - original draft (supporting)

Search for more papers by this author
Kent Olsen

Kent Olsen

Department of Research & Collections, Natural History Museum Aarhus, Aarhus, Denmark

Contribution: Conceptualization (supporting), Data curation (lead), Formal analysis (supporting), Funding acquisition (supporting), ​Investigation (equal), Methodology (equal), Project administration (equal), Resources (equal), Writing - original draft (equal), Writing - review & editing (equal)

Search for more papers by this author
First published: 15 May 2024

Abstract

Wildlife that share habitats with humans with limited options for spatial avoidance must either tolerate frequent human encounters or concentrate their activity into those periods with the least risk of encountering people. Based on 5259 camera trap images of adult wolves from eight territories, we analyzed the extent to which diel activity patterns in a highly cultivated landscape with extensive public access (Denmark) could be explained from diel variation in darkness, human activity, and prey (deer) activity. A resource selection function that contrasted every camera observation (use) with 24 alternative hourly observations from the same day (availability), revealed that diel activity correlated with all three factors simultaneously with human activity having the strongest effect (negative), followed by darkness (positive) and deer activity (positive). A model incorporating these three effects had lower parsimony and classified use and availability observations just as well as a ‘circadian' model that smoothed the use-availability ratio as function of time of the day. Most of the selection for darkness was explained by variation in human activity, supporting the notion that nocturnality (proportion of observations registered at night vs. day at the equinox) is a proxy for temporal human avoidance. Contrary to our expectations, wolves were no more nocturnal in territories with unrestricted public access than in territories where public access was restricted to roads, possibly because wolves in all territories had few possibilities to walk more than few hundred meters without crossing roads. Overall, Danish wolf packs were 6.5 (95% CI: 4.6–9.6) times more active at night than at daylight, which make them amongst the most nocturnally active wolves reported so far. These results confirm the prediction that wolves in habitats with limited options for spatial human avoidance, invest more in temporal avoidance.

Introduction

Worldwide, humans heavily influence the populations and ecosystem functions of large carnivores (Ripple et al. 2014). Human impacts on large carnivore populations are not only manifested through lethal interactions, but also indirectly by inducing costly avoidance behaviors which result in effective habitat loss through spatial avoidance or reduced foraging efficiency (Musiani et al. 2010, Kuijper et al. 2016, Ordiz et al. 2021). As is the case for other wildlife that are or have been subject to human persecution, avoidance of people by large carnivores should be considered as an adaptive anti-predator strategy directed towards our species (Frid and Dill 2002, Lasky and Bombaci 2023). Accordingly, many large carnivorous mammal species with no or few natural enemies, but an evolutionary history of human persecution, try to minimize human encounters by avoiding habitats with high risks of encounters with humans, especially during periods when people are most active (Kautz et al. 2021, Ripari et al. 2022, Smith et al. 2022, Thorsen et al. 2022).

Populations of some large mammalian carnivore species are increasingly recolonizing human-impacted landscapes following legal protection, as is the case for grey wolf Canis lupus (‘wolf' hereafter), lynx Lynx lynx and brown bear Ursus arctos in Europe (Chapron et al. 2014). As a result, knowledge about behavioral adaptions to human presence in habitats heavily used by people is important to predict where populations may establish in the future, how they may cope and how they will behave in relation to humans in such habitats. In that regard, one may hypothesize that in landscapes with uniformly high human activity, reduced activity during those periods of the day where humans are most active, might represent an optimal avoidance strategy. This is simply because spatial avoidance of people might be difficult for predators that patrol large home ranges in search of their prey, which they sometimes pursue over long distances during hunting episodes. As humans are strongly diurnal, in landscapes with few options for spatial avoidance, increased nocturnally would be a logical outcome, as has been shown for mammals in general (Gaynor et al. 2019, Lewis et al. 2021).

The wolf represents an interesting species to study behavioral adaptations to human presence in a formerly heavily persecuted top-predator mammal species, now re-colonizing ever more cultivated and densely humanly impacted landscapes (Boitani et al. 2022). Where wolves and humans coexist, humans represent the predominant source of mortality for wolves (Mech 2017). This is even the case for protected populations, where traffic mortality and illegal killings may dominate causes of mortality, and where both of which are likely to be a function of human encounter rates in space and time (Suutarinen and Kojola 2018, Nowak et al. 2021, Sunde et al. 2021). Under such circumstances, wolves should minimize their probability of encounters with humans. Spatially, wolves generally avoid using areas near human infrastructures, especially during the daytime (Theuerkauf et al. 2003a, 2007, Mancinelli et al. 2019, Carricondo-Sanchez et al. 2020, Smith et al. 2022), unless those infrastructures serve a purpose, such as to facilitate the movement of wolves in the landscape (Zimmermann et al. 2014, Bojarska et al. 2020, Kautz et al. 2021). In Alberta, Canada, wolves registered with wildlife cameras became more nocturnal after initiation of a control program, resulting in less activity overlap with prey species (Frey et al. 2022). In Spain, wolves registered with wildlife cameras were more diurnally active in a site with restricted human access compared to two control sites with no human access restrictions (Martínez-Abraín et al. 2023). In areas where undisturbed foraging habitat is available in relative excess, spatial avoidance of humans might represent a low-cost strategy, as foraging activities can be just directed from one spatial location to another without modifying the foraging effort per se. In the Bialowieza Forest in Poland, for example, wolves showed spatial avoidance of humans, but their diel activity appeared to be driven by the activity of their prey and not human activity (Theuerkauf et al. 2003a, b2003b). As wolves are expected to synchronize their diel cycle to optimize their foraging efficiency, temporal human avoidance through alteration of the diel cycle, may therefore be considered a last-resort strategy used when opportunities for spatial avoidance have been exhausted. In such habitats, wolves are left with the choice of either avoiding humans by reducing activity overlap or adapt (habituate) to human presence while active. Wolves that establish territories in densely cultivated landscapes with considerable human activity may therefore cope by reducing their temporal activity overlap with humans, by becoming more nocturnal than in habitats with less human activity (Martínez-Abraín et al. 2023, Petridou et al. 2023). Conversely, if temporal human avoidance is costly (e.g., in terms of reduced foraging efficiency), one should expect a reduction in nocturnality over time, unless counterbalanced by the costs of human encounters (Martínez-Abraín et al. 2023).

The level of temporal human avoidance (possibly proxied by nocturnality) of wolf populations in cultivated landscapes is of biological interest as well as of relevance for management, as it may provide insight into the behavioral adaptability of wolves to co-exist with humans in shared habitats and how this may develop in coming years. Despite its relevance for understanding the behavioral adaptions of wolves to human presence and its ecological consequences in shared habitats, diel activity patterns have so far not been described for wolves in the highly cultivated landscapes in west-central Europe. Even more fundamentally, the proximate ecological and environmental drivers (humans, prey, light conditions) that shape their timescape selection have not been analyzed on a population level within a statistical framework where their total and partial effects are disentangled.

For animals in general (Gilbert et al. 2023) and wolves in particular (Theuerkauf 2009), decisions about the timing of periods of activity depend on multiple drivers. In simplified form, these can be categorized as 1) an underlying, circadian rhythm, which is a genetically hard-wired entity formed by selection (Yerushalmi and Green 2009), and dynamic factors to which wolves respond: 2) ambient conditions (light, temperature. etc.), 3) foraging opportunities (prey activity) and 4) predation risk (for wolves synonymous with humans) (Fig. 1). Disentangling the effects of dynamic drivers on diel activity patterns is statistically challenging due to confounding and counteracting factors. For instance, wolves and prey may both be responding to the activity of their shared super-predator, humans (Visscher et al. 2023) that is highly diurnally active. Increased activity during periods of darkness may therefore in principle be a result of a genuine preference for being more active in darkness, selection for periods where prey is most active or when humans are less active. Accordingly, a successful analysis must be able to tease apart the effects of predictors representing competing hypotheses.

Details are in the caption following the image

Direct (black arrows) and indirect (grey arrows) drivers of diel activity patterns in wolves in habitats shared with humans: human activity, prey activity, light conditions, and general circadian rhythm.

In this paper, we aim first to describe the underlying diel activity patterns of a wolf population in a highly cultivated landscape (Denmark/west-central Europe) and second to identify the drivers behind the spatial and seasonal variation in diel activity. As the options for spatially avoiding humans during their normal activity patterns are presumed to be highly limited, we expect wolves from this population to invest heavily in temporal avoidance of humans.

First, we describe the diel activity profiles of Danish wolves at different seasons, representing different night lengths. Second, by means of a resource selection approach (use-availability), we test the extent to which variation in wolf diel activity (timescape selection) is explained by variation in 1) light conditions, 2) prey activity and 3) human activity (each predictor separately as well as all three together). We would consider a predictor as a significant potential driver of wolf timescape selection if it remained significant, when accounting for the two other predictors. Furthermore, we evaluated the ability of the three predictor variables' (in isolation and in combination) to predict variation in diel activity as the statistical models' classification success in use versus availability observations, compared to the classification success obtained by a (general additive) model that fitted the diel activity as a simple smoothing function of time from midnight to midnight (representing a basic circadian activity function).

Third and finally, since our results indicate nocturnal activity to be a temporal mechanism to avoid humans, we analyzed conditional nocturnality as a binary outcome (observed at night versus day) relative to night length under different social conditions (territory occupied by a single individual, pair or pack), and monitoring aim (wolf population monitoring specifically or wildlife in general) considering random variation between territories. If nocturnality is a plastic response to reduced options to avoid humans spatially, we would expect wolves in territories with restricted public access to be less nocturnal than wolves in territories with free public access. At the population level, we would expect wolves from this population to express some of the highest levels of nocturnality reported worldwide, as Danish landscapes offer very limited opportunities to avoid humans spatially when active.

Material and methods

Study area and population

Wolves were extirpated from Denmark during the 18th century (the last vagrant was killed in 1813). In modern times, the first immigrants from the Central European Lowland population were registered in 2012, and reproduction was first observed in 2017. By August 2023, the Danish sub-population numbered about 30 adult individuals, including at least two packs, three pairs and two resident single individuals (Sunde et al. 2023b). Despite strict legal protection, the mortality rate is high (2012–2019: 46% per year) and suspected to be mainly caused by illegal killings (Sunde et al. 2021).

Wolves are currently confined to the Jutland peninsula (29 778 km2) in Denmark, where the human population density is 87 km2, and the land surface cover comprises 12% developed, 61% farmland, 13% forest (almost all planted and managed) and 10% heathland. Wolf occupation patterns on a 10 × 10 km level correlate positively with forest and heathland cover (Mayer et al. 2022). Resident wolves maintain home ranges within core areas of forest and heathland, surrounded by farmland, through which wolves readily cross when moving between forest parts to hunt for prey. Home range sizes for pairs and packs are approximated to 100–200 km2, estimated from registered signs (scats, kills, photos etc.) (Olsen et al. 2020) and data from a single GPS-collared individual (Sunde et al. 2023a).

In Denmark, the public has free access to nature areas on governmental-owned land apart from a few wildlife reserves. In privately owned forests, the public has access on foot and bike along forest roads. In all territories, few forest or heathland areas are located more than 500 m from the nearest road.

Population monitoring and data collection

Since 2017, Natural History Museum Aarhus and Aarhus University have monitored all wolves in Denmark for the Danish Environmental Protection Agency, following the standards defined by Reinhardt et al. (2015). The occurrence and turnover of individuals is registered from genetic markers obtained from scat, hair, saliva or urine samples collected by systematic patrolling of forest roads and by snow tracking (active monitoring) as well as saliva samples from livestock kills obtained by the Danish Nature Agency.

A territory was defined as the area patrolled by a single wolf, pair or pack for minimum six months following Reinhardt et al. (2015). The core areas and approximate territory extensions were estimated from the distribution of wolf signs (scats, tracks, kills, photos etc.) within the landscape (Sunde and Olsen 2018). As part of the monitoring, with permission from the landowners, we placed wildlife cameras in places known (from the appearance of footprints, scats, or other signs) or suspected (leading lines in the landscape which from experience are known to be used by wolves when commuting, e.g. forest roads) to be used by the wolves within the territories. At locations with public access, visitors were informed about the presence of the cameras through signs containing project information and our contact details. We used cameras with fast trigger times able to record fast-moving species, that recorded videos and/or multiple pictures (see the Supporting information material for a list of models and settings used in the different territories). Cameras were usually visited every two to six weeks, checking battery levels and changing memory cards. The raw image files were uploaded to a server with restricted access upon which all memory cards were formatted before reuse. Where possible, the wolves on the images were identified to age defined as pup (born same calendar year) or adult (not a pup, hence all grown-up wolves observed January–June were coded as adults) and coded in the database. If multiple wolves on the same photo or video sequence were identified to different age or individual, they were registered as different records in the database. Prior to the analyses, such doublets or triplets were removed, so only one unique camera observation entered the analysis as observation unit.

As the cameras were placed to maximize the number of wolf observations, sampling effort was concentrated in the central parts of the territories where wolf sign concentrations were highest. Cameras aimed at recording wolves were usually placed along trails and forest roads used by wolves when traversing their territories and at places with high density of scats and footprints, that indicated frequent use by wolves at a given time. In a subset of the territories we also had cameras placed in the terrain, optimized to register all large and medium sized mammal species. For wolf population monitoring purposes, observations from both types of surveys were entered in the Danish national database of wolf observations. The effort expended in terms of camera days was not registered in this database. Observations of general wildlife were logged in a separate database, which also included information on the effort expended in terms of the number of camera days (26 210 in total). Due to resource constraints, this database only contained a subset of the total number of camera observations available in the raw data. The wolf data used for this analysis were therefore drawn from the first database. The number of different camera locations, resulting in wolf observations varied from 26 to 198 per territory (median: 51) and the total area covered (100% minimum convex polygon) by cameras delivering wolf data for the analysis, ranged from 6.5 to 79.3 (median: 21.3) km2 per territory (Table 1, Fig. 2).

Details are in the caption following the image

. Map of Denmark showing the distribution of the territories in the analysis. Polygons and dots indicate distribution of camera observations for each territory.

Table 1. Wolf territories and wildlife camera observations selected for the analysis. Access: rules for public access (R: restricted, F: free access). Fenced: whether the entire or major part of the surveyed area was surrounded by wildlife fence (Y: yes, N: no). Pos.: number of different wildlife camera positions contributing with wolf or human + deer images. Area: area (km2) covered by a 100% minimum convex polygon, surrounding all camera locations resulting in wolf images. Pups: number of images only showing pups (June–December). Unkn. : number of images only showing wolves for which age could not be determined (not included in analyses). Adults Total: number of images showing at least one adult wolf (divided on social status: Pack, Pair or single territory holder). Deer: number of images showing deer (all species). Humans: number of images showing human observations (incl. vehicles). Days: number of camera days (only available for deer and human observations). For more territory information, see the Supporting information
Territory information Wolf observations Deer and human observations
Territory Access Fenced Pos. Area Pups Unkn. Pack Pair Single Adults total Deer Humans Pos. Days
Hovborg R Y 198 16.3 1,125 124 2,593 102 3 2,698 5,373 6,496 332 13,397
Klosterhede F N 38 22.1 0 0 0 80 79 159 . .
Nørlund F N 36 79.3 0 0 0 0 86 86 1,729 3,052 73 3,014
Skjern R N 54 20.5 423 1 449 86 0 535 . .
Ll. Vildmose R Y 26 9.2 0 0 0 0 252 252 . .
Ulfborg-1 F N 47 6.5 82 9 107 30 0 137
Ulfborg-2 F N 172 37.8 184 22 262 313 90 665 4,213 5,469 577 9,799
Ulfborg-3 F N 65 50.4 0 2 0 355 372 727
Sum: 636 1,814 158 3,411 966 882 5,259 11,315 15,017 982 28,210

Selection of observations for analyses

We selected wolf camera trap data separated by minimum 5 minutes from eight independent territories from six areas (one territorial area was occupied by three different constellations of individuals during different time periods: Table 1, Supporting information). This selection resulted in 5259 camera observations of adults, 1814 observations only showing pups (representing five litters from four territories), and 158 observations where the age could not be determined (excluded from the analyses). Of the 5259 observations of adult wolves, 3280 (62%) originated from cameras for monitoring wolves, 1257 (24%) from cameras for monitoring general wildlife and 722 (14%) from cameras where the initial purpose had not been recorded. As any dependence between observations within territories was accounted for in the statistical analyses by stating territory as random effect, we decided to use the full data set rather than a reduced data set based on observations separated by 30 minutes, as often recommended to avoid serial dependence of observations (Tobler et al. 2008, Ferretti et al. 2023). Under all circumstances, increasing the minimum sampling interval from 5 to 30 minutes only reduced the data set by 250 observations, and did not change the outcome of any of the statistical analyses.

Two territories were only occupied by single individuals, two territories were occupied some of the time by single individuals, later by pairs, two were first surveyed while occupied by a pair that later became a pack, and two surveyed during periods while occupied by single individuals, pairs and packs (Table 1, Supporting information). Three of the six territory areas (five of eight territories) had free public access, while access was restricted in the remaining three (one unfenced military area, two privately owned nature areas enclosed by wildlife fences through which no ungulates could pass: Table 1, Supporting information).

Digitized data on wildlife and human activity was available from three of the six territory areas (five of eight territories: Table 1). As ungulates, especially cervids (Cervidae, ‘deer' hereafter) constitute the main and most selected prey type of wolves in central Europe (Nowak et al. 2011, Jedrzejewski et al. 2012, Wagner et al. 2012, Newsome et al. 2016), we used 11 315 camera observations of deer (species composition: red deer Cervus elaphus: 60%, roe deer Capreolus capreolus: 31%, fallow deer Dama dama: 4%, unidentified deer species: 5%) to represent diel activity of prey. The 15 017 observations of humans were divided between 48% pedestrians, 16% bicyclists, 31% motorized vehicles and 5% horse riders.

Seasonal definition

To account for seasonal variation, we divided the year into quartiles: November–January (mean day length at 56˚N: 8.47 hours; range: 7.92–9.68 hours), February–April (11.97; 9.22–14.77), May–July (16.01; 14.83–16.55) and August–October (12.57; 9.73–15.32). These not only contrasted the two three-month periods with the shortest and longest daylengths, but also provided a good division of the ecological and reproductive annual cycle for wolf packs that have young offspring in May–July, mobile offspring (frequenting rendezvous sites) in August–October, increasingly independent offspring through November–January, and a pre-parturition period from February–April, when last year's offspring have attained full independence.

Statistical analysis of diel activity patterns of wolves, deer and humans

For each three-month period, we quantified the general variation in diel activity of juvenile and adult wolves (W hereafter), deer (D) and humans (H), by modelling the relative frequency of observations per one-hour-interval from midnight to midnight (0: 00:00–00:59, 1: 01:00–01:59, etc.). We used the R package ‘mgcv' ver. 1.8 to fit generalized additive models (GAM) with beta distribution and logit-link, an adaptive cyclical cubic smoothing spline for time, and territory ID as random effect (Zuur 2012, Simpson 2014, Wood 2017). Models were visually validated by plotting standardized model residuals against fitted values (Zuur 2012, Wood 2017).

As data for human and deer activity was not available for all areas (Table 1), extrapolation was necessary. As the season-specific diel activity curves for humans were highly correlated between the different study areas (Supporting information), we produced one season-specific diel activity function based on all data pooled. Among the deer, diel activity correlated less between fenced and unfenced areas (Table 2, Supporting information). As we know from GPS-data that red deer in fenced areas move shorter hourly distances around dusk and dawn than red deer in unfenced nature areas (Sunde and Mortensen unpubl.), we created one activity distribution for deer based on data from all three areas (‘Deer-total', abbreviated to ‘DT'), and one differentiated between fenced and unfenced areas (‘Deer-local', abbreviated to ‘DL').

Table 2. Correlation matrix (Pearson's r) of the diel activity levels throughout the year for adult wolves (W), humans (H) and deer (D, divided on DT: total observations, DL(F): fenced area and DL(UF): unfenced areas) and whether the sun was below or above the horizon (ND, scored as night = 1, day = 0). The matrix is based on 8,760 hourly observations, covering the entire year (24 per day from 1 Jan. to 31 Dec) for each of which is derived the time and date specific sun angle and activity level values of W, H and D (Fig. 4)
ND W H DT DL(F) DL(UF)
ND 1.00 0.78 −0.78 0.15 0.47 0.00
W 0.78 1.00 −0.78 0.47 0.76 0.30
H −0.78 −0.78 1.00 −0.25 −0.53 −0.07
DT 0.15 0.47 −0.25 1.00 0.71 0.96
DL(F) 0.47 0.76 −0.53 0.71 1.00 0.52
DL(UF) 0.00 0.30 −0.07 0.96 0.52 1.00

To quantify the extent to which diel activity levels of adult wolves, deer and humans were associated with light conditions and correlated internally throughout the year, we created a correlation matrix comprised by 24 × 365 = 8760 hourly time observations, covering the entire year (1 January–31 December). For each hourly time observation, we assigned light conditions (categorical variable: night [sun angle < 0˚, coded as 1] versus day [sun angle ≥ 0, coded as 0], ‘ND') and the relative diel activity of W, H, DT and DL (separated between fenced and unfenced areas) as predicted for each of the four seasons by the GAM-models.

Statistical analysis of predictors of timescape selection

We analysed the extent to which wolves selected to be active as function of ND, H and DT/DL by means of a resource selection function (RSF) based on use-availability design (Boyce et al. 2002). For each wolf observation (used) we created 24 pseudo-observations, representing every hour for the same date and geographical location of the wildlife camera image.

We used mixed effect binary logistic regression (response variable: used versus unused) with territory ID as random intercept. As models with camera type as random effect resulted in similar results as models without camera type, we did not include camera type as random effect. We evaluated the extent to which a fixed effect (ND, DT or DL, H) predicted variation in diel activity from the magnitude of its selection coefficient in the logistic regression equation (Boyce et al. 2002) and considered it important if deviating significantly (p < 0.05) from 0. To establish the raw effects, we first ran models containing each of the three types of fixed effects in isolation (four different models in total as deer activity was represented by two alternative variables). To establish the extent to which the magnitude of a given fixed effect remained (or was reduced) when controlling for one or both two other competing fixed effects, we furthermore ran models representing all possible combinations of main effects of ND, H and DT/DL. If a fixed effect remained statistically significant when controlling for the other two fixed effects, we took it as indication that it contributed genuinely to explaining the variation in diel activity. We used the R package ‘mgcv' ver. 1.8 to fit generalized mixed models (GLM) with binomial distribution, logit-link, and territory ID as random effect (Zuur 2012, Simpson 2014, Wood 2017). Models were visually validated by plotting standardized model residuals against fitted values (Zuur 2012, Wood 2017).

To explore the predictive power of H, ND, and DT/DL (in isolation as well as in combination), we calculated Somers' D, which is a nonparametric index of a model's ability to correctly classify the dependent variable, derived as D = 2 (AUC – 0.5) where AUC is the area under the model's receiver operation curve. If D = 1, all observations are correctly classified by the model, whereas D = 0 indicates a non-informative model. To compare the predictive effects of the three fixed-effects variables with a model based on mere smoothing of the basic circadian pattern, we contrasted the D-values of the models based on different combinations of the three fixed effects with those of a mixed effects logistic GAM-model where activity was smoothed as function of time of day. The GAM-model was fit with an adaptive cyclical cubic smoothing spline for time, and territory ID as random effect (Zuur 2012, Simpson 2014, Wood 2017). If Somers' D for a fixed effects model approached the D-value of the GAM-model, we took that as an indication that the fixed effects model was able to capture most of the diel activity variation. For all models considered, we calculated AICc. We considered a difference in AICc-values (∆AICc) > 7 as substantial evidence for the model with the lowest AICc-value to have most support in data (Burnham et al. 2011)

Analysis of variation in nocturnality

Using nocturnality as a proxy for temporal human avoidance, we modelled the conditional probability that a wolf image was registered at night (sun angle < 0°; pn) as opposed to at day (sun angle ≥ 0°) as a logistic regression function, where we included the logit-transformed night length (logit [NL], the logit-transformed proportion of the day the sun is below the horizon) as a nuisance variable to adjust for variable night length. From the logistic regression equation, we derived conditional nocturnality (N) as the difference on a logit-scale between the predicted probability that an observation under given circumstances would be nocturnal, relative to what should be expected from the length of the night (N = logit [pn] − logit [NL]). From simple algebra it follows that the activity selection ratio between night and day (SRN:D: how many more times wolves were observed per time unit at night as opposed to daytime) can be derived as RN:D = exp [N]. At the equinox (where NL = 0.5, hence logit [NL] = 0), this expression could be simplified to N(NL = 0.5) = logit [pn] = ln (pn/[1 − pn]) ⇔ RN:D = pn/(1 − pn). As an example, if a sample of study subjects was observed equally often at day and night, standardized at the equinox, pn[NL=0.5] = 0.5 and SRN:D = 0.5/(1 − 0.5) = 1. In comparison, situations where 80% and 90% of the observations at the equinox were nocturnal corresponded to wolves being 4 (0.8/0.2) and 9 (0.9/0.1) times more active during darkness than in daylight, respectively.

The analysis was run as a mixed effect logistic regression function with territory ID as random intercept, using the GLIMMIX procedure in SAS 9.4 with a logit link function, binomial error distribution, using Satterthwaite's approximation to estimate degrees of freedom. As fixed effects we included logit [NL]), social status (single wolf, pair or pack), human access (restricted or free), camera purpose (wolf monitoring, general wildlife monitoring, not registered). As models with camera type as random effect resulted in similar results as models without camera type, we did not include camera type as random effect.

We first ran the full model, evaluating the statistical significance of the fixed effects by means of type-III F-statistics. Since the type-III analysis revealed no effect of human access, we reran the model without this predictor. Finally, we tested for a possible significant interaction term between social status and night length. All models had a generalized Χ2/df ratio between 1.00 and 1.03, suggesting perfect residual fit, hence no adjustment for overdispersion was necessary.

Results

General diel activity patterns

Throughout the year, adult wolves showed a bimodal activity pattern, peaking at ca 18–20 in the evening and at ca 4–6 in the morning and a dip in the middle of the day (Fig. 3). On a seasonal basis, the inactivity period was shortest (ca 10–16: 6 hours) in November–January and longest (ca 8–17: 9 hours) in May–October (Fig. 3). During August–October, pups were slightly more diurnal than adults as their activity peaked 1–2 hours later in the morning and 1 hour earlier in the evening. During November and December their activity patterns had converged towards that of the adults (Fig. 3).

Details are in the caption following the image

Diel activity patterns (1 = average) for adult (95% confidence zones indicated with darker color) and juvenile (95% confidence zones indicated with lighter color) wolves, divided into three-month periods. Estimates and 95% confidence zones are based on GAM-models. Black lines indicate diel activity patterns of adult wolves for each of the eight territories.

Throughout the year, human activity peaked between 10 and 15, with the activity period being shortest from November to January and longest from May to October (Fig. 4). Like wolves, deer showed a bimodal activity pattern, with activity peaks around 5 and 17–18 from November to April and 7–8 and 20–21 from May to October (Fig. 4).

Details are in the caption following the image

Diel activity patterns (1 = average) for adult wolves (same functions as in Fig. 3), deer (all species) and human activity divided into three-month periods. Estimates and 95% confidence zones are based on GAM-models.

Throughout the entire year, the predicted diel activity levels of adult wolves correlated strongly positively (r = 0.78) with periods of darkness and equally strongly, but negatively, with predicted human activity (Table 2). Predicted human activity in turn correlated equally strongly negatively with periods of darkness (r = −0.78). Predicted wolf activity correlated modestly positively with predicted deer activity (DT, r = 0.47) (Table 2).

Predictors of timescape selection

Wolves selected for the dark periods, periods with high deer activity and low human activity (Fig. 5). All three predictors remained statistically significant if entered together (Fig. 5), suggesting that they all contributed to the variation in wolf activity. However, the magnitude of the selection coefficients of nd was reduced by 60% in models that included H, indicating that most of the selection for dark hours could be contained by temporal variation in human activity. In comparison, the effect size of H was only reduced by 33% in models that included nd. Selection strengths for ND and H were not affected by the inclusion of DT or DL in the models. Similarly, selection strengths of DT and DL were only modestly reduced by including ND and/or H (Fig. 5).

Details are in the caption following the image

Selection coefficients with 95% confidence errors for timing of diel activity in relation to variation in the following four predictors (from left): human activity (H), night-day (nd: sun angle < 0° versus sun angle ≥ 0°), Deer activity derived for the total data combined (DT) and local deer activity differentiated between fenced and unfenced populations (DL). In the top of the figures, estimates are shown for isolated effects (no other variables in model); in the bottom of the figures, estimates are shown for the partial effects in models including all three fixed effect predictors. All effects shown in the figure were highly statistically significant (p < 0.0001).

In line with the effects of all three predictors being statistical significant when entered alongside in the same model, the fixed-effects models that included all three classes of predictors (H + nd + DT and H + nd + DL ) had substantially higher support than the best model comprised by only one (H: ∆AICc ≥ 325) or two (H + DL or H + DT: ∆AICc ≥ 140) factors (Table 3). Notably, the two top models comprising the three fixed effects had more support in data (∆AICc > 76) and approximately equal classification power as the ‘time' model that modelled activity (use-availability ratio) as a function of time of the day (all models: Somers' D = 0.39: Table 3).

Table 3. Parsimony (AICc: Akaike's information criterion corrected for small sample size) and ability to correctly classify use and availability observations expressed as AUC (Area under the curve) and Somers' D (D = 2[AUC − 0.5]) of alternative models. D/DTime indicates the ratio in predictive ability of fixed effects models (comprised by different combinations of nd, DL/DT or H) compared to a model that smooths the diel activity as a general additive model (GAM) function of time from midnight to midnight. Abbreviations; Time: diel activity smoothed as function of time from midnight to midnight; nd: night or day (sun below or above the horizon), DT: deer activity estimated from total sample (all study areas), DL: deer activity divided between fenced and unfenced areas; season: (May–July, August–October, November–January, February–April). All models had territory ID as random intercept
Model AICc ∆AICc AUC D D/DTime
ND + DL + H 41164.9 0.0 0.696 0.391 0.99
ND + DT + H 41171.3 6.3 0.693 0.385 0.99
Time (GAM-model) 41247.5 82.5 0.696 0.393 1.00
DL + H 41304.9 140.0 0.690 0.381 0.97
DT + H 41321.2 156.3 0.687 0.373 0.95
ND + H 41360.7 195.8 0.680 0.361 0.92
H 41496.8 331.9 0.677 0.354 0.91
ND + DL 41657.2 492.3 0.691 0.382 0.98
ND + DT 41702.9 538.0 0.688 0.375 0.96
ND 41825.3 660.4 0.665 0.331 0.85
DL 43698.6 2533.7 0.599 0.199 0.51
DT 43912.5 2747.5 0.574 0.149 0.38
No predictors 44161.8 2996.9 0.500 0.000 0.00

If entered as individual effects, H had most support in data and the highest classification success (D = 0.36), followed by nd (D = 0.33), DL (D = 0.20) and DT (D = 0.15: Table 3). Of the two alternative measures of deer activity, DL had far more support than DT in models including no other fixed effects (∆AICc = 214: Table 3), whereas the difference was more modest for models that also included H and nd (∆AICc = 6.4: Table 3).

Nocturnality

The probability that a camera observation was registered at night (sun angle > 0°) varied significantly as a function of night length, social status, and the survey purpose of the camera (Table 4), but not between territories with free and restricted public access (F1,7.23 = 0.13, p = 0.73). The interaction term between social status and night length was not statistically significant (additive effect to model stripped of the effect of public access: F2,5259 = 1.38, p = 0.25).

Table 4. Fixed and random effects predictors of the probability that a wolf observation on wildlife cameras is registered at night (sun angle < 0). NL: night length (proportion of the calendar date sun angle was < 0). Territories with restricted public access are indicated by ‘(R)'
Predictors B SE df t p Type-III test
Fixed effects:
Intercept 1.981 0.214 32.13 9.27 < 0.0001
Logit (NL) 1.472 0.083 5259 17.76 < 0.0001 F1,5259 = 315, p < 0.0001
Status (pack vs single) 0.545 0.170 164.6 3.22 0.002 F2,394.6 = 6.50, p = 0.0017
Status (pair vs single) 0.102 0.138 1017 0.74 0.458
Aim (general vs unkn.) −0.187 0.142 5259 −1.32 0.189 F2,5259 = 12.37, p < 0.0001
Aim (wolf vs unkn.) −0.658 0.145 5259 −4.55 < 0.0001
Individual territories (relative to population mean):
Klosterhede −0.455 0.200 15.85 −2.28 0.037
Skjern (R) −0.294 0.178 12.24 −1.65 0.124
Hovborg (R) −0.224 0.181 12.55 −1.24 0.237
Nørlund −0.077 0.238 15.63 −0.32 0.752
Ulfborg-1 −0.010 0.237 15.5 −0.04 0.967
Ulfborg-2 0.170 0.166 10.86 1.02 0.329
Ulfborg-3 0.213 0.170 11.68 1.25 0.235
Ll. Vildmose (R) 0.677 0.220 16.4 3.08 0.007

The slope of the logistic regression line relative to night length was significantly steeper than 1 (Table 4; test of H0: b = 1: t5259 = [b − 1]/SEb = [1.472 − 1]/0.083 = 5.69, p < 0.0001), meaning that RN:D increased with increasing night length. Hence, for packs, pn varied from 69% at the summer solstice (sun below horizon 31% of the time, SRN:D = 5.0) to 96% at the winter solstice (sun below horizon 69% of the time, SRN:D = 10.8) (Fig. 6a). Packs were slightly more nocturnal than pairs and single wolves (Table 4), as packs at the equinox were SRN:D = 6.5 (95% CI: 4.6–9.6) times more active at night than in daytime (pn = 88% nocturnal observations) compared to SRN:D = 4.5 (95% CI: 3.2–6.5, pn = 84%) for pairs and SRN:D = 4.1 (95% CI:2.8–5.9, pn = 81%) for single wolves (Fig. 6b). The overall nocturnality selection ratio at the equinox (disregarding social status) was SRN:D = 5.1 (95% CI: 3.6–7.1) or pn = 83% nocturnal observations.

Details are in the caption following the image

(A) Nocturnality (shown for adult wolves from packs with 95% confidence zones) expressed as the proportion of wildlife camera observations registered at sun angles < 0° as a logistic regression function night length (NL: proportion of the day with sun angles < 0°). The dotted line indicates the line representing equal activity by day and night. (B) Selection coefficients for nocturnality at the equinox for adult wolves on territory held by a pack, pair, or a single individual. The coefficients (95% confidence intervals indicated) represent the log-odds ratio of the difference in activity between periods of day light and dark at the equinox (i.e. a coefficient of 2 indicate that wolves are exp [2] = 7.39 times more active at night than at day).

Cameras positioned to monitor wolves resulted in fewer nocturnal observations (across social status: SRN:D = 4.7, 95% CI: 3.3–6.6 or pn = 82%) than cameras mounted for monitoring of all species of wildlife (SRN:D = 7.5, 95% CI: 5.1–11 or pn = 88%) and cameras with an unknown purpose (SRN:D = 9.0, 95% CI: 6.0–14 or pn = 90%) (Table 4).

The coefficient/SE-ratio of the covariance parameter of random variation (0.133/0.087) did not suggest overall significant variation in nocturnality between territories (z = 1.52, p = 0.13), although predictions for one territory (Klosterhede) stood out as significantly less, and another territory (Lille Vildmose) as significantly more nocturnal than the population mean (Table 4).

Discussion

Methodological considerations

To our knowledge, this is the first study of diel activity in wolves based on camera trap data from multiple independent territories, using statistical analyses to account for possible variation between territories. As such, this study is probably also the first to provide reliable estimates of wolf activity patterns on population level based on camera trap data. Even though diel variation in camera trap observations is considered a reliable indicator of diel activity variation in carnivorous mammals (Lashley et al. 2018), estimates of diel variation based on camera trap data are sensitive to temporal variation in spatial distribution. As wolves generally spend more time resting in the core area than in the peripheral parts of their home range (Okarma et al. 1998, Mancinelli et al. 2018), and select more for habitats with low human disturbance risk when resting at daytime than at night (Mancinelli et al. 2019, Carricondo-Sanchez et al. 2020, Smith et al. 2022), wildlife cameras placed at locations used disproportionately for resting may therefore overestimate the magnitude of diurnal activity and vice versa. As our cameras were concentrated in the core areas of the territories (median area covered by cameras: 21.3 km2) and in forest habitats, it is conceivable that our data underestimated activity periods when wolves patrolled and hunted in the more peripheral parts of their home range and in open habitats. In that case, Danish wolves are probably even more nocturnally active than suggested by our data. Apparent differences in nocturnality between wolves of different social status may reflect differential spacing patterns rather than any real difference in diel activity. On the hand, if active wolves used trails and forest roads less during daytime than at night (Kautz et al. 2021), it would skew the diel distribution in the opposite direction. We found the opposite pattern, because cameras mounted away from roads to register all species of wildlife resulted in a higher proportion of nocturnal observations than cameras positioned specifically for monitoring wolves. As many cameras mounted for monitoring wolves were placed where wolves marked, they likely captured more occasions where wolves were socially active than cameras positioned to register all species of wildlife. In conclusion, any sampling bias due spatial location of the cameras likely resulted in underestimation of nocturnality in Danish wolves, at least in relation to foraging.

Diel variation patterns and its predictors

Territorial wolves in Denmark followed the same diel activity pattern as has been described in most other wolf populations, i.e. with activity peaks after dusk and before dawn and a 6–10 hour activity dip during the middle of the day (Theuerkauf et al. 2003b, 2007, Eriksen et al. 2011, Ogurtsov et al. 2018, Mori et al. 2020, Frey et al. 2022, Martínez-Abraín et al. 2023, Petridou et al. 2023). On a seasonal basis, the activity cycle was stretched towards longer inactivity periods during summer than winter. This points towards that the diel activity rhythm partly, but not fully, followed daylength, as also illustrated by stronger selection for dark periods the longer the night lasted. We also found that young pups (August–October) appeared to be less nocturnally active than adults, probably reflecting a phase in their life-history before they started following the older pack members.

This study suggests that analyzing predictors of diel activity variation as a resource selection function is an efficient method to identify and tease apart the effects of multiple, inter-correlated predictor variables, representing alternative hypotheses for the ecological drivers that mold diel activity patterns. In our opinion, this approach provides a strong and deductively robust analytical supplement to the widely used practice of estimating and comparing diel activity kernel distribution of a maximum of two samples at a time (Ridout and Linkie 2009, Zimmermann et al. 2016, Meredith and Ridout 2017). The results of the use-availability analysis show that the variation in diel activity could be expressed as a combined function of light conditions (selection for darkness), human activity (avoidance) and high deer activity (selection), and that these three predictors in combination fitted data at least as well as a (circadian) ‘time' model that fitted activity variation as a smoothed function of time of the day. Hence, our results align with the suggestion by Theuerkauf (2009) based on his analysis of between-population variation that wolf diel activity patterns are driven by fear as well as foraging profitability. It is worth noting that human activity provided stronger predictions of wolf activity than did light conditions and even less so, deer activity. This might be taken as an indication that temporal avoidance of humans is the strongest single predictor of timing of activity in wolves in Denmark. The fact that human activity explained about 60% of the selection strength for night versus day supports the widespread notion that nocturnality in wolves (Theuerkauf 2009, Frey et al. 2022, Martínez-Abraín et al. 2023) and other wildlife (Gaynor et al. 2019, Lewis et al. 2021) is largely a strategy to avoid humans.

Even though most of the effect of daylight on timescape selection was contained in the timing of human activity, wolves still selected for night when adjusting for the effect of other variables. The adaptive basis for selection for darkness per se(after correcting for human and prey activity) is unknown but may in principle also reflect an antipredator response to humans that in evolutionary time (before invention of artificial light) would probably have posed a much-reduced threat from encounters under darkness than in daylight. An alternative adaptive explanation would be that wolves would have higher hunting success upon encounters with prey during the night than by day as suggested by Theuerkauf (2009).

Many studies have, to a varying extent, found diel activity periods of wolves to overlap with activity periods of their prey, but without being able to conclude decisively whether the correlation was driven by variation in prey activity or other factors (Theuerkauf et al. 2003b, Eriksen et al. 2011, Mori et al. 2020, Petridou et al. 2023). Positive selection for deer activity, even when accounting for human activity and light conditions in this study, support the biologically plausible suggestion that the diel activity periods of wolves also are governed by the activity periods of their prey. The fact that differential deer activity in fenced and unfenced areas correlated closer with wolf activity than deer activity averaged over all areas, may support the assertion that increased wolf activity during activity periods of their prey is a stimuli-driven response rather than a genetically hard-wired pattern embedded in the circadian rhythm. To conclude convincingly on the extent to which wolf activity patterns positively respond to prey activity and whether this is behaviorally plastic versus genetically determined, it would be necessary to repeat the analysis using data from study areas where there were demonstrably marked differences in diel activity profiles of deer, which were not available for this study.

Variation in nocturnality

As expected at a latitude where night length varied seasonally from 31% to 69% of the diel cycle, the proportion of observation registered at night increased with increasing night length (from 69% at summer solstice to 96% at winter solstice for packs). From this it also followed that nocturnality should be measured in relation to day length, or as minimum sampled equally across the year.

Nocturnal observations varied between 81% (single wolves), 84 % (pairs) and 88% (packs) of all observations at the equinox, equaling an odds-ratio in activity difference between night and day (SRN:D) between 4.1 (single wolves) and 6.5 (packs). On this basis, wolves in Denmark appear to be amongst the most nocturnally active wolves registered so far. In comparison, other camera trap studies have reported 63% (SRN:D = 1.7) and 79% (3.8) nocturnality in a human restricted and human accessible nature area in Spain (Martínez-Abraín et al. 2023), 69% (2.2) in a forest nature reserve in western Russia (Ogurtsov et al. 2018), 73% (2.7) in forests across Croatia (Blašković et al. 2022) and 83% (4.9) in habitats with low human disturbance and 97% (32) in habitats subject to high human disturbance in Greece (Petridou et al. 2023). Based on telemetry studies, Theuerkauf (2009) found nocturnal activity in 11 populations to vary from 48% (0.92: Alaska) to 82% (4.6: Italy, average 68%, 1.7). High nocturnality in Danish wolves supports our predictions for individuals to select nocturnality to avoid human encounters when living in landscapes where spatial avoidance is difficult (Gaynor et al. 2019, Lewis et al. 2021). Contrary to our expectations, wolves from territories with restricted public access were no less nocturnal than wolves from territories with free public access, as reported from Spain (Martínez-Abraín et al. 2023), Poland (Theuerkauf 2003a), Canada (Frey et al. 2022) and Greece (Petridou et al. 2023). A plausible reason could be that in Denmark, even those areas with restricted access fail to provide wolves with sufficient undisturbed habitat for foraging activities during the daytime due to the dense forest road networks and fragmented forest stands in all the surveyed territories. The significantly higher nocturnality than the population mean found in the wolf occupying the totally fenced-off territory of Lille Vildmose, might possibly be sampling-related, because wildlife cameras in all other territories were concentrated in forest habitats in the core areas of their home range, which may have been used more during daytime than at night (Mancinelli et al. 2018), resulting in under-sampling at night compared to the true patterns of activity. A higher propensity to use the peripheral parts of the territory primarily for hunting (resulting in underestimation of nocturnality of individuals responsible for hunting) could also possibly explain higher nocturnality in packs compared to pairs and single wolves, because parents and older pack members may allocate a higher proportion of their activity in the core area when with pups compared to wolves without offspring (Ciucci et al. 1997, Rio-Maior et al. 2018).

Standardized nocturnality as index of temporal human avoidance in wolves

This study adds to the mounting evidence that nocturnality is a powerful indicator of investment in temporal human avoidance in wolves (Theuerkauf 2009, Frey et al. 2022, Martínez-Abraín et al. 2023) as well as in wildlife in general (Gaynor et al. 2018, Gilbert et al. 2023). On this basis, we propose that standardized measures of nocturnality obtained from wildlife cameras can be used as a cheap and easily applied measure of temporal human avoidance in space and time. From an analytical perspective, modelling and testing for conditional variation in nocturnality (in this case social status and human access) is far simpler than modelling conditional variation in human avoidance based on use-availability designs, as the first type model enables predictors to be tested as main effects whereas the latter requires them to be tested as interaction terms, which often results in problems with model convergence. Thanks to their extensive use in wolf population monitoring programs, wildlife camera-based data on nocturnality are available from a wide range of populations covering multiple ecological gradients of landscape (human density, forest cover etc.) and human-induced mortality (legal and illegal persecution, traffic). Longitudinal data on nocturnality within populations or territories could be used to test the suggestion by Martínez-Abraín et al. (2023) that natural selection will promote less nocturnality in wolves in protected populations in the future.

We also propose that, as an elaboration of Theuerkauf's (2009) initial nocturnal activity index that was simply the raw proportion of activity during night (on the assumption of equal day and night length in samples), standardizing this proportion explicitly to the equinox based on a logistic regression analysis with night length as predictor provides a convenient comparative measure across populations and their living environments.

Acknowledgements

– Thanks go to our dedicated volunteers, who spend countless of hours on servicing our wildlife cameras. Ruth Morrison Svensson and Tony Fox kindly polished the English text. Tony Fox provided thoughtful comments that strongly improved the final version of the manuscript.

Funding

– The camera trap surveys have been economically supported by the Danish Environmental Protection Agency (National Danish Wolf Monitoring Program: KO, PS), Aage V. Jensen's Naturfond (Lille Vildmose territory: PS, KO), Klelund Aps. (Hovborg territory: PS, KO) and Fonden Frands Christian Frantsens legat (Ulfborg territory: SAK). The analysis and writing of the paper were supported by a donation from 15 Juni Fonden (PS, KO, RMM).

Author contributions

Peter Sunde: Conceptualization (lead); Data curation (supporting); Formal analysis (lead); Funding acquisition (lead); Investigation (lead); Methodology (lead); Project administration (equal); Resources (lead); Validation (equal); Visualization (lead); Writing – original draft (lead); Writing – review and editing (lead). Sofie Amund Kjeldgaard: Conceptualization (supporting); Data curation (equal); Formal analysis (supporting); Investigation (equal); Methodology (equal); Project administration (supporting); Resources (supporting); Software (supporting); Supervision (supporting); Validation (lead); Visualization (supporting); Writing – original draft (supporting); Writing – review and editing (supporting). Rasmus Mohr Mortensen: Formal analysis (equal); Methodology (supporting); Visualization (equal); Writing – original draft (supporting) Kent Olsen Conceptualization (supporting); Data curation (lead); Formal analysis (supporting); Funding acquisition (supporting); Investigation (equal); Methodology (equal); Project administration (equal); Resources (equal); Writing – original draft (equal); Writing – review and editing (equal).

Transparent peer review

The peer review history for this article is available at https://www.webofscience.com/api/gateway/wos/peer-review/wlb3.01251.

Data availability statement

Data are available from the Dryad Digital Repository: https://doi.org/10.5061/dryad.1ns1rn92h (Sunde et al. 2024).