Group living is a behavioral characteristic in some mammalian species, but in the Order Carnivora only 10–15% species live in groups, a strategy that entails specific selective pressures (Creel and Macdonald 1995). In social and cooperative species of carnivores, most individuals live in social units (i.e. groups, packs, clans, prides) with complex dynamics that affect parameters such as litter size, pup and adult survival, dispersal and, ultimately, population dynamics; group size can also influence prey choice, kill rates and interactions among conspecifics and with other species. In addition, reproductive output can be determined by group living through reproductive suppression of subordinate individuals (Macdonald and Kays 2005).

Wolf populations are organized in cooperative social units named packs (Mech and Boitani 2003). Packs of wolves are highly territorial and maintain exclusive areas by means of visual, olfactory and acoustic communication. Their basic composition is a breeding pair, the dominant male and female that usually monopolize reproduction, plus their offspring and other adults or sub-adult wolves (Mech 1999). Wolf pack size ranges from two to more than twelve individuals, and varies seasonally from lower in late winter to higher in summer after parturitions (Mech and Boitani 2003). It may also vary widely among populations (Fuller et al. 2003), and variation in pack size is one of the determinants of wolf population size (Hayes and Harestad 2000, Fuller et al. 2003, Apollonio et al. 2004), together with prey biomass and territory density (Kittle et al 2015, Mech and Barber-Meyer 2015). Pack size variation and composition affects pup survival and reproductive success (Harrington et al. 1983, Peterson et al. 1984), as well as kill rates, food intake, predation rates and interactions with neighboring groups (Vucetich et al. 2002, Metz et al. 2011).

The cohesiveness of individuals within a pack has been related to the capability of killing large prey, and the outcome of competition with scavengers (Schmidt and Mech 1997, Vucetich et al. 2004). Cohesion among wolves in a pack is affected by the number of individuals, the composition of the group (i.e. age and sex of the members), the season, and also by abiotic factors like snow depth (Fuller 1991, Mech and Boitani 2003). Pack cohesion can influence the estimates of pack size, and thus ultimately of population size (Chapron et al. 2016). Estimates of pack size have been commonly derived from direct observations, or counting the track-sets of packs in snow (e.g. compilation by Fuller et al. 2003), and more recently from non-invasive genetic sampling at rendezvous sites and along tracks (Stenglein et al. 2011, Liberg et al. 2012). Nevertheless, potential factors affecting the counts of individuals in packs as well as variation and uncertainty have not often been described in detail (Fuller et al. 2003, Barber-Meyer and Mech 2015).

A currently isolated wolf population inhabits humanized landscapes of the Iberian Peninsula, in Spain and Portugal (Chapron et al. 2014). The population, which includes about 350 wolf packs (MAGRAMA 2016, Torres and Fonseca 2016), is strictly protected in Portugal (listed as Endangered, Cabral et al. 2005), and is under culling and hunting management in most of its Spanish range (Quevedo et al. 2019). Several authors have discussed pack size in this population (Barrientos 2000, Llaneza et al. 2009, Fernández-Gil et al. 2010, Blanco and Cortés 2012, Fernández-Gil 2013) but evaluation of determinants and variability of pack size are generally lacking. Our goals were estimating average pack size and evaluating its seasonal and long-term variation by using direct observations and track sets, while exploring factors that could affect those estimates.

Figure 1.

Localities included in the study. Winter data (blue dots) were collected in the Cantabrian Range (n = 42 localities), while summer data (red dots) were collected in the Duero Plateau and Montes de León mountain range (n = 22 localities). Grey shading shows a 10 × 10 km grid covering the approximate distribution of the Iberian wolf population in Spain and Portugal (sensu Chapron et al. 2014).

Material and methods

Winter data

We collected wolf observations from 2000 to 2018 in winter, between November and April, i.e. between the abandonment of rendezvous-sites (the places where pups are fed until they can travel with the rest of the pack) and the births of the following breeding season.

We collected data on winter travelling packs (sensu Messier 1985, Ballard et al. 1995), i.e. the number of wolves travelling together (Schmidt et al. 2008), a metric commonly used in demographic studies of wolves (Fuller et al. 2003). We searched for track sets in the snow and used spotting scopes at dawn to look for wolves from vantage points. We focused on areas regularly used by wolves, mostly ridges, mountain passes and unpaved roads. We retained for analyses observations that met the following criteria: 1) direct observation of two or more travelling wolves lasting at least 10 min; 2) track sets on snow of at least 1 km including at least two wolves (i.e. the minimum number of wolves in a pack; Fuller et al. 2003, Ausband et al. 2014). Criteria to determine the number of individuals through direct observations and snow tracks were thus roughly equivalent in terms of traveled distances (Frame et al. 2004). We discarded for example observations of lone wolves, unclear or short track sets or observations of resting packs. All winter data were collected by either one of two observers: any locality could be searched by any of both, but at different dates. Each record included the number of individuals, location, date, observer and method (direct observation versus snow tracking). We used vantage points consistently and searched for snow track sets at specific locations; thus for a given combination of pack and year we assumed that data obtained at a specific location and year belonged to the same pack.

Data on snow cover were gathered from the NASA National Snow and Ice Data Center (Hall and Riggs 2016a, b). Snow cover was measured at each pack location and closest available date (i.e. maximum of eight days of difference) as the proportion of terrain covered by snow in 500 m cells and, when data was missing at that resolution, in 0.05° cells (about 4.3 × 5.5 km). Data on snow depth was not available.

Table 1.

Description of variables considered in generalized linear mixed models (GLMMs).

Variables and data analyses

We hypothesized that some methodological factors can affect the counts of individuals in a pack, e.g. counting method (two classes in winter) and observer; estimates can also be affected by abiotic factors (snow) that influence cohesion among individuals in a pack. We looked for long term trends by using year as a continuous variable along the studied periods in summer and winter. We also expected that the number of wolves in a group would decrease along both seasons because of mortality and dispersion, so we considered the month in which data were gathered as predictor, either as continuous (linear effects) or as a categorical (non-linear effects) variable (Table 1). We fitted generalized linear mixed models (GLMM; Poisson error distribution, log link function) to the number of wolves (n wolves) as response variable in winter, and observer, method, month, year and snow cover as predictors. We excluded from analysis those records lacking reliable data on snow cover. In summer, we fitted separate models to the number of adults and sub-adults, and pups (n ads and n pups, respectively), using month and year as predictors. We entered locality as random factor in all GLMMs. We used Akaike information criteria (AIC) and AIC weight (AICw) for model selection (Burnham and Anderson 2004, Burnham et al. 2011). Models were fitted using R package lme4 (Bates et al. 2015, < www.r-project.org>). We computed an autocorrelation function ‘acf’ in R (2019) with the data of the maximum number of wolves per pack in winter.

Figure 2.

(A) Box plots of the number of wolves per observation in winter (n=253 data), (B) box plot of number of adult/sub-adult wolves and pups per observation in summer at rendezvous sites (n=186 and 150 for adults and pup, respectively).

We considered pack-year as the unit to estimate average pack size: the social unit at a given locality and the year in which data were obtained (Apollonio et al. 2004, Jedrzejewski et al. 2007). In winter, data between November and the following April were assigned to a unique winter season. We estimated average pack size from the largest count in each pack during any given year, following Śmietana and Wajda (1997) and Jedrzejewski et al. (2000). The full data set is available in Supplementary material Appendix 1 Table A1.

Results

Discussion

Our results of average pack size in winter (4.2 ± 1.7) were consistent with most European data, and even with reports along the south-eastern current range of the species in North America, e.g. Ontario and Wisconsin (Table 3). Summer estimates, on the other hand, showed much more variation, probably due at least partly to variability in methods. Winter methods seemed more consistent across studies and populations. Overall, we found wide variability in reporting certainty; several studies did not include the variance of their estimates (Table 3). Our dataset, based on a long series that used consistent methods, allowed reporting the variance of estimates, exploring changes in winter and summer wolf pack size, and evaluating some potential determinants of the counts. Note however that we obtained winter and summer data in relatively disjunct areas, where the staple diet of wolves is in principle different (Cuesta et al. 1991); therefore winter and summer estimates are not strictly comparable. North American data showed wider variation of winter pack size than European data (Table 3), probably reflecting wider variation in management of the studied populations (i.e. variation among years and zones in harvest and culling rates), and perhaps wider variation of prey base (Hayes and Gunson 1995, Fuller et al. 2003). Our winter estimates relied on so-called traveling packs, which have been implicitly regarded as roughly equivalent of actual pack size (Fuller et al. 2003). In winter, variation in pack cohesion may confound estimates of pack size based on traveling packs, especially in large ones (Peterson et al. 1984) that can split temporarily into foraging groups (Jedrzejewski et al. 2002). However, actual pack and traveling pack sizes do not differ significantly in winter, even in large packs (Dale et al. 1995). We expected high pack cohesion during winter in our study areas, as wolves must cope with unpredictable resources (mainly wild prey), both in space and time. Although it has been suggested that abiotic factors such as snow depth can affect cohesion (Fuller 1991), the most relevant factors affecting group cohesion are the same that drive group living in wolves, i.e. resource dispersion and competition with scavengers (Schmidt and Mech 1997, Vucetich et al. 2004, MacNulty et al. 2012, 2014). Usually high-ranking individuals do most kills, which means that few wolves commonly capitalize most predatory events (Sand et al. 2006), except when hunting very large prey (MacNulty et al. 2014). A considerable proportion of individuals in packs are pups from the previous breeding season, which do not usually hunt but have to remain close to the others to access food. Indeed, most wolf populations subsist primarily on prey they hunt, and thus cohesion should be high when the group travels along the territory during the period not occupied in feeding the pups at rendezvous sites. It is conceivable that only where important resources are predictable, at least spatially (e.g. farm offal, which at the same time does not require refined hunting skills), wolf groups may show looser cohesion, even during winter (Boitani 1992).We found that average pack size declined 22% in winter, a decline that may be explained by mortality and dispersal (Fuller et al. 2003). Human-related mortality can cause marked population declines over the winter, yet our study is one of the few that followed intra-seasonal changes in wolf pack size (Mech 1977, Jedrzejewska et al. 1996, which showed lower and higher declines, respectively, compared with our study; see also Table 3). A decline in winter pack size may also be due to temporal separation of the breeding pair from the pack at the onset of the mating season, in late winter. We did find a slightly higher frequency of packs of two wolves in February–April compared to November–January (21%, n = 79 pack-years, and 15%, n = 81 pack-years, respectively; Supplementary material Appendix 1 Table A1), but we do not expect that temporal separation of breeding pairs fully explained the observed late winter decline in pack size. In addition, the overall frequency of winter packs composed by just two wolves in our study (18%, 160 pack-years) was somewhat lower than in other studies (e.g. 25% in Adams et al 2008; 31% in Kittle et al. 2015; metrics estimated from the data reported in both studies).

Table 2.

Candidate models (GLMM, Poisson distribution, log link function, locality entered as random factor) fitted to number of wolves per pack in winter (n wolves n=240), and n ads (n=186) and n pups (n=150) in summer. AIC indicates Akaike information criterion; ΔAIC is the difference between best model (lowest AIC) and candidate models; AICw are AIC weights. We included parameter estimates β and standard errors (SE) for the best winter model.

Figure 3.

Box plot of winter pack size from 2000 to 2018. We used the highest count of each pack and year for the plot (n = 160 packs-year).

Table 3.

Compilation of worldwide estimates of wolf pack size; mean ± SD (* indicates SE). We calculated averages where possible in those studies that did not provided them directly (see column Notes). DO = direct observation; EH = elicited howling; SNT = snow-tracking; Misc = miscellaneous; Q = questionnaires; RT = radio-tracking; TS = tracks; NIG = non-invasive genetic sampling. Column n indicates pack-years.

Continued.

There were fewer estimates of summer wolf pack size in the literature (Table 3), despite that summer observations can provide estimates of reproductive success. Our summer estimates included only metrics from packs that bred successfully, i.e. those that raised pups to late summer. Thus they do not indicate average number of pups per pack because wolf populations include substantial though variable proportion of non-breeding and unsuccessful packs: 15% in protected areas without lethal management (e.g. Denali National Park, Alaska, Mech et al. 1998), and up to 20% in protected areas with some lethal management (Adams et al. 2008). Mitchell et al. (2008) found that smaller packs living in areas with high human-caused mortality rates in the Rocky Mountains of the U.S. had lower probability of raising pups, and there were between 16 and 28% of unsuccessful packs.

Monitoring wolf pack size while clearly reporting methods, sample sizes, season and determinants of seasonal and inter-annual changes, remains an important aspect of population dynamics of this highly social species (Liberg et al. 2012, Chapron et al. 2016). Non-invasive genetic sampling has been recently used to ease estimates of wolf abundance at moderate spatial scales (Marucco et al. 2009, Stransbury et al. 2014), and multiple methods of population monitoring had been recently proposed (Ausband et al. 2014, Jiménez et al. 2016). However, estimating and monitoring the number of wolves that cooperate in a given territory are elusive tasks (Stenglein et al. 2011), which would always benefit from the natural history insights of direct observations and counts of snow track sets in the diverse contexts of wolf populations (Table 3). Particularly so in those situations when more logistically complex or expensive procedures (genetic sampling, radio-telemetry) cannot be used to guide estimates. The importance of using robust estimates of wolf pack size is emphasized by the fact that wolf management in many populations worldwide relies on hunting and culling (Boitani 2003). Moreover, wolf culling programs were implemented in many areas with poor understanding of population dynamics and ecological effects (Gehring et al. 2003, Wallach et al. 2009, Rutledge et al. 2010, Creel et al. 2015), an scenario likewise suggested for the wolf population in Spain (Quevedo et al. 2019).