Territory quality determines social group composition in Ethiopian wolves Canis simensis


  • Lucy A. Tallents,

    Corresponding authorSearch for more papers by this author
  • Deborah A. Randall,

    1. Wildlife Conservation Research Unit, Zoology Department, University of Oxford, Recanati-Kaplan Centre, Tubney House, Abingdon Road, Tubney, Abingdon, Oxon, OX13 5QL, UK
    Search for more papers by this author
  • Stuart D. Williams,

    1. Wildlife Conservation Research Unit, Zoology Department, University of Oxford, Recanati-Kaplan Centre, Tubney House, Abingdon Road, Tubney, Abingdon, Oxon, OX13 5QL, UK
    Search for more papers by this author
  • David W. Macdonald

    1. Wildlife Conservation Research Unit, Zoology Department, University of Oxford, Recanati-Kaplan Centre, Tubney House, Abingdon Road, Tubney, Abingdon, Oxon, OX13 5QL, UK
    Search for more papers by this author

Correspondence author. E-mail: lucy.tallents@linacre.oxon.org


1. We contrast the value of four different models to predict variation in territory size as follows: resource density (the ideal free distribution), population density, group size and intruder pressure (relative resource-holding potential). In the framework of the resource dispersion hypothesis, we test the effect of resource abundance and spatial variation in resource distribution on the age/sex composition of social groups.

2. We explore these drivers of territory size and group size/composition in Ethiopian wolves Canis simensis in the Bale Mountains, Ethiopia, using fine-scale distribution maps of their major prey species based on satellite-derived vegetation maps.

3. The number of adult males is correlated with territory size, while prey density, wolf population density and intruder pressure are not associated with territory size. On average, each additional adult male increases territory size by 1·18 km2.

4. Prey abundance increases with territory size (average biomass accumulation of 6·5 kg km−2), and larger territories provide greater per capita access to prime foraging habitat and prey.

5. The age/sex composition of wolf packs is more closely related to territory quality than territory size. Subordinate adult females are more likely to be present in territories with greater proportions of prime giant molerat Tachyoryctes macrocephalus habitat (i.e. >80% of Web Valley territories and >20% in Sanetti/Morebawa), and more yearlings (aged 12–23 months) occur in territories with greater overall prey biomass.

6. Wolf packs with restricted access to good foraging habitat tend to defend more exclusive territories, having a lower degree of overlap with neighbouring packs.

7. The greater per capita access to prey in large groups suggests a strong evolutionary advantage of collaborative territorial defence in this species, although the relative costs of territorial expansion vs. exclusion depend upon the spatial distribution of resources. We propose a model whereby territory size is determined by the number of adult males, with the presence of subordinate females and yearlings dependent on the quality of habitat, and the abundance and distribution of prey, incorporated within territory boundaries.


Various models have been proposed to explain variation in territory size and group composition among conspecific breeding units. These range from the ecological, such as resource density (the ideal free distribution, Fretwell & Lucas 1969) and dispersion (the resource dispersion hypothesis (RDH), Macdonald 1983), to the social, including constraints such as the resource-holding potential of residents vs. intruders (Stamps 1990). Here, we investigate the power of these models to explain patterns in Ethiopian wolf Canis simensis (Rüppell 1835) territory size and overlap, and the size and age/sex composition of wolf social groups.

If animals operate under an ideal free distribution, all territories will have the same per capita food availability (Fretwell & Lucas 1969) for the primary occupants. At a population and individual level, smaller home ranges are therefore predicted in areas of higher resource abundance, provided the resources are uniformly distributed or accessible (Johnson et al. 2002). This has been demonstrated in many and diverse species such as Townsend’s voles Microtus townsendii (Taitt & Krebs 1981), ovenbirds Seiurus aurocapillarus (Smith & Shugart 1987) and brown bears Ursos arctos (McLoughlin et al. 2000). However, this model may not pertain to situations where territories are contiguous or perennially occupied. Range contraction in response to increasing resources may be undesirable because neighbour pressure may prevent re-expansion or result in escalating conflict at times of food stress (Patterson 1980; von Schantz 1984). In such situations, population density rather than resource abundance may determine territory size through increasing intruder pressure (Wolff 1993). Alternatively, the resource-holding potential of residents relative to intruders (Stamps 1990) may become the prime influence on territory size. Intruders are more likely to be tolerated in ranges where, or at times when, food is abundant, and this is likely to alter the degree of exclusivity of territories. The relationship between population density, resource availability, territory size and overlap is potentially intricate, and may influence, and be influenced by, group size and composition in social species. Thus, we consider in parallel the drivers of variation in social group size and composition.

The minimum defensible territory is the smallest area sufficient to contain the resources necessary for survival and reproduction of the minimum breeding unit (Krebs & Davies 1993) and configured to minimize cost/benefit ratios (Davies & Houston 1984). Social groups may form when benefits accrue from group living through, for example, cooperative hunting (Packer & Ruttan 1988) or communal breeding (Solomon & French 1997). However, these benefits are not necessarily a prerequisite of group formation. In heterogeneous environments, the minimum defensible territory may encompass sufficient resources to give additional animals a degree of food security (Carr & Macdonald 1986), allowing spatial groups (sensuMacdonald 1983) to form. The RDH (Macdonald 1983) posits that while territory size will depend upon the dispersion of foraging patches, group size will be determined by patch richness. Our use of the term ‘patch’ here does not necessitate an arbitrary delineation of patches in an unproductive matrix; the RDH may also be applied to a continuous surface of foraging resources (Blackwell 2007).

Under a contractionist scenario (Kruuk & Macdonald 1985), the presence of additional, non-breeding individuals within a given minimum defensible territory may vary with temporal fluctuations in resource availability. Where resources are more homogeneous, groups may instead adopt an expansionist strategy, relying on greater corporate strength to increase territory size beyond the minimum required by a breeding unit (Kruuk & Macdonald 1985). The spatial extent of foraging habitats could also influence group size if interference competition reduces food intake of subordinates.

Here, we contrast the following factors as determinants of territory size: (i) density of food resources, (ii) population density, (iii) group size and (iv) intruder pressure. Laying the groundwork for our analysis of group composition, we explore how the availability of foraging habitat and food scales with territory size. We test whether the following measures of territory quality can predict group size and composition: (i) total prey abundance in the territory, (ii) the areal extent of different foraging habitats within each territory and (iii) spatial variability in prey abundance. We investigate the benefits of territorial expansion and how territorial overlap varies with resource abundance.

Ethiopian wolves Canis simensis in the highly productive Afroalpine grasslands of the Bale Mountains represent a revealing system to investigate socio-ecological factors influencing group composition and territory configurations. Packs of 2–13 adults (≥2 years old) and yearlings (12–23 months) breed cooperatively and defend a common territory. Where food resources are rich, packs occupy small, stable home ranges that neatly tessellate, covering all available habitat (Sillero-Zubiri & Macdonald 1997), and territory size is correlated with pack size (Sillero-Zubiri & Gottelli 1995b; Marino 2003). In contrast, individuals in resource-poor areas have larger territories and are members of smaller packs. Resources other than prey, such as water, mates and denning sites, are unlikely to be limiting: numerous alpine streams/lakes characterize the high plateaux, suitable denning sites abound, and multi-male groups, floater females, and opportunities for extra-pack copulations (Sillero-Zubiri, Gottelli & Macdonald 1996, Randall et al. 2007) ensure the ready availability of mates.

We test our hypotheses on data from three distinct sub-populations of Ethiopian wolves, from areas that vary in the abundance and distribution of the wolves’ prey. Three rodent species, the giant molerat Tachyoryctes macrocephalus (Rüppell 1842) and murine rodents Arvicanthis blicki (Frick 1914) and Lophuromys melanonyx (Petter 1972), are endemic to the southern highlands of Ethiopia (Yalden & Largen 1992) and comprise 88·8% of the wolves’ diet by volume (Sillero-Zubiri & Gottelli 1995a).

Previous work indicates that Ethiopian wolves adopt an expansionist strategy (Marino 2003) and that relative intruder pressure is a factor in territorial defence. Sillero-Zubiri & Gottelli (1995b) remark that wolf territories expand when neighbouring pack size falls, and larger groups usually win territorial disputes (Sillero-Zubiri & Macdonald 1998). Given that the contribution to territorial defence varies by age/sex class (Sillero-Zubiri & Macdonald 1998) and the costs and benefits of territorial expansion may depend on age and social status, adult males and dominant females are likely to exert the greatest outward pressure with subordinate females/yearlings having less impact. We examine how benefits might accrue to members of larger social groups, promoting the evolution of the expansionist strategy in this species.

Given the site fidelity and foraging behaviour of Ethiopian wolf packs (Sillero-Zubiri & Gottelli 1995b), we predict that this translates into greater tolerance of range overlap where prey densities are high and spatially homogeneous. Packs defending low-quality territories may have to maintain exclusive use of a larger proportion of their range to secure reliable access to sufficient resources. We therefore test whether the degree of overlap between neighbouring territories is higher in areas of abundant, uniformly distributed food.

In addition, we investigate whether shared areas differ in quality from areas that are exploited by only one wolf pack. Two opposing scenarios can be imagined. Areas of range overlap may represent low-quality zones that do not merit the defence effort required to expel intruders. Packs are predicted to invest more in acquiring and defending high-quality areas, maintaining exclusive use of only the best patches. Alternatively, ranges might overlap in the most productive areas, which attract the greatest number of incursions and therefore have the highest defence costs. In the latter case, the benefits of exclusion may also be lower because additional competitors are less likely to depress prey acquisition rates to the extent where it is not worth sharing them. As scant data exist to choose between these alternatives a priori, we simply contrast the productivity of overlap vs. exclusive areas.

Materials and methods

Determining social group size and composition

Twenty-eight wolf packs (nine each in the Web Valley and Sanetti, 10 in Morebawa) were followed for between one and four breeding seasons. Pack compositions were determined through repeated observations of social groups when they congregated during early morning greetings. All calculations disregard wolves below 1 year of age. Population density for each sub-population was the total number of wolves divided by the combined area of all territories, minus overlap areas. Intruder pressure was the mean number of adult males plus dominant females in packs immediately adjacent. Relative intruder pressure was the ratio of neighbour to resident adult males plus dominant females.

Mapping wolf territories

Wolf locations were collected during follows of individuals and social groups and recorded using global positioning system (GPS) units (Garmin 12; Garmin International Inc., Olathe, KS, USA). Our focus on social and patrolling behaviour and opportunistic collection of waypoints at other times resulted in a bias in the distribution of wolf locations. This precludes intra-territorial analysis of ranging behaviour with these data, so here we only map external boundaries. In such cases, strict independence of sequential observations is less critical (Kernohan, Gitzen & Millspaugh 2001), and the removal of autocorrelated points can erode biologically useful information (de Solla, Bonduriansky & Brooks 1999, Börger et al. 2006).

Territory boundaries were determined using Ranges 7 (v0.811; Anatrack Ltd, Wareham, UK). Boundaries were defined as minimum concave polygons (Harvey & Barbour 1965) using an Edge Restriction value of 0·5× the home-range width. This reduced the inclusion of land such as lava promontories, branches of wetland and areas used by adjacent packs into which residents did not venture.

Ethiopian wolves are known to range more broadly in the mating season (Sillero-Zubiri & Gottelli 1995b); therefore, locations for the calendar month during which mating occurred in a given site/season were excluded. The months used to determine territories depended on the timing of breeding; locations were used from the 11 calendar months directly following the month in which most occurrences of mating were observed. Occasional extra-territorial forays were removed by restricting polygons to the most central 98% of observations, using the recalculated arithmetic mean centre option in Ranges 7. This degree of outlier removal was sufficient to remove four forays by single individuals that went well beyond the normal limits of the territory, and further outlier removal was avoided to prevent erosion of areas for which only sparse foraging records existed although the perimeter was clearly patrolled.

Examination of territory size vs. number of locations indicated that a minimum of 100 points collected over the 11 months was required to approach an asymptote in territory size; therefore, territories with fewer than 100 locations were excluded (17 of 69 pack-years). Territory boundaries were determined for 28 packs, for one to four breeding seasons per pack depending on the level of monitoring.

Exclusive areas were defined as the central part of the territory used purely by the occupying pack, which did not overlap with the territories of adjacent packs. This does not preclude the presence of individuals from other packs in these areas but does exclude areas that were used and defended regularly by neighbours. Boundaries were stable over the duration of this study (apart from during the social upheaval caused by the rabies outbreak mentioned later) and neighbouring packs did indeed exploit overlap areas concurrently with residents. The area of entire territories was calculated in Ranges 7, and of overlap/exclusive areas in ArcGIS (v9.0; Environmental Systems Research Institute Inc., Redlands, CA, USA). Adequate home-range data exist to delineate exclusive and overlap areas for five site/year combinations (Web 2002–2003 and 2004–2005, Morebawa 2003–2004, Sanetti 2002–2003 and 2003–2004).

Quantifying habitat and prey distribution in territories

The first step in modelling rodent distributions was to develop a map of Afroalpine vegetation using remotely sensed multi-spectral imagery (Tallents 2007). Unsupervised classification isolated 23 vegetation classes. Quantitative vegetation surveys were used to describe each class, and all classes for which sufficient ground-truthing data existed (n = 17) were statistically distinct in terms of substrate, species composition, height and/or vegetation cover. The majority of classes identified discrete vegetation communities, while some represented mixed pixels at the interface of different vegetation types. This vegetation map was used as a predictor variable, alongside topography and plant productivity, to model the spatial abundance and distribution of the rodents. Vegetation class was a significant predictor of the abundance of all three species.

Prey distribution maps quantified the number of rodents, by species, in each 28·5-m grid cell, and the total biomass of prey present in each grid cell. A separate map was produced for each breeding season, based on known changes in rodent abundance in parallel with seasonal trends in plant productivity (Tallents 2007).

All GIS analyses of habitat and prey abundance were performed in ArcGIS. The area of each habitat type per territory was extracted using Tabulate Area. The total number of prey individuals per territory and total summed biomass per territory were calculated using Zonal Statistics As Table. This function also gave the grid cell maximum, mean and standard deviation (SD) of prey abundance pixel values: the latter indicates the degree of spatial heterogeneity of prey, or how patchily they were distributed.

Categories of habitat quality were created by amalgamating classes with similar prey abundance for each rodent species separately (coded as AB, LM and TM), for the murine rodents together (Mur) and for prey biomass as a whole (Bio), Table 1. For each rodent taxon, the habitat quality class with the lowest (or zero) rodent density (quality class ‘0’) was defined as unproductive, while quality classes ‘1’ through ‘4’ denote increasing rodent densities.

Table 1.   Foraging habitat quality categories, defined by prey abundance. Habitat quality codes in the text and Table 2 are composed of the species indicator (AB) coupled with the density indicator from the column title (3) to give a composite code (AB3). Density varies from negligible or absent (0) to highly abundant (4)
SpeciesMean rodent abundance (rats ha−1)
  1. *Estimates for quality levels 3 and 4 combined.

AB: Arvicanthis blicki 0·1–8·48·5–18·1>18·2*27·9
LM: Lophuromys melanonyx 0·1–13·713·8–26·2>26·3*60·7
Mur: murid rodents00·1–24·925·0–36·6>36·7*72·8
TM: Tachyoryctes macrocephalus<0·90·9–1·81·9–3·6>3·7*12·1
Bio: biomass (kg ha−1)<1·51·6–3·33·4–6·26·3–9·0>9·1

Statistical analysis

When sufficient data were available for more than 1 year of a pack’s history, all values for territory, pack composition and prey attributes were averaged to provide a single summary measure per pack, avoiding pseudoreplication. Therefore, pack (or territory) is the sampling unit in all tests, with a sample size of 28. Stability in wolf pack and territory sizes in the absence of epizootics supports this approach (Sillero-Zubiri & Gottelli 1995b). Two packs (Sodota and Wolla) consisted of a single pair that failed to breed. These packs were clear outliers in several analyses, and we removed them from these analyses to ascertain patterns of resource availability in packs that breed successfully.

Where factors of interest varied among sites, this relationship is described first, to provide context for the hypothesis tests. Relationships between two continuous variables (e.g. territory and pack size) were tested using Spearman’s rank correlations. Mann–Whitney tests were used to compare the effect of categorical predictor variables with two classes (e.g. prey richness of overlap and exclusive areas), and Kruskal–Wallis tests were used to assess the effect of categorical predictor variables with more than two classes (e.g. site). Combinations of categorical and continuous variables (e.g. site and territory size) were tested with general linear models (GLMs) with a Gaussian error distribution. All statistical tests were executed in SPSS (v13; SPSS Inc., Armonk, NY, USA), and figures were created using R (v 2.11.1, R Development Core Team, 2010).


The three Bale Mountains study sites, Web Valley, Morebawa and Sanetti Plateau, are depicted in Fig. 1. Sites varied in the abundance and distribution of the three prey species (Tallents 2007) and in their overall biomass density, Fig. 2. Morebawa and Sanetti had similar levels of prey biomass density, with the lower abundance of A. blicki and T. macrocephalus in high-altitude Sanetti compensated for by greater prevalence of L. melanonyx. The Web Valley represented an exceedingly rich area, with c. 75% greater biomass per square kilometre than the higher altitude sites, Fig. 2d. Giant molerats contributed more to prey biomass in the Web Valley (37·8% vs. 23·5% in Morebawa and 12·9% in Sanetti). Prey biomass was more homogeneously distributed in Sanetti (SD of kg ha−1 pixel values in Web: 2·60; Morebawa: 3·04; Sanetti: 2·16; Fmax59,824 = 1·98, < 0·001).

Figure 1.

 Digital elevation model of the Bale Mountains, showing study sites. Elevations from the Shuttle Radar Topography Mission imagery (Geographic Coordinate System, WGS84) are clipped at 2900 m (black) to delineate the Afroalpine zone. Maximum elevation is 4385 m (white).

Figure 2.

 Rodent densities by study site, based on 2002–2005 NDVI-adjusted predictions (Tallents 2007). Densities are individuals per square kilometre and biomass is kg km−2. Bars show mean ± SD, (a) AB = Arvicanthis blicki, (b) LM = Lophuromys melanonyx, (c) TM = Tachyoryctes macrocephalus and (d) biomass of all species combined. MOR, Morebawa; SAN, Sanetti Plateau; WEB, Web Valley.

Table 2 presents the relationship between site, territory size and the area of each habitat quality class. The three study sites differed in the extent of unproductive habitats (Mur0, TM0 and Bio0; equating to swamps and heather slopes) within territories. Web had the least of all the unproductive habitat categories, while Morebawa had the greatest extent of habitats classed as unsuitable for murine rodents, and Sanetti had the most habitat from which molerats were largely absent. After accounting for the effect of site, all territories had roughly the same area of unproductive habitats regardless of territory size. The area of low-quality habitat in terms of total prey biomass (Bio1) remained constant in all territory sizes, regardless of site.

Table 2.   Site and territory size as predictors of the area of foraging habitat quality class (see Table 1 for definitions)
Habitat qualityGLM ModelTerritory size PCoefficient (% of territory)95% CI of coefficientSite PSite order (decreasing)R2Error d.f.
  1. N = 28 packs. NS, not significant; N/A, not applicable. †P-value for site/territory size interaction, coefficients based on mean value of other main effect.

AB1Territory size + site0·0240·090·01–0·17<0·001San, Mor, Web0·7024
AB2Territory size<0·0010·670·50–0·83NSN/A0·7126
AB34Territory size + site0·0050·210·07–0·35<0·001Web, Mor, San0·4824
LM1Territory size0·0010·380·18–0·59NSN/A0·3326
LM2Territory size<0·0010·370·21–0·54NSN/A0·4226
LM34Territory size + site0·0180·230·04–0·420·006Web, San, Mor0·3024
Mur0SiteNSN/AN/A0·001Mor, San, Web0·3925
Mur1Territory size + site0·0040·070·03–0·12<0·001San, Mor, Web0·5624
Mur2Territory size<0·0010·450·24–0·65NSN/A0·4026
Mur34Territory size<0·0010·360·19–0·53NSN/A0·4126
TM0SiteNSN/AN/A<0·001San, Mor, Web0·6325
TM1Territory size + site0·0120·300·07–0·530·036San, Mor, Web0·4024
TM2Territory size + site + (territory size*site)0·008†0·250·13–0·37N/AMor, Web, San0·7922
TM34Territory size + site + (territory size*site)<0·001†0·200–0·54N/AWeb, Mor, San0·8822
Bio0SiteNSN/AN/A0·001Mor, San, Web0·3825
Bio1SiteNSN/AN/A0·026San, Mor, Web0·1925
Bio2Territory size + site0·0030·400·15–0·65<0·001San, Mor, Web0·5724
Bio3Territory size + site0·0010·350·17–0·530·047Web, Mor, San0·4424
Bio4Territory size + site + (territory size*site)0·012†0·080–0·28N/AWeb, Mor, San0·7622
Bio34Territory size + site + (territory size*site)0·030†0·420·14–0·71N/AWeb, Mor, San0·7322

In general, the proportion of each habitat quality category within a territory varied by site but not by territory size, reflecting the differing productivity of the three sites. Territories in Web had the greatest proportion of Bio34 (Kruskal–Wallis χ2 = 20·1, < 0·001, n = 28) and the lowest proportion of unproductive habitats (Bio0: Kruskal–Wallis χ2 = 17·7, < 0·001, n = 28), while Sanetti had the lowest proportion of good foraging habitats, and Morebawa had the highest proportion of unproductive land.

Mean pack size (adults plus yearlings) was 6·4 ± 2·3 SD (range 2–12, Table 3). The number of adult females varied little, with 13 of 28 packs (46·4%) having two adult females, one pack having three and the remaining 50% each having a single adult female. All other age/sex categories contributed to variation in pack size, and the number of adult males was not significantly correlated with the abundance of other age/sex categories.

Table 3.   Pack size and composition by site, averaged over all pack-years
SitenAFAMYFYMYUPack size
  1. Numbers indicate mean ± SD. Minima and maxima for all sites are indicated in the final row. A, adult; Y, yearling; F, female; M, male; U, unknown.

Web91·3 ± 0·52·9 ± 1·41·0 ± 0·91·6 ± 1·306·8 ± 3·3
Morebawa101·6 ± 0·73·6 ± 1·10·8 ± 1·10·2 ± 0·30·1 ± 0·36·1 ± 1·9
Sanetti91·6 ± 0·53·0 ± 0·71·1 ± 0·90·7 ± 0·806·4 ± 1·8
Overall281·5 ± 0·63·2 ± 1·10·9 ± 0·90·8 ± 1·00 ± 0·26·4 ± 2·3
Range 1–31–50–30–30–12–12

Wolf population densities (adults plus yearlings) were roughly equivalent in Web (1·27 adults plus yearlings km−2 in 2002–2003) and Sanetti (1·23 km−2 in 2003–2004), and lower in Morebawa (0·86 km−2 in 2003–2004). Mean territory size for the two lone breeding pairs in the Web Valley, Sodota and Wolla was 3·6 km2, predicted to contain 18 340 murine rodents and 1684 giant molerats; a total biomass of 2987 kg. This may indicate the minimum defensible territory in optimal habitat.

Determinants of territory size

Of the alternative hypotheses for drivers of territory size (prey density, wolf population density, pack size and intruder pressure), the number of adult males plus the dominant female was the only significant correlate of territory size (Table 4), with larger multi-male groups being associated with larger territories. Territory size increased in parallel with most age/sex categories (Fig. 3), but the trend was steepest for adult males (plus the dominant female). Furthermore, once the influence of adult males had been accounted for, no other age–sex class or combination of classes had a significant impact on territory size.

Table 4.   Correlates of territory size. Spearman’s rank correlations based on 28 packs
Dependent variableIndependent variablesρP
  1. AM, adult male; AF, adult female.

Territory sizeMean prey biomass density−0·1640·404
Wolf population density−0·1880·337
Intruder pressure0·3430·074
Relative intruder pressure−0·2760·155
Total pack size0·3300·086
AMs + dominant AF0·4560·015
Figure 3.

 Territory size is determined by the number of adult males. Figures show territory size against: (a) pack size, (b) number of subordinate females, (c) number of adult males plus the dominant female and (d) number of other wolves in the pack (subordinate females plus yearlings).

Relationship between territory size and quality

Larger territories had greater areas of all habitat classes except the unproductive habitats and Bio1, Table 2. The relationship between territory size and the area of prime foraging habitat (TM34, Bio4 and Bio34) depended on site (interaction term, Table 2). The area of both TM34 and Bio4 per territory was relatively constant in Sanetti and Morebawa, but in the Web Valley larger territories secured proportionately greater areas of the best foraging habitats. In all sites larger territories secured greater areas of Bio34, but the slope was steepest in Web, where almost entire territories consisted of good habitat, while the slope of the relationship was intermediate in Morebawa and shallowest in Sanetti, Fig. 4.

Figure 4.

 Area of the highest quality foraging habitat (Bio34, see Table 1) increases with territory size. Crosses indicate packs in Morebawa; filled circles are Sanetti packs, and open circles are Web Valley packs.

The relationship between territory size and total number of prey varied by prey species and by site, Fig. 5. Larger territories always held more murine rodents: this relationship was site invariant for L. melanonyx (territory size: F1,24 = 130·6, < 0·001, R2 = 0·83, n = 28, Fig. 5b), but the number of A. blicki within a territory increased steeply with territory size in Web, and more gradually in Sanetti (territory size/site interaction: F2,22 = 8·8, P = 0·002, R2 = 0·93, n = 28, Fig. 5a). Giant molerats within a territory remained roughly constant in Sanetti but increased sharply with territory size in Web and at an intermediate rate in Morebawa (territory size/site interaction: F2,22 = 3·6, P = 0·043, R2 = 0·91, n = 28, Fig. 5c). The species-specific patterns resulted in differing rates of prey accumulation by site, with biomass increments per unit area greater in Web than in Morebawa or Sanetti (territory size/site interaction: F2,22 = 8·3, P = 0·002, R2 = 0·91, n = 28, Fig. 5d).

Figure 5.

 Territory size predicts total prey abundance. Crosses indicate packs in Morebawa; filled circles are Sanetti packs, and open circles are Web Valley packs. Figures show: (a) AB = Arvicanthis blicki, (b) LM = Lophuromys melanonyx, (c) TM = T. macrocephalus and (d) biomass.

Determinants of pack composition

The biomass available per wolf at any one time, averaged over all breeding packs, was 729·9 ± 266·3 kg SD (n = 26). Pack size (adults plus yearlings) was more strongly correlated with total rodent biomass in a territory than with territory size (total biomass: Spearman ρ28 = 0·57, < 0·001; territory size: Spearman ρ28 = 0·40, P = 0·037). The correlation between biomass and the number of adult males did not reach significance (Spearman ρ28 = 0·34, P = 0·081).

The relatively constant number of adult females per pack means that there was no significant correlation between number of females and total number of rodents in a territory (Spearman ρ28 = −0·13, P = 0·51). The presence of an additional adult female could only be predicted once the single pack with three adult females (Fulbana in Morebawa) was removed from the dataset. A second adult female was more likely to be found in territories that had a higher proportion of the best giant molerat habitat (TM34), after site differences in habitat had been accounted for (residuals from the GLM of per cent cover of TM34 against site: F1,25 = 6·5, P = 0·017, R2 = 0·17, n = 27, Fig. 6).

Figure 6.

 Prime molerat habitat per territory determines the number of adult females. The proportion of the territory that is prime habitat for giant molerats is contrasted between packs with a single, dominant adult female (empty boxes) and packs with an additional, subordinate adult female (shaded boxes). MOR, Morebawa; SAN, Sanetti Plateau; WEB, Web Valley. Boxes illustrate median and interquartile range, whiskers show full range.

The number of yearlings, however, was positively correlated with rodent abundance (yearling females: ρ28 = 0·48, P = 0·01; yearling males: ρ28 = 0·61, < 0·001). More yearling females were found in territories in which biomass was distributed homogeneously (SD of biomass per territory: Spearman ρ28 = −0·42, P = 0·026). The number of yearling males was positively related to the mean biomass within a territory (i.e. a measure of prey richness that does not change with territory size: Spearman ρ28 = 0·39, P = 0·039). Site was a significant predictor of the number of yearling males in a pack (Kruskal–Wallis χ2 = 7·3, P = 0·026, n = 28) with the most in the Web Valley and the least in Morebawa. No other age/sex categories varied significantly by site.

Benefits of the expansionist strategy

Wolf density within territories decreased as territory size increased (Spearman ρ26 = −0·63, P = 0·001). Sodota and Wolla had small territories but high biomass per wolf because of their small pack size. When these two outliers were removed, biomass per wolf increased with territory size (Spearman ρ26 = 0·53, P = 0·005). For giant molerats, this relationship was only significant after accounting for site differences in molerat abundance (site: F2,22 = 24·2, < 0·001; territory size: F1,22 = 4·8, P = 0·04; R2 = 0·68, n = 26). Sanetti had the fewest molerats per wolf, and Web the most. For murine rodents combined and for total biomass, prey availability per wolf remained similar in each site, with differences between packs explained partly by territory size.

After removing Sodota and Wolla, territories were configured to contain equal amounts of the best foraging habitat (Bio4) per wolf, within the constraints of site differences in its distribution. However, larger territories in all sites had proportionately greater areas of the next best foraging habitat (Bio3) per wolf (Spearman ρ26 = 0·39, P = 0·048).

Influence of habitat quality on territorial overlap

Larger territories tended to have larger exclusive areas (Spearman ρ28 = 0·77, < 0·001), so the percentage area used solely by the resident pack remained relatively constant with territory size (mean = 67·6% ± 20·8 SD, Spearman ρ28 = 0·17, P = 0·39). Territory size and the size of exclusive areas were not significantly different among sites, but the size of overlap areas between neighbouring packs did vary significantly (Kruskal–Wallis χ2 = 6·8, P = 0·033, n = 28, Table 5), with the largest shared areas in Morebawa, and the smallest in Sanetti. In Sanetti and Morebawa, but not in resource-rich Web, the degree of exclusivity was negatively correlated with the proportion of the territory covered by the best foraging ground (Bio34: Spearman ρ18 = −0·66, P = 0·003, Fig. 7).

Table 5.   Territory size by site. Numbers indicate mean ± SD (range), averaged over all years for each pack. Areas of convex polygons are presented for comparison with previous studies.
Territory size (concave polygons, km2)6·43 ± 2·40 (2·96–10·18)8·61 ± 3·21 (3·96–14·39)7·04 ± 1·77 (3·75–8·89)7·41 ± 2·64 (2·96–14·39)
Exclusive area (km2)4·28 ± 2·43 (1·51–9·07)5·38 ± 3·32 (0·65–12·34)5·68 ± 1·65 (3·56–8·12)5·12 ± 2·57 (0·65–12·34)
Proportion exclusive (%)63·5 ± 16·4 (42·4–89·2)59·1 ± 24·1 (16·5–87·9)81·1 ± 15·1 (60·4–100·0)67·6 ± 20·8 (16·5–100·0)
Territory size (convex polygons, km2)7·58 ± 2·71 (3·00–12·18)11·62 ± 4·54 (4·00–19·63)8·38 ± 1·85 (5·63–11·16)9·28 ± 3·65 (3·00–19·63)
Figure 7.

 Territory exclusivity vs. territory quality. Exclusivity is quantified by the proportion of the territory that does not overlap with neighbours, and quality is measured by the proportional extent of the best foraging habitat, Bio34 (see Table 1 for a definition). Crosses indicate packs in Morebawa; filled circles are Sanetti packs, and open circles are Web Valley packs. Ovals indicate the negative trend in Morebawa and Sanetti (continuous line), and lack of a relationship in Web (dashed line).

Areas of territory overlap had significantly more predictable and higher average biomass than did exclusive areas (overlap areas: 6·73 kg ha−1 ± 2·19 SD; exclusive areas: 6·41 kg ha−1 ± 2·35; Wilcoxon Z5 = −2·02, P = 0·043, for both mean and SD of biomass). The trend for overlap areas to harbour a greater proportion of the richest biomass habitats (Bio34) approached significance (overlap areas: 58·2% ± 29·8 SD; exclusive areas: 54·0% ± 30·9; Wilcoxon Z5 = −1·75, P = 0·08).


Determinants of territory size

Ethiopian wolf territory size in this optimal habitat was unrelated to prey density at the level of both populations and packs. The predictions of the ideal free distribution were not upheld: there was no evidence of an inverse relationship between prey density and territory size, and per capita access to prey was not evenly distributed across territories. Likewise, there was no compelling evidence to indicate that territory size was governed by wolf population density or the degree of intruder pressure.

Our analyses suggest that the primary determinant of territory size in this species is the number of adult males plus the dominant female, given her lead role in territorial defence (Sillero-Zubiri & Macdonald 1998). Previous research demonstrated a correlation between pack size and territory size (Sillero-Zubiri & Gottelli 1995b; Sillero-Zubiri et al. 2004), but the results presented here show that it is the number of adult males (plus the dominant female) that determines territory size, not pack size as a whole. The lack of influence of subordinate adult females in this study may partly be an artefact of low variation: previous work that found a positive impact of adult females on territory size included more packs with multiple subordinate females (J. Marino, pers. comm.). Splendid fairy wrens Malurus splendens exhibit a similar pattern, with territory size determined by the number of males in the cooperatively breeding group, and no relationship with the number of females (Brooker & Rowley 1995).

Social group composition vs. territory size and quality

The number of subordinate adult females and yearlings was correlated with the abundance and spatial predictability of foraging habitats and prey. This matches the prediction of the RDH that the number of ‘secondary’ animals in a group will be determined by resource richness (Johnson et al. 2002). The direction of causation is unclear; it may be that yearlings and adult females adjust to resources through higher juvenile survival (Tallents 2007) and lower female dispersal on high-quality territories. Alternatively, an increase in group size may allow packs to secure more foraging habitat and biomass per wolf by capturing higher quality areas. This latter mechanism implies that territory boundaries are very labile in response to changes in pack size, regardless of the stability of neighbouring packs and the stationary nature of optimal foraging patches. Subordinate adult females do actively contribute to territorial defence, although to a lesser extent than adult males or dominant females, and yearlings of both sexes also occasionally participate, although only males urine-mark (Sillero-Zubiri & Macdonald 1998). Therefore, augmentation of patrolling group size by subordinate females and yearlings is likely to help prevent encroachment of neighbouring packs and facilitate expansion. However, contiguous territories have a degree of spatial inertia that would make it difficult to shift boundaries in response to high recruitment. This inertia argues for territory quality to be driving higher juvenile survival and female philopatry, rather than an increase in these age/sex classes allowing the adjustment of boundaries to incorporate higher quality foraging habitat. We infer that territorial expansion by adult males allows for future group augmentation.

The occurrence of additional adult females only in high-quality territories suggests that the dominant female may be more tolerant of a potential reproductive rival when foraging competition is reduced, or alternatively, it may be more difficult for her to control or evict a rival when alternative foraging sites abound. We construe the significant effect of foraging habitat, but not rodent abundance per se, as further support for this tolerance hypothesis, as reduced encounters between adult females could minimize the aggression that may lead to subordinate eviction. These data do not allow us to determine the direction of causation, but as argued earlier, the inertia of territory boundaries suggests that pack composition (via recruitment and dispersal) is probably determined by resources within a territory rather than vice versa. Positive correlations between the number of females and territory quality (number of acacia trees) have also been documented for vervet monkeys Cercopithecus aethiops (Lee & Hauser 1998).

For yearlings, different aspects of prey abundance appear to govern the numbers of males and females. While more of both sexes occur in territories with higher prey productivity, yearling females are also more likely to be found in territories where the distribution of prey biomass is more spatially predictable. A sociological explanation could be that in territories with more clumped resources, and hence greater potential for interference competition or monopolization by dominants, yearling females may have lower survival rates or be more likely to disperse before reproductive maturity. However, foraging interference is infrequent, with dominant animals easily controlling subordinates (Sillero-Zubiri & Gottelli 1995a).

Advantages of expansion

We confirm Marino’s (2003) findings that the expansionist strategy (sensuKruuk & Macdonald 1985) of Ethiopian wolf packs in the Bale Mountains does improve their access to prey. Prime giant molerat habitat represented a greater proportion of larger territories in the Web Valley. As a result, and because territory size increased disproportionately with pack size in all sites, larger territories had greater per capita rodent abundance and areas of prime foraging habitat. Wolves in large packs are therefore at a competitive advantage, although the resources available to an individual will depend upon factors such as dominance rank and foraging ability, as well as social group size. This potential benefit of living in larger packs has implications for the evolution of sociality in this species: strong selective pressure exists to maintain male philopatry. However, there appears to be an upper limit on pack size of 12–13 animals. This indicates that subordinates in very large packs may gain more through dispersal than from remaining to further expand their natal territory. Dispersers may gain better breeding opportunities or decreased intra-pack competition (as presented as a general case by Macdonald & Carr 1989). By securing greater per capita areas of prime habitat, large social groups may also profit from being buffered against stochastic variation in rodent abundance, even if the food intake rate of individuals is not enhanced (Packer 1986; Thurber & Peterson 1993). These differences in territory quality should lead to increased fitness of wolves residing in large territories, some aspects of which are explored in Tallents (2007). In contrast, studies on grey wolves in North America found either no effect (Messier 1985), or a negative influence (Hayes et al. 2000) of pack size on per capita food availability. The different socio-ecology of grey wolves and Ethiopian wolves probably explains this discrepancy. Grey wolves are group hunters that have higher hunting success in smaller packs but share the kill, causing members of larger packs to incur a disproportionately higher cost (Hayes et al. 2000). In contrast, the intake rate of Ethiopian wolves dining unaccompanied is far less likely to be affected by the size of their social group.

Expansion vs. exclusion

Wolves in the Bale Mountains appear to adopt a range of strategies in response to differences in the richness and predictability of their prey, and resultant shifts in the balance of the costs and benefits of expansion. The benefits are greatest in the Web Valley, where access to good foraging habitats increases linearly with territory size, and accessible prey biomass rises at a more rapid rate per unit area than in Morebawa or Sanetti. In Morebawa, larger territories secure more of the intermediate quality habitats, while maintaining a core extent of the best foraging habitats. In contrast, the benefits of expansion are lowest in Sanetti, where fewer of the wolves’ favoured prey, giant molerats, are secured per unit area, and the extent of good foraging habitats does not increase with territory size because prime habitat is distributed more sparsely in a matrix of less productive land. In Morebawa and Sanetti, the degree of exclusivity seems to depend upon territory quality: packs defending areas dominated by good foraging habitats are more tolerant of overlap (or less capable of monopolizing the rich resources), while packs with more restricted access to prey maintain more exclusive use (potentially because of lower intrusion rates). In Web, the degree of overlap may instead be determined by relative intruder pressure, with a tendency for territories to be more exclusive when packs are large relative to their neighbours (L. Tallents, pers. obs.). Placing more emphasis on monopolizing resources may be a more adaptive strategy for wolves on Sanetti, where expansion gives lower, although still apparent, gains.

These results are consistent with a model for food-maximizing, solitary species developed by McLoughlin et al. (2000), which predicts decreasing home-range overlap as habitat quality decreases. The match in predicted and observed patterns of overlap for this food-maximizing group-living species suggests that Ethiopian wolf territoriality is governed by some of the same adaptive pressures. Similar patterns occur in other species: for example, Eide, Jepsen & Prestrud (2004) report lower Arctic fox Alopex lagopus range overlap where prey are sparsely distributed, and female California voles Clethrionomys rufocanus allow increased overlap when given supplemental food, but defend exclusive areas when food is in short supply (Ostfeld 1986; Ims 1987). Messier (1985) documents decreased overlap between grey wolf Canis lupus packs in areas of low moose density.

Areas shared by neighbouring Ethiopian wolf packs are more resource-rich than those monopolized by a single pack, a finding mirrored in black-footed ferrets Mustela nigripes (Jachowski et al. 2010) and black bears Ursus americanus (Horner & Powell 1990). The greater richness of overlap areas probably leads them to be more heavily contested, and therefore harder to monopolize.

A model for Ethiopian wolf territoriality

Uniting the results mentioned earlier, we propose the following conceptual model of territoriality in Ethiopian wolves: (i) territory size is determined by the number of adult males; (ii) the presence of subordinate females is dependent on sufficient amounts of high-quality foraging habitat and (iii) the amount and distribution of prey within the territory determines the survival and dispersal of juveniles. We demonstrate separately that territory quality has an impact on juvenile survival (Tallents 2007). The advantages of male philopatry demonstrated here, coupled with enhanced productivity of packs in high-quality territories, support the evolution of sociality in this species.


We are grateful to the Ethiopian Wildlife Conservation Organisation and the Bale Mountains National Park for permission to work in the BMNP, and to Gedlu Tessera, Mustefa Dule and staff of the EWCP for field data collection. Claudio Sillero, Jorgelina Marino and Paul Johnson gave helpful insights, and LT’s travel was supported by grants from British Airways and Lufthansa. Comments from two anonymous reviewers greatly improved the structure of the manuscript.