Interactions among terrestrial carnivores involve a complex interplay of competition, predation, and facilitation via carrion provisioning, and these negative and positive pathways may be closely linked. Here, we developed an integrative framework and synthesized data from 256 studies of intraguild predation, scavenging, kleptoparisitism, and resource availability to examine global patterns of suppression and facilitation. Large carnivores were responsible for one third of mesocarnivore mortality (n = 1,581 individuals), and intraguild mortality rates were superadditive, increasing from 10.6% to 25.5% in systems with two versus three large carnivores. Scavenged ungulates comprised 30% of mesocarnivore diets, with larger mesocarnivores relying most heavily on carrion. Large carnivores provided 1,351 kg of carrion per individual per year to scavengers, and this subsidy decreased at higher latitudes. However, reliance on carrion by mesocarnivores remained high, and abundance correlations among sympatric carnivores were more negative in these stressful, high-latitude systems. Carrion provisioning by large carnivores may therefore enhance suppression rather than benefiting mesocarnivores. These findings highlight the synergistic effects of scavenging and predation risk in structuring carnivore communities, suggesting the ecosystem service of mesocarnivore suppression provided by large carnivores is strong and not easily replaced by humans.

We searched the Web of Science database for all relevant studies published to date in February 2018 (search terms specified below). We adopted a “snowball” approach, wherein we also reviewed relevant literature cited in studies located from Web of Science literature searches. After completing this process, we conducted a secondary search in Google Scholar and examined the first 200 papers (sorted by relevance) to determine whether additional papers with useable data were identified. This secondary search did not return additional useable papers. When publications reported findings from more than one study area or from distinct time periods with different carnivore communities, we recorded these as separate studies. Conversely, we pooled multiple publications into a single study in cases when findings from the same area and carnivore community were reported in more than one publication. For each study included in our database, we recorded the continent, latitude and longitude, study area size, season(s) of data collection, years and duration of study, dominant habitat type, and vegetative cover class. Study area sizes reported by authors were recorded when possible. When size was not stated, we referred to study area maps, delineations of reserves, or other boundaries referred to in the study to estimate the study area size. Dominant habitat was classified into one of 11 categories (see Table S1 in Supporting Information). Descriptions of the dominant vegetation were used to classify vegetation as open, closed, or mixed, with “mixed” chosen when the proportion of open and closed habitats appeared to be roughly equal.

We conducted five separate literature searches to obtain data for different analyses:

1. Suppression. Studies reporting mortality and abundance relationships among smaller and larger carnivores were located using the terms “intraguild predation AND carnivor*”, “mortality AND cause AND (telemetry OR collar) AND carnivor*,” and “interspecific killing AND carnivor*.” We restricted studies to regions with carrion-provisioning large carnivores, defined as large-bodied (>15 kg; Hunter 2011) non-ursid carnivores, as ursids (bears) tend to monopolize carcasses until thoroughly depleted (Allen et al. 2015). We restricted telemetry studies to those that collared subordinate carnivores (hereafter, “mesocarnivores”), defined as a carnivore that co-occurs with a larger, dominant carnivore. Each record consisted of a species pair for each study. We recorded the species names, total number of collared mesocarnivores (N_t), number of mortalities caused by each carnivore species (N_c), number of mortalities of unknown cause (N_u), and total number of mortalities (N_m). For each unique study-species pair combination, the proportion of known-caused mortality caused by the larger carnivore (MortProp) was calculated as:

               Eqn. 1

The proportion of individuals with known fate killed by the larger carnivore (mortality rate, MortRate) was calculated as:

                 Eqn. 2

In cases when only data on mortalities was reported, MortRate could not be calculated. The numerators and denominators of each equation were used to construct binomial models of mortality risk (see Statistical Analyses below). Information on the sex and age class (adult, sub-adult, or juvenile) of mesocarnivore was recorded when known. When mortality rates for juveniles were reported separately from adults, they were excluded, as most studies were of adults only. When studies allowed estimation of annual rates of cause-specific mortalities, these rates were also recorded.  

We recorded correlations in abundance among species pairs (AbundCor) along with the sample size when available. We used only metrics that range from -1 to 1, which consisted of Pearson correlation coefficients, Spearman’s rank correlation coefficients, species interaction factors (SIFs), and standardized path coefficients from structural equation models.

To test the carnivore cascade hypothesis, each species in the “suppression” data table was ranked according to its position in the dominance hierarchy of the carnivore guild in the study area. Top carnivores were ranked 1, which consisted of grey wolves, cougars (Puma concolor), Eurasian lynx (Lynx lynx), dingoes (Canis dingo), African lions (Panthera leo), tigers (Panthera tigris), spotted hyenas (Crocuta crocuta), and grizzly bears (Ursus arctos) in our dataset. Mid-size carnivores had a rank of 2 (e.g., coyotes, jackals [Canis spp.], bobcats [Lynx rufus], Canada lynx [L. canadensis]), smaller carnivores had a rank of 3 (e.g., foxes, small wild cats), and occasionally a small carnivore was assigned a rank of 4.  For example, in Alaska wolves and grizzly bears = 1, wolverines (Gulo gulo), coyotes, and Canada lynx = 2, red foxes = 3, and American marten (Martes americana) = 4. Aside from the top carnivores, the ranks of the other species could differ among systems. For example, American marten were ranked as 3 in a study in northeastern Oregon, where foxes and fishers (Pekania pennanti) were absent and the guild consisted of cougars (rank 1), coyotes and bobcats (rank 2), and marten (rank 3). For African large carnivores, lions and spotted hyenas were assigned a rank of 1, and leopards (Panthera pardus), wild dogs (Lycaon pictus), and cheetahs (Acinonyx jubatus) were assigned a rank of 2. Although leopards occasionally kill wild dogs and cheetahs, it was more appropriate to have all three species ranked directly below lions and hyenas for testing the carnivore cascade hypothesis, because wild dogs and cheetahs interact with lions and hyenas far more often than they interact with leopards (e.g., Mills & Gorman 1997; Gorman et al. 1998; Durant 2000a). We calculated the difference in ranks among pairs (RankDiff), and we classified pairs in one of three categories (PairClass): (1) large-meso (rank 1 versus 2), (2) meso-small (rank 2 versus 3), or (3) large-small (rank 1 versus 3). Because of the small sample size of species ranked 4 (n = 2), these cases were excluded from analyses using RankDiff or PairClass.

2. Scavenging. Studies reporting the proportion of mesocarnivore diet comprised of carrion and visitation rates to carcasses were located using the search terms “carrion AND scaveng*,” “mesopredator AND scaveng*,” and “carnivor* AND scaveng*.” As above, we excluded studies from areas that lacked carrion-provisioning large carnivores. We further restricted studies to those in which scavenging by mesocarnivores was confirmed by observation or in cases where the authors provided rationale for considering ungulate remains to be from scavenging rather than predation. For each study of scavenging, we recorded scavenger species, large carnivore(s) present, carrion source (when known), sample size and type, and carrion type. We recorded two metrics of carrion use depending on the study design: (1) CarDiet, the proportion of the diet comprised of carrion (from studies of scat or stomach contents), or (2) CarVisit, the proportion of carcasses visited (from observations at carcass sites).

3. Kleptoparasitism. Studies reporting data on kleptoparasitism among carnivores were located using the search terms “kleptoparasitism AND carnivor*.” We recorded species names for each victim-thief pair, sample size (number of kills monitored), the number of kills stolen, and densities of each species when reported. The proportion of kills kleptoparisitised was calculated as the number stolen divided by the number monitored (KleptRate).

4. Carrion provisioning. To estimate the amount of carrion provided to scavengers by large carnivores, we searched the literature for studies of kill rates. We first used the terms “carnivor* AND kill rate,” and “carnivor* AND predation rate,” but these searches yielded too few relevant results. We modified our search terms to specify the large carnivores present in each major ecosystem or continent (e.g., “wolf OR Canis lupus AND kill rate,” “lion OR Panthera leo AND kill rate”). We recorded species-specific kill rates for each carnivore and prey species pair. The total biomass (kg) of carrion provided by an individual carnivore per year (CarProv) was calculated as:

CarProv=365*k_i*b_i*r                    Eqn. 3

Where k_i is the number of individuals of ungulate prey species i killed per individual carnivore per day, b_i is the body mass of each ungulate species, and r is the proportion of carrion biomass remaining after initial abandonment by the carnivore. The body mass of each prey species was obtained from each study based on the age and sex classes consumed, or from mean adult body masses in the PanTHERIA database if not reported (Jones et al. 2009). The proportion of carrion biomass remaining after abandonment was a constant (0.2788) calculated as the average across 26 studies reporting this metric based on visual observations (95% CI = 0.21-0.34).

            To obtain a rough estimate of carrion provided by large carnivores per unit area for comparison to the biomass of live small prey for mesocarnivores (see below), we recorded all carnivore densities found in examined studies and supplemented with records from PanTHERIA (n = 70 estimates for 12 large carnivore species). We then multiplied CarProv by the average density for each carnivore species to obtain an estimate of carrion provided by each large carnivore species per km² over the course of a year.

5. Small prey biomass. When available, we recorded density (individuals/hectare) of rodents and lagomorphs reported in studies from the above literature searches, as these taxa are the primary year-round resources for most mesocarnivores (Macdonald & Nel 1986; Feldhamer et al. 2007). However, this data was often missing from carnivore studies. We therefore searched for studies of small prey density in regions where information was lacking using targeted searches for specific prey species or taxa based on knowledge of the primary prey bases in each ecosystem (e.g., hare OR Lepus AND abundance OR density). Thus, this search is not considered comprehensive but intended to provide a representation of small mammal biomass in areas throughout the world. We entered small mammal density data for each species (or, species combinations when reported together), season, and year reported in the study. We then calculated average small prey density for each study, summing across species when densities of multiple small mammal species were estimated within a study. Small mammal density estimates were multiplied by the average adult body mass (obtained from the PanTHERIA database) to calculate biomass (kg per km²).

The following files have been uploaded for this dataset. Excel files have a worksheet with the data and a Metadata worksheet that describes each data field.

Appendix S1. Citations of studies included in the meta-analysis.

Table S1. Numbers of studies with each dominant habitat type in the study area, and classification of vegetation as closed, mixed, or open (.docx file)

Table S2. Attributes of studies included in the meta-analysis (.xlsx file)

Table S3. Attributes of species included in the meta-analysis (.xlsx file)

Table S4. Intraguild mortality rates of radiocollared mesocarnivores (.xlsx file)

Table S5. Abundance correlations among pairs of larger and smaller carnivores (.xlsx file)

Table S6. Proportion of mesocarnivore diet comprised of ungulate carrion (.xlsx file)

Tabls S7. Visitation rates of mesocarnivores to ungulate carcasses (.xlsx file)

Table S8. Rates of kleptoparasitism by large carnivores (.xlsx file)

Table S9. Kill rates of large-bodied prey by large carnivores and estimates of carrion biomass provided (.xlsx file)

Table S10. Estimates of small prey (rodent and lagomorph) densities (.xlsx file)