Environmental factors influencing spotted hyena and lion population biomass across Africa

Abstract The spotted hyena (Crocuta crocuta Erxleben) and the lion (Panthera leo Linnaeus) are two of the most abundant and charismatic large mammalian carnivores in Africa and yet both are experiencing declining populations and significant pressures from environmental change. However, with few exceptions, most studies have focused on influences upon spotted hyena and lion populations within individual sites, rather than synthesizing data from multiple locations. This has impeded the identification of over‐arching trends behind the changing biomass of these large predators. Using partial least squares regression models, influences upon population biomass were therefore investigated, focusing upon prey biomass, temperature, precipitation, and vegetation cover. Additionally, as both species are in competition with one other for food, the influence of competition and evidence of environmental partitioning were assessed. Our results indicate that spotted hyena biomass is more strongly influenced by environmental conditions than lion, with larger hyena populations in areas with warmer winters, cooler summers, less drought, and more semi‐open vegetation cover. Competition was found to have a negligible influence upon spotted hyena and lion populations, and environmental partitioning is suggested, with spotted hyena population biomass greater in areas with more semi‐open vegetation cover. Moreover, spotted hyena is most heavily influenced by the availability of medium‐sized prey biomass, whereas lion is influenced more by large size prey biomass. Given the influences identified upon spotted hyena populations in particular, the results of this study could be used to highlight populations potentially at greatest risk of decline, such as in areas with warming summers and increasingly arid conditions.


| INTRODUC TI ON
The spotted hyena (Crocuta crocuta Erxleben) and the lion (Panthera leo Linnaeus) are two of the most widespread and abundant large mammalian carnivores in Africa (Hatton et al., 2015, and references therein). Nevertheless, despite their abundance, the lion's IUCN Red List classification is Vulnerable, with a decreasing population trend, and a population estimate in Africa of close to 20,000 individuals (Bauer et al., 2016). The spotted hyena fares better but its population is also decreasing, with an estimate of 27,000-47,000 individuals (Bohm & Höner, 2015). It is consequently critical to understand the factors influencing the population biomasses of these two species, to quantify better the limitations on their populations, aiding knowledge of drivers of population decline, and informing strategies for conservation. Given the current threat of climate change impacts, the focus of this paper will be on ecological factors, including the abundance of prey, the potential for competition, climate (temperature and precipitation variables), and vegetation cover.
Higher carnivore population density is primarily controlled by high prey biomass (Carbone & Gittleman, 2002) but in general, including the African savannah, predator biomass increases at a lower rate to prey biomass (Hatton et al., 2015). The densities of lion and spotted hyena are positively correlated with prey biomass (Périquet et al., 2015) and prey density (Cooper, 1989), respectively  found that predator population densities were correlated with biomass of preferred prey and biomass of preferred prey weight range (i.e., biomass of prey weighing 190-550 kg in the case of lion; Hayward & Kerley, 2005). Van Orsdol et al. (1985) found that it was prey biomass during the season of lowest food availability that was most important food metric influencing lion population density.
Population size may also be influenced by competition (Carbone & Gittleman, 2002), with a reduction in preferred prey resulting in suboptimal foraging (Hayward & Kerley, 2008), predation, and disease (Kissui & Packer, 2004). Spotted hyena and lion compete with each other, and with other carnivores, for food through both interference competition (direct interactions, Amorós et al., 2020;Kruuk, 1972) and exploitation competition (the use of the same resource by different species, Amorós et al., 2020;Hayward & Kerley, 2008;Périquet et al., 2015). The success of interference competition depends on factors such as the numbers of spotted hyenas relative to the numbers of lions at a carcass, and whether or not an adult male lion is present (Cooper, 1991;Höner et al., 2002). It is therefore important to explore whether competition with other carnivores influences population biomass of spotted hyena and lion.
Competition may be reduced through temporal partitioning (carnivores being active at different times of the day, Hofer, 1998;Mills, 1998;Périquet et al., 2015;Schaller, 1972) or through spatial partitioning, such as different carnivore species occupying different types of vegetation (Schaller, 1972). Targeting of different prey age classes, in addition to frequency of scavenging, also separates carnivore species (Mills, 1990).
As well as prey biomass and competition, other factors may influence lion and spotted hyena populations. For example, Celesia et al. (2010) suggested that rainfall, temperature, and elevation were more important influences upon lion density. Ogutu and Dublin (2002) also found that rainfall influences lion density. The triangle area of the Maasai Mara National Reserve has lower lion density than elsewhere in the reserve. There, the low precipitation in the dry season has an indirect affect due to the resulting low food availability. In the wet season, the area becomes waterlogged, which results in greater disease prevalence (Ogutu & Dublin, 2002).
Precipitation is also important for spotted hyena populations.
Although spotted hyenas may obtain much of their water requirements from fresh carcasses in arid areas, population density is higher in areas of reliable water sources (Cooper, 1989). Examples of arid areas with low population density of spotted hyena include the Kalahari Gemsbok National Park in South Africa (Mills, 1990) and the Tsauchab River Valley in Namibia (Fouche et al., 2020). In addition, Gasaway et al. (1991) suggested that arid conditions may reduce spotted hyena populations if prey is scarce and most of the food comes in the form of desiccated carcasses. Temperature may also be important for spotted hyenas, which are inactive during the hottest parts of the day . More specifically, Cooper (1990) found that spotted hyena individuals were unable to hunt in temperatures above about 20°C.
Given the aforementioned potential influences of biomass and climate upon spotted hyena and lion populations, in addition to the reduction of competition through various mechanisms of niche separation, any changes to prey biomass, temperature, precipitation, or vegetation openness due to climate change or human influence may be a concern for future populations. For example, Wolf and Ripple (2016) demonstrated that declining populations of prey species may increase the vulnerability of carnivore populations, although the lion and spotted hyena were not among the most vulnerable species.
Similarly, Sandom et al. (2017) stated that declining prey is of particular threat to large felids (weighing more than 15 kg), primarily relying upon larger prey. As for changing climatic conditions, an example is the African wild dog (Lycaon pictus). There is concern that switching from diurnal to nocturnal activity in response to hot temperatures by the African wild dog may be insufficient to make up for lost daytime hunting during denning seasons, indicating potential negative impacts of future rise in temperature for this species (Rabaiotti & Woodroffe, 2019).
In this paper, the variation in population biomass density in the spotted hyena and the lion and their corresponding environmental correlates (other predator biomass, prey biomass, temperature, precipitation, and vegetation cover) is explored. With the notable exceptions of Cooper (1989), , Celesia et al. (2010), andPériquet et al. (2015), few studies have attempted this type of synthetic analysis, leaving many aspects of the detailed interactions of spotted hyenas and lions to be explored. This paper therefore takes the novel approach of looking at the influences across spotted hyena and lion populations in Africa, and the breadth of this study means it can provide a much wider overview on factors influencing spotted hyena and lion population decline.

| Sites and data
The influences of environmental variables upon spotted hyena and lion biomass were investigated from 14 published sites across Africa ( Figure 1). From these sites, data were obtained on predator and prey biomass, temperature metrics, precipitation metrics, and vegetation cover.
The predator and prey population biomass density data were obtained from a database in Hatton et al. (2015), who collated animal abundance data from the literature for locations across Africa. Sites were excluded from the present study if spotted hyena were absent, if the abundance of a species was uncertain, if spotted hyena abundance was combined with that of another hyaenid, or if the boundary of the site could not be determined. In total, 30 datasets were included in the biomass analyses from different years spanning 1962 to 2009 ( Figure 1 and Table 1).
In addition to spotted hyena and lion biomass, the biomasses of other large predators were collected and combined for each site.
Large predators are here regarded as those with an adult body mass of over 20 kg. In Africa, there are seven large mammalian predators: spotted hyena, brown hyena (Parahyaena brunnea), striped hyena (Hyaena hyaena), lion, leopard (Panthera pardus), cheetah (Acinonyx jubatus), and African wild dog. However, striped hyena was not included as data for this species are scarce. This is with the exception of the Tarangire National Park, Tanzania, where striped hyena abundance data were provided in lieu of brown hyena abundance (Hatton et al., 2015, and references therein). The striped hyena is solitary and occurs at low densities (Hofer & Mills, 1998), so its exclusion from the present study should not greatly influence the results. Hatton et al.'s (2015) database includes biomasses of potential prey species over 5 kg in weight. Prey were split into five body size classes, following the distinctions of Périquet et al. (2015): very small (<20 kg), small (20-120 kg), medium (120-400 kg), large (400-600 kg), and very large (>600 kg).
Unless otherwise stated in the original publications or by Hatton et al. (2015), the boundaries of the sites were taken to be the entire area, that is, the entire national park, national reserve, game reserve, or district. The Serengeti ecosystem datasets in Hatton et al. (2015) were derived from a number of different publications; therefore, the boundaries of this site were taken from a map of the Serengeti ecosystem (Hopcraft, 2008).
The climate variables used were as follows: maximum temperature of the warmest month, minimum temperature of the coolest month, temperature seasonality (as standard deviation), precipitation of the wettest month, precipitation of the driest month, and precipitation seasonality (as the coefficient of variation). All data F I G U R E 1 Location of sites used in the spotted hyena and lion biomass analyses. 1. Amboseli National Park, Kenya.
are from WorldClim (Hijmans et al., 2005) and were derived from interpolated records of climate data recorded between the years 1950-2000. The variables were taken from the bioclimatic dataset at a resolution of 2.5 min. Each temperature and precipitation value was taken from the center of each site. The center point of each site was the point where the median latitude and longitude intersected.
Median latitude was calculated from the most northerly and southerly latitudes of each location. The same was performed for longitude. This was done using Image Landsat Google Earth Pro (2013).
The vegetation data are taken from the University of Maryland Global Land Cover Classification at 1-km resolution (Hansen et al., 1998(Hansen et al., , 2000 and obtained by the Advanced Very High Resolution  Table 2). The percentage cover of each classification was calculated from the transect counts. Some transects fell over pixels classed as water, cropland, or bare ground. These were excluded from the percentage calculations as it was assumed that spotted hyenas and lions would not be regularly inhabiting these areas.
Full details of the biomass, climate, and vegetation data for each site are included in the https://doi.org/10.5061/dryad.prr4x gxmj.

| Statistical analyses
The relationships between key variables (prey biomass, predator biomass, temperature variables, precipitation variables, and vegetation cover) were analyzed initially to enable an appropriate statistical analyses strategy. In many cases where it is appropriate to test the responses of species population data to environmental influences, where there may be several dependent variables, multiple regression analyses is often used (see Carrascal et al., 2009). Spearman rank order correlations revealed significant correlations between many of the independent environmental variables of interest here (Table A1), meaning that multiple regression analysis could not be applied to determine the relationship between dependent and independent variables, since it cannot accommodate multicollinearity (Carrascal et al., 2009;Mac Nally, 1996. Similarly, the presence of 16 independent variables in this study also prevented the use of hierarchical partitioning (see Olea et al., 2010).
Partial least squares (PLS) regression was therefore chosen. In a comparison of three statistical tests (multiple regression, principal components analysis followed by multiple regression, and PLS), Carrascal et al. (2009) found that PLS performed better under multicollinearity, even with low sample sizes. PLS is also ideal in the current study given that there are 16 independent variables and 30 datasets: PLS is useful when "the number of predictor variables is similar to or higher than the number of observations" (Carrascal et al. (2009, p. 682).
Prior to the assessment of the influence of environmental conditions upon spotted hyena and lion biomass, the biomass, temperature, and precipitation datasets were base-10 logarithmically transformed to reduce skew and to avoid autocorrelation. Some datasets contained values of zero that could not be log transformed.
Where this was the case, the value of zero was converted to a value a unit of magnitude lower than the lowest nonzero value in the dataset. For example, if the lowest value was one, the zero was converted to 0.1, and then base-10 logarithmically transformed.
The vegetation cover data are expressed as percentages and therefore could not simply be logarithmically transformed.
Percentage data suffer from the Unit Sum auto-correlation problem whereby the value of one variable is dependent on the value of the other variables that are used to calculate the percentage

Site
Year (season)

Amboseli National Park, Kenya 2007
Hluhluwe iMfolozi National Park, South Africa 1982 TA B L E 1 Sites from Hatton et al.'s (2015) database included in the spotted hyena and lion biomass analyses (Aitchison, 1982;Pollard et al., 2006). To avoid this, the vegetation data were transformed by the centered log-ratio, following Kucera and Malmgren (1998) and Pollard et al. (2006): where g is the geometric mean of the vegetation category counts for each site, x is the count value of each vegetation category, and d is the number of vegetation categories. The ratio of a vegetation category count and the geometric mean was then calculated and base-10 logarithmic transformed: where clr is the centred log-ratio, and log10 is the base-10 logarithmic transformation.
In the present study, each PLS produced a p-value and r 2 value.
The strength of association of each independent variable with the dependent variable was indicated by the standardized coefficients.
The results were assessed for outliers and leverage points. A site was classed as an outlier if its standardized residual had a value greater or less than two. A site was deemed as a leverage point if its value fell beyond the vertical leverage reference line (LRL), which was calculated by: where m is the number of components in the PLS, and n is the number of observations (Minitab Inc., 2010).
To assess the effect of underlying variation in the data included in the PLS models, each model was re-run 29 times, excluding one which would indicate that there was a consistent relationship between the dependent and independent variable, regardless of which sites were included in the model.
2m n TA B L E 2 Vegetation classes and descriptions from the University of Maryland Global Land Cover Classification at 1-km resolution (Hansen et al., 1998(Hansen et al., , 2000, and classes used in the present study

| RE SULTS
The r 2 and p-values of the PLS regressions for both spotted hyena and lion are summarized in Table 3 (2003). Of these, only Lake Manyara was originally identified as a leverage point for PLS 2b ( Figure A4), again justifying inclusion of all the sites that fell just beyond the LRL.
For re-runs of PLS 1b (spotted hyena as the dependent variable), the confidence intervals of the standardized coefficients are low, ranging from 0.01 for closed vegetation cover, to 0.02 for minimum temperature of the coolest month (Table 4). This indicates that confidence can be placed in the results, as no single site alters the results. Again, this is different for the re-runs of PLS 2b (lion as the dependent variable), which are mostly larger than for PLS 1b. The confidence intervals ranged from 0.01 for very large prey biomass, to 0.19 for temperature of the warmest month (Table 5).
In suggesting that these hold little importance in explaining the variation in spotted hyena biomass. Despite this, all these variables have coefficients from some runs that are close to zero. There is therefore no indication that any variables are consistently and strongly related to lion biomass, contrasting with the many more variables strongly related to spotted hyena biomass in PLS 1b. Additionally, Hatton et al. (2015) noted that the prey abundances recorded from the Kalahari were higher than previous estimates, so there were fewer predators than may have been expected given the prey biomass. This variation in prey abundance may be due to the correlation between prey and rainfall, the latter being unpredictable in the area (Mills, 1990). The extreme leverage value and the potential lag of predator abundance behind prey abundance mean that removing the site and re-running the models (PLS 1b for spotted hyena biomass and PLS 2b for lion biomass) was an appropriate decision. Biomass of medium-sized prey has the strongest overall influence on spotted hyena biomass. Despite spotted hyenas being adaptable in the prey that they target (Hayward, 2006;Mills, 1990), this result is corroborated by known preferred prey weights of 56-182 kg (Hayward & Kerley, 2008), equivalent to small-to mediumsized prey in this study.

| D ISCUSS I ON
Similarly, the positive association between lion biomass and large-sized prey biomass may be explained by the fact that lions most commonly target prey weighing 190-550 kg (Hayward & Kerley, 2005), equivalent to medium-to large-sized prey species in this study. Further, large prey provide more energy intake for large predators, which is necessary to offset energy expended, including that expended while hunting (which is particularly high for predators of large body mass such as lion, Carbone et al., 2007).
The positive association with very small-size prey biomass re- lion predate small-to large-sized prey species, during periods when these species are unavailable, lion will survive on very small-sized prey, namely Thomson's gazelle (Schaller, 1972). The great significance of very small-sized prey species may also reflect their importance in allowing the survival of lion when preferred (larger) prey are unavailable. A further example is seen in the importance of warthog (Phacochoerus africanus, here classed as small-sized prey), in the diet of lions . Further research is required to better understand within-species carnivore abundance patterns in relation to the size and abundance of their prey base (following Carbone et al., 2011;Hatton et al., 2015).
Reliance of very small-size prey may have implications for interspecies competition, particularly lion's and spotted hyena's competition with other carnivores that preferentially target smaller prey (e.g.,

African wild dog, Hayward & Kerley, 2008). This was observed in
the Kafue National Park in Zambia: As larger sized prey populations decreased, smaller sized prey became more important in predator diets, meaning that there was more overlap in the diets of different predators (Creel et al., 2018). This evidence may therefore be used to highlight at-risk populations (not only of spotted hyena and lion, but other predators, too) due to increased competition from prey structure changes that have necessitated reliance upon smaller prey.
The relationship between prey biomass and spotted hyena and lion biomasses agrees with Hatton et al. (2015) in that predator density and biomass are positively correlated. It also agrees with other studies, such as Cooper's (1989) (Fuller & Murray, 1998), and tiger (Panthera tigris) density and prey density in India (Karanth et al., 2004). In addition to prey biomass, PLS 1b and 2b suggest that there are other hitherto undocumented influences upon spotted hyena and lion abundance, explored below.
The minimum temperature of the coolest month has a strong positive relationship with spotted hyena biomass, suggesting that spotted hyena is averse to the very coldest temperatures, that is, spotted hyena populations are greater when winter temperatures are warmer. Although Cooper (1990) and  indicated that spotted hyenas are inactive during the warmest part of the day, this result does not conflict with those studies, as the coolest month temperatures in this study range from 5.4 to 16.8°C (Hijmans et al., 2005). These temperatures are lower than the maximum tolerated temperature for hunting of 20°C (Cooper, 1990).
The maximum temperature of the warmest month has a negative relationship with spotted hyena biomass, although its potential influence is lower than for winter temperatures. This is supported by Cooper (1990)  summer temperatures of sites included in the present study are all above 20°C, ranging from 25.1 to 33.7°C (Hijmans et al., 2005), although spotted hyena may circumvent this to an extent through crepuscular or nocturnal activities (Cooper, 1990; Celesia et al. (2010) found that a different temperature variable (mean annual temperature) is positively correlated with lion density.
Precipitation has some influence upon the spotted hyena, notably adverse effects caused by very dry conditions such as lack of available water bodies. In addition, hot and dry conditions may lead to more rapid desiccation of carcasses, which are themselves important sources of water for spotted hyena, especially in periods of drought (Cooper, 1990;Cooper et al., 1999). This ability to source water from carcasses may be one of the reasons for the limited influence of precipitation.
None of the precipitation variables appear to influence lion biomass, in contrast to Celesia et al.'s (2010) finding that lion density is positively correlated with mean annual rainfall. However, this variable is different to the ones used in the present study (precipitation of the wettest month, precipitation of the driest month, and precipitation seasonality).
Perhaps unexpectedly, open vegetation cover was found to have a strong negative relationship with spotted hyena abundance, while semi-open vegetation cover has a positive relationship. The spotted hyena often hunts by pursuing its prey (Kruuk, 1972;Mills, 1990), so it would seem logical that open conditions should be optimal but this is not the case. Moreover, there appears to be no consistent vegetation preference for den location, with dens having been observed in open grassland (Amboseli Airstrip, Kenya), plains rather than wooded areas (Serengeti and Ngorongoro Crater, Tanzania), savannah and forest patches (Comoé National Park, Côte d'Ivoire), and patches of shrub and isolated trees (Namibia-Naukluft Park, Namibia, Faith, 2007;Henschel et al., 1979;Korb, 2000;Kruuk, 1972;Tilson et al., 1980).  Table 2).
In contrast to the spotted hyena, semi-open vegetation cover is negatively associated with lion biomass. Indeed, even in individual sites, spatial partitioning has been observed between spotted hyenas and lions. For example, in the Serengeti, spotted hyenas occupy the plains and woodland borders while lions occupy the plains, but are most frequently within wooded grassland (Schaller, 1972).
However, this in itself presents a problem as wooded grassland is classed as semi-open vegetation in the present study. Additionally, Périquet et al. (2015) suggested that some vegetation cover is needed to allow lions to ambush their prey.
An alternative explanation for the relationship with vegetation cover may lie in the limitations of the dataset. The data were collected between the years 1981 and 1994 (Hansen et al., 1998(Hansen et al., , 2000, and so any change in vegetation before or after this time period was not recorded. This dataset is nevertheless preferable to obtaining vegetation data from a multitude of sources, as at least the data classification is consistent between sites (Hansen et al., 1998(Hansen et al., , 2000. Small-scale differentiation (below the 1-km resolution of the dataset, Hansen et al., 1998Hansen et al., , 2000 at individual sites may be another explanation.  (Cooper et al., 1999;Höner et al., 2002;Kruuk, 1972;Mills, 1990;Schaller, 1972). Therefore, any negative influence of other predators may be largely cancelled out by spotted hyena and lion succeeding in other competitive interactions. Additionally, as suggested by the aforementioned findings from PLS 1b and 2b, and those of other studies, environmental partitioning (including vegetation, prey preference, and time of activity) may limit the negative impact of other predators upon spotted hyena and lion abundance (Hofer, 1998;Mills, 1990Mills, , 1998Périquet et al., 2015;Schaller, 1972). However, lack of influence of lions upon spotted hyenas is contrary to studies that suggest low lion populations likely lead to greater or more stable spotted hyena populations. For example, Green et al. (2019) observed that in the Talek  Although the five variables discussed above are the only ones that have a consistent positive or negative association with lion biomass, many of the coefficients of these variables are close to zero, depending upon the site removed from the PLS 2b re-runs. The overall lack of consistency between runs suggests that the conditions influencing lion biomass are site-specific or that there are additional influences that were not considered in the analyses. This is backed up by the low r 2 values on some of the PLS runs, which suggest that a large proportion of the variation in lion biomass is not explained by the model. This is in contrast to PLS 1b, which consistently has high r 2 values, associated with spotted hyena biomass.
One potential influence that was not included in the models is disease, which may influence population sizes and may be a factor in PLS 1b, given that there are datasets from Ngorongoro Crater, from five different years. This has been observed in the Ngorongoro Crater, where an outbreak of stable flies (Stomoxys calcitrans) in 1962 (Fosbrooke, 1963, cited in Kissui & Packer, 2004, unknown diseases in 1994 and 1997, and a tick-borne disease and the canine distemper virus (Kissui & Packer, 2004) have all impacted the lion population.
The same area has also witnessed short-term declines in spotted hyena population density through an outbreak of Streptococcus equi ruminatorum in 2002-3 (Höner et al., 2006). This resulted in an increased mortality rate and associated population decline. The disease also became more prevalent with greater interspecific competition and lower prey density, indicating the importance of food availability in influencing the impact of disease (Höner et al., 2012).
A further potential influence is humans. However, it is difficult to quantify human impact in a way that can be included in the model.
Woodroffe (2000) assessed impacts of humans by including densities of people in states, districts, and counties in the study. However, this approach is not suitable in the present study given that the sites are not restricted by political boundaries, but are instead conservation areas. Further, as all sites included in this study have some type of protected status (e.g., national park, conservation area, game reserve; Table 1), this may also influence biomass: Variation in management of protected areas has been found to influence lion population sizes across Africa (Lindsey et al., 2017).  (Elliot & Gopalaswamy, 2016, and further discussed by Braczkowski et al., 2020). This may be of particular concern for species of low population size. Indeed, Mills (1990) acknowledged that nocturnal species in particular, such as the brown hyena, are difficult to count and therefore difficult to determine density and biomass. Ideally, the study would be conducted with modern population estimates of predators and prey from all sites.
Further enhancement would be for the study to be repeated with additional sites, particularly targeting those in western Africa to assess whether the predator populations maintain the same relationship with environmental variables in a wider geographical area.
Given that the Kalahari Gemsbok National Park was removed, there were also only three sites located in southern Africa (although this did account for seven datasets given observations from multiple years at two of these sites). However, the study did analyze the information that is currently available from numerous sites in eastern and southern Africa. Finally, the information gained here, particularly regarding the influences upon spotted hyena biomass could be used to inform conservation efforts. For example, populations that are potentially vulnerable to population decline might be identified, such as in areas with warmer summers and drier conditions, as the model indicates that spotted hyena biomass is negatively influenced by these variables.
In summary, the results indicate that spotted hyena is the more sensitive of the two species to environmental conditions in terms of impacts on biomass. This is surprising given the plasticity of spotted hyena's behavior, such as switching from crepuscular to nocturnal activity, and changing the vegetation that individuals occupied in response to increased human presence in the Talek region of the Maasai Mara National Reserve (Boydston et al., 2003).
The results of the study have potential implications for conservation, particularly of spotted hyena. Any increasing aridity and warmer summer temperatures with climate change are of concern, especially as spotted hyena are inactive during the hottest parts of the day (Cooper, 1990;; increasing summer temperature may limit the time during which they can hunt and thus limit food intake. Further, changes in vegetation, especially a decrease of semi-open vegetation through land management should also be considered as a potential negative threat to spotted hyena populations. Changes in the extent of closed vegetation may also be a factor influencing lion population biomass.

This work was supported by a Natural Environment Research
Council studentship (NE/L002485/1) through the London NERC Doctoral Training Partnership. We are grateful for the detailed comments provided by the Associate Editor and by Professor Matthew Hayward and another anonymous Reviewer.