Interpretation according to a climate–vegetation response model suggested that acclimation of forage plants to increasing temperature had resulted in temperature-enhanced plant productivity, initially increasing food availability and supporting transient synchronous increases in population abundance of both blue wildebeest and zebra, and selective grazers. As acclimation of plants to concurrently rising minimum (nocturnal) temperature (T_min) took effect, adjustments in metabolic functionality occurred involving accelerated growth activity at the cost of storage-based metabolism. Growth-linked nitrogen dilution and reduced carbon-nutrient quality of forage then resulted in phases of subsequently declining herbivore populations. Over the long term (1910–2010), progressive plant functionality shifts towards accelerated metabolic growth rather than storage priority occurred in response to T_min rising faster than maximum temperature (T_max), thereby cumulatively compromising the carbon-nutrient quality of forage, a key resource for selective grazers.

The results of analyses thus revealed consistency between herbivore population trends and levels of forage quantity and quality congruent with expected plant metabolic responses to climate effects. Thus, according to the climate-vegetation response model, climate effects were implicated as the ultimate cause of large herbivore population performance in space and over time.

Conservation implications: In its broadest sense, the objective of this study was to contribute towards the enhanced understanding of landscape-scale functioning of savanna systems with regard to the interplay between climate, vegetation and herbivore population dynamics.

The long-term and spatially widespread nature of population trends, however, suggested that over-arching landscape-scale influences may not as yet have been accounted for in currently available studies. A greater role of climate effects as a long-term landscape-scale factor was therefore implicated. In this context, dry-season rainfall and ambient temperature were considered important climate variables. Ambient temperatures have been increasing progressively over the past century (Figure 1), with minimum (nocturnal) temperature (T_min) having increased more steeply than maximum (daytime) temperatures (T_max) (Houghton et al. 1996). Accordingly, in this study we explored to what extent climate–vegetation responses could plausibly explain diverse and divergent spatiotemporal patterns in population performance of eight large herbivore species with divergent forage requirements in a semi-arid savanna system. Preliminary assessments had revealed large herbivore species groups with similar within-group population trends over time. These appeared to be matched by commonalities of habitat preferences (with regard to altitude, underlying geology, northern compared with southern KNP) and features of nutritional ecology [Seydack et al. 2012; for roughage bulk feeders such as buffalo and waterbuck, see Hofmann (1973, 1989); for short-grass preference grazers such as blue wildebeest and zebra, see Bodenstein, Meissner and Van Hoven (2000) and Grange and Duncan (2006); for selectively feeding antelope species such as sable, roan, tsessebe and eland, see Skinner and Chimimba (2005)]. The study involved three components:

• establishing correlations (patterns) between herbivore population performance trends and climate features (rainfall, temperature)

• application of the climate–vegetation response model (Seydack et al. 2012), which defined the expected effects of climate on key indices of forage availability for herbivores (process) towards the interpretation of herbivore population performance patterns in space and over time

• exploring the degree of consistency with which climate-controlled forage quantity and quality can explain correlations between climate and herbivore population trends.

FIGURE 1: Southern hemisphere surface air temperature anomalies between 1850 and 2009.

The park is longitudinally subdivided into granitic substrates forming relatively nutrient-poor sandy soils in the west and more nutrient-rich basalt-derived clay soils in the east (Venter et al. 2003). The savanna vegetation on nutrient-poor substrates tends to be dominated by trees of Combretaceae (e.g. Combretum and Terminalia species) and Caesalpineaceae, whereas those on more nutrient-rich clay soils are dominated by trees of Mimosaceae – especially Acacia species (Gertenbach 1983; Venter et al. 2003). The vegetation of the northern KNP is characteristically dominated by mopane (Colophospermum mopane) in tree or shrub form (Gertenbach 1983; Venter et al. 2003). The herbaceous layer of the KNP is dominated by C₄ grass species (Kennedy, Biggs & Zambatis 2003) and the more nutrient-rich savanna vegetation types on clay soils carry dense stands of nutritious, high-bulk grasses (Venter et al. 2003).

Animal density distribution patterns were smoothed by converting the point-based census data to a 0.5-km grid using kernel densities with a 5-km search radius. The kernel function was based on the quadratic kernel function described in Silverman (1986). Mean altitudes for units of analysis were derived from a digital elevation model (Environmental Systems Research Institute 2006). Long-term rainfall was based on spline interpolation of long-term average rainfall of 23 ranger posts distributed across the park. Mean nutrient status ratings for spatial units of analysis were computed based on geology (Venter et al. 2003) and vegetation indicators of nutrient availability according to the landscape vegetation units of Gertenbach (1983). For the same geological substrate soil nutrient status was further differentiated according to the predominance of the following tree species in order of decreasing nutrient availability indication: fine-leaved Acacia species > Colophospermum mopane (clay soil) > Combretum species (Scholes & Walker 1993; Venter et al. 2003). Weighted means according to proportional spatial representation of soil nutrient availability ratings were calculated for each unit of analysis. The following soil nutrient availability ratings were applied:

• 1: Sandveld communities
• 2: granites with prominence of Combretum species
• 3: granites with prominence of Colophospermum mopane; vegetation associated with Lebombo north and south
• 4: thickets of Sabi and Crocodile Rivers
• 5: vegetation communities on gabbro, shales and amphibolites
• 6: basalts with dominance of Colophospernum mopane
• 7: basalts with prominence of Acacia species.

Multiple regression analyses of population densities on altitude and soil nutrient status were performed for the northern and southern KNP separately, corrected for spatial autocorrelation as required. Three types of regression analysis were applied, namely ordinary least squares, spatial lag model and spatial error model (Anselin 2005). The spatial neighbourhood had to be defined for each of the two spatial regression models: neighbouring polygons of all those sharing sides (rook contingency) or neighbouring polygons sharing a side or only a vertex (queen contingency). The spatial regression decision process based on the Lagrange multiplier test statistic was used to identify which of the models produced the most reliable results (Anselin 2005). Whenever one of the spatial models was selected, owing to the presence of spatial autocorrelation, Akaike information criteria were used to identify whether the rook or queen contingency models gave the best result.

For the analysis of association of climate and herbivore population performance between 1944 and 2003 we reconstructed the temporal population trends for the groups of wildebeest/zebra, selective grazer species, buffalo and waterbuck from historical observations or population estimates (1944–1976) and census data (1977–2003). The temporal population performance trends were interpreted according to the following four categories, with assigned rank variables given in parentheses: decreasing (1); remaining low (2); high or sustained at relatively high levels (3); increasing (4). Spearman rank correlation analyses were performed between these rank variables and T_max, T_min, annual rainfall and mean monthly dry-season rainfall (April–September). Variable means for groups of two consecutive years were used, resulting in n = 30 for the period of analysis spanning 60 years. Analyses were differentiated according to northern and southern KNP in respect of the blue wildebeest and zebra group and performed in respect of three time lag periods (zero-, two- and four-year time lags).

Since the Spearman rank analyses implicated that temporal population trends were linked to climate, we further performed a series of repeated measures ANOVAs to explore space–time (climate) interactions. For this purpose we defined strata formed by grouping neighbouring units of analyses (compartments as described above). These strata, defined with regard to subregions on a north-to-south gradient and altitude groupings, thus contained replicates represented by neighbouring units of sampling (Legendre & McArdle 1997). Partly depending on data availability, three separate repeated measures ANOVAs were performed in respect of population density estimates determined for the relevant compartments and species or species groups from aerial survey census data:

• For blue wildebeest and zebra, data were used from the subregions north (150 m a.s.l. – 550 m a.s.l.), central lower (150 m – 350 m), central upper (350 m – 550 m), southern lower (150 m – 350 m) and southern upper (350 m – 650 m), for the years 1980, 1986, 1990, 1993, 2002 and 2008.

• Further analyses of population density data from the northern KNP for the four selective grazer species, blue wildebeest and zebra involved three altitudinal groupings (low, middle and upper altitude) and the years 1977, 1980, 1986, 1990, 1993 and 2005. Because granites and basalts stretched over divergent altitudinal ranges and the distributions of certain species were concentrated on specific substrates, analyses specific for species–substrate combinations were performed: sable antelope on granites, roan and tsessebe on basalts, and blue wildebeest, zebra and eland over both substrates combined (altitudinal ranges for granites or basalts combined: low: 150 m a.s.l. – 300 m a.s.l.; mid-altitude: 300 m – 400 m; upper: 400 m – 550 m).

• Of the selective grazers only sable antelope occurred at sufficient population densities in both the northern and the southern KNP. Analyses were performed for the years 1980, 1986, 1990, 1993 and 2005, and the following spatial units: upper, middle and lower altitudes in the northern KNP, upper and lower altitudes in the central KNP, and upper south-facing as well as north-facing, middle and lower altitudes in the southern KNP.

Subregions along the north-to-south gradient represented areas of increasing rainfall (Venter et al. 2003) and altitudinal groupings have implications for the prevailing temperature regimes according to lapse rates of –0.777 °C.100 m^-1 (T_max) and –0.465 °C.100 m^-1 (T_min), respectively (Schulze 1995). Given the extent of global warming over the past century (Figure 1), these lapse rates imply the equivalent to a lowering of the average altitude of the KNP of between 100 m and 200 m. The analyses described in this section were aimed at revealing patterns in population performance of the relevant herbivore species in respect of climate variables in space (north–south gradient: rainfall; altitude: temperature) and over time (climate data: South African Weather Service), as well as spatiotemporal interactions.

At the plant species level, andropogonoid grass species are physiologically predisposed to cope with pronounced seasonality (Osborne 2008). The associated physiological capacity of species of Andropogoneae for sustained metabolic performance at variable and lower resource levels (dry-season growth) underpins their metabolic scope for storage of metabolites (TNC and minerals). Andropogonoid grass species accordingly tend to produce forage of relatively high carbon-nutrient quality (Seydack et al. 2012).

Plants in relative SMP mode are comparatively stress tolerant in order to cope with resource level fluctuations. As such their metabolism is capable of functioning at relatively low temperature and water availability/optima (SMP-tw functional efficiency). Such metabolic capacity is, however, associated with relatively high storage-to-growth allocation priority and comparatively high maintenance respiration costs (i.e. they are temperature sensitive). In contrast, plants in relative RMP mode are geared to the maximum use of surplus resource conditions, with functionality at relatively high temperature and water resource level optima and high growth-to-storage priority (Seydack et al. 2012). Thus, abundance of resources such as water, nutrients and temperature enhances RMP mode productivity, but increasing productivity (increased RMP-TW functionality) progressively reduces plant tissue quality for herbivores through nutrient dilution (An et al. 2005; Dwyer et al. 2007; ’t Mannetje 1982, 1984; Mutanga et al. 2004; Van Soest, Mertens & Deinum 1978; Wilson 1982; Wilson & Ford 1971), reduction in digestibility (’t Mannetje 1984; Van Soest et al. 1978; Wilson 1982; Wilson & Ford 1973) and reduced allocation to storage (low TNC content; Estiarte & Peñuelas 1999; Herms & Mattson 1992; Van Soest et al. 1978).

Three main groups of process effects are involved in the climate–vegetation response model (CVRM): adjustments in plant metabolic performance (SMP-RMP functionality) settings shaping the spatial distribution of forage resources (CVRM-1), SMP-to-RMP temperature acclimation responses (CVRM-2), and spatiotemporal interactions in these effects (CVRM-3).

According to the climate–vegetation response model both shorter-term transient temperature acclimation pulses (Figure 1: increasing temperature towards 1940; increasing towards 1960; increasing after 1975) and longer-term shifts towards advanced RMP functionality settings are predicted to have occurred in response to temperature trends over the past century (Figure 1; Seydack et al. 2012).

• CVRM-2.1: Pulses of increased forage availability due to increased productivity of quality forage induced by pseudo-cyclic phases of temperature increases (Figure 1: increasing towards 1940; increasing towards 1960; increasing post-1975). In response to rising T_max, transient pulses of SMP productivity (SMP-P temperature acclimation phase state) occur when the temperature optima for plant functionality are still relatively low (tw metabolic activity). However, as acclimation to T_min takes effect, increased RMP-TW metabolic activity of plants results in reductions in forage quality (low metabolic allocation to storage and increased nitrogen dilution) and thus forage availability. The complete SMP–RMP–SMP temperature acclimation sequence involves four temperature acclimation phase states: 1: SMP-t (SMP functionality at relatively low temperatures; Np ↓, Nq ↓, CNq ↑↑); 2: SMP-P (enhanced SMP productivity due to rising T_max; Np ↑, Nq ↑, CNq ↑, SMP-tw efficiency ↑); 3: SMP-T (SMP functionality at relatively higher/increased temperatures; Np ↑, Nq ↑, CNq ↓); 4: RMP-T (RMP functionality at high temperatures; Np ↑, Nq ↓, CNq ↓↓, SMP-tw efficiency ↓), reverting to phase states 3: SMP-T (lower altitudes) or 1: SMP-t (upper altitudes) as ambient temperatures decrease again.¹

• CVRM-2.2: Long-term progressive shifts towards relatively advanced RMP functionality settings in response to cumulative acclimation to increasing T_min and declining diurnal temperature ranges. Trends of progressively deteriorating carbon-nutrient quality and increasing nitrogen productivity to nitrogen quality ratios (Np/Nq ↑) over the past century were accordingly anticipated (Figure 1).

These temperature acclimation responses were expected to have resulted in transient pulses of increased forage availability (increased landscape-scale availability of forage items of favourable nitrogen and carbon-nutrient quality), as well as the progressive long-term decline of carbon-nutrient quality of forage and an increasing nitrogen productivity/nitrogen quality ratio associated with advanced RMP functionality settings (Seydack et al. 2012).

• From a situation prevailing before about 1920–1930, where population abundance of selective grazers and bulk feeders was of similar orders of magnitude, the former generally declined progressively over subsequent decades, whereas the latter had generally increased over this time span (Figure 2: following Ludorf 1918; Pienaar 1963).

• A pronounced declining phase occurred for 10 or more years preceding the mid-1950s; involving, particularly, sable antelope, roan antelope, tsessebe, wildebeest and eland. The only grazers not to show this trend were buffalo. Population decreases were more severe in the southern and central KNP (Steyn 1953–1961; Steyn 1954). After this decline, only sable antelope sustained reasonable population numbers south of the Olifants River (Figure 2). Main source habitat centres and prospering populations of selectively feeding antelope species were consistently recorded in respect of higher-lying areas in the western boundary zone and further west of the park [Joubert 1976; Archives (Skukuza) 1902–1945].

• Zebra population numbers surpassed those of wildebeest during 1955–1970 and stabilised at higher levels than the latter, especially in the northern KNP (Joubert 2007a; Steyn 1954; Figure 2).

FIGURE 2: Population estimates for six of the eight species studied to show long–term population trends for (a) selective grazers and (b) bulk feeders in the northern and the southern Kruger National Park.

FIGURE 3: Population numbers of (a) blue wildebeest, (b) zebra, (c) buffalo and (d) waterbuck based on censuses carried out between 1978 and 2006 (northern and southern Kruger National Park).

FIGURE 4: Population numbers of (a) sable antelope, (b) roan antelope, (c) tsessebe and (d) eland based on censuses carried out between 1954 and 2006 (northern and southern Kruger National Park).

FIGURE 5: Distribution of census observation localities in relation to altitude for sable antelope (black dots) and eland (white dots) during (a) 1980/1986 combined and (b) 2005, and for roan antelope (black dots) and tsessebe (white dots) during (c) 1980/1986 combined and (d) 2005.

FIGURE 6: Long-term population trends reconstructed from historical records and extant census data for (A) blue wildebeest and zebra in southern Kruger National Park (bold arrows) and trend deviation for northern populations (broken line arrow), and (B) selectively feeding antelope species: sable, roan, tsessebe and eland (bold arrows; thinner arrows for sable and eland only).

FIGURE 7: Annual means of mean monthly minimum temperatures and maximum-minimum temperature differences (Tmax–Tmin = diurnal temperature range) at (a) Messina, northwest of the Kruger National Park and (b) Pretoriuskop (southern KNP at upper altitudes; South African Weather Service).

FIGURE 8: Population densities of blue wildebeest and zebra (1980–2008: Kruger National Park), differentiated according to subregions of divergent metabolic mode settings (CVRM-1).

FIGURE 9: Population densities of four selective grazers, (a) sable antelope, (b) tsessebe, (c) roan antelope, (d) eland and two large herbivore species selecting for nitrogen–quality patches, (e) blue wildebeest and (f) zebra in the Kruger National Park (1977–2005 or 2008) differentiated over three altitude classes.

FIGURE 10: Population densities of sable antelope in subregions of the Kruger National Park (1980–2005), differentiated into broad classes of altitude.

The crudely synchronous population cycles of blue wildebeest and zebra and the selective grazers (Figure 6), but long-term inverse population trends of these two groups (Figures 2a–b) represented a seemingly paradox result. However, the long-term increase of blue wildebeest and zebra, as nitrogen quality patch selectors, and the concurrent decline of antelope species selecting for carbon-nutrient quality plant parts (Figure 2a–b) were consistent with the predictions of the climate–vegetation response model. According to this model, increasing temperatures over the last eight to nine decades (with pronounced phases of temperature increases during approximately 1920–1940 and after about 1975: Figures 1 and 7) would have progressively enhanced nitrogen productivity and resulted in higher availability of nitrogen quality forage in the landscape for blue wildebeest and zebra owing to the predominance of the SMP-T phase state. (CVRM-2.1: nitrogen not diluted by high levels of storage-based carbohydrates, which is postulated to have applied before 1940 when the SMP-t acclimation phase state was expected to have predominated.) Concurrently, however, congruent with predictions of the climate–vegetation response model, acclimation of plants to progressively rising T_min (nocturnal warming) since the 1910s (Figure 1) resulted in cumulative shifts of SMP-to-RMP functionality settings (CVRM-2.2). The associated metabolic allocation adjustments from storage towards growth priority were associated with reductions in the storage-based carbon-nutrient quality of forage in the landscape. Accordingly, the landscape-scale availability of carbon-nutrient quality forage, a key resource for selective grazers, would have declined (ongoing SMP-to-RMP temperature acclimation: CVRM-2.2), resulting in the long-term decline of their population numbers in the KNP.

According to the climate-vegetation response model forage quality is determined by the position of plant functionality on the SMP–RMP metabolic performance mode gradient (spatially: metabolic mode settings: CVRM-1; temporally: temperature acclimation sequences: CVRM-2). The SMP mode permits plant functionality at relatively low temperature and water availability levels, which supports storage-based nutrient accumulation and growth-curbed metabolic activity, resulting in comparatively high-quality forage. Plants in the RMP mode perform under conditions of high water and temperature levels, which results in low-quality forage. With the climate–vegetation response model being central to an explanatory framework, we proceeded to apply an approach of pattern/process matching where we explored the extent of consistency with which climate impacts on key nutritional resources (process) could explain spatiotemporal correlations between climate and herbivore population trends (pattern).

Whereas buffalo and waterbuck, grouped as roughage bulk feeders, seemed to be tracking forage quantity largely determined by rainfall (Table 2; Mills et al. 1995; Owen-Smith & Ogutu 2003), temperature and dry-season rainfall were identified as influential in determining the availability of quality forage in the landscape important for selective grazers and those that prefer short grasses. The importance of dry-season rainfall for these species has been identified in previous studies (Dunham et al. 2004; Owen-Smith & Ogutu 2003). However, the effectiveness of dry-season rainfall in determining the availability of quality forage in the landscape appears to be interdependent with the prevailing plant metabolic performance mode settings (spatial scale) and temperature acclimation responses (temporal scale), as controlled spatiotemporally by gradients of water availability and temperature effects.

According to the climate–vegetation response model, prevalence of the SMP mode in the arid northern KNP results in pronounced SMP-to-RMP temperature acclimation sequences (CVRM-2.1) in response to pulses of temperature increases [Figure 7a; broadly synchronous with temperature trends of the southern hemisphere (Figure 1)], thereby inducing accentuated population cycles of short-grass preference grazers (Figures 3a, b and 6) and selective grazers (Figures 4 and 6). Such pronounced temperature responses are expected since SMP mode plant functionality is sensitive to the impact of nocturnal warming on respiration (metabolic performance trade-off relationships) and pronounced phase states of temperature acclimation are associated with disproportionately enhanced (in SMP-P phase state) or reduced (RMP-T phase state) SMP-tw efficiency in responding to dry-season rainfall and producing high-quality forage in general. In the central KNP, with intermediate RMP–SMP mode settings and higher actual dry-season rainfall, a postulated favourable product of SMP-tw efficiency and dry-season rainfall seems to support the relatively sustained population performance of blue wildebeest and zebra. The higher rainfall in the southern KNP underpinned relative RMP mode settings (CVRM-1), which are inherently associated with poorer forage quality (lower carbon-nutrient quality, due to reduced storage allocation, and lowered nitrogen quality, due to dilution by structural carbon resulting from high TW growth priority). Here selective grazers, mainly sable antelope, were sustained only at the highest altitudes and because of reasonable amounts of dry-season rainfall and the availability of carbon-nutrient quality forage procurable from andropogonoid grass species. Extending the emerging pattern to near tropical conditions (RMP-TW metabolic mode settings), it follows that selective grazers can persist under such circumstances only if substantial dry-season rainfall occurs and suitable andropogonoid grass species are commonly available as forage plants. This seems to be the basis for the stable persistence of roan antelope in moist dystrophic West African savannas receiving 800 mm rainfall or more per year (De Bie 1991). Blue wildebeest and zebra do not occur under such conditions where extreme nitrogen dilution of forage items is encountered (RMP-TW functionality).

Whereas blue wildebeest and zebra populations performed poorly at extremes of the rainfall gradient (low rainfall: nitrogen dilution by storage products associated with SMP functionality and nitrogen dilution by structural carbon at higher rainfall: RMP-TW functionality), selective grazer species may cope with such conditions. At the arid extreme, SMP modes of plant functionality support forage with carbon-nutrient quality (given sufficiently low ambient temperatures or high altitude) and browsing then seems to be particularly advantageous (eland; roan antelope in more arid parts of West Africa; De Bie 1991). At the mesic end of the gradient, sufficient amounts of dry-season rainfall and andropogonoid grass species form the basis of their persistence. Blue wildebeest and zebra populations perform well under intermediate SMP–RMP plant functionality (central KNP), where nitrogen quality of forage is favoured by relative SMP-tw efficiency of plants, reasonable amounts of dry-season rainfall, growth activity curbed through moisture stress (SMP-Tw functionality) and nutrient-rich substrates. However, over the long term, populations of selective grazers declined (Figure 2a) because of plants’ progressive acclimation to rising temperatures (since about 1910: Figure 1), which compromised storage-based carbon-nutrient quality of available forage (CVRM-2.2). In contrast, rising ambient temperature, by enhancing nitrogen productivity under conditions of SMP-Tw plant functionality, increased the availability of forage with adequate nitrogen quality, expectedly thereby favouring increased population densities of blue wildebeest and zebra (Figure 2b).

Prey selection for both buffalo and waterbuck was inversely dependent on rainfall averaged over three preceding years (Owen-Smith & Mills 2008a). Both buffalo and waterbuck seemed more vulnerable to malnutrition, and thus to predation, during periods of low rainfall (Owen-Smith & Mills 2006; Owen-Smith 2008; Owen-Smith & Mills 2008a). Despite heavy predation by lions on waterbuck (Mills et al. 1995), the resultant mortality did not seem to be additive and more important than the negative effects of drought, and populations increased during normal years despite heavy predation on the species by lions and other predators (Pienaar 1969). This also appears to be the case for buffalo. As forage resources improved after the 1991–1992 drought, both species showed recovery in abundance (Figures 3c-d). Prey selection on these species is thus considered rather to be a consequence of malnutrition and not the ultimate cause of declines in abundance.

Wildebeest and zebra were found to be more susceptible to predation under conditions of above-average rainfall. This was attributed to prey selection shifting towards these species owing to their susceptibility to predation being increased through the effect of higher vegetation cover facilitating hunting success of lions (Owen-Smith & Mills 2008a). Reduced exposure to predation was apparently associated with increasing abundance of these two species. Also, higher predation pressure on adult wildebeest than on zebra (Owen-Smith & Mills 2008b) was believed to have resulted in higher population numbers for zebra than for wildebeest in recent decades (Joubert 2007b). These patterns were interpreted to support the contention that wildebeest and zebra, the primary prey species selected by lions in the KNP (Owen-Smith & Mills 2008b), were held below the food ceiling through additive predation (Owen-Smith 2008). However, following from the results of this study, some of these patterns can be interpreted alternatively with reference to the dynamics of nitrogen quality of the forage selected by these two species. Accordingly, the reduced population performance during periods of above-average rainfall is attributed to reduced nitrogen quality due to nitrogen dilution associated with high grass productivity. Between 1955 and 1970 the abundance of zebra commenced to exceed that of wildebeest and stabilised at higher population levels. According to our study, this was interpreted as resulting from zebra being tolerant to relatively lower nitrogen quality forage than wildebeest, as such quality was expected to decline with increasing nitrogen productivity and the associated progressively increasing TW growth priority over this time span (climate–vegetation response model).

The provision of numerous artificial water points has been implicated to have benefited common water-dependent species, resulting in the influx of blue wildebeest and zebra into the range of the less water-dependent selective grazer species (Gaylard, Owen-Smith & Redfern 2003). Population declines of selective grazer species were consequently attributed to increased predation owing to the associated build-up of lion numbers in their ranges (Harrington et al. 1999). However, in spite of extensive water provisioning, the populations of blue wildebeest and zebra continued to fluctuate and the prolonged period of relatively low population densities of these species since the early 1990s in the northern KNP (Figures 3a–b) did not result in the recovery of selective grazer populations. Evidence of a marked shift in prey selection by lions towards the selectively feeding antelope species around 1987 was reported by Owen-Smith and Mills (2008a). The 2–3-fold increase in relative prey selection for these species was considered adequate to explain a similar rise in adult mortality observed after 1986 (Owen-Smith & Mason 2005; Owen-Smith & Mills 2008a). This shift in predation towards the selectively feeding antelope species was apparently induced by (1) lowered abundance and hence lower availability of buffalo as prey following the 1982–1983 drought and (2) coupled with reduced susceptibility of wildebeest and zebra over the period of low rainfall that set in during 1987. With all three principal prey species less available, lions apparently turned towards alternative prey species, including the selectively feeding antelope species and waterbuck (Owen-Smith & Mills 2008a). However, although the involvement of predation in the decline of selectively feeding antelope species after 1986 is implicated, it is difficult to decouple the interactive predator–prey dynamics from the extrinsically driven context provided by changing food resource conditions (Owen-Smith et al. 2005; Owen-Smith & Mills 2006). The results of this study support an interpretation of the decline of these antelope species involving an interaction between compromised forage quality and a resulting increased vulnerability to predation as a consequence. As accelerated rising temperatures since about 1980 (Figures 1 and 7) increasingly compromised carbon-nutrient quality of forage selected by selectively feeding antelope species (climate–vegetation response model), their nutritionally linked condition apparently declined and had involved most adults of the populations by 1986/1987. Rapid population declines ensued, with no signs of recovery to date (Figures 4a-d). In contrast, as food resource conditions (rainfall-based forage quantity) improved for waterbuck – a species heavily predated on by lions – their population numbers progressively increased (Figure 3d), whilst the population numbers of the selectively feeding antelope species remained low. According to the interpretation of this study the absence of recovery of these antelope populations is the result of continued low levels of carbon-nutrient quality of available forage due to sustained elevated ambient temperatures (Figures 6 and 7). Although impacts of predation on the small remaining populations of selectively feeding antelope species cannot be ignored (Grant et al. 2002), we conclude that even though proximate impacts of predation are considered of significance, the ultimate cause of population performance trends of these species appears to be related to the climatically compromised carbon-nutrient quality of their forage.

The results of pattern/process matching conducted in this study revealed remarkable consistency between herbivore population trends and key indices of forage quantity and quality, shaped in accordance with expectations of plant responses to climate (climate–vegetation response model). With regard to all eight species studied, their targeting by lion predation through shifting prey selection, as implicated by Owen-Smith and Mills (2008a), apparently coincided with time periods of compromised forage quantity or quality as predicted by climate–vegetation response processes in this study. Population trends of these species were found to be in apparent synchrony with features of a changing climate (rainfall, dry-season rainfall, and maximum and minimum ambient temperatures). Furthermore, some of the herbivore population trends, such as the lack of recovery of populations of the selectively feeding antelope species over recent years but population increases for waterbuck over the same period, cannot be readily explained with reference to predation, but conform to climatically driven forage availability as outlined in this study. We conclude that ultimate causality of the observed population trends of the species studied is linked to landscape-scale climate–vegetation responses controlling the availability of suitable forage (bottom-up regulation). Proximate impacts of predation additionally shaped the basic population dynamics in a complex and interactive manner (top-down impacts).

Anselin, L., 2005, Exploring spatial data with GeoDa^TM: A workbook, University of Illinois, Urbana-Champaign.

Archives, Kruger National Park (Skukuza), Annual reports, Sabi and Shingwitsi game reserves, 1902–1925; Kruger National Park, 1926–1945.

Archives, Skukuza, Warden’s annual reports: Kruger National Park, unpublished reports, 1946–1952.

Bodenstein, V., Meissner, H.H. & Van Hoven, W., 2000, ‘Food selection by Burchell’s zebra and blue wildebeest in the Timbavati area of the Northern Province Lowveld’, South African Journal of Wildlife Research 30(2), 6–72.

Buckland, S.T., Anderson, D.R., Burnham, K.P. & Laake, J.L., 1993, Distance sampling estimating abundance of biological populations, Chapman, London.

De Bie, S., 1991, Wildlife resources of the West African savanna, Wageningen Agricultural University Papers 91–92, Wageningen.

Dewar, R.C., Medlyn, B.E. & McMurtrie, R.E., 1999, ‘Acclimation of the respiration/ photosynthesis ratio to temperature: Insights from a model’, Global Change Biology 5, 615–622. http://dx.doi.org/10.1046/j.1365-2486.1999.00253.x

Dunham, K.M., Robertson, E.F. & Grant, C.C., 2004, ‘Rainfall and the decline of a rare antelope, the tsessebe (Damaliscus lunatus lunatus), in Kruger National Park, South Africa’, Biological Conservation 117, 83–94. http://dx.doi.org/10.1016/S0006-3207(03)00267-2

Dwyer, S.A., Ghannoum, O., Nicotra, A. & Von Caemmerer, S., 2007, ‘High temperature acclimation of C₄ photosynthesis is linked to changes in photosynthetic biochemistry’, Plant, Cell and Environment 30, 53–66. http://dx.doi.org/10.1111/j.1365-3040.2006.01605.x, PMid:17177876

East, R., 1984, ‘Rainfall, soil nutrient status and biomass of large African savanna mammals’, African Journal of Ecology 22, 245–270. http://dx.doi.org/10.1111/j.1365-2028.1984.tb00700.x

Environmental Systems Research Institute (ESRI), 2006, ESRI data and maps: Elevation and image data, ESRI, Redlands, California, USA.

Estiarte, M. & Peñuelas, J., 1999, ‘Excess carbon: The relationship with phenotypical plasticity in storage and defense functions of plants’, Orsis 14, 159–203.

Fritz, H. & Duncan, P., 1994, ‘On the carrying capacity for large ungulates of African savanna ecosystems’, Proceedings of the Royal Society of London B 256, 77–82. http://dx.doi.org/10.1098/rspb.1994.0052, PMid:8008761

Gaylard, A., Owen-Smith, N. & Redfern, J., 2003, ‘Surface water availability: Implications for heterogeneity and ecosystem processes’, in J.T. du Toit, K.H. Rogers & H.C. Biggs (eds.), The Kruger experience: Ecology and management of savanna heterogeneity, pp. 171–188, Island Press, Washington DC.

Gertenbach, W., 1983, ‘Landscapes of the Kruger National Park’, Koedoe 26, 9–121.

Grange, S. & Duncan, P., 2006, ‘Bottom-up and top-down processes in African ungulate communities: Resources and predation acting on the relative abundance of zebra and grazing bovids’, Ecography 29, 899–907. http://dx.doi.org/10.1111/j.2006.0906-7590.04684.x

Grant, C.C., Davidson, T., Funston, P.J. & Pienaar, D.J., 2002, ‘Challenges faced in the conservation of rare antelope: A case study on the northern basalt plains of the Kruger National Park’, Koedoe 45(2), 45–62.

Harrington, R., Owen-Smith, N., Viljoen, P.C., Mason, D.R. & Funston, P.J., 1999, ‘Establishing the causes of the roan antelope decline in the Kruger National Park, South Africa’, Biological Conservation 90, 69–78. http://dx.doi.org/10.1016/S0006-3207(98)00120-7

Heitkönig, I.M.A. & Owen-Smith, N., 1998, ‘Seasonal selection of soil types and grass swards by roan antelope in a South African savanna’, African Journal of Ecology 36, 57–70. http://dx.doi.org/10.1046/j.1365-2028.1998.114-89114.x

Herms, D.A. & Mattson, W.J., 1992, ‘The dilemma of plants: To grow or defend’, Quarterly Review of Biology 67(3), 283–335. http://dx.doi.org/10.1086/417659

Hofmann, R.R., 1973, The ruminant stomach. Stomach structure and feeding habits of East African game ruminants, Vol. II, East African Literature Bureau, Nairobi.

Hofmann, R.R., 1989, ‘Evolutionary steps of ecophysiological adaptation and diversification of ruminants: A comparative view of their digestive systems’, Oecologia 78, 443–457. http://dx.doi.org/10.1007/BF00378733

Houghton, J.T., Meira Filho, L.G., Callander, B.A., Harris, N., Kattenberg A. & Maskell, K., (eds.), 1996, Climate change 1995: The science of climate change, Cambridge University Press, Cambridge.

Joubert, S.C.J., 1976, ‘The population ecology of the roan antelope (Hippotragus equinus equinus) (Desmarest, 1804) in the Kruger National Park’, DSc thesis, Faculty of Science, University of Pretoria.

Joubert, S.C.J. & Bronkhorst, P.H.L., 1977, ‘Some aspects of the history and population ecology of the Tsessebe Damaliscus lunatus lunatus in the Kruger National Park’, Koedoe 20, 125–145.

Joubert, S., 2007a, The Kruger National Park: A history, Vol. I., High Branching, Johannesburg.

Joubert, S., 2007b, The Kruger National Park. A History, Vol. II., High Branching, Johannesburg.

Kennedy, A.D., Biggs, H. & Zambatis, N., 2003, ‘Relationship between grass species richness and ecosystem stability in Kruger National Park, South Africa’, African Journal of Ecology 41, 131–140. http://dx.doi.org/10.1046/j.1365-2028.2003.00391.x

Kruger, J.M., Reilly, B.K. & Whyte, I.J., 2008, ‘Application of distance sampling to estimate population densities of large herbivores in Kruger National Park’, Wildlife Research 35, 371–376. http://dx.doi.org/10.1071/WR07084

Le Roux, E., 2010, ‘Habitat and forage dependency of sable antelope (Hippotragus niger) in the Pretoriuskop region of the Kruger National Park’, MSc thesis, School of Animal, Plant & Environmental Sciences, University of the Witwatersrand.

Legendre, B. & McArdle, B.H., 1997, ‘Comparison of surfaces’, Oceanologica Acta 20, 27–41.

Ludorf, J.F., 1918, Report of the Game Reserve Commission, Government Printer, Pretoria.

Macandza, V.A., Owen-Smith, N. & Cross, P.C., 2004, ‘Forage selection by African buffalo in the late dry season in two landscapes’, South African Journal of Wildlife Research 34(2), 113–121.

Magome, H., Cain, J.W.I., Owen-Smith, N. & Henley, S.R., 2008, ‘Forage selection of sable antelope in Pilanesberg Game Reserve, South Africa’, South African Journal of Wildlife Research 38(1), 35–41. http://dx.doi.org/10.3957/0379-4369-38.1.35

‘t Mannetje, L., 1982, ‘Problems of animal production from tropical pastures’, in J.B. Hacker (ed.), Nutritional limits to animal production from pastures, pp. 67–85, Commonwealth Agricultural Bureaux, Farnham Royal.

‘t Mannetje, L., 1984, ‘Nutritive value of tropical and subtropical pastures, with special reference to protein and energy deficiency in relation to animal production’, in F.M.C. Gilchrist & R.I. Mackie (eds.), Herbivore nutrition in the subtropics and tropics, pp. 56–66, The Science Press, Craighall.

Mduma, S.A.R., Sinclair, A.R.E. & Hilborn, R., 1999, ‘Food regulates the Serengeti wildebeest: a 40-year record’, Journal of Animal Ecology 68, 1101–1122. http://dx.doi.org/10.1046/j.1365-2656.1999.00352.x

Mills, M.G.L., Biggs, H.C. & Whyte, I.J., 1995, ‘The relationship between rainfall, lion predation and population trends in African herbivores’, Wildlife Research 22, 75–88. http://dx.doi.org/10.1071/WR9950075

Mutanga, O., Prins, H.H.T., Skidmore, A.K., Van Wieren, S., Huizing, H., Grant, R. et al., 2004, ‘Explaining grass-nutrient patterns in a savanna rangeland of southern Africa’, Journal of Biogeography 31, 819–829. http://dx.doi.org/10.1111/j.1365-2699.2004.01072.x

Osborne, C.P., 2008, ‘Atmosphere, ecology and evolution: what drove the Miocene expansion of C₄ grasslands?’, Journal of Ecology 96, 35–45. PMid:18784799, PMid:2517376

Owen-Smith, N., 2008, ‘Changing vulnerability to predation related to season and sex in an African ungulate assemblage’, Oikos 114, 602–610. http://dx.doi.org/10.1111/j.0030-1299.2008.16309.x

Owen-Smith, N. & Ogutu, J.O., 2003, ‘Rainfall influences on ungulate population dynamics’, in J.T. du Toit, H.C. Biggs & K.H. Rogers (eds.), The Kruger experience: Ecology and management of savanna heterogeneity, pp. 310–331, Island Press, Washington DC.

Owen-Smith, N. & Mason, D.R., 2005, ‘Comparative changes in adult versus juvenile survival affecting population trends of African ungulates’, Journal of Animal Ecology 74, 762–773. http://dx.doi.org/10.1111/j.1365-2656.2005.00973.x

Owen-Smith, N., Mason, D.R. & Ogutu, J.O., 2005, ‘Correlates of survival rates of ten African ungulate populations: Density, rainfall and predation’, Journal of Animal Ecology 74, 774–778. http://dx.doi.org/10.1111/j.1365-2656.2005.00974.x

Owen-Smith, N. & Mills, M.G.L., 2006, ‘Manifold interactive influences on the population dynamics of a multispecies ungulate assemblage’, Ecological Monographs 76, 73–92. http://dx.doi.org/10.1890/04-1101

Owen-Smith, N. & Mills, M.G.L., 2008a, ‘Shifting prey selection generates contrasting herbivore dynamics within a large-mammal predator-prey web’, Ecology 89(4), 1120–1133. http://dx.doi.org/10.1890/07-0970.1, PMid:18481536

Owen-Smith, N. & Mills, M.G.L., 2008b, ‘Predator-prey size relationships in an African large-mammal food web’, Journal of Animal Ecology 77, 173–183. http://dx.doi.org/10.1111/j.1365-2656.2007.01314.x, PMid:18177336

Parrini, F., 2006, ‘Nutritional and social ecology of the sable antelope in a Magaliesberg Nature Reserve’, PhD thesis, School of Animal, Plant & Environmental Sciences, University of the Witwatersrand.

Pienaar, U de V., 1963, ‘The large mammals of the Kruger National Park: Their distribution and present-day status’, Koedoe 6, 1–37.

Pienaar, U de V., 1969, ‘Predator-prey relationships among the larger mammals of the Kruger National Park’, Koedoe 12, 108–176.

Scholes, R.J. & Walker, B.H., 1993, An African savanna: Synthesis of the Nylsvley study, Cambridge University Press, Cambridge. http://dx.doi.org/10.1017/CBO9780511565472

Schulze, R.E., 1995, Hydrology and agrohydrology: A text to accompany the ACRU 3.00 agro-hydrological modeling system, University of Natal (Department of Agricultural Engineering), Pietermartizburg.

Seydack, A.H., Grant, C.C., Smit, I.P., Vermeulen, W.J., Baard, J. & Zambatis, N., 2012, ‘Climate and vegetation in a semi-arid savanna: development of a climate-vegetation response model linking plant metabolic performance to climate and the effects on forage availability for large herbivores’, Koedoe 54(1), Art. #1046, 12 pages. http://dx.doi.org/10.4102/koedoe.v54i1.1046

Silverman, B.W., 1986, Density estimation for statistics and data analysis, Chapman and Hall, New York.

Sinclair, A.R.E., Dublin, H. & Borner, M., 1985, ‘Population regulation of Serengeti wildebeest: A test of the food hypothesis’, Oecologia 65, 266–268. http://dx.doi.org/10.1007/BF00379227

Skinner, J.D. & Chimimba, C.T., 2005, The mammals of the Southern African subregion, 3rd edn., Cambridge University Press, Cambridge/Cape Town.

Smit, I.P.J., 2011, ‘Resources driving landscape-scale distibution patterns of grazers in an African savanna’, Ecography 34, 67–74. http://dx.doi.org/10.1111/j.1600-0587.2010.06029.x

Smuts, G.L., 1972, ‘Seasonal movements, migration and age determination of Burchell’s zebra (Equus burchelli antiquorum, H. Smith, 1841) in the Kruger National Park’, MSc thesis, Faculty of Science, University of Pretoria.

Steyn, L.B., 1953–1961, ‘Annual reports: Kruger National Park’, unpublished reports, Skukuza, South Africa.

Steyn, L.B., 1954, Report of the supervisor of game censuses in 1954 in the Kruger National Park, South African National Parks, South Africa.

Van Soest, P.H., Mertens, D.R. & Deinum, B., 1978, ‘Preharvest factors influencing quality of conserved forage’, Journal of Animal Science 47, 712–720.

Venter, F.J., Scholes, R.J. & Eckhardt, H.C., 2003, ‘The abiotic template and its associated vegetation patterns’, in J.T. du Toit, K.H. Rogers & H.C. Biggs (eds.), The Kruger experience: Ecology and management of savanna heterogeneity, pp. 83–129, Island Press, Washington DC.

Viljoen, P.C., 1996, Ecological aerial surveys in the Kruger National Park, report 04/1995, Kruger National Park (Scientific Services), Skukuza.

Wan, S., Xia, J., Liu, W. & Niu, S., 2009, ‘Photosynthetic overcompensation under nocturnal warming enhances grassland carbon sequestration’, Ecology 90(10), 2700–2710. http://dx.doi.org/10.1890/08-2026.1, PMid:19886480

Watson, L.H. & Owen-Smith, N., 2000, ‘Diet composition and habitat selection of eland in semi-arid shrubland’, African Journal of Ecology 38, 130–137. http://dx.doi.org/10.1046/j.1365-2028.2000.00229.x

Whyte, I.J., 2007, Rare antelope in the Kruger National Park: Trends, current status, and options for future management, report 03/2007, South African National Parks, Skukuza.

Whyte, I.J. & Joubert, S.C.J., 1988, ‘Blue wildebeest population trends in the Kruger National Park and the effect of fencing’, South African Journal of Wildlife Research 18(3), 78–87.

Wilson, J.R., 1982, ‘Environmental and nutritional factors affecting herbage quality’, in J.B. Hacker (ed.), Nutritional limits to animal production from pastures, pp. 111–131, Commonwealth Agricultural Bureaux, Farnham Royal.

Wilson, J.R. & Ford, C.W., 1971, ‘Temperature influences on the growth, digestibility, and carbohydrate composition of two tropical grasses, Panicum maximum var. trichoglume and Setaria sphacelata, and two cultivars of the temperate grass Lolium perenne’, Australian Journal of Agricultural Research 22, 563–571. http://dx.doi.org/10.1071/AR9710563

Wilson, J.R. & Ford, C.W., 1973, ‘Temperature influences on the in vitro digestibility and soluble carbohydrate accumulation of tropical and temperate grasses’, Australian Journal of Agricultural Research 24, 187–198. http://dx.doi.org/10.1071/AR9730187