Comparative changes in adult vs. juvenile survival affecting population trends of African ungulates

Authors

  • NORMAN OWEN-SMITH,

    Corresponding author
    1. Centre for African Ecology, School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Wits 2050, South Africa
    Search for more papers by this author
  • DARRYL R. MASON

    1. Centre for African Ecology, School of Animal, Plant and Environmental Sciences, University of the Witwatersrand, Wits 2050, South Africa
    2. South African National Parks, Private Bag X402, Skukuza 1350, South Africa
    Search for more papers by this author
    • Present address: Darryl R. Mason, Queensland Parks & Wildlife Service, PO Box 2316, Mount Isa, Qld. 4825, Australia.


Prof. N. Owen-Smith, School of Animal. Plant and Environmental Sciences, University of the Witwatersrand, Wits 2050, South Africa. Fax: +27 11 717 6454; E-mail: norman@gecko.biol.wits.ac.za

Summary

  • 1Among large mammalian herbivores, juvenile survival tends to vary widely and may thus have a greater influence on population dynamics than the relatively constant survival rates typical of adults. However, previous studies yielding stage-specific survival rates have been mostly on temperate zone ungulates and in environments lacking large predators.
  • 2Annual censuses coupled with assessments of population structure enabled annual survival rates to be estimated for the juvenile, yearling and adult segments of nine ungulate species in South Africa's Kruger National Park. Four of these populations persisted at high abundance after initial increases (zebra, wildebeest, impala and giraffe), while five showed progressive declines during the latter part of the study period (kudu, waterbuck, warthog, sable antelope and tsessebe).
  • 3The magnitude of the reduction in adult survival between periods showing contrasting population trends was similar to or greater than the corresponding change in juvenile survival for five of the nine species. Accordingly alterations in population phase, from increasing to stable or stable to declining, were brought about mostly through reduced survival within the adult segment. Elevated predation risk may have been responsible.
  • 4Estimates were derived of the relative survival rates of juveniles, yearlings and adult segments associated with zero population growth, and the survival differential between adult males and females, for all nine species. Stage-specific survival rates appeared dependent on body mass, but with some anomalies. The sex difference in adult survival showed no obvious relation with sexual size dimorphism.
  • 5For large mammalian herbivores, assessments of relative elasticities of stage-specific survival rates on population growth are problematic for several reasons. Sensitivity to corresponding increments in either survival or mortality rates provides a better basis for ecological or adaptive interpretation. Survival rates of adults seem to vary more over multiyear periods compared with mainly annual fluctuations in juvenile survival. More studies are needed on tropical species and in environments retaining large predators to support generalizations about factors influencing ungulate life-history patterns.

Introduction

The relative contributions of different life history stages to population dynamics is of fundamental theoretical importance (Caswell 2001), and can aid causal interpretations of changes in abundance (Varley & Gradwell 1960; Sibly & Smith 1998; Albon et al. 2000). Among large, long-lived mammals, the elasticity of adult survival on population trend is generally substantially greater than that of equivalent relative changes in fecundity or juvenile survival (Heppell, Caswell & Crowder 2000). However, the survival rate of adult ungulates, particularly prime-aged females, generally remains relatively constant, while juvenile survival shows wide annual variability (Gaillard, Festa-Bianchet & Yoccoz 1998; Gaillard et al. 2000). Accordingly, variable juvenile survival could in practice be the ‘key factor’ mostly responsible for fluctuations in population abundance. Gaillard & Yoccoz (2003) suggest that these patterns could be the outcome of evolutionary canalization counteracting temporal variability in the fitness component most influential on population growth, i.e. adult female survival.

The most accurate estimates of stage-specific survival rates are obtained from long-term studies on individually marked animals (Gaillard et al. 1998). However, with a single exception, published studies using this approach have been restricted to temperate-zone ungulates. Many temperate ungulate populations are hunted or harvested, and as a result intrinsically increasing rather than limited by environmental resources. Furthermore, most temperate-zone studies have been conducted in environments lacking large predators. Hence the extent to which the observed demographic patterns are more widely representative of ungulates is uncertain, particularly with respect to the diversity of species inhabiting the African savannas.

Annual counts of large ungulate populations within South Africa's Kruger National Park spanning two decades, supported by surveys of the sex and age structure of these populations for part of this period, enabled survival estimates to be derived for specific population segments through this period. Moreover, while certain species increased and then persisted at high abundance over this period, others declined progressively towards greatly reduced abundance levels (Owen-Smith & Ogutu 2003). A common breakpoint in population trend for most of these species pivotal around 1987 (Piepho & Ogutu 2003) suggested some general influence on their dynamics. Rainfall variability clearly contributed to the pattern, but seemed inadequate alone to explain the contrasting trends manifested by different species (Ogutu & Owen-Smith 2003). The drastic crash by the roan antelope Hippotragus equinus (Desmarest) population in the northern section of the park was associated with a substantial reduction in adult survival, implicating elevated predation by lions Panthera leo L. following an influx of zebra as centrally involved (Harrington et al. 1999). We now examine the demographic mechanisms associated with the altered population trends for a further nine ungulate species. In particular, we seek to establish the relative contributions of variability in adult vs. juvenile survival. Furthermore, we provide estimates of the combination of class-specific survival rates conferring zero population growth for these species in a predator-rich environment. Finally, we assess the implications of these findings for comparative life history analysis of such large, long-lived mammals.

Materials and methods

the source data

The procedure used in the ‘ecological aerial surveys’ conducted annually in the Kruger National Park (KNP) between 1977 and 1996 has been described in detail elsewhere (Viljoen & Retief 1994; Redfern et al. 2002; Owen-Smith & Ogutu 2003), and will merely be summarized here. Adjacent transects by fixed-wing aircraft systematically covered almost the entire 20 000 km2 extent of the KNP between 1980 and 1993, and provided partial regional coverage in other years. All animals seen from the air were noted by four observers, in addition to the pilot and a recorder, with these data entered on to maps in the early years and later into a laptop computer coupled with GPS coordinates. Surveys were undertaken during the mid dry season between May and August when visibility conditions were best. The park perimeter was entirely fenced from 1975 until 1993, precluding dispersal movements beyond park boundaries. Thereafter sections of the western boundary fence adjoining private wildlife reserves were removed, allowing animal movements in both directions, but the effect on the much larger ungulate populations in the KNP is expected to be negligible. Among the nine species considered, Burchell's zebra Equus burchelli (Gray), blue wildebeest Connochaetes taurinus (Burchell), impala Aepyceros melampus (Lichtenstein) and giraffe Giraffa camelopardalis L. persisted at high abundance after 1986 (Ogutu & Owen-Smith 2003). Five species declined after 1986 to less than a third of their peak abundance, including greater kudu Tragelaphus strepsiceros (Pallas), common waterbuck Kobus ellipsiprymnus (Ogilby), warthog Phacochoerus aethiopicus (Pallas), sable antelope Hippotragus niger (Harris) and tsessebe Damaliscus lunatus (Burchell). As an index of the reliability of the count totals, coefficients of variation around linear trends fitted separately to the pre- and post-1986 data ranged from 2 to 3% for zebra and wildebeest to almost 10% for impala and nearly 13% for warthog.

Population structure was recorded annually between 1983 and 1996 by ground surveys over the full extent of the park, conducted between August and October and, except for impala, by one of us (DRM) until 1991 (Mason 1990). Where possible, animals seen were classified as either adult (> 2 years) male, adult (> 2 years) female, yearling (1–2 years) male, yearling (1–2 years) female, or juvenile (< 1 year). Juveniles, mostly aged between 6 and 12 months, were readily distinguished from animals older than a year for the six antelope species and for warthog, all of which have narrow birth seasons. The foaling period of zebra is somewhat spread, and observations after 1991 appeared unreliable in separating juveniles from yearlings, and hence had to be discarded for distinguishing survival in these two segments. For giraffe, calves less than a year old tend to be isolated from their mothers and from herds, and thus undersampled. Accordingly, juvenile recruitment was represented by yearling animals, in accordance with the 1·5–2-year birth interval (Hall-Martin & Skinner 1978). For wildebeest, waterbuck, impala and tsessebe, yearling females could not be distinguished reliably from adult females, hence the yearling total was calculated by doubling the number of yearling males, assuming an even sex ratio, with a corresponding deduction from the adult female class. For tsessebe, yearlings of both sexes could not be distinguished reliably from adults, and so were included in the ‘adult’ category. The relevant information derived from these surveys comprised the annual juvenile/adult female and yearling/adult female ratios, as well as the male/female ratio in the adult segment. Analysis was restricted to the 12-year period 1983–94, when survival estimates were judged to be most reliable.

Additional data on sex and age structure were available for kudu and wildebeest from earlier studies. For kudu, annual survival estimates over the period 1974–84, based on individually recognizable animals within two study areas (Owen-Smith 1990), were re-calculated including additional population structure records from outside the study areas, while omitting the distinction between prime and old females possible only in the study areas. Classified counts of wildebeest in the Central region of the KNP were undertaken between 1978 and 1983 by Whyte (1985). Calf and yearling proportions were derived by combining his samples for July and October

data analysis

Deriving survival estimates by balanced accounting

When the population totals in two successive years are known together with the annual recruitment into the population, the annual mortality loss can be estimated by simple balanced accounting, provided emigration and immigration are precluded: Nt = Nt–1 + bNt − dNt–1, where Nt = population size at time t, and b and d are the effective crude birth and death rates. Recruitment is represented by the juvenile segment aged less than 1 year at the time. Accordingly, the annual survival rate St between years t − 1 and t for all animals alive in year t − 1 is {Nt(1 − Jt)}/Nt–1, where Jt = juvenile proportion in the population at time t. Similarly, if both the juvenile and yearling proportions are known, survival into the ‘adult’ segment, represented by animals older than 2 years in year t, is SA,t = [Nt{1 − (Yt + Jt)}]/{Nt–1 (1 − Jt–1)}, where Yt = yearling proportion at time t. The accuracy of these estimates depends on the precision and repeatability of the aerial counts and on the reliability of the population structure samples. Juveniles tend be less visible from the air than other age classes, hence aerial counts could underestimate the annual population increase when the proportion of juvenile increases, and vice versa, thereby biasing estimates of annual adult survival. However, long-term estimates of adult survival are unaffected, because the annual biases cancel out. For 1994 when the count did not encompass the entire park, population totals were projected from the proportional population changes recorded in those regions that were covered.

Estimates of juvenile survival obtained from the juvenile/adult female ratio represent the outcome of fertility failures, fetal abortion and mortality between birth and the time of the survey. They also exclude offspring mortality associated with the death of the mother. Reductions in adult fecundity generally take place only after juvenile survival has been substantially reduced in large mammals, and are relatively minor among species producing just a single offspring (Eberhardt 1977, 2002; Gaillard et al. 2000). For example, for wildebeest in the Serengeti the adult pregnancy rate dropped from 0·95 to 1·0 while the population was growing to about 0·85 after the population had stabilized at high abundance, a change of at most 15% (Mduma, Sinclair & Hilborn 1999). Accordingly, changes in the juvenile/adult female ratio can be ascribed entirely or mostly to changes in juvenile survival. Juvenile survival estimates from ratios relative to adult females are also subject to bias from the proportion of females assigned to the ‘adult’ segment that were actually 2 years of age, and thus pre-parturient for most ungulate species. This was corrected using a stage-structured model formulated in a spreadsheet, as described below. Note that the procedure we used should not be confused with the structured demographic accounting described by Brown et al. (1993), a more rigorous procedure that requires complete population enumeration.

adjustments for variable population structure using a spreadsheet model

Five population segments were distinguished in the spreadsheet model: juveniles (< 1 year), yearlings (1–2 years), subadults (2–3 years), adult females (> 3 years) and adult males (> 3 years), with the adult segment subdivided further between males and females. The model was initiated sufficiently far in advance of the estimation period to allow the arbitrary starting population structure to adjust to the population trends shown prior to the period over which survival estimates were derived. Further assumptions were needed concerning relative survival rates between the yearling, subadult and adult stages, and the relation between the survival rate of adult males and that of adult females. Survival rates from yearling to subadult, and from subadult to adult, were assumed to be identical in both sexes to the survival rate among adult females (informed by the prior study on kudu, Owen-Smith 1993). Population structures were invariably biased towards females, indicating that adult males survived less well than adult females. To estimate sex-specific survival rates within the adult segment, we assumed that male survival covaried with female survival, but with male survival more strongly depressed by adverse conditions than that of females (cf. Toigo & Gaillard 2003). Specifically, SAm = (SAf)z, with the value of the power coefficient z adjusted until the adult sex ratio in the modelled population matched the mean adult sex ratio from the field classifications of population structure, for each species.

From 1983 on (but earlier for wildebeest and kudu with prior population structure data), the numbers of juveniles and yearlings each year were assigned using the observed ratios of these classes to females older than 2 years. The annual survival rates of adult females were then adjusted to replicate the observed changes in count totals between years. Juvenile survival was estimated directly from the ratio of juveniles to females aged two or more years for wildebeest, which are primiparous at 2 years of age in the Kruger Park (Whyte 1985). Warthog also usually give birth first at 2 years of age, but produce a typical litter of three (Mason 1982), so juvenile survival was estimated by dividing the juvenile/adult female (> 2 years) ratio by 3. For species with first reproduction at 3 years of age, juvenile survival was estimated from the ratio of juveniles to females in the modelled adult segment aged 3 years or older. Giraffe first reproduce only after 4 years of age, hence juvenile survival was estimated from the ratio of yearlings to adult females older than 4 years, adjusted for a 1·5-year minimum birth interval, thus spanning a 2-year period. Accordingly, the age class 3–4 years was also distinguished as subadult in the spreadsheet model for giraffe. For zebra, which cannot maintain annual births because of a 12-month gestation period, an adjustment was made to juvenile survival to allow for a mean pregnancy rate among adult females of 0·80 per year (Smuts 1976). Juveniles of most species were aged about 6–11 months at the time of the population structure surveys, so that juvenile survival estimates cover a period from pre-birth to shortly after weaning.

Yearling survival was estimated from the number of yearlings in year t relative to the number of juveniles in year t − 1 indicated in the spreadsheet model, for all species except giraffe. Yearling survival is pivotal on the age of 1 year, i.e. it spans the period between c. 0·5 and c. 1·5 years of age. The spreadsheet model yielded further estimates of annual survival for adult males, for the combined adult segment including both males and females, and for broader segments combining subadults, or yearlings plus subadults, with adults, by comparing the numbers within the appropriate classes between successive years.

adjustments for errors in count totals

Count totals are believed to represent between 60% and 85% of the true population depending on the species, but less for warthog (Redfern et al. 2002). Random census errors together with possible annual variation in the extent of the undercount bias affects estimates of annual adult survival, as well as producing negative autocorrelation between successive estimates of annual adult survival (McNamara & Harding 2004). Although mean survival estimates for adults over extended periods are unaffected (except marginally by edge effects), the error variance in these estimates is amplified, with a resultant reduction in statistical power to detect differences in survival between time periods. Two alternative procedures for dealing with the problem were followed: (1) factoring out the effect of the negative autocorrelation on survival estimates derived from the original count data statistically, or (2) transforming the original count data to suppress the negative autocorrelation prior to deriving survival estimates.

The data transformation entailed calculating a weighted running average, inline image = 0·67Nt + 0·33Nt+1, where inline image= adjusted population estimate, and Nt = recorded count total for year t. Stronger smoothing using a three-point weighted average introduced positive serial correlation into the adult survival estimates for most species, and hence was rejected. The two-point averaging effectively interpolates between successive pairs of count totals, with the weighting shifting the interpolation point towards the count total for the current year. This transformation effectively ‘reigned in’ the most extremely deviant estimates of adult survival, without much altering the annual survival estimates during periods when a constant population trend was maintained.

statistical procedures

Based on a prior statistical analysis (Piepho & Ogutu 2003), the study period was split into two contrasting phases of population trend: (1) the period prior to 1987, when most species were increasing in abundance, and (2) the post-1986 period, when populations were generally declining. To negate effects of autoregressive errors on adult survival estimates and their standard errors, survival estimates were contrasted between these phases using proc autoreg in SAS version 8·2 (SAS Institute 2001). This procedure employs the Yule–Walker method (or generalized least squares) to augment the regression model with an autoregressive model for the random error, through jointly estimating both the regression coefficients and the autoregressive model parameters (Gallant & Goebel 1976; Harvey 1981; Judge et al. 1985). A dummy number in the regression model represented the two phases of population growth. Estimates were thereby obtained of survival differences between these periods plus confidence limits and associated probability levels, as well as mean survival rates, their standard deviations and coefficients of variation (CVs), and the first order autocorrelation remaining in the residual error.

To assess demographic contributions to the change in population growth rate between the two periods, the λ (defined as Nt/Nt−1) generated by the mean juvenile recruitment and survival rates for the first phase, assuming a constant environment, was obtained from a spreadsheet model structured as described above after a stable age distribution had been achieved, for each population. The effect on λ of either (1) reducing the survival rates for both juveniles and yearlings to their mean values during the second period, or (2) lowering adult survival to its mean value during the second period, was then determined. For tsessebe, the yearling contribution was approximated, while for giraffe the juvenile (< 1 year) contribution was approximated. Survival rates for different demographic segments associated with conditions of zero population growth were obtained graphically, from the intersections of lines joining survival estimates for each phase of population trend with the perpendicular zero axis.

Results

population structure

Population samples were numerically substantial for the more common species, but became somewhat small towards the end of the study period for the two rarest species, sable and tsessebe (Table 1). Nevertheless, standard deviations for juvenile and yearling ratios for these two species did not differ from those for more abundant ungulates. The adult sex ratio was most strongly biased towards females for kudu, waterbuck and sable, and least so for giraffe and zebra.

Table 1.  Summary of the annual count totals, population structure samples and derived measures for the nine ungulate species
SpeciesRange in annual sample sizeJuveniles per adult* femaleYearlings per adult* femaleAdult males per adult female*
Count totals (1980–93)Population structure (1983–94)MeanSDMeanSDMean
  • *

    Females older than 2 years.

  • †Juveniles represented by 2-year-old animals.

Stabilizing species
 Zebra21 454–33 1641119–22860·3370·0340·1700·0200·675
 Wildebeest 8 568–14 601 779–36810·4940·1040·2910·0750·631
 Impala91 884–137 0553608–69450·4870·1810·3480·1640·471
 Giraffe 4 122–5759 436–12310·2020·0710·731
Declining species
 Kudu 3 127–10 760 348–21540·3400·1260·2390·0710·408
 Waterbuck 1 419–5042 263–11380·3960·1580·2450·1020·399
 Warthog   721–4354 117–8911·2120·5480·5930·2830·577
 Sable   856–2240  29–3110·3910·1140·2590·0800·427
 Tsessebe   222–1163  53–2700·3090·1350·511

variability in annual survival estimates

Juvenile survival was least variable for zebra and wildebeest, and most variable for impala and warthog, while CVs indicated greatest relative variability for warthog and tsessebe (Table 2). The variation in calf survival was surprisingly high for giraffe, considering its large body size, and possibly influenced by changes in the proportion of females giving birth each year. Standard deviations in yearling survival tended to be similar to those for juvenile survival, despite being prone to greater sampling error, while CVs were mostly less for yearling survival because of a higher mean survival. Standard deviations for adult survival were generally lower than those for juvenile survival, although not for zebra and warthog when estimates of adult survival were derived from the original count data. With the effect of census error on the adult survival estimates suppressed by the transformation of count totals, adult survival always appeared less variable than juvenile survival, but with the difference in the respective standard deviations not large in several cases. CVs were consistently much lower for adult than for juvenile survival, because of the substantial difference in mean survival rates between these two stages. Standard deviations in survival estimates were generally greater for the pre-1987 period, which included a severe drought year and corresponding reduction in annual survival for most population segments, than after 1986 (Table 3).

Table 2.  Statistical variability in the survival estimates
SpeciesDataSDCVFirst order autocorrelation*
OverallMean within phasesOverall
  • *

    First order autocorrelation in the residual error after fitting the model incorporating phase differences in survival rates.

  • Orig = estimated from original count data, Trans = estimated from transformed count data.

Juvenile survival
Stabilizing populations
 Zebra 0·0460·0470·100−0·627
 Wildebeest 0·0910·0810·180   0·474
 Impala 0·2110·2130·343   0·341
 Giraffe 0·1410·1390·357   0·540
Declining populations
 Kudu 0·1650·1660·384   0·490
 Waterbuck 0·1730·1740·378   0·144
 Warthog 0·1840·1910·442−0·070
 Sable 0·1160·1220·274−0·189
 Tsessebe 0·1520·1040·407−0·170
Yearling survival
Stabilizing populations
 Zebra 0·0530·0430·106−0·024
 Wildebeest 0·1290·1090·216−0·039
 Impala 0·1560·1590·213   0·201
Declining populations
 Kudu 0·1370·1040·200   0·401
 Waterbuck 0·1930·2020·328−0·157
 Warthog 0·1970·2060·390−0·238
 Sable 0·0920·0700·119   0·063
Adult survival
Stabilizing populations
 ZebraOrig0·0500·0480·054−0·565
Trans0·0280·0190·031   0·278
 WildebeestOrig0·0830·0590·094−0·143
Trans0·0820·0490·094   0·101
 ImpalaOrig0·1250·1300·158−0·362
Trans0·1120·1000·124   0·051
 GiraffeOrig0·0790·0820·088−0·221
Trans0·0570·0590·063   0·262
Declining populations
 KuduOrig0·1390·1160·170−0·286
Trans0·1030·0740·127−0·160
 WaterbuckOrig0·1600·1530·204−0·320
Trans0·1200·0970·154−0·004
 WarthogOrig0·2130·2020·340   0·152
Trans0·1590·1390·258   0·302
 SableOrig0·1080·1000·136−0·264
Trans0·0900·0740·115−0·006
 TsessebeOrig0·1080·0950·141−0·437
Trans0·0830·0710·110−0·348
Table 3.  Comparative differences in annual survival rates between the pre-1987 and post-1986 periods. Sample n indicates number of years spanned within each period
SpeciesnDataSurvival, pre-1987 (mean ± SD*)Survival, post-1986 (mean ± SD*)Difference ± 95% CIP
  • *

    SDs unadjusted for autocorrelation.

  • Estimates of means, confidence limits and P-values corrected for autocorrelation.

  • Orig = estimated from original count totals, Trans = estimated from transformed count totals.

Juveniles
Stabilizing populations
 Zebra4, 5 0.474 ± 0.0670.454 ± 0.023−0.020 ± 0.036   0.224
 Wildebeest8, 8 0.545 ± 0.0680.456 ± 0.092−0.089 ± 0.119   0.133
 Impala4, 8 0.651 ± 0.3170.569 ± 0.148−0.082 ± 0.364   0.622
 Giraffe4, 8 0.415 ± 0.0820.371 ± 0.157−0.044 ± 0.232   0.678
Declining populations
 Kudu9, 8 0.496 ± 0.1980.372 ± 0.119−0.124 ± 0.238   0.285
 Waterbuck4, 8 0.505 ± 0.1530.431 ± 0.182−0.074 ± 0.277   0.562
 Warthog4, 8 0.466 ± 0.2570.393 ± 0.154−0.073 ± 0.264   0.550
 Sable4, 8 0.439 ± 0.1520.422 ± 0.106−0.017 ± 0.157   0.811
 Tsessebe4, 8 0.534 ± 0.0900.293 ± 0.109−0.241 ± 0.142   0.003
Yearlings
Stabilizing populations
 Zebra4, 5 0.542 ± 0.0620.476 ± 0.018−0.066 ± 0.075   0.074
 Wildebeest4, 8 0.526 ± 0.1250.668 ± 0.090  0.148 ± 0.118   0.019
 Impala4, 8 0.757 ± 0.1670.709 ± 0.155−0.048 ± 0.261   0.689
Declining populations
 Kudu9, 8 0.761 ± 0.0870.625 ± 0.121−0.146 ± 0.140   0.042
 Waterbuck4, 8 0.590 ± 0.1360.593 ± 0.225  0.003 ± 0.261   0.978
 Warthog4, 8 0.515 ± 0.1860.484 ± 0.214−0.031 ± 0.245   0.784
Adults (both sexes > 2 years)
Stabilizing populations
 Zebra4, 8Orig0.955 ± 0.0630.912 ± 0.041−0.043 ± 0.037   0.031
Trans0.954 ± 0.0090.909 ± 0.023−0.045 ± 0.033   0.014
 Wildebeest8, 8Orig0.941 ± 0.0680.821 ± 0.052−0.120 ± 0.059< 0.001
Trans0.939 ± 0.0530.811 ± 0.044−0.128 ± 0.059< 0.001
 Impala4, 8Orig0.820 ± 0.2110.778 ± 0.071−0.042 ± 0.138   0.506
Trans0.822 ± 0.1540.775 ± 0.064−0.047 ± 0.151   0.504
 Giraffe4, 8Orig0.909 ± 0.1000.886 ± 0.073−0.024 ± 0.111   0.649
Trans0.907 ± 0.0580.887 ± 0.060−0.020 ± 0.100   0.666
Declining populations
 Kudu9, 8Orig0.896 ± 0.1480.734 ± 0.061−0.162 ± 0.096   0.003
Trans0.884 ± 0.0950.735 ± 0.037−0.149 ± 0.069< 0.001
 Waterbuck4, 8Orig0.892 ± 0.2410.738 ± 0.092−0.154 ± 0.169   0.069
Trans0.885 ± 0.1360.727 ± 0.073−0.158 ± 0.140   0.031
 Warthog4, 8Orig0.751 ± 0.2580.565 ± 0.172−0.186 ± 0.276   0.163
Trans0.736 ± 0.1470.541 ± 0.135−0.195 ± 236   0.095
 Sable4, 8Orig0.870 ± 0.1010.758 ± 0.099−0.112 ± 0.116   0.058
 Trans0.861 ± 0.0500.747 ± 0.082−0.115 ± 0.108   0.039
 Tsessebe4, 8Orig0.837 ± 0.1030.730 ± 0.092−0.107 ± 0.094   0.029
Trans0.819 ± 0.0610.722 ± 0.075−0.097 ± 0.076   0.019

A strong negative autocorrelation between successive estimates of juvenile survival was apparent for zebra, while for giraffe, kudu and wildebeest, annual juvenile survival was quite strongly positively autocorrelated (Table 2). For other species, the autocorrelation was effectively neutral. For yearling survival, the residual autocorrelation was generally neutral, except for kudu. Adult survival estimates derived from the original count totals were negatively autocorrelated for all species except warthog, and most strongly for zebra and tsessebe. Transforming the count data largely suppressed the serial correlation in adult survival estimates.

contrasts in survival rates between phases of population trend

All nine ungulate species showed lowered survival post-1986 compared with the period prior to 1987 in all life history stages, except for yearling wildebeest and juvenile waterbuck (Table 3). However, the difference in juvenile survival between the two periods was statistically significant only for tsessebe, although quite substantial also for impala, waterbuck and giraffe. Yearling survival differed significantly between these phases in the case of wildebeest, kudu and sable. All species, except for impala, giraffe and warthog, showed significant or almost significant reductions in adult survival between the earlier and later phases of population trend, for estimates derived from the transformed count data. Even when adult survival rates were calculated from the original count totals, probability levels for the differences between periods remained under 0·10 for these species. For warthog, the difference in adult survival between the two phases appeared quite large, but did not attain statistical significance due to high estimation error. As expected, mean adult survival estimates for the two phases were altered little by the data transformation.

Among the species persisting at high abundance, giraffe and impala showed mainly a reduction in juvenile survival between the two phases of population trend (Fig. 1). For zebra and wildebeest, the absolute change in adult survival slightly exceeded that among juveniles, while for zebra the estimated reduction in yearling survival was also greater than that among juveniles. For wildebeest, yearling survival was anomalously higher after the population had increased. Among the species showing substantial population declines, the decrease in adult survival appeared greater than that in juvenile survival for kudu, sable, waterbuck and warthog. Tsessebe showed a substantial reduction in adult survival, but with the magnitude of the change in juvenile survival even greater. Sable showed a large decrease in yearling survival, despite little change in juvenile survival, and kudu an even larger reduction in yearling survival than that for sable. For all species, the projected reduction in survival of adult males between the two phases was greater than that among adult females, as a consequence of the prior assumption made about how these rates were related. The change in annual survival for the aggregate grouping of adults, subadults and yearlings of both sexes (which is free of assumptions about how class-specific rates were related) also exceeded the magnitude of the change in juvenile survival between the phases of population growth for four of the declining species.

Figure 1.

Contrast in mean annual survival rates between the 1983–86 and 1987–94 periods, in relation to the mean annual population growth rate shown during these phases of population trend, for each demographic segment. For kudu and wildebeest, mean survival estimates for the 1978–82 period are additionally shown.

Survival rates within the adult and immature segments were widely disparate for zebra and giraffe, the two largest species, but closely similar for warthog, the species producing multiple young per litter (Fig. 1). Other species appeared intermediate.

contributions to depressed population growth

For seven of the nine species, the reduction in population growth rate (λ) between the pre-1987 and post-1986 periods was brought about primarily through reduced adult survival (Fig. 2). Giraffe and impala were the exceptions, with the change in immature survival contributing more to the altered population trend than the smaller change in adult survival. For all five of the declining species, as well as for wildebeest, reduced adult survival lowered λ by 0·1 or more, sufficient to transform a growing population into a declining one even with no change in immature survival.

Figure 2.

Contribution to the reduction in λ between the pre-1987 and post-1986 phases of population trend of changed adult survival vs. changed survival in the immature segment, i.e. juveniles plus yearlings combined.

survival rates for zero population growth

Adult survival estimates for zero population growth (obtained graphically as shown in Fig. 1) showed a negative relationship with juvenile recruitment (r =–0·902), represented by the juvenile/adult female ratio, as expected (Fig. 3a). The adult class included animals > 4 years of age for giraffe, >2 years for warthog, and > 3 years for all other species, while the measure of recruitment incorporates differences among species in fecundity as well as in juvenile mortality. An even tighter correlation was evident when survival rates after the juvenile stage were compared with juvenile recruitment (Fig. 3b, r = −0·974). The two largest species, giraffe and zebra, exhibited the highest adult survival rates, coupled with lowest recruitment rates. For both of these species, the mean birth interval exceeds 1 year, although only marginally so for zebra. Warthog, with high recruitment due to a typical litter of three, fell at the opposite extreme. Impala, the smallest among the medium-sized bovids, showed the lowest adult survival, coupled with the highest juvenile recruitment, for this subgroup. Among the remaining bovids, adult survival rates for population stasis clustered between 0·83 and 0·85, associated with juvenile/adult female ratios ranging between 0·43 and 0·62. For overall survival beyond the juvenile stage, sable showed the highest rate of 0·84, and impala the lowest rate of 0·79, at population stasis. Wildebeest showed a juvenile ratio almost as high as that of impala, an outcome of their early primiparity plus anomalous pattern in yearling survival.

Figure 3.

Survival rates of particular demographic segments associated with population stasis for each ungulate species. Juvenile recruitment is the ratio of offspring to adult females. (a) adults (both sexes combined) vs. juvenile recruitment; (b) all animals older than juveniles, vs. juvenile recruitment; (c) yearling vs. juvenile survival; (d) adult male vs. adult female survival (line indicates equality).

For impala, high juvenile survival was coupled with high yearling survival (Fig. 3c). For zebra, low juvenile recruitment was associated with low survival among yearlings, almost below that for warthog. Sable showed a high yearling survival counterbalancing low juvenile survival. For other bovids, juvenile survival rates of about 0·45–0·55 were coupled with yearling survival rates of about 0·6–0·7 at population stasis. The estimated sex differential in survival was small for zebra and giraffe, and between 0·08 and 0·14 per year for other species (Fig. 3d).

Discussion

demographic patterns

The annual coefficient of variation in juvenile survival exceeded that in adult survival for all nine of the KNP ungulate populations, in conformity with the general pattern for large herbivore populations established by Gaillard et al. (1998, 2000). Mean CVs in these survival rates appeared only moderately greater than the mean values for bovids reported by Gaillard et al. (2000), despite the fundamentally cruder estimation of adult survival rates from the demographic accounting procedure: for juveniles, mean of 0·318 in Kruger vs. 0·304 elsewhere; for yearlings, means of 0·225 vs. 0·143; for adults, mean of 0·120 from the transformed Kruger data, vs. 0·073 for prime-aged females and 0·178 for senescent females from the above reviews.

In contrast, the extent of the reduction in annual adult survival between the two phases of population trend shown was similar to or greater than that in juvenile survival for four of the five species with declining populations, and for wildebeest among the species persisting at high abundance. This was the case even though survival estimates for the pre-1987 phase for most of these populations were restricted to just 4 years, including the severe drought year of 1982–3 associated with reduced survival across all stage classes for most of these species. Accordingly the shifts in population trend between these periods, especially from roughly stable to declining, were brought about largely through decreased adult survival.

The survival estimates from KNP refer to the entire adult segment, i.e. all animals older than 2 years, as prime-aged and senescent animals could not be distinguished within the adult class. Gaillard et al. (1998) highlighted constancy especially in the survival rates of prime-aged females, compared with variable juvenile recruitment, as typical of large mammalian herbivores. Findings from the prior kudu study in the KNP, based on individually recognizable animals, were concordant with this pattern (Owen-Smith 1990). Survival rates of prime-aged female kudus changed little with density or rainfall, even during the severe 1982–3 drought, while the survival of females older than 6 years was fairly strongly influenced by changing rainfall relative to density. The mean survival rate of 0·91 estimated for prime-aged female kudus in this detailed study was associated with an annual survival rate among all adult females, including old animals, of 0·87 (re-calculated from data presented by Owen-Smith 1993). The estimated annual survival for adult female kudu obtained from balanced accounting for the pre-1987 period (0·88) is almost identical to the above estimate based on individually marked animals, supporting the lack of bias in the accounting procedure.

Population regulation occurring mainly through increased mortality among adults has been reported both for African buffalo Syncerus caffer (Sparrman) and for wildebeest in the Serengeti National Park in Tanzania (Sinclair 1974; Mduma et al. 1999), in both cases ascribed to nutritional limitations despite the abundance of predators. A change in susceptibility to mortality among adults following population stabilization has been recorded also for red deer Cervus elaphus L. on the Island of Rum, where predators are lacking (Albon et al. 2000). Changes in survival within the total adult segment could be entirely or largely the result of a shift in age structure of the population towards older animals with lowered survival rates (Festa-Bianchet, Gaillard, & Côté 2003). However, the magnitude of the reduction in adult survival among all of the declining populations, as well as for wildebeest in KNP, and its persistence over several years, makes it unlikely that it was restricted to the senescent segment. This implicates predation as centrally involved, specifically by lion, for which kills are concentrated largely on the adult segment of medium-sized ungulates (Pienaar 1969; Mills & Biggs 1993). The decline and subsequent persistent low density of several ungulate species in Namibia's Etosha National Park was likewise ascribed to high adult and yearling mortality, associated with high predation plus periodic outbreaks of anthrax (Gasaway et al. 1996).

comparative survival rates

For the Kruger Park ungulates, population stasis was associated with a juvenile recruitment (i.e. postweaning offspring: adult female ratio) of about 0·5 among the five medium–large bovids (adult body mass 80–200 kg). Because this measure incorporates maternal infertility and prenatal mortality, the actual survival of juveniles from birth to weaning could perhaps be 10% higher, thus corresponding closely with the mean juvenile survival rate for all bovids of 0·547 reported by Gaillard et al. (2000). The juvenile loss was counterbalanced by an annual survival rate among adults of both sexes combined of about 0·84, or for the broader class encompassing all animals beyond the juvenile stage of about 0·80. Specifically for the adult female segment, survival rates fell within the range 0·84–0·89 for these medium-sized species. The two largest ungulate species, giraffe and zebra, showed somewhat lower juvenile recruitment, and correspondingly higher survival within the adult class, about 0·90 or more. The two smallest species, impala and warthog, displayed higher juvenile recruitment and lower adult survival than the larger species. Survival rates among yearlings were generally about mid-way between those of juveniles and animals older than 2 years, but with much interspecific variability. Zebra showed surprisingly low survival in both the yearling and juvenile classes, closely similar to that for warthog. This probably reflects the concentration of lion predation on immature zebra (Mills & Shenk 1992).

Monthly survival rates for adult wildebeest in the Serengeti indicate an average annual survival close to 0·90 (Mduma et al. 1999), somewhat higher than shown by wildebeest in Kruger. Correspondingly, the mean calf survival rate of about 0·40 indicated by Mduma, Hilborn, & Sinclair (1998) for the Serengeti population is somewhat lower than that for Kruger wildebeest. For giraffe in the Serengeti, annual survival rates were estimated to be 0·95 among adults, 0·92 among subadults, 0·88 among yearlings, and 0·52 among juveniles to 6 months of age, also higher than the Kruger estimates, but with the population growing at 6% per year at that time (Pellew 1983). Sinclair (1995) reported a juvenile survival of 0·45 for giraffe, and juvenile survival rates under 0·40 for other medium-sized ungulates in the Serengeti. Among temperate zone ungulates, higher survival rates for prime-aged females (0·92–0·97 per year) were frequently associated with high juvenile survival, indicating growing populations (Gaillard et al. 1998). In the broader data set of ungulate populations assembled by Gaillard et al. (2000), the mean survival of prime-aged females was 0·874 (n = 57 populations), i.e. no different from the Kruger Park populations. The mean survival rate of yearlings in Kruger Park appears much lower than for ungulates elsewhere –0·650 prior to 1987, vs. 0·872 from Gaillard et al. (2000) – perhaps because the Kruger estimates span an earlier age range (0·6–1·6 years vs. 1·0–2·0 years of age).

Lower survival among adult males than among adult females is widely evident from adult sex ratios biased towards females (e.g. Owen-Smith 1993; Toigo & Gaillard 2003), but few estimates of the magnitude of the sex difference in mortality have been available for unhunted populations, and especially for populations subject to natural predation. For most of the Kruger bovids, the sex ratio among adults older than 2 years of about 0·4–0·5 males per female yielded a differential in sex-specific survival of between 0·12 and 0·14 per year. This represents almost a doubling in male mortality rate relative to mortality among adult females. The sex difference in annual mortality was rather less in zebra (0·025 per year), and giraffe (0·02 per year), but still led to a substantial deviation from parity in the adult sex ratio (0·67 males per female for zebra, 0·73 for giraffe). Wildebeest were intermediate, with a moderate bias in adult sex ratio (0·63) associated with a moderate sex difference in adult survival (0·08). The sex difference in adult survival for the Kruger populations did not show any clear relation with the extent of sexual size dimorphism, supporting the conclusion reached previously by Owen-Smith (1993) and Toigo & Gaillard (2003). The lion kill records assembled by Pienaar (1969) indicate a generally higher vulnerability of male ungulates to predation, compared with females.

implications for life history theory

The effects of variation in stage-specific survival rates on the overall population growth rate λ are commonly assessed using the proportional elasticity (∂ logλ/∂ logai,j, where ai,j is the survival rate between stages i and j; Caswell 2001). This measure standardizes variation relative to the mean survival rate, which is the ecological outcome of particular circumstances. Moreover, for ungulates the mean survival rate among juveniles tends to be about half that among adults, so that the same decrement in survival, as a proportion of the population segment, constitutes a smaller change relative to the mean survival for adults than for juveniles. Coupled with the fact that, for such long-lived, and mostly monotocous species, adults invariably constitute a greater proportion of the population, the elasticity of population growth is automatically much greater for changes in adult survival than for equivalent proportional changes in juvenile survival (see Albon et al. 2000 for a further discussion). Moreover, an analysis based on proportional changes in mortality, rather than in survival, could lead to a different interpretation (Link & Doherty 2002). The change in adult survival rates between different phases of population growth in KNP represented effectively a doubling of annual adult mortality. The proportional increase in juvenile mortality was much less, except for tsessebe. The neutral resolution for such large, long-lived species would seem to be the absolute magnitude of the change in survival or mortality rates (necessarily identical), which in life history analysis is simply the sensitivity ∂λ/∂ai,j.

Fixation upon the annual cycle for assessing variability in survival rates is also questionable for long-lived mammals with generation times spanning several years. Highlighting this are the high CVs in annual juvenile recruitment shown by the two largest ungulate species we considered, zebra and giraffe, despite their relatively stable populations. These CVs probably reflect the fact that females of these species have birth intervals exceeding 1 year, with synchrony among females in the year when they give birth potentially arising through environmental entraining (documented for the African elephant Loxodonta africana by Moss 2001). As annual recruitment typically amounts to only about 10% for monotocous species reproducing annually, changes in juvenile survival need to persist for several years to have much influence on population abundance. Factors influencing adult survival, such as trophic interactions with food resources and predator populations, and incidence of infectious diseases, vary over multiyear periods, while juvenile survival is more prone to annual variability in resources or weather conditions.

Gaillard & Yoccoz (2003) suggest that relative constancy in the survival of adult females could be a result of evolutionary canalization in the face of environmental variability, with adults effectively sacrificing offspring survival under adverse conditions in order to ensure their own future reproductive prospects. However, females in the Kruger ungulate populations continued to support offspring even when their own survival was substantially reduced. If the future survival of adults becomes uncertain due to a high risk of predation, residual reproductive value becomes lowered relative to the fitness contribution from current reproduction (Pianka & Parker 1973). Questions thus arise about the extent to which the life history patterns of various northern ungulates may have adjusted to a lack of large predators over many generations. Cohort effects may also lower the survival prospects of adults born under adverse conditions, as well as the reproductive success of these animals (Albon, Clutton-Brock & Guinness 1987; Albon, Clutton-Brock & Langvatn 1992), with the life history consequences depending on variability in these measures relative to their influence on fitness.

Conclusions

Findings highlighted the overwhelming importance of changes in adult survival rates for bringing about shifts in ungulate population trends, from increasing to stable, and especially stable to declining phases. This contrast with the general pattern documented for temperate-zone ungulates may be due to the central involvement of a predator preying largely upon the adult segment of most of these populations. Further long-term studies are needed, particularly on tropical species, and populations in temperate or arctic regions subject to predation, to provide a broader basis for generalizations about the typical life-history patterns of large mammalian herbivores and the evolutionary forces that have shaped them.

Acknowledgements

We are indebted to the Scientific Services division of the Kruger National Park, in particular D. Pienaar, H. Biggs, and I. Whyte, for making available the census data and population structure records after 1991. This paper benefited from the critical comments on earlier drafts by J.-M. Gaillard and M. Festa-Bianchet, and advice on statistical issues by L.P. Fatti.

Ancillary