- Wiley Online Library
Transcription
- Wiley Online Library
Journal of Animal Ecology 2010, 79, 109–116 doi: 10.1111/j.1365-2656.2009.01601.x Stochastic demography and population dynamics in the red kangaroo Macropus rufus Niclas Jonzén1*, Tony Pople2, Jonas Knape1 and Martin Sköld3 1 Department of Theoretical Ecology, Ecology Building, Lund University, SE-223 62 Lund, Sweden; 2Invasive Plants and Animals, Biosecurity Queensland, Department of Primary Industries and Fisheries, GPO Box 46, Brisbane, Qld 4001, Australia; and 3Department of Economics, Statistics and Informatics, Örebro University, SE-701 82 Örebro, Sweden Summary 1. Many organisms inhabit strongly fluctuating environments but their demography and population dynamics are often analysed using deterministic models and elasticity analysis, where elasticity is defined as the proportional change in population growth rate caused by a proportional change in a vital rate. Deterministic analyses may not necessarily be informative because large variation in a vital rate with a small deterministic elasticity may affect the population growth rate more than a small change in a less variable vital rate having high deterministic elasticity. 2. We analyse a stochastic environment model of the red kangaroo (Macropus rufus), a species inhabiting an environment characterized by unpredictable and highly variable rainfall, and calculate the elasticity of the stochastic growth rate with respect to the mean and variability in vital rates. 3. Juvenile survival is the most variable vital rate but a proportional change in the mean adult survival rate has a much stronger effect on the stochastic growth rate. 4. Even if changes in average rainfall have a larger impact on population growth rate, increased variability in rainfall may still be important also in long-lived species. The elasticity with respect to the standard deviation of rainfall is comparable to the mean elasticities of all vital rates but the survival in age class 3 because increased variation in rainfall affects both the mean and variability of vital rates. 5. Red kangaroos are harvested and, under the current rainfall pattern, an annual harvest fraction of c. 20% would yield a stochastic growth rate about unity. However, if average rainfall drops by more than c. 10%, any level of harvesting may be unsustainable, emphasizing the need for integrating climate change predictions in population management and increase our understanding of how environmental stochasticity translates into population growth rate. Key-words: climate, elasticity, mammals, matrix models Introduction Life-history evolution and population dynamics are two central fields of ecology where the original theories were deterministic but we have gained tremendously in general understanding by considering stochastic models (Lande, Engen & Sæther 2003). The relevance of models capturing the random variation experienced by natural populations goes beyond ecological theory and is motivated by the need to evaluate the population consequences of human impact such as harvesting (Lande, Engen & Sæther 1995; Jonzén, Ripa & Lundberg 2002) as well as climate change (Boyce et al. 2006). The latter will expose organisms to novel environmental conditions and potentially affect the magni*Correspondence author. E-mail: [email protected] tude and frequency of environmental events (Karl, Knight & Plummer 1995; Easterling et al. 2000) that shape the life history and demography of species. Large terrestrial mammalian herbivores are widespread and some of them are found in strongly fluctuating environments such as rainfall-driven savannas and grasslands (Nowak 1991). Despite living in stochastic environments and thus showing fluctuating vital rates (Gaillard et al. 2000), the demography and population dynamics are often studied using deterministic matrix models (Caswell 2001). Deterministic analyses of long-lived species have shown that the elasticity of population growth rate to adult survival is stronger than for juvenile survival and fecundity (Gaillard, FestaBianchet & Yoccoz 1998; Lebreton 2006). However, adult survival is known to be less variable than juvenile survival (Gaillard et al. 1998) and large annual variation in an element 2009 The Authors. Journal compilation 2009 British Ecological Society 110 N. Jonze´n et al. calculations, we assume that rainfall is the only driver of population change through its effect on survival and reproduction. There are of course other external factors that will affect the population growth rate, but this is still a reasonable assumption given the lack of significant limiting interactions with predators and ⁄ or parasites within Australia’s sheep rangelands (Caughley, Shepherd & Short 1987; Pople & Grigg 1998). The annual survival rate was modelled as a logistic function of annual rainfall (R): sk ¼ ck eak þbk R ; 1 þ eak þbk R 1 (a) 0·8 0·6 0·4 0·2 STOCHASTIC DENSITY-INDEPENDENT MODEL We used a demographic model with three age classes (0–12 months, 12–24 months and 24+) and a 1-year time step between t and t + 1. In drought conditions, red kangaroo populations often show in order what is typical of large mammals (Hanks 1981): decreased juvenile survival, delayed age at sexual maturity, reduced fecundity and decreased adult survival. Such a sequence has been found throughout Australia (Frith & Sharman 1964; Newsome 1965; Shepherd 1987; Pople 1996). Food supply as measured by pasture biomass or approximated by rainfall is correlated with the rate of increase (Caughley, Bayliss & Giles 1984; Bayliss 1985, 1987; Cairns & Grigg 1993). To simplify P(reproduction|rainfall) 0 0 Materials and methods ðeqn 1Þ where k refers to either adults (k = a) or juveniles (k = j). Due to data limitations, we only implemented two survival schedules: juvenile survival from the first- to the second-age class (when the young become independent),and adult survival between the later-age classes (Fig. 1a, Table 1). Details on the estimation of survival are given in Tenhumberg et al. (2004). The reproductive biology of red kangaroos has been well studied (see review by Tyndale-Biscoe & Renfree 1987). Breeding may, if environmental conditions are good, occur year round and females give birth to c. 1Æ5 young per year. As we are only modelling the female segment of the population, the number of female offspring produced per reproducing female is 0Æ75. The proportion of females giving birth at a given age is, however, dependent on the environment. Frith & Sharman (1964) studied how the proportion of mature females at a given age varied between areas differing in the amount of annual rainfall. We used data from the study by Frith and Sharman to interpolate the (increasing) proportion of mature females within the three age classes as a function of annual rainfall (Fig. 1b, Appendix S1, Supporting Information). To generate variability in the vital rates, we sampled annual rainfall with replacement from a 123-year-long time series on annual precipitation in Menindee at the edge of Kinchega National Park (32 S, 142 E). As a comparison, we also drew annual rainfall from a lognormal distribution, with parameters estimated from the Menindee data set. However, the results were not affected by the alternative procedures and we therefore present the analysis based on resampling only. Rainfall statistics are given in Fig. 2. Survival with a small elasticity may affect the population growth rate more than a small change in an element having high elasticity but less variability (Sæther 1997). Still, both empirical (e.g. Benton, Grant & Clutton-Brock 1995; Morris et al. 2008) and theoretical (Orzack & Tuljapurkar 1989) work suggests that long-lived species are often less sensitive to variability in vital rates than short-lived species. How sensitive they are to changes in environmental variability has, however, not been thoroughly investigated (but see Davis, Pech & Catchpole 2002). Hence, there is clearly a need for quantitative analyses to establish how variation in the environment affects population growth rate by changing means and variability in demographic rates (Sæther 1997). Understanding and predicting variation in population growth rate is also a challenge to population management. In South Australia, aerial surveys are conducted annually to estimate population abundance of the red kangaroo and the western grey kangaroo (Macropus fuliginosus, Desmarest; Grigg et al. 1999). The counts in a given year are used to set the harvest quota in the following year. A less costly alternative would be to decrease the survey frequency and to predict population abundance using a population model, data from previous surveys, and relevant covariates such as rainfall (Hauser, Pople & Possingham 2006). However, if population managers are going to rely on model predictions rather than survey data in some years, it becomes crucial to understand how stochastic rainfall translates into population growth rate (Jonzén et al. 2005), which is what we are addressing here. In this study, we make use of the approach developed by Tuljapurkar and co-workers (e.g. Tuljapurkar 1990; Tuljapurkar, Horvitz & Pascarella 2003) to calculate the elasticity of the stochastic growth rate with respect to the mean and variability in vital rates (i.e. survival and fecundity). We apply the methods to a stochastic environment model (Lebreton 2006) of the red kangaroo (Macropus rufus, Desmarest), a species inhabiting the arid and semi-arid zone of Australia, where the environment is characterized by unpredictable and highly variable rainfall. The elasticities with respect to the vital rates are compared to the elasticities with respect to the mean and variability of rainfall and we also evaluate how the stochastic growth rate changes with harvesting and changes in rainfall patterns. 1 100 200 300 400 500 600 700 800 100 200 300 400 500 Rainfall (mm) 600 700 800 (b) 0·8 0·6 0·4 0·2 0 0 Fig. 1. (a) Survival of individuals of age above (solid line) or below (dotted line) 12 months as a logistic function of annual rainfall. Parameter values are given in Table 1. (b) Probability of reproduction at age 12–24 months (dotted line) and 24+ months (solid line) as a function of annual rainfall. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 Stochastic demography and population dynamics 111 Table 1. Parameter values used in the logistic survival functions STOCHASTIC GROWTH RATE AND ELASTICITIES Survival parameters Value(s) Interpretation aj aa bj ba cj ca )3 0Æ1 0Æ014 0Æ01 0Æ8 0Æ99 Juvenile survival constant Adult survival constant Rain effect on juvenile survival Rain effect on adult survival Maximum juvenile survival Maximum adult survival Finally, we also tested the performance of a demographic model having 13-age classes, starting from newborn and each class covering 3 months except from the final group, which includes all animals being 36 months or older (when all animals are likely to be sexually mature), and we let the time step between time t and t + 1 be 3 months. However, the results did not change and we therefore only report results using the simple model having three age classes. To summarize, we let N(t) be the 3 · 1 population vector at time t and X(t) is the 3 · 3 stochastic transition matrix. The mapping from time t ) 1 to time t is given by NðtÞ ¼ XðtÞNðt 1Þ; ðeqn 2Þ where the temporal variation is due to stochastic rainfall affecting survival rates and the probability of reproduction at a given age. Rainfall (mm) 800 (a) 400 Frequency 1900 1920 1940 Year 1960 1980 2000 XðtÞ0 vðtÞ vðt 1Þ ¼ XðtÞ0 vðtÞ ; 1 ðeqn 5Þ of population structure vectors and reproductive vectors respectively, where kk1 denotes ‘1-norm’, i.e. summation of the absolute value of all elements and u(0) = v(T) = 1 ⁄ 3Æ(1, 1, 1) (or some other arbitrary initial vectors). All model properties were obtained using numerical simulations in MATLAB (The MathWorks Inc. 2000). We were interested in separating the elasticity of ks with respect to the mean lij and the standard deviation rij of vital rate Xij: @ log ks ; @ log lij @ log ks ¼ : @ log rij ESl ij ¼ ðeqn 6Þ ðeqn 7Þ where Æv(t), u(t)æis a scalar product and Cij is the perturbation of the Xij element. We get the elasticity of ks with respect to the mean of Xij by setting Cij(t) = lij. Similarly, by setting Cij(t) = Xij(t) ) lij, we get the elasticity with respect to the standard deviation. 5 0 0 100 200 300 400 500 Rainfall (mm) 600 700 800 (c) 0·5 ACF ðeqn 4Þ X 1 T vi ðtÞCij ðtÞuj ðt 1Þ ; T!1 T kðtÞhvðtÞ; uðtÞi t¼1 10 0 –0·5 –1 0 XðtÞuðt 1Þ ; kXðtÞuðt 1Þk1 uðtÞ ¼ Dij ¼ lim (b) 15 1 We also generated the sequences The sum of these two elasticities equals the so-called stochastic elasticity (Tuljapurkar et al. 2003), which confounds the effects of changing mean and variability of the vital rates (Haridas & Tuljapurkar 2005). The full elasticity matrix was computed numerically following Tuljapurkar (1990): 200 20 ðeqn 3Þ ESr ij 600 0 1880 For the model given by eqn 2, we computed the stochastic growth rate ks from 1 jNðtÞj ln ðCaswell 2001and references thereinÞ: lnðks Þ ¼ lim t!1 t jNð0Þj 2 4 6 Time lag (years) 8 10 Fig. 2. (a) Time series, (b) histogram and (c) autocorrelation function of total annual rainfall in Menindee. The dotted lines in (c) indicates the approximate 95% confidence interval. The average annual rainfall is 242 mm, the standard deviation is 107 mm and the coefficient of variation 0Æ44. THE EFFECT OF HARVESTING AND CLIMATE CHANGE Red kangaroos are harvested for meat and skins and, since 1984, each of the mainland States has offered annual harvest quotas that are a constant proportion (10–20%) of the estimated population size (Pople & Grigg 1998; Grigg & Pople 2001). Modeling studies have suggested that harvest rates of c. 10–20% of kangaroo populations are sustainable in the long term (Caughley 1987; Hacker, McLeod & Druhan 2003; Pople 2003) but, in some years and areas, red kangaroo populations face annual harvest fractions above 30% (Pople 2006). We were interested in contrasting elasticity patterns between an unharvested and a harvested population. In the harvested population, we decreased the annual survival rate of the oldest age class by 30%. Primarily adults are harvested as harvesters are paid according to carcass and skin size. The harvest tends to be male biased, so a harvest rate of 30% for a female population is particularly conservative. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 112 N. Jonze´n et al. Australian rainfall varies substantially, especially in the dry and semi-dry areas inhabited by the red kangaroo. Projections of future rainfall patterns are uncertain with models simulating either increase or decrease across large areas of Australia (Whetton et al. 2005). To study the effect of possibly changing rainfall patterns due to climate change, we increased or decreased the observed values of annual rainfall by a constant times the mean value and, again, sampled from data with replacement. An alternative procedure would be to multiply each value by a constant. The latter procedure does not only change the average rainfall but also the variance. Because it is not known which procedure is most consistent with future global change, we evaluated both methods. To get a general idea about how harvest affected the stochastic growth rate under different rainfall scenarios, we varied the annual harvest fraction between 0 and 0Æ4 and, for each combination of harvest fraction and average rainfall, we estimated the stochastic growth rate. Finally, we note that anthropogenic modification of the biosphere is projected to increase climatic variability, resulting in more variable rainfall and more frequent extreme temperatures (Karl et al. 1995; Easterling et al. 2000).We therefore estimated the elasticity of ks to a change in the standard deviation of annual rainfall (see Appendix D in Haridas & Tuljapurkar (2005) for the numerical procedure). For a comparison, we also estimated the elasticity of ks to a change in mean rainfall. SERIAL AUTOCORRELATION There is no evidence of serial autocorrelation in the rain data and decreasing the prediction interval makes no difference as the autocorrelation with lag 1 month is only c. 0Æ1. However, as environmental serial autocorrelation can have a sizeable effect on population growth rate (Tuljapurkar & Haridas 2006) and we want to make sure that our results are robust to any misspecification of the autocorrelation structure, we nevertheless simulated rainfall from a process where log of rainfall was assumed to follow an AR(1), having the observed (on a log scale) mean and variance as expectations and let the autoregression parameter be 0Æ8. That did not change the elasticity patterns. Results The stochastic rainfall translates into variation in vital rates and especially juvenile survival, and the reproduction in age classes 2 and 3 fluctuates considerably (Fig. 3). The adult survival does not fluctuate much, reaching 5% of the maximum possible variance given its mean (see Morris & Doak 2004), compared to 19% in juvenile survival. The age distribution fluctuates heavily but, on average, c. 30% of the population belong to age class 1, 10% to age class 2 and 60% are found in age class 3 (Fig. 4). In Fig. 5, we present the elasticities with respect to the l r mean, ESij , and the variability, ESij , of the matrix elements. The stochastic growth rate is far more sensitive to a proportional change in the mean value when compared to the standard deviation of a matrix element. Summing up all the standard deviation elasticities, we get an estimate of the effect of variability on population growth rate: Tr ¼ X r ESij ¼ 0026: ij There is coupling between the summed elasticities with respect to the mean (Tl) and variability (Tr) of the matrix elements such that Tl + Tr = 1 (Haridas & Tuljapurkar 2005). Hence the value of Tr found here is relatively small. The stochastic growth rate (with no harvest) was estimated to be 1Æ056, which should be compared to the deterministic growth rate derived from the average matrix (k = 1Æ068), where the elements of the average matrix are the expected survival and reproductive rates. To what extent this difference is negligible or important depends on the context. The variability in vital rates does, however, induce variation in the population growth rate (range = 0Æ50–1Æ61, CV = 20%), which is important for population management (see Discussion). (b) x 104 2·5 (a) 15 000 1·5 1 5000 0·5 0 0 0·5 Juvenile survival 0 0 1 (c) x 104 3 Frequency Frequency 2 10 000 (d) 2 0·5 Adult survival 1 x 104 1·5 2 1 1 0 0 0·5 0·5 Fecundity age class 2 1 0 0 0·5 Fecundity age class 3 1 Fig. 3. Histogram of (a) juvenile annual survival rate, (b) adult annual survival rate, fecundity (number of female offspring per female) in (c) age class 2 and (d) age class 3. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 Stochastic demography and population dynamics 113 0·8 0·7 Age distribution 0·6 0·5 0·4 0·3 0·2 Fig. 4. An example of the temporal fluctuation in age structure. Age class 1 (open circles): mean = 0Æ28, SD = 0Æ047, CV = 0Æ17. Age class 2 (filled triangles): mean = 0Æ11, SD = 0Æ039, CV = 0Æ35. Age class 3 (filled circles): mean = 0Æ60, SD = 0Æ076, CV = 0Æ13. 0·1 0 900 910 In an unharvested population survival in age class 3 has l the highest ES followed by survival from age class 1 to 2, fecundity in class 3, survival from class 2 to 3 and fecundity in age class 2. Harvesting 30% of age class 3 increases the magl nitude of ES of survival of the two youngest age classes and of fecundity in the final age class, but the relative ranking of the elasticities remains (Fig. 5). With the survival harvesting l in age class 3 again has the highest ESij , but is now followed by in age class 3. The latter also has the highest fecundity Sr Eij . In accordance with previous simulation studies, the model kangaroo population analysed here can sustain an annual 920 930 940 950 960 Time (years) 970 980 990 1000 harvest fraction of c. 20% (Fig. 6). However, if average rainfall drops by more than c. 10%, any level of harvesting may be unsustainable, as even an unharvested population cannot maintain itself. A caveat is that there is no density dependence so there is no harvest compensation to halt any longterm declines. As can be seen in Fig. 6, a given increase in average rainfall has less effect than a given decrease on the stochastic growth rate. Hence, we need to reduce harvesting if the average annual rainfall declines, but we should not increase harvesting to the same extent if the average annual rainfall were to increase. This is a result of Jensen’s inequality (Ruel & Ayres 1999) and the concave relationship between Fig. 5. Proportional change of the stochastic growth rate for a proportional change in the mean value (a, b) or the standard deviation (c, d) of survival (a, c) and fecundity (b, d). The solid lines represent an unharvested population and the dotted line a population where 30% of the individuals aged 24 months or more are harvested. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 114 N. Jonze´n et al. 20 1·2 1·2 1·2 10 1·1 1·1 1·1 5 1 1 0·9 1 0 0·9 0·9 0·8 0·7 0·6 0·8 1 –5 0·7 0·6 0·5 0·4 0·3 5 6 0· 0· –20 0 0·2 0· –15 7 0· 8 –10 0·9 Change in average rainfall (%) 15 0·05 0·1 0·15 0·4 0·2 0·3 0·25 0·3 0·35 0·4 Harvest fraction Fig. 6. The effect of harvesting (annual fraction of age class 3) and a change in the average annual rainfall on the stochastic growth rate. rainfall and the stochastic growth rate. The latter can be seen by noting that – for a given harvest fraction – the distance between contour lines increases with increasing rainfall (Fig. 6). The concave relationship also implies that increased variability in vital rates will decrease the stochastic growth rate, as shown by the negative standard deviation elasticities (Fig. 5c,d). Reducing the average rainfall by multiplying the observed mean with a constant gives very similar results. Compared to the results reported in Fig. 6, a change in average rainfall does not change the stochastic growth rate to the same extent. However, the overall picture is almost identical (not shown). Finally, we evaluated the effect on the stochastic growth rate of a proportional disturbance of the standard deviation in annual rainfall. In contrast to the weak effect of a proportional disturbance of the variability in vital rates, the elasticity of ks to a change in the standard deviation of annual rainfall turned out to be far more important and was estimated to )0Æ10. This should, however, be compared with the elasticity of ks to a change in mean rainfall, which was estimated to 0Æ55. Harvesting 30% of age class 3 increased the elasticities with respect to the mean and the standard deviation of rainfall to 0Æ78 and )0Æ14 respectively. The increase in the elasticity with respect to the mean caused by harvest mainly stems from an increase in the elasticity with respect to juvenile survival. The decrease of the elasticity with respect to the standard deviation stems from changes of roughly equal magnitude in the elasticities with respect to juvenile survival, adult fecundity and adult survival, the first two negative and the later positive. Discussion In most studies on elasticity in random environments, the socalled stochastic elasticity, ES, is calculated (Caswell 2001). However, ES confounds the effects of changing mean and variability of the vital rates (Haridas & Tuljapurkar 2005) and we have therefore, following Tuljapurkar et al. (2003), separated the effects of perturbing the mean and variability of the matrix elements. Applying this framework to a model population of the red kangaroo, a classical example of population dynamics strongly affected by stochastic rainfall (Caughley et al. 1987), we show that the elasticity to variability in vital rates is very low compared to the elasticity to the mean values. This may seem surprising considering the importance of the highly variable rainfall for kangaroo demography, but as the coefficient of variation of adult survival rate is rather small (Fig. 3), we should not expect a proportional change in the standard deviation to be very important in relation to a proportional change in the mean (see Fig. 5; see also Morris et al. (2008)). A change in the standard deviation of juvenile survival could be pronounced, but its effect is hampered by the relatively stable and productive third age class (Fig. 4). The weak effect of variability in vital rates in the red kangaroo is in accordance with a recent study (Morris et al. 2008) showing that short-lived species are predicted to be more strongly affected by increased variability in vital rates relative to long-lived species. To calculate the elasticity of the stochastic growth rate with respect to the mean and variability in vital rates is motivated by the fact that climate models suggest changes in not only average conditions but also in variability (Karl et al. 1995; Easterling et al. 2000). Hence, we should map from changes in the interesting environmental variables via vital rates to population change (Boyce et al. 2006). As noted by Morris et al. (2008), the data needed are rarely available and our study is important in showing that the generally low elasticity to vital rate variability in long-lived species (Morris et al. 2008) does not necessarily imply that those species are insensitive to changes in environmental variability. The overall effect of a change in the variability of rainfall is a combination of mean and standard deviation elasticities for all vital rates affected by rainfall (Haridas & Tuljapurkar 2005), and the estimated elasticity with respect to the standard deviation of rainfall is comparable to the elasticities with respect to the mean of all vital rates but the survival in age class 3. Hence ignoring the effect of increased variability in rainfall is not motivated by referring to the low standard deviation elasticities. However, one should bear in mind that the elasticity with respect to the mean rainfall is five times the elasticity with respect to the standard deviation of rainfall, i.e. keeping track of the average conditions is far more important in a long-lived species such as the red kangaroo. This is in line with some previous empirical studies on long-lived species (e.g. Benton et al. 1995; Morris et al. 2008). Even though long-lived species in general seem to be less sensitive to environmental fluctuations than short-lived species, there are likely to be interspecific differences in how they respond to environmental stochasticity. The ability of a life history to handle stochasticity is dependent on the amount of environmental variation, how it affects the vital rates and on the details of the life history, as well as on the generation time (Orzack & Tuljapurkar 1989). To contrast 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 Stochastic demography and population dynamics 115 the case of the red kangaroo to another long-lived mammal, consider the red deer (Cervus elaphas) on Rum in Scotland. This population also shows weak effects of stochasticity on elasticity patterns, but has a different life history than the red kangaroo in that there are trade-offs between current fecundity, future fecundity and survivorship (Benton et al. 1995). Incorporating these costs of reproduction gives negative elasticities with respect to survival in most age classes. Several other studies have highlighted the importance of covariance among vital rates (e.g. Sæther & Bakke 2000) and failing to take that into account can lead to misleading conclusions (Morris & Doak 2004; Coulson, Gaillard & Festa-Bianchet 2005). Vital rates in red kangaroos are positively correlated due to the effect of rainfall but female red kangaroos show intriguing adaptations to unpredictable fluctuations between periods of dry and wet conditions that do not require the inclusion of trade-offs. For instance, females can delay the progress of pregnancy such that they can rapidly replace a young that died early or left the pouch (Tyndale-Biscoe & Renfree 1987), and the cost of reproduction is therefore very low. It is reasonable to assume that the low standard deviation elasticities reported here are the results of an evolutionary past where the plasticity in vital rates was under strong selection. The reproductive strategy in the red kangaroo is especially well adapted to unpredictable fluctuations, and there is much more to be gained (in fitness terms) by increasing the average of vital rates rather than further minimizing the fluctuations (see also Gaillard & Yoccoz 2003). However, an increased harvest fraction makes the population more sensitive to a decrease in average rainfall, mainly through increasing the importance of juvenile survival, or to an increase in the variability of rainfall through increasing the sensitivity of adult fecundity and juvenile survival. Hence, to maintain a sustainable harvesting strategy, it is important to keep track of the nature of changes in climate and their effects on vital rates. In conclusion, the red kangaroo system illustrates an example where climate effects are indirect and mediated by the intrinsic processes of birth, death and age structure (Krebs & Berteaux 2006). By making a rigorous analysis that explicitly partial out the effect of mean and variability in vital rates due to stochastic rainfall, we have shown that red kangaroos are likely to be affected not only by any reduction in average rainfall through its negative effect on all vital rates, but also by increased variability in rainfall. The latter may be important also in long-lived species if variability is high and many vital rates are affected. Acknowledgements We are very grateful to K. Al-Khafaji, C.V. Haridas, S. Tuljapurkar, S. Beissinger and two anonymous reviewers who provided comments and suggestions that significantly improved this study. Carol Horvitz kindly provided the MATLAB code used in Tuljapurkar et al. (2003). N.J. and J.K. were financially supported by the Swedish Research Council. References Bayliss, P. (1985) The population dynamics of red and western grey kangaroos in arid New South Wales, Australia. I. Population trends and rainfall. Journal of Animal Ecology, 54, 111–125. Bayliss, P. (1987) Kangaroo dynamics. Kangaroos: Their Ecology and Management in Sheep Rangelands of Australia (eds G. Caughley, N. Shepherd & J. Short), pp. 119–134. Cambridge University Press, Cambridge. Benton, T.G., Grant, A. & Clutton-Brock, T.H. (1995) Does environmental stochasticity matter? Analysis of red deer life histories on Rum. Evolutionary Ecology, 9, 559–574. Boyce, M.S., Haridas, C.V., Lee, C.T. & the NCEAS Stochastic Demography Working Group. (2006) Demography in an increasingly variable world. Trends in Ecology and Evolution, 21, 141–148. Cairns, S.C. & Grigg, G.C. (1993) Population dynamics of red kangaroos (Macropus rufus) in relation to rainfall in the South Australian pastoral zone. Journal of Applied Ecology, 30, 444–458. Caswell, H. (2001) Matrix Population Models: Construction, Analysis, and Interpretation, 2nd edn. Sinauer Associates, Sunderland, MA. Caughley, G. (1987) Ecological relationships. Kangaroos: Their Ecology and Management in Sheep Rangelands of Australia (eds G. Caughley, N. Shepherd & J. Short), pp. 159–187. Cambridge University Press, Cambridge. Caughley, G., Bayliss, P. & Giles, J. (1984) Trends in kangaroo numbers in western New South Wales and their relation to rainfall. Australian Wildlife Research, 11, 415–422. Caughley, G., Shepherd, N. & Short, J. (eds) (1987) Kangaroos: Their Ecology and Management in Sheep Rangelands of Australia. Cambridge University Press, Cambridge. Coulson, T., Gaillard, J.M. & Festa-Bianchet, M. (2005) Decomposing the variation in population growth into contributions from multiple demographic rates. Journal of Animal Ecology, 74, 789–801. Davis, S.A., Pech, R.P. & Catchpole, E.A. (2002) Populations in variable environments: the effects of variability in a species’ primary resource. Philosophical Transactions of the Royal Society of London. Series B: Biological Sciences, 357, 1249–1257. Easterling, D.R., Meehl, J., Parmesan, C., Chagnon, S., Karl, T.R. & Mearns, L.O. (2000) Climate extremes: Observations, modeling, and impacts. Science, 289, 2068–2074. Frith, H.J. & Sharman, G. (1964) Breeding in wild populations of the red kangaroo, Megaleia rufa. CSIRO Wildlife Research, 9, 86–114. Gaillard, J.M. & Yoccoz, N.G. (2003) Temporal variation in survival of mammals: a case of environmental canalization. Ecology, 84, 3294–3306. Gaillard, J.M., Festa-Bianchet, M. & Yoccoz, N.G. (1998) Population dynamics of large herbivores: variable recruitment with constant adult survival. Trends in Ecology and Evolution, 13, 58–63. Gaillard, J.M., Festa-Bianchet, M., Yoccoz, N.G., Loison, A. & Toı̈go, C. (2000) Temporal variation in fitness components and population dynamics of large herbivores. Annual Reviews in Ecology and Systematics, 31, 367– 393. Grigg, G.C. & Pople, A.R. (2001) Sustainable use and pest control in conservation: kangaroos as a case study. Conservation of Exploited Species (eds J. Reynolds, G. Mace, K. Redford & J. Robinson), pp. 403–423. Cambridge University Press, Cambridge. Grigg, G.C., Beard, L.A., Alexander, P., Pople, A.R. & Cairns, S.C. (1999) Survey of kangaroos in South Australia 1978-1998: a brief report focusing on methodology. Australian Zoologist, 31, 292–300. Hacker, R.B., McLeod, S.R. & Druhan, J. (2003) Final Report for Project #D8003. Evaluating Alternative Management Strategies for Kangaroos in the Murray Darling Basin. NSW Department of Agriculture, Dubbo, NSW. Hanks, J. (1981) Characterization of population condition. Dynamics of Large Mammal Populations (eds C.W. Fowler & T.D. Smith), pp. 47–73. Wiley and Sons, New York. Haridas, C.V. & Tuljapurkar, S. (2005) Elasticities in variable environments: properties and implications. The American Naturalist, 166, 481–495. Hauser, C.E., Pople, A.R. & Possingham, H.P. (2006) Should managed populations be monitored every year? Ecological Applications, 16, 807–819. Jonzén, N., Ripa, J. & Lundberg, P. (2002) A theory of stochastic harvesting in stochastic environments. The American Naturalist, 159, 427–437. Jonzén, N., Pople, A.R., Grigg, G.C. & Possingham, H.P. (2005) Of sheep and rain – large-scale population dynamics of the red kangaroo. Journal of Animal Ecology, 74, 22–30. Karl, T.R., Knight, R.W. & Plummer, N. (1995) Trends in high frequency climate variability in the 20th century. Nature, 377, 217–220. Krebs, C.J. & Berteaux, D. (2006) Problems and pitfalls in relating climate variability to population dynamics. Climate Research, 32, 143–149. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116 116 N. Jonze´n et al. Lande, R., Engen, S. & Sæther, B.-E. (1995) Optimal harvesting of fluctuating populations with a risk of extinction. The American Naturalist, 145, 728–745. Lande, R., Engen, S. & Sæther, B.-E. (2003) Stochastic Population Dynamics in Ecology and Conservation. Oxford University Press, Oxford. Lebreton, J.D. (2006) Dynamical and statistical models of vertebrate population dynamics. Comptes Rendus Biologies, 329, 804–812. Morris, W.F. & Doak, D.F. (2004) Buffering of life histories against environmental stochasticity: accounting for a spurious correlation between the variabilities of vital rates and their contributions to fitness. American Naturalist, 163, 579–590. Morris, W.F., Pfister, C.A., Tuljapurkar, S., Haridas, C.V., Boggs, C.L., Boyce, M.S., Bruna, E.M., Church, D.R., Coulson, T., Doak, D.F., Forsyth, S., Lee, C.T. & Menges, E.S. (2008) Longevity can buffer plant and animal populations against changing climatic variability. Ecology, 89, 19–25. Newsome, A.E. (1965) Reproduction in natural populations of the red kangaroo, Megaleia rufa (Desmarest), in central Australia. Australian Journal of Zoology, 13, 735–759. Nowak, R.M. (1991) Walker’s Mammals of the World, Vol. 2, 5th edn. Johns Hopkins University Press, Baltimore, MD. Orzack, S.H. & Tuljapurkar, S. (1989) Population dynamics in variable environments. VII. The demography and evolution of Iteroparity. The American Naturalist, 133, 901–923. Pople, A.R. (1996) Effects of harvesting upon the demography of red Kangaroos in Western Queensland. PhD thesis, The University of Queensland, Brisbane. Pople, A. (2003) Harvest Management of Kangaroos During Drought. Unpublished report to New South Wales National Parks and Wildlife Service, Dubbo, NSW. Pople, A.R. (2006) Modelling the Spatial and Temporal Dynamics of Kangaroo Populations for Harvest Management. Final report to the Department of Environment and Heritage, Canberra. Pople, A.R. & Grigg, G.C. (1998) Commercial Harvesting of Kangaroos in Australia. Environment Australia, Canberra. Available at: http://www. environment.gov.au/biodiversity/trade-use/wild-harvest/kangaroo/harvesting/ index.html, accessed 21 July 2009. Ruel, J.J. & Ayres, M.P. (1999) Jensen’s inequality predicts effects of environmental variation. Trends in Ecology and Evolution, 14, 361–366. Sæther, B.-E. (1997) Environmental stochasticity and population dynamics of larger herbivores: a search for mechanisms. Trends in Ecology and Evolution, 12, 143–149. Sæther, B.–.E. & Bakke, Ø. (2000) Avian life history variation and contribution of demographic traits to the population growth rate. Ecology, 81, 642–653. Shepherd, N. (1987) Condition and recruitment of kangaroos. Kangaroos: Their Ecology and Management in Sheep Rangelands of Australia (eds G. Caughley, N. Shepherd & J. Short), pp. 135–158. Cambridge University Press, Cambridge. Tenhumberg, B., Tyre, A.J., Pople, A.R. & Possingham, H.P. (2004) Do harvest refuges buffer kangaroos against evolutionary responses to selective harvesting. Ecology, 85, 2003–2017. The MathWorks Inc. (2000) Using MatLab, Version 6. The MathWorks Inc., Natick, MA, USA. Tuljapurkar, S. (1990) Population Dynamics in Variable Environments. Springer-Verlag, New York, NY. Tuljapurkar, S. & Haridas, C.V. (2006) Temporal autocorrelation and stochastic population growth. Ecology Letters, 9, 327–337. Tuljapurkar, S., Horvitz, C.C. & Pascarella, J.B. (2003) The many growth rates and elasticities of populations in random environments. American Naturalist, 162, 489–502. Tyndale-Biscoe, C.H. & Renfree, M.B. (1987) Reproductive Physiology of Marsupials. Cambridge University Press, Cambridge. Whetton, P.H., McInnes, K.L., Jones, R.N., Hennessy, K.J., Suppiah, R., Page, C.M., Bathols, J.M. & Durack, P.J. (2005) Australian Climate Change Projections for Impact Assessment and Policy Application: A Review. CSIRO Marine and Atmospheric Research Paper, 001. CSIRO Marine and Atmospheric Research, Aspendale, Vic. Received 19 March 2009; accepted 3 July 2009 Handling Editor: Tim Benton Supporting Information Additional supporting information may be found in the online version of this article. Appendix S1. Probability of reproduction within the three age classes as a function of annual rainfall. Please note: Wiley-Blackwell are not responsible for the content or functionality of any supporting materials supplied by the authors. Any queries (other than missing material) should be directed to the corresponding author for the article. 2009 The Authors. Journal compilation 2009 British Ecological Society, Journal of Animal Ecology, 79, 109–116