Dr R. Shine, School of Biological Sciences A08, University of Sydney, NSW 2006, Australia. Fax: 61-2-9351-5609. E-mail: firstname.lastname@example.org
1. Many organisms rely upon stored energy reserves to support reproduction and do not initiate breeding until their reserves exceed some ‘reproductive threshold’. However, the determinants of such thresholds are poorly understood; for example, we do not know if they are fixed (invariant) or respond dynamically to fluctuations in resource availability.
2. An 8-year field study on water pythons (Liasis fuscus Peters 1873) in tropical Australia shows that individual female pythons adjust their reproductive thresholds in response to annual variation in prey abundance.
3. In every year of the study, female pythons that reproduced were in better condition (mass relative to body length) than non-reproductive females. However, in years with low abundance of rats, female pythons reproduced despite being in relatively poor condition. Indeed, the mean condition of reproductive pythons in one particularly ‘bad’ year was as low as the mean condition of non-reproductive females in a ‘good’ year.
4. Clutch sizes were slightly reduced in a ‘bad’ year, but the main effect of the lowered threshold was greater emaciation of the females after laying.
5. Recapture records of marked snakes show that the annual variation in thresholds is due to flexibility of individual females, not to differential representation of cohorts with different reproductive thresholds.
6. The dynamic adjustment of threshold levels fits well with predictions from life-history models and is likely to be a widespread phenomenon.
One fundamental aspect of life-history variation is energy storage, and the possibility it allows for a temporal dissociation between the accrual of a resource and the expenditure of that resource in reproduction. Although many organisms (‘income breeders’ in Drent & Daans 1980 terminology) gather the energy used for breeding during the reproductive season, most animals rely to some extent on ‘capital’ (stored energy) to support reproductive expenditure. This ‘capital-breeding’ tactic is particularly common in ectotherms, especially long-lived species (e.g. Doughty & Shine 1998). For such an animal, the ‘decision’ as to whether or not to reproduce in a given year appears to depend primarily upon the magnitude of stored reserves; reproduction will occur only when the reserves exceed some threshold level. Such thresholds have been clearly documented in several taxa (e.g. DeRouen et al. 1994; Naulleau & Bonnet 1996).
Although the idea that many animals delay reproduction until they have sufficiently large energy reserves is a simple and uncontroversial one (e.g. Frisch 1984; DeRouen et al. 1994; Ebbinge & Spaans 1995), we know relatively little about the determinants of such thresholds. Theoretical models suggest that the threshold should be determined by optimal clutch size issues, and particularly by non-linearities in the relationship between the costs and benefits of additional reproductive investment (e.g. Schaffer 1974; Bull & Shine 1979). In the current paper, we address a central issue in this field: is the threshold constant for a given species, or does it vary through time depending on the organism's perception of reproductive costs and resource availability?
Previous empirical work has tended to assume that the threshold level is invariant and this may often be true in relatively stable environments. However, logic suggests that organisms exposed to stochastic fluctuations in resource availability might well benefit (i.e. enhance their lifetime reproductive success) by adjusting the reproductive threshold to local conditions. For example, suppose that conditions change such that the total energy stores for an organism always remain slightly below the existing threshold. An organism that adheres to the original threshold level thus never reproduces, and eventually dies through senescence or misadventure. In contrast, an organism that adjusts (i.e. lowers) its threshold in response to the lower resource level, will produce at least some offspring. Similarly, elevating the threshold in response to a short-term increase in resources may enhance post-reproductive energy stores (and, thus, maternal survival).
Snakes offer excellent model systems to investigate these questions; females of most species rely on existing energy stores to fuel reproduction and in many cases must accumulate reserves over several years prior to each reproductive episode (e.g. Seigel & Ford 1987; Brown 1991). Also, many snake species inhabit environments that provide spatially and temporally variable prey abundance, and there is evidence of year-to-year and place-to-place variation in reproductive frequencies in many such taxa (e.g. Glissmeyer 1951; Diller & Wallace 1984; Seigel & Ford 1987). Temporal variation in resource availability is particularly high in many Australian habitats, because of the dominant role played by the El Nino-Southern Oscillation system (e.g. Taylor & Tulloch 1985). Consequently, the Australian fauna and flora show diverse and striking adaptations that enable them to deal with this stochasticity (e.g. Newsome 1966; Flannery 1994). One such system involves water pythons (Liasis fuscus) in tropical Australia; our studies have revealed strong year-to-year fluctuations in the availability of prey (native rats), driven by annual climatic variability (Madsen & Shine 1996a,b, 1999) and have documented strong demographic responses of the pythons to these stochastic variations (Shine et al. 1997; 1998a). Thus, the water pythons provide a good system in which to examine whether the reproductive threshold in female snakes remains constant, or responds dynamically to resource fluctuations.
Materials and methods
Study species and area
Water pythons (Liasis fuscus) are large (to 3 m, 5 kg) non-venomous snakes that occur over a wide area of tropical Australia (e.g. Cogger 1992). This species is very abundant on the Adelaide River floodplain, 60 km east of Darwin in the Northern Territory. Our previous papers on this system have described the climate, topography and environment of the area around Fogg Dam, the main focus of our studies (e.g. Madsen & Shine 1996a). Briefly, the area is in the wet-dry tropics; it is hot year-round, but rainfall is concentrated in a 4-month wet-season (December to March), with resultant seasonal flooding. The snakes feed primarily on dusky rats (Rattus colletti Thomas 1904), a native plaguing rodent species that shows massive annual fluctuations in abundance depending upon local rainfall patterns (Redhead 1979; Madsen & Shine 1999). Reproductive rates of the pythons vary considerably among years, with rat abundance determining python feeding rates, body condition and the proportion of adult-size female snakes that reproduce each year (Shine & Madsen 1997). However, the clutch sizes of reproducing pythons show little annual variation, suggesting that the snakes delay reproduction until they have stored enough energy to support production of a full complement of eggs (Shine & Madsen 1997).
Over an 8-year period (1990–97), we captured pythons by spotlighting at night. Captured snakes were measured, weighed and individually marked before being released at their site of capture the following day. Their sex was determined by probing the hemipenial sacs in the tailbase. Reproduction is highly seasonal in this species (Madsen & Shine 1996b), and reproductive (pre-ovulatory, gravid or immediately post-ovipositional) females were easily recognizable by body shape during the months of July, August and September. Developing ovarian follicles and oviductal eggs can be detected through manual palpation. In some years, gravid females were retained in captivity (mean duration = 30·6 days) until they produced the clutch (Madsen & Shine 1996b; Shine et al. 1997), at which time they were reweighed and released. Females of this species do not feed while they are carrying eggs (Shine & Madsen 1997). The present paper is based on our data for body sizes (snout–vent length = SVL), masses, clutch sizes and egg sizes of 445 reproductive female pythons collected from July to September over 8 years.
We do not have experimental data on the ways in which energy stores influence a snake's ‘decision’ as to whether or not to reproduce. However, we have extensive information on the animals’ body condition over a long period of time, encompassing years when the pythons varied considerably in this trait (Shine & Madsen 1997). Also, we know that annual variation in body condition was closely linked to rat abundances and feeding rates of the pythons, and that much of the variation in body condition was attributable to energy storage in abdominal fat bodies (Shine & Madsen 1997). Hence, we can quantify the body condition of reproductive and non-reproductive snakes in each year of the study. If the reproductive threshold remains constant, we expect that the condition scores of reproductive snakes will remain fairly constant through time; whereas if the threshold changes in response to food availability, we expect to see considerable annual variation in the condition indices of reproductive, as well as non-reproductive snakes.
To obtain an index of body condition (mass relative to length) in pythons, we calculated residual scores from the general linear regression of (ln-transformed) mass to snout–vent length. This residual score will be positive for a fatter than usual snake and negative for a more-than-usually slender animal. We analysed the data using the software programs statview 4·5 and superanova (AbacusConcepts 1991, 1995) on a PowerMacintosh 9600/350 computer.
Adult female water pythons showed considerable annual variation in body condition over the period from 1990 to 1997 (Fig. 1). Importantly, this was true for reproductive as well as non-reproductive snakes (Fig. 1, using single-factor anova with year as the factor; for non-reproductive females, F7,638 = 29·0, P = 0·0001; for reproductive females, F7,515 = 41·4, P = 0·0001). The annual mean condition scores of reproductive females mirrored those of non-reproductive females, yielding a highly significant correlation in condition scores between the two groups of females (n = 8, R = 0·96, P = 0·0002).
Reproductive female snakes exhibited significantly higher condition indices (i.e. were more heavy-bodied) than non-reproductive females in every year of our study (Table 1 and Fig. 2). This consistency supports the notion of a ‘condition threshold’ that determines whether or not a female will become reproductive; otherwise, we would not expect to see such a consistent displacement in mean condition indices between the reproductive and non-reproductive animals. The position of this threshold (which must lie at a condition index intermediate between that of reproductive and non-reproductive animals: see Fig. 1) must have shifted among years, because the magnitude of annual variation in mean condition indices was so great (Fig. 1). Thus, for example, the mean condition of the reproductive females in 1997 (a year when most animals were very thin) was similar to that of the non-reproductive females in 1991 (a year when the snakes were much fatter, Fig. 1). Indeed, a direct comparison of these two groups of snakes shows that their body shapes were almost identical (Fig. 3; slopes F1,163 = 0·61, P = 0·44; intercepts F1,163 = 0·06, P = 0·81). That is, the snakes exposed to low feeding opportunities in 1997 initiated reproduction at a threshold body condition much lower than that in evidence in 1991 (when food was more plentiful). A threshold was apparent in both years (i.e. reproductive females were significantly heavier-bodied than non-reproductive females within each year, Fig. 2), but the position of the threshold shifted substantially. These two years were not atypical; for example, similar analyses of covariance showed that mass relative to body length of reproductive females in two other ‘poor’ years (1990 and 1996) was not significantly different from that of non-reproductive females in three other ‘good’ years (1991, 1992 and 1995; P > 0·07 in all comparisons).
Table 1. Comparisons between body shapes (mass relative to length) of reproductive vs.non-reproductive adult (>140 cm snout–vent length) female water pythons in each year of the study. The Table gives values from single-factor heterogeneity of slopes tests and ancovas, with reproductive status as the factor, ln snout–vent length as the covariate, and ln pre-ovipositional mass as the dependent variable. No ancova test was performed for data from 1997, because of the significant heterogeneity in slopes
Heterogeneity of slopes test
Equality of intercepts test
There are two plausible mechanisms by which the reproductive threshold could change so dramatically among years (Fig. 1). Either (1) individual females display remarkable plasticity in this respect and adjust their reproductive threshold to prevailing conditions; or (2) individual snakes retain invariant thresholds, but the composition of the reproductive population changes through time. For example, it might be true that groups of females (perhaps particular age groups or genetic subsets of the larger population) differ in thresholds, and also differ in the conditions that elicit reproduction. This phenomenon could engender patterns similar to those evident in Fig. 1. We can test between these two explanations with our data on individually marked females that were captured in two reproductive years. Three-hundred-and-ninety of our 516 records of the body condition of reproductive snakes (75%) refer to females that were only captured once as reproductive, but the remaining 126 records are from 55 females recaptured as reproductive in more than 1 year. The hypothesis of reproductive plasticity predicts that these individuals will vary in condition in the same way as the population as a whole; whereas the alternative notion suggests that these females should remain consistent in body condition each time that they reproduce.
Individual females do indeed vary considerably in condition from one reproductive episode to the next and follow the same trajectory in this respect as does the shift in populational mean values. That is, the condition of individual reproductive females changed in the same way as the population mean. To avoid pseudo-replication, we restricted our statistical analysis of this pattern to only one randomly selected record per snake, and only one comparison between any two particular years (e.g. 1990 vs. 1991, 1990 vs. 1992, 1990 vs. 1995). The resulting 21 comparisons reveal a highly significant correlation in between-year change in condition of individual reproductive females vs. the mean (Fig. 4a, n = 21, R = 0·84, P = 0·0001). We also compared these between-year changes in individual condition to the mean between-year changes for non-reproductive females; again, the correlation was highly significant (Fig. 4b, n = 21, R = 0·82, P = 0·0001).
Another way to evaluate whether the among-year variation in reproductive thresholds results from plasticity by individual females, is to look in more detail at individual responses between particular years when the overall population mean ‘reproductive threshold’ increased (e.g. from 1994 to 1995), was approximately stable (e.g. from 1991 to 1992) or decreased (e.g. from 1992 to 1993: see Fig. 1). Again, the strong result is most females changed their ‘reproductive threshold’ in the same way as the population mean (Fig. 5). Thus, the data presented in Fig. 4 and 5 provide strong support for the ‘individual plasticity’ hypothesis.
Figure 5 also reveals considerable variation among females in their ‘reproductive thresholds’. The individual threshold appears to be relatively constant over time relative to the population mean; a female that exhibits high body condition in one reproductive episode will be similarly heavy-bodied relative to other females when she reproduces again (Fig. 5). To test the statistical significance of this effect, we calculated residual scores from a general linear regression of the body condition score for an individual female vs. the mean body condition for all reproductive females in that year; thus, we could factor out the effect of annual variation in this parameter. A one-factor anova with female number as the factor revealed significant variation among females (F54,125 = 2·20, P = 0·001). Thus, although mean values for body condition in the entire population fluctuate considerably through time and individual females change appreciably in this respect, their body condition, nonetheless, tends to remain consistent relative to that of other females.
The mass of a reproductive female snake consists of two components: the clutch and the female's body. Thus, there are two pathways by which individual females could change their pre-oviposition body condition (i.e. mass relative to length) from one reproductive episode to the next: the change may involve either (or both) the clutch mass, or the female's body mass exclusive of the clutch. In turn, we can imagine two possible scenarios for a reproducing female python in poor condition. First, she could produce a ‘normal’ clutch, with the inevitable consequence that her body reserves after laying will be substantially reduced (i.e. she will be more-than-usually emaciated after laying). The second possibility is that she will reduce her clutch size, but maintain her post-oviposition condition.
We can examine these possibilities with data on body condition, mean egg size and the clutch size of the wild-caught gravid females that laid their eggs in captivity, shortly after capture, in 5 years of our study (1991, 1992, 1993, 1995 and 1997). The clear result from these analyses (below) is that female pythons utilize both of these pathways, but that the reduction in clutch size is minor compared to the reduction in body condition.
Body condition after oviposition
Female pythons were in much poorer condition (i.e. weighed less relative to snout–vent length) in ‘bad’ years than in ‘good’ years. This is evident when we compare the condition of these captives between years when the females were in good condition (1991, 1992 and 1995) with years when they were in poor condition (1993 and 1997, see Fig. 1). ‘Bad’ years were associated with much lower condition of the female after laying (one-factor anova with ‘good’ vs. ‘bad’ year as factor, F1,139 = 14·46, P = 0·0002), as well as prior to oviposition (see Fig. 1; one-factor anova, F1,139 = 17·87, P = 0·0001). This result reflects the high overall correlation between body condition scores of these pythons prior to vs. after oviposition (n = 140, R = 0·88, P = 0·0001); that is, female pythons that were unusually slender-bodied prior to oviposition were also unusually thin (relative to other females) after laying their eggs. This correlation suggests that clutch masses relative to maternal size remained relatively consistent among years, whereas the degree of maternal emaciation was more variable.
Do female pythons in relatively poor condition produce smaller and/or fewer eggs than females in good condition? We evaluated this possibility by comparing egg sizes and clutch sizes to the degree of maternal emaciation after laying. Because clutch size depends to a large degree on maternal body size (Madsen & Shine 1996b), we factored out variation due to maternal size by calculating residual scores from the general linear regression of clutch size against maternal snout–vent length. We quantified maternal body condition in the same way, using post-ovipositional rather than pre-ovipositional maternal mass (because the mass of the eggs is included in the latter measure, thus generating an artifactual correlation). The analysis revealed a significant positive correlation between female post-ovipositional condition and relative clutch size (Fig. 6a, n = 140, R = 0·43, P = 0·0001). That is, female pythons that are in relatively good body condition after laying, tend to produce larger clutches than do females in poor condition. Will the female's body condition also affect the mass of her eggs? We found that, indeed, the mean egg mass did shift with maternal condition after laying: mean egg masses of highly emaciated females were slightly lower than those of females in better condition (Fig. 6b, n = 121, R = 0·20, P = 0·027; note that sample sizes are lower than for the comparison of clutch sizes, because of missing data for egg masses from some clutches).
Do these relationships translate into year-to-year variation in reproductive output, associated with the annual variation in reproductive threshold? To evaluate this possibility, we compared mean egg mass between years when the females were in good condition (1991, 1992 and 1995) with years when they were in poor condition (1993 and 1997, see Fig. 1). A one-factor anova (with ‘good’ vs. ‘bad’ year as the factor) revealed no significant difference in mean egg mass (F1,120 = 0·15, P = 0·69). Similarly, we compared relative clutch sizes (i.e. number of eggs relative to maternal body length) between years when the females were in good condition with years when they were in poor condition (Fig. 1). A one-factor ancova (with ‘good’ vs. ‘bad’ year as the factor, maternal snout–vent length as the covariate and clutch size as the dependent variable) confirmed a small, but statistically significant decline in brood size (relative to maternal body length) when females were in poor condition (slopes homogeneous –F1,136 = 0·63, P = 0·43; intercepts –F1,137 = 5·1, P = 0·026, Fig. 7). Thus, the lower reproductive condition indices for pre-ovipositional female pythons in ‘bad’ years (Fig. 1) is partly due to a slight reduction in brood size and egg mass, but mostly reflects a decrease in maternal reserves (i.e. carcass mass post-oviposition). The end result of decreases in these two traits—clutch mass as well as maternal carcass mass—is that the ratio of these two factors (relative clutch mass = RCM) did not differ between ‘good’ vs. ‘bad’ years (RCM based on the ratio of clutch mass to female post-oviposition mass, one factor anova with ‘good’ vs. ‘bad’ year as factor, F1,135 = 0·29, P = 0·59).
Like many other kinds of organisms that depend upon stored reserves for reproduction (‘capital-breeders’), female pythons in our study population do not reproduce unless they have attained a threshold level of body-condition (and, thus, energy reserves). However, we are unaware of any previous demonstration that this threshold value changes with time, in concert with (and presumably, in response to) temporal variations in prey availability. Although the current paper does not include data on prey abundance, our previous analyses have provided strong evidence for a causal link between rat numbers and python condition, mediated via feeding rates of the snakes (Shine & Madsen 1997). Thus, the water pythons provide the first clear evidence of the dynamic nature of this body-condition threshold for breeding. Although the mechanism for this adjustment remains unknown, it seems that the snakes somehow assess prey availability and modify their reproductive output accordingly.
Why do female water pythons adjust their reproductive thresholds to prevailing conditions? A lowering of the reproductive threshold corresponds to an increase in reproductive effort, because more emaciated females may be less likely to survive to breed again (e.g. Madsen & Shine 1992, 1995, 1998b). Mathematical models that link survival, growth and lifetime reproductive success indicate that optimal reproductive effort increases when: (i) the female has a low probability of survival to the next reproductive opportunity; and (ii) she will grow relatively little over that period, so that her clutch size will not be elevated by a larger maternal body size (Shine & Schwarzkopf 1992). Low rat availability is likely to have both of these effects on water pythons. Years when rats are scarce are characterized by relatively low growth rates and reduced survival rates of the pythons (Madsen & Shine 1998a). Thus, a female python that encounters low prey availability may well enhance her lifetime production of offspring by lowering her reproductive threshold; the ‘cost’ (to her survival and/or later growth) may well be less than the ‘benefit’ (i.e. she will produce at least one clutch during her lifetime). In contrast, a relatively high reproductive threshold (ensuring good maternal condition after laying) may well be the optimal tactic when prey numbers are high.
The shift in maternal condition after laying, as well as in reproductive output (clutch size), shows that the snakes must somehow change the ‘rules’ they use to allocate reproductive expenditure relative to energy storage. At first sight, it might seem that the kind of information-processing that would be required would be too sophisticated for a snake. However, female snakes already rely upon linking their reproductive output to a factor that is a variable rather than a constant: maternal body size. As a female grows larger, her clutch size increases (e.g. Fitch 1970). We do not know the mechanism for this effect, but it certainly involves matching fecundity to body size. It is only a small extra step to take body condition into account also.
What is the sampling period over which the female assesses resource levels? Our data suggest that it may be greater than one year. The most interesting comparison is between 1996 and 1997, 2 years of very low rat abundances and consequently, poor body condition for the pythons (Fig. 1). Females were equally emaciated in both years, but the proportion of females that reproduced increased dramatically from 1996 (20%) to 1997 (45%) (χ2 = 9·93, 1 d.f., P < 0·002). This shift suggests that most females were reluctant to substantially lower their reproductive threshold after just 1 year of low food availability (1996), but were prepared to do so after two such years in succession (1996 and 1997). A recent experimental study has shown that females of a long-lived lizard species modify their reproductive allocation in the current year, based upon their experience of resource availability during the previous breeding season (Doughty & Shine 1998). The water pythons of Fogg Dam may well be doing the same.
The ability to adjust reproductive thresholds in this way may be widespread, especially in species that experience substantial temporal variation in resource availability. Previous studies provide evidence of exactly this kind of adjustment; for example, Seigel & Fitch (1985) documented annual variation in clutch size relative to maternal body size in several snake species and found that years of higher fecundity in one species (Diadophis punctatus Linné 1766) were those with weather conditions (high rainfall) that increased the snakes’ opportunities to forage. The parallel with water pythons (where rainfall drives rat availability, Madsen & Shine 1999) is very strong, even though the system used by Seigel and Fitch (small fossorial earthworm-eating colubrid snakes in Kansas) is very different to the one that we have studied in Australia.
Experimental studies provide further evidence of facultative shifts in snake reproduction. When exposed to lowered food availability in captivity, some snake species respond by reducing clutch size, but not maternal condition after laying (e.g. Elaphe guttata Linné 1766, Seigel & Ford 1991), whereas others reduce both reproductive output and maternal post-oviposition condition (e.g. Thamnophis marcianus Baird & Girard 1853, Ford & Seigel 1989). Reproductive frequency was unaffected in both these cases (i.e. all females reproduced), but data from field studies on other snake species suggest that a decreased proportion of females will reproduce when food supply is reduced (e.g. Diller & Wallace 1984). Offspring size remains unaffected by maternal nutrition in most species investigated to date (Seigel & Fitch 1985; Ford & Seigel 1989; Seigel & Ford 1991; Gregory & Skebo 1998), but increases with higher food supply in at least one taxon (Andren & Nilson 1983). Clutch sizes change in response to variations in prey availability in several lizard species (e.g. James & Whitford 1994) and vary through time in some populations of snakes (e.g. Seigel & Fitch 1985). In yet other snakes, reproductive traits are relatively stable from year to year despite considerable variation in weather conditions (e.g. Plummer 1983). This diversity is consistent with life-history theory; the optimal response to lowered resource levels will depend on the demography of the population and on the probable duration of the resource shortage relative to the reproductive lifetime of a female snake.
The discovery that reproductive thresholds of water pythons are flexible and change according to prey availability, builds upon our earlier conclusions concerning the effects of annual rodent fluctuations on python feeding rates, body condition and reproductive output (Shine & Madsen 1997). In the earlier study, we showed that the proportion of adult-sized female pythons reproducing each year varies as a function of prey availability, but that reproductive output per clutch remains relatively invariant. Our more detailed analyses in the current paper reveal subtle variation in reproductive output per clutch, but much greater variation in reproductive thresholds and, thus, in maternal body condition after laying. Nonetheless, it is important to note that the plasticity in these traits is not so great as to obscure the overall relationship between prey availability and the proportion of female pythons that reproduce in a given year (Shine & Madsen 1997). In species in which all adult females reproduce each year, annual variation in resource availability might engender much greater shifts in reproductive output and maternal body condition (e.g. Ford & Seigel 1989); whereas in the water pythons, a considerable proportion of adult females do not reproduce if their energy reserves are too low.
Overall, our results fit well with emerging paradigms on the importance of energy storage as an important dimension of life-history variation among species and among years within single populations. Reproducing animals adjust their allocation ‘decisions’ in surprisingly complex and sophisticated ways, and we expect that long-term studies on reproductive ‘tactics’ will provide many additional examples of the flexible adjustment of energy resources between storage and reproduction.
We thank G. Bedford, B. Cantle, P. Fisher, P. Harlow, P. Osterkamp and J. Osterkamp for field assistance, and K. Levy for logistical support. The study was funded by the Australian Research Council.
Received 16 April 1998;revisionreceived 24 August 1998