Sunny side up: lethally high, not low, nest temperatures may prevent oviparous reptiles from reproducing at high elevations

Transcription

1 Blackwell Science, LtdOxford, UKBIJBiological Journal of the Linnean Society The Linnean Society of London, Original Article R. SHINE ET AL. THERMAL CONSTRAINTS ON REPTILE OVIPARITY Biological Journal of the Linnean Society, 2003, 78, With 3 figures Sunny side up: lethally high, not low, nest temperatures may prevent oviparous reptiles from reproducing at high elevations RICHARD SHINE*, MELANIE J. ELPHICK and ELIZABETH G. BARROTT Biological Sciences A08, University of Sydney, NSW 2006, Australia Received 5 May 2002; accepted for publication 6 September 2002 Oviparous (egg-laying) lizards and snakes generally inhabit warmer climates than do related viviparous (livebearing) taxa. This pattern is widely attributed to the failure of oviparous reproduction in cold climates, but the thermal regimes of potential nest-sites above and below the elevational cut-off for oviparous reproduction have never been quantified. We studied oviparous (Bassiana duperreyi) and viviparous (Eulamprus heatwolei) scincid lizards at such a site in the Brindabella Range of south-eastern Australia. Miniature data-loggers monitored temperatures of nest-sites and lizards in midsummer, partway through the incubation period of eggs in natural nests. Our results contradict the simplistic notion that mean nest temperatures determine this elevational limit for oviparity. Instead, potential nest-sites with average temperatures suitable for embryogenesis in Bassiana are available well above the threshold elevation. However, thermal minima decrease consistently with elevation and thus the maximum temperature needed for any given mean incubation temperature increases rapidly with elevation. Potential nest-sites above the elevational threshold can only attain mean temperatures high enough to sustain embryogenesis by having lethally high thermal maxima. Such nest-sites are available close to the soil surface, but cannot support development. In contrast, behavioural thermoregulation allows viviparous lizards to maintain high mean body temperatures concurrently with relatively low maximum temperatures, regardless of elevation. Paradoxically, oviparous reptiles may be restricted to low elevations not because nests that provide appropriate mean incubation temperatures are unavailable further up the mountain, but because eggs laid in such shallow nests would overheat The Linnean Society of London, Biological Journal of the Linnean Society, 2003, 78, ADDITIONAL KEYWORDS: geographical distribution lizard oviparity thermal viviparity. INTRODUCTION Consistent associations between environmental factors and life-history traits, especially if present in many distantly related lineages of organisms and in many different parts of the world, provide some of the strongest evidence for Darwinian interpretations of life-history diversity (e.g. Harvey & Pagel, 1991). Many such patterns have been identified, and in earlier years were often accorded the status of laws or rules. For example, colder climates are associated with larger body size in endotherms (Eisenberg, 1981), and a higher incidence of brooding in marine invertebrates (Gallardo & Penchaszadeh, 2001). Similarly, *Corresponding author. rics@bio.usyd.edu.au habitat productivity has been identified as a correlate of the incidence of polygyny (Emlen & Oring, 1977), communal breeding (Brown, 1987) and clutch sizes (Geffen & Yom-tov, 2000) in birds. Although some of these patterns are striking, they are rarely so clear-cut as to permit direct identification of threshold conditions: that is, where a specific life-history strategy is viable in one area but fails in an adjacent site. The association between reptilian reproductive modes (oviparity vs. viviparity) and climate (warm vs. cold) offers an exceptional opportunity in this respect. Oviparous (egg-laying) lizards and snakes are restricted to relatively warm areas, with the proportion of viviparous (live-bearing taxa) increasing dramatically with increasing elevation or latitude (e.g. Mell, 1929; Weekes, 1933; Shine & Berry, 1978). This association is extraordinarily consistent across conti- 325

2 326 R. SHINE ET AL. nents (e.g. it is seen in Africa, Asia, Australia, North and South America: Tinkle & Gibbons, 1977) and across diverse reptilian lineages representing >100 separate evolutionary transitions from oviparity to viviparity (Blackburn, 1982, 1985; Shine, 1985). The reason for this association presumably involves thermal biology. In severely cold areas, viviparity allows developing eggs (retained in utero) to be kept warm by their mother s thermoregulatory behaviour, and thus to avoid detrimentally low nest temperatures (Mell, 1929; Sergeev, 1940; Shine, 1983, 1985). Oviparity is precluded in such areas because soil temperatures are too low to permit successful embryonic development and hatching. Despite the virtually universal acceptance of this model to explain the absence of oviparous squamates from severely cold climates, there are remarkably few empirical data on the temperatures of potential nestsites in such regions (Andrews, 2000). Such data are essential if we are to evaluate reasons for the absence of egg-layers in cold environments. To understand how the distribution of oviparous reptiles is constrained by cold, we need to compare potential nest-sites immediately above and below the thermal threshold of a species distribution. This is difficult in practice, because distributional limits are rarely so clear-cut. The task can be facilitated by focusing on elevation rather than latitude as the determining factor, because of the steeper thermal clines in the former situation. Ideally, then, we need to find the upper distributional limit in an area where similar habitat features (and thus, potential nest-sites) extend into higher, colder regions than those occupied by the study species. We have found such an area in the Brindabella Range of south-eastern Australia. An oviparous scincid lizard, Bassiana duperreyi, is abundant at lower elevations but absent from otherwise similar areas >1630 m (Pengilley, 1972). At the highest locality we have recorded for this taxon (Mount Ginini), the lizards nest in large numbers at the lower end of a ski run, but are never seen further up the run (pers. obs.). Extensive searching over four years has revealed >1000 eggs (corresponding to the clutches of >200 females) at elevations <1630 m on the site, but no eggs above this point. The ski-run extends directly up the north-east side of the mountain to 1762 m, with no obvious habitat discontinuity between the sites occupied by Bassiana and those further up. This location thus provides an ideal opportunity to compare the thermal regimes of nest-sites used by Bassiana to those potentially available above the upper elevational limit of this taxon. Another distantly related oviparous lizard species also breeds at the bottom of Mount Ginini but no higher (the agamid Tympanocryptis diemensis) whereas at least five live-bearers are found all the way up to the summit (Egernia whitii; Eulamprus heatwolei; E. tympanum; Pseudemoia entrecasteauxii; Tiliqua nigrolutea: pers. obs.). This distributional pattern suggests that reproductive mode is causally involved in whatever mechanism prevents Bassiana from reproducing successfully at higher elevations. In support of this inference, Bassiana eggs that were translocated to artificial nest-sites at higher elevations on Mount Ginini exhibited lower hatching success than eggs maintained at lower elevations (Shine, 2002a). Why are these higher-elevation nests unsuitable for oviparous reproduction? We conducted a study to quantify thermal regimes above and below the upper elevational limit for Bassiana. In particular, we ask: (1) Does the elevational difference simply involve a shift in mean incubation temperatures, such that potential nest-sites above the cut-off are too cool to permit embryonic development? This hypothesis predicts that the threshold elevation should correspond to the mean nest temperature below which hatching will not occur (Sergeev, 1940) or that hatching will occur but produce suboptimal hatchling phenotypes (Shine, 1995). (2) Does thermal variance in potential nest-sites shift with elevation also? Embryogenesis in Bassiana is affected by thermal variance as well as mean temperature, and females select nest-sites that display high thermal variance as well as high mean temperatures (Shine & Harlow, 1996; Shine, Elphick & Harlow, 1997). On the other hand, embryos would be killed if temperatures fell above or below their tolerance levels (Muth, 1980; Packard & Packard, 1988). It is thus of interest to examine elevation-related shifts in maximum and minimum as well as mean nest temperatures, and the relationship between these variables. (3) How do thermal regimes experienced by embryos inside viviparous reptiles shift with elevation? Because viviparous reptiles extend all the way to the top of Mount Ginini, we can compare the thermal regimes experienced by uterine embryos in such taxa to those experienced by eggs of an oviparous form at the same location. Again, both the mean and variance of incubation temperatures are of interest. To answer these questions, we gathered data on thermal regimes at a range of elevations for: (1) An array of potential nest-sites, including exposed as well as heavily sheltered locations. These sites allowed us to quantify conditions in all available nesting locations, including ones that were less wellinsulated (and thus, with higher and more variable temperatures) than the sites actually used; also, sites that occurred at higher elevations than any actual nests.

3 THERMAL CONSTRAINTS ON REPTILE OVIPARITY 327 (2) Actual nest-sites of Bassiana over a range of elevations. (3) Gravid females of the viviparous lizard Eulamprus heatwolei, over a range of elevations. As for potential nest-sites, these locations included areas above the upper elevational limit for oviparous reproduction, but within the usual range of viviparous species. MATERIAL AND METHODS STUDY SPECIES AND AREA The three-lined skink, Bassiana duperreyi, is a medium-sized (to 80 mm snout-vent length) insectivore that is abundant in montane areas of southern and eastern Australia (Cogger, 1992). Because of the relatively short warm season in montane areas, oviposition and hatching are highly synchronized in oviparous species. Female Bassiana lay a single clutch of 3 9 eggs in early summer (December), usually under logs or rocks (Pengilley, 1972). The eggs hatch 2-3 months later, shortly prior to the onset of severely cold weather in autumn (Pengilley, 1972; Shine, 1983). Many clutches are communal, containing the eggs of up to 20 females (Pengilley, 1972; Shine & Harlow, 1996). The slightly larger viviparous lizard Eulamprus heatwolei (highland water skink; to 100 mm SVL) is sympatric with Bassiana over a large area, but extends into higher colder regions as well (to >1800 m: Pengilley, 1972). We have studied both species extensively in the Brindabella Range 40 km west of Canberra, Australian Capital Territory (Shine, 1983, 1995; Shine & Harlow, 1993, 1996). The climate is hot in summer (January mean temperature 25.9 C) but cold in winter (July mean 10.4 C) with intermittent snow cover (Green & Osborne, 1994; Australian Bureau of Meteorology). Nesting by Bassiana is concentrated in open areas among eucalypt forests. Many of these open areas are anthropogenically derived (especially, clearways cut for powerlines, and ski-runs at higher elevations: Shine et al., 1997; Shine, Barrott & Elphick, 2002a). POTENTIAL NEST-SITES In spring (November) 2000, we laid out sets of four grey concrete pavers (Boral Masonry: cm, 10 cm thick) on flat ground at each of four locations: Picadilly Circus (1240 m asl; E, S) and on the lower (1615 m), middle (1660 m) and upper (1720 m) slopes of Mount Ginini ( E, S). The Picadilly site was in a 60 m wide anthropogenic clearing under powerlines, whereas the Ginini site was a 50-m-wide ski-run. The replicate pavers within each site were spaced 7 29 m apart (depending on proximity of trees), and each was partially dug into the underlying soil so that it was firmly embedded. Thermal regimes in natural nests of Bassiana are affected not only by their elevation, but also by their degree of exposure to solar radiation, and by the size and type of cover items (logs, rocks: Shine et al., 1997; Shine, Barrott & Elphick, 2002). To reduce the effects of this latter confounding factor, we used standardized cover items (above). These pavers have similar thermal characteristics to natural cover items, and are readily used as nest-sites by Bassiana (unpubl. data). To standardize exposure to sunlight, all pavers were in open areas (as is typical for all natural nests of Bassiana: Shine et al., 1997, 2002a, b). Pavers at the two lower localities (Picadilly and the base of Mount Ginini) were <1 m from natural nests of Bassiana, whereas those at higher elevations were placed in habitats and orientations that closely resembled natural nesting areas. To verify that sun exposure was similar along the elevational gradient, we took hemispherical (180 ) photographs with a 35-mm camera and fisheye lens (Canon F1 7.5 mm) placed on a paver and pointing directly upwards. The resultant photographic records of the size, shape and location of gaps in the forest canopy were scanned and the digital images analysed using GLA software (Gap Light Analyser v.2.0, Frazer, Canham & Lertzman, 1999). This program calculates attributes such as canopy structure and gap light transmission exposure. We also took similar photographs at natural nest-sites, to compare our pavers with natural nests in these respects. The GLA calculations confirmed that our pavers were exposed to similar radiation intensity as were natural nests. The four pavers at the base of Mount Ginini did not differ significantly from eight nearby natural nests in the percentage of open sky (vs. canopy) above them (means 65 vs. 61%, F 1,9 = 1.26, P = 0.29), in the duration of direct sunlight falling on the paver each day (means 577 vs. 612 min d -1 ; F 1,9 = 1.14, P = 0.31) or in total incident radiation (means 39.0 vs mol m -2 d -1 ; F 1,9 = 0.23, P = 0.64). At Picadilly, the pavers were in slightly more open areas than the nests (because some nests at this site were partially shaded), but overlapped considerably (% open canopy, means 76 vs. 56%; F 1,18 = 19.60, P < 0.001; duration of direct sunlight, means 711 vs. 577 min d -1, F 1,18 = 7.18, P < 0.02; total incident radiation, means 43.0 vs mol m 2 d -1 ; F 1,18 = 7.39, P < 0.02). Comparing among pavers only, canopy openness was greatest for the highest-elevation site (near the top of Mount Ginini, 89%), next greatest for the lowest site (Picadilly, 76%) and lowest for the intermediate sites (63 and 65%; F 3,11 = 10.54, P < 0.002). The duration of direct sunlight and total incident radiation showed the same pattern (mean values for duration = GT 759, PC 718, GB 586, GM 551 min d -1, F 3,11 = 76.98, P < ; mean values for total radia-

4 328 R. SHINE ET AL. tion = GT 43.5, PC 43.0, GM 39.3, GB 39.0 mol m -2 d -1 : F 3,11 = 20.36, P < ). Thus, canopy openness, duration of direct sun and total incident radiation varied among sites, but did not change consistently with elevation. The most important of these variables for thermal regimes (total incident radiation) varied only slightly among elevations (mean values mol m -2 d -1 ). Thermal data-loggers (Thermochron ibutton, Dallas Semiconductor, Dallas, Texas, USA; diameter 15 mm, height 6 mm, mass 3.3 g; and Hobo-temp H8 and XT models with external leads, Onset Computer Co., Massachusetts) were used to record temperatures every 15 min at three locations associated with each paver: on the paver s upper surface (and thus exposed to direct sunlight for most of the day), directly under the paver (where eggs would usually be laid) and 30 cm deep underground directly beneath the paver (simulating a nest much deeper than any we have recorded in the Brindabella Range: Shine et al., 1997). These locations were chosen to encompass the range of potential nest-sites available to lizards at each site, in terms of the thickness of cover items. The exposed thermochron measured regimes that would be experienced by extremely superficial nests whereas the underground probe measured conditions for eggs buried deep under the soil surface. Physical attributes of an exposed thermochron (size, colour, etc.) have only a minor influence on its thermal regimes (Vitt & Sartorius, 1999; Shine & Kearney, 2001). Data on temperatures at these locations were gathered during the period January 2001; the weather was fine and warm throughout this period (mean daily maxima 31.4 C, minima 14.1 C). Because of high temporal correlation of temperatures within any given nest, even a brief sample such as this should provide a robust indication of thermal regimes experienced during the incubation period (Shine & Harlow, 1996; Shine et al., 1997, 2002a). Even if this assumption is violated, comparisons between thermal conditions measured simultaneously at different elevations should be reliable. BODY TEMPERATURES OF LIZARDS To monitor the body temperatures of viviparous female lizards at the same sites over the same time period, we erected two circular open-topped nylon arenas ( Space Pop, Smash Enterprises, Melbourne; 48 cm diameter, 56 cm deep) at each elevation. Damp potting mix was provided as a substrate, with logs for shelter and to provide an elevated basking platform. Crickets were provided as food. We placed two recently captured gravid female water skinks (Eulamprus heatwolei) in each arena after attaching thermochron data-loggers to the mid-dorsal surface of each animal using super-glue (Loctite 406, Loctite Australia, Caringbah, NSW, Australia). To reduce mass of the data-loggers, they were removed from their metal canisters and plastic-dipped before being attached to the animal (Robert & Thompson, 2002). The final package weighed <1.5 g, whereas the lizards averaged 13.9 g. The animals showed no overt effect of the thermochrons, and moved about freely within the arenas. The data-loggers were set to record temperatures every 10 min. After 6 days, the thermochrons were removed and the lizards released at their initial sites of capture. Calibration trials show that the temperatures recorded by external thermochrons are very similar to internal body temperatures of the lizards carrying them (Robert & Thompson, 2002). NATURAL NESTS Over a 7-year period ( to ), we placed thermal data-loggers in a total of 170 natural Bassiana nests, spread among three elevations. These comprised 75 nests at Coree Flats (1050 m asl; E, S), 82 nests at Picadilly Circus (1240 m) and 13 nests on the lower slopes of Mount Ginini (1615 m). Temperatures were monitored from early December (<1 week after laying) to hatching (generally in March/ April). We calculated mean, minimum and maximum temperatures for each nest over the entire incubation period. DATA ANALYSIS Data were analysed using Statview 5.0 and SuperANOVA 1.1 on an Apple Macintosh G4 computer. Assumptions of statistical tests were checked prior to analysis. The assumption of equal variances was violated for temperature data collected at potential nest-sites (at every elevation, thermal regimes deep underground were less variable than for exposed probes, and no transformation could remedy the problem). We thus used a non-parametric test (Kruskal Wallis) to examine these data. To quantify the relationship between mean and maximum temperatures for potential nests at each site, we used the combined data for environmental temperatures (i.e. ground surface plus under paver plus 30 cm underground) at each elevation. This procedure was designed to assess the range of thermal conditions potentially accessible to a nesting female lizard at any given elevation. RESULTS Sample sizes for some thermal measurements were reduced by data-loggers failing to record, being displaced by itinerant wombats, or becoming detached from lizards. Thus, we obtained thermal data on three

5 THERMAL CONSTRAINTS ON REPTILE OVIPARITY 329 lizards from three sites but only one lizard from the bottom of Mount Ginini. We obtained either three or four data-sets for all environmental variables at all locations except for 30 cm underground on the middle slope of Ginini (one data-set only). Nonetheless, variation among replicates within locations and sites was small (see below), so these losses will have little effect on overall patterns. POTENTIAL NEST-SITES Thermal profiles showed considerable diel variation. Soil temperatures on the ground surface reached >45 C during the day and fell below 15 C at night (Fig. 1). Deep-soil temperatures showed the least fluctuation, and nest temperatures (i.e. those under pavers) were intermediate in this respect. Because heating rates were slow, temperatures under pavers did not peak until late afternoon (Fig. 1). At every elevation, deep-soil probes recorded lower mean temperatures, lower maximum temperatures, higher minimum temperatures, and less thermal variation than did probes in more exposed situations (Fig. 2; Kruskal Wallis tests show P < 0.05 for every such test). Temperatures under pavers (potential nest sites) were intermediate between deep-soil probes and exposed data-loggers in all of these respects (Fig. 2). The effects of elevation were weaker than these location-within-elevation effects. Mean temperatures Temperature o C lizard ground surface nest underground Time (h) Figure 1. Diel variation in temperatures on a fine warm day (15 January 2001) at 1660 m asl (middle slopes of Mount Ginini) in the Brindabella Range. The graph shows output from miniature data-loggers that were (a) attached to the dorsal surface of a gravid female of a viviparous lizard species (Eulamprus heatwolei) in an outdoor arena, and thus free to thermoregulate ( lizard ); (b) glued to the upper surface of a grey concrete paver ( ground surface ); (c) placed under the paver, where eggs would typically be laid ( nest ); or (d) buried 30 cm deep under the paver ( underground ). decreased at higher elevations (for deep-soil temperatures, F 3,7 = 4.91, P < 0.04; for nest, F 3,11 = 22.04, P < ; for exposed, F 3,12 = 12.33, P < 0.001). Standard deviation in temperature was not affected by elevation (all P > 0.34), nor was maximum temperature (all P > 0.05). Minimum temperatures were much lower at higher elevations for deep-soil probes (F 3,7 = 8.79, P < 0.01) but this effect was weaker for nest temperatures (F 3,11 = 3.46, P = 0.055) and not evident for exposed (soil-surface) temperatures (Fig. 2; F 3,12 = 0.80, P = 0.52). To summarize, mean and minimum environmental temperatures were generally lower at higher elevations, whereas maximum temperatures were unaffected (Fig. 2). Data on potential nest-sites (i.e. combining readings from all data-loggers, including ground surface as well as under pavers and well-buried) revealed strong associations between mean, minimum and maximum temperatures. At each elevation, potential nest-sites with higher mean temperatures also had higher maxima (Picadilly, N = 11, r = 0.91, P < 0.001; Ginini bottom, N = 10, r = 0.95, P < 0.001; Ginini middle, N = 9, r = 0.97, P < 0.001; Ginini top, N = 12, r = 0.99, P < ). These relationships between mean and maximum temperatures differed among elevations, with a given mean value corresponding to a higher maximum value at higher elevations than at lower elevations (ANCOVA slopes F 3,34 = 0.35, P = 0.79; intercepts F 3,37 = 20.78, P < ). The opposite pattern was evident for the relationship between mean and minimum temperatures within each elevation. Potential nest-sites with higher mean temperatures had lower not higher minima (Picadilly, N = 11, r = 0.88, P < 0.001; Ginini bottom, N = 10, r = 0.83, P < 0.004; Ginini middle, N = 9, r = 0.77, P < 0.02; Ginini top, N = 12, r = 0.93, P < ; comparing these relationships among elevations, ANCOVA slopes F 3,34 = 1.54, P = 0.22; intercepts F 3,37 = 10.45, P < ). This counter-intuitive result reflects differences among sites in exposure: well-insulated sites (such as those 30 cm underground) had high minima and low maxima, whereas more exposed sites (such as those on the ground surface) had low minima and high maxima. The negative correlations between mean and minimum temperatures indicate that average values were determined primarily by maxima rather than minima, in turn due to the much greater range of maximum than minimum temperatures within each site (typically about fourfold greater, with ranges of 40 vs. 10 C). This analysis shows that minimum soil temperatures decline with increasing elevation (Fig. 2), and that potential nest-sites within any given elevation display only a modest range of variation in minimum temperatures. In contrast, variation in factors such as cover-item thickness and the degree of exposure to

6 330 R. SHINE ET AL. Mean temperature ( o C) Minimum temperature ( o C) 27.5 a c Elevation (m) lizard ground surface nest underground Figure 2. Thermal regimes at four elevations in the Brindabella Range. The Figures show means and associated standard errors calculated over a period of six days (11 16 January 2001) for (a) mean temperature (b) the standard deviation of mean temperature (c) minimum temperature, and (d) maximum temperature, as measured by miniature data-loggers in four places. These units were either attached to dorsal surfaces of highland water skinks in outdoor arenas ( lizard ); glued to the upper surface of a grey concrete paver ( ground surface ); placed under the paver, where eggs would be laid ( nest ); or buried 30 cm deep under the paver ( underground ). See text for statistical analysis of these data. solar radiation engender much greater variation in maximum temperatures. An inevitable consequence of these two results is that the only way that a highelevation nest-site can exhibit a high mean temperature is to have a very high maximum temperature. In practice, this will characterize a nest-site close to the soil surface, with high exposure to solar radiation. In contrast, the higher minimum temperatures of lowerelevation sites allow the same mean value with a lower maximum, and hence may permit successful embryonic development under a thicker or more shaded cover-object. The central result is that all else being equal, any given mean temperature will be associated with a higher maximum temperature in a highelevation than in a lower-elevation nest-site. BODY TEMPERATURES OF LIZARDS Unlike environmental regimes, the body temperatures of female Eulamprus were largely unaffected by Maxiumm temperature ( o C) Standard deviation ( o C) b d Elevation (m) elevation (Fig. 2; for mean temperature, elevation effect F 3,6 = 2.05, P = 0.21; for minimum temperature, F 3,6 = 0.32, P = 0.81; for maximum temperature, F 3,6 = 1.60, P = 0.29). The sole exception was standard deviation, which was slightly lower at the bottom of Ginini than elsewhere (Fig. 2; F 3,6 = 6.21, P = 0.03; but no posthoc tests were significant, and the aberrant location was the one that was represented by only a single individual). Thus, the overall pattern was that lizard body temperatures were similar across the different elevations, whereas environmental temperatures were not. NATURAL NESTS Despite the 565 m range in elevation across our three study sites ( m), our 7-year data set showed that natural nests at Coree Flats, Picadilly Circus and at the base of Mount Ginini exhibited virtually identical mean temperatures (19.4, 19.7, 19.8 C, respec-

7 THERMAL CONSTRAINTS ON REPTILE OVIPARITY 331 tively; one-factor ANOVA on ln-transformed data to remove variance heterogeneity, F 2,167 = 0.99, P = 0.38). However, mean maxima were higher at the higherelevation sites (32.2, 34.5, 37.2 C; F 2,167 = 8.39, P < ), and mean minima were lower (13.1, 12.1, 10.3 C; F 2,167 = 22.49, P < ). That is, nests at the highest elevation (Mount Ginini) exhibited similar mean temperatures to lower-elevation nests, despite significantly lower night-time minima, by achieving higher daytime maximum temperatures. Closer inspection reveals complex relationships between mean, minimum and maximum temperatures. As was the case for potential nest-sites (see above), a higher mean temperature was associated with a higher maximum temperature (Fig. 3; 1050 m, N = 75 nests, r = 0.71, P < ; 1240 m, N = 82 nests, r = 0.82, P < ; 1615 m, N = 13 nests, r = 0.58, P < 0.04), and the relationship between mean and maximum temperatures differed among elevations. ANCOVA with elevation as the factor, mean nest temperature as the covariate and maximum temperature as the dependent variable, confirmed that maxima were higher, relative to mean temperature, for nests at higher rather than lower elevations (Fig. 3; Maximum temperature ( C) CT max developmental zero 1050m 1240m 1615m Mean temperature ( C) Figure 3. The relationship between mean temperature and maximum temperature for natural nests in the Brindabella Range (numbers show elevations in m asl). The horizontal line at a maximum temperature of 40 C represents the critical thermal maximum (CTmax), the level likely to be lethal to eggs. The vertical line at 16.5 C ( developmental zero ) shows the minimum temperature below which embryogenesis ceases in this species. Thus, development can occur successfully only in the thermal region to the lower right of this Figure, bounded by these two lines (see text for further explanation). The graph shows data for natural nests (separately for three sites, at 1050, 1240 and 1615 m) monitored over a seven-year period. Higher mean values were associated with higher maximum values, and the relationship between mean and maximum nest temperatures varied with elevation (see text). slopes F 2,164 = 0.35, P = 0.70; intercepts F 2,166 = 9.57, P < ; posthoc tests have all P < 0.05). However, the relationship between mean nest temperature and minimum nest temperature was much weaker (1050 m, N = 75 nests, r = 0.11, P = 0.34; 1240 m, N = 82 nests, r = 0.13, P = 0.24; 1615 m, N = 13 nests, r = 0.29, P = 0.33). As for the potential nest-sites, this result reflected the relative constancy of minimum temperatures across nests within each elevation. Minimum temperatures for nests were about 3 C cooler at the highest than the lowest elevation site (see above), so that in order for nests to exhibit approximately equal mean temperatures over this elevational range (as they did: see above), maximum nest temperatures averaged about 4 or 5 C higher at the highest elevation than at the lowest one (Fig. 3). Analysis of data (elevations combined) from natural nests for which we recorded hatching success showed that a higher proportion of eggs hatched successfully from nests with higher mean incubation temperatures (N = 73 nests, r = 0.30, P < 0.015). The four coolest nests (<18 C) did not produce any viable hatchlings, with the proportion of successful eggs increasing at higher mean temperatures (50% for 18 C, 66% for 19 C, 72% for 20 C, 81% for 21 C). Hatching success was low for the hottest nests, however (31% from four nests with a maximum temperature of >40 C, vs. 80% from 25 nests with maxima C). DISCUSSION Our study provides the first quantitative information on thermal regimes available above and below the upper elevational threshold for oviparous reproduction in reptiles. Thermal regimes did indeed shift with elevation in ways that would affect the viability of embryos laid in a nest (as also shown experimentally, by the lower viability of eggs translocated to sites higher on Mount Ginini: Shine, 2002a). However, the nature of these thermal shifts was more complex than we expected. Most published discussions on the restriction of oviparous reptiles to warmer climates have simply taken as self-evident the fact that potential nest-sites at high elevations are too cool to permit embryonic development through to hatching (e.g. Tinkle & Gibbons, 1977; Shine, 1985). Our data challenge this assumption, and suggest instead that the relationship between mean, minimum and maximum temperatures plays a crucial role in limiting oviparous reproduction to low-elevation nests. Thus, the adaptive significance of viviparity lies not in a simple increase in mean incubation temperatures for the embryos, but in the (mobile) female s ability to break the mathematical link between mean and maximum temperatures that applies to any fixed point (such as a nest). This link between mean and maximum is espe-

8 332 R. SHINE ET AL. cially strong when (as was the case for our data), minimum values were relatively invariant within any given elevation. We looked only at thermal factors, and do not doubt that other variables also modify nest-site availability. For example, the fact that oviparous reptiles at high elevations need to lay their eggs in relatively superficial nests (e.g. under relatively thin cover items) in order to experience sufficiently high mean temperatures (Sexton & Claypool, 1978; see Fig. 2) means that eggs in such nests not only are exposed to high thermal variance, but also may be more prone to desiccation or predation than those under thicker shelter (Andrews, 2000). Biotic factors may also play a role. For example, Bassiana in the Brindabella Range are broadly sympatric with large and ferocious bulldog ants of the genus Myrmecia. At low elevations, many shelter-items (logs, rocks) partially or fully exposed to sunlight serve as suitable nest-sites for both lizards and ant colonies, and thus are not a limited resource. At the base of Mount Ginini, however, these thermal conditions can only be obtained under large thin rocks with full sun exposure. Because such rocks are scarce, most shelter thousands of large aggressive ants, severely curtailing their availability as lizard nests (pers. obs.). Thus, the availability of potential nestsites may decrease at higher elevations not because there are no thermally suitable sites, but because many such sites are occupied by competing taxa and thus are not available for lizard nesting. How do elevation-induced changes in nest temperatures impact on embryonic development in Bassiana? Soil temperatures in summer are unlikely to fall below the critical thermal minima for Bassiana eggs, regardless of elevation. Mean minima remained above 10 C in natural nests and above 5 C even in fully exposed positions for potential nest-sites (see above), whereas eggs of Bassiana in the laboratory readily tolerate temperatures around 0 C (Shine, 1983, 2002b). Thus, the elevational limit for successful reproduction in this species is unlikely to be set by minimum soil temperatures. However, our data suggest that both mean and maximum nest temperatures may influence the suitability of nest-sites. Many natural nests exhibit either mean values too low for embryogenesis, or maximum temperatures that are perilously close to lethal values (Fig. 3). Hatching success, developmental rate and hatchling phenotypes are all enhanced by incubation at relatively high mean temperatures (Shine, 1983, 1995), and clearly, maximum nest temperatures must remain below lethal levels also. Extensive laboratory studies reveal that embryogenesis in Bassiana ceases below a mean incubation temperature of 16.5 C (Shine & Harlow, 1996); thus, temperatures below this level will not support development. In natural nests in the field, the minimum temperature for successful hatching was 18 C and the maximum about 38 C (see above). Laboratory data confirm that eggs cannot tolerate incubation temperatures above about 40 C (unpubl. data). Although we do not know how long eggs need to be exposed to such temperatures to experience reduced viability, even a relatively brief exposure to this temperature may be deleterious. Thus, female Bassiana need to lay their eggs in sites with minimum nest temperatures above about 0 C, mean temperatures above 16.5 C (and probably >18 C) but with maximum nest temperatures remaining below 40 C. The first of these criteria (minimum temperature) is satisfied over the entire range of elevations that we studied, but the two latter criteria are more challenging because the strong positive correlation between the mean and the maximum makes it difficult to find sites (especially at high elevations) where a sufficiently high mean does not entail a lethally high maximum. Figure 3 compares these thermal requirements of Bassiana embryos to the incubation regimes recorded in natural nests monitored over a 7-year period at different elevations. In this diagram, the range of conditions that allow successful development are bounded by the vertical line for mean temperature (below which development ceases) and the horizontal line for maximum temperature (above which the eggs would overheat). Unlike the temperatures under pavers (designed to mimic standardized natural nests) that are shown in Figure 2, mean temperatures in real nests were virtually identical across a wide range of elevations (see above for mean values). This difference between real nests under rocks and potential nestsites under pavers reflects the highly non-random attributes of the real nests. Both the pavers and the real nests experienced lower thermal minima at higher elevations, so that the maintenance of high mean temperatures in real nests at high elevations was due to the female lizards selection of thermally favourable locations for egg-laying. Female Bassiana are highly selective of thermal regimes in potential oviposition sites (Shine et al., 1997) and thus presumably seek out thermally optimal sites (high mean, low maximum) for egg deposition. Even at relatively low elevations, female Bassiana select nest-sites based on thermal variance as well as mean temperature, probably because hot, variable regimes accelerate embryogenesis and optimize hatchling phenotypes (Shine et al., 1997). There is significant spatial variation in thermal regimes under cover items at any given elevation, as a function of factors such as slope, aspect and cover-item thickness and heat-retention (e.g. logs vs. rocks). Thus, even very high elevations may have occasional sites where eggs can experience high mean temperatures without lethally high maxima. The wide scatter

9 THERMAL CONSTRAINTS ON REPTILE OVIPARITY 333 around the trend lines in Figure 3 provides evidence of such heterogeneity, although much of this scatter reflects year-to-year variation in weather conditions and hence overestimates the potential for maternal nest-site selection in any one year to decouple the mean from the maximum. Regardless, the significant relationships between mean and maximum temperatures within natural nests (Fig. 3) as well as within potential nest-sites, show that females cannot completely escape this constraint. At all elevations that we examined, maximum temperatures inside some of the natural nests approached lethal levels (40 C), whereas mean temperatures inside other nests approached the minimum for viable incubation (Fig. 3). Nests at higher elevations had higher maxima relative to the mean (Fig. 3), suggesting that the necessary combination of a high mean temperature and low maximum temperature becomes increasingly scarce at higher elevations. Indeed, this situation is a mathematical inevitability given the fact that thermal minima decrease with elevation, and show relatively little variation within any single elevation. Thus, the thermal shift with elevation involves more than a simple decrease in mean incubation temperature. If we restrict attention to replicate nest-sites with similar orientations and thicknesses (i.e. under our pavers), then mean temperatures do indeed decline with increasing elevations (Fig. 2, nests). However, much warmer sites are available, under more superficial cover items (Fig. 2, ground surface). Female Bassiana take advantage of this diversity and are able to find nest-sites with high mean temperatures even at high elevations; but in order to do so in the face of decreasing minima, they have to use sites where daily temperatures reach high maximum values (Fig. 3). Behavioural thermoregulation (heliothermy) allows gravid females of viviparous species to maintain the same advantageous combination of thermal means and variances as are available in (some) low-elevation nest-sites (Fig. 2). Although sample sizes for viviparous females were small in the present study, our results are supported by an extensive published literature (including studies on the Brindabella skinks) showing that mean and maximum body temperatures of heliothermic squamates are relatively unaffected by local climatic conditions (Avery, 1982; Shine, 1983; Greer, 1989). Although the combination of a high mean temperature and a relatively low maximum temperature is difficult or impossible to attain in potential nest-sites at higher elevations (Fig. 3), gravid female lizards can maintain such regimes even at elevations far above the upper limit for oviparous reproduction (Fig. 2). We thus conclude that the factor restricting oviparous reptiles to relatively low-elevation nesting sites is not a simple absence of potential nests with sufficiently high mean temperatures for embryogenesis. Although mean temperatures of the ground surface (the upper limit of available nest temperatures) declined with elevation, they always remained well above the mean temperatures of successful nests. Thus, female Bassiana prepared to use superficial nests could find many sites warm enough for incubation, but maximum temperatures in such a nest would exceed the eggs thermal tolerance. This constraint is a direct consequence of lower minimum (overnight) temperatures in high-elevation sites, such that any given mean incubation temperature is necessarily accompanied by a higher maximum temperature than would be the case for a low-elevation nest. These results suggest that the fundamental reason why viviparous reptiles are able to reproduce in severely cold climates is their ability via behavioural thermoregulation to break the link between mean and maximum incubation temperatures for their offspring. Simple mathematics means that any fixed point must be subject to that link: if the mean is to remain constant, a lower minimum must entail a higher maximum. Eggs are immobile and thus cannot escape that relationship. However, there is considerable spatial heterogeneity in temperatures at any given time of day (Fig. 1), such that a gravid female viviparous reptile can move around to select appropriate thermal environments. The jagged trace of maternal body temperatures during daylight hours (Fig. 1) reflects shuttling heliothermy (Huey & Slatkin, 1976), with consequent regulation of a mean temperature much closer to the maximum than can be attained by any immobile egg. Our interpretation thus differs from previous work on high-elevation reptiles, which has emphasized the fact that maternal body temperatures are on average higher than nest temperatures (Shine, 1983; Qualls, 1997; Andrews, 2000). Although this was true at high but not low elevations in our study (Fig. 2), our data suggest that this difference reflects an elevational shift in the relationship between mean and maximum incubation temperatures. Contrary to intuition and previous speculation, higher elevations do not lack potential nest-sites with mean incubation temperatures high enough to sustain embryogenesis. Paradoxically, the reason that oviparous reptiles cannot reproduce in cold climates may be that although nests with high mean temperatures are available, any eggs laid in those sites would overheat. ACKNOWLEDGEMENTS We thank K. Robert for introducing us to thermochrons, D. Hochuli for creative suggestions, and helpful families (R. and D. Barrott, D. Elphick) for

10 334 R. SHINE ET AL. accommodation during fieldwork. This study was funded by the Australian Research Council. REFERENCES Andrews RM Evolution of viviparity in squamate reptiles (Sceloporus spp.): a variant of the cold-climate model. Journal of Zoology (London) 250: Avery RA Field studies of body temperatures and thermoregulation. In: Gans C, Pough FH, eds. Biology of the Reptilia, Vol. 12. New York: Academic Press, Blackburn DG Evolutionary origins of viviparity in the Reptilia. I. Sauria. Amphibia-Reptilia 3: Blackburn DG Evolutionary origins of viviparity in the Reptilia. II. Serpentes, Amphisbaenia, and Icthyosauria. Amphibia-Reptilia 6: Brown JL Helping and communal breeding in birds: ecology and evolution. Princeton, NJ: Princeton University Press. Cogger HG Reptiles and amphibians of Australia, 4th edn. Sydney: Reed Books. Eisenberg JF The mammalian radiations. London: Athlone Press. Emlen ST, Oring LW Ecology, sexual selection, and the evolution of mating systems. Science 197: Frazer GW, Canham CD, Lertzman KP Gap Light Analyzer (GLA), Version 2.0: Imaging software to extract canopy structure and gap light transmission indices from true-color fisheye photographs, users manual and program documentation. Burnaby, British Columbia: Simon Fraser University. Gallardo CS, Penchaszadeh PE Hatching mode and latitude in marine gastropods: revisiting Thorson s paradigm in the southern hemisphere. Marine Biology 138: Geffen E, Yom-tov Y Old endemics and new invaders: alternative strategies of passerines for living in the Australian environment. Behavioral Ecology and Sociobiology 47: Green K, Osborne W Wildlife of the Australian snow country. Sydney: Reed Books. Greer AE The biology and evolution of Australian lizards. Chipping Norton: Surrey Beatty and Sons. Harvey PH, Pagel MD The comparative method in evolutionary biology. Oxford: Oxford University Press. Huey R, Slatkin M Costs and benefits of lizard thermoregulation. Quarterly Review of Biology 51: Mell R Beitrage zur Fauna Sinica. IV. Grundzuge einer ökologie der chinesischen reptilien und einer herpetologischen tiergeographie Chinas. Berlin: Walter de Gruyter. Muth A Physiological ecology of desert iguana (Dipsosaurus dorsalis) eggs: temperature and water relations. Ecology 61: Packard GC, Packard MJ The physiological ecology of reptilian eggs and embryos. In: Gans C, Huey RB, eds. Biology of the Reptilia, Vol. 16. New York: Alan R. Liss, Pengilley R Systematic relationships and ecology of some lygosomine lizards from southeastern Australia. PhD Dissertation. Canberra: Australian National University. Qualls CP The effects of reproductive mode and climate on reproductive success in the Australian lizard, Lerista bougainvillii. Journal of Herpetology 31: Robert K, Thompson MB Reconstructing thermochron ibuttons to reduce size and weight as a new technique in the study of small animal thermal biology. Herpetological Review in press. Sergeev AM Researches in the viviparity of reptiles. Moscow Society of Naturalists Jubilee Issue: Sexton OJ, Claypool L Nest sites of a northern population of an oviparous snake, Opheodrys vernalis (Serpentes, Colubridae). Journal of Natural History 12: Shine R Reptilian viviparity in cold climates: testing the assumptions of an evolutionary hypothesis. Oecologia 57: Shine R The evolution of viviparity in reptiles: an ecological analysis. In: Gans C, Billett F, eds. Biology of the Reptilia, Vol. 15. New York: John Wiley and Sons, Shine R A new hypothesis for the evolution of viviparity in reptiles. American Naturalist 145: Shine R. 2002a. Reconstructing an adaptationist scenario: what selective forces favor the evolution of viviparity in montane reptiles? American Naturalist 160: Shine R. 2002b. Eggs in autumn: responses to declining incubation temperatures by the eggs of montane lizards. Biological Journal of the Linnean Society 76: Shine R, Barrott EG, Elphick MJ Some like it hot: effects of forest clearing on nest temperatures of montane reptiles. Ecology 82: Shine R, Berry JF Climatic correlates of live-bearing in squamate reptiles. Oecologia 33: Shine R, Elphick MJ, Harlow PS The influence of natural incubation environments on the phenotypic traits of hatchling lizards. Ecology 78: Shine R, Harlow P Maternal thermoregulation influences offspring viability in a viviparous lizard. Oecologia 96: Shine R, Harlow P Maternal manipulation of offspring phenotypes via nest-site selection in an oviparous reptile. Ecology 77: Shine R, Kearney M Field studies of reptile thermoregulation: how well do physical models predict operative temperatures? Functional Ecology 15: Tinkle DW, Gibbons JW The distribution and evolution of viviparity in reptiles. Miscellaneous Publications of the Museum of Zoology, University of Michigan 154: Vitt LJ, Sartorius SS HOBOs, tidbits and lizard models: the utility of electronic devices in field studies of ectotherm regulation. Functional Ecology 13: Weekes HC On the distribution, habitat and reproductive habits of certain European and Australian snakes and lizards, with particular regard to their adoption of viviparity. Proceedings of the Linnean Society of NSW 58: