Climate change impacts on fitness depend on nesting habitat in lizards

Transcription

1 Functional Ecology 2011, 25, doi: /j x Climate change impacts on fitness depend on nesting habitat in lizards Wen-San Huang*,1 and David A. Pike 2 1 Department of Zoology, National Museum of Natural Science, 1 Kuan-Chien Rd., Taichung, 404, Taiwan; and 2 School of Marine and Tropical Biology, James Cook University, Townsville 4811 Australia Summary 1. Through nest-site selection, mothers exert control over the incubation environment to which eggs are exposed, which in turn can affect offspring fitness. The strong relationship between offspring quality and incubation temperature in many ectotherms suggests that contemporary climate change could modify the fitness benefits gained from such behaviour. 2. We used life-history data from a tropical lizard (the long-tailed skink, Eutropis longicaudata) to show that a natural dichotomy in nesting habitat directly influences fitness, but that higher ambient temperatures in recent years are having a disproportionate impact on optimal nest sites. 3. Gravid lizards either lay eggs beneath rocks in natural habitat or inside a concrete wall in human-modified habitat, where incubation temperatures are higher. Consequently, females using artificial habitat produced larger hatchlings that matured earlier and had higher rates of survival than conspecifics using natural habitat. 4. Ambient temperatures have impacted artificial habitats disproportionately over the last decade by increasing nest temperatures in artificial habitat three times that of natural habitat (1Æ5 vs. 0Æ5 C, respectively). This has reversed nest-site quality by lowering offspring viability of artificial nest sites, but has increased offspring viability of natural nest sites. 5. Climate change has overridden the fitness benefits derived from nesting in artificial habitats, but has caused a resultant increase in the fitness benefits derived from nesting in natural habitats. Laboratory incubation experiments confirm that these patterns are attributed to temperature. 6. Our study highlights the interactive effects of disparate human environmental impacts on fauna; by creating the concrete wall, human habitat modification initially conferred fitness benefits by increasing incubation temperatures, but human-induced climate change has raised nest temperatures above the point at which fitness is reduced. 7. Contemporary climate change is altering the location and availability of optimal nesting habitat, thereby changing the ability of nesting females to adaptively manipulate the phenotypes of their offspring. Consequently, human-induced climate change can lead to some habitats becoming ecological traps. Key-words: environmental effects, Eutropis longicaudata, growth, Mabuya longicaudata, microhabitat temperatures, offspring fitness, phenotypic plasticity, survival Introduction *Correspondence author. wshuang@mail.nmns.edu.tw Developing eggs are extremely sensitive to their incubation environment, and in oviparous species that show little or no parental care, selecting nest sites with conditions suitable for embryonic development can enhance egg survival, and thus fitness. However, within the range of incubation environments that successfully produces viable offspring, the phenotypes of emerging hatchlings can vary substantially with environmental variables, including temperature and moisture (Blouin-Demers, Weatherhead & Row 2004; Brown & Shine 2004; Shine 2004; Brakefield & Reitsma 2008). This phenotypic variation can influence traits, such as, body size, sex and behaviour (Elphick & Shine 1998; Bragg, Fawcett & Bragg 2000; Qualls & Shine 2000; Morjan 2003; Brakefield & Reitsma 2008). For example, studies of Ó 2011 The Authors. Functional Ecology Ó 2011 British Ecological Society

2 1126 W.-S. Huang & D. A. Pike natural nests in the field have revealed that hatchling phenotypes are affected by maternal nest-site selection (Muth 1980; Packard & Packard 1988; Burger 1993; Brown & Shine 2004) and that such selection can influence neonate survivorship (Kolbe & Janzen 2001). This evidence suggests that nesting females evaluate the incubation environment and are searching for nest sites that will enhance the fitness of both themselves and their offspring (Goldsbrough, Hochuli & Shine 2004). Contemporary changes in climate (e.g. temperature, rainfall) could therefore influence the abundance and distribution of suitable nesting sites and modify the phenotypes of hatchlings from these sites. Changes in microhabitat temperatures can have positive or negative impacts on incubating eggs. The potential positive influences of temperature on hatchling phenotype can only occur inside the temperature range within which eggs can successfully incubate (Packard & Packard 1988; David et al. 1998). Incubation temperatures above species-specific ranges substantially decrease the viability of embryos (e.g. Packard & Packard 1988; Lo wenborg et al. 2010). Consequently, if nest sites currently experience temperatures at the lower end of the range within which eggs can hatch, the impacts of contemporary climate change could actually increase fitness parameters until temperatures exceed this threshold. After reaching this threshold, fitness will be substantially reduced. Recent work in this area has focused on sex ratios of species in which incubation temperature influences offspring sex (e.g. Mitchell et al. 2008; Telemeco, Elphick & Shine 2009). Despite the urgent need to understand how climate impacts fitness, estimating survival rates from eggs incubated at different temperature and humidity regimes is inherently difficult; consequently, such studies are rare (Brown & Shine 2004). Part of the problem is that the nesting habits of most ectotherms are unknown (e.g. Doody, Freedberg & Keogh 2009; Schwanz et al. 2010). This is compounded by the logistical difficulties associated with quantifying the variation in nest-site conditions used by females, and understanding how that variation influences the survival not only of eggs, but of the resultant offspring over longer time-scales (e.g. Pike et al. 2008; Lo wenborg et al. 2010). A further limitation is that ambient temperatures (for which historical data are widely available) are not equivalent to microhabitat conditions (Kearney, Shine & Porter 2009), such as those experienced by incubating eggs. This greatly limits our ability to obtain biologically relevant data associated with these questions. Recent advances in mechanistic niche modelling have circumvented this latter constraint by using broad-scale climate data to model nest temperatures and predict the consequences of ambient temperatures on offspring traits (e.g. sex; Mitchell et al. 2008; Kearney & Porter 2009). Despite this recent advance, empirically derived data on how temperature influences the life-history trait of interest ultimately are needed. This reinforces the need for long-term studies on maternal nest-site selection in the field, including a detailed understanding of the consequences of nest-site selection for hatchling phenotypes, growth rates, and most importantly, survival. We used long-term field data on tropical long-tailed skinks [Eutropis (Mabuya) longicaudata] to understand how nest-site choice influences fitness, and whether contemporary climate change is influencing the availability of optimal nest sites (Fig. 1). Female long-tailed skinks nest in two distinct types of microhabitats where they either bury their eggs beneath rocks in natural habitat or lay them inside the concrete retaining walls in human-modified habitat. We first show that these two microhabitats differ in thermal and hydric regimes, and this in turn influences female reproductive success (incubation duration, egg survival and hatchling phenotype) and offspring fitness (growth, age at maturity and survival). We then show that contemporary climate change is having a disproportional impact on human-modified habitat, such that temperatures inside nests are increasing at faster rates than nests in natural habitat. Ultimately, this is changing the location and availability of optimal nest sites in the landscape. Materials and methods STUDY SPECIES AND SITES We studied long-tailed skinks on Orchid Island, located off the southeastern coast of Taiwan (45 km 2 ; N, E) from 2001 to 2009 (Fig. 1). This generalist lizard nests in two distinct microhabitats: (i) natural grassland lowland rainforest ecotones where skinks bury their eggs beneath rocks; (ii) and concrete retaining walls where skinks leave their eggs exposed inside plastic drain pipes (pipe diameter is c. 10 cm). We studied these two habitat types (separated by 150 m) on Orchid Island. Because adult females generally stay within 100 m of their initial capture site, these individuals are extremely faithful to individual nesting habitats (W.-S. Huang, unpublished data). The natural site contained 300 rocks, and the artificial site contained 1200 holes spread along 2 km of road. Females mature at 98Æ1 mm snout-vent length (SVL) and lay 2 13 eggs per clutch (mean = 6Æ5 eggs) from February August (Huang 1994, 2006a). After nesting, female long-tailed skinks typically do not remain with the eggs; however, females from Orchid Island nesting inside the artificial holes guard their eggs during incubation (but do not coil around them) to deter predation by egg-eating snakes (Huang 2006b, 2007). Attendant females remain with the eggs for at least one week (mean duration = 22Æ5 days), after which time they begin to leave (Huang Fig. 1. A long-tailed skink (Eutropis longicaudata) on Orchid Island, Taiwan. Photograph by W.-S.H.

3 Fitness mediated nest-site selection and climate change b, 2007). Because this behaviour is not displayed by females nesting in natural habitats, we firstestablished whether maternalbehaviour (i.e. remaining with the clutch for <1 vs. >1 week) influenced reproductive success or the incubation environment of nests laid in humanmodified habitat (see below for details on these attributes). analgesic was smeared on the toes and lizards were given a maximum of two marks). We recaptured these individuals from May to August over the next 5 years ( ) and used these data to calculate growth rates, age at maturity and survival. DOES NEST LOCATION INFLUENCE FITNESS? Reproductive success In the year 2001, we monitored nests during the peak reproductive period (May August) to evaluate reproductive success between the two habitat types. Daily monitoring consisted of looking beneath all rocks and searching all artificial holes for freshly laid eggs. Upon finding a nest, we recorded oviposition date, clutch size, egg mass (±0Æ01 g), and whether the female was present or absent in the nests in artificial habitat. Hatchlings from nests in artificial habitat remain inside their nest hole for at least 2 days after hatching, which allowed us to hand-capture them. By contrast, hatchlings from natural nests leave almost immediately, so towards the end of incubation we encircled natural nests with plastic fencing to trap emergent hatchlings until our next nest check. This allowed us to capture all hatchlings emerging from each clutch. Upon capture, we measured the following body size attributes of each hatchling: SVL, body mass, tail length, head length, head width and head depth. Hatchlings were too small to determine their sex in the field, and thus we do not consider sex in our analyses. We calculated egg survival (i.e. hatching success) by counting the number of emergent hatchlings and dividing this by the total number of eggs in the clutch. We used one-way ANOVAs to compare clutch size, incubation duration and egg survival between nesting habitats (natural and artificial). We used mixed effects ANOVAs with clutch identity as a random factor and habitat type as a fixed factor to compare differences in egg mass, incubation duration, hatchling SVL, body mass, relative tail length and relative head length between habitats. For the tail and head length comparisons, we used SVL as the covariate. Preliminary analyses revealed that head measurements were intercorrelated (all r >0Æ49, P <0Æ001), so we only retained head length for analysis. Prior to all analyses we checked that the data met the assumptions of parametric statistical tests. To quantify incubation conditions, we recorded nest temperatures and humidity at 6-h intervals daily (i.e. four readings per day resulting in readings per clutch, depending on incubation duration). Both variables were measured by placing a probe (Lutron Electronic Enterprise Co. Ltd., Taipei, Taiwan) next to the clutch. The resultant temperature and humidity readings were averaged throughout the incubation period for each individual nest at each time period. We first tested whether the presence of attendant females influenced nest temperatures or humidity (using nests in artificial habitat only) and then tested for differences between artificial vs. natural nests. For these analyses, we used repeated-measures ANOVAs with maternal attendance or nest type as factors (respectively) and daily temperature or humidity readings as the dependent variable. For each nest, we averaged the temperature and humidity readings taken at 6-h intervals across the incubation period, giving four readings per nest representing the average diel cycle. Offspring fitness In the year 2001, we individually marked 200 hatchlings from 50 clutches, half of which were from each habitat type (using toe-clips, Growth rates and age at maturity We calculated growth rates for each recaptured individual over yearly intervals (SVL, mm per year) and tested whether growth rates slowed with increasing body size using linear regression. To test for differences in relative growth rate (adjusted for body size) between nest types, we used ANCOVA with nest type as the factor, the midpoint body size between the longest recapture intervals (SVL or body mass) as the covariate, and annual growth rate over the same interval as the dependent variable. To test whether there were consistent differences in body size during each year of the study, we used a two-way ANOVA with year and habitat type as factors, and body size as the dependent variable. We estimated age at sexual maturity for hatchlings emerging from both habitat types using a von Bertalanffy (1938) growth model. We fitted mark recapture data to a known-age model using the nonlinear, least-squares regression procedure in the software JMP 5.0.1, where asymptotic length and intrinsic growth rates were initially seeded with best guesses (after Frazer, Gibbons & Greene 1990). Growth models were run separately for natural and artificial nests, allowing us to construct separate age-size curves for each group. We then used the mean size of hatchlings from each nest type and the minimum size at which females mature (98Æ1 mm; Huang 2006a) to estimate the age at which female hatchlings from each nest type would reach sexual maturity. Offspring survival We pooled recapture data for each year to estimate survival using a Cormack-Jolly-Seber (CJS) model implemented in program MARK (version 5.1; White & Burnham 1999). Our main interest was in survival between hatchlings born in artificial and natural habitats, rather than in estimating recapture rates. We developed a set of candidate models that held recapture rates constant or differed between nest types and that tested for constant survival, time-dependent survival, differences in survival between nest types or interactions between these variables. We derived an estimate of lack of fit for the global (i.e. most parameterized) model in our candidate set using program RELEASE implemented in MARK. There was some evidence that the global model did not fit the data well (v 2 =18Æ48, d.f. = 12, P =0Æ10), so we adjusted our models for overdispersion (based on the variance inflation factor, cˆ = 1Æ54) using quasi-likelihood AIC (QAIC c ) values prior to model selection (Lebreton et al. 1992; Burnham & Anderson 1998). QAIC c values were used to select the best approximating (hereafter, best) model for the data, based on the principles of parsimony and trade-offs between under- and overfitting models (Burnham & Anderson 1998). The best-supported models are those that make up the top 90% of Akaike weights and have relative deviations from the best model of less than two (i.e. DQAIC c < 2; Burnham & Anderson 1998). We used the highestranking candidate model to estimate survival and compared annual survival between hatchlings from artificial vs. natural nests over time using a paired t-test. Survival estimates incorporate the probabilities of dying and emigrating (termed apparent survival), which in our study equates to local persistence in the study area (White & Burnham 1999).

4 1128 W.-S. Huang & D. A. Pike IS CONTEMPORARY CLIMATE CHANGE ALTERING NEST-SITE QUALITY? Ambient temperatures We downloaded mean monthly air temperatures for Orchid Island from the Taiwanese Central Weather Bureau for For each year, we calculated the median, maximum and variance in air temperature during the peak lizard nesting season (May August) and used linear regression to determine whether ambient temperatures were dependent upon year. Nest temperatures First, we established whether nest temperatures were related to ambient temperatures in similar ways between habitat type. We used ANCO- VA with nest type as the factor, ambient temperature as the covariate and nest temperature as the dependent variable for minimum, maximum, and mean temperatures and thermal variance. Significant interaction terms between habitat type and temperature would indicate that ambient temperatures influence nest temperatures in different ways. Next, we tested whether nest temperatures (minimum, maximum, mean and thermal variance) influenced hatchling phenotype (egg hatching success, SVL and body mass) using data from both nest types for (N = 129 artificial nests, 86 natural nests). To determine whether temperatures inside the artificial and natural nests have changed over the course of our study, we regressed mean nest temperatures collected from the field with year ( ) for each nest type. We then used year as a continuous variable and compared nest temperatures between habitat types using ANCOVAs with year as the covariate. We were most interested in whether there was a significant interaction between habitat type and year for each variable, which would indicate that nest temperatures are changing in different ways over time in each habitat. We then split our long-term data set into two discrete periods based on nest temperatures: early years which had relatively cooler nest temperatures ( ) and late years, which had relatively warmer nest temperatures ( ). We used a two-way repeated-measures ANOVA with habitat type and time period as factors and temperatures throughout the 24-h cycle to determine whether nest temperatures changed significantly between time periods. Changes in hatchling phenotype We tested whether hatchling phenotype changed in similar ways over time using ANCOVA with year as the covariate and hatchling attributes (egg hatchingsuccess, SVL, ormass) asthe dependent variable. Aswith the thermal data above, we were most interested in whether there was a significant interaction between habitat type and year for each variable. Table 1. Comparisons of reproductive success and hatchling body size attributes between Eutropis longicaudata nests laid beneath rocks in natural habitat (N = 25) vs. artificial habitat (N = 25). Egg survival is the percentage of eggs hatching from each clutch. Values are for 2001 only and are presented as ranges followed by means ± SE in parentheses Variable Natural habitat Artificial habitat F-value d.f. P-value Clutch size (n) 2 11 (6Æ1 ± 1Æ4) 1 13 (6Æ5 ± 1Æ2) 0Æ60 1, 47 0Æ44 Egg mass (g) 0Æ7 1Æ4(1Æ0 ±0Æ4) 1Æ0 1Æ4(1Æ3 ±0Æ3) 2Æ85 48, 199 <0Æ0001 Incubation duration (days) (31Æ6 ± 0Æ39) (26Æ9 ± 0Æ4) 71Æ63 1, 47 <0Æ0001 Egg survival (%) 33Æ3 100 (67Æ1 ±3Æ1) 66Æ7 100 (86Æ0 ±1Æ5) 40Æ60 1, 47 <0Æ0001 Hatchling SVL (mm) 32Æ0 35Æ9 (34Æ5 ±0Æ2) 35Æ1 39Æ2 (37Æ4 ±0Æ2) 4Æ65 48, 199 <0Æ0001 Hatchling body mass (g) 0Æ60 1Æ28 (0Æ92 ± 0Æ03) 0Æ89 1Æ29 (1Æ22 ± 0Æ03) 23Æ93 48, 199 <0Æ0001 Hatchling tail length (mm) 51Æ0 66Æ0 (62Æ3 ±1Æ1) 51Æ0 71Æ0 (65Æ2 ±1Æ1) 1315Æ89 48, 199 <0Æ0001 Hatchling head length (mm) 9Æ0 10Æ3 (10Æ3 ±0Æ1) 8Æ9 11Æ3 (11Æ1 ±0Æ1) 7Æ15 48, 199 <0Æ0001 SVL, snout-vent length. (c) (d) Fig. 2. Mean daily thermal regimes, thermal variation, (c) hydric regimes and (d) hydric variation (± SE) inside Eutropis longicaudata nests laid in artificial vs. natural habitats. Daily regimes are averaged over 6-h intervals from May to August Note that values at 0 h are shown twice to better depict the entire 24-h cycle, but these data were only included once in the analyses.

5 Fitness mediated nest-site selection and climate change 1129 Wethen testedwhether hatchingsuccess changed between time periods using a two-way ANCOVA with nest type (artificial vs. natural) and time period ( vs ) as factors, the total number of eggs within each clutch as the covariate and the number of eggs hatching as the dependent variable. We also tested whether SVL and body mass changed over time, using a two-way ANCOVA with habitat type (artificial vs. natural) and time period (warm vs. cool years) as factors, and meanclutchvaluesofsvl orbodymassasdependent variables. Finally, in 2009 we conducted a laboratory experiment to understand how different thermal regimes influence egg survival and offspring phenotype. We collected 15 clutches of freshly laid eggs, each of which contained six eggs (N = 90 eggs), and randomly allocated two eggs from each clutch to the following temperature treatments: (i) low (daily range: C, mean temperature: 29Æ7 C), (ii) moderate (daily range: C, mean temperature: 30Æ6 C), (iii) and high (daily range: C, mean temperature: 31Æ1 C). These temperature regimes cycled through the range of temperatures daily in a manner similar to nests in the field and matched temperatures from natural nest sites, artificial nest sites during cooler years and artificial nest sites during warmer years, respectively. All eggs from each treatment were placed into a single container containing moist vermiculite, but eggs were separated by at least 3 cm. Ambient humidity inside the container was maintained at 90%. We recorded whether each egg hatched, the incubation duration, and measured the phenotype of hatchlings as described above. We compared egg survival among the three incubation treatments using a contingency table analysis and compared incubation duration and phenotypes among treatments using one-way ANOVAs. Results INFLUENCE OF MATERNAL ATTENDANCE (c) Female attendance time (i.e. < or >1 week) had no influence on clutch size and mass, incubation duration, egg survival or hatchling body size (in all cases P > 0Æ34). Likewise, nest thermal and hydric regimes did not differ with maternal attendance time (temperature: means of 30Æ2 ±1Æ2 vs. 30Æ1 ±1Æ4 C, respectively; nest type, F 1,43 =1Æ16, P =0Æ29, nest type * time interaction, F 3,129 =1Æ00, P =0Æ39; humidity: means of 92Æ8 ±0Æ9% vs. 92Æ1 ±0Æ6%, respectively; nest type, F 1,43 =1Æ75, P =0Æ19, nest type * time interaction, F 3,129 =0Æ50, P =0Æ68). Thus, maternal attendance did not significantly influence any of the variables we investigated, and any subsequent differences between natural and artificial nest sites are unlikely to be a function of maternal behaviour. DOES NEST LOCATION INFLUENCE FITNESS? Reproductive success Fig. 3. Mean body size (snout-vent length, SVL; ± SE) of Eutropis longicaudata emerging from nests laid in artificial vs. natural habitats. Data for 2001 represent body size at hatching, and data for show how these hatchlings subsequently grew. Von Bertalanffy growth curves for E. longicaudata hatchlings incubated in artificial habitats (dotted line) vs. natural habitats (solid line). The horizontal line indicates the minimum size of sexually mature females and the vertical lines show the estimated age at maturity for artificial (dotted line) vs. natural nests (solid line). (c) Annual survival estimates (± SE) of E. longicaudata hatching from artificial vs. natural habitats in Note that in some cases error bars are too small to be visible. In the year 2001, we recorded data from 25 nests in artificial habitat and 25 nests in natural habitat from egg laying until hatching. Although clutch size was similar between habitat types, eggs from nests in artificial habitat were larger, hatched almost 5 days earlier and had 18Æ9% higher survival than eggs from natural habitat (Table 1). Mean incubation temperatures of nests in artificial habitat were significantly higher than those in natural habitat (habitat type, F 1,68 =17Æ10, P =0Æ0001; habitat type * time interaction, F 3,204 =1Æ29, P =0Æ28; Fig. 2a). Mean temperature during the hottest part of the day was 30Æ5 ±0Æ3 C for natural nests and 31Æ4 ± 0Æ3 C for artificial nests. Additionally,

6 1130 W.-S. Huang & D. A. Pike nest temperatures in artificial habitat were significantly higher than nests in natural habitat in terms of overall minimum (F 1,68 =459Æ12, P <0Æ0001), maximum (F 1,68 =1876Æ45, P <0Æ0001) and mean temperatures (F 1,68 =1222Æ07, P <0Æ0001), as well as thermal variance (F 1,68 =636Æ93, P <0Æ0001; Fig. 2b). Likewise, humidity inside artificial nests was significantly higher than natural nests (habitat type, F 1,68 =5Æ99, P =0Æ017; habitat type * time interaction, F 3,204 =2Æ37, P =0Æ07; Fig. 2c). Humidity typically decreased with increasing nest temperature, with overall humidity during the hottest part of the day averaging 89Æ7 ±0Æ4% for natural nests and 91Æ6 ±0Æ6% for artificial nests. Humidity of nests in artificial habitat was significantly higher than nests in natural habitat in terms of overall minimum (F 1,68 =5Æ64, P =0Æ020), maximum (F 1,68 =3Æ57, P =0Æ063) and mean temperatures (F 1,68 =5Æ99, P =0Æ017), but was not different in terms of hydric variance (F 1,68 =0Æ001, P =0Æ98; Fig. 2d). Because of the low diel variation in humidity (generally <4%), we do not consider this variable in further analyses. Offspring body size Eggs incubating in artificial habitat hatched earlier than those from natural habitat, and the hatchlings from artificial habitat were larger than those from natural habitat in terms of SVL, body mass, relative tail length and relative head length (Table 1). Growth rates and age at maturity Of the 100 hatchlings released from each nest type, we recaptured 68 individuals from artificial habitat and 86 individuals from natural habitat at least once from 2002 to There was a significant negative correlation between midpoint SVL and growth rate, indicating that growth rates declined with increasing hatchling body size (artificial nests: R 2 = )0Æ39, P <0Æ001; natural nests: R 2 = )0Æ32, P <0Æ001). Hatchlings from both habitat types grew at the same rate relative to their body size (ANCOVA, F 1,151 =0Æ01, P =0Æ99). Despite this, however, hatchlings born in artificial habitat were consistently larger than were hatchlings from natural habitat in each year (two-way ANOVA; nest type: F 1,342 =217Æ02, P <0Æ0001; year: F 5,342 =86188Æ90, P <0Æ0001; nest * year interaction: F 5,342 =14Æ61, P <0Æ0001; Fig. 3a). Based on the minimum size of sexually mature females, the von Bertalanffy growth models predicted that hatchlings born in artificial habitat attained sexual maturity 4 months earlier than hatchlings from natural habitat (i.e. at 1Æ58 years or 19 months vs. 1Æ9 years or 23 months; Fig. 3b). Of the female hatchlings that reached maturity and were subsequently found nesting, a greater proportion nested in the habitat in which they were born than in the opposite habitat (n = 52 females: 35 born in artificial habitat nested there, eight born in natural habitat nested there, six born in artificial habitat nested in natural habitat and three born in natural habitat nested in artificial habitat; comparing individuals nesting in natal habitat vs. switching habitats, v 2 =22Æ23, d.f. = 1 and P <0Æ0001); 82Æ7% of females remained faithful to the habitat in which they were born. Offspring survival The model with most support from our mark recapture data (87%) incorporated differences in hatchling survival between habitat types over time (i.e. from 2002 to 2006; Table 2). The next highest-ranking model was one where survival varied with time, but not by habitat type (Table 2). None of the other candidate models carried much weight (Table 2). There was considerable support for a real difference between the two highest-ranking models (DQAIC c =3Æ92), and consequently we used the highest-ranking model to estimate survival parameters. Average annual survival rates of hatchlings were significantly higher for individuals born in artificial habitat (paired t =3Æ30, P =0Æ03; Fig. 3c). On average, hatchlings emerging from nests in artificial habitat had 24% higher survival from 2002 to 2006 than those hatching from natural habitat (Fig. 3c). IS CONTEMPORARY CLIMATE CHANGE ALTERING NEST-SITE QUALITY? Ambient temperatures Median ambient temperatures on Orchid Island have increased significantly during the lizard nesting season from 1940 to 2009 (N = 66 years, data were unavailable for 1945 and 1946; R 2 =0Æ13, P =0Æ003; Fig. 4a). During this time, Table 2. Candidate models used to assess the influence of nest type (artificial vs. natural) on hatchling Eutropis longicaudata survival rates. Hatchlings were born and marked in 2001 and recaptured from 2002 to We ranked models according to Akaike Information Criterion (AIC) with adjustment for over dispersion (QAIC c ). The best-supported models have DQAIC c values <2 Survival Recapture QAIC c DQAIC c QAIC c weight Model likelihood QDeviance Parameters Time * Nest Constant * Nest 619Æ43 0 0Æ87 1Æ00 78Æ11 12 Time Constant * Nest 623Æ35 3Æ92 0Æ12 0Æ14 92Æ58 7 Time Constant 629Æ64 10Æ21 0Æ01 0Æ01 100Æ94 6 Constant * Nest Constant * Nest 633Æ62 14Æ Æ04 4 Time * Nest Constant 636Æ16 16Æ Æ98 11 Constant Constant * Nest 641Æ35 21Æ Æ81 3 Constant Constant 644Æ23 24Æ Æ72 2 Constant * Nest Constant 645Æ76 26Æ Æ22 3

7 Fitness mediated nest-site selection and climate change 1131 increased during this period (R 2 =0Æ09, P =0Æ014; Fig. 4b), as did monthly thermal variance (R 2 =0Æ08, P =0Æ022; Fig. 4c). Nest temperatures (c) Fig. 4. Ambient temperatures averaged throughout the Eutropis longicaudata nesting season (May August) on Orchid Island, Taiwan from 1940 to Shown are monthly median temperatures, mean monthly maximum temperatures and (c) mean monthly thermal variance. Data for 1945 and 1946 were unavailable. there was an increase in ambient temperature of 0Æ065 C per decade, which amounts to an overall average increase of 0Æ455 C. Likewise, mean maximum monthly temperatures We found that the relationship between ambient and nest temperatures differed significantly between habitat types for minimum (F 1,211 =16Æ68, P <0Æ0001) and maximum temperatures (F 1,211 =7Æ53, P =0Æ007) and thermal variance (F 1,211 =23Æ06, P <0Æ0001), but not for mean temperatures (F 1,211 =2Æ23, P =0Æ14; see Fig. S1 in Supporting Information). Nest temperatures significantly influenced egg hatching success rates and hatchling SVL, but not hatchling body mass (Table 3). From the year 2001 to 2009, mean nest temperatures increased significantly in artificial habitat (R 2 =0Æ27, P < 0Æ0001), but not in natural habitat (with one statistical outlier removed, R 2 =0Æ05, P =0Æ30). We found significant interactions between habitat type and year for minimum (F 1,211 =68Æ53, P <0Æ0001; Fig. 5a), maximum (F 1,211 =43Æ17, P <0Æ0001; Fig. 5b) and mean temperatures (F 1,211 =50Æ79, P <0Æ0001; Fig. 5c), as well as thermal variance (F 1,211 =40Æ26, P <0Æ0001; Fig. 5d). In nests laid in artificial habitat, temperatures are becoming warmer and more variable over time, and this increase is occurring faster than in nests laid in natural habitat (Fig. 5). These thermal variables increased in the artificial habitat beginning in 2006, whereas this pattern was not apparent in the natural habitat (Fig. 5). When we divided these years into two distinct time periods ( vs ), we found that mean nest temperatures varied significantly between nest type and time period (habitat type: F 1,211 =1630Æ90, P <0Æ0001; time period: F 1,211 =314Æ70, P <0Æ001; interaction: F 1,211 =149Æ95, P < 0Æ0001; Fig. 6a). Nest temperatures in both artificial and natural habitats increased in more recent years (Fig. 6a), but the relative increase in nest temperature was three times greater in artificial habitat than in natural habitat (i.e. 1Æ5 vs. 0Æ5 C, respectively; Fig. 6b). Changes in hatchling phenotype We found significant interactions between habitat type and year for egg hatching success (F 1,211 =68Æ53, P <0Æ0001; Table 3. Relationships between hatchling Eutropis longicaudata phenotypes and nest temperatures (n = 75 nests in artificial habitat and 25 in natural habitat). P-values are from individual regressions between phenotype and nest temperature, using data from both artificial and natural habitats. Significant results are indicated in bold Minimum nest temperature Maximum nest temperature Mean nest temperature Nest thermal variance Hatchling phenotype R 2 P R 2 P R 2 P R 2 P Egg hatching success 0Æ01 0Æ51 0Æ09 0Æ01 0Æ06 0Æ01 0Æ18 <0Æ0001 Snout-vent length 0Æ47 <0Æ0001 0Æ28 <0Æ0001 0Æ31 <0Æ0001 0Æ08 0Æ006 Body mass 0Æ04 0Æ06 0Æ02 0Æ21 0Æ02 0Æ17 0Æ01 0Æ56

8 1132 W.-S. Huang & D. A. Pike (c) (d) Fig. 5. Eutropis longicaudata nest temperatures in artificial vs. natural habitats from 2001 to Shown are overall minimum, maximum and (c) mean temperatures along with (d) thermal variance (all ±SE). The solid trend line is for artificial habitat, and the dashed trend line is for natural habitat. Fig. 6. Mean nest temperatures inside Eutropis longicaudata nests laid in artificial vs. natural habitat, averaged over 6-h intervals, and shown for May August and Relative increases in E. longicaudata nest temperatures located in artificial vs. natural habitat from 2001 to 2005 and from 2006 to 2009, respectively. Note that values at 0 h are shown twice to better depict the entire 24-h cycle, but these data were only included once in the analyses. Error bars are not shown because they are smaller than the symbols. Fig. 7a) and hatchling SVL (F 1,211 =43Æ17, P <0Æ0001; Fig. 7b), but not for hatchling body mass (F 1,211 =50Æ79, P <0Æ0001; Fig. 7c). In artificial habitat, egg hatching success rates steadily decreased during our study, but in natural habitats these rates increased (Fig. 7a). By contrast, hatchling SVL stayed constant in the artificial habitat, but increased in natural habitat (Fig. 7b). Body mass showed the same relative pattern, but the change over time was not significant (Fig. 7c). We found that hatching success differed between habitat type and time period in different ways (habitat type: F 1,96 =0Æ73, P =0Æ39; time period: F 1,96 =2Æ23, P =0Æ14; interaction; F 1,96 =21Æ27, P <0Æ0001); hatching success in the artificial habitat decreased by 28% in more recent years, while hatching success in the natural habitat increased by 14% (Fig. 8a). However, SVL showed the opposite pattern (habitat type: F 1,92 =165Æ26, P <0Æ0001; time period; F 1,92 =6Æ56, P =0Æ012; interaction; F 1,92 =10Æ68, P = 0Æ002; Fig. 8b), where hatchlings from natural habitat increased in SVL in more recent years, but SVL did not change in the artificial habitat (Fig. 8b). There was no significant change in body mass between time periods (F 1,92 =2Æ94, P =0Æ09; time period: F 1,92 =2Æ88, P =0Æ09; interaction; F 1,92 =2Æ47, P =0Æ12; Fig. 8c). A direct comparison of hatching success and hatchling phenotype between habitat types in recent years ( ) revealed similar hatching success rates (ANCOVA, F 1,97 =1Æ94, P =0Æ17; Fig. 8a) and hatchling body mass (F 1,94 =3Æ62, P =0Æ06; Fig. 8c), but that hatchlings from the artificial habitat still have larger SVLs than those in the natural habitat (F 1,94 =162Æ99, P <0Æ0001; Fig. 8b,c). Finally, our laboratory experiment revealed that incubation temperature strongly influenced hatching success (v 2 =10Æ30, d.f. = 2, P <0Æ0001; Fig. 9a); egg survival was highest at moderate incubation temperatures (mimicking nest temperatures in artificial habitat during relatively cool years) and was lowest in our high-temperature incubation treatment (mimicking nest temperatures in artificial habitat in relatively warm years; Fig. 9a). Incubation temperature also had significant effects on incubation duration (F 2,43 =17Æ55, P <0Æ0001; Fig. 9b), SVL (F 2,43 =52Æ41, P <0Æ0001;

9 Fitness mediated nest-site selection and climate change 1133 (c) (c) Fig. 7. Mean egg hatching success, hatchling snout-vent length (SVL) and (c) body mass (all ±SE) of Eutropis longicaudata from 2001 to 2009 shown for nests incubating in artificial vs. natural habitat. The solid trend line is for artificial habitat and the dashed trend line is for natural habitat. Fig. 9c), and body mass (F 2,43 =3Æ73, P =0Æ03; Fig. 9d). Hatchlings from the coolest incubation treatment were took longest to hatch and were smallest, while hatchlings from the two warmer treatments hatched earlier and were larger and nearly equivalent in size (Fig. 9b,c). Post-hoc tests indicated that in all cases the low treatment was significantly different from the other two treatments (P <0Æ05), but that in all cases the moderate- and high-temperature treatments were similar (P > 0Æ31). Discussion Our study provides field-based evidence that nest-site selection by tropical long-tailed skinks influences both offspring and maternal fitness, but that contemporary climate change has altered the location and availability of optimal nest sites. Gravid females nest within two distinct microhabitats Fig. 8. Changes in hatching success (percentage of eggs hatching), hatchling snout-vent length and (c) hatchling body mass for Eutropis longicaudata hatching from artificial or natural nests in the early ( ) vs. late ( ) years of our study. Shown are mean values ±SE. (beneath the rocks in natural habitat or inside the retaining walls), and by using artificial habitat for nesting females increased the survival of their eggs and produced larger offspring that hatched earlier (Table 1; Fig. 3a), matured earlier (Fig. 3b) and had higher rates of survival (Fig. 3c) than females nesting in natural habitat. Thus, during the early years of our study females could increase their fitness by nesting in human-modified habitats. However, as our study progressed the ambient temperatures in human-modified began increasing above those of natural habitats, thereby reversing this pattern of nest-site quality. The consequence has been a very rapid reduction in the quality of the artificial habitat, but an increase in the quality of the natural habitat. Because natural habitats are being modified at alarming rates world-wide, these findings have widespread implications for oviparous species persisting in human-dominated landscapes.

10 1134 W.-S. Huang & D. A. Pike (c) (d) Fig. 9. Hatchling phenotype of Eutropis longicaudata eggs incubated under laboratory conditions at low (mean temperature: 29Æ7 C, range: C), moderate (30Æ6 C, range: C) or high temperatures (31Æ1 C, range: C) for hatching success of eggs (N =30 treatment), incubation duration, (c) snout-vent length and (d) body mass. depicts the percentages of eggs hatching in each treatment, shows incubation duration ±SE, and (c) and (d) show mean body size ±SE. Available nesting sites differed in two ways that plausibly explain the mechanism behind the difference in fitness between the two nesting habitats: (i) females nesting inside the artificial wall remained with their eggs during incubation; (ii) and nests in artificial habitat were warmer and more humid than natural nests (and potentially differ in other, unmeasured ways). Nest attendance by long-tailed skinks deters whole-clutch predation by egg-eating snakes (Huang 2006b), but none of the clutches in the current study were preyed upon in their entirety (Table 1), and we found no evidence that the duration of female nest attendance influenced either incubation conditions (e.g. incubation duration, temperature and humidity) or offspring phenotype. Thus, it seems unlikely that the unique maternal attendance behaviour displayed by females nesting in artificial habitats has any functional role except to deter predators from eating eggs (counter to some other nest-guarding lizard species; reviewed in Shine 1988). Consequently, the fitness differences that we observed are likely due to direct influences of the incubation environment, and our laboratory experiments further support this assertion (Fig. 9). Our findings are consistent with other studies, showing that warmer and or more humid incubation environments produce larger and faster-growing offspring (Blouin-Demers, Weatherhead & Row 2004; Brown & Shine 2004; Radder & Shine 2007). We suspect that temperature is the more important variable because humidity varied only slightly throughout the day (Fig. 2c,d). Future studies looking at the interactive effects of temperature and humidity on offspring phenotype would be worthwhile. Individuals hatching earlier in the season have substantial advantages over those hatching relatively later in the season, mainly because early-hatching individuals have longer activity periods before the onset of cold weather than do laterhatching conspecifics (Olsson & Shine 1997; Shine & Elphick 2001; Warner & Shine 2005). For example, in our study area, small invertebrates (the major food source for hatchlings) are in low availability late in the active season (September January; W.S. Huang, unpublished data). Because head size is related to the size of prey that lizards can ingest (Schoener 1967), smaller hatchling lizards may be at a disadvantage when capturing prey late in the season. Although recent studies have documented catch-up growth, whereby smaller hatchlings grow at faster rates than larger conspecifics (thus eliminating any initial body size advantage at hatching; Radder, Warner & Shine 2007), in our study hatchlings from both nest types grew at equal rates relative to their body size. The end result was that individuals from natural nests were consistently smaller than those from artificial nests at any given point in time (Fig. 3a). Because body size (rather than age) influences sexual maturation in reptiles, this consistent body size advantage explains the differences in age at sexual maturity (Fig. 3b) and plausibly increases the number of lifetime breeding attempts (Olsson & Shine 1997; Warner & Shine 2007). Why did temperatures inside the concrete wall increase steadily throughout our study, while this did not occur in nest sites located in natural habitat? The concrete wall is located in an area with open canopy cover and is exposed to direct sunlight, while the rocks under which skinks lay their eggs in natural habitats are located along the forest edge and are thus more shaded. Consequently, ambient temperatures influenced nest temperatures differently in each habitat type (see Fig. S1). The concrete wall is much larger in volume than the small rocks used for nesting in natural habitats, and larger objects can absorb and retain heat more efficiently than smaller objects (especially those located on soil, which acts as a heat sink). Thus, it is unsurprising that these two microhabitats have responded to increased ambient temperatures in different ways. Artificial habitats can play important roles in the ecology, evolution and behaviour of species. In some cases, modified habitats support higher population densities than their

11 Fitness mediated nest-site selection and climate change 1135 natural counterparts (e.g. Germaine & Wakeling 2000), even though individuals may use both habitat types in similar ways (e.g. Kolbe & Janzen 2002). Anthropogenic habitat modification can influence nesting microenvironments by providing increased nest temperatures for some species (e.g. Sartorius, Vitt & Colli 1999; Shine, Barrott & Elphick 2002; Angilletta, Sears & Pringle 2009; Lo wenborg et al. 2010), but decreased nest temperatures for others (Mrosovsky, Lavin & Godfrey 1995; Kolbe & Janzen 2002). This can cause some habitats to become ecological traps, where the quality of an attractive habitat is reduced but animals still select that habitat because it was previously beneficial (e.g. Gates & Gysel 1978; Kolbe & Janzen 2002; Robertson & Hutto 2006). In our study, the available dichotomy of nesting sites was created through human habitat modification, namely construction of an artificial wall to prevent erosion in This wall was colonized rapidly by long-tailed skinks, some of which subsequently developed a unique maternal nest-guarding behaviour (Huang 2006b, 2007). A majority of hatchling female long-tailed skinks continued to nest in their natal habitat once reaching maturity, despite lowered quality of the artificial nest sites. This demonstrates that human habitat modification can benefit fauna in ways that enhance fitness, but that climate change can alter microhabitats disproportionately in these habitat types such that they rapidly become unsuitable for nesting by reducing egg viability. We speculate that the presence of the artificial wall has increased population size over the past decade by providing novel and abundant nesting sites since Further increases in temperature could result in complete loss of this important nesting habitat (in terms of producing viable offspring), which could have cascading effects throughout the population by slowing population growth rates, potentially eroding the unique nest-guarding behaviour displayed by long-tailed skinks nesting in artificial habitats. Alternatively, females nesting early in the season (when temperatures are cooler) may be able to continue using the artificial wall for nesting, but could switch to nesting beneath rocks in natural habitat as temperatures increase seasonally. Our study highlights the interactive effects of disparate human environmental activities on fauna and demonstrates how contemporary climate change can cause human-modified habitats to become ecological traps. Acknowledgements We thank C.H. Chang and several volunteers for help with fieldwork. Fred Janzen, Betsy Roznik and Dan Warner provided helpful comments on an earlier draft, and Carryn Manicom provided advice on survival analyses. Funding was provided by the National Science Council of Taiwan (NSC B MY3). All work was conducted under the National Museum of Natural Science Protocol Permit NMNSHP and adhered to the animal ethics protocols of the National Museum of Natural Science, Taichung, Taiwan and Pingtung County, Taiwan. References Angilletta Jr, M.J., Sears, M.W. & Pringle, R.M. (2009) Spatial dynamics of nesting behavior: lizards shift microhabitats to construct nests with beneficial thermal properties. Ecology, 90, von Bertalanffy, L. (1938) A quantitative theory of organic growth. Human Biology, 10, Blouin-Demers, G., Weatherhead, P.J. & Row, J.R. (2004) Phenotypic consequences of nest-site selection in black rat snakes (Elaphe obsoleta). Canadian Journal of Zoology, 82, Bragg, W.K., Fawcett, J.D. & Bragg, T.B. (2000) Nest-site selection in two eublepharid gecko species with temperature-dependent sex determination and one with genotypic sex determination. Biological Journal of the Linnean Society, 69, Brakefield, P.M. & Reitsma, N. (2008) Phenotypic plasticity, seasonal climate and the population biology of Bicyclus butterflies (Satyridae) in Malawi. Ecological Entomology, 16, Brown, G.P. & Shine, R. (2004) Maternal nest-site choice and offspring fitness in a tropical snake (Tropidonophis mairii, Colubridae). Ecology, 85, Burger, J. (1993) Colony and nest site selection in lava lizards Tropidurus spp. in the Galapagos Islands. Copeia, 1993, Burnham, K.P. & Anderson, D.R. (1998) Model Selection and Inference: A Practical Information-Theoretic Approach. Springer-Verlag, New York. David, J.R., Gibert, P., Gravot, E., Petavy, G., Morin, J.-P., Karan, D. & Moreteau, B. (1998) Phenotypic plasticity and developmental temperature in Drosophila: analysis and significance of reaction norms of morphometrical traits. Journal of Thermal Biology, 22, Doody, J.S., Freedberg, S. & Keogh, J.S. (2009) Communal egg-laying in reptiles and amphibians: evolutionary patterns and hypotheses. Quarterly Review of Biology, 84, Elphick, M.J. & Shine, R. (1998) Longterm effects of incubation temperatures on the morphology and locomotor performance of hatchling lizards (Bassiana duperreyi, Scincidae). Biological Journal of the Linnean Society, 63, Frazer, N.B., Gibbons, J.W. & Greene, J.L. (1990) Exploring Fabens growth interval model with data on a long-lived vertebrate, Trachemys scripta (Reptilia: Testudinata). Copeia, 1990, Gates, J.E. & Gysel, L.W. (1978) Avian nest dispersion and fledging success in field forest ecotones. Ecology, 59, Germaine, S.S. & Wakeling, B.F. (2000) Lizard species distributions and habitat occupation along an urban gradient in Tucson, Arizona, USA. Biological Conservation, 97, Goldsbrough, C.L., Hochuli, D.F. & Shine, R. (2004) Fitness benefits of retreat-site selection: spiders, rocks and thermal cues. Ecology, 85, Huang, W.-S. (1994) Report on egg clutch size of the long-tailed skink, Mabuya longicaudata from Taiwan. Journal of Taiwan Museum, 47, Huang, W.-S. (2006a) Ecological characteristics of the skink, Mabuya longicaudata, on a tropical East Asian island. Copeia, 2006, Huang, W.-S. (2006b) Parental care in the long-tailed skink, Mabuya longicaudata on a tropical Asian island. Animal Behaviour, 72, Huang, W.-S. (2007) Costs of egg caring in the skink, Mabuya longicaudata. Ecological Research, 22, Kearney, M.R. & Porter, W.P. (2009) Mechanistic niche modelling: combining physiological and spatial data to predict species ranges. Ecology Letters, 12, Kearney, M.R., Shine, R. & Porter, W.P. (2009) The potential for behavioral thermoregulation to buffer cold-blooded animals against climate warming. Proceedings of the National Academy of Sciences, 106, Kolbe, J.J. & Janzen, F.J. (2001) The influence of propagule size and maternal nest-site selection on survival and behaviour of neonate turtles. Functional Ecology, 15, Kolbe, J.J. & Janzen, F.J. (2002) Impact of nest-site selection on nest success and nest temperature in natural and disturbed habitats. Ecology, 83, Lebreton, J.-D., Burnham, K.P., Clobert, J. & Anderson, D.R. (1992) Modeling survival and testing biological hypotheses using marked animals: a unified approach with case studies. Ecological Monographs, 62, Lo wenborg, K., Shine, R., Kärvemo, S. & Hagman, M. (2010) Grass snakes exploit anthropogenic heat sources to overcome distributional limits imposed by oviparity. Functional Ecology, 24, Mitchell, N., Kearney, M.R., Nelson, N.J. & Porter, W.P. (2008) Predicting the fate of a living fossil: how will global warming affect sex determination and hatching phenology in tuatara? Proceedings of the Royal Society. Biological Sciences, 275, Morjan, C.L. (2003) Variation in nesting patterns affecting nest temperatures in two populations of painted turtles (Chrysemys picta) with temperaturedependent sex determination. Behavioral Ecology and Sociobiology, 53,