Comparative Reproductive Allometry of Syntopic Black-Knobbed Sawbacks (Graptemys nigrinoda) and Alabama Map Turtles (Graptemys pulchra) in the Alabama River, with Comparison to Three Congeners
Abstract
In many freshwater turtle species, increasing adult female body size is accompanied by increases in both clutch size and egg size. Because clutch size and egg size represent competing sinks for the increasing abdominal space that can be devoted to reproductive potential as a female grows larger, both are hypoallometrically related to body size, i.e., each increases at less than the rate at which body size is increasing. I studied the reproductive allometry of 2 species of the map turtle and sawback genus (Emydidae: Graptemys) in syntopy in the Alabama River. Relative to congeners, Graptemys nigrinoda is a small-bodied species and Graptemys pulchra is a large-bodied species. In the study populations, adult female G. pulchra were on average 23% larger in plastron length and 125% larger in body mass and there was no overlap in size. Both species had positive but hypoallometric relationships of both clutch size and egg size with body size, a pattern reported for 3 congeners previously, thus further demonstrating the competing sinks of increases in both variables and refuting the optimal egg size hypothesis. For interspecific comparisons using analysis of covariance, there was no significant effect of species on clutch size after accounting for the effect of plastron length. The effect of species on egg width and width of the pelvic aperture was significant, with G. pulchra having relatively wider eggs and pelvic apertures, perhaps related to the more domed carapace and larger body sizes of their hatchlings. Compared with 3 congeners, the 2 Alabama River species had relatively small clutches of relatively large eggs. Latitudinal effects on reproductive output may be important in Graptemys, with more-northerly populations tending toward larger clutches of smaller eggs. There is a need to analyze more data from more populations with log-log analyses of reproductive allometry to confirm the trend. There is also a need to consider shape allometry in future analyses; in particular, the hypoallometry of both shell height and shell width relative to shell length.
Female body size is an important correlate of reproductive output in animals (Blueweiss et al. 1978; Stearns 1992). Variation in adult female size is substantial and is thus an important consideration in studies of the life history of reptile populations. For example, in a review of turtle populations, 0.72 was a relatively constant and size-independent mean ratio of carapace lengths of the smallest and largest reproductive females (Shine and Iverson 1995). The difference in body size between smallest and largest body sizes stems in part from indeterminate or post maturational growth of individuals and in part from variation in body size at maturation, as influenced by juvenile growth rates (Halliday and Verrell 1988; Congdon et al. 2013). Because the 0.72 mean ratio reported for turtles is for a linear measurement, the difference is magnified with regard to three-dimensional body volume and thus mass. On average, the largest turtle in a hypothetical population is not only 39% longer (by the ratio 1/0.72 = 1.39) but also up to 39% wider and 39% higher domed (although allometry reduces these percentages). The largest female would thus be expected to have as much as 2.7 times more body volume (1.393 = 2.7, i.e., length × width × height) than the smallest female. There is thus substantial opportunity for female turtles to invest more energy in clutch production with increased body size.
Optimal egg size theory predicts that females will show little variation in egg size and that the variation will not be related to body size, while clutch size will increase linearly with the size of the female (Smith and Fretwell 1974; Brockelman 1975). In many turtle species, however, females produce larger eggs at larger body size, which runs contrary to optimal egg size theory and may occur because anatomical constraints of the pelvis or the gap between the carapace and plastron preclude egg size optimization in smaller females (Tucker et al. 1978; Congdon and Gibbons 1987; Long and Rose 1989; Clark et al. 2001). If total female reproductive output scales isometrically with body size (i.e., if total output scales with increasing abdominal volume) and if females increase both egg size and clutch size at larger body sizes, then neither will be expected to increase isometrically with body size (Ryan and Lindeman 2007; Iverson et al. 2019). Dual hypoallometry of egg size and clutch size, in which both parameters increase at a significantly slower rate than female body size, has been reported for 10 turtle species (Ryan and Lindeman 2007; Macip-Ríos et al. 2012; Naimi et al. 2012; Fehrenbach et al. 2016; Iverson et al. 2019).
Three of the 14 species of map turtles and sawbacks (genus Graptemys) are among the turtle species that have had their reproductive output analyzed for evidence of allometry. Results have been largely congruent, with a pattern by which females increase their investment in clutches as they grow larger by increasing both clutch size and egg width, such that neither parameter can increase isometrically with female size (Table 1). However, little has been reported regarding the reproduction of either of the species of Graptemys that are endemic to the Mobile Bay drainage of southern and central Alabama, northeastern Mississippi, and northwestern Georgia. Lahanas (1982) reported the only data on clutch and egg sizes for the black-knobbed sawback, G. nigrinoda, based on 7 clutches in the Tensaw Delta, where the species is relatively large in body size compared with populations from more lotic, upstream localities (Fehrenbach et al. 2016). Lovich et al. (2014) reported on clutch and egg sizes for the Alabama map turtle, G. pulchra, based on 2 clutches excavated along the lower Tallapoosa River, the easternmost major tributary within the Mobile Bay system. The data from these 2 studies included no information on the size of the females that laid the 9 clutches reported. Coleman (2020) reports on clutch sizes and female body sizes for 6 clutches of G. pulchra.
I radiographed gravid females of both endemic Alabama drainage species of Graptemys in order to examine the relationships of clutch size, egg width, and pelvic aperture width to female body size. I also compared the 2 sympatric species in statistical models to determine whether or not they differ with regard to how these variables are influenced by body size. Finally, I added data from the 3 Graptemys species previously studied for reproductive allometry, which allowed me to make similar comparisons across 5 species that span the diversity of the genus.
METHODS
From 2008 to 2013, turtles were captured in the Alabama River in Autauga and Dallas counties, Alabama (site description in Lindeman 2016) using hoopnets modified as basking traps, which were suspended under angled logs and branches that the turtles used as basking sites (Lindeman 2014), and unbaited fykenets (Vogt 1980a). Turtles were measured for midline plastron length (PL) with a flexible plastic ruler, weighed with Pesola spring scales to determine body mass (BM) to the nearest 1 g (if < 1 kg) or the nearest 10 g (if ≥ 1 kg), and marked using notches in marginal scutes of the carapace (Cagle 1939). For each species, I calculated the ratio of smallest to largest adult female using PL.
From 15–21 May 2009, 1–8 July 2011, 12–18 June 2012, and 21–25 June 2013, turtles determined to be gravid via palpation of their inguinal regions were transported to a veterinary clinic for radiographic determination of clutch and egg size. Film radiographs were taken through 2012 and digital radiographs were taken in 2013. Egg widths (XREW) and pelvic aperture width (PAW) of each radiograph were measured with Vernier calipers using correction for a size standard placed on the anterior part of the plastron of a female placed upside down on the radiography table (standard US quarter; Graham and Petokas 1989). I used mean and maximum XREW per clutch to examine overall variation in egg width as well as the largest egg each female would have to pass through her pelvic opening. To test for a possible bias in radiographic method, I compared log10-transformed mean and maximum XREW and log10 PAW of film and digital radiographs in an analysis of covariance (ANCOVA) with log10 PL as the covariate for the G. nigrinoda data set (low sample size of digital radiographs precluded a similar analysis for G. pulchra).
Mean PL, XREW, and PAW are reported with standard errors (SE). No G. nigrinoda were radiographed more than once, but 1 female G. pulchra was radiographed twice in consecutive years and entered in the data set only once for analysis, using mean values that were highly similar for all parameters (identical PL, difference of 1 egg in clutch size, and XREW means and maxima and PAW that each differed by < 1 mm).
For each log-log regression slope relating clutch size, XREW, or PAW to PL, 95% confidence intervals were computed and used to identify hypoallometry, i.e., significant departure from isometry (failure of the confidence limits to include the value expected under isometry; Ryan and Lindeman 2007). For G. pulchra, I conducted the analyses relating clutch size to PL both with and without data from 6 females from the Cahaba River radiographed by Coleman (2020); I also used ANCOVA to determine whether there was a difference in the relationship between the 2 sites. Because Coleman measured midline PL with calipers, I used the equation PLruler = –1.75 + 1.04 × PLcaliper (P.V.L., unpubl. data, 2017) to estimate PL as measured using a flexible ruler from caliper measurements and to make Coleman's measurements comparable to mine.
I used ANCOVA to compare log-log regressions of clutch size, XREW, and PAW on PL between the 2 Alabama River species. The 3 F-values returned in ANCOVA output were based on Type I (sequential) sums of squares and evaluated the influences on the dependent variable of a) log PL, b) species, and c) their interaction, with the interaction term constituting a test for heterogeneity of the slopes of the 2 species (McDonald 2014).
I also used ANCOVA to make similar comparisons in a data set expanded to include data from G. versa (Lindeman 2005), G. geographica (Ryan and Lindeman 2007), and G. sabinensis (Fehrenbach et al. 2016). Graptemys pulchra represents a clade of 5 species that are characterized by the exceptional megacephaly of adult females; G. nigrinoda, G. versa (data from Lindeman 2005), and G. sabinensis (data from Fehrenbach et al. 2016) represent the largest clade of 8 species, which are characterized by adult females exhibiting either mesocephaly (moderately broad head width; G. versa) or microcephaly (narrow head width; G. nigrinoda and G. sabinensis); and G. geographica (data from Ryan and Lindeman 2007) is the outgroup species of the genus and also has mesocephalic adult females (Lamb et al. 1994; Lindeman 2000; Stephens and Wiens 2003; Praschag et al. 2017; Thomson et al. 2018). For G. versa and G. geographica, PAW was measured from radiographs with correction for magnification as described above; egg width data from excavated nests of G. geographica in Ryan and Lindeman (2007) were not used in the present study. In order to identify which of the 5 species differed significantly in their log PL–corrected dependent variable, I evaluated general linear models (GLM) to which 4 nominal variables (also known as dummy variables) denoting 4 of the 5 species were added (McDonald 2014). While phylogenetic comparative methods, which make statistical corrections for the nonindependence of related species data points, have proven useful in previous studies of trophic morphology and body size in Graptemys (Lindeman and Sharkey 2001; Lindeman 2008), they were not employed in the present case because the data points I analyzed represent individual females rather than species, and only 5 of the 14 species of Graptemys have been studied for allometric scaling of reproductive variables, greatly reducing potential degrees of freedom for correlation or ANCOVA analyses.
Each of the GLM models produced an equation for predicting its log-transformed dependent variable based on a global, genus-wide slope for log PL that applied to the species that was left out of the nominal variables, plus 4 other slope parameters that denoted how much to add or subtract to the dependent variable to account for each of the species that was coded with a nominal variable. Because each nominal variable denoting a species is evaluated with a t-statistic that determines whether the added or subtracted parameter is significant relative to the default species, each of these models was run in all 5 possible permutations of 4 nominal variables, i.e., leaving each species out once in a permutation. The t-statistics generated compare each of 10 possible pairs of species; thus statistical significance for the results of these analyses was evaluated with a simultaneous Bonferroni correction to α (0.05/10 = 0.005).
In the present study, gravid female G. nigrinoda and G. pulchra did not overlap in body size (see “Results”), which is potentially problematic for statistical comparisons of the 2 species. ANCOVA assumes a linear relationship of the dependent variable to the covariate for both groups being compared, although they may have a curvilinear relationship. The other 3 Graptemys species also only partially overlap one another and the 2 species from Alabama in body size. However, log-log transformation of variables for regression, besides being necessary for investigation of allometry, should also alleviate concerns about potential curvilinear relationships of variables. All analyses were conducted in S-PLUS (Insightful Corporation 2006).
RESULTS
Radiographic and Morphometric Data. — I radiographed 31 gravid G. nigrinoda and 11 gravid G. pulchra (one twice in different years). There was no overlap in body sizes between gravid G. nigrinoda (PL range, 135–180 mm; BM range, 523–954 g) and gravid G. pulchra (PL range, 185–207 mm; BM range, 1400–1940 g; Table 2, Fig. 1). For PL, the ratio of smallest to largest female was 0.72 for G. nigrinoda and 0.89 for G. pulchra. Gravid G. pulchra were on average 23% longer in PL and 129% heavier in mass than gravid G. nigrinoda, nearly identical to figures reported earlier for pooled gravid and nongravid females of the populations (23% and 125%, respectively; Lindeman 2016).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 19, 2; 10.2744/CCB-1399.1
Mean clutch sizes were 4.0 for G. nigrinoda (range, 2–8) and 5.4 for G. pulchra (range, 4–7), i.e., clutches averaged having 36% more eggs in the latter species (Table 2). For G. nigrinoda, the mean and maximum XREWs and PAWs of 19 film radiographs did not differ from those of 12 digital radiographs in ANCOVA (F1,27 = 0.012–0.11; p = 0.74–0.91), and slopes for film and digital radiographs were homogeneous (F1,27 = 0.02–0.11; p = 0.75–0.89), thus XREW and PAW data from the 2 methods were pooled within each species. Individual eggs ranged in width from 18.9 to 26.0 mm in G. nigrinoda and from 22.3 to 27.2 mm in G. pulchra. Egg width within a clutch was on average 10% larger (per-clutch means) or 11% larger (largest eggs) in G. pulchra compared with G. nigrinoda (Table 2). Similarly, PAW was on average 23% greater in G. pulchra than in G. nigrinoda (Table 2).
Intraspecific Correlation Analyses. — The log-log correlations of clutch size, EW, and PAW with female PL are summarized for both species in Table 3. In G. nigrinoda, clutch size was significantly related to PL with a log-log slope (2.67) that was slightly hypoallometric but not different from isometry. The relationships of XREW to PL, which were marginally nonsignificant for within-clutch means and significant for the largest eggs in clutches, were strongly hypoallometric in both cases. The relationship of PAW to PL was also marginally nonsignificant and hypoallometric. In ANCOVA, slopes for the relationships of XREW and PAW to PL were significantly heterogenous for within-clutch means (F1,58 = 4.61, p = 0.036) and approached significant heterogeneity for the largest eggs within a clutch (F1,58 = 3.31, p = 0.074), with the slope for PAW being steeper in both comparisons.
In G. pulchra, clutch size was marginally nonsignificantly related to PL with a log-log slope of 2.98 that was almost perfectly isometric, albeit with very wide 95% confidence limits (–0.03–5.99; Table 3). Slopes for the relationship of XREW (means and largest eggs) and PAW to PL were not significant. In ANCOVA, slopes for the relationships of XREW and PAW to PL were not significantly heterogenous for within-clutch means (F1,18 = 0.24, p = 0.63) or for the largest eggs within a clutch (F1,18 = 0.47, p = 0.50), although the slope for PAW was steeper in both comparisons.
When I analyzed the correlation of clutch size with PL for a data set combining my Alabama River data with data from 6 females in the Cahaba River (mean clutch size = 9.2, range 4–13; mean estimated PLruler = 225.3, range 176–255; Coleman 2020), Site was not a significant contributor to the model in ANCOVA (F1,13 = 2.66, p = 0.13), slopes for the 2 data sets were homogenous (F1,13 = 0.04, p = 0.85), and clutch size was significantly positively related to clutch size (F1,13 = 69.95, p < 0.0001). The log-log slope for the pooled data set of 17 females was 3.14 (95% confidence limits, 2.32–3.96), i.e., it remained isometric.
Interspecific Comparisons, Alabama River Species. — A log-log ANCOVA found no departure from homogeneity of slope between the 2 species for clutch size as influenced by female PL (F1,38 = 0.09, p = 0.76), with no difference between the 2 species in PL-corrected clutch size (F1,38 = 1.18, p = 0.29). Results for mean XREW showed homogeneity of slopes (F1,38 = 0.13, p = 0.72) and approached significance for interspecific differences in PL-adjusted egg width (F1,38 = 3.96, p = 0.053). Homogeneity of slopes with significant interspecific differences was found for both maximum XREW (homogeneity of slopes, F1,38 = 0.76, p = 0.39; interspecific differences, F1,38 = 5.81, p = 0.021) and PAW (homogeneity of slopes, F1,38 = 0.000008, p = 0.998; interspecific differences, F1,38 = 5.23, p = 0.028); female G. pulchra produced relatively wider eggs and had relatively wider pelvic apertures. In all of these analyses, the covariate PL was found to be a highly significant influence (all F1,38 > 28, p < 0.0001).
Interspecific Comparisons, 5 Graptemys Species. — The log-log slopes relating clutch size, mean and maximum XREW, and PAW were found to be homogeneous in ANCOVAs analyzing data from all 5 of the studied species of Graptemys, with highly significant effects of both species and the covariate, female PL (Table 4; Figs. 2 and 3). In GLM analyses (Table 5), the 2 Alabama River species, G. nigrinoda and G. pulchra, had relatively small clutches that did not differ, while G. geographica and G. versa had relatively large clutches that did not differ, and G. sabinensis had intermediate clutch sizes that were statistically similar in PL-corrected size to only those of G. pulchra. The only northern species analyzed, G. geographica, had smaller mean egg width compared with the 4 southern, Gulf Coastal species, which did not differ; results were similar for maximum egg widths except that G. geographica and G. versa also did not differ after Bonferroni correction to α (p = 0.013). In G. pulchra, PAW was large compared with the other 4 species, which did not differ from one another, although G. pulchra was significantly larger in PAW only compared with G. geographica and G. sabinensis, with comparison of G. pulchra with the other 2 species approaching but not achieving significance under the Bonferroni correction to α (p = 0.019 for comparison with G. nigrinoda and p = 0.0051 for comparison with G. versa).
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 19, 2; 10.2744/CCB-1399.1
Citation: Chelonian Conservation and Biology: Celebrating 25 Years as the World's Turtle and Tortoise Journal 19, 2; 10.2744/CCB-1399.1
In GLM analyses, global log-log slopes for clutch size, mean and maximum XREW, and PAW were positive, significant, and hypoallometric (Table 4). The slopes for mean and maximum XREW were less steep than the slope for PAW; thus the difference between egg width and the pelvic aperture (egg clearance) was greater in larger females, as can be seen in comparison of the 2 panels in Fig. 3.
DISCUSSION
The mean clutch size of G. nigrinoda from the Alabama River (4.0) is 27% smaller than the mean of 7 G. nigrinoda clutches (5.5) reported by Lahanas (1982), a difference that can probably be attributed to the difference in body size between the specimens from the Alabama River (to 180 mm PL) and Lahanas' specimens from the Tensaw Delta (to 202 mm PL; Lindeman 2008). Small clutch sizes are similarly typical of the 2 other members of the sawback clade: G. flavimaculata (mean, 4.7; Horne et al. 2003) and G. oculifera (mean, 3.6; Jones 2006). The mean clutch size of G. pulchra from the Alabama River (5.4) is comparatively small, both relative to clutch sizes of larger-bodied G. pulchra from the upper Cahaba River (mean, 9.2; Coleman 2020) and those of the other 4 Graptemys species in the megacephalic clade, all of which also reach larger female body sizes than do the G. pulchra of the Alabama River (see below): G. barbouri (means of 5.2 and 8.8; Cagle 1952; Ewert and Jackson 1994; Ewert et al. 2006); G. ernsti (mean, 7.2; Shealy 1976); G. gibbonsi (mean, 7.5; Vogt et al. 2019a); and G. pearlensis (mean, 6.4; Vogt et al. 2019b).
The results of analyses of reproductive allometery in syntopic populations of G. nigrinoda and G. pulchra in the Alabama River only partially mirrored results in previous studies of 3 species of Graptemys in which dual hypoallometric increase of clutch size and egg width with increasing female size was the rule (Lindeman 2005; Ryan and Lindeman 2007; Fehrenbach et al. 2016; Tables 1 and 3). The correlation of clutch size with female body size was highly significant for G. nigrinoda but marginally nonsignificant for G. pulchra, likely due to a small sample size. However, after including additional data from the Cahaba River, the correlation was significant. Both log-log slopes were near the isometric expectation of 3 rather than being closer to half that value as in the previous studies. The correlation of egg width with female body size was significant for maximum egg width per clutch and approached significance for mean egg width per clutch in G. nigrinoda, which exhibited significantly hypoallometric relationships of the variables, while neither relationship was significant for G. pulchra.
The lack of consistent results, both between the 2 Alabama River species and with previously studied congeners, may stem from a combination of the effects of low sample sizes and the possibility of interannual environmental effects on reproductive output. Low sample sizes are associated with many of the greater departures from patterns observed in other studies of reproductive allometry in turtles (Iverson et al. 2019). Overall, results for G. nigrinoda (n = 31) were more similar to those from previous studies of Graptemys species than were those of G. pulchra (n = 11). Data were collected during 4 yrs of a 5-yr span and at variable times of the active season, so any variation in factors (e.g., climate, prey availability, clutch order) that may have promoted greater or lesser reproductive output in any 1 yr may also have contributed to the departure of the results from those of previous studies.
The 2 sympatric species in the Alabama River were shown to have similar relationships of clutch size, mean and maximum egg width, and pelvic aperture width to female body size, as the slopes relating clutch size to body size were homogeneous and there was no difference between the species in how clutch size scaled with body size. Tests for slope homogeneity for mean and maximum egg width and pelvic aperture width similarly produced very high p-values (≥ 0.72), but eggs and pelvic apertures were both absolutely and relatively wider in the larger-bodied species, G. pulchra. The differences in egg and pelvic aperture widths may relate to the slightly more domed shells of G. pulchra, which could require larger eggs (2%–3% greater carapace height, not including the carapacial keels; P.V.L., unpubl. data, 2008–2013). The differences also may allow for production of slightly larger hatchlings in G. pulchra. Producing larger hatchlings in G. pulchra would give their hatchlings a head start on the larger sizes that they achieve relative to G. nigrinoda at the study site (23% longer and 125% heavier in females and 10% longer and 36% heavier in males; Lindeman 2016). Comparative data on hatchling sizes in these 2 species are not yet available.
Mature females of 3 coexisting species of Graptemys overlap broadly in body size in the upper Mississippi River (Vogt 1980b). In contrast, megacephalic females of G. pulchra, G. gibbonsi, and G. pearlensis all achieve considerably larger body sizes than do sympatric micro-cephalic sawback species (G. nigrinoda, G. flavimaculata, and G. oculifera, respectively; Lindeman 2008). In the Alabama River population sampled for the present study, the ratio of smallest to largest gravid G. nigrinoda was 0.75, just above the average reported for all turtles (0.72) by Shine and Iverson (1995). Similarly, ratios for other Graptemys species studied previously are 0.71 for G. versa in the South Llano River of Texas, 0.74 for G. geographica in Lake Erie in Pennsylvania, and 0.71 for G. sabinensis in the West Fork Calcasieu and Mermentau rivers of Louisiana (Lindeman 2005; Ryan and Lindeman 2007; Fehrenbach et al. 2016). The ratio for G. pulchra in the Alabama River (0.89) is therefore a decided outlier, exceeding even the highest ratios reported by Shine and Iverson (1995) in their review of turtle mature body sizes (0.83, for populations of the emydid species Terrapene ornata and Chrysemys picta). Furthermore, Coleman's (2020) 6 gravid females in the Cahaba River have a ratio of 0.69.
The possibility that the upper and lower limits of female size reported here do not adequately represent the full range of adult body size for G. pulchra at the Alabama River site must be considered. Within the range of body sizes of radiographed G. pulchra, 12 of 23 captures were of gravid females. The largest female G. pulchra captured for the present study was 15 mm smaller than the largest female measured in a large series of museum collections (PL 222 mm, a female collected in the Coosa River in Talladega County; Lindeman 2008) and 48 mm smaller than the largest female reported by Coleman (2020; estimated at 255 mm if measured with a ruler). Within the megacephalic clade of Graptemys, the 4 other species have been recorded with midline plastron lengths as great as 215 mm (G. pearlensis), 228 mm (G. barbouri), 231 mm (G. ernsti), and 242 mm (G. gibbonsi; Lindeman 2008; P.V.L., unpubl. data, 1994–1996, 2015–2018). There is no reason to suspect that the hoopnet basking traps used missed larger female G. pulchra (no female G. pulchra were captured in fykenets), however, because larger turtles of other species have regularly been caught in such traps (Lindeman 2014; P.V.L., unpubl. data, 2011–2018). At the lower extreme, 8 females ranging in PL from 168 to 184 mm (mean, 176.0 mm) were captured between the dates of 17 June and 6 July and were not gravid; thus females slightly smaller than the smallest gravid female were well sampled. It appears that the Cahaba River turtles studied by Coleman (2020) are simply a large-bodied population of the species. Reasons for the interpopulational difference in body size, which may include differences in growth rate, age at maturation, food resource availability, hydrologic differences (i.e., dammed vs. undammed rivers), and/or survivorship and longevity, remain to be investigated.
The analyses of 5 species of Graptemys for which data are available reveal high similarity in the 2 sympatric species from Alabama, in contrast to the differences that were found in analyses that included just these 2 species, and found significant differences in egg width and pelvic aperture width. Compared with their other congeners, the 2 Alabama species have relatively small clutches of relatively wide eggs and relatively large pelvic apertures for their body sizes. A contrasting trend is exhibited by G. geographica in Pennsylvania, which exhibits relatively large clutches of relatively narrow eggs and a relatively small pelvic aperture. The other 2 species, G. versa from Texas and G. sabinensis from Louisiana, are intermediate to these patterns.
The interspecific comparisons support the existence of anatomical constraints on egg size (Tucker et al. 1978; Congdon and Gibbons 1987) in these 5 species because increased egg width was associated with increased pelvic aperture width and vice versa. Comparison of the slopes for the egg width vs. the pelvic aperture width regressions suggests that the constraint is progressively lessened with body size in Graptemys populations, however, as the clearance (i.e., the difference between egg width and pelvic aperture width; Lovich et al. 2012; Rothermel and Castellón 2014; Kern et al. 2016) becomes substantially greater in larger turtles. In other words, there is heterogeneity in the rates of increase of egg width and pelvic aperture width relative to the rate of increase in female size, with egg width being slower. The possibility that medium- to large-sized females are freed from constraint and able to produce optimal egg sizes that cease to increase with the size of the female (Ryan and Lindeman 2007; Rollinson and Brooks 2008; see fig. 4iii in Lindeman 2020) has yet to be rigorously investigated but would result in heterogenous rates of increase, with pelvic aperture width having a steeper log-log slope (and better fit to the PL) than egg width. A thorough investigation of this hypothesis would require a very large sample size from a population; the sample should either be collected all in 1 yr, or relatively evenly spread over multiple years, with year taken into account in statistical modelling. An alternative hypothesis is that egg size continues to increase with female body size in the larger females, but at a slower rate due to the increase in clutch size that also takes place (Ryan and Lindeman 2007).
There are 2 primary determinants of body size in females of Graptemys populations: latitude (for the 3 species that range widely from south to north in the Mississippi River drainage) and relative head width (Lindeman 2008). Northern populations of G. geographica, G. ouachitensis, and G. pseudogeographica have large-bodied females, while at southern latitudes where the other 11 species have more restricted ranges, microcephalic females are small compared with megacephalic females, with mesocephalic females being intermediate. Clutch size and egg width data for Graptemys populations (summarized in Lindeman 2013; see also Vogt et al. 2019a, 2019b) suggest a trend similar to the pattern revealed in the 5-species ANCOVAs of the present study, in which G. geographica from a northern site in Pennsylvania contrasted with the other 4 southern species, having large clutches of relatively small eggs. Genus-wide, all reports of clutch size averages > 10 eggs are from G. geographica, G. ouachitensis, and G. pseudogeographica populations at middle to northern latitudes. This is in spite of the fact that females of 5 species of the Gulf Coastal megacephalic clade, to which G. pulchra belongs, achieve similar shell lengths to the females of northern populations of G. geographica, G. ouachitensis, and G. pseudogeographica (and are more massive; P.V.L., unpubl. data, 2002–2020). Eggs for other northern populations of the 3 widespread species are generally small, while those of the species of the megacephalic clade in the south are generally large, although their log-log relationships with female body size cannot be reconstructed from the published data. Latitudinal trends toward larger clutches of smaller eggs at northern latitudes have also been reported for turtles in general (Iverson et al. 1993) and the wide-ranging coastal species, Malaclemys terrapin, the sister species to Graptemys (Lovich et al. 2018). Iverson et al. (1993) discussed several hypotheses that might explain selection for smaller eggs at northern latitudes.
More data from the 3 wide-ranging species are necessary to fully investigate latitudinal trends, particularly with inclusion of southern populations, which have been little studied for these 3 species. I predict that detailed allometric assessment of reproductive parameters of populations of these 3 species along a latitudinal gradient would show intraspecific patterns of larger clutches of relatively smaller eggs in more-northern populations, suggestive of a clutch-size–egg-size trade-off that is mediated by climatic effects (see reviews of hypotheses in Iverson et al. 1993 and Lovich et al. 2018). Such a pattern also suggests biomechanical considerations, as the pelves of southern species must be able to accommodate their relatively wider eggs. Enlargement of the pelvis may compromise its locomotory or other functions (Congdon and Gibbons 1987).
In addition, other aspects of egg size (egg length, egg volume or mass) should also be investigated in future studies of reproductive allometry in Graptemys, but will require measurements taken on deposited clutches (Iverson et al. 2019). In G. geographica in a Pennsylvania population, egg width and egg mass showed similarly hypoallometric relationships to female size, while egg length was not related to female size (Ryan and Lindeman 2007). Egg length also did not correlate with female size in G. versa (Lindeman 2005), but no other assessment of reproductive allometry in Graptemys using other egg size measurements has been conducted.
The fact that pelvic aperture width did not scale isometrically with plastron length (global slope for 5 species, 0.69, with an upper limit on the 95% confidence limit of 0.86) is problematic. This result likely stems from the fact that turtles change in body proportions as they grow, i.e., they do not grow in length, width, and height at equal rates; they become relatively longer and less domed (Mosimann 1958; Fish and Stayton 2014). Hence width and height of the shell—the 2 parameters most likely to relate to the size of the pelvic opening—are hypoallometric relative to shell length. Given allometric changes in the 3-dimensional shape of female turtles, future investigations of reproductive allometry should include analyses relating reproductive variables to shell width and height as well as shell length in order to better refine our ideas of what are the expected slopes under isometry. Ideally, we should use slopes and their confidence limits from log-log regressions to analyze reproductive variables and distinguish hypoallometry from isometry via consideration of overall body shape of females rather than just their relative length.
Variation in gravid female body size, labeled by midline plastron lengths (PL, in mm), in Graptemys nigrinoda (top row) and Graptemys pulchra (bottom row) from the Alabama River. The 5 G. nigrinoda shown encompass 71% of the total range observed for PL (135–180 mm), while the 4 G. pulchra shown include the largest and smallest individuals in the study.
The log-log relationship of clutch size to female plastron length in 5 species of Graptemys. Dotted lines are species-specific regression lines (denoted to the left with the first three letters of each specific epithet), while the solid line is a global regression line from general linear modeling of the relationship (Table 3).
The log-log relationships of mean egg width (XREW) and pelvic aperture width (PAW) to female plastron length in 5 species of Graptemys. Dotted lines are species-specific regression lines (denoted to the left with the first 3 letters of each specific epithet) while the solid lines are global regression lines from general linear modelling of the relationships (Table 3). Results for largest eggs in clutches were similar to those for mean egg width (see Table 3).
Contributor Notes
Handling Editor: Will Selman
