Reproduction and Nesting of the Endangered Ringed Map Turtle, Graptemys oculifera, in Mississippi
ABSTRACT
Reproduction and nesting in the ringed map turtle (Graptemys oculifera) were investigated in the Pearl River of west-central Mississippi in 1995 and 1996. Nesting occurred from mid-May until mid-July but peaked in mid-June. Minimum carapace length of females at sexual maturity was 130 mm, but mean size at maturity was between 130 and 140 mm. Mean and modal CSs were 3.7 and 3. Larger females were gravid earlier than smaller ones and both egg and CS declined as the nesting season progressed. CS was positively correlated with both female carapace length and body mass. Mean egg length, width, and mass were 38.8 mm, 22.7 mm, and 11.8 g, respectively. Hatchlings averaged 35.5 mm carapace length and 8.9 g in mass. Annual clutch frequency may range from 0.96 to 1.42, and a minimum of ca. 60% of females reproduced on an annual basis. Predators destroyed an average of 86% of nests each year. Major vertebrate nest predators were armadillos, raccoons, and fish crows. Invertebrate egg predators destroyed an additional 24% of eggs known to be fertile.
Many turtles of conservation interest are inhabitants of rivers, but as Jackson and Walker (1997) pointed out, we know less about the ecology of riverine turtles than of almost any other chelonian group. Graptemys, with 12 described species, is a genus of small to medium-sized turtles that inhabit rivers and streams of Gulf of Mexico drainages in central and eastern North America (Ernst et al. 1994). One of these, the ringed map turtle (Graptemys oculifera), is restricted to the Pearl River watershed of Mississippi and Louisiana, where it occurs primarily in the main channels of the Pearl River and its largest tributary, the Bogue Chitto River. Although G. oculifera has been listed as a threatened species by the US Fish and Wildlife Service since 1986 and is listed as endangered by the IUCN Red List (IUCN 2004), little information is available on its nesting biology and reproduction.
Females mature at a relatively small size (115–120 mm plastron length) and at a relatively late age (10 years) (Jones and Hartfield 1995). Cagle (1953) described a nesting attempt and provided data from a gravid female indicating that she would have produced 2 clutches in a season. Anderson (1958) briefly described the locations of nests on a sandbar in the Pearl River and provided information on the movements of hatchlings at emergence. Kofron (1991) examined a series of museum specimens and concluded that ovulation and nesting occurred from May through July, that hatching occurred in August, and that CS (CS) was 2–3 eggs. Given the paucity of data on reproduction in G. oculifera and its conservation status, the objectives of this study were to acquire information on female body size, CS, clutch frequency, egg size, hatchling size, nesting season, the incubation period, nesting success, and nest predation in the species.
METHODS
The study was conducted on the Pearl River near Ratliff Ferry, Madison County, Mississippi (Fig. 1), an area previously identified as having a high density of G. oculifera (Jones and Hartfield 1995). The river at the study site ranged from 40 m in width in side channels to approximately 200 m in bendways. Water depths ranged from < 2 to > 9 m. The 30-year (1962–1992) average discharge of the Pearl River at Jackson (ca. 45 km downstream) was 4627 cubic feet per second (Plunkett et al. 1993). Highest flows occurred in March and April, and lowest in September. The highest mean monthly temperatures recorded over a 54-year period (1939–1993) at Jackson International Airport, ca. 38 km southwest, occurred in July and the greatest precipitation occurred in March (Ruffner and Blair 1981; Wood 1996).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[195:RANOTE]2.0.CO;2
Females were captured in open-topped traps, constructed of either 1-inch nylon mesh in a metal frame, 1-inch poultry netting, or 5 × 10–cm wire mesh, which were attached to limbs and logs used as basking sites by the turtles. Trapping began on 1 May in 1995 and 30 April in 1996 and continued until no gravid females were present in weekly samples and/or nesting was no longer observed in the study area. Approximately 15 traps were used each day for 4 days during each week of sampling.
Captured turtles were permanently marked by drilling holes in the marginals using Cagle's (1939) numbering system. Straight-line carapace length (CL) and maximum plastron length (PL) were measured to the nearest millimeter with aluminum timber calipers. Midline plastron length (MPL) was measured to the nearest 0.1 mm with dial calipers. Body mass (BM) was measured to the nearest gram using a portable electronic balance.
Females were palpated and gravid individuals were transported to a veterinarian for radiography (Gibbons and Greene 1979), which was used to determine CS and clutch frequency. Gravid females found while attempting to nest were measured and marked but were not X-rayed.
Eleven sandbars were monitored for nesting activity by G. oculifera (Fig. 1). Available nesting habitat, measured using a hipchain, was considered to be the area between the river and the treeline at the top of the bar. Both the percentage of vegetative cover and open sand on each bar were estimated by point counts. Most sandbars in the study area were used for camping by recreational boaters for at least part of the nesting season. This occurred sporadically during the week and was greatest on weekends and major holidays. Human presence on the 11 monitored sandbars was recorded daily during the week throughout the 1996 nesting season.
I surveyed sandbars several times each day for intact nests, which were located either by direct observation of nesting females or by following tracks to nest sites. Given the frequency with which sandbars were surveyed, most nests found in the absence of females were from ca. 2 to 12 hours old. Nests were excavated and for each egg the following was recorded: egg length (EL) and egg width (EW), both measured to the nearest 0.1 mm, and egg mass (EM), measured to the nearest 0.1 g. Eggs were then reburied in the nest cavity. Nest sites were plotted on a scale map of the sandbar and keyed to numbered aluminum tags attached to nearby trees. Both distance and compass bearing from the tags were recorded to facilitate relocation of the nests.
Physical characteristics of G. oculifera nest sites were measured for intact nests and when females were discovered while nesting. These included straight-line distances of the nest site from the river and to the closest vegetation, whether the latter was herbaceous or woody, the number of both herbaceous and woody stems within a 0.5-m circle centered on the nest site, the percent shade at 1 m above the nest site as determined with a concave densiometer, soil particle diameter determined using a soil grain sizing folder (Forestry Suppliers No. 77332), and substrate type (sand or sand/soil mixture). Nest cavity depth and diameter were measured when no alteration of these dimensions had been made while excavating the nest cavity. Substrate temperature at the nest site and air temperature 1 m above the substrate in both open and shaded conditions were measured when females were encountered while nesting.
Temperatures were monitored for 10 nests in each year of the study with a HOBO-XT7 data logger (Onset® Computer Corporation) attached to a 1.8-m cable with a thermistor probe. The data logger was sealed in a plastic container and buried ca. 1 m from the nest cavity with the probe inserted at the approximate midpoint of the clutch. Data loggers were programmed to record nest temperatures at the same time each day at 48-minute intervals (30 measurements/day) throughout the incubation period.
Nests were caged in plastic mesh, which prevented predation and permitted an investigation of the incubation period. The mesh was folded into an open box and buried open-side down over the clutch ca. 3 cm below the soil surface. All caged nests were checked for evidence of hatching at 55 days after oviposition and daily thereafter until pipping occurred. The incubation period is defined for this report as the time between oviposition and pipping of the eggs, and emergence as the time when hatchlings moved up through the sand from the nest cavity to beneath the plastic mesh. The latter is reported as both days from clutch deposition and as days since pipping. Eggs that did not pip were returned to the lab and examined under a dissecting microscope to determine if development had occurred.
Caged nests were checked daily to determine if predators had attempted to excavate them. Predators were identified by tracks left in the sand near the nests, and only the first predation attempt for each nest was included in subsequent data analyses. Other G. oculifera nests that had been destroyed by predators (raided nests) were also counted daily on each of the 11 sandbars. The predator, if identifiable from tracks at the nest site, was recorded for each raided nest. Eggshells left by predators at nest sites were removed from the sandbars daily so that raided nests were not counted twice.
Single eggs from 32 clutches and the entire complement of 19 clutches were incubated in the lab in 1996 to investigate the relationship between egg and hatchling size. Eggs were buried up to approximately two-thirds of their width in separate numbered compartments of covered plastic trays filled with moistened vermiculite. Trays were placed in circulated-air incubators (Hovabator Model 1583) at 28.5°–29.5°C and misted weekly to reduce evaporative water loss. CL, PL, and BM were measured for all hatchlings upon emergence from the nest cavity or after yolk absorption in lab-incubated turtles. Hatchlings were then released near the sandbar from which the eggs originally came.
Statistical tests were performed using Statistica (Statsoft, Inc. 1999). Variables were transformed to their natural logarithms where appropriate to meet the assumptions of parametric tests. Comparison-wise error rates were set at alpha = 0.05. Sequential Bonferroni adjustments for multiple comparisons were calculated as in Rice (1989) with an experiment-wise error rate set at alpha = 0.10 (Chandler 1995). Unadjusted p-values are reported, but cases in which adjustment for multiple tests altered significance are so indicated. Figures were produced in part using AXUM (Trimetrix Inc. 1999). All means are followed by 1 standard deviation.
RESULTS
Female Body Size Variation
I captured 705 individual female G. oculifera ranging from 111–215 mm CL during the 2 years of the study. Using the size of the smallest gravid female (130 mm CL) as an estimate of the minimum size at maturity, 670 (95%) of the captured females were potentially adults (Table 1). Of these, 175 were recaptured at least once, and several were recaptured on multiple occasions, resulting in 896 total captures of adult females during this study.
Jones and Hartfield (1995) estimated from a von Bertalanffy growth curve that female G. oculifera matured at a MPL of ca. 115–120 mm. Based on the relationship between MPL and CL found in this study (MPL = 0.8395CL + 4.9139; F(1, 917) = 21,642.7, p < 0.0001) an MPL of 115 mm occurs in females with a CL of approximately 131 mm and an MPL of 120 mm in females with a CL of approximately 137 mm. Mean size at maturity, then, based on both the growth curve and the reproductive data, which appear to be concordant, probably occurs between 130 and 140 mm CL, even though only approximately 7% of captured females in this size range were gravid (Fig. 2). Larger females were also more likely to be gravid when captured than were smaller individuals (Fig. 2), but it is not clear whether this resulted from a relatively greater fecundity of large females or because larger females were more difficult to capture when not gravid.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[195:RANOTE]2.0.CO;2
Body size and capture date were significantly correlated in gravid females (CL: Spearman's r = −0.124, p = 0.04; BM: r = −0.205, p = 0.0006). A plot of CL and capture date (Fig. 3) indicated that females found gravid both early and late in the sampling period were larger than those gravid during the middle part of the period. The sampling period was therefore arbitrarily divided into quarters (before 20 May, 20 May–8 June, 9 June–28 June, after 28 June) and gravid female body size (CL, BM) was compared among them. Females gravid during the first quarter were significantly larger in both CL and BM than those gravid during the second and third quarters (Mann-Whitney U-tests, all p < 0.001), but did not differ from those gravid during the fourth quarter (p > 0.10). Gravid females collected during the second, third, and fourth quarters, however, did not differ in size (Mann-Whitney U-tests, all p > 0.17).
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[195:RANOTE]2.0.CO;2
Oviposition Period
Gravid females were present in trapping samples from 1 May to 12 July in 1995 and from 30 April to 15 July in 1996. The proportion of gravid females in weekly samples ranged to as high as 69% (Fig. 4). The frequency of gravid females in samples peaked during both the third week of May and the second week of June in 1995. The former, however, was based on a relatively small sample size (Fig. 4). In 1996, the number of gravid females captured gradually increased to a maximum during the second week of June. The frequency of gravid females in trapping samples declined markedly in both years after that date (Fig. 4). The first G. oculifera nest was found on 12 May in 1995 and on 14 May in 1996, and the last on 17 July in 1995 and on 19 July in 1996. The length of the nesting season was therefore 67 days in both years.
Citation: Chelonian Conservation and Biology 5, 2; 10.2744/1071-8443(2006)5[195:RANOTE]2.0.CO;2
Female G. oculifera did not appear to retain eggs for long periods of time. Thirty-three females first captured while gravid were recaptured 11–43 days later when they were no longer gravid, presumably having deposited their clutches. Eighteen of these females (54.5%) were recaptured within 15 days of their initial capture. Similarly, 27 females initially captured when not gravid were recaptured 7–42 days later and were found to be gravid. Seventeen of these females (63%) were recaptured within 21 days of their initial capture and apparently became detectably gravid within that 3-week period. Using the minima of these ranges, female G. oculifera may only be detectably gravid with a particular clutch, as determined using both palpation and radiography, for a little more than 2.5 weeks.
CS
Mean CS determined from radiography was 3.78 ± 1.48 (range 1–10, mode = 3, n = 86) in 1995 and 3.59 ± 1.02 (range 1–8, mode = 3, n = 159) in 1996. There was no difference in CS between years when body size (CL) was accounted for (Analysis of Covariance [ANCOVA], F(1, 243) = 1.03, p = 0.311). The combined mean for both years was 3.66 ± 1.20. Mean CS for 134 intact G. oculifera nests found during the study was 3.42 ± 1.11 (range 1–7, mode = 3), which was not significantly different from that observed for X-rayed females (Mann-Whitney U-test; Z = −1.830, p = 0.067).
Both CL and BM were positively and significantly related to CS (CS = 0.054CL − 5.143; F(1, 243) = 245.6, p < 0.001; CS = 0.004BM + 0.882; F(1, 243) = 320.7; p < 0.001), with BM explaining a higher proportion of the variation (r2 = 0.569) in CS than CL (r2 = 0.503). The CS–body size regressions indicated that an increase of 1 additional egg occurred for each approximately 225 g increase in BM or for each approximately 19 mm increase in CL.
CS declined through the nesting season, even when body size (CL and BM) was accounted for (partial r = −0.182, p = 0.006) which may have resulted from fewer eggs in second or third clutches. However, 10 females with 2 measurable clutches in a single season showed no significant difference in CS between their first and second clutches (Wilcoxon matched-pairs test: Z = 0.534, p = 0.593).
Clutch Frequency
Thirteen females were gravid twice in either 1995 (5 of 50 [10%] captured at least twice) or 1996 (8 of 74 [10.8%] captured at least twice) as indicated by radiography and/or capture during nesting attempts. Mean annual clutch frequency (Frazer et al. 1989) for this population was therefore 1.10 and mean annual egg production (mean CS × mean clutch frequency) was 4.03. Mean interclutch interval was 23.3 days (range 12–44 days), but females recaptured more than 21 days apart may have nested during the intervening period. Mean time between clutches if the longest intervals (28, 33, 41, 44 days) were omitted was 16.7 ± 3.8 days.
Mean annual clutch frequency was probably underestimated because females were not recaptured often enough to provide a sample size sufficient to adequately evaluate this trait. A second estimate can be obtained from females initially captured during the first half of the nesting season (prior to 15 June) and recaptured between 12 and 21 days later (range of adjusted interclutch interval). This eliminates females initially captured late in the nesting season when they may have been postreproductive and those recaptured at intervals greater than the hypothesized interclutch period given above. Mean annual clutch frequency using data only from these females (n = 36) was 1.25 and mean annual egg production was 4.58. This, however, ignores females that met the capture–recapture requirements but that were never found gravid (n = 17). Six of these 17 females had weight losses of 21–68 g between captures. If these relatively large weight changes are assumed to have resulted from clutch deposition, and ignoring the 11 females with no weight changes, then mean annual clutch frequency was 1.42 and mean annual egg production was 5.20. If the remaining females that were never found gravid are included, then mean clutch frequency was 0.96 and mean annual egg production was 3.51. Therefore, mean annual clutch frequency may range from 0.96–1.42 and mean annual egg production from 3.51–5.20. It should be emphasized, however, that these are minimum estimates because females that were not gravid when captured may have been so either before or after their capture date, and it is therefore likely that some clutches were missed.
If the average internesting interval was ca. 17 days and if females are assumed to have produced clutches every 17 days, then some females may have produced more than 2 clutches per year. One female, gravid when captured on 20 May 1996, was found attempting to nest 44 days later on 3 July. If the 20 May date is assumed to be her first clutch of the season, she could have produced a second clutch by approximately 18 June and a third by the first week of July. A second female, one of the largest (210 mm CL) captured during the study, was found attempting to nest on 5 June 1996. Because large females tend to be gravid both earlier and later than smaller individuals, this nesting attempt may have represented her second clutch of the season, with the first on approximately 20 May. She was recaptured again while attempting to nest on 16 July. Assuming a 17-day internesting interval, this female may have produced a third clutch around 22 June and potentially a fourth clutch of the season by 16 July.
Eighty-three females were captured in both 1995 and 1996, and only 12 (14.4%) were gravid in both years. Three of these were found at least once on sandbars while gravid but prior to depositing eggs, so their CSs were not determined. CSs between years did not differ significantly (Wilcoxon matched pairs test: Z = 0.534, p = 0.593) for the 9 females with measurable clutches. The 12 females were gravid from 36 days earlier to 19 days later in 1996 than in 1995, and mean date of capture was 4.00 ± 14.46 days earlier in 1996. Dates of capture when gravid were not significantly different between the 2 years (Wilcoxon matched pairs test: Z = 1.098, p = 0.272).
The timing and frequency of recaptures may have also precluded a reliable estimate of the proportion of the population that was gravid in both years. As indicated above, female G. oculifera appeared to nest at approximately the same time each year, and the probability of capturing a female while she was gravid was higher for females captured prior to 15 June. Therefore, to estimate the proportion of the female population that was gravid in both years, only females that were known to have been gravid at least once, that were captured prior to 15 June in both years and that were recaptured in 1996 within ca. 19 days of when they were gravid in 1995 (mean ± 1 SD of difference in capture dates when gravid in 1995 and 1996) were considered. Seventeen females met these requirements, and of these, 10 (58.8%) were gravid in both years.
Some females, as indicated above, reproduced annually. However, recapture data also suggest that some females may not reproduce every year. Two females were captured 4 times in 1996, and were not gravid at any capture. The first female (154 mm CL) was initially captured 7 May, was recaptured 13 days later on 20 May, was recaptured again on 3 June after a period of 14 days, and was last captured 15 days later on 18 June. She may have produced a clutch after her last capture, but because females are apparently detectably gravid for ca. 2.5 weeks, the clutch would have been a relatively late one that would have been produced sometime after 5 July which is almost at the end of the nesting season. A second female (149 mm CL) was first captured on 30 May, was recaptured 12 days later on 11 June, was recaptured again 15 days later on 26 June, and was recaptured a final time 14 days later on 9 July. Body size of these 2 females indicated that they were probably mature and their recapture histories suggest that they did not produce clutches in 1996.
Egg Size
The 136 intact G. oculifera nests found during the study contained 456 eggs (Table 2). Mean egg size per clutch did not differ significantly between years but did among months (Table 2). Eggs deposited in May were significantly larger than June eggs, which were significantly larger than those from July (Table 2). Mean EL per clutch in 23 females captured while nesting was negatively correlated with date of nesting (Spearman's r = −0.507, p = 0.014), even when female body size (CL, BM) was accounted for (partial r = −0.740, p = 0.0001), but neither of the other 2 egg size measures were significantly related to clutch deposition date (EW: Spearman's r = 0.126, p = 0.568; EM: r = −0.199, p = 0.361).
Incubation Period
The first emergence of G. oculifera hatchlings in 1995 was from an unmarked nest on 27 July, and in 1996 from a caged nest on 2 August. There was no difference between years (Table 3) in the length of time between egg deposition and pipping (Mann-Whitney U-test: Z = −0.019, p = 0.985) and in time between deposition and emergence (Z = −1.706, p = 0.88) for caged nests. Data were therefore combined for both years, and mean length of time was 64.4 ± 4.7 days between deposition and pipping, 76.3 ± 7.7 days between deposition and emergence, and 12.0 ± 5.5 days between pipping and emergence. Eggs incubated in the lab, however, pipped in a significantly shorter period of time than those incubated in the field in 1996 (Mann-Whitney U-test: Z = −4.69, p < 0.0001).
Relationships of Egg, Hatchling, and Female Sizes
There were no differences in CL, BM, or PL between field and lab-reared hatchlings from the same nests (Mann-Whitney U-tests: CL, Z = 0.303, p = 0.763; PL, Z = 0.546, p = 0.585; BM, Z = 1.023, p = 0.305), nor were there any hatchling size differences between years (CL, Z = 0.320, p = 0.749; PL, Z = 0.713, p = 0.476; BM, Z = −0.230, p = 0.818). Data from both years and from both lab and field-reared nests were therefore combined (Table 4). Mean hatchling CL, PL, and BM (n = 77) per clutch were all negatively correlated with clutch deposition date (CL: Spearman's r = −0.237, p = 0.037; PL: Spearman's r = −0.361, p = 0.001; BM: r = −0.288, p = 0.011) although there was no significant correlation between CL and clutch deposition date when adjustment for multiple tests was made. When mean egg size (EL, EW, EM) per clutch was accounted for there was no significant correlation between hatchling size and date of clutch deposition (PL: partial r = 0.015, p = 0.898; BM: partial r = 0.061, p = 0.608).
There were significant correlations between female body size and mean EW per clutch for the 23 females discovered while nesting, but none for the other measures of egg size (Table 5). Twelve of these clutches produced hatchlings. Mean hatchling CL per clutch was weakly correlated with female CL and PL (Table 5). After adjusting for multiple tests, however, there were no significant correlations between female and hatchling sizes.
Nest Densities
I found 403 G. oculifera nests (intact and depredated) in 1995 and 476 in 1996 (Table 6). The surface area of the 11 monitored sandbars was composed of an average of 39% open sand, 38% herbaceous vegetation, and 23% woody vegetation, primarily black willow (Salix nigra). The number of nests per sandbar did not differ significantly between years (Wilcoxon matched pairs test: Z = 1.56, p = 0.120), even though sandbar 6, which had over 11% of all nests in 1995, had fewer than 4% in 1996 (Table 6). The density of G. oculifera nests per sandbar ranged from approximately 1 nest per 4 m2 to < 1 nest per 200 m2 (Table 6). Most sandbars with an area of less than 1000 m2 had relatively high densities of turtle nests, i.e., 1 nest per 5–20 m2 of sandbar. Larger sandbars had much lower densities, ranging from 1 nest per approximately 70–160 m2. The number of nests per sandbar was correlated with sandbar area (Spearman's r = 0.78, p = 0.005) but not with the amount of open sand nor with the amount of either herbaceous or woody vegetation present (all p > 0.23).
Sandbars 12 and 21 had fewer and sandbar 5 had more nests (Table 6) than predicted by the relationship between area and number of nests per sandbar (nests = 0.0091[area] + 20.19; r2 = 0.64, F(1, 9) = 16.23, p < 0.003). The relatively few nests on sandbar 12 may have resulted from its frequent use by recreational boaters. During the 1996 nesting season, boaters were observed at 4 of the 11 sandbars on 33 weekdays and were present for almost the entire day on approximately half of those dates. Twenty-seven of these observations (82%) were from sandbar 12. If weekends are included, during which most large sandbars in the study area were occupied by camping boaters, humans were present on sandbar 12 at least part of the day for 46 of the 67 days (69%) that G. oculifera were nesting in 1996. Sandbar 21 was infrequently visited by humans, but had an extensive growth of Salix nigra approximately 5 m from the edge of the river which formed a ring ca. 3 m wide and which bordered ca. 80% of the sandbar adjacent to the river. This may have made the sandbar appear unsuitable as a nesting site to female G. oculifera observing it from the river. Sandbar 5, although relatively large, had relatively little open sand (Table 6) and did not appear to be optimal nesting habitat. This sandbar had once been part of a lateral and point bar system adjacent to the main channel, but had become partially isolated as a result of the river changing its course. This resulted in less scouring during floods and an increase of vegetative growth. Although sandbar 5 no longer had large expanses of open sand, it was still one of the largest sandy areas in the lower part of the study area, and may have been frequently used by G. oculifera females because of the absence of better nesting areas in the immediate vicinity.
Timing of Nesting
Female G. oculifera were observed nesting between 0815 and 1700 hrs. Of 27 G. oculifera observed excavating a nest chamber or laying eggs, 8 (29.6%) were found between 1000 and 1100 hours, and 19 (70.4%) were nesting before 1200 hours. Surface soil temperatures for nesting females ranged from 23° to 49°C. No evidence of nocturnal nesting was observed in either of the 2 field seasons.
Nest Sites
There were no differences in G. oculifera nest site characteristics (Table 7) between years or months (Mann-Whitney U-tests, all p > 0.06). Most nests were constructed in fine sand approximately 18 m from the water at locations with a canopy cover averaging approximately 37% and within 1 m of vegetation (Table 7). There were usually both herbaceous and woody stems within 50 cm of the nest cavity, which averaged approximately 6 cm in diameter and 12 cm in depth (Table 7).
Nest Temperatures
The mean nest temperature for 10 nests with more than 1000 individual temperature observations (> 30 days) was 27.26° ± 0.84°C. Temperature extremes for these 10 nests ranged from 13.7° to 37.0°C (Table 8). Means of each of the 30 daily temperatures were calculated for each nest over the entire period during which the data loggers were active (Table 8). Nests reached their average maximum temperatures at 1535 hours and average minimum temperature at 0633 hours. The mean difference between highest and lowest mean daily temperatures was 6.36° ± 1.24°C and ranged from 5.05° to 9.25°C. There were significant correlations between low mean nest temperature and number of woody stems around the nest site (Spearman's r = 0.77, p = 0.007), between mean high nest temperature and distance of the nest site from water (r = 0.70, p = 0.025), and between canopy cover and both time of the highest mean temperature (r = −0.66, p = 0.039) and time of lowest mean temperature (r = −0.69, p = 0.026). However, the latter 3 were not significant after adjustments were made for multiple comparisons.
Nest Predation
Approximately 86% of the caged G. oculifera nests were attacked by vertebrate predators (Table 9), but none of the attacks were successful. Armadillos (Dasypus novemcinctus) and raccoons (Procyon lotor) were the most frequent predators identified at both caged and raided nests. Fish crows (Corvus ossifragus) were significant predators at raided nests (Table 9). Predator frequencies did not differ between years for caged (Kruskal–Wallis test: H = 1.11, p = 0.293) or raided nests (H = 0.10, p = 0.753). Frequencies of known predators were marginally nonsignificant between the 2 groups of nests (both years combined: H = 3.94, p = 0.05) probably because there were fewer predation attempts by fish crows on caged nests. This likely occurred because caging nests disturbed a relatively large area of sand and may have made nest sites unrecognizable to fish crows. Predation attempts on caged nests occurred as late as 69 days after the eggs were deposited, but over 42% of all predation attempts were made within the first 24 hours, and 81% within the first 14 days following oviposition. Nests upon which predation attempts had been made (n = 100) had significantly greater canopy cover (36.6% vs. 16.3%) than those on which no predation attempts had been made (n = 15) (Mann-Whitney U-test: U = 365.0, Z = −3.2, p = 0.001) but did not differ for the other nest sites characteristics measured (all p > 0.10). This may have been a result of an edge effect on nest predation rates, which has been observed for other turtle species (Temple 1987).
Egg Development
Sixty-one clutches containing a total of 206 eggs were caged to exclude vertebrate predators in 1995. When these eggs were checked after 55 days, only empty egg shells remained for 11 eggs from 7 clutches, so their developmental status could not be ascertained. Thus, 195 eggs, all incubated in the field, were examined in 1995. In 1996, 217 eggs from 65 clutches were examined. Approximately 37% of these were incubated in the laboratory and the remainder incubated in the nest cavity (Table 10).
The frequencies of fertile and undeveloped eggs differed significantly between years (χ2 = 153.14, df = 1, p < 0.0001), with almost 60% of the 1995 eggs showing no evidence of development whereas almost 97% of 1996 eggs were fertile, as indicated by the presence of embryos or live hatchlings (Table 10). In 1995, the entire complement of 22 clutches showed no evidence of development, whereas only 1 clutch in 1996 did not have any eggs that developed. At least 2 of the failed 1995 clutches were inundated by high water in early June, and a third appeared to have been desiccated by plant rootlets. All but 1 of the 19 remaining failed clutches were deposited in June, and 14 of these were deposited between 5 and 16 June during a cool rainy period. Approximately 22% of eggs produced viable (emerged) hatchlings in 1995, whereas over 80% produced viable hatchlings in 1996. The differences in development and emergence of eggs between 1995 and 1996 did not appear to result from incubating some eggs in the laboratory in 1996. There were no differences in 1996 between lab and field-reared eggs in the proportions of fertile vs. undeveloped eggs (χ2 = 0.24, df = 1, p = 0.626), between pipped vs. fertile but unpipped eggs (χ2 = 0.01, df = 1, p = 0.936), nor between the number of hatchlings that emerged vs. developed hatchlings that died before emergence (χ2 = 1.71, df = 1, p = 0.191).
Egg Predation
Approximately half of the fertile eggs that developed but that did not show signs of pipping were destroyed by Solenopsis molesta, a native species of fire ant that appeared to attack eggs just prior to and immediately following pipping (Table 11). Larvae of the dipteran Tripanurga (= Metaposarcophaga) importuna, a sarcophagid fly, were also important predators, particularly in 1995. The larvae were present in many of the clutches that did not show any evidence of development, and were frequently found infesting eggs that appeared to have been unpipped. These, however, may have first been attacked by S. molesta, which apparently gained entry into eggs by chewing holes through the shell, a behavior which has also been noted in Solenopsis invicta, the red imported fire ant (Moulis 1997). Approximately 30% of eggs that did not show signs of pipping contained dead embryos. Approximately equal numbers of turtles that pipped but did not emerge were either killed by Tripanurga or S. molesta or died before emerging from the nest (Table 11). Of the 211 field-incubated eggs known to be fertile in this study, 50 (23.7%) were destroyed by invertebrate predators prior to either pipping or emergence. Several clutches when first checked for evidence of pipping contained only empty shells and were likely destroyed by invertebrates.
DISCUSSION
CS and Frequency
Graptemys oculifera is a relatively small species with an apparently low annual reproductive potential. Female CL at maturity (130 mm) is less than that of many other species in the genus (e.g., Graptemys ouachitensis, 150–160 mm; Graptemys geographica, 190 mm; Graptemys ernsti, 212 mm; and Graptemys flavimaculata, 159 mm; Ernst et al. 1994; Horne et al. 2003), and the mean CL of gravid females found in this study (165 mm) was less than that of G. flavimaculata (190 mm; Horne et al. 2003), its sister species (Lamb et al. 1994; Stephens and Wiens 2003). Its CS (3.7) is the smallest yet reported for any Graptemys, which range from 4.8 for G. flavimaculata (Horne et al. 2003) and 5.5 for Graptemys nigrinoda (Lahanas 1982) to 14.1 for Graptemys pseudogeographica (Vogt 1980).
Horne et al. (2003) using radiography and Shelby et al. (2000) using ultrasound on the same population of G. flavimaculata found what they termed an abnormally low reproductive frequency of 1.16 and 1.17, respectively. The reproductive frequency of G. oculifera estimated using radiography was 1.10 and may have ranged from 0.96–1.42 (see Results). The number of clutches produced by other Graptemys range from 2 to 3 (G. ouachitensis, G. pseudogeographica; Vogt 1980), to 3 (G. nigrinoda; Lahanas 1982), to as many as 6 (G. ernsti; Shealy 1976). These studies derived their estimates from dissections and therefore are probably maximum estimates of average clutch frequency for those species, whereas my estimate for G. oculifera, which is derived from radiography and recapture data, is likely a minimum estimate. The problems inherent in using radiography and multiple recaptures of adult females to determine clutch frequency have been discussed elsewhere (Tucker 2001; Horne et al. 2003) and likely apply equally well here. On the other hand, both Iverson and Smith (1993) and Tucker (2001) have used peaks in the distribution of gravid females captured throughout the nesting season to estimate clutch frequency. If this is equally applicable to G. oculifera, then Fig. 4 supports the hypothesis that the majority of females nest only once during the year. Shelby et al. (2000) suggested that the low reproductive frequency in G. flavimaculata might have resulted in part from a disruption of the hormonal system because of chemical pollution. It is not known whether a similar problem affects G. oculifera at Ratliff Ferry or if some unknown selective pressure favors an extremely low reproductive potential in this species. A study investigating the endocrine system of G. oculifera similar to that of Shelby et al. (2000) would be useful in investigating this question.
Some female G. oculifera apparently skip reproduction in certain years, a phenomenon that has also been observed in a number of other turtles (see Congdon et al. 1987). The probability of skipping a year of reproduction was apparently size-related in Chrysemys picta, where large individuals reproduced every year but smaller turtles skipped years (Christens and Bider 1986). There were not enough recapture data to determine whether this also happened in G. oculifera, but the body size data in Fig. 2 indicates that larger females were much more likely to have been gravid when captured than were smaller individuals. This suggests that the majority of larger females may be gravid on an annual basis whereas smaller turtles are gravid on a more irregular schedule.
Variation in Body Size and Reproduction
Gravid G. oculifera found early in the nesting season tended to be larger than those gravid later in the year, which has also been noted in G. ernsti (Shealy 1976) and in some populations of Chelydra serpentina (Hammer 1969; Petokas and Alexander 1980). However, this was not observed in other snapping turtle populations (Congdon et al. 1987; Iverson, et al. 1997), nor in Emydoidea blandingii (Congdon et al. 1983). Why this occurred is unclear, but may have involved either the relative quantity of resources available for reproduction or the relative efficiency with which clutch development was completed. If the former is true, then larger, early-nesting females may have had sufficient fat reserves at the beginning of the summer activity period to complete development of their clutches, whereas those nesting later in the season may have been less efficient in acquiring and storing resources the previous fall and therefore had to forage in spring in order to finish clutch development. This implies that larger females were more efficient either at foraging or at storing lipids than were smaller individuals. However, if most of the energy allocated to an initial clutch in the current year came from resources harvested the previous year, as has been suggested for C. picta (Congdon and Tinkle 1982), then both large and small females should already have had sufficient reserves for their initial clutches prior to the beginning of the nesting season, and would be expected to initiate egg-laying at approximately the same time. Although a few small females did nest early in the season (Table 1), most females ovipositing during that period were larger individuals. Alternatively, because basking may be integral to the maturation of eggs, probably by increasing metabolic rates, larger females may have been able to maintain higher body temperatures more efficiently in early spring and matured clutches more quickly than smaller females, thus laying eggs earlier in the season. Additional data on the energetics of egg production in G. oculifera will be necessary to test these hypotheses. In either case, early nesting may be advantageous because offspring would have longer to grow before the onset of cooler temperatures in the fall, assuming that overwinter survival in neonates is positively correlated with body size.
Both clutch and egg size of G. oculifera declined during the nesting season, even when controlling for the effects of body size. A decrease in CS during the nesting season has been observed in other turtles (Apalone mutica, Plummer 1977; Trachemys scripta, Pseudemys floridana, Kinosternon subrubrum, Gibbons et al. 1982; Chelydra serpentina, Iverson et al. 1997), both with and without controlling for body-size effects, and a seasonal decline in egg size was observed in both Malaclemys terrapin (Montevecchi and Burger 1975) and Chelydra serpentina (Iverson et al. 1997). Both egg mass and CS decreased seasonally across multiple clutches in individual Chrysemys picta (Iverson and Smith 1993), so the seasonal decline in these parameters observed in the Ratliff Ferry population of G. oculifera may have occurred because eggs and clutches found later in the nesting season were from second or even third clutches. Thus the observed seasonal decline in clutch and egg size in G. oculifera may mean that clutch frequency was actually much higher than indicated by the recapture data. Alternatively, if these later eggs and clutches were primarily from initial (and presumably the only) clutches in a season, then the resources available for reproduction in a given year apparently vary greatly among females, allowing some to produce large eggs and clutches early in the nesting season, but limiting others to smaller eggs and clutches later in the season. If the latter is true, then the annual reproductive output in this species may be controlled largely by extrinsic environmental factors, as has been suggested for other turtles (Gibbons 1982; Gibbons et al. 1982). Additional data on clutch frequency and reproductive energetics will be necessary to evaluate these hypotheses.
Egg and hatchling size were correlated as were EW and female body size, so it was expected that female body size and hatchling size would also be correlated. However, in this study this was not the case. The lack of correlation between female size and hatchling size may have resulted from either variation in incubation conditions among clutches and its subsequent effects on hatchling sizes (see Packard et al. 1985; Packard et al. 1987), or from small sample size (n = 12), which included only a limited range of female and egg sizes. Either of these 2 factors could easily have overshadowed any variation in hatchling sizes because of differences in adult female sizes. With a larger range in sizes and greater numbers of both females and eggs, and with the latter incubated under identical conditions, hatchling and female sizes may well prove to be correlated.
Nesting and Nest Predation
All nests of G. oculifera found during the study were on sandbars, in contrast to those of G. flavimaculata, which were found on open areas along the riverbank as well as sandbars (Horne et al. 2003). These riverbank areas were used because of the frequent occupation of sandbars by humans, effectively preventing their use by nesting females. Female G. oculifera did not use riverbanks as nesting sites because, even though the Ratliff Ferry area is heavily used by recreational boaters and human occupation of sandbars occurred with some frequency, the large number of sandbars in the study area seemed to provide ample undisturbed nesting habitat. However, in areas of the Pearl River where few sandbars are present, as occurs in the lower Pascagoula River where Horne et al. (2003) studied G. flavimaculata, or on individual sandbars in high use areas, such as sandbar 12, human disturbance could have a significant impact by forcing turtles to nest in suboptimal habitats.
Production of hatchlings from monitored nests varied significantly during the study, ranging from ca. 22% in 1995 to ca. 78% in 1996. This resulted in part from the large percentage of undeveloped eggs in 1995, when almost 60% failed to develop. Eggs which show no evidence of development are usually considered to be infertile (e.g., Burger 1977; Congdon et al. 1983; Christens and Bider 1986), but in most of these studies fewer than ca. 10% of eggs were in this condition. It seems unlikely that infertility was responsible for failure of the large number of eggs in 1995 because this would require annual variation in G. oculifera fertility rates ranging from 40% to almost 100%. Eggs were handled in the same manner in both years, so it is improbable that the variation was a result of methodological differences. It seems more likely that this resulted from the effects of temperature or moisture on development. If this were the case, the eggs were affected soon after oviposition because the contents were still bright yellow and none showed evidence of somatogenesis. A large proportion of the failed clutches were deposited in June, when mean temperatures were ca. 1.5°C cooler than the long-term average (National Weather Service, Jackson, unpubl. data). Whether this minor variation in temperature could have resulted in failure of such a large number of G. oculifera eggs will have to be determined in laboratory experiments.
The relatively large number of G. oculifera nests destroyed by predators is similar to that found in several other studies of North American turtles (e.g., Moll and Legler 1971; Shealy 1976; Burger 1977; Vogt 1980; Congdon et al. 1993) with ca. 86% of the nests attacked by vertebrates and ca. 24% of the remaining eggs destroyed by invertebrates. Armadillos have not been identified as significant predators of North American turtle nests (Wilbur and Morin 1987), but were one of the principal predators of Trachemys scripta nests in Panama (Moll and Legler 1971). The armadillo is a relatively recent addition to the fauna of Mississippi, first reported east of the Mississippi River in 1943 (Lowery 1943) and occupying two-thirds of the state by 1974 (Humphrey 1974). Mississippi also has a large population of raccoons (J. Watkins, Mississippi Department of Wildlife, Fisheries, and Parks, pers. comm.), and the fish crow also appears to be expanding its range (Madge and Burn 1994). Vertebrate predation experienced by G. oculifera today may therefore be significantly higher than it was historically.
Both of the primary invertebrate species implicated in the destruction of G. oculifera eggs and hatchlings have been previously identified as predators of turtle nests. Tripanurga importuna preyed upon recently pipped hatchlings of G. pseudogeographica (Vogt 1981), destroying 36% of the turtles from 23 clutches. Ants have also been implicated in the destruction of turtle hatchlings in several studies (e.g., Burger 1977; Landers et al. 1980; Congdon et al. 1983; Moulis 1997). Solenopsis molesta, the primary ant predator found in this study, also preyed upon Trachemys scripta eggs in Panama (Moll and Legler 1971). Solenopsis invicta, the red imported fire ant, was expected to be a potential predator of hatchling G. oculifera, because it has been implicated as a predator on a variety of turtle species (reviewed in Moulis 1997), and it was present in the vicinity of all sandbars. However, it was responsible for the destruction of only 1 clutch and was not a significant cause of mortality in the study area. Mount (1981) noted that S. invicta had difficulty establishing colonies in excessively sandy soils, which may explain why few G. oculifera nests, which are constructed primarily in almost pure sand, were depredated by this species.
Although the Ratliff Ferry population of G. oculifera appears to be stable at present, its future is not entirely secure. Turtles are long-lived species, and the demographic traits that have evolved with longevity often result in populations that have a limited ability to adapt to increases in mortality, including a chronic increase in the mortality of neonates (Congdon et al. 1993). Predation rates found in this study, although similar to that reported in other studies, are a reason for concern, particularly because one of the major egg predators (D. novemcinctus) is a recently arrived component of the fauna, the second (P. lotor) has increased substantially over the state in the last few years, and the third (C. ossifragus) appears to be expanding both its range and numbers. If G. oculifera cannot adapt to increased nest predation and there is no concomitant increase in juvenile or adult survivorship, the long-term prospects for the survival of this species do not appear to be high. The low reproductive frequency of this species, if it is real, is a serious concern, particularly if it has resulted from an unknown chemical stressor in the watershed. Finally, although there did not appear to be a serious conflict between turtle nesting and human usage of sandbars in the Ratliff Ferry area, this has the potential to become a serious problem in the future, especially if the recent trend toward the establishment of summer-long tent camps continues. These multitent encampments, which are usually set up in April and remain until early September, not only disrupt female nesting behavior but also cover much of the available nesting habitat, especially on small sandbars.
Map of the study area on the Pearl River, Mississippi. The numbered areas are the 11 sandbars monitored for Graptemys oculifera nesting activity. The inset shows the location of the study area in west-central Mississippi.
Proportion of gravid and nongravid adult (≥ 130 mm CL) female Graptemys oculifera in 10-mm CL size classes. Numbers above bars are sample sizes.
Mean CL of gravid Graptemys oculifera by capture week for both years of the study combined. Numbers are sample sizes for each weekly interval.
Proportion of gravid Graptemys oculifera in weekly trapping samples for each of the 2 years of the study. Numbers represent adult females captured during each week.
