| |
MULTIVARIATE COMPARISONS
Canonical discriminant analysis (SAS Institute, 1988) was used to determine the distinctiveness of C. canni and its two closest relatives, C. novaeguineae and C. reimanni (Rhodin, 1994b; Georges et al., 2002). All raw data were expressed as ratios to remove ontogenetic variation from the dataset. Head measurements were expressed as a ratio of head length (HL); head length and all shell measurements were expressed as a ratio of carapace length (CL). A number of composite variables were defined as combinations of three or more raw measurements, but none were retained by the subsequent analyses. Descriptions of all measurements utilised in this study are in Appendix 2.
An initial analysis was performed between all members of the C. longicollis group, however only the analysis between the three closest species is presented in graphical form. The generalised multivariate distances between all taxa in the C. longicollis group, obtained from a discriminant analysis with all raw measurement ratios included, are presented in Table 3 as an indication of their general morphological similarity. Note that the distinction between C. canni and C. novaeguineae is substantial (34.12 units), and greater than the distance between Chelodina reimanni and C. novaeguineae (19.95 units), in support of our recognition of C. canni and C. novaeguineae as separate species.
Stepwise selection (significance level for entry = 0.05; for removal = 0.10) was used to obtain a subset of the original variables that provided the best discrimination. This yielded the subset of head measurements HL, PW, HWT, HWJ, HT, VT and the shell measurements CW4, CW8, V1, V2, PLF and PLR. Clearly, both head shape and shell shape are well represented in the final formula that provided the best discrimination (Fig. 4). A total of 63.2% of the among-groups variation was explained by the first canonical variate. This variate provides the bulk of the discrimination between C. reimanni and C. canni, also contributing to the discrimination between these two and C. novaeguineae. The second canonical variant provided further discrimination between C. novaeguineae and the other two forms (Fig. 4). An indication of the strength of discrimination is given by cross-validation (SAS Institute, inc., Cary, NC, 1988), although it does rely on assumptions of normality, unlikely to be strictly upheld because not all animals were the same overall size and because growth is allometric. Nevertheless, only four of the 85 animals in the analysis were misclassified. One C. novaeguineae was misclassified as C. reimanni, and vice versa, one C. canni was misclassified as C. reimanni and one as C. novaeguineae.
|
|
The most influential variable in the discrimination was PWF (Partial R2 = 0.70; F=95.1, p<0.0001) followed by PW/HWT (Partial R2 = 0.45, F=32.2, p<0.0001) and IO/OD (Partial R2 = 0.35, F=21.0, p<0.0001), so differences in plastron shape were the most influential in providing overall discrimination among the three taxa. This is consistent with the fact that C. canni and C. novaeguineae are readily distinguished by differences in plastron shape and head shape. There was no clear partition of the raw variables in terms of their association with one or the other canonical variates so we could not carry our interpretation further.
NATURAL HISTORY NOTES
Kennett et al. (1992) described the eggs, hatchlings, habitat, and basic natural history of C. canni. A topotypic female in the senior author's collection produced one egg in September of 1993, which subsequently hatched after 88 days incubation time. In addition, one of the non-topotypic females produced ten eggs after oxytocin injection in August of 1994. The eggs were white and oval, with hard shells. Four of the ten eggs showed no signs of life, and after one month were discarded. Six of them hatched after 95-98 days at ca. 28-29C in sterile, moist vermiculite. The six live eggs averaged 33.2 x 21.9 mm (ranges, 32.8-33.8, and 21.6-22.2 mm), and the hatchlings averaged 33.6 mm carapace length (range, 31.0-35.2 mm). Two died soon after hatching for unknown reasons. The remaining hatchlings (now subadults) from both sites are identical in every character. Our eggs and hatchlings are only slightly larger than those described by Kennett et al. (1992)
The range of C. canni overlaps that of C. rugosa, Elseya dentata, Elseya latisternum, Emydura subglobosa and Emydura tanybaraga, but it has been collected syntopically with only C. rugosa (Covacevich et al., 1990). The Gangalidda people at Old Doomadgee call C. canni "bungarra mali", the stinking turtle (Covacevich et al., 1990). The local Aboriginal people at Jinduckin rear Mataranka call C. canni "nganymalin", meaning smelly armpit turtle (Kennett et al., 1992), owing to the distinctive pungent odour produced by the turtle when handled. This defence mechanism is common for this group of turtles as demonstrated again by the Gogodala people of Balimo Papua New Guinea, calling a local form of C. novaeguineae "ipudinapi" meaning "little smelly turtle" (Robert Danaya, pers. com).
|
|
