Ecomorphological correlates of microhabitat selection in two sympatric Asian box turtle species (Geoemydidae: Cuora)

Ecomorphological correlates of microhabitat selection in two sympatric Asian box turtle species (Geoemydidae: Cuora) Journal: Canadian Journal of Zoology Manuscript ID cjz-2016-0218.r2 Manuscript Type: Article Date Submitted by the Author: 23-May-2017 Complete List of Authors: Xiao, Fanrong; Hainan Normal University, College of Life Sciences Wang, Jichao; Hainan Normal University, College of Life Sciences Shi, Haitao; Hainan Normal University, College of Life Sciences Long, Zaizhong; Hainan Normal University, College of Life Sciences Lin, Liu; Hainan Normal University, College of Life Sciences Wang, Wei; Hainan Normal University, College of Life Sciences Keyword: keeled box turtle, Cuora mouhotii, Indochinese box turtle, Cuora galbinifrons, leaf litter, microhabitat, rock crevice

Page 1 of 24 Canadian Journal of Zoology Ecomorphological correlates of microhabitat selection in two sympatric Asian box turtle species (Geoemydidae: Cuora) Fanrong Xiao, Jichao Wang, Haitao Shi, Zaizhong Long, Liu Lin, and Wei Wang F. Xiao, J. Wang, H. Shi, Z. Long, L. Lin, W. Wang: College of Life Sciences, Hainan Normal University, Haikou 571158, China H. Shi: Chengdu Institute of Biology, Chinese Academy of Sciences, Chengdu 610041, China (e-mail: haitao-shi@ 263. net) Corresponding authors: H. Shi, Hainan Normal University, Longkun south road no. 99, Haikou, Hainan province, China. Tel: 0898-65803298 (e-mail: haitao-shi@ 263. net)

Page 2 of 24 Ecomorphological correlates of microhabitat selection in two sympatric Asian box turtle species (Geoemydidae: Cuora) F. Xiao, J. Wang, H. Shi, Z. Long, L. Lin, and W. Wang Abstract: Closely related species that co-occur in homogeneous environments often possess differing morphologies, which can result in niche divergence that minimizes interspecific competition. In the present study, we examined the relationship between the ecomorphological characteristics and microhabitat selection of two Asian box turtle species, the keeled box turtle Cuora mouhotii (Gray, 1862) and Indochinese box turtle C. galbinifrons (Bourret, 1939), which have sympatric distributions in the rainforest of Hainan, China. We found that C. mouhotii had a relatively flat shell and preferred microhabitats with rock crevices and steep slopes in the field, whereas C. galbinifrons had a domed shell and was restricted to microhabitats of deciduous leaves under bamboo growing on gentle slopes. We conclude that morphological divergence allows the two Cuora spp. to use different microhabitats and, thereby, to successfully co-occur. Key words: keeled box turtle, Cuora mouhotii, Indochinese box turtle, Cuora galbinifrons, leaf litter, microhabitat, rock crevice.

Page 3 of 24 Canadian Journal of Zoology Introduction One of the most important objectives of ecomorphology is to determine the underlying mechanisms of species co-occurrence in relation to their morphological and ecological characteristics, both in the present and over evolutionary time (Motta and Kotrschal 1992; Vanhooydonck et al. 2000). Generally, co-occurring species either possess similar morphologies, owing to convergent evolution from adaptation to the same habitat (Grant 1972), or divergent morphologies that result in niche divergence and minimize interspecific competition (Hutchinson 1959; Brown and Bowers 1985). In particular, closely related species that co-occur in homogeneous environments should be morphologically distinct to partition limited resources, such as microhabitats, and, thus, to reduce competitive interactions and, ultimately, achieve co-occurrence. Numerous studies on vertebrate taxa have shown that ecomorphology correlates with habitat selection among sympatric species (e.g., mammals: Davies et al. 2007; birds: Guillemain 2002; reptiles: Pounds 1988; Lindeman 2000; Vanhooydonck et al. 2000; Higham and Russell 2010; and fish: Robinson and Wilson 1994). However, such studies have typically correlated morphological features with habitat classes, and only a few have quantified ecological characteristics, such as microhabitat (Lindeman 2000; Vanhooydonck et al. 2000). Both the keeled box turtle (C. mouhotii Gray, 1862; formerly Pyxidea mouhotii) and Indochinese box turtle (C. galbinifrons Bourret, 1939) are distributed in south China, including Hainan Province (Zhao and Adler 1993). C. mouhotii is typically found in moist evergreen forests at low altitude (Wang et al. 2011a; Wangyal et al.

Page 4 of 24 2012; Struijk et al. 2016), whereas C. galbinifrons is found in similar forests but at high altitude (Wang et al. 2011b). Previous studies have reported the occurrence of natural hybrids in Hainan Province (Shi et al. 2005), which implies that, in some areas, the distributions of the two species overlap, and the species have since been reported to co-occur at 700 914 m in the Hainan Diaoluoshan Natural Reserve by Wang et al. (2011a,b). Interestingly, the morphology of C. mouhotii is extremely divergent from its sympatric species, C. galbinifrons, as well as from all other species in the genus. For example, C. mouhotii possesses a relatively flat carapace for a member of Cuora, with a flattened top, whereas C. galbinifrons has a highly domed carapace. In general, carapace shape is correlated with the habitat preference of turtles: aquatic turtles have flattened carapaces, whereas terrestrial turtles have highly domed carapaces, and semiaquatic turtles have carapaces with intermediate morphology (Romer 1967; Claude et al. 2003; Bonnet et al. 2010; Benson et al. 2011). However, previous studies about the microhabitat preference of flat-topped turtle species remain scarce, such as C. mouhotii. Additionally, although C. galbinifrons has been considered to exhibit a semiaquatic habit (Ernst and Barbour 1989), Wang (2007) and Wang (2008) tracked nine individuals in Diaoluoshan Natural Reserve for two years, using radio telemetry, and found that the species never entered the water. Therefore, the ecological function of the two species divergent morphologies has been unclear. Recently, one individual of C. mouhotii was found in a rock crevice at the Deo Ca Protected Forest, Phu Yen Province, Vietnam (Ly et al. 2013). In contrast, C.

Page 5 of 24 Canadian Journal of Zoology galbinifrons is more frequently observed in microhabitats of deciduous leaves under dense bamboo (Wang et al. 2011b). Therefore, in the present study, we quantified the microhabitat use of co-occurring C. mouhotii and C. galbinifrons at Diaoluoshan Natural Reserve. We hypothesized that the two species would occupy different microhabitats, owing to their divergent shell morphology. Materials and methods Ethics statement This study was approved by the Animal Research Ethics Committee of Hainan Provincial Education Centre for Ecology and Environment, Hainan Normal University (HNECEE-2014-002), and was carried out in strict accordance with the institutional guidelines. Fieldwork was carried out with permission from the Diaoluoshan Forest Bureau. No turtles were sacrificed for this study or incurred injury or death while in the traps. Microhabitat selection We studied the microhabitat selection of C. mouhotii and C. galbinifrons at the Diaoluoshan Natural Reserve in Hainan Province, China (18 43 53 N, 109 52 10 E) from April 2015 to February 2016. The field site was located in a rainforest between 860 and 914 m in altitude, where the average highest temperature is 28 C in July, the average lowest temperature is 15.4 C in January, and the annual rainfall ranges from 1870 to 2760 mm.

Page 6 of 24 The turtles were captured in traps baited with salty fish or rancid pork skin; thereafter, a radio-tracking transmitter (RI-2B, 216.000 216.999 MHz; Holohil Systems, Ltd., Caro Ontario, Canada) was fitted to the last costal scute of each turtle. The size (diameter height = 21 8 mm), mass (6 g), and location of the transmitters would not have prevented turtles from using any of the microhabitats. Then the turtles were released at the site of capture. The turtles were tracked daily from 09:00 to 16:00, using a handheld receiver (TRX-1000S, 216.000 216.999 MHz; Wildlife Materials International, Inc., Murphysboro, Illinois, USA) with a three-element folding antenna (Wildlife Materials, Inc.), and the locations were recorded using a handheld Magellan Triton 400 Global Positioning System (Magellan Navigation, Inc., Santa Clara, USA). Radio-tracked turtles were tracked for an average of 164 days. All radio-tracking transmitters were carefully removed from the turtles shells when the tracking study was complete. We hypothesized that the locations at which the turtles remained for >3 d indicated preferred microhabitats and quantified 14 microhabitat characteristics within two different-sized quadrats (10 10 m and 1 1 m) at each preferred site. In the large quadrats, we quantified (1) slope gradient ( ), (2) slope position (1 = upper slope, 2 = middle slope, 3 = lower slope; slope position refers to the location of slope occupied by turtle in the mountain, which was classified as upper slope, middle slope or lower slope when the turtle occur near peak, middle or low position of the mountain, respectively), (3) canopy cover (%), (4) number of fallen logs, (5) number of bamboo clumps, (6) number of tree holes, and (7) number of rock crevices; and in

Page 7 of 24 Canadian Journal of Zoology the small quadrats, we quantified (8) bamboo density (plants/m 2 ), (9) herbage height (cm), (10) herbage density (plants/m 2 ), (11) deciduous leaf cover (%), (12) deciduous leaf thickness (cm), (13) stone cover (%), and (14) type of concealment (1 = rock crevice, 2 = tree hole, 3 = fallen log, 4 = grass, 5 = bamboo clump, 6 = deciduous leaves, 7 = bare ground). In total, we collected data from 44 preferred sites from seven C. mouhotii individuals and from 27 preferred sites from five C. galbinifrons individuals. Chi-squared tests (SPSS 16.0; SPSS, Inc., Chicago, IL, USA; same below) were used to assess whether the two species differed in their preference of slope position and type of concealment. The Kolmogorov-Smirnov test was performed to test the normality of the 12 numeric microhabitat variables. One-way analysis of variance (ANOVA) was subsequently used to test whether the two species differed in slope gradient, herbage height, and deciduous leaf cover, and the Mann-Whitney U test was used to determine whether the two species differed in any the nine remaining microhabitat variables. Furthermore, discriminant function analysis (DFA) was performed to assess the differences in the 12 numeric microhabitat variables between the two species. The stepwise method of DFA was performed to determine the best variables significantly separating the two species. In addition, we measured the slope gradient at stopover sites at which the turtles remained for <1 d, using a clinometer. Whether the two species differed in slope gradient was examined using one-way ANOVA.

Page 8 of 24 Morphological measurements Morphological measurements were taken from all the experimental individuals (n = 9 for C. mouhotii, n = 8 for C. galbinifrons), as well as from other individuals that were more recently encountered in the field (nine individuals for C. mouhotii and five individuals for C. galbinifrons), using a Vernier caliper (accuracy to 0.02 mm). For each individual, we measured carapace length, carapace width (measured at the sixth marginal scute), and carapace height (measured at the sixth marginal scute). In addition, we also calculated the ratio of carapace height to width (R) as a measure of shell contour (Domokos and Várkonyi 2008). After the normality and homogeneity of variance were tested for all morphological measurements, one-way ANOVA was used to test whether the two species differed in carapace length, and to remove the effect of body size, one-way analysis of covariance (ANCOVA; carapace length as the covariate) was performed to test whether the two species differed in any of the other morphological variables. Results Microhabitat selection No individuals of either C. mouhotii or C. galbinifrons were found in the water during the field study. Among the 14 microhabitat variables, slope gradient, number of rock crevices, deciduous leaf cover, deciduous leaf thickness, stone cover (Table 1), and type of concealment (χ 2 = 69.69, df = 6, P < 0.0001) significantly differed

Page 9 of 24 Canadian Journal of Zoology between the two species. This finding indicates that C. mouhotii was more commonly observed on rocky substrates and steep slopes (Table 1) and typically selected rock crevices for concealment (56.82%; Fig. 1B), whereas C. galbinifrons was usually observed in deciduous leaves on gentle slopes (Table 1) and selected deciduous leaves for concealment (93.93%; Fig. 1C). The stepwise DFA indicated that three variables were selected which significantly discriminate C. mouhotii and C. galbinifrons (stone cover, slope gradient, and bamboo density; Table 2), and the efficiency of the analysis was high (62 of 71 quadrats - 87.3% were correctly classified). The distribution of the quadrats on the DFA axis showed that the microhabitats selected by C. mouhotii were mostly distinct from those selected by C. galbinifrons, and only a small overlap between the two species (Fig. 2). For instance, both C. mouhotii and C. galbinifrons selected the same terrestrial habitat with high-canopy cover (Table 1; Fig. 1A). In addition, the slope gradients of the stopover sites of the two species differed significantly (one-way ANOVA; F 1, 324 = 128.214; P < 0.0001), with C. mouhotii (mean ± SD: 30.3 ± 8.7, range: 5 60 ) occurring on steeper slopes than C. galbinifrons (22.3 ± 7.5, 5 40). This result was in accordance with the slope gradients from the preferred sites. Morphological measurements The carapace length of the two species did not differ significantly (one-way ANOVA; Table 3). However, one-way ANCOVA, with carapace length as the

Page 10 of 24 covariate, indicated that the carapace height and ratio of carapace height to width of the two species all differed significantly, whereas the carapace width of the two species was similar (Table 3). These findings indicated that C. mouhotii had a relatively flat shell, whereas C. galbinifrons had a highly domed one. In addition, the plastron hinge of C. mouhotii only allowed the shell to close partially, whereas that of C. galbinifrons allowed the shell to close completely; and C. mouhotii also possessed a carapace with serrated posterior edges and a flattened top, whereas C. galbinifrons possessed a carapace with smooth edges. Discussion In agreement with previous studies, we found that both adult C. mouhotii and C. galbinifrons were terrestrial species and selected the same high-canopy cover habitat (Wang 2007; Lian 2009); however, we also observed that the microhabitats of the two species were significantly different. Notably, C. mouhotii individuals were mostly observed in areas with greater stone cover and often selected rock crevices for concealment, where they hid for up to 20 d, whereas C. galbinifrons individuals were mostly observed in deciduous leaf litter under bamboos, where they often hid for long periods, as well. In addition, C. mouhotii also occupied microhabitats with slopes that were much steeper than those of the microhabitats occupied by C. galbinifrons. This suggests that the two species co-occur by occupying different niches. Understanding the relationship between microhabitat selection and morphology is critical for understanding ecomorphological adaption (Vanhooydonck et al. 2000;

Page 11 of 24 Canadian Journal of Zoology Rivera 2008; Rivera et al. 2014). In the present study, we found that body size (carapace length) did not differ between C. mouhotii and C. galbinifrons, but the shell of C. mouhotii was significantly more flat than that of C. galbinifrons. Our microhabitat experiment confirmed that C. mouhotii prefers and selects microhabitats with rock crevices, and since flattened bodies often allow animals to fit into narrow crevices (Miles 1994) and because previous studies have reported that terrestrial turtles with flat carapaces, such as the African pancake tortoise (Malacochersus tornieri Siebenrock, 1903), are adapted to lifestyles in rocky crevices (Ireland and Gans 1972; Malonza 2003), we believe that the relatively flat shell of C. mouhotii is an adaptation for using microhabitats with rock crevices. In addition to possessing a flatter shell than C. galbinifrons, the flat-topped carapace of C. mouhotii was important in its adaptation to the rock crevice microhabitats, as in several species of the genus Homopus Duméril and Bibron, 1835 (Ernst and Barbour 1989; Bonin et al. 2006). We also observed that C. mouhotii has a posteriorly serrated carapace, which might help prevent the turtles from being dragged out of rocky crevice by predators. In addition, C. mouhotii occupied microhabitats with slopes that were much steeper than those of the microhabitats occupied by C. galbinifrons. This difference may be correlated with the shell shapes of the two species. Since flat shells impart a lower center of gravity (Domokos and Várkonyi 2008) and thus increase stability on steeper surfaces, the relatively flat shell of C. mouhotii could also be an adaptation for climbing on smooth rocks and inclined substrates; however, this morphofunctional

Page 12 of 24 hypothesis should be confirmed by further investigation. In contrast to the relatively flat shell of C. mouhotii, the highly domed shell of C. galbinifrons is typical of terrestrial turtles, and both the present and previous studies (Wang 2007) have reported that C. galbinifrons inhabits terrestrial habitats. Indeed, the species is well adapted to terrestrial environments, since its shell can tolerate much stronger mechanical forces, such as those from terrestrial predators (Greene 1988; Stayton 2011); and although the highly domed shell may also restrict the species to hiding in soft leaves and to microhabitats with relatively gentle slopes, the species possesses several alternative defensive features. For example, C. galbinifrons also has a single transverse hinge across the middle of its plastron that allows complete retraction and protection of its extremities from predators (Pritchard 2008), and the color and pattern of the species carapace is similar to that of deciduous leaves, which may function to reduce predator detection via camouflage (Fig. 1C; Stevens and Merilaita 2009; Xiao et al. 2016). Moreover, the smooth marginal scutes of C. galbinifrons allow it to burrow deeper than C. mouhotii under deciduous leaves. Notably, we found that C. galbinifrons was a leaf-dwelling species. However, many terrestrial species that hide under leaf litter, such as the black-breasted leaf turtle (Geoemyda spengleri Gmelin, 1789; R = 0.54; Benson et al. 2011), forest cane turtle (Vijayachelys silvatica Henderson, 1912; R = 0.51; Whitaker and Vijaya 2009), and spiny turtle (Heosemys spinosa Gray, 1831; R = 0.53; Spinks et al. 2012), have flat shells. Therefore, it is actually quite surprising that the highly domed C. galbinifrons also uses this microhabitat. One possible way in which C. galbinifrons

Page 13 of 24 Canadian Journal of Zoology can mitigate the disadvantage of its tall shell is that the coloration of C. galbinifrons is somewhat similar to that of other leaf-dwelling species, as well as to that of leaves (i.e., camouflage). It is also possible that the leaf litter in the habitat of C. galbinifrons affords high cover and is thick enough to hide its highly domed shell. Intriguingly, other Cuora spp. possess different shell morphologies and occupy diverse habitats. For example, both C. galbinifrons (R = 0.70) and the yellow-margined box turtle (C. flavomarginata Gray, 1863; R = 0.65; Benson et al. 2011) have highly domed shells and occupy terrestrial habitats, i.e., evergreen forests (Lue and Chen 1999), whereas the Malayan box turtle (C. amboinensis Daudin, 1802), which also has a relatively domed shell (R = 0.61; Benson et al. 2011), is aquatic (Ernst 1989; Joyce and Gauthier 2004). Similarly, C. mouhotii, which has a relatively flat shell (R = 0.58), occupies terrestrial environments with rocky crevices; whereas, the Chinese three-striped box turtle (C. trifasciata Bell, 1825), golden-headed box turtle (C. aurocapitata Luo and Zong, 1988), and Pan s box turtle (C. pani Song, 1984), which are morphologically similar and possess very flat and streamlined shells (R = 0.48, 0.45, and 0.47, respectively; Zhang et al. 1998) that are typical of aquatic species (R under approx. 0.6; Domokos and Várkonyi 2008), are completely aquatic (Zhang et al. 1998; Cheung 2007). Phylogenetic analysis will be needed to elucidate the patterns of shell evolution and habitat diversification of Cuora spp. in the future. In summary, we found that the two species occupy different niches. C. mouhotii prefers microhabitats with rock crevices and was more commonly observed on slopes with steep inclines in the field, for which its flattened shell is well suited, whereas the

Page 14 of 24 dome-shelled of C. galbinifrons restricts the species to hiding in deciduous leaves with bamboos and to microhabitats with relatively gentle slopes. Therefore, we conclude that morphological divergence allows the two Cuora spp. to partition the available microhabitats, and, thereby, to successfully co-occur. Acknowledgements We are grateful to F. Kong, Z. Zhao, and Q. Wang for assistance with field work. We are also grateful to the editors, R.M. Brigham and H. Guderley, and an anonymous reviewer for comments that improved this article. This study was supported by the National Natural Science Foundation of China (nos. 31260518 to J.W. and 31372228 to H.S.).

Page 15 of 24 Canadian Journal of Zoology References Benson, R.B.J., Domokos, G., Várkonyi, P.L., and Reisz, R.R. 2011. Shell geometry and habitat determination in extinct and extant turtles (Reptilia: Testudinata). Paleobiology, 37: 547 562 Bonin, F., Devaux, B. and Dupre, A. 2006. Turtles of the World. Johns Hopkins University Press, Baltimore, MD, USA. Bonnet, X., Delmas, V., El-Mouden, H,, Slimani, T., Sterijovski, B., and Kuchling, G. 2010. Is sexual body shape dimorphism consistent in aquatic and terrestrial chelonians? Zoology, 113: 213 220. Brown, J.H., and Bowers, M.A. 1985. Community organization in humming birds: relationships between morphology and ecology. The Auk, 102: 251-269. Cheung, S.M. 2007. Ecology, conservation trade of freshwater turtles in Hong Kong and southern China, with particular reference to critically endangered Cuora trifasciata. D. Phil. thesis, Department of Ecology and Biodiversity, The University of Hong Kong, Hong Kong SAR, China. Claude, J., Paradis, E., Tong, H., and Auffray, J.C. 2003. A geometric morphometric assessment of the effects of environment and cladogenesis on the evolution of the turtle shell. Biol. J. Linn.Soc. 79:485 501. Davies, T., Meiri. S., Barraclough, T., and Gittleman, J. 2007. Species co-existence and character divergence across carnivores. Ecol. Lett. 10: 146 152. Domokos, G., and Várkonyi, P.L. 2008. Geometry and selfrighting of turtles. Proc. R. Soc. Lond. B Biol. Sci. 275: 11 17. Ernst, C.H., and Barbour, R.W. 1989. Turtles of the World. Smithsonian Institution Press

Page 16 of 24 Washington, DC, USA. Grant, P.R. 1972. Convergent and divergent character displacement. Biol. J. Linn.Soc. 4:39-68. Greene, H.W. 1988. Antipredator mechanisms in reptiles. In Biology of the reptilia. Edited by C. Gans and R.B. Huey. New York: Alan R. Liss, 16:1 152 Guillemain, M., Fritz, H., Guillon, N., and Simon, G. 2002. Ecomorphology and coexistence in dabbling ducks: the role of lamellar density and body length in winter. Oikos, 98: 547-551. Higham, T.E., and Russell, A.P. 2010. Divergence in locomotor performance, ecology, and morphology between two sympatric sister species of desert-dwelling gecko. Biol. J. Linn.Soc. 101: 860-869. Hutchinson, G.E. 1959. Homage to Santa Rosalina or why are there so many kinds of animals? Am.Nat.93: 145 159. Ireland, L.C., and Gans, C. 1972. The adaptive significance of the flexible shell of the tortoise, Malacochersus tornieri. Anim.Behav. 20: 778-781. Joyce, W.G., and Gauthier, J.A. 2004. Palaeoecology of Triassic stem turtles sheds new light on turtle origins. Proc. R. Soc. Lond. B Biol. Sci. 271: 1 5. Lian, Y. 2009. The microhabitat use and home range of keeled box turtle (Cuora mouhotii). M.Sc. thesis, College of Life Science, Hainan Normal University, Haikou,China. Lindeman, P.V. 2000. Resource use of five sympatric turtle species: Effects of competition, phylogeny and morphology. Can. J. Zool. 78: 992-1008. Lue, K.Y., and Chen, T.H. 1999. Activity, movement patterns, and home range of the yellow-margined box turtle (Cuora flavomarginata) in northern Taiwan. J. Herpetol. 33:590-600.

Page 17 of 24 Canadian Journal of Zoology Ly, T., Hoang, H.D., and Stuart, B.L. 2013. Occurrence of the endangered Keeled Box Turtle, Cuora mouhotii, in southern Vietnam. Chelonian Conserv. Biol. 12: 184-187. Malonza, P.K. 2003. Ecology and distribution of the pancake tortoise, Malacochersus tornieri in Kenya. J. East Afr. Nat. Hist. 92:81-96. Miles, D.B. 1994. Covariation between morphology and locomotory performance in Scleropine lizards. In Lizard ecology: historical and experimental perspectives. Edited by L.J. Vitt and E.R. Pianka. Princeton: Princeton University Press, pp. 207 235 Motta, P.J., and Kotrschal, K.M. 1992. Correlative, experimental, and comparative evolutionary approaches in ecomorphology. Neth. J. Zool. 42:400-415. Pounds, J.A. 1988. Ecomorphology, locomotion, and microhabitat structure: patterns in a tropical mainland Anolis community. Ecol.Monogr. 58: 299-320. Pritchard, P.C.H. 2008. Evolution and structure of the turtle shell. In Biology of turtles. Edited by J. Wyneken, M.H. Godfrey, and V. Bels. CRC Press, Boca Raton. pp. 45 83. Rivera, G. 2008. Ecomorphological variation in shell shape of the freshwater turtle Pseudemys concinna inhabiting different aquatic flow regimes. Integr. Comp. Biol.48: 769 787. Rivera, G., Davis, J.N., Godwin, J.C., and Adams, D.C. 2014. Repeatability of habitat-associated divergence in shell shape of turtles. Evol. Biol.41: 29 37. Robinson, B.W., and Wilson, D.S. 1994. Character release and displacement in fishes: a neglected literature. Am. Nat. 144: 596-627. Romer, A.S. 1967. The vertebrate story. University of Chicago Press, Chicago, USA. Shi, H., Parham, J,F., Simison, W.B., Wang, J., Gong, S., and Fu, B. 2005. A report on the hybridization between two species of threatened Asian box turtles (Testudines: Cuora) in the

Page 18 of 24 wild on Hainan Island (China) with comments on the origin of 'serrata'-like turtles. Amphibia-Reptilia, 26: 377-381. Spinks, P. Q., Thomson, R. C., Hughes, B., Moxley, B., Brown, R., Diesmos, A., and Shaffer, H. B. 2012. Cryptic variation and the tragedy of unrecognized taxa: the case of international trade in the spiny turtle Heosemys spinosa (Testudines: Geoemydidae). Zool. J. Linn. Soc. 164: 811-824. Stayton, C.T. 2011. Biomechanics on the half shell: functional performance influences patterns of morphological variation in the emydid turtle carapace. Zoology, 114: 213 223. Stevens, M., and Merilaita, S. 2009. Animal camouflage: current issues and new perspectives. Philos. Trans. R. Soc. Lond. B Biol. Sci. No. 364: 423 427. Struijk, R.P.J.H., McCormack, T.E.M., Nguyen, T.T., Pham, T.V., Stumpel, J.B.G., Wang, J., and Auer, M. 2016. Intergradation between Cuora mouhotii mouhotii (Gray, 1862) and Cuora mouhotii obsti Fritz, Andreas & Lehr, 1998 with notes on the species geographical distribution and phenotypic variation. Sauria, 38: 31 47. Vanhooydonck, B., Van Damme, R., and Aerts, P. 2000. Ecomorphological correlates of habitat partitioning in Corsican lacertid lizards. Funct. Ecol.14: 358-368. Wang, J. 2007. Microhabitat use and home range of Indo-Chineses Box turtle (Cuora galbinifrons). M.Sc. thesis, College of Life Science, Hainan Normal University, Haikou,China. Wang, J., Gong. S., Shi, H., Liu, Y., and Zhao, E. 2011a. Reproduction and nesting of the endangered keeled box turtle (Cuora mouhotii) on Hainan Island, China. Chelonian Conserv. Biol. 10:159-164. Wang, J., Shi, H., Xue, C., Wang, L., and Zhao, E. 2011b. Population Densities of Cuora

Page 19 of 24 Canadian Journal of Zoology galbinifrons at Diaoluoshan Nature Reserve, Hainan Island, China. Sichuan Journal of Zoology 30:471-474. Wang, L. 2008. The activity rhythm and microhabitat use of Indo-Chineses box turtle (Cuora galbinifrons). M.Sc. thesis, College of Life Science, Hainan Normal University, Haikou,China. Wangyal, J. T., Wangchuk, D., and Das, I. 2012. First Report of Turtles from the Himalayan Kingdom of Bhutan. Chelonian Conserv. Biol. 11:268-272. Whitaker, N., Vijaya, J. 2009. Biology of the forest cane turtle, Vijayachelys silvatica, in South India. Chelonian Conserv. Biol. 8: 109-115. Xiao, F., Yang, C., Shi, H., Wang, J., Sun, L., and Lin, L. 2016. Background matching and camouflage efficiency predict population density in four-eyed turtle (Sacalia quadriocellata). Behav.Processes, 131: 40 46. Zhang, M., Zong, Y., and Ma, J.1998. Fauna Sinica (Reptilia 1): General accounts of Reptilia, Testudoformes and Crocodiliformes. Science Press, Beijing, China. Zhao, E., and Adler, K. 1993. Herpetology of China. Society for the Study of Amphibians and Reptiles in cooperation with Chinese Society for the Study of Amphibians and Reptiles. Oxford, Ohio, USA.

Page 20 of 24 Table 1 Characteristics of microhabitat used by C. mouhotii and C. galbinifrons. Values indicate the mean ± SD measurements for each species, as well as the statistical significance of their difference. Quadrat size C. mouhotii C. galbinifrons one-way ANOVA or Mann- Whitney U test Slope gradient ( ) 10 10 m 33.55 ± 8.5 26.81 ± 6.25 F = 12.679, P = 0.001 Canopy (%) 10 10 m 75.45 ± 10.39 80.19 ± 6.12 Z = -1.74, P = 0.082 No. fallen logs 10 10 m 2.84 ± 1.78 2.85 ± 1.23 Z =-0.491, P = 0.623 No. bamboo clumps 10 10 m 4.95 ± 2.47 4.00 ± 2.62 Z = -1.791, P = 0.073 No. tree holes 10 10 m 1.25 ± 0.87 0.89 ± 0.75 Z =-1.759, P = 0.079 No. rock crevices 10 10 m 5.45 ± 4.81 0.19 ± 0.48 Z = -5.846, P < 0.0001 Bamboo density (plants/m 2 ) 1 1 m 4.43 ± 4.37 3.85 ± 5.19 Z =-1.007, P = 0.314 Herbage height (cm) 1 1 m 41.01 ± 23.59 38.70 ± 16.68 F = 0.203, P = 0.654 Herbage density (plants/m 2 ) 1 1 m 4.41 ± 2.49 4.33 ± 2.97 Z = -0.552, P = 0.581 Decid. leaf cover (%) 1 1 m 48.07 ± 20.78 80.93 ± 14.08 F = 52.566, P < 0.0001 Decid. leaf thickness (cm) 1 1 m 6.18 ± 3.38 17.56 ± 24.88 Z = -5.779, P < 0.0001 Stone cover (%) 1 1 m 58.48 ± 35.21 2.78 ± 2.25 Z = -6.185, P < 0.0001 Table 2 Stepwise discriminant analysis of microhabitat variables of two Cuora spp. Variable No. Variables Discriminant coefficients Wilk s λ F P 1 Stone cover 0.962 0.51 66.29 < 0.0001 2 Slope gradient 0.424 0.463 39.429 < 0.0001 3 Bamboo density 0.36 0.432 29.382 < 0.0001

Page 21 of 24 Canadian Journal of Zoology Table 3 Difference in the morphology of two Cuora spp. Values indicate the mean ± SD measurements for each species, as well as the statistical significance of their difference (one-way ANOVA for carapace length and one-way ANCOVA for the remaining parameters, with carapace length as the covariate). C. mouhotii C. galbinifrons F P Sample size 18 13 Carapace length (mm) 162.91 ± 18.47 168.77 ± 8.61 1.12 0.298 Carapace width (mm) 114.24 ± 7.82 113.98 ± 4.73 1.77 0.194 Carapace height (mm) 65.81 ± 4.19 79.26 ± 3.67 101.53 < 0.0001 Carapace height / width 0.58 ± 0.03 0.70 ± 0.03 113.65 < 0.0001

Page 22 of 24 Figure Legends Fig. 1 The shared macrohabitat and divergent preferred microhabitats of two sympatric Cuora spp. in Hainan, China. A, the high-canopy cover habitat shared by both C. mouhotii and C. galbinifrons. B, the preferred microhabitat of C. mouhotii (rock crevices), with a turtle visible in the crevice (arrow). C, the preferred microhabitat of C. galbinifrons (deciduous leave litter under bamboo clumps), with a turtle visible under the leaf litter (arrow). All photos were taken by F. Xiao. Fig. 2 Distribution of the quadrats on the DFA axis.

Page 23 of 24 Canadian Journal of Zoology Fig. 1 The shared macrohabitat and divergent preferred microhabitats of two sympatric Cuora spp. in Hainan, China. 382x821mm (300 x 300 DPI)

Page 24 of 24 Fig. 2 Distribution of the quadrats on the DFA axis 105x75mm (300 x 300 DPI)