What the skull and scapular morphology of the dugong (Dugong dugon) can tell us: sex, habitat and body length?

The dugong (Dugong dugon, Müller) is an endangered marine mammal species. We examined the relationship between sex, habitat and body length based on the skull and scapular morphology and morphometrics of 81 dugong samples in Thailand. A total of 58 parameters from the skull and scapula (25 from the cranium, 23 from the mandible and 10 from the scapula) as well as tusks were used in this study. Data were analyzed by univariate analysis, followed by discriminant analysis and multivariate linear regression. Here we show, 100% and 98.5% accuracy rates for sexing using large tusks and the skull, respectively. Scapular morphology using the caudal border tubercle and coracoid process showed 91.30% and 96.15% accuracy rates for identifying males and females. Skull morphometrics could categorize dugong habitat, i.e. living in the Andaman Sea or Gulf of Thailand, with 100% accuracy. Moreover, our model could be used to estimate body length with coefficient of determination (R 2) of 0.985. The results of our study showed that skull morphology and morphometric measurements could be used as a tool for sex identification, location identification and estimation of body length. But scapular morphology is the best tool for sex identification in dugongs.

on sexual dimorphism of dugong skulls. This is possibly due to limitations on the number of dugong skulls and related records that are available for study. Moreover, Spain et al. also noted that the relationship between the development of upper second incisors and skull size possibly could be used to distinguish the sex of mature dugongs 18 . However, they noted that, in the case of female dugongs with fully erupted tusks, this cannot be used for identifying sex. Therefore, the primary purpose of our study was to use the skull and scapular morphology of dugongs to identify sex, body length and habitat. Information in this study may be beneficial in relation to dugong biology, conservation and forensics. For example, if dugong body remains are found and the sex cannot be identified or the actual body length measured due to decomposition, a method based on the discrimination functions from our study can be used to determine the sex or body length.

Material and Methods
Samples and history records. Samples were obtained from the Animal Anatomy Museum, Phuket Marine Biological Center, Phuket, Thailand (Supplement 1). A total of 81 dugong samples were taken from two different habitats, the Andaman Sea (n = 70) and the Gulf of Thailand (n = 11). These samples included skull bones from 37 males, 39 females and 5 of unknown sex. A total of 44 separate permanent tusks were also used in this study, categorized into two sizes, short (length shorter than 7 cm, n = 18) and long tusks (length longer than 7 cm,  Table 2). (cra. = cranial, cau. = caudal, dor. = dorsal, le. = left, ri. = right, ven. = ventral). dorsal view (C). However, the female cranium appears larger than that of the male. Comparative morphology of the male and female mandible, left lateral view (D) and dorsal view (E), also did not find markers for sex indentification. The female mandible was also larger than the male mandible. (cra. = cranial, cau. = caudal, dor. = dorsal, le. = left, ri. = right, ven. = ventral).   Table 3. Description of measurements taken from the dugong scapula (see Fig. 3). . It was difficult to use small tusks for sex identification, even when observing median plane sections; dental pulp (p) was similar between both sexes. In large permanent tusks there was a clear difference between sexes; dental pulp in males was triangle-shaped (similar to small tusks), but female tusks were filled with dentine, with only a small amount of dental pulp. n = 26). The recorded data used in this study included sex (male or female), body length, and habitat when alive (the Andaman Sea or the Gulf of Thailand). Animal bones from the Animal Anatomy Museum which were used in this study did not require approval by the Animal Ethics Committee, Faculty of Veterinary Medicine, Chiang Mai University.

Morphometric measurement parameters and morphology.
A total of 58 parameters were used in this study: 25 parameters from the cranium (Table 1 and Fig. 1), 23 from the mandible ( Table 2 and Fig. 2) and 10 from the scapula ( Table 3 and Fig. 3). These parameters were adapted from literature reviews of dugongs and other mammal species 8,10,11,18,26 . Measurements were obtained using either an osteometric board or digital vernier caliper to the nearest 0.1 mm. Each measurement was taken two times, at 48 h intervals, and was recorded as mean value.
The morphology of the skull was studied to establish reference hallmarks that could be used to distinguish the skull and scapula of male vs. female dugongs. Moreover, the presence of deciduous tusks was recorded (Fig. 1), and the outer and inner morphology of permanent tusks was observed to compare between males and females as well. For the study of the inner morphology, tusks were cut through the median plane.
Study design and statistical analysis. The body length of the mature dugong ranges from 2.2-2.4 m.
However, some previous studies used dugongs with body lengths of only 2.0 m for their investigation 18,27 . But our study was not limited to body lengths of 2.0-2.5 m because a good and efficient discrimination function should be able to be applied not only on adults but on immature animals as well. Moreover, some literature has reported that the dugong skull and skeleton continues to grow after puberty 18,22 . For this reason, in this study we included all skulls and scapulas to create discrimination functions that could be used for sex and habitat identification and body length estimation. All 58 parameters (48 from the skull and 10 from the scapula) were subjected to a Shapiro-Wilk normality test using the R program and were presented as mean ± SD with units in centimeters. Statistical analysis comparing between sexes (male and female) and habitats (Andaman Sea and Gulf of Thailand) was performed using either a t-test for normally distributed parameters or Mann-Whitney U test for non-normally distributed parameters. All of the parameters in the scapula showed a non-normal distribution,   Table 4. Measurement data (mean ± SD) taken from dugong skulls (cranium and mandible) and scapulas, by sex. All parameters are in cm, except superscript # indicates units in degrees. * Non-normally distributed parameter using Mann-Whitney U test for testing a significant difference. and the data set of mophometrics for dugongs from the Gulf of Thailand, a small population, also resulted in a non-normal distribution. Analysis of differences between habitats was performed using Mann-Whitney U test.
To establish equations for sex or habitat determination, we used ratios which represented the proportion of each parameter within the same bone (for example, BCC value was divided by other parameter values of the dugong cranium). In this study, we compared the effectiveness of the morphometrics of the cranium, mandible, skull (cranium and mandible) and scapula in creating a function for assigning sex or habitat. These ratio data sets were analyzed through a stepwise discriminant analysis with leave-one-out classification based on four parts: (i) cranium, (ii) mandible, (iii) skull and (iv) scapula. In order to find correlations between the morphometrics of the skull and scapula and dugong body length, a multivariate linear regression model was used to create an equation for estimating the body length of dugongs based on these parameters of skull and scapular morphology. The average body lengths of dugongs in this study ranged from 0.88 to 3.64 m. In all analyses, a p-value < 0.05 was considered to be statistically significant.

Results
Descriptive morphology of the skull, tusk and scapula, comparing males and females. The skull (cranium and mandible) morphology of males and females was similar, and no significant hallmarks were found that could be used as sex characteristics. However, male skulls were smaller than female skulls, even though the body was larger or longer. An example is shown in Fig. 4, comparing the cranium and mandible of a male dugong whose body length was 3.64 m with a female dugong with body length of 2.59 m.
The existence of the deciduous tusks for dugongs was explored. Overall, 41% of males and 54% of females presented deciduous tusks. However, 100% of dugongs with body size more than 2.31 m did not presented deciduous tusks while 77% of dugongs with body length less than 2.31 m presented deciduous tusks.
Both outside and median plane sections of small tusks in males and females were similar (Fig. 5). The median plane showed dental pulp in a triangular shape in both sexes. The proximal part of large tusks in females was more convex than in males, and the median plane view showed a clear difference between sexes. Male tusks had dental pulp in a triangular shape, while females had a small amount of dental pulp. From this morphology, large permanent tusks of dugongs showed clear differences between males and females, with 100% accuracy from a total of 27 large permanent tusks. While small permanent tusks were not able to be used for sex identification, 39% (7/18) were accurately identified as male tusks, while none could be identified as tusks from females.
The scapula of the dugong is flat with a triangular shape, slightly curved inward toward the medial surface. This study found two significant hallmarks on the scapula that could be used as a tool for sex identification in dugongs (Fig. 6). The first was on the caudal boarder of the scapula: males presented a caudal border tuberosity, which was absent in females. The second was on the coracoid process: in males this presented as a short and thick shape, while in females it presented as longer and thinner compared with males. These two characteristics had a 91.30% (21/23) accuracy rate for identifying males and a 96.15% (25/26) accuracy rate for females. However, we note that these two characteristics can be used only in mature scapulas (all epiphyses on the scapula are closed).
Morphometric correlation. The correlation among the parameters of skull bones, including cranium and mandible, was performed by Pearson's correlation analysis (Fig. 7A). It was found that most correlations of these parameters were positive, with the highest positive relationship observed between MH and CH. A negative relationship could be found between MA1, MA2 and PA and the others; the highest negative relationship was noted between MA2 and CrPMH. Some parameters of skull morphometrics had a linear relationship with the body length of dugongs, especially MSL and BMW with 0.62 and 0.59, respectively. All parameters among scapular morphometrics showed a positive relationship; GCW and SL showed the highest relationship with 0.94 (Fig. 7B). The highest correlation of scapular morphometrics and body length was 0.61, for SW. However, SA2 was excluded from this correlation analysis because some values were missing. Moreover, we could not find a relationship (p = 0.1024) between dugong tusk length and body length for either sex (Fig. 8).
Sex determination using skull and scapular morphometric equations. A total of 76 dugong samples, including males (n = 37) with body length between 1.16-3.64 m (2.17 ± 0.46) and females (n = 39) with body length between 0.88-2.81 m (2.26 ± 0.46), were included in this study. Another 4 dugongs were excluded due to unknown sex. The morphometric measurements proved that both the cranium and mandible from males were smaller compared with females. Out of a total of 48 parameters (Table 4) used to compare between males and females, 29 of 43 length parameters for females were higher (p > 0.05) than for males, and 4 of 5 angles for males were higher than for females. However, there was no significant difference of any parameters in both the cranium and mandible, while scapular morphometrics had two significant variables, SW and SNW, which showed a remarkable difference between males and females (Table 3). Ratios obtained from morphometrics of the cranium, mandible and skull were analyzed by stepwise discrimination to generate equations for sex determination; each equation consisted of three, two and thirteen variables for the cranium, mandible and skull, respectively ( Table 5). As shown in Fig. 9, the distribution of discriminant values (DV) obtained from the equations and the boundary decision for sexing was set at zero of DV. The DV distribution of the cranium, mandible and scapula had large overlapping areas between males and females (Fig. 9A,B,D), whereas that of the skull had a smaller overlapping area (Fig. 9C)  were correctly determined, while females had an incorrect prediction. In contrast, the equation obtained from the scapula displayed the lowest accuracy (68.1%) and precision (68.1%) ( Table 5).

Habitat identification using skull and scapular morphometric equations. A total of 80 dugongs
were taken from two different habitats -69 dugongs from the Andaman Sea (male = 33, female = 33, sex unknown = 4, and body length = 2.20 ± 0.48 m) and 11 dugongs from the Gulf of Thailand (male = 4, female = 6, sex unknown = 1, and body length = 2.21 ± 0.28 m). One dugong was excluded from the study due to missing habitat records. Out of a total of five parameters, three parameters from the cranium (BW, NCW, ZAL) and one parameter from the mandible (CW) had significantly higher values in skulls from the Andaman Sea, while one parameter from the mandible (CL) showed a significantly higher value in skulls from the Gulf of Thailand (Table 6). In the scapula, we found that SN exhibited a significant difference, i.e. greater in dugongs from the Andaman Sea. As shown in Table 7, three equations from a stepwise discriminant analysis based on the morphometrics of the cranium, mandible and skull, included five, six and sixteen variables, respectively. The boundary decision for habitat determination was zero of DV. The distributions of DV from three equations are displayed in Fig. 10; overlapping areas were observed for the cranium and scapula but not for the mandible and skull. The results of habitat identification based on morphology revealed that the skull equation possessed the highest accuracy (100%) and precision (100%), while the cranium and mandible equations had an accuracy of 93.8% and 97.3%, respectively. The scapula appeared to have a low accuracy for identifying habitat.

Body length determination using skull morphometric equation. The average body length of 80
dugongs was 2.20 ± 0.46 m, ranging from 0.88 to 3.64 m. One dugong was excluded from the study because the record of body length was missing. According to the correlations between body length and other parameters (Fig. 7), most did not exceed a moderate correlation. To acquire an effective equation for estimation of dugong body length, multivariate linear regression was performed based on the morphometrics of the cranium, mandible and scapula. As shown in Table 8, the fittest model with the highest adjusted correlation coefficient of 0.985 was a skull-based equation, followed by the cranium, mandible and scapula with correlation coefficients of 0.746, 0.381 and 0.371, respectively. The error of the skull-based equation was 6.5%. In the skull-based equation, all 12 variables displayed a significant coefficient at p < 0.05. Figure 11 illustrates the agreement between body length and predicted body length obtained from each equation. We found that almost all of the values in the skull-based equation were on the diagonal line, indicating high accuracy and precision.

Discussion
The results from this study have implications for dugong conservation, ecology and osteology, and can be used to monitor changes in the population structure, i.e. the number, size and sex of dugongs. Sex and body length frequency distribution can provide life-table information on living dugongs. Moreover, the results of this study expanded the basic knowledge of dugong osteology, revealing sexual dimorphism in the skull and scapula. For example, estimation of the sex ratio of the dugong population would provide the current status of the dugong, leading to the creation of an appropriate conservation plan. Since the first published report on the morphometric differences between male and female dugongs in 1976, no studies have been performed on this topic. The highlight of our study was demonstrating that skull morphometrics could distinguish between males and females with a 96.9% accuracy rate, while using large permanent tusks gave 100% accuracy. Skull measurements were different between dugongs living in the Andaman Sea and the Gulf of Thailand with 100% accuracy. Moreover, some skull and scapular parameters can be used to estimate body length with coefficient of determination (R 2 ) of 0.985, while scapular morphology using the caudal border tuberosity and coracoid process showed a 91.30% accuracy rate for identifying males and a 96.15% accuracy rate for identifying females. Additionally, we noted that the age range for dugongs can be predicted based on the presence or absence of deciduous tusks. Overall, we have shown that skull morphology and morphometric measurements and scapular morphology can be used as a tool for sex identification, habitat identification and body length estimation.

Skull/scapula and sex identification.
In other species (e.g. human, canine and leopard) the skull can be used for sex identification, either from reference hallmarks or a morphometric equation 8,[28][29][30] . In this study we showed that sexual dimorphisms can clearly separate males and females by using large permanent tusk morphology, scapular morphology and a skull morphometric equation. Large permanent tusks can be used as a tool for sex identification with a 100% accuracy rate; the scapula could be used to distinguish males and females with 91.3% and 96.1% accuracy, respectively; and skull morphometrics (using parameters from the cranium and mandible) showed a 96.7% accuracy rate. Our study did not find any significantly different parameters of the skull between sexes, but most parameters were higher in females. Scapular width and caudal scapular notch width was greater in females than in males (p < 0.05), but a morphometric equation from scapular parameters gave a poor accuracy rate (68.1%). A previous study 18 reported that 6 out of 26 parameters were significantly different between sexes, but only the anterior snout width was larger in males; the other 5 parameters (snout length, pterygoid-frontal depth, anterior mandible depth, extra-mandibular chin width and mandibular length) were significantly higher  Table 6. Measurement data (mean ± SD) taken from dugong skulls (cranium and mandible) and scapulas from two habitats (the Andaman Sea and the Gulf of Thailand). All parameters are in cm, except superscript # indicates units in degrees.
in females. However, we noted that the previous study had overlapping data, without a clear cut-off value, so this feature may not provide complete separation between sexes. A discrimination function is the best tool for use in sex identification in case large tusks are absent or cutting through tusks is prohibited. We created a discrimination function from three different data sets: cranium, mandible and skull. The results show that skull morphometrics had the highest accuracy (96.7%) and precision (96.9%) rate, while using only parameters from the cranium or mandible did not give higher accuracy and precision. The morphometrics of the dugong skull gave an effective equation for sex identification, as shown in Table 5. Dugong scapulas also exhibited sexual dimorphism that can be used for identifying sex with a high accuracy rate as well.
Skull/scapula and habitat identification. In many species (e.g. lion, leopard and sea turtle), an animal's habitat has an effect on skeleton size 11,28,31,32 . Our study also looked for a relationship between skull morphology and location. We found that skulls from dugongs living in the Andaman Sea seemed to be larger than skulls from dugongs living in the Gulf of Thailand. This study found that the base width of the cranium, zygomatic arch length, premaxillae angle of the cranium and mandibular condyle width of dugongs from the Andaman Sea were significantly higher compared with dugongs from the Gulf of Thailand. Only mandibular condyle length was significantly higher in dugongs from the Gulf of Thailand. Also, the minimum width of the scapula in dugongs living in the Gulf of Thailand was significantly lower (p < 0.05) compared with dugongs living in the Andaman Sea. After analyzing the discrimination functions from four different data sets -cranium, mandible, skull and scapula -the skull was found to show the highest accuracy and precision rates, of 100%. But the cranium and mandible also showed high accuracy rates of over 90%, while the scapula showed the lowest accuracy rate (73.7%). The morphometrics of the dugong skull provided an effective equation for habitat identification, as shown in Table 7.
Skull/scapula and body length estimation. Body length in Sirenia is widely used for estimating body weight 33,34 and maturation 22,35 . A previous study demonstrated a correlation between skull measurements, body length and field weight in Florida manatees 36 . However, only four parameters were used, including basal skull length, occipital condyle width, foramen magnum width and foramen magnum height. The highest correlation (R 2 = 0.90) was between foramen magnum width and field weight. Spain and Heinsohn 25 studied the correlation between 26 parameters from the skull and body length in 52 dugong samples from Australia (26 mature and 26 immature), and found the best parameter to be condylo-premaxillary length. Their proposed equation was: Y = 54.214 + 1.650X + 0.110X 2 , where Y is the body length from the tip of the snout to the notch in the tail fluke and X is the condylo-premaxillary length, both in centimeters. However, the authors did not report the correlation rate from this equation. In our study, 57 parameters were used (25 from the cranium, 23 from the mandible and 9 from the scapula) and four models were generated from various parameters of the cranium, mandible, skull and scapula. A model from the skull to estimate body length using 12 parameters showed the highest correlation (98.5%) compared with the other three models, as shown in Table 8.
Not only skull morphology can be used as a tool for estimating dugong body length. Burgess et al. 35 reported that male dugongs with body length lower than 240 cm usually had unerupted tusks, while erupting tusks are usually found in dugongs whose body length is approximately 240-259 cm, and dugongs with body length over than 260 cm typically have erupted tusks. However, our study did not find this correlation; but we did find a correlation between tusk length and body length, i.e. there was a low correlation with body length. Study Limitations. Our study had many limitations. First, we only had a small number of dugong samples from the Gulf of Thailand, conforming to a previous study which reported few dugongs living in the Gulf of Thailand compared with the Andaman Sea 4, 34, 37 . Only 11 animals from the Gulf of Thailand were used to create an equation to differentiate between the two different locations; thus there was a low power of confidence, even though the result showed very high accuracy (100%). However, it is possible that dugongs living in the two different locations had different skull morphometrics, because the average body length of 69 dugongs taken from the Andaman Sea (2.20 ± 0.48 m) and 11 dugongs taken from the Gulf of Thailand (2.21 ± 0.28 m) was not significantly different. Dugong age was also a limitation; we did not know the true age of these animals, so our study was confined by these two limitation in finding a correlation between skull morphometrics and age. Other studies have tried to estimate dugong maturation based on body length, which defined mature size as longer than 2.2-2.4 m 18,27 . However, the 44 stranded dugongs used in this study which were found around the coast of Thailand had a lower average size (2.16 ± 0.51 m) compared with those reports 18    dugongs from different habitats also had a different average size. A previous study reported that out of a total of 101 stranded dugongs in Thailand 34 , 47.5% (48/101) were 2.0-3.0 m long, 31.7% (32/101) were 1.5-2.0 m long and 20.8% (21/101) were less than 1.5 m long. The last limitation was obviously seen when we categorized dugong body length into six groups: lower than 1 m (n = 1), 1.00-1.50 m (n = 2), 1.51-2.00 m (n = 27), 2.01-2.50 m (n = 26), 2.51-3.00 m (n = 23) and over 3 m (n = 1). The distribution of dugong body length was not equal; only one dugong was longer than 3 m and only one was shorter than 1 m (see details in Supplement 1). Nevertheless, we strongly believe that our study of dugong body length showed high reliability because of the large number of dugong skeletons, possibly serving as an accurate representation of a real dugong population.
Take-Home Message. Study of the morphology of the dugong skull, scapula and other bones can provide useful information in relation to dugong biology, conservation and anatomy. However, the main limitation for studying dugongs is the number of bone samples available in collections and reliable recorded data of these samples. In addition to the presence or absence of deciduous tusk, another one of our ideas that we would like to share with all scientists is to estimate age using skull and scapular morphology and morphometric parameters, because age estimation of stranded dugongs is an important issue for ecology, biology and conservation. In many kinds of animals, studies have used bone morphology or morphometrics for age estimation. Stansfield demonstrated using skull morphology to estimate African elephant age by using an age reference line 12 . In dugongs, research has shown many parameters related to age, such as incisor dentinal growth layer groups 38 , body length 39 or closure of cranial sutures 40 . However, there is always a question about estimating the real age of dugongs. Our results showed that an equation or model created from many skull parameters could be used for sex and habitat identification and body length estimation, with a very high accuracy rate. We also believe that skull morphology can be used to estimate age and will give a high accuracy rate as well. However, to create this model we had to have dugong skulls for which we knew the actual ages. In contrast, scapular morphometrics showed poor discrimination function but presented excellent hallmarks for sex identification. Studies of other bones might reveal some surprising results and provide valuable insights for dugong biology, conservation and forensic science.

Conclusion
From this study we learned that dugong skull and scapular morphology and morphometrics can be used for sex and habitat identification as well as body length estimation. A discrimination function from skull morphometrics and two markers on the scapula from our study can be used at both the field and laboratory level for dugong conservation and forensic science. We also noted that a discrimination function that used many parameters from the cranium and mandible gave a better result and accuracy rate than using a single bone or single parameter. Due to the size of bones used in this study and the outcome of all measurement results, we believe that dugongs living in the Gulf of Thailand are of smaller size than dugongs living in the Andaman Sea. But to prove this hypothesis, further studies need to be performed.