Differentiating three Indian shads by applying shape analysis from digital images
Dwivedi, Arvind (2020), Differentiating three Indian shads by applying shape analysis from digital images, Dryad, Dataset, https://doi.org/10.5061/dryad.02v6wwq0m
Study areas and sample collection
Samples of T. ilisha were collected from Digha and Kakinada on the coast of Bay of Bengal and Bharuch on Narmada River. River Narmada originates as Narmada Kund, located at Amarkantak and flows through the Gulf of Khambhat into the Arabian Sea. Samples of T. toli were collected from Okha, Gulf of Kachchh on the Arabian Sea coast (Singh, 2002). Samples of H. kelee were collected from Rajahmundry on the Godavari River which flows to Southeast coast and join the Bay of Bengal.
A total of 120 samples comprising of T. ilisha, T. toli and H. kelee were collected based on taxonomic key characters (Whitehead, 1985). The sample size and related information for three species selected to document morphometric variations are presented in Table 1 and Figure 1.
As per IUCN status, T. ilisha is under Least Concern (LC), and T. toli and H. kelee are under Not Evaluated (NE) category. Fish samples were obtained from wild, directly from the commercial catches. Sites from where fishes were collected fell outside Protected Areas (PAs) and therefore no permits were required from State Forest and Wildlife Department.
Meristic counts were followed as per Whitehead (1985). Fin rays were counted under transmitted light using a stereoscopic microscope. Results from previous studies (Whitehead, 1985; Quddus et al., 1984a; Narejo et al., 2008; Jawad et al., 2011b; Antony et al., 2005; Losse, 1968) were also included in this study for comparison between species.
Landmark based morphometric analysis
Digital images of individual fish, labelled with a specific code for identification, placed on laminated graph sheets, body posture and fins were teased into a natural position, was captured using a Cyber shot DSC-W300 digital camera (Sony, Japan).
Two–dimensional Cartesian coordinates of 13 landmarks were recorded on the left view of each specimen (Figure 2). Data were generated from digitized images using a combination of three different softwares: tpsUtil was used for converting graphics images in to ‘tps’ format; tpsdig was used for fixing landmarks on the images and also setting scale for the image; PAST (Hammer et al., 2001) was used for generating truss data based on the landmarks (Rohlf, 2006). A total of 78 inter landmark morphometric characters were extracted by measuring distances between landmarks.
The data were log-transformed (Strauss, 1985) and size effect was removed by using the following (Elliott et al., 1995):
Madj = M(Ls/L0)b
where M is the original measurement, Madj the size adjusted measurement, L0 the standard length of the fish, Ls the overall mean of standard length for all fish from all samples in each analysis, and b estimated for each character from the observed data as the slope of the regression of log M on log L0 using all fish from each group.
Standard length (SL, character code 1-6) was not used in the final analysis because SL was used as a basis for transformation (Mamuris et al., 1998) and thus 77 morphometric characters were retained. Univariate analysis of variance (ANOVA) was performed to assess the significant difference among the three species. The significant variables (p<0.01) were subjected to Principal Component Analysis (PCA) and Discriminant Function Analyses (DFA). Box plot was prepared for each discriminant morphometric characters. Range bar of ratios of traditional morphometric data with standard length, extracted from differentiating landmark based morphometric characters were prepared. Canonical Variate (CV) plot was generated from significant variables to visualize the variations among groups. Squared Mahalanobis distance among the three species was computed (Puillandre et al., 2012) and distances between the centroids of the three species were visualized by performing non-metric multidimensional scaling (Legendre & Legendre, 1998). An unweighted per-group method with arithmetic averaging (UPGMA) cluster analysis (Hair et al., 1998) based on the Squared Mahalanobis distance between the group centroids was applied to determine the similarity/dissimilarity between species. Statistical analyses for morphometric data were performed using the software packages viz. SPSS version 16, JMP version 9 and Excel (Microsoft Office 2007).