To refer to this article use this url: http://ctoz.nl/vol86/nr03/a01
Morphometry
Raw data. The PCA based on raw data shows a strong size-dependency of most characters, with the first axis (Eigenvalue = 12.0) explaining 54.6% of the variation and the second one (Eigenvalue = 3.52) only 16.0% (Fig. 2A). All characters of the Iberian population were size-dependent (except LA and PS), while the only size-dependent characters in the Congolese population were WL, NS, WWP, VCL, DLS and VCS (Table 1A). The size-relationship in the PCA plot was also more evident for the Iberian population, whose specimens showed a wider size-range representativeness than the Congolese ones. Nevertheless, the two populations clearly form separate groups in the PCA plot, and turned to be significantly different (ANOSIM, Global R = 0.435, significance level = 0.1%).
Fig. 2. Principal Component Analyses plots. a. Based on raw data. b. Based on size independent data. c. Based on measurement proportions. |
Table 1. Relationships with worm size (as body width with parapodia) in Oxydromus okupa sp. nov., with bold characters indicating significant differences. A. Morphometric measurements. B. Taxonomically relevant proportions. Measurement abbreviations as in Fig. 1 . Coeff: Pearson correlation coefficient; p: significance level. |
The intra-population average distances for the Iberian and Congolese worms were 29.9% and 7.10%, respectively, whilst the inter-population dissimilarity was 48.67%. PS, LA and PP, and NS, WL and PP most contributed to the Iberian and Congolese intra-population similarity, respectively, whilst LA, PS and PP most contributed to the inter-population dissimilarity (SIMPER, Table 2).
Table 2. List of the ten most contributing raw measurements to the intra-population similarities and inter-population dissimilarity based on the SIMPER analyses. Measurement abbreviations as in Fig. 1. Av.Value: Average value; Av.Sq.Dist: average square distance; Sq.Dist/SD: square distance divided by standard deviation; Contrib%: percentage of contribution; Cum.%: cumulative percentage of contribution. |
All averaged character measurements showing significant differences were higher in the Iberian than in the Congolese population, except for NS and DLS, with the most remarkable difference being at LA (one-way ANOVA, Table 3).
Table 3. Comparative table of the raw morphometric measurements (µm) in the Iberian and Congolese populations. Differences expressed as percentages, with bold characters indicating significant differences (according to Benjamini & Hochberg, 1995) that may be higher (normal text) or lower (italics) in the Iberian than in the Congolese population, respectively. Measurement abbreviations as in Fig. 1. Min: minimum; Max: maximum; Mean: average ± standard deviation; F: Fisher’s F index; p: significance level. |
Based on raw data, the discriminant function included the variables DLS, PP, PNCLL and LA (Table 4). This function correctly classified the 100% of individuals from both localities (Fig. 3) and the success probability in the cross validation was of 98%.
Table 4. Results of the discriminant analyses based on raw data, size-independent data and measurement proportions. F: Fisher’s F index; Lambda: Rao approach to the Wilks’ Lambda test; p: significance level; Coeff.: Standardized coefficients for the variables included in the inter-population discriminant functions, arranged in a decreasing order, according to their contribution to the inter-population discrimination; Congo, Iberian: coefficients of the selected variables in the classification functions for the Congolese and Iberian populations, respectively. Intercept: Intercept of the classification functions for each dataset. Measurement abbreviations as in Fig. 1. |
Size-independent data. For the purpose of this analysis, and taking into account that the size range was well represented in the Iberian population, only LA and PS raw data were considered as size-independent (Table 1A). The PCA plot based on size-independent data revealed again two clearly different groups corresponding to the two studied populations (Fig. 2B). Despite the axes being less representative than those obtained for the raw data (Eigenvalues = 6.61 and 3.49, variation explained = 31.5% and 16.6%, respectively for axis 1 and 2), the two populations were more clearly distinguishable and showed more significant differences than the raw data (ANOSIM, global R = 0.539, significance level = 0.1%).
Based on these size-independent characters, the intra-population average distance within the Iberian and Congolese populations was 21.0% and 11.1%, respectively, whilst the inter-population distance was 50.3%. PS, DAPE and LA, and WL, NS, VCS, and PP most contributed to the Iberian and Congolese intra-population similarity, respectively, whilst the Congolese vs. Iberian dissimilarity was mainly explained by LA, DAPE, DLS, PP, and PS (SIMPER, Table 5).
Table 5. List of the ten most contributing size-independent measurements to the intra-population similarities and inter-population dissimilarity based on the SIMPER analyses. Measurement abbreviations as in Fig. 1. Av.Value: Average value; Av.Sq.Dist: average square distance; Sq.Dist/SD: square distance divided by standard deviation; Contrib%: percentage of contribution; Cum.%: cumulative percentage of contribution. |
When comparing the averaged character measurements, all those showing significant differences were higher in the Iberian than in the Congolese population (particularly, LA, PS, PP and DAPE), except for NS, WWP, DLL, and DLS in particular (one-way ANOVA, Table 6).
Table 6. Comparative table of the size-independent morphometric measurements in the Iberian and Congolese populations. Differences expressed as percentages, with bold characters indicating significant differences (according to Benjamini and Hochberg, 1995) that may be higher (normal text) or lower (italics) in the Iberian than in the Congolese population. Measurement abbreviations as in Fig. 1. Min: minimum; Max: maximum; Mean: average ± standard deviation ; F: Fisher’s F index; p: significance level. |
Based on size-independent data, the discriminant function included the variables WL, NS, DLS, PP, PNCLL, and LA (Table 4). This function correctly classified the 100% of individuals from both localities (Fig. 3) and the success probability in the cross validation was 100%.
Character proportions. Nine of the 20 character proportions analysed showed significant negative (n = 7) or positive (n = 2) correlations with size, indicating allometric relationships in the Iberian population, while only one was significantly negatively correlated with size in the case of the Congolese population (Table 1B).
In the PCA based on character proportions, Axes 1 (eigenvalue = 5.38) and 2 (eigenvalue = 3.13) explained 26.9% and 15.6% of the variation, respectively. The PCA plot also highlighted a marked clustering for the individuals of the two populations under study (Fig. 2C), the results being slightly less discriminant than the previous ones but equally highly significant (ANOSIM, Global R = 0.421, significance level = 0.1%).
The average intra-population distance for the Iberian and Congolese populations were 21.4% and 11.1%, respectively, whilst the average inter-population distance was 46.0%. DAPE/HL, DCSS/DLS, and LA/HL, and WL/WW, NS/WW, DCSS/DLS, and LA/LH were the most informative proportions for the intra-population similarity in Iberian and Congolese worms, respectively, whilst the intra-population dissimilarity was mainly explained by DCSL/DLL, LA/HL, NS/NW, and DAPE/HL (SIMPER, Table 7).
Table 7. List of the ten most contributing measurement proportions to the intra-population similarities and inter-population dissimilarity based on the SIMPER analyses. Measurement abbreviations as in Fig. 1. Av.Value: Average value; Av.Sq.Dist: average square distance; Sq.Dist/SD: square distance divided by standard deviation; Contrib%: percentage of contribution; Cum.%: cumulative percentage of contribution. |
All averaged character proportions showing significant differences, except for NS/WW and WWP/WW, were higher in the Iberian than in the Congolese population, most of them with differences higher than 10% and particularly higher than 20% in the case of DCSS/DLS (one-way ANOVA, Table 8).
Based on character proportions, the discriminant function included the variables WL/WW, NS/WW, DCSS/DLS, PS/HL, and DAE/DPE (Table 4). This function correctly classified the 100% of individuals from both localities (Fig. 3) and the success probability in the cross validation was 96%.