Values of the six karyological parameters are shown in S4. ANOVAs performed on karyological variables revealed significant differences among populations, but SNK failed to detect discrete groupings of populations (S5), apart from the centromeric index of Chromosome III, where Mediterranean and Atlantic populations (with the exception of CRo) were assigned to different groupings. Furthermore, CRo was uniquely characterised for centromeric index of Chromosome II, and PUo for relative lengths of Chromosomes I and III, and centromeric index of Chromosome III.
The plot of the first two nMDS dimensions separated CRo and PUo (OTUs D and E, respectively) better than morphological data (Figs 2(, 3(); the stress value of the ordination was slightly lower as well (7.24%). Gaussian clustering of standardized dimensional karyological data detected four clusters (S3B). The classification error rate with respect to the OTUs was less than that corresponding to morphological data (30%). Two Gaussian clusters corresponded almost perfectly to CRo and PUo, respectively (S3B), and nearly all individuals were assigned to the expected cluster with a probability of ≥ 95%. Once again two specimens from PUo could not be assigned to any cluster as their probability did not reach the assumed threshold; conversely an individual from COo was assigned to the same cluster as individuals from PUo. The remaining individuals that could be assigned to a cluster with a probability ≥ 95% were subdivided into two clusters that roughly resembled their Mediterranean or Atlantic origin. Only one Atlantic (from FEo) and eight Mediterranean specimens (four from PPx and four from CHx) were misassigned. Furthermore, with the exceptions of CRo and TJx (where all individuals were assigned), the proportion of unclassified individuals ranged from about 7% (ARx) up to 80% (CSo and PLo).