Uncertainties and risks in delimiting species of Cambeva (Siluriformes: Trichomycteridae) with single-locus methods and geographically restricted data

Cambeva contains species with complex taxonomy or poorly delimitated in terms of morphology and geopraphic distribution. We conducted an extensive review of Cambeva populations from coastal drainages of Southern to Southeastern Brazil to evaluate species geographic limits with an integrative analysis including morphological and molecular data (COI). We test if two single-locus methods, Bayesian Poisson Tree Processes (bPTP) and Generalized Mixed Yule Coalescent (GMYC), are efficient to delimit species boundaries in Cambeva by the comparison with the diagnosable morphological units. Using GMYC, we also evaluated the combination of tree and molecular clock priors to reconstruct the input phylogeny and assessed how well the implemented model fitted our empirical data. Eleven species were identified using a morphological diagnosability criterion: Cambeva balios, C. barbosae, C. botuvera, C. cubataonis, C. davisi, C. guaraquessaba, C. iheringi, C. tupinamba, and C. zonata and two treated as undescribed species. In contrast with previous knowledge, many of them have wider distribution and high intraspecific variation. Species delimitation based on single-locus demonstrated incongruences between the methods and strongly differed from the morphological delimitation. These disagreements and the violation of the GMYC model suggest that a single-locus data is insufficient to delimit Cambeva species and the failure may be attributable to events of mitochondrial introgression and incomplete lineage sorting.