Abstract: The effects of evolution models on success rates of species identification via DNA barcoding were evaluated based on specimens of 44 Noctuidae species collected from Wuling Mountain, Hebei, northern China. Different phylogenetic tree reconstruction methods, including the neighbor-joining, the maximum parsimony, the maximum likelihood and the Bayesian inference, were used to reconstruct phylogenetic trees based on the COI gene sequences of these species, and success rates of different models (12 models for the neighborjoining, 7 models for the maximum likelihood method, and 2 models for the Bayesian inference) were evaluated systematically. The results showed that there were no significant differences in success rates of species identification among 12 different models for the neighborjoining method, but significant differences in success rates were found among different evolution models for both maximum likelihood and Bayesian methods. As a comparison, the maximum parsimony method which is not based on evolution models obtained stable success rates of species identification. The 6 evolutionary models shared by the neighborjoining and maximum likelihood methods obtained different success rates between these two methods. In addition, we found that singletons in the database largely reduced success rates of species identification, suggesting that deep intraspecific sampling be required in the future studies of DNA barcoding.

Key words: Lepidoptera, Noctuidae, COI gene, evolution model, DNA barcoding, phylogenetic trees, species identification