loading page

Supervised machine learning to determine species boundaries reveals the scale of mislabelling in Dicroglossid frogs across the Eastern Indomalaya and Southeastern Palearctic
  • +7
  • Siti N. Othman,
  • Yuran Dong,
  • Boyang Yu,
  • Tae Eun Um,
  • Dallin B. Kohler,
  • Vishal Prasad,
  • Deyatima Ghosh,
  • Xiaoli Zhang,
  • Jiechen Wang,
  • Amaël Borzée
Siti N. Othman
Nanjing Forestry University
Author Profile
Yuran Dong
Nanjing Forestry University
Author Profile
Boyang Yu
Nanjing University
Author Profile
Tae Eun Um
Nanjing Forestry University
Author Profile
Dallin B. Kohler
Nanjing Forestry University
Author Profile
Vishal Prasad
Nanjing Forestry University
Author Profile
Deyatima Ghosh
Nanjing Forestry University
Author Profile
Xiaoli Zhang
Nanjing Forestry University
Author Profile
Jiechen Wang
Nanjing University
Author Profile
Amaël Borzée
Nanjing Forestry University

Corresponding Author:[email protected]

Author Profile

Abstract

Widespread amphibians are frequently impacted by incorrect species assignments and inaccurate taxonomic labelling in open databases and published studies, leading to confusion regarding species boundaries and missing information for population trends at the ecozone level. We examined three genera in the family Dicroglossidae (Fejervarya, Hoplobatrachus and Quasipaa) across the Indomalaya and Palearctic ecozones, where unresolved species boundaries and the introduction of exotic populations are significant issues. Alternative to conventional phylogeny, we employed the Bayes Factor Delimitation method and nested sampling to assess 18 comparative delimitation models for 15 verified species in GenBank, using 4,044 bp unlinked multilocus data from 487 individual Asian Dicroglossids across both ecozones. Our best-supported model delineated 17 clades across three genera, highlighting Fejervarya as the genus with the shallowest clade assemblages, with two clades within each of the F. kawamurai, F. multistriata, and F. limnocharis species complexes, all diverging below 3.70 Mya in the Northern Indomalaya and Southeastern Palearctic. Through adjusted ecological niche modelling with species delimitation data, we verify that precipitation of the driest month, elevation, and mean diurnal temperature range, are key variables influencing the distribution of F. kawamurai and F. multistriata in mainland East Asia. Cross validation using call analysis supported significant differences in call durations and intervals between F. kawamurai [Clade II] and F. multistriata [Clade II] in their overlapping range in the eastern Yangtze River Basin indicated the existence of subspecies boundaries even in the absence of physical barriers and validated the presence of F. multistriata along the central coast of China, i.e. Shanghai. Our findings also showed the clustering of the southern Palearctic Fejervarya clade, including the population in the Jiangsu Plains, within the Yangtze group of F. kawamurai, and delineated the northern limit of F. multistriata in the vicinity of the Huai River.