Fig. 3 Population structure of H. tibetana (a, b, c), H. digitata (d, e, f), H. gregoryi (g, h, i) and H. platon(j, k, l, m), including results of PCA (a, d, g, j) and admixture (b, e, h, k) analyses based on genomic data. Best fit K in the admixture estimates was indicated with black boxes except for H. gregoryi , for which the methods outlined in Evanno et al. (2005) did not lead to conclusive results. The colour and the code represented the minimal basins that the sample was assigned to. (c, f, i, l) Sampling sites of the four target species in this study. The inner pie chart indicates the distribution of inferred admixture clusters with best fit K (except forH. gregoryi , for which K3 was selected) of each species at the given sites; Outer ring pie charts indicate the distribution of haplogroups calculated by the MJ network based on mtCOI (more details see Fig. S22). Colourful patches on the map show the minimal basins as categorized in this study. (m) Paleo-drainage rearrangement of the Yangtze Basin and sampling location of H. platon in this area. The inner pie chart showed the distribution of inferred admixture clusters with K4. Outer ring pie charts indicated the distribution of haplogroups calculated by the MJ network based on mtCOI. The Paleo-drainage maps were reproduced from (Deng et al. 2021a).