Phylogenetic analysis
To analyze mtDNA sequences, the concatenated 1111-bp Cyt b and CR sequence data from Jiangsu were aligned with NCBI sequences. One hundred and seventeen unique haplotypes were identified from the data set combining NCBI and 78 individuals, and 128 unique haplotypes were identified from the data set combining NCBI and 173 individuals. Then, 79 unique haplotypes were identified by collapsing the concatenated sequences from 2015 and 2016. For the 117 haplotypes, the best-fit model from Bayesian analysis was the GTR+I+G model. The software Jmodeltest v2.1.7 identified the TrN+I+G model as most appropriate from ML analyses. The GTR+I+G and TIM3+I+G models were identified as most appropriate from BI and ML analyses of 128 haplotypes. The HKY+I+G and TPM2uf+I+G models were suitable from BI and ML analyses of 79 haplotypes, respectively. The two trees were all paraphyletic and consensus was from an overall perspective with credible bootstrap values for all relationships among branches. Dunlins from a stopover in Jiangsu shared haplotypes or formed sister relationships with haplotypes from six subspecies, including C . a. actites , C .a. kistchinski , C . a. sakhalina , C. a. hudsonia , C. a. pacifica , and C. a. arcticola .
Haplotypes from Jiangsu and the breeding areas combined were clustered into two lineages, Alaskan and Beringian lineages, based on the percentage of speculated subspecies, which suggested that dunlins staging in Jiangsu migrated from Beringian and Alaskan areas. The corresponding numbers of individual dunlins are in parentheses in Fig. 2, and the numbers of dunlin individuals at the end of each branch are shown in Tables S1 and S2. Fifty-six individuals (78 in total) were inferred into the two lineages in 2015, and the proportion identified was 71.79%. One hundred and thirty-six individuals (173 in total) were inferred into two lineages in 2016, and the proportion identified was 78.61%. Twenty-two individuals in 2015 and 37 individuals in 2016 could not be identified. The proportion of Alaskan birds increased from 58.97% in 2015 to 63.58% in 2016, and the proportion of Beringian birds increased from 12.08% in 2015 to 15.02% in 2016. Simultaneously, the proportion of unknown birds decreased from 28.21% in 2015 to 21.39% in 2016. Three subspecies, C. a. arcticola , C . a. sakhalina , and C . a. kistchinski , were more common in Jiangsu than other subspecies (Fig. 2). Moreover, distinct from the phylogeny shown in Fig. 2a, there were some haplotypes found only in Jiangsu that were assigned to a single clade of unknown subspecies in Fig. 2b. Notably, some of the dunlin individuals from Alaskan breeding areas detected at Jiangsu had a close phylogenetic relationship with C. a. hudsonia.