Migratory route analysis
Under the Bayesian framework, autumn and spring migration routes were inferred for three individual L. c. superciliosus (Figure 4, Figure S1.4). Full annual data were retrieved from two tags, and similarity was observed between their autumn and spring routes. One tag stopped recording on the continent due to cell failure three months after deployment. Despite uncertainties involved with location estimates, higher mean probabilities for migratory locations were observed among the three individuals both in the western part of the Japanese archipelago and along the CRECS to northern Taiwan. The probability map and the median migratory tracks suggest that birds migrated across the ECS between these two regions and did not detour over the Korean Peninsula. The data from the two successfully retrieved geolocators revealed that both shrikes had wintered in the Sunda Islands, specifically on Java and Flores.
Discussion:
The results from the three approaches supported the LDD hypothesis for genetic divergence of the Brown Shrike in the Japanese archipelago (Figure 1a, 5a-b). A phylogenetic split between the southern and archipelagic clades was inferred in the gene networks (Figure 2), suitable glacial breeding habitat for L. c. superciliosus was estimated to have existed around the CRECS by the SDM (Figure 3), and its current migratory route was observed between the CRECS and the Japanese archipelago across the ECS (Figure 4). Our results cannot reflect gradual expansion of a population via normal dispersal over a terrestrial landscape. This is because, during the glacial periods around which the divergence was estimated, the gulf separating the Japanese archipelago, the Korean Peninsula and the CRECS was relatively wide (Figure 5; Matsuzaki et al., 2019; Ota, 1998), compared to the LGM to which we projected our SDM (Figure 3). LDD colonisation in a stepping-stone manner along the Ryukyu islands from the CRECS or Taiwan may also be less likely. This is supported by our results showing that most of the Ryukyu islands were predicted to have been unsuitable during a glacial period (Figure 3b-d), and the migratory route was not inferred along them (Figure 4), unlike the routes reported for other species (Shiu et al., 2006; Koike, Hijikata, Higuchi, 2016). Meanwhile, since northern Taiwan was connected to the CRECS during this period (Figure 5), the inferred migration aligns with our prediction. Therefore, within our hypothesis-testing framework, we imply that our results reflect the LDD-driven divergence of a Japanese migratory subspecies across the ECS.