Supporting information
Figure S1 Sampling sites of Syngnathus schlegeli along the coastline of China (QD=Qingdao; HZW=Haizhouwan; ZS=Zhoushan; ZH=Zhuhai; HK=Haikou). Background gradient of color indicate water depth.
Figure S2 Presence records of Syngnathus schlegeli used in this study. All distribution data (black points, red points, blue points, and green points) were used to develop species distribution model. Records in red represent Yellow Sea clade (56 presence records), records in blue correspond to East China Sea clade (23 records), and records in green indicate South China Sea clade (29 records). Presence records of S. schlegeli along the coastal waters of mainland China were divided into three distinct clades based on the present genetic cluster results and previous phylogeographic studies conducted in the same region (Donget al. 2012; Liu et al. 2007; Ni et al. 2012; Xuet al. 2009).
Figure S3 Multi-collinearity analysis results between eleven marine predictor variables. bathymetry: depth of the seafloor; bio8: mean annual SSS (psu); bio9: SSS of the freshest month (psu); bio10: SSS of the saltiest month (psu); bio11: annual range in SSS (psu); bio12: annual variance in SSS (psu); bio13: mean annual SST (°C); bio14: SST of the coldest month (°C); bio15: SST of the warmest month (°C); bio16: annual range in SST (°C); bio17: annual variance in SST (°C). SSS presents sea surface salinity and SST means sea surface temperature. Predictors in same box are highly correlated (i.e. pairwise Pearson’s correlation coefficients |r | > 0.7) and only one was chosen for subsequent analyses.
Figure S4 True skill statistic (TSS) (a) and the area under the receiver operating characteristic curve (AUC) (b) values of ten modelling algorithms. Dashed lines represent cutoff values for TSS (0.75) and AUC (0.90). ANN: artificial neural network; CTA: classification tree analysis; FDA: flexible discriminant analysis; GAM: generalized additive model; GBM: generalized boosting model; GLM: generalized linear model; MARS: multiple adaptive regression splines; Maxent: maximum entropy; RF: random forest; SRE: surface range envelop. Each algorithm was run ten times and results are expressed as mean ± standard error.
Figure S5 Continuous predictions of habitat suitability forSyngnathus schlegeli under present-day (a) and Last Glacial Maximum (LGM) (b, c, d) climate conditions. CCSM3, ECBilt-Clio, and HadCM3M2 indicate different general circulation models used to simulate paleoclimates.
Table S1 References of distribution range of Syngnathus schlegeli
Table S2 Singular values and percent explained for relative warps based on the geometric morphometric data
Table S3 Statistics of properties of the RAD sequences after filtering
Table S4 Statistics of the depth and coverage of reads after mapping to the reference genome
Table S5 Statistics of SNPs after filtering
Table S6 Population pairwise F ST between different S. schlegeli populaitons base on the cytb andsnx33 gene
Table S7 Statistics of selected genes detected by selection sweep