Seascape Genomics
Neutral population structure in N. lapillus (N=3,820 loci) was summarized into 11 PCs that explained 100% of the variation in the dataset. The best model describing neutral population structure included two environmental predictor variables (env_PC1: p=0.0104; env_PC2: p=0.0129, Table 2). This model was significant, but only explained a small proportion of variation in the genetic dataset (p=0.0005, R2adj=0.1862; Table 4; Fig. 5). The outlier dataset (N=24 loci) was summarized into five PCs, explaining 87.35% of the variation in the dataset. Here, the best model contained three environmental predictor variables (env_PC1: p=0.0100; env_PC2: p=0.0003; env_PC3: p=0.0019; Table 2) and one geographic spatial variable (dbMEM_2: p=0.0016; Table 2; Fig. 6). This model was also significant, but in contrast to the neutral model, explained a larger proportion of the variation in the genetic dataset (p=0.0001, R2adj=0.5756; Table 4; Fig. 5). Partial RDAs attributed most of this variation to the environmental predictors (R2adj=0.2975, p=0.0022), with a smaller and non-significant proportion attributed to the geographic vector (R2adj=0.0004, p=0.4369).
Neutral population structure in S. umbilicalis (N=11,457 loci), was summarized into 11 PCs that explained 100% of the variation in the dataset. The best model describing neutral population structure included two environmental predictor variables (env_PC2: p=0.0091; env_PC3: p=0.0299; Table 2), two geographic spatial variables (dbMEM_2: p=0.0366; dbMEM_3: p=0.0037; Table 2; Fig. 6), and one larval connectivity vector (AEM5: p=0.0268; Table 2; Fig. 6). This model was significant, but only explained a small proportion of variation in the genetic dataset (p=0.0013, R2adj=0.1739; Table 4; Fig. 7). The largest proportion of explainable variation was attributed to the geographic vectors (R2adj=0.0422, p=0.2020), with smaller proportions attributed to the larval connectivity vector (R2adj=0.0281, p=0.2874) and environmental variables (R2adj=0.0207, p=0.2969), however, none of these partitioned models were significant. The outlier dataset (N=143 loci), was also summarized into 11 PCs, explaining 100% of the variation in the dataset. The best model contained two geographic spatial vectors (dbMEM_2: p=0.0142; dbMEM_3: p=0.0016; Table 2; Fig. 6) and one larval connectivity vector (AEM5: p=0.0356; Table 2). This model was also significant and explained a relatively large proportion of the variation in the genetic dataset (p=0.0009, R2adj= 0.4559; Table 4; Fig. 7). Most of the explainable variation in this model was attributed to geographic vectors (R2adj=0.3940, p=0.0008), with a much smaller and non-significant proportion attributed to the larval dispersal vector (R2adj=0.0459, p=0.1353).