Integration of population genetics with oceanographic models reveals strong connectivity among coral reefs across Seychelles

Many countries with tropical reef systems face hard choices preserving coral reefs in the face of climate change on limited budgets. One approach to maximising regional reef resilience is targeting management efforts and resources at reefs that export large numbers of larvae to other reefs. However, this requires reef connectivity to be quantified. To map coral connectivity in the Seychelles reef system we carried out a population genomic study of the Porites lutea species complex using 241 sequenced colonies from multiple islands. To identify oceanographic drivers of this connectivity and quantify variability, we further used a 2 km resolution regional ocean simulation coupled with a larval dispersal model to predict the flow of coral larvae between reef sites. Patterns of admixture and gene flow are broadly supported by model predictions, but the realised connectivity is greater than that predicted from model simulations. Both methods detected a biogeographic dispersal barrier between the Inner and Outer Islands of Seychelles. However, this barrier is permeable and substantial larval transport is possible across Seychelles, particularly for one of two putative species found in our genomic study. The broad agreement between predicted connectivity and observed genetic patterns supports the use of such larval dispersal simulations in reef system management in Seychelles and the wider region.

included in initial analysis (excluding the outgroup identified on tree which had a higher divergence (>=5%) to the genome reference than the rest (~1-2% divergence)): A) Admixture plots for K=2 & 3, mean cross-validation error for 100 runs of ADMIXTURE: K2= 0.237; K3= 0.242; B) Principal component analysis of genetic variation in Porites cf.lutea, the variance explained by PC1 and PC2 is 1.98% and 1.11% respectively.Sites are coloured lightest in the south, darkest in the northern Seychelles.
Figure S3: Mean transport probabilities between island groups across the seasonal cycle.Here, we are referring to the Aldabra and Farquhar Groups as the 'Outer Islands'; the Amirante Islands and Southern Coral Group as 'Amirantes', and the islands on the Seychelles Plateau as the 'Inner Islands'.
Figure S4: Left: Time-mean potential connectivity between pairs of reef groups in Seychelles, expressed as transport probability (the y and x axes respectively indicate source and destination sites).Right: Implicit connectivity between pairs of reef groups in Seychelles, expressed as the median number of generations of dispersal separating pairs of reef groups (i.e. the median number of generations required for the backward cumulated implicit connectivity to exceed 0.5).Labels refer to island groups: Inner (Inner Islands), SCG (Southern Coral Group), AMI (Amirante Islands), ALP (Alphonse Group), FAR (Farquhar Group), and ALD (Aldabra Group).

Figure S1 :
Figure S1: The percentage of SNPs for each sample that were different from the reference genome, for: a) All 241 samples and b) excluding the main outgroup which has >5% divergence from the reference genome.Samples were grouped based on the phylogenetic clade they fell into in Fig 2a.Calculation for each point: (Total snps -snps that were low coverage or heterozygous)*the number of genome reference sites.

Figure S2 :
Figure S2: Population structure for the 220 samples(182, included in initial analysis (excluding the outgroup identified on tree which had a higher divergence (>=5%) to the genome reference than the rest (~1-2% divergence)): A) Admixture plots for K=2 & 3, mean cross-validation error for 100 runs of ADMIXTURE: K2= 0.237; K3= 0.242; B) Principal component analysis of genetic variation in Porites cf.lutea, the variance explained by PC1 and PC2 is 1.98% and 1.11% respectively.Sites are coloured lightest in the south, darkest in the northern Seychelles.

Figure S5 :
Figure S5: PCA to identify consistent clusters of reefs identified through Infomap.Points represent reef groups, and are coloured according to PC1.Reef groups that cluster together in PC1-PC2 space are frequently identified by Infomap as belonging to the same module.

Figure S7 :
Figure S7: As in Figure S6, but based on parameters for A valida.

Figure S8 :
Figure S8: Comparison of gene flow estimates from BayesAss analysis for the X clade, using the same sample set but with different sized groupings.Original analysis separated the specific sampling sites at Aldabra (a & c) but the results presented in this study use the broader grouping (b & d).

Figure S9 :
Figure S9: Gene flow estimates from BayesAss analysis for the full 220 sample set (A&D), the X clade (B&E) and the Y clade (C&F).Data for these plots is available here.

Table S1 :
Sample sites and quantities.241 samples were sequenced, 21 of these were not included in analysis because they had significant divergence from the reference genome.

Table S2 :
Pairwise Fst calculated for the X clade using R Package HIEFSTAT (colourmap with green being the relative highest Fst, red being relative lowest)

Table S3 :
Pairwise Fst calculated for the Y clade using R Package HIEFSTAT (colourmap with green being the relative highest Fst, red being relative lowest)

Table S4 :
Sample groupings for BayesASS analysis based on the 130 blue clade samples only and presented in Figure3of this study (Alphonse data not plotted due to low sample size).