Fig 1.
(A) Locations of sampled populations where mean midsummer month sea surface temperature differed by up to ~3°C. (B) Principal component analysis of water quality and temperature parameters at the sampled locations. Winter.T—10% quantile of winter temperature, Summer.T– 90% quantile of summer temperature, Daily.T– 90% quantile of daily temperature range, Phos–total dissolved phosphorus, Chl–chlorophyll, NO3 –nitrate, Secchi–Secchi depth (water clarity). Locations are colored according to summer temperature as in panel A. (C) Principal component analysis of genome-wide genetic variation (inset–Acropora millepora). Centroid labels are initial letters of population names as in panel A. (D) ADMIXTURE plot of ancestry proportions with K = 2 (the lowest cross-validation error was observed with K = 1). Analyses on panels C and D were based on 11,426 SNPs spaced at least 2.5 kb apart and not including FST outliers.
Fig 2.
Estimated demography of A. millepora populations on the GBR.
(A) Arc-plot of migration rates among populations reconstructed from population genomic data. Inset: ∂a∂i model used: ancestral population splits into two populations of unequal sizes (N1 and N2) some time T in the past, these populations exchange migrants at different rates depending on direction. (B) Migration rates according to the biophysical model. On panels A and B, the arcs should be read clockwise to tell the direction of migration; line thickness is proportional to the migration rate. (C) Correlation between log-transformed biophysical and genetic migration rates (Mantel r = 0.58, P = 0.05). (D) Box plot of effective population sizes inferred by the split-with-migration model (panel A) across all population pairs and bootstrap replicates. (E) Historical effective population sizes inferred by stairwayPlot for the Keppel population and pooled Sudbury, Orpheus and Magnetic populations (GBR). The line is median of 200 bootstrap replicates, light shaded area is 95% credible interval, dark-shaded area is 75% credible interval.
Fig 3.
Modeling coral metapopulation persistence under warming.
(A, C, E): Mean fitness, relative to maximum attainable with perfect heritability. (B, D, F): Mean phenotype (thick lines) and modeled temperatures (thin noisy lines). (A, B): Settings for the most efficient selection (perfect heritability, narrow tolerance). (C, D): Settings for the least efficient selection (low heritability, broad tolerance). (E, F): Intermediate heritability and tolerance settings (Esd = 1, σ = 1) with no migration. Warm-adapted populations (W and M) are shown as red-tint traces, populations from mild thermal regime (S and O) are green-tint traces, and the cool-adapted population (K) are the blue traces. Note close similarity between traces for pairs of populations pre-adapted to the same temperature (W, M and S, O). (G) Sensitivity of populations to random thermal anomalies increases under warming. Modeled temperature anomalies are shown as grey line, fluctuations in populations’ fitness–as colored lines (residuals from loess regression over fitness traces at Esd = 1, σ = 1; Wilkie: orange line, Keppel: blue line). The sign of temperature anomalies is inverted to better reveal the correspondence between rise in temperature and drop in fitness. Mutation rate was 1e-6 per locus per gamete in all simulations shown.
Fig 4.
Effect of mutation rate on population persistence.
(A, C, E): Mean fitness, relative to maximum attainable with perfect heritability. (B, D, F): Mean phenotype (thick lines) and modeled temperatures (thin noisy lines). Mutation rate (mu) per locus per gamete is listed above the graphs; effect sizes of new mutations were drawn from a normal distribution with mean 0 and standard deviation 0.2°C. Adaptation to local thermal conditions and initial adaptive response based on genetic rescue happen efficiently even under low mutation rate (1e-7), but further evolution is only possible at high mutation rate (1e-5). All simulations shown share intermediate selection efficiency settings: Esd = 1, σ = 1.
Fig 5.
Larger population size and finer genetic architecture facilitate population persistence under warming.
(A, C, E, G): Mean fitness, relative to maximum attainable with perfect heritability. (B, D, F, H): Mean phenotype (thick lines) and modeled temperatures (thin noisy lines). Number of QTLs (N qtl) and population sizes (Ne) are listed above graphs (K population size is five-fold smaller in all cases). With 100 QTLs, their effect sizes are proportionally smaller to enable the same total genetic variance as with 10 QTLs. Both larger population size (C, D) and finer genetic architecture (E, F) improve population persistence, and combination of the two might enable populations to adapt indefinitely (G, H). All simulations shown share intermediate selection efficiency settings: Esd = 1, σ = 1, and mu = 1e-6.