Extended Data Fig. 5: Uncertainty in model projections.

Each greenhouse gas emission scenario has large uncertainty in the rate and amount of sea level rise²⁷, as shown for the (a) low RCP2.6, (b) intermediate RCP4.5 and (c) very high RCP 8.5 scenario. This implies for example that within the low scenario, there is a small possibility that cumulative SLR will substantially exceed the SLR of the median projection for the intermediate scenario (95% RCP2.6 = 1.05 m, 50% RCP4.5 = 0.84 m in 2125). (d) This uncertainty in climate projections also leads to large uncertainty in the amount of projected population and habitat loss, as depicted by the error bars reflecting the mean loss for the 5% and 95% percentiles of climate projection within each emission scenario (n = 100 replicate simulations). For example, within RCP4.5 the median projection is 21% habitat loss in 2125, but there is a small (5%) probability that habitat loss will be 60% on these islands under this scenario (a threefold difference). (e) Uncertainty in population and habitat loss projections due to stochastic processes (environmental and demographic stochasticity) is much smaller, as can be seen from the error bars which reflect the 5%-95% confidence intervals around the mean projected population and habitat loss (across the n = 100 replicate model runs for each island and scenario using median climate projections).