Comparing projected regional sea-level change from CMIP5 against observations

Projections of regional sea-level change based on CMIP simulations are essential for coastal management, adaptation, and mitigation of impacts, as they serve as the foundation for producing policy-relevant reports like those from the IPCC (Intergovernmental Panel on Climate Change). Although evaluating future estimates of sea-level change against observations is important for asserting confidence in projected values, this has only been done for global-mean sea-level change and multiple tide gauge locations. In this study, we evaluate projections of regional sea-level change for the entire range of satellite altimetry measurements. Projections can only be evaluated against observations when a significant length of data has been collected. For instance, the latest IPCC assessment report (AR6) provides projections from 2020, which leaves only five years of data for comparison with recent observations up to 2025. Therefore, we turn to the previous IPCC report (AR5), based on CMIP5, as this allows for a longer overlapping period (2007–2022). We evaluate the IPCC AR5 projections of total sea-level change and its two main components—mass changes due to melting land ice and sterodynamic changes—against satellite altimetry, gravity measurements, and ocean reanalysis products, respectively. Over short periods and on regional scales, the large internal climate variability in observed sea levels obscures the underlying trend, hindering model comparisons. To address this, we use three methods to reduce internal climate variability in the observations before comparing them to the projections: low-frequency component analysis, multivariate regression analysis, and self-organizing maps. Error metrics indicate that low-frequency component analysis performs best at reducing internal climate variability, and we proceed with this method. The projected and observed total sea-level trends agree well for the overlapping period: in 98% of the ocean area, satellite observations fall within one standard deviation of AR5 projections for the middle-of-the-road emissions scenario (RCP4.5). We find that, in general, the AR5 projections underestimate total and mass-driven sea-level change, while they overestimate the sterodynamic component. The largest discrepancies occur in regions with strong ocean currents, such as the Kuroshio Current and the Southern Ocean, where models appear to inadequately capture sea-level trends. Overall, our results provide confidence in future sea-level trends as estimated in AR5, particularly over longer time periods and broad regions.