Glacial isostatic adjustment in the Red Sea: Impact of 3-D Earth structure

Sea-level history in the Red Sea region has commonly been interpreted as an accurate proxy for global mean sea level (GMSL), which can be used to constrain global ice volumes and inform a diverse range of regional paleoclimate studies. Previous modeling work has demonstrated, however, that glacial isostatic adjustment (GIA) processes may introduce significant departures from GMSL in this region. The GIA signal is a complex combination of deformational, gravitational, and rotational effects arising from shifting ice and water surface loads and consists of long-wavelength effects superimposed by a short-wavelength signal associated with “continental levering” – a response to local meltwater loading. In this study, we revisit the effects of GIA in the Red Sea region using Earth models characterized by 3-D variations in mantle viscoelastic structure and lithospheric thickness, focusing on the period from Last Glacial Maximum (LGM) to present day. We find that the presence of the Red Sea Rift and low-viscosity upper mantle acts to amplify sea-level rise associated with continental levering, yielding a more pronounced rise for sites toward the center of the Red Sea during deglaciation in comparison to GMSL. In contrast, coastal sites experience a net GIA-induced sea-level fall that acts in opposition to the GMSL rise. Furthermore, while the maximum departure from GMSL occurs close to LGM for northern sites, the GIA signal can peak as late as 14 ka for southern sites. Simulations based on 3-D Earth models tend to show a smaller departure from GMSL than 1-D predictions for most coastal sites, including at the Strait of Bab el-Mandeb at the mouth of the Red Sea, and typically peak at ∼10 m. The opposite is true for sites close to the rift: at these locations the difference between the 1-D and 3-D simulations can reach ∼20 m. We therefore conclude that any mapping from local sea level to GMSL is both location and time dependent and cannot be captured by a simple linear scaling.