Coral-Based Sea Surface Salinity Reconstructions and the Role of Observational Uncertainties in Inferred Variability and Trends

Climate observations in much of the tropical oceans are scarce during most of the 20th century, so paleoclimate proxies are needed to understand the full range of natural climate variability. Past proxy studies have focused primarily on sea surface temperatures, but there are comparatively few salinity reconstructions. Such reconstructions can extend our understanding of hydroclimate across the tropical oceans, including variability in precipitation, evaporation, and ocean circulation. Here we compile a network of salinity-sensitive coral delta O-18 records, then apply a reduced-space method based on empirical orthogonal function analysis to reconstruct annual tropical salinity anomalies over the 20th century. A comparison of surface salinity data sets, including reanalyzes (SODA2/3, Ocean ReAnalysis System 5 (ORAS5), Global Ocean Data Assimilation System) and objective analyses (Institute of Atmospheric Physics (IAP), EN4, Delcroix), show large discrepancies in the spatial structure, temporal evolution, and importance of the leading modes of variability. Two salinity data sets, IAP and ORAS5, are retained for climate reconstruction. Our coral-based salinity reconstructions reveal significant long-term trends over the 20th century, which are likely associated with hydrological cycle intensification and possibly a weakening of the Walker Circulation. These reconstructions also capture the spatial and temporal patterns of salinity anomalies associated with the El Nino-Southern Oscillation, Interdecadal Pacific Oscillation, and Atlantic Multidecadal Oscillation. Ultimately, this approach can enhance our understanding of tropical hydroclimate prior to the observational era.