SO-CHIC Cheng, L., and Coauthors, 2020: Improved Estimates of Changes in Upper Ocean Salinity and the Hydrological Cycle. J. Climate, 33, 10357–10381

November 5, 2020 Lijing Cheng, Kevin E. Trenberth, Nicolas Gruber, John P. Abraham, John T. Fasullo, Guancheng Li, Michael E. Mann, Xuanming Zhao, and Jiang Zhu

Ocean salinity records the hydrological cycle and its changes, but data scarcity and the large changes in sampling make the reconstructions of long-term salinity changes challenging. Here, we present a new observational estimate of changes in ocean salinity since 1960 from the surface to 2000 m. We overcome some of the inconsistencies present in existing salinity reconstructions by using an interpolation technique that uses information on the spatiotemporal covariability of salinity taken from model simulations. The interpolation technique is comprehensively evaluated using recent Argo-dominated observations through subsample tests. The new product strengthens previous findings that ocean surface and subsurface salinity contrasts have increased (i.e., the existing salinity pattern has amplified). We quantify this contrast by assessing the difference between the salinity in regions of high and low salinity averaged over the top 2000 m, a metric we refer to as SC2000. The increase in SC2000 is highly distinguishable from the sampling error and less affected by interannual variability and sampling error than if this metric was computed just for the surface. SC2000 increased by 1.9% ± 0.6% from 1960 to 1990 and by 3.3% ± 0.4% from 1991 to 2017 (5.2% ± 0.4% for 1960–2017), indicating an acceleration of the pattern amplification in recent decades. Combining this estimate with model simulations, we show that the change in SC2000 since 1960 emerges clearly as an anthropogenic signal from the natural variability. Based on the salinity-contrast metrics and model simulations, we find a water cycle amplification of 2.6% ± 4.4% K⁻¹ since 1960, with the larger error than salinity metric mainly being due to model uncertainty.

Publication DOI: https://doi.org/10.1175/JCLI-D-20-0366.1.