Reconstructing historical salinity fields over the 20th century using a data-driven method and Argo data

Ocean salinity is a fundamental variable that determines seawater density and, therefore, stratification and oceanic dynamics. To understand how salinity is affected and how it contributes to ocean processes, its variability must be studied, particularly through in situ observations. Unfortunately, while temperature observations were limited during the 20th century, salinity observations were even sparser, as some instruments were designed to measure temperature only. The development of the Argo observing system since 2002 has improved sampling and reduced the disparity between both variables, enabling better assessment of salinity variability over the past 20 years at interannual to decadal scales. In this study, we estimate salinity covariability with temperature from the Argo period to reconstruct monthly subsurface salinity fields, in the tropical Pacific between 1930 and 2001, leveraging temperature observations. The analysis is performed using the data-driven RedAnDA method, which combines Data Assimilation, Analog Prediction, and Reduced-space Interpolation, first validated using synthetic data. We reconstruct the 20th century interannual variability of salinity associated with ENSO events both at the surface and in the subsurface. Notably, thanks to the coupling with temperature, the representation of stratification and its modulation by vertical salinity gradients is enhanced. This new method and product provide for the first time the possibility to extend the hydrological time series consistently in the past, offering potential new insights into mechanisms generating decadal variability in the Pacific.