Impact of Rain-Adjusted Satellite Sea Surface Salinity on ENSO Predictions From the GMAO S2S Forecast System

Abstract

Previous research has shown that assimilating satellite sea surface salinity (SSS) has improved initialization of coupled El Niño/Southern Oscillation (ENSO) forecasts. However, most of these assimilation techniques have either removed the freshwater bias by correcting to monthly mean fields of subsurface observations or ignored it altogether. In this paper, we explore the impact of accounting for the satellite SSS fresh bias by first estimating, then removing the near-surface salinity gradient from the satellite SSS using the Rain Impact Model (RIM [Santos-Garcia et al., 2014, https://doi.org/10.1002/2014jc010137]). This diffusivity model is calculated using collocated satellite rainfall and SSS estimates. Two ocean reanalyses are produced, one assimilating RIM data, which removes the fresh bias at the surface (SSS_RIM), and the other experiment retains this bias (CONTROL). Both reanalyses additionally assimilate all conventional ocean observations. Comparison of SSS_RIM versus CONTROL shows that the thermocline is deeper for the SSS_RIM, allowing this reanalysis to store more heat. Removing the fresh bias destabilizes the water column for the SSS_RIM experiment, allowing enhanced mixing, and more heat storage. ENSO forecasts initiated from April reanalyses from 2015 to 2021 are consistently warmer for SSS_RIM than for the CONTROL. For all but one instance (2017), these SSS_RIM forecasts are closer to observations than the CONTROL. These results argue that operational coupled forecast centers should reevaluate bias-correcting the satellite SSS using monthly gridded fields of in situ salinity, but rather they should utilize observed rainfall to estimate coincident near surface salinity gradients.