[21-Feb-2024] Hall, S. and Bayler, E.
Presented at the 2024 Ocean Sciences MeetingSalinity, a fundamental ocean state parameter, governs the Arctic Ocean's upper density structure, influencing physical dynamics, sea-ice cover, and convection processes. The paucity of in situ salinity measurements, however, limits comprehensive understanding of the Arctic Ocean state and processes. Satellite-based salinity observations extend spatiotemporal coverage, but those retrievals sample only the top few centimeters ('skin surface') and ice-free regions. In addition to land, dispersed sea-ice, and ice-edge contamination due to low spatial resolution, satellite L-band sea-surface salinity measurements are adversely affected by the cold temperatures of the polar ocean, which reduces the salinity signal-to-noise ratio. This study employs salinity observations from several in situ (0-5m) sources, satellite salinity observations, and an independent objectively analyzed air-sea forcing product for ancillary information to convert satellite-derived skin salinity to near-surface bulk salinity using a machine-learning-based approach, the Generalized Additive Model (GAM). The satellite salinity observations are provided from the European Space Agency's Soil Moisture – Ocean Salinity (SMOS) mission and the National Aeronautics and Space Agency's Soil Moisture Active Passive (SMAP) mission. Major steps of this methodology include: co-locating satellite and in situ measurements within 50 km and 3.5 days; computing the salinity biases and characterizing their statistics; then training the GAM with in situ salinity to convert the satellite-based skin salinity observations to bulk salinities. For assessing the fidelity of this bulk-salinity computation methodology and algorithm to compute bulk salinity, this study characterizes the statistics of the skin-effect and bias-corrected near-surface salinities with co-located derived near-surface bulk salinities and in situ observations to demonstrate improvements. This research addresses satellite salinity high-latitude retrieval biases, enables the assimilation of those high-latitude satellite salinity observations into numerical modeling, and contributes to validating, verifying, and operationalizing the National Oceanographic and Atmospheric Administration's Unified Forecast System's global coupled model.