Deep Learning-Based Temperature and Salinity Reconstruction and derived Geostrophic Currents in the Pan-Arctic Ocean with 3-Day resolution on EASE-Grid 2.0 at 25km (Version 1)

This dataset collection provides observation-constrained reconstructions of Arctic Ocean hydrography and diagnostic geostrophic circulation for the period 2011–2021. The datasets were generated using a Long Short-Term Memory neural network trained with in situ hydrographic profiles and satellite-derived surface information, including sea surface temperature, sea surface salinity, and absolute dynamic topography. The framework estimates temperature and salinity anomalies relative to GLORYS12V1 reanalysis, which are then used to obtain reconstructed three-dimensional temperature and salinity fields on 102 WOA standard depth levels at 3-day temporal resolution. The collection includes one pan-Arctic product on a 25 km EASE2 grid and four regional 6.25 km products covering the main Arctic gateways: Bering Strait, Davis Strait, Fram Strait, and the Barents Sea Opening. The datasets also include model-based uncertainty estimates, surface input fields and absolute dynamic height, supporting studies of Arctic freshwater variability, circulation change, and climate model evaluation.