Artificial Intelligence for Stable Isotope Tracers (AISIT): a database of δ18O and ancillary data in the Pan-Arctic [in review]

The Arctic is undergoing rapid change. Near-surface air temperatures are rising at around four times the global average, ocean temperatures are increasing, sea ice is thinning, retreating rapidly, and growing younger on average, while pan-arctic river run-off and glacial melt are both increasing. The increased fluxes of nutrient-loaded freshwater into the Arctic significantly impact marine systems both locally and globally. Arctic nutrient supplies underpin marine productivity not only within the Arctic Ocean, but also in temperate and tropical regions via oceanic exports. This makes the tracing of freshwater sources from land, the cryosphere and the atmosphere into and throughout the ocean a critical priority for understanding the global consequences of Arctic climate change. When combined with salinity measurements, stable oxygen isotope ratios (δ18O) in seawater are a powerful tracer for distinguishing freshwater contributions from meteoric sources (precipitation, rivers, glaciers) and sea-ice. While major previous efforts have compiled extensive δ18O observations in other regions, no up-to-date, centralised, accessible, and curated pan-Arctic compilation exists, despite the rapid growth in observational coverage over the recent decades. Here, we present a unified pan-Arctic seawater δ18O database comprising observations collected north of 60 °N, compiled from public repositories, peer-reviewed literature, and datasets contributed directly by colleagues across the scientific community. To unify these records, we harmonised metadata, applied standardised quality-control procedures, and integrated records into a single Artificial Intelligence (AI) compatible dataset adhering to FAIR (Findable, Accessible, Interoperable, and Reusable) data principles. We further added available barium (Ba) concentration data as another freshwater tracer, capable in some conditions of further de-convolving Arctic river run-off into separate sources. The resulting database comprises 45,714 unique δ18O measurements, paired with salinity measurements for non-freshwater samples, and 5531 unique Ba measurements spanning 1967–2025 (Thorpe-Morgan et al., 2026). This dataset provides a useful resource for quantifying sources of freshwater in the Arctic Ocean, in a format which facilitates data-driven and machine-learning approaches to spatio-temporally-sparse environmental datasets.