WP 2 Case studies towards AI benchmarks

This work package uses two case studies to test the AISIT database in real-world scientific applications and to improve its readiness for use in AI-based research.

2.1 Svalbard case study

The Svalbard region provides an ideal focus area due to its rich, well-documented freshwater datasets and long history of research. Data from recent field campaigns in and around Ny-Ålesund, combined with historical observations, include measurements of freshwater tracers alongside key environmental information such as temperature and salinity.

This case study will compare traditional and AI-based approaches to mapping freshwater inputs from Svalbard and distributions in the Barents Sea. Feedback from this work will help refine the structure, metadata, and usability of the AISIT database.

2.2 Pan-Arctic

To understand freshwater patterns across the entire Arctic, this case study will use existing Arctic-wide datasets and model outputs to reconstruct large-scale freshwater distributions. By combining tracer relationships with ocean circulation information, we will explore how well AI methods can capture Arctic-wide freshwater dynamics.

Lessons learned from this work will guide further improvements to the AISIT database, particularly in how large-scale and dynamical information is represented.