Med-CORDEX phase 3: AI-based empirical and statistical downscaling

In the early 2010s, Med-CORDEX was established to advance regional climate modelling. At that time, statistical downscaling was developed alongside dynamical downscaling. The rapid emergence of AI in the early 2020s has renewed interest in statistical downscaling.

In 2023, the Med-CORDEX Empirical Statistical Downscaling team convened to develop innovative methods based on emerging AI technologies. Within the community and following the recent recommendations developed in Chun et al. (2026), the key objective for the next five years is to develop and operationalize advanced AI-driven methods grounded in empirical scaling relationships related to coupled oceanic and atmospheric processes.

The envisaged coordinated actions include:

Organizing and maintaining the Med-CORDEX simulation catalog to integrate with downscaling workflows.
Developing new AI approaches for linking statistical downscaling outputs by creating prototype models.
Designing and implementing algorithms for non-stationary signal extraction for impact study applications (e.g., agriculture, water resources).

The team is also investigating AI infrastructures better suited for probabilistic approaches and exploring unique skill sets that combine empirical scaling with modern AI hardware such as GPUs and TPUs.

Points of Contact: Sun Kwok (UWE), Leonardo Aragão (CMCC), Matias Ezequiel Olmo (BSC), Christoforus Bayu Risanto (Vatican Observatory)