Med-CORDEX phase 3: seasonal-to-decadal forecast

The Med-CORDEX community has successfully used RCMs to downscale GCM projections. However, while uninitialized projections are valuable for assessing long-term climate evolution, they diverge from observed natural variability, limiting their usefulness for seasonal and decadal adaptation planning. At these timescales, reliable forecasts require predicting both long-term trends and natural variability, which is only possible through initialized forecasts. Climate predictions are currently produced mainly at global, coarse resolution (0.5–1°), limiting their regional relevance. High-resolution RCMs (~0.1°) can bridge this gap by better representing extreme events, particularly in complex regions such as the Mediterranean, where realistic local feedback mechanisms are critical. Dynamical downscaling can also be complemented by less computationally demanding statistical methods and machine-learning tools, which can capture nonlinear processes (Aragão et al. 2026). Extending the Med-CORDEX framework to seasonal-to-decadal forecasts, through a coordinated multi-model downscaling exercise, would improve regional climate information and support decision-making. Three main benefits are expected:

an improved large-scale signal from initialized coupled GCMs driving RCMs,
better representation of local-scale forcing through higher RCM resolution,
added value from the RCM initialization compared with non-initialized RCMs.

This combined improvement could be crucial for representing weather–climate extremes, where large-scale forcing is modulated by local processes while initialized GCMs capture “in-phase” natural variability.

Points of Contact: Silvio Gualdi (CMCC), Javier Garcia-Serrano (U. Barcelona)