Digital Twins for Medicanes
Bridging Earth Observation and High-Resolution Modelling
Building on the EO-based monitoring and modelling tools developed in Phase 1, MEDICANES Phase 2 introduces Digital Twins for Medicanes — a major step forward in the integration of satellite Earth Observations and high-resolution numerical simulations. Rather than treating EO and models as separate tools, Digital Twins create a deep and continuous synergy between the two, enabling virtual replicas of real medicane events that are directly comparable to satellite imagery.
From Dual Approach to Deep Synergy
The traditional approach in medicane research has relied on a dual strategy: satellite EO systems for observation and numerical models for simulation. Phase 2 goes beyond this by using EO emulators — specifically the RTTOV fast radiative transfer code — to synthesise what a satellite would see if it were observing the output of a high-resolution model run. This makes it possible to directly compare model simulations and real satellite imagery on the same terms, uncovering subtle discrepancies and validating model performance at an unprecedented level of detail.
What the Digital Twins Will Do
The Digital Twin work is organised in four interconnected research tasks:
Numerical simulations using the Meso-NH atmospheric model — run at very high resolution, down to large-eddy simulations at ~100 m horizontal scale — are downscaled from the ECMWF IFS global analysis. The RTTOV radiative transfer code then emulates optical, thermal infrared, and passive microwave satellite observations directly from the 4D model output, including detailed cloud microphysics (ice crystals, snow, graupel, droplets, and rain).
Emulated EO products from the Digital Twins are quantitatively compared with actual satellite observations using standard metrics for brightness temperature and cloud structure. Retrieval algorithms developed in Phase 1 — for warm core detection and convective properties — are applied to both real and emulated imagery, producing comparable diagnostics of cyclone intensity (maximum sustained winds, central pressure) and thermodynamic structure.
SAR (Synthetic Aperture Radar) imagery from Sentinel-1 reveals fine-scale structures in the ocean surface — boundary layer wind rolls, wave patterns, and sea surface temperature gradients — that are invisible to other sensors. Statistical analysis of both SAR backscatter and Digital Twin surface variables will characterise the length scales and directionality of these coherent structures. This task will be conducted partly through a research visit at the ESA Earth System Science Hub at ESRIN (Frascati, Italy), maximising the synergy between EO and modelling expertise.Beyond statistical analysis, this task will also explore the direct emulation of SAR backscatter signatures from Digital Twin surface variables, aiming to produce synthetic SAR-like imagery that can be compared to Sentinel-1 observations at the level of individual wave patterns and wind streaks.
Medicanes are rare — only a handful occur each year. To overcome the limitations imposed by this small sample, the project will generate synthetic Digital Twins: numerical experiments in which intense Mediterranean cyclones that did not become medicanes are nudged towards medicane conditions (e.g. by increasing sea surface temperatures or enhancing latent heat release), producing physically plausible but not yet observed events. These synthetic events serve two concrete purposes: enriching the training datasets for AI-based near-real-time tracking and intensity estimation tools — which currently suffer from the small number of observed medicane cases — and providing physically consistent extreme scenarios as input to the Hazard & Impact Assessment framework, enabling risk analysis beyond the historical record.
Key Outputs
The EO datasets, model outputs, and diagnostics produced within the MEDICANES project are made publicly available through the AERIS data catalogue. Phase 2 outputs will be progressively added to the catalogue as they become available.