Resolution Sensitivity of Rossby Wave Breaking and Warm Conveyor Belts in Global ICON Simulations

Forecast busts over Europe—periods of abnormally low predictive skill—are often associated with extreme weather events and linked to misrepresented upper-level dynamics, including latent heating from mesoscale convective systems (MCSs), Rossby wave breaking, and warm conveyor belt (WCB) outflow. This study investigates how explicitly resolving mesoscale processes affects the simulation of these key mechanisms in global ICON ensemble forecasts at grid spacings ranging from 40 km down to 2.5 km. As a test case, we analyze a forecast bust from ECMWF’s Integrated Forecasting System (IFS) related to the development of Storm Dennis (February 2020), the second-most intense North Atlantic winter storm of the past 150 years, and compare ICON with IFS.

We find a systematic improvement in forecast skill with finer grid spacing. Coarse-resolution simulations reproduce the forecast bust and fail to capture the correct trough–ridge pattern, while convection-permitting simulations more accurately represent upper-level potential vorticity anomalies, WCB structure, and cyclone development.

Our analysis reveals a multi-stage chain of error growth arising from several interacting factors. Large initial-condition uncertainties over the North Pacific provide a background sensitivity, but the strongest early error growth occurs over the central United States, coinciding with a period of deep convection from MCSs. Convection-permitting simulations produce stronger and more coherent MCSs, leading to enhanced negative PV injection near 250 hPa and substantially reduced Rossby wave activity errors. In contrast, coarser-resolution simulations exhibit weaker or misplaced MCSs, resulting in larger errors in the upper-tropospheric flow. These midlatitude convective differences subsequently modulate the intensity and orientation of downstream WCBs over the North Atlantic. The WCB then amplifies the pre-existing errors, linking the central-U.S. convective phase to the eventual European forecast bust.

Overall, our results demonstrate that mesoscale processes over North America—especially MCS-driven PV perturbations—play a key role in setting the predictability of the North Atlantic flow regime during Storm Dennis. Convection-permitting global simulations improve the representation of these processes and offer a physically consistent pathway toward reducing forecast busts in high-impact weather situations. To assess the robustness and generality of these findings, additional case studies are currently being analyzed.