ERA5-based present-day ICON simulations with variable polar resolution

Project: PolarRES: Global variable-resolution ICON simulations with polar refinements - The EU Horizon 2020 project PolarRES (https://polarres.eu) aims to improve regional climate information to support impact assessments in the Arctic and Antarctic. To this end, high-resolution climate projections are performed based on physically plausible storyline scenarios. A key objective of the project is to assess the impact of enhanced spatial resolution in polar regions for both present-day and future climates using global variable-resolution climate models. This modelling strategy enables a process-based quantification of how increased polar resolution affects large-scale atmospheric circulation and linkages between polar regions and lower latitudes.

Therefore, within this project we conducted bipolar variable-resolution atmosphere-only simulations using the ICON (Icosahedral Nonhydrostatic) model. The simulations were conducted as part of PolarRES WP2 and were supported by the European Union’s Horizon 2020 research and innovation programme under grant agreement No. 101003590.
This work utilised computational resources provided by the Deutsches Klimarechenzentrum (DKRZ) through allocations granted via the Alfred Wegener Institute (AWI) as part of the Helmholtz shareholder resources budget, as well as the AWI high-performance computing facility Albedo. Additionally, ERA5 reanalysis data were accessed through the DKRZ data pool, made available by DKRZ Data Management.
Summary: To evaluate the added value of regional refinement in the ICON model, this experiment focuses on present-day (PD) climate simulations using ERA5 boundary conditions. These simulations serve as a baseline for assessing how increased spatial resolution over polar regions affects the representation of key climate processes, large-scale circulation, and polar–midlatitude linkages. Comparing the uniform and refined configurations enables a systematic evaluation of the performance, internal consistency, and potential improvements introduced by two-way nested domains in a controlled present-day setup.

This experiment comprises three atmosphere-only simulations conducted with version 2.6.6 of the ICON model as part of WP2 of the EU Horizon 2020 project PolarRES. All simulations follow an AMIP-style setup and are forced with boundary conditions from ERA5, representing present-day climate conditions. The simulations differ only in their horizontal resolution over the polar regions.

PD_ERA_UNREF was performed using a globally uniform R3B5 grid, corresponding to a horizontal resolution of approximately 52.6 km. The two refined simulations, PD_ERA_REF_ARCTIC and PD_ERA_REF_ANTARC, are based on the same R3B5 base grid but include two nested domains over the Arctic and Antarctic, respectively. Each includes an intermediate-resolution nest (R3B6, ~26.3 km) north or south of 50° latitude, and a high-resolution nest (R3B7, ~13.2 km) beyond 57°N or 57°S. Two-way nesting was used to allow feedback from the refined domains to the global domain. All simulations used 90 vertical levels with a model top at approximately 75 km. The time step was halved with each nesting level to ensure numerical stability. All physical parameterisations were applied consistently across domains and chosen to perform robustly across the resolution range.

The simulations were initialised on 1 January 1984 using ERA5 data and integrated for 31 years. The first year is discarded as spin-up, yielding 30 years of output (1985–2014). Sea surface temperatures and sea ice concentrations were prescribed daily from ERA5 for the period 1984–2014. While sea ice concentration is prescribed, the model calculates sea ice thickness prognostically. Time-varying greenhouse gas concentrations (CO₂, CH₄, N₂O, and CFCs) follow CMIP6 historical forcings.

The ICON dynamical core is based on Zängl et al. (2015) and the model code is available via https://www.icon-model.org/ (release note for version 2.6.6: https://gitlab.dkrz.de/icon/icon-model/-/blob/release-2024.07-public/RELEASE_NOTES.md?ref_type=heads). All three simulations used the ecRad radiation scheme (Hogan et al., 2018), a single-moment cloud microphysics scheme following Doms et al. (2011) and Seifert (2008), and a convection scheme based on Tiedtke (1989) and Bechtold et al. (2008). Turbulent processes are represented by a prognostic TKE-based turbulence scheme (Raschendorfer, 2001). Orographic drag is parameterised following Lott and Miller (1997), and non-orographic gravity wave drag is based on Orr et al. (2010). Parameter settings for subgrid-scale orographic and non-orographic gravity wave drag were guided by Köhler et al. (2021), with adaptations to our model resolution. The land surface is represented using the TERRA component for soil-vegetation-atmosphere transfer (Schrodin and Heise, 2001), with topography derived from the GLOBE dataset (Hastings et al., 1999) at ~1 km native resolution.

Post-processing included horizontal interpolation of selected variables to a regular 0.5° × 0.5° lat-lon grid for the global domain (UNREF), and to 0.125° × 0.125° for the high-resolution refined domains. Vertical interpolation to 19 pressure levels was applied to data on native model levels.