Climate Modeling of TRAPPIST-1e with a Global Storm Resolving model: high resolution for habitability studies

Recent advances have made kilometer-scale Earth climate modeling possible, yet global exoplanet simulations still rely on coarse resolution models (>100 km) that require explicit parametrization of convection and clouds that introduce significant uncertainties. These parameters are critical for tidally locked planets, where we can only observe the terminator: the boundary between the day and night side. The presence or absence of high clouds at this location can determine our ability to characterize a planet’s atmosphere [1].

A study with ICON-Sapphire, an Earth Global Storm Resolving model exhibited a convergence of different climate statistics when reaching kilometer scale on an aquaplanet [2]. Resolution was shown to strongly impact the tropical circulation and humidity as well as the amount of cloud liquid water and ice content in the simulated climate. In this study we look at the changes that kilometer scale modeling brings to a tidally locked exoplanet.

We focus on TRAPPIST-1e, a rocky planet slightly smaller than Earth orbiting in the habitable zone of an ultra-cool red dwarf star 40 light-years away. We simulate its atmosphere at 5 km horizontal resolution using ICON-Sapphire [3]. To do so, we adapted the model to reflect TRAPPIST‑1e’s size, rotation, stellar irradiation, and an idealized

atmospheric composition consistent with the THAI model intercomparison project [4]. To ensure long-term stability, we incorporated other modifications, including the use of artificial ozone heating to stabilize the temperature of the stratosphere.

We examine how planetary parameters shape the simulated climate of a tidally locked exoplanet, with emphasis on high clouds at the terminator. Comparing our convection-resolving simulation with lower resolution simulations from the existing literature and our own experiments, we assess how kilometer-scale modeling alters atmospheric circulation and cloud processes. We also look at the general energy balance of the planet since we know from aquaplanet simulations that kilometer scale resolution can influence it through albedo due to changes in the cloud cover.

Development and model spin-up were done at around 100 km resolution on the VSC-5 system of Austrian Scientific Computing (ASC) running on 20 to 48 CPU nodes at a decadal timescale to reach a steady atmospheric state from which to start the high resolution simulations. The computationally demanding kilometer-scale simulations are being executed on GPUs on Leonardo, the pre-exascale EuroHPC supercomputer, hosted by CINECA. To facilitate this cross-platform development and ensure performance portability, we developed and deployed containerized versions of ICON, enabling seamless compilation and execution on diverse CPU and GPU architectures.

This work highlights the potential of high resolution exoplanet climate modeling to help refine the interpretation of future observational data and shows how an already existing complex earth system model can be reshaped and used for new applications with a relatively low development effort.

[1] Komacek, Thaddeus D., Thomas J. Fauchez, Eric T. Wolf, and Dorian S. Abbot. The Astrophysical Journal Letters 888, no. 2 (2020).

[2] Peinado Bravo, A., D. Klocke, and B. Stevens. Journal of Advances in Modeling Earth Systems 18 (2): (2026).

[3] Hohenegger, Cathy, Peter Korn, Leonidas Linardakis, et al. Geoscientific Model Development 16 (2): 779–811 (2023).

[4] Fauchez, Thomas J., Martin Turbet, Eric T. Wolf, et al. Geoscientific Model Development 13 (2): 707–16. (2020).