High-Resolution Climate Modeling
Much of the uncertainty in climate projections stems from the need to empirically parameterize processes on scales smaller than can be resolved explicitly in global climate models, such as clouds, convection, and turbulence. Recent advances in computational resources and modeling now allow simulations at higher resolutions than were previously possible, such that mesoscale processes (at horizonal scales of 10-100 km) can be represented in global models. These new high-resolution models provide an unprecedented opportunity to understand the two-way interactions between mesoscale processes, which are important in the dynamics of weather systems, and the large-scale climate system. Our research uses a hierarchical modeling approach across a range of resolutions and configurations in the CESM and ICON climate models to understand (1) the role of frontal processes in atmosphere-ocean coupling in midlatitudes and the potential for improved climate predictions from higher resolution models that resolve these processes and (2) differences in tropical climate and its response to external forcings between high-resolution configurations with resolved deep convection and coarse-resolution configurations with parameterized deep convection. Our overarching goals are both to understand the fundamental dynamics of cross-scale interactions between the mesoscale and global scale and to develop new modeling and machine learning methods to bring the added value of representing mesoscale processes into coarse-resolution global models, which are still needed to broadly investigate climate change scenarios.
We run high-resolution climate simulations with the ICON model on the GPU partition of the CSCS supercomputer Alps. The corresponding research was recently highlighted in a BBC documentary about the Alps supercomputer. external page [Watch on YouTube]
Current Projects
The pattern effect on global climate with explicit versus parameterized deep convection - Clarissa Kroll, Robb Jnglin Wills
The equilibrium climate sensitivity (ECS), the warming following a doubling in CO2 concentrations, is the cornerstone upon which climate mitigation and adaption strategies are built. However, its true value is highly uncertain, primarily due to uncertainty in the cloud feedback. The atmospheric response to the "pattern effect", the spatial pattern of ocean surface warming, is a substantial source of uncertainty in this regard, as the convection- and gravity-wave-mediated propagation of temperature anomalies from near the surface in convective regions into the upper atmosphere has strong impacts on global cloud cover. The pattern effect is thus important for setting the global radiative feedback and ECS. Accurately modeling the ocean warming pattern and the atmospheric response to this warming pattern is key for accurate projections of the climate response to climate change scenarios. In this project, we compare the simulated atmospheric response to key patterns of ocean temperature anomalies in low- and high-resolution ICON simulations. Specifically, we investigate the upscale influence of the newly resolved small-scale processes on the large-scale circulation and climate in the high-resolution simulations and the corresponding influence on the simulation of the pattern effect compared to the low-resolution simulations. The pattern effect is an ideal test case for how increased resolution can improve the simulation of climate change, because it can benefit from explicitly resolved deep convection.
Finished Projects
The role of subtropical moisture transport in shaping precipitation biases in the Intertropical Convergence Zone — Clarissa Kroll, Robb Jnglin Wills
In external page Kroll et al. 2025, we investigate the physical mechanisms underlying the formation of the double Intertropical Convergence Zone, a persistent precipitation bias observed across multiple generations of climate models, with a particular focus on the ICON model. By examining the moisture budget and comparing simulations ranging from conventional coarse resolutions to 5 km grids, we identify a key physical mechanism: ICON transports excessive moisture vertically out of the subtropical boundary layer before it can reach the equator. This premature vertical export is driven by the model’s parameterized turbulence and shallow convection schemes, which account for processes that remain unresolved even at kilometer-scale resolution.
Because the overexpressed vertical moisture transport reduces near-surface moisture entering the tropics, the tropics become too dry, deep convection weakens, and the large-scale Walker circulation slows down. The resulting circulation change allows spurious deep convection to occur over the Eastern Pacific, producing the spurious “double-ITCZ" rain band. Importantly, our analysis shows that the overexpressed moisture export from the boundary layer persists at all tested resolutions and remains active even when deep convection is switched off at 5 km.
We further tested a previously proposed tuning approach that increases evaporation to address the tropical moisture and precipitation biases. While this improves the rainfall pattern, it also alters trade-wind strength, moisture source regions, and the global circulation, illustrating how treating symptoms rather than the root cause can degrade physical consistency. Together, these findings provide a physically grounded explanation of the double-ITCZ bias in ICON and point to concrete pathways for improving climate models by using combined turbulence and shallow convection parameterizations for future model generations.
The impact of lower stratospheric heating perturbations on atmospheric moisture and deep convection - Clarissa Kroll
Volcanic eruptions lead to heating perturbations in the tropopause and lower stratosphere, as would geoengineering relying on stratospheric aerosol injection. This has impacts on moisture transport into the stratosphere both in the form of water vapor and in the form of frozen moisture originating from deep convection. Whereas the water vapor transport is controlled by the larger-scale tropical cold-point temperatures, and therefore sufficiently described by conventional coarse-scale simulations, a realistic description of frozen moisture transport by deep convection is only achieved with km-scale simulations. We study both moisture transport pathways using a hybrid approach: For the study of water vapor entry values we rely on coarse-scale large ensemble simulations for volcanic eruptions of different magnitudes (external page Kroll et al., 2021). For the frozen moisture entry within overshooting convection, we narrow in and select one volcanic eruption magnitude to investigate with km-scale simulations. The km-scale simulations then allow us to study the effects of heating perturbations on single deep convective systems (external page Kroll et al. 2023, external page Kroll et al. 2024).
We have synthesized the findings from coarse and high resolution studies by combining them in one thermodynamically based description of the involved processes. This formulation allows a fast estimation of changes in stratospheric moisture content. The findings can be used to improve upon the currently used parameterizations of overshooting convection, estimate the impact of future eruptions, or to bring the stratospheric moisture increase, i.e. after the 2022 Hunga eruption, into context (external page Kroll and Schmidt 2024).