The pursuit of global kilometer-scale modeling in atmosphere science presents a dual challenge: comprehensive models become computationally prohibitive and analytically intractable, while idealized models lack essential physical forcing, creating a critical gap in atmospheric research. To bridge this divide, this work develops and validates the innovative ACC-LMARS-SW model, a novel theoretical framework designed for efficient, high-resolution global atmospheric dynamics studies. The model features three synergistic innovations. First, its physics kernel introduces real-world topography and a real-world-equivalent season-aware radiative forcing scheme into the shallow-water equations, significantly enhancing physical realism while retaining conceptual clarity. Second, it employs a variable-resolution stretched cubed-sphere grid that concentrates computational resources on regions of interest , achieving sub-kilometer local resolution without prohibitive global cost. Third, the dynamical core is fully redesigned for massive parallelism using the OpenACC standard, leveraging the efficiency of the Low-Mach Approximate Riemann Solver (LMARS) to harness GPU acceleration.

A suite of numerical experiments demonstrates the model's capabilities. 1) Mesh efficacy: Stretched-grid simulations show reduce error in target areas compared to uniform-resolution runs in classic tests, better resolving nonlinear eddy interactions. 2) Physical fidelity: A global simulation at ~1.5 km average resolution (with ~500 m over the South China Sea) forced by real topography and idealized radiation spontaneously generates a horizontal kinetic energy spectrum featuring distinct  k^-3 and k^-5/3 power-law segments , which is a hallmark of real atmospheric scale interactions. 3) Computational performance: GPU implementation achieves up to a 70x speedup versus estimated serial CPU execution, making global kilometer-scale theoretical experiments feasible on a single workstation.

In conclusion, ACC-LMARS-SW successfully integrates algorithmic design, mesh technology, and high-performance computing to create an efficient and physically insightful "numerical laboratory." This approach provides the high resolution and computational efficiency needed to study fundamental multi-scale dynamics, such as cross-scale energy transfer, dynamical responses to orography, and mesoscale eddy dynamics.

How to cite: Li, Z. and Chen, X.: Accelerated LMARS Shallow Water Model with a Stretched Cubed Sphere Grid at Sub-km Scales, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-4665, https://doi.org/10.5194/egusphere-egu26-4665, 2026.

Radiative transfer plays a key role in the atmosphere’s thermodynamics as photons from the Sun interact with the atmosphere and with other photons being emitted by the earth itself. In climate modeling and numerical weather prediction, it is very common to simplify radiative transfer into a one-dimensional, purely vertical, phenomenon for the sake of saving computational resources. This is because solving the radiative transfer equation requires building and solving what is essentially a five-dimensional problem (three spatial dimensions and two angular dimensions). Improvements in computational resources have resulted in higher-resolution simulations for both climate and weather modeling, and this increase in resolution makes the assumption of purely vertical radiative transfer more and more difficult to justify. However, these improvements to computing power make it possible to attempt to solve a three-dimensional radiative transfer equation as part of a larger atmospheric model and in doing so take into account the three-dimensional effects that are commonly neglected. Doing this efficiently requires the use of adaptive mesh refinement, which while commonly used in the CFD community, is not a technique that is widely used in simulations of the atmosphere. In this work, Jexpresso.jl, an open source flexible general conservation law solver, is extended to solve the full equations of Radiative transfer. Adaptive mesh refinement in both space and angle is then used
to make it possible to speed up solving the three-dimensional radiative transfer equation and couple the solution to an atmospheric model which uses adaptive mesh refinement around clouds. The objective is to allow for improvements in the prediction of atmospheric flow behaviors around clouds and the resulting radiative heat fluxes, which are highly sensitive to cloud cover.

How to cite: Tissaoui, Y., Stechmann, S., Wang, H., and Marras, S.: Toward three-dimensional radiative transfer computations via adaptive mesh refinement in both space and angle, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-13120, https://doi.org/10.5194/egusphere-egu26-13120, 2026.

Various grids are being used to discretize the sphere for general circulation models of ocean and atmosphere, coupled to sea ice and land models. Depending on the numerics (or the network architecture for machine learning-based models) some grids are better suited and more widely used but no single grid meets all requirements. The HEALPix grid was originally invented for applications in cosmology and designed for spectral transform efficiency, hierarchical nested ordering and equal-area grid cells. Here, we present several variants: The OctaHEALPix grid, and the octaminimal Gaussian and Clenshaw-Curtis grids. The OctaHEALPix grid inherits all the properties of the original HEALPix grid but based on a single square matrix instead of 12 used for the base pixels (faces) of the HEALPix grid. This eliminates singularities for the 8 corner cells with 7 instead of 8 neighbouring cells. The nested hierarchy of the OctaHEALPix grid is therefore a quadtree from the coarsest to the finest zoom level. Data on the OctaHEALPix grid can be arranged as a single square matrix, representing an equal-area map projection of the sphere, allowing for interpolation-free visualisation and data storage. However, the HEALPix grids do not provide an exact quadrature in the Legendre transform, such that transform errors are higher than with Gaussian, or equi-angle latitude-based grids using the Clenshaw-Curtis quadrature. The inexact transform with the HEALPix grids does not pose any problems in simulations where other sources of error dominate. We present the grid-flexible spectral transform implemented in the atmospheric circulation model SpeedyWeather.jl that simultaneously supports all ring-based, equi-longitude grids both on CPU (including multithreading) and GPU. The OctaHEALPix grid is also favourable for machine learning-based models like diffusion models based on the UNet architecture which only require custom boundary conditions. The other HEALPix variant we present is the octaminimal Gaussian grid, which imposes Gaussian latitudes on the OctaHEALPix grid which reduces transform errors while preserving more of the HEALPix’s equal-area and hierarchy properties. Similarly, the octaminimal Clenshaw-Curtis grid uses regular latitudes with the Clenshaw-Curtis quadrature. Simulations based on these grids are presented for coupled climate simulations and implications for hybrid numerical and machine learning-based models are discussed.

How to cite: Klöwer, M., Gelbrecht, M., Leland, J., Groenke, B., and Hotta, D.: Variants of HEALPix grids for global climate modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15163, https://doi.org/10.5194/egusphere-egu26-15163, 2026.

We introduce a new suite of idealized test cases for the dynamical cores of atmospheric General Circulation Models. The tests were developed for the 2025 Dynamical Core Model Intercomparison Project (DCMIP-2025) which took place at the National Center for Atmospheric Research (NCAR) in June 2025 alongside a summer school (https://dcmip.org). The test suite was designed to probe (a) the impact of topography on atmospheric flows at various scales (see also the EGU 2026 companion paper led by Timothy Andrews); (b) a convection (squall line) test case; and (c) idealized experiments for atmospheric machine learning emulators.

This presentation focuses on the squall line test case, which challenges nonhydrostatic dynamical cores to accurately simulate moist convection and squall line dynamics at kilometer-scale resolutions. The scenario incorporates simplified moisture feedbacks using a warm-rain Kessler parameterization and is implemented across a range of horizontal (0.25–4.0 km) and vertical (250–500 m) grid spacings on a reduced-radius sphere. We assess three leading nonhydrostatic dynamical cores: the nonhydrostatic version of the Department of Energy/NCAR ‘Spectral Element’ model (called HOMME or SE), the ‘Model for Prediction Across Scales’ (MPAS), and NOAA’s Finite-Volume cubed-sphere dynamical core FV3. These are all available within NCAR’s Community Atmosphere Model (CAM) which is the atmospheric component of the Community Earth System Model (CESM).

Our analysis examines the numerical convergence characteristics, model-to-model variability, and the influence of core-specific dissipation mechanisms on the simulated convective storms. The results demonstrate the test case’s potential as a rigorous benchmark for future core development and model intercomparison efforts. This work contributes to advancing robust evaluation frameworks for atmospheric models in the kilometer-scale, nonhydrostatic regime.

How to cite: Jablonowski, C., Androski, N., Andrews, T., and Hughes, O.: DCMIP-2025: Introducing an Idealized Squall Line Test Case for Nonhydrostatic Dynamical Cores, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-8455, https://doi.org/10.5194/egusphere-egu26-8455, 2026.

Spectral methods have a well-established history in numerical weather prediction. However, stable formulations for high-order discontinuous Galerkin (DG) methods in both horizontal and vertical directions have only recently been developed using split-form DG (cf. Waruszewski et al. and Souza et al.).

This presentation provides an overview of the implementation of split-form DG methods for the compressible Euler equations, utilizing various formulations on both triangular and quadrangular spherical grids. Key focus areas include split-form versions on triangular grids and the linear algebra associated with HEVI-like (Horizontally Explicit, Vertically Implicit) temporal integration schemes. To manage vertically propagating sound waves implicitly, temporal integration is performed using W-Rosenbrock methods with an approximate Jacobian in the radial direction.

The framework is implemented in the Julia package CGDycore.jl, leveraging KernelAbstractions.jl and MPI for parallel execution across diverse architectures. We present numerical comparisons using the Held-Suarez test case, alongside comprehensive weak and strong scaling results for different computing environments.

How to cite: Knoth, O. and Artiano, M.: A discontinuous Galerkin weather dycore for triangular and quadrangular grids, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-14161, https://doi.org/10.5194/egusphere-egu26-14161, 2026.

Atmospheric dynamical cores solve the equations of motion that describe the fluid flow at the resolved scales. Commonly used dynamical cores, including the Department of Energy's (DOE) ‘Higher Order Methods Modeling Environment’ (HOMME) dynamical core, also known as a ‘Spectral Element’ (SE) dynamical core, and the National Center for Atmospheric Research's (NCAR’s) ‘Model for Prediction Across Scales’ (MPAS), contained two approximations of the fluid dynamics equations. They are called the Shallow-atmosphere and Traditional (SA+T) approximations. These approximations discard the so-called Nontraditional Coriolis terms. Omitting these terms significantly biases the dynamical response to diabatic heating, inducing up to 10% relative error in the wind field as linear model results suggested. The biases induced by these approximations are expected to be most severe in tropical regions, and to become more severe at finer grid spacings. We present preliminary evidence that removing these approximations may help mitigate the double Intertropical Convergence Zone (ITCZ) bias that is present in many climate models.

We present dynamical core intercomparisons between the shallow-atmosphere version of the HOMME and MPAS dynamical cores and their deep-atmosphere variants which remove the SA+T approximations. We present an overview of the modifications made to the HOMME and MPAS dynamical cores. Aquaplanet simulations made using NCAR’s Community Earth System Model (CESM) show that for sea surface temperature distributions that produce a double ITCZ, removing the SA+T approximations induces equatorward shifts in precipitation. We will examine sensitivities to physics-dynamics coupling strategies and parametric sensitivity to the deep convection scheme. We also examine the performance of MPAS, which collocates physics columns with dynamics columns, with HOMME, which simulates physics on a sparser finite-volume grid. Our simulations are done at a nominal 1º grid spacing, providing conclusive evidence that the SA+T approximations induce climatologically significant biases at the coarser grid spacings used in workhorse climate model configurations.

How to cite: Hughes, O., Ong, H., Herrington, A., Lauritzen, P., Guba, O., Taylor, M., and Jablonowski, C.: Examining the Contribution of Deep-Atmosphere Dynamical Cores to Mitigating Double ITCZ Bias in Aquaplanets, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15662, https://doi.org/10.5194/egusphere-egu26-15662, 2026.

Normal modes of hydrostatic primitive equations on the sphere decompose the linearized flow into the slow propagating Rossby, fast propagating inertio-gravity (IG), and equatorial Kelvin and MRG waves, characterized by intermediate speeds. However, this simple characterization is no longer valid in the nonlinear system, where wave-wave, wave-mean flow interactions, and adiabatic forcing can significantly alter the wave speeds.

A number of methods aimed at determining the slowly evolving component of the nonlinear flow (so called slow manifold) were proposed under the umbrella terms ”nonlinear normal mode decomposition” (NNMD) or "nonlinear normal mode initialization" (NNMI). In the classical NNMD, Rossby waves are considered slow a priory, while the slow IG part consists of the the unbalanced component of the flow slaved to the Rossby modes. More precisely, in classical NNMD, slow IG waves are those that would be stationary in a nonlinear flow consisting of the slow modes only. While this approach is very successful in suppressing high-speed gravity waves, it also suppresses slowly propagating linearly unbalanced flow and large scale tropical circulation.

We propose a new method of flow decomposition into the slow and fast components based on computing instantaneous phase speeds of normal modes in the nonlinear system. The method does not make assumptions on the composition of the slow manifold. Instead, the decomposition is obtained from a constraint optimization problem that minimizes the norm of the fast component, while requiring that the slow manifold does not contain modes propagating faster than the selected cutoff speed. We apply the method to reanalysis data, demonstrate its efficiency, and analyze the composition of the slow manifold as function of the cutoff speed. In particular, when the cutoff is chosen to be approximately equal to the fastest linear Rossby wave speed, most of the tropical circulation and significant part of unbalanced modes are retained in the slow manifold.  

How to cite: Vasylkevych, S., Chen, J., Holube, K., Zagar, N., and Lunkeit, F.: New nonlinear normal modes flow decomposition based on instantaneous phase speeds in the normal modes framework, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-15174, https://doi.org/10.5194/egusphere-egu26-15174, 2026.

Weather and climate models require atmospheric transport to be numerically stable, accurate, efficient, and mass-conserving. Adaptively Implicit-Explicit (AdImEx) time stepping can provide mass conservation and stability while taking long time steps for efficiency. However, challenges persist: (1) strongly reduced accuracy for large Courant numbers, and (2) a lack of preservation of uniform fields with substage time step sizes varying in space. We introduce AdHImEx, a novel AdImEx approach that resolves these issues, delivering an unconditionally stable, conservative, efficient advection scheme that preserves unity and increases temporal accuracy for large Courant numbers from first- to second-order. Remarkably, it requires only a single implicit solve per time step, performed with an iterative matrix solver, while retaining the full efficiency and third-order accuracy of the explicit time stepping in regions with small Courant numbers. Two-dimensional results will be shown.

How to cite: te Winkel, A., Weller, H., Kühnlein, C., and Kent, J.: AdHImEx: Adaptively High-order Implicit-Explicit Time Stepping for Atmospheric Transport with Long Time Steps, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-3905, https://doi.org/10.5194/egusphere-egu26-3905, 2026.

Accurate and efficient time integration is a critical component of Numerical Weather Prediction (NWP) and climate modelling. Current approaches in the Met Office’s next‑generation dynamical core, GungHo, rely on a serial-in-time, low-order, iterative semi‑implicit timestepping scheme coupled with Flux‑Form Semi‑Lagrangian (FFSL) transport scheme. As model resolutions and computational scales increase, exploiting parallelism in the time dimension is becoming increasingly important.

This work evaluates the potential of time‑parallel Deferred Correction (DC) methods as a viable alternative to these traditional schemes. I examine two strategies for introducing temporal parallelism:

• Revisionist Integral Deferred Correction (RIDC), which parallelises across correction sweeps, and
• Spectral Deferred Correction (SDC) with time‑parallel (diagonal) preconditioners, which parallelises across collocation nodes.

Results are presented for a suite of standard dynamical core test cases for the compressible Euler equations. These experiments demonstrate how time‑parallel DC algorithms can improve temporal accuracy and offer meaningful opportunities for reducing wall‑clock time.

Together, these developments highlight the promise of DC-based integrators as a pathway toward more scalable and efficient time-stepping in next‑generation atmospheric models.

How to cite: Brown, A.: A comparison of time-parallel deferred correction methods for atmospheric modelling, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-19200, https://doi.org/10.5194/egusphere-egu26-19200, 2026.

In the last decade, there has been growing interest in developing new dynamical cores for climate and weather simulations based on the Discontinuous Galerkin (DG) approach. To achieve accuracy, efficiency, and stability, various numerical formulations of the Euler equations are currently being explored.

In this talk, we present novel structure-preserving methods for different formulations of the Euler equations. These include models based on different thermodynamic variables, such as potential temperature, internal energy, or total energy, as well as the Exner pressure formulation and the vector-invariant form. These methods are developed within the flux-differencing DG framework, with a specific focus on the efficient implementation of split forms for both conservative and non-conservative terms.

The implementation is carried out entirely in Julia and is integrated within the Trixi.jl and CGDycore.jl ecosystems. We will discuss the different architectural approaches used in these packages and showcase numerical results, specifically focusing on the baroclinic instability test case to compare the different formulations.

How to cite: Artiano, M., Knoth, O., Spichtinger, P., and Ranocha, H.: Structure-Preserving Methods for the Euler Equations, EGU General Assembly 2026, Vienna, Austria, 3–8 May 2026, EGU26-18731, https://doi.org/10.5194/egusphere-egu26-18731, 2026.