TL;DR
This paper introduces a novel framework for evaluating implicit-explicit (IMEX) time-stepping schemes in atmospheric models. By using a linearized non-hydrostatic system of normal modes, the framework offers a more rigorous assessment of scheme stability, dispersion, and dissipation compared to previous methods, enabling the development of more robust and efficient atmospheric models.
What Problem Does This Paper Address?
Accurately simulating atmospheric dynamics requires numerical models to handle both slow (e.g., Rossby, gravity) and fast (e.g., acoustic) waves. Explicit time-stepping schemes are limited by the Courant-Friedrichs-Lewy (CFL) condition imposed by these fast waves, necessitating very small time steps and high computational costs. Implicit-explicit (IMEX) schemes offer a solution by treating fast waves implicitly while keeping slow waves explicit, balancing accuracy and computational efficiency. However, evaluating the stability and performance of these IMEX schemes has been challenging. Existing evaluation frameworks, often based on simplified 2D acoustic systems, do not fully capture the complexities of real atmospheric dynamics, potentially leading to schemes that appear stable in these simplified tests but are unstable in more realistic scenarios. This paper addresses this gap by proposing a new, more comprehensive evaluation framework.
Methods
The core of this paper is the development of a new evaluation framework based on a linearized non-hydrostatic system of normal modes. This system is derived from the atmospheric dynamics equations and includes a full spectrum of wave types (acoustic, gravity, and Rossby) and is linearized around a hydrostatic reference state. The authors then use this system to analyze the stability, dispersion, and dissipation properties of IMEX Runge-Kutta (RK) schemes.
Key methodological contributions include:
- Normal Mode Analysis: Instead of simplified 2D acoustic systems, the framework utilizes a more complex system of normal modes that better represents the behavior of waves in atmospheric models.
- Spacetime Operator Construction: The authors construct a spacetime operator from the discretized system of equations and IMEX schemes, allowing for the analysis of their numerical stability.
- Stability Diagrams: Stability diagrams are generated by examining the eigenvalues of the spacetime operator, revealing regions of stable and unstable time steps and resolutions.
- Dispersion and Dissipation Analysis: The framework also allows for the evaluation of how well IMEX schemes preserve wave frequencies (dispersion) and damp wave amplitudes (dissipation).
- Development of New Schemes: Using this framework, the authors develop and analyze a family of second-order, low-storage, high-CFL IMEX RK schemes.
Datasets
Not discussed in the paper. The paper focuses on theoretical analysis and numerical simulations of idealized systems rather than using observational or reanalysis datasets.
Key Insights and Ideas
- Framework Selectivity: The new framework based on normal modes is significantly more selective than previous 2D acoustic systems. Schemes that appear stable in the 2D acoustic system can be unstable when evaluated with the normal mode system, highlighting the limitations of simpler tests.
- IMEX Scheme Design: The framework provides crucial insights for designing IMEX schemes. The authors demonstrate how to develop second-order, low-storage, high-CFL IMEX RK schemes by carefully constructing the implicit part of the scheme.
- Wave Behavior in IMEX: The analysis reveals that IMEX schemes can have varying impacts on different wave types. While gravity and Rossby waves are generally well-preserved, acoustic waves can experience significant dissipation or dispersion, which can be tuned by modifying the implicit table of the scheme.
- Importance of Acoustic Wave Representation: The paper argues that even though acoustic waves have low energy, their accurate representation is crucial for maintaining hydrostatic balance in atmospheric models, underscoring the need for careful IMEX scheme evaluation.
Key Results
- M1 Scheme Instability: The M1 scheme, which appeared stable in 2D acoustic system tests, was found to be unstable in the new normal mode framework for practical time steps and resolutions.
- Development of M2 Schemes: A family of second-order, low-storage, high-CFL IMEX RK schemes (M2 schemes) was developed. These schemes offer improved stability and performance characteristics.
- M2be Scheme Stability: The M2be scheme, utilizing a backward-Euler implicit stage, demonstrated the most robust stability, allowing for significantly larger time steps compared to other schemes.
- M2cn Scheme Non-Dissipation: The M2cn scheme, with a Crank-Nicolson implicit stage, was found to be non-dissipative for hyperbolic problems, offering a different trade-off between stability and dissipation.
- Vertical Resolution Independence: For some schemes (ARK2(2,3,2)(2) and M2b), stability was found to be independent of vertical resolution, which is a desirable characteristic for high-resolution models.
Limitations and Future Work
- Linearized System: The framework is based on a linearized system. While this simplifies analysis, it doesn't fully capture the nonlinear behavior of real atmospheric models.
- Dry Atmosphere: The analysis is performed for a dry atmosphere. Future work could extend this to include moist atmospheric processes.
- Comparison with Boussinesq Equations: The authors suggest that an extension of this work would be to investigate the framework's selectivity using compressible Boussinesq equations, similar to other studies.
- Tuning Dispersion and Dissipation: The paper highlights that dispersion and dissipation of acoustic waves can be tuned by modifying the implicit table of a method, suggesting further exploration of this tuning capability.
- Nonlinear Stability: While the framework provides insights into linear stability, a full assessment of nonlinear stability would require further investigation.
Summary updated