Despite significant progress, challenges persist in identifying and correcting errors within individual ESM components, exacerbated by complex inter-component interactions. Moreover, limitations in the observational network hinder the full constraint of ESMs. Efforts to assess, test, and enhance models have reduced major systematic errors, yet challenges persist while new ones have emerged. Ongoing research and development efforts focus on enhancing the accuracy, seeking a better representation of different processes in ESMs by identifying and correcting systematic errors, but also considering the efficiency, and scalability of ESM components, including the dynamical core, physics, and their coupling.
This session invites contributions related to the development, testing, and application of innovative techniques for ESMs, as well as contributions that provide a deep understanding of the nature and causes of systematic errors in ESMs. Topics encompass governing equations, horizontal and vertical discretizations, structure-preserving methods, time-stepping schemes (including parallel in time schemes), advection schemes, adaptive multi-scale models, model errors across space and time scales, hierarchies of models, physics-dynamics and physics-physics cross-component coupling, initialized predictions, climatology of weather prediction models, data assimilation methodologies, the use of ML to identify errors and detect causal connections, stochastic parameterization for uncertainty representation, and verification diagnostics and metrics for characterizing systematic errors and process understanding across various modeling communities (regional and global km-scale modeling).
Global ocean simulations at very high resolution are extremely time consuming. Representing sub-mesoscale eddies on a numerical grid requires local resolutions of around 600m and is currently only possible in simulations over a few weeks or months. We will investigate two approaches to increase the throughput of ICON-O with the aim of enabling sub-mesoscale resolving simulations of climatologically relevant timescales.
The first approach replaces the current Adams-Bashforth time stepping method with parallelizable spectral deferred correction (SDC) methods. SDC is an iterative method that computes the stages of a fully implicit Runge-Kutta method by multiple “sweeps’’ with a low-order integrator, often an implicit-explicit Euler. It delivers arbitrary, tunable order of accuracy and possesses good stability properties. Proper selection of method parameters allows for small-scale parallelization of each sweep, using threads up to the number of computed stages. We will investigate stability, accuracy and efficiency for a parallel SDC implementation in ICON-O and the research code SWEET. Benchmark results on a single node of JUSUF at Jülich Supercomputing Center demonstrate speedups over the currently used Adams-Bashforth-2 algorithm.
The second approach is based on super-resolution techniques from image enhancement. We propose a dynamic super-resolution technique, where the numerical solution is frequently modified by a U-net-type neural network to correct it towards the restriction of a higher resolution simulation. For the Galewsky test case we demonstrate that our approach can deliver L2 errors comparable to a 10km resolution on a 20km resolution mesh while correctly conserving mass.
How to cite: Ruprecht, D., Freese, P., Götschel, S., Lunet, T., Lapolli, F. R., Korn, P., Witte, M., Kadow, C., and Schreiber, M.: Improving performance of ICON-O through parallel-in-time integration and dynamic super-resolution, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-1213, https://doi.org/10.5194/egusphere-egu25-1213, 2025.
This presentation focuses on numerical solutions for Initial Value Problems (IVPs) involving linear PDEs dominating the time step size, as is the case for dynamical cores. We investigate using Rational Approximation of Exponential Integration (REXI). REXI replaces sequential time-stepping with a sum of rational terms, enabling parallelization and exploiting additional scalability on supercomputers for spatially limited problems.
We introduce the "unified REXI" method, showing its algebraic equivalence to methods developed decades ago, such as implicit Runge-Kutta methods, Cauchy-contour integration, and direct approximations. Our studies involve basic test cases for dynamical cores, offering a detailed numerical investigation, discussion, and comparisons of REXI methods. We address numerical issues and propose workarounds where feasible. Performance comparisons are conducted using nonlinear shallow-water equations on a rotating sphere on high-performance computing systems.
In addition to exposing more parallelism for faster solutions, we evaluate resource efficiency at prescribed accuracy. Our findings reveal that diagonalized lower-order Gauss Runge-Kutta methods (formulated as REXI) achieve a 64x reduction in computational resource requirements compared to prior work.
How to cite: Schreiber, M. and Brown, J.: Rational Approximation of Exponential Integration (REXI) for dynamical cores, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6960, https://doi.org/10.5194/egusphere-egu25-6960, 2025.
Global numerical modeling is entering the era of kilometer-scale, non-hydrostatic, and AI-powered. Meanwhile, heterogeneous computing is the trend in HPC. As a strong candidate for the next-generation global kilometer-scale GCM, the A-grid dynamical core based on LMARS (Low Mach number Approximate Riemann Solver) needs to address the following three issues: 1. The strong-gradient problem is particularly significant at high resolutions. Although LMARS can use higher-order numerical schemes, it cannot guarantee monotonicity. 2. The large discrepancy between vertical and horizontal grid spacings severely limits the time integration step size of the non-hydrostatic model. 3. Classic models are written in FORTRAN, and algorithms designed for CPUs may not be suitable for GPUs. This study builds a prototype model LMARSpy with specific solutions to these issues: addressing the strong-gradient problem through a high-order monotonicity limiter, to solve the non-hydrostatic problem with a conserving vertical implicit solver, and building a Python-based high-performance computing platform to address heterogeneous computing challenge. A series of benchmark tests show that: 1. The monotonicity limiter effectively eliminates non-physical oscillations and maintains high-order accuracy in discontinuous regions, with a computational cost increase of only 10.4% on GPUs. 2. The vertical implicit solver relaxes CFL limitations in cases where vertical grid spacing is much larger than horizontal grid spacing, improving computational efficiency by at least an order of magnitude. 3. The Python high-performance computing platform supports the efficient operation of the dynamical core on both CPU and GPU computing platforms. The performance of a single GPU-based system can rival large computing clusters with over 325 standard CPU cores. Last but not least, with the PyTorch backend built-in, LMARSpy is born with efficient compatibility with machine learning.
How to cite: Wei-Kang, Z. and Xi, C.: A GPU-Ready LMARS-Based Nonhydrostatic Dynamical Core with a Monotonicity Limiter and a Vertical Implicit Solver, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2801, https://doi.org/10.5194/egusphere-egu25-2801, 2025.
The currently available computing power severely limits the spatial resolution in chemistry climate simulations, even on upcoming exascale machines. This is mainly due to the large number of prognostic variables which includes chemical tracers. To further enhance the reliability and accuracy of climate projections, small scales need to be better resolved.
Adaptive methods enable a continuously re-adjusted focus of computational power in time and space. This increases the achievable level of detail considerably, while reducing the time to solution and resource consumption. However, adaptivity requires a sophisticated selection of adaptation criteria, algorithms, memory layouts, and communication patterns to fully utilize modern HPC infrastructures.
Additionally, discontinuous Galerkin methods promise to increase the effective resolution, i.e. by employing high order polynomials, so that prognostic variables are better resolved, even on coarser meshes. Most of the additional computation is done locally, so that the overall algorithm is ideally suited for parallel execution. The typical lack of robustness of higher-order methods can be remedied by utilizing state-of-the-art entropy stable schemes.
In this conference contribution, we will present the setup and the interfaces between MESSy, Trixi.jl and t8code as well as results from a prototypical simulation that showcases the interaction, application, and challenges of dynamic adaptive meshes.
Here, MESSy is a software framework that allows to integrate multiple numerical model components to build regional and global chemistry climate models. t8code is a parallel mesh management library written in C++. Finally, Trixi.jl is a computational fluid dynamics solver, build around a modern Discontinuous Galerkin method, and written in the Julia programming language.
This work was performed within the project ADAPTEX (ADAPtive Earth system modelling with strongly reduced computation time for EXascale-supercomputers), which aims to evaluate the potential benefit of dynamic adaptive meshes in ESM.
How to cite: Hartung, K., Geihe, B., Hergl, C., Holke, J., Jöckel, P., Markert, J., and Schlottke-Lakemper, M.: Towards dynamic adaptive mesh refinement in Earth system models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-15654, https://doi.org/10.5194/egusphere-egu25-15654, 2025.
In this contribution, we present a non-standard, functional analytic framework to rigorously analyze the properties of the semi-implicit second-order finite volume discretization for the compressible Euler equations developed by Benacchio and Klein (2019). In experiments, this method shows favorable stability properties for geophysically relevant benchmarks and is capable of approximating the pseudo-incompressible and/or hydrostatic limit regime without changing the underlying discretization.
Our main results are the consistency and stability of the implicit projection step, which we achieve by choosing discontinuous velocity and continuous pressure variables. As a consequence, the classical divergence is replaced by its natural analogue, i.e., by line integrals along the boundary of a dual cell.
In contrast to preceding work, we consider general quadrilateral and cuboid meshes, and we provide an interpolation operator that is compatible with the natural divergence on the dual grid.
Aiming for a rigorous stability estimate of the overall scheme, we furthermore discuss a choice of advection operator that ensures compatibility.
How to cite: Hastermann, G. and Klein, R.: Analysis of a Seamless Semi-Implicit Finite Volume Method for Atmospheric Flows, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19435, https://doi.org/10.5194/egusphere-egu25-19435, 2025.
Sea ice is an import part of Earth's climate system. Yet, an accurate, highly resolved simulation of sea ice dynamics remains challenging. As the development of faster processors has slowed down, a turn to more specialized hardware is needed to achieve more accurate simulations at higher resolutions. Graphics processing units (GPUs) offer an order of magnitude higher floating-point performance and efficiency compared to CPUs. However, their full utilization often also requires significant engineering effort. Therefore, several frameworks have emerged in recent years which aim to simplify general-purpose GPU programming. In particular, heterogeneous compute frameworks such as SYCL and Kokkos make it possible to develop a unified code base that works accross GPUs and CPUs. Similarly, machine learning frameworks like PyTorch combine an easy to use interface with highly specialized backends that can make it possible to transparently exploit new hardware features to accelerate large-scale linear algebra workloads. Furthermore, their use provides a simple path-way to integrate machine learning components into simulations.
In this talk, we compare available options for the GPU-parallelizaton of the novel sea-ice code neXtSIM-DG. Its dynamical core is based on higher-order finite elements for the momentum equation and discontinous Galerkin elements for the advection. This makes the code highly parallezible. We discuss characteristics of our discretization and its consequences for the GPU implementation. For the full port of the dynamical core we use Kokkos as, based on our assessement, it combines usability with good performance. With moderate changes compared to the OpenMP-based CPU code, the new implementation achieves a sixfold speedup on the GPU while being as fast as the reference on the CPU.
How to cite: Jendersie, R., Lessig, C., and Richter, T.: Development of a GPU-accelerated, Finite Element based Dynamical Core for Sea Ice, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5694, https://doi.org/10.5194/egusphere-egu25-5694, 2025.
Impact of a New Submesoscale Parameterization Scheme on the Simulation of Kuroshio Extension in a High-Resolution OGCM
Ziyi Zhang1,2, Bo An1, Zhiwei Zhang3 and Yongqiang Yu1,2
- Key Laboratory of Earth System Numerical Modeling and Application, Institute of Atmospheric Physics, Chinese Academy of Sciences, Beijing 100029
- University of Chinese Academy of Sciences, Beijing 100049
- Ocean University of China, Qingdao 266100
ABSTRACT
In this study, we investigate the impact of a new submesoscale parameterization scheme developed by Zhang et al. (2023; Zhang23 parameterization hereafter) on the simulation performance of the regional version of the 1/10° ocean model LICOM3.0 in the North Pacific. Specifically, a set of numerical experiments with and without Zhang23 parameterization scheme are conducted to study changes in thermal and dynamic characteristics in the Kuroshio Extension region resulting from the submesoscale processes. Compared to the control experiment, the deeper bias in the winter mixed layer depth(MLD) is significantly reduced in the sensitivity one with Zhang23 parameterization in the Kuroshio Extension region, by 26.7m (24.0%) on average. The simulated Kuroshio Extension shifts southward by about one degree of latitude from 36.5°N to 35.5 °N, closer to the observations.
The simulated SST in the sensitivity run is much cooler than in the control run in the KE region. This contradicts the change in net surface heat flux, implying that the internal dynamical mechanism dominates these changes. Further analyses suggest that the vertical heat fluxes caused by parameterized submesocale processes are mainly concentrated at 39-40°N, 140-150°E in winter. This results in significant warming in the upper 100 m and cooling below, contributing to mixed layer restratification. The ocean heat content decreases in the region where the most energic mesoscale eddy exist in KE region, resulting in a weaker meridional thermal gradient at 37°N. This leads to a southward shift of the ocean front in the upper ocean and KE axis, which weakens meridional heat transport, and thus exacerbates the SST cooling north of 37°N. Although the vertical heat flux of the Zhang23 parameterization occurs primarily in winter, the cold SST anomalies in the subsurface are maintained throughout the year via the reemergence mechanism. The parameterization scheme directly affects winter temperatures by promoting upward heat transport, which warms the surface and cools the subsurface, leading to a southward shift and weakening of the WBC. In other seasons, it influences temperatures via current changes, maintaining a cooling anomaly in the north surface and subsurface layers. Overall, this results in a highly variable surface layer and a relatively stable cold subsurface layer.
This study demonstrates the good performance of Zhang23 submesoscale parameterization in improving the simulation of Kuroshio extension in the eddy-resolving OGCM. It reveals the key role of vertical heat transport by submesoscale processes and other oceanic dynamical mechanisms in modulating the sea temperature and its seasonal variation in the Kuroshio Extension region.
How to cite: Zhang, Z., An, B., Zhang, Z., and Yu, Y.: Impact of a New Submesoscale Parameterization Scheme on the Simulation of Kuroshio Extension in a High-Resolution OGCM, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2978, https://doi.org/10.5194/egusphere-egu25-2978, 2025.
We describe the sensitivity of an ensemble of integrations of the HadGEM3-GC5 coupled model, with a ¼° ocean and N96 (~130km) atmosphere, to settings in the ocean component that have been demonstrated to affect the numerical mixing in forced simulations. This configuration is closely related to the UK contribution to CMIP7. The ensemble is integrated for 60 years with constant present-day greenhouse forcing.
The ocean surface temperature has a robust response to numerical mixing, with increased mixing leading consistently to warming over the global ocean by up to 0.5°C, while reducing mixing cools the surface by a similar degree. The response of the surface air temperature is closely similar to that of the SST, and is seasonally amplified at high latitudes in the respective winter. We present large-scale ocean and atmospheric metrics, and discuss mechanisms for the counter-intuitive sign of the sensitivity of surface temperatures in these simulations to numerical mixing in the ocean, in which more mixing warms the surface and vice versa. This sensitivity is significant, since it is comparable with the surface temperature changes expected in a simulation with historical or future greenhouse scenario forcings, and we shall speculate on the implications for modelling future climates.
How to cite: Megann, A., Copsey, D., Walsh, A., and Tooth, O.: The sensitivity of a climate model to numerical mixing in its ocean component, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4105, https://doi.org/10.5194/egusphere-egu25-4105, 2025.
The double Inter-Tropical Convergence Zone (ITCZ) feature is a prominent bias in the precipitation distribution simulated by most weather and climate models. Its persistence over several CMIP generations without substantial improvements is one factor motivating the investigation of whether higher resolution and the discard of parameterizations can solve the misrepresentation of large-scale atmospheric circulation. In this work, we analyze uncoupled simulations with the Icosahedral Nonhydrostatic Weather and Climate Model, ICON, in the eXtended Predictions and Projections configuration (XPP), spanning horizontal resolutions from 160 km up to 5 km. Although improvements in the precipitation bias are apparent due to the better representation of orography starting at 40 km horizontal resolution, we demonstrate that the double ITCZ feature persists over the entire resolution spectrum – even if deep convective parameterization is discarded at 5 km. In a subsequent sensitivity study, we can tie the emergence of the double ITCZ to biases in the near surface humidity arising from the turbulence scheme. We perform a physically motivated parameter optimization to correct for the bias in near surface specific humidity and can reduce tropcial precipitation biases over the entire resolution hierarchy. Our findings not only showcase the benefits of models supporting various resolutions but also underline the importance of further developing indispensable parameterizations.
How to cite: Kroll, C., Schneidereit, A., Jnglin Wills, R., Kornblueh, L., and Niemeier, U.: Parameterization adaption needed to unlock the benefits of increased resolution for the ITCZ in ICON, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7001, https://doi.org/10.5194/egusphere-egu25-7001, 2025.
Recently, hierarchical finite element spaces were constructed using the subdivision algorithm
to generalize well-established concepts in isogeometric analysis to irregular meshes, for example of
the sphere. This talk illustrates the suitability of these function spaces for structure-preserving
simulations of the shallow water equations with local refinement. We focus on quantifying the
observed accuracy gains and the scaling of the compute time by comparison to established finite
elements without local refinement.
A key benefit of the hierarchical spaces is that they maintain a discrete de Rham complex
under local refinement and thus yield structure-preserving numerical methods. This property is
crucial for discretisations of atmospheric models as it allows us to locally increase the resolution
whenever necessary without changing characteristic features of the simulated systems, such as mass,
total energy or vorticity. Thus, the evolution of the discrete system mirrors the continuous system
faithfully and the simulation results are physically meaningful. We will present simulation results
that confirm these theoretical results.
How to cite: Piel, R. and Bauer, W.: An adaptive, structure-preserving numerical method for the rotating shallow water equations using hierarchical finite elements generated through subdivision, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-449, https://doi.org/10.5194/egusphere-egu25-449, 2025.
The limitations of high-performance computing (HPC) often impose significant constraints on the development and performance of numerical models. While double precision ensures high accuracy in traditional models, it comes with substantial computational costs. Lower precision can reduce computational expenses but may introduce round-off errors that degrade model accuracy. The Quasi double-precision (QDP) algorithm addresses these errors by maintaining corrections, thereby improving result accuracy. Building on previous work where the QDP method demonstrated effective enhancement in the dynamics core of the Model for Prediction Across Scales-Atmosphere (MPAS-A), this study explores the potential of extending the QDP approach to both the dynamics core and tracer components. The impact of the QDP method on these two components is explored through a series of planned experiments using idealized and real-data cases. By applying QDP to both the dynamics and tracer components, this research aims to assess the joint effects on model accuracy and efficiency, seeking to demonstrate the feasibility of further enhancing model performance without significantly increasing computational resources. This study offers a promising path toward more cost-effective simulations in numerical weather prediction and climate modeling.
How to cite: Lai, J. and Wang, L.: Enhancing Numerical Model Accuracy with Quasi Double-Precision: Extending the Application to Dynamics and Tracers in MPAS-A, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-2641, https://doi.org/10.5194/egusphere-egu25-2641, 2025.
In this study, the characteristics of Western North Pacific tropical cyclones (TCs) simulated in 1985 to 2014 by a climate system model FGOALS-f3-H that participated in the HighResMIP are evaluated. The simulations from the stand-alone atmospheric and fully coupled versions are inspected to explore the effects of ocean coupling. Results show that the model can capture the main TC characteristics in tracks, amounts, and genesis locations. The seasonal and interannual variation of TC genesis frequency (TCGF) are reasonably simulated, as well as the wind-pressure relationship (WPR), horizontal structures, and TC-induced precipitation. However, some obvious biases remain in both versions, mainly in the TC intensity and TCGF distributions. The simulation shows a much smaller number of super typhoons (SSTY) and tropical storms (TS), and a larger number of severe typhoons (STY), typhoons (TY), and severe tropical storms (STS). The TCGF shows underestimated biases in the west of 140E with less genesis of all categories, and overestimated biases in the east of 140E with more genesis of STY, TY, and STS, corresponding to the biases of intensity distributions. The overestimated TCGF biases are more obvious in coupled simulation which can be explained by the TC genesis potential index (GPI). The stronger biases of relative humidity, potential intensity and vertical wind shear contribute to the higher GPI biases together with warm sea surface temperature (SST) biases and weak Western Pacific subtropical high (WPSH). The model shows a stronger seasonal cycle in atmospheric simulation which is improved in coupled simulation but remains stronger. Meanwhile, the TCGF interannual variation in atmospheric simulation shows a better correlation with the observation (coefficient of 0.44 vs 0.14). The annual TCGF shows similar patterns in empirical orthogonal function (EOF) analysis, and the first principal component (PC1) is relevant to the Nino3.4 index, suggesting that TCGF is simultaneously modulated by the El Niño-Southern Oscillation (ENSO). Therefore, the poor interannual variation in coupled simulation can be attributed to a weak ENSO. For WPR, the simulation shows a larger wind speed with same sea level pressure, but the coupled simulation shows a closer WPR to the observation. The simulated horizontal structures of winds and pressure are similar in atmospheric and coupled simulation, while the warm core of intense TCs is stronger in coupled simulation. The simulated precipitation is overestimated which can be weakened in coupled simulation. The coupled simulation can also capture the observed SST cooling of TCs, while the atmospheric simulation cannot, showing more reasonable atmosphere-ocean interactions. Overall, ocean coupling can improve some details of the simulated TCs but not the climatology due to ocean biases and coupled interactions, the better TC simulation in a coupled model requires improvements on both atmospheric and ocean models.
How to cite: Guo, Y. and Yu, Y.: The simulation of tropical cyclones in the Western North Pacific by a climate system model FGOALS-f3-H and the effects of ocean coupling, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-3460, https://doi.org/10.5194/egusphere-egu25-3460, 2025.
Fully implicit timestepping methods have several potential advantages for atmosphere/ocean simulation. First, being unconditionally stable, they degrade more gracefully as the Courant number increases, typically requiring more solver iterations rather than suddenly blowing up. Second, particular choices of implicit timestepping methods can extend energy conservation properties of spatial discretisations to the fully discrete method. Third, these methods avoid issues related to splitting errors that can occur in some situations, and avoid the complexities of splitting methods. Fully implicit timestepping methods have had limited application in geophysical fluid dynamics due to challenges of finding suitable iterative solvers, since the coupled treatment of advection prevents the standard elimination techniques. However, overlapping Additive Schwarz methods, as introduced for geophysical fluid dynamics by Cotter and Shipton (2023), provide a robust, scalable iterative approach for solving the monolithic coupled system for all fields and Runge-Kutta stages. In this study we investigate this approach applied to the rotating shallow water equations, facilitated by the Irksome package (Farrell et al, 2021) which provides automated code generation for implicit Runge-Kutta methods. We compare various schemes in terms of accuracy and efficiency using an implicit/explicit splitting method, namely the ARK2 scheme of Giraldo et al (2013), as a benchmark. This provides an initial look at whether implicit Runge Kutta methods can be viable for atmosphere and ocean simulation.
References:
Cotter, Colin J., and Jemma Shipton. "Mixed finite elements for numerical weather prediction." Journal of Computational Physics 231, no. 21 (2012): 7076-7091.
Farrell, Patrick E., Robert C. Kirby, and Jorge Marchena-Menendez. "Irksome: Automating Runge–Kutta time-stepping for finite element methods." ACM Transactions on Mathematical Software (TOMS) 47, no. 4 (2021): 1-26.
Giraldo, Francis X., James F. Kelly, and Emil M. Constantinescu. "Implicit-explicit formulations of a three-dimensional nonhydrostatic unified model of the atmosphere (NUMA)." SIAM Journal on Scientific Computing 35, no. 5 (2013): B1162-B1194.
How to cite: Bauer, W. and Cotter, C. J.: Fully implicit timestepping methods for the rotating shallow water equations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4566, https://doi.org/10.5194/egusphere-egu25-4566, 2025.
Terrestrial vegetation affects the atmosphere through both biogeochemical and physical processes. Radiative forcing is affected by carbon uptake and release, while albedo, evapotranspiration, and surface roughness also depend on vegetation. In Earth system models (ESMs), vegetation growth is often represented by prognostic simulations of the leaf area index (LAI). MIROC-ES2L, an ESM version of the MIROC climate model, simulates a larger El Niño-Southern Oscillation (ENSO) amplitude compared to simulations that prescribe observed LAI.
To investigate the cause of this enhanced ENSO amplitude in MIROC-ES2L, we compared the feedback processes contributing to El Niño growth in two experiments: one with observed LAI and the other with model-prognosed LAI. The prognosed LAI experiment showed stronger zonal advection, Ekman, and meridional advection feedbacks, associated with warmer sea surface temperatures (SSTs) in the eastern equatorial Pacific.
Sensitivity experiments were conducted to identify the regions where LAI contributes most significantly to SST changes. These experiments constrained LAI to observed values in specific regions, while using model-prognosed values elsewhere. The results show that the ENSO amplitude is particularly sensitive to LAI over South America, where the model overestimates LAI along the west coast.
We conclude that the underlying mechanism is as follows: increased LAI over South America induces surface cooling due to enhanced latent heat release. This modifies the tropospheric circulation, weakening the local Walker circulation over the Andes and consistently altering the SST distributions. The resulting SST warming off the coast of Peru enhances the convective response to SST anomalies, strengthening the effective Bjerknes feedback and amplifying ENSO.
How to cite: Hashimoto, K., Hajima, T., Tatebe, H., Kataoka, T., and Miura, H.: Land Components Control the ENSO Representation in the Earth System Model MIROC-ES2L, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-4636, https://doi.org/10.5194/egusphere-egu25-4636, 2025.
Monsoon low-pressure systems (LPSs) during the Indian Summer monsoon season are crucial synoptic-scale phenomena over the Indian subcontinent. An average of about 14 LPSs per summer monsoon season form contributing approximately 60-70 percent of rainfall over large parts of India. Many previous modelling studies have shown a systematic bias of a southward shift in LPS activity and a dry bias over the monsoon core zone of ISM. This study is an attempt to account for these biases in the NCAR Community Earth System Model (CESM2.1.3) simulation. Our present-day simulation using the fully coupled configuration of the model shows a southward shift in LPS activity compared to ERA5 reanalyses, consistent with prior studies showing similar biases in CMIP5 (Coupled Model Intercomparison Project Phase 5) model simulations relative to various reanalysis datasets.
We show that biases in simulating LPS activity in CESM2.1.3 are related to a southward bias in the latitude of the low-level westerly jet as well as a dry bias over northwestern India. The simulated monsoon low-level jet, typically located poleward of the zero absolute vorticity contour, is more east-west oriented and biased southward in the model simulation compared to ERA5. This is due to enhanced meridional advection of negative absolute vorticity over the central Indian Ocean below the zero absolute vorticity contour from increased cross-equatorial flow and reduced meridional advection of positive absolute vorticity over the western Arabian Sea above the zero absolute vorticity contour. Further, it is likely that a bias towards larger dry air intrusion from the north and west of Arabian Sea into India leads to a southward displacement of the monsoon low-level jet. This shift adversely affects LPS genesis and growth in northern and northwestern India, leading to a southward bias in LPS activity in model simulations as compared to reanalysis data. We also offer evidence to show that the southward shift in LPS activity and the dry air intrusion into northwest India are nearly common biases across the three generations (CMIP3, CMIP5 and CMIP6) of climate models, contributing to a dry bias over the monsoon core zone of the Indian summer monsoon.
How to cite: Singh, M., Bala, G., and Seshadri, A. K.: Biases in model simulations of Indian Summer monsoon low-pressure systems , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-5990, https://doi.org/10.5194/egusphere-egu25-5990, 2025.
The upper portion of atmospheric models typically employ a damping (sponge) layer to absorb upward-propagating wave energy, thereby preventing spurious wave reflections from the rigid model top. In the Icosahedral Nonhydrostatic Weather and Climate Model (ICON), implicit Rayleigh damping is applied on the vertical velocity following the approach of Klemp et al., 2008. In high-resolution simulations, the waves resolved in the model can increase substantially. Insufficient damping leads to more frequent model crashes due to numerical instability. While increasing the amount of damping might be a simple and intuitive adjustment, excessive damping can also induce back reflections from the damping layer itself, causing spurious standing oscillations. It is therefore crucial to carefully adjust the damping settings with sensitivity experiments.
We analyze a set of ICON simulations with different damping settings. These are 1-month simulations with a horizontal resolution of 10 km. The goal is to provide insights into the optimal damping settings that can improve numerical stability in high-resolution simulations without compromising the atmospheric representation. Evaluation of the mean wind field outside of the damping layer shows no significant changes across different damping settings. However, spurious standing oscillations are observed in the tropical stratosphere. Further examination demonstrates that these oscillations align with the vertical grids, an indication that they can possibly be caused by vertical discretization error rather than back reflections from the damping layer. Investigation of Eliassen-Palm (EP) flux using Transformed Eulerian Mean (TEM) analysis also shows no significant change in the mean EP flux at the damping layer height due to changing damping coefficient. This suggests no major change in back reflections from the damping layer. Overall, our current results show that increasing the amount of damping could improve numerical stability with no indication of severely altered atmospheric representation over the 1-month time frame. Further analysis through long-term simulations will be necessary to assess the possible impact on longer timescales.
References
Klemp, J. B., Dudhia, J., and Hassiotis, A. D. (2008). An Upper Gravity-wave Absorbing Layer for NWP Applications. Mon. Wea. Rev., 136(10), 3987–4004. doi: 10.1175/2008MWR2596.1
How to cite: Wong, T. C., Kroll, C., and Jnglin Wills, R.: Impact of vertical velocity damping on the numerical stability and atmospheric representation in ICON, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-6374, https://doi.org/10.5194/egusphere-egu25-6374, 2025.
To improve the prediction skill of Korea Meteorological Administration climate prediction system (KMA-GloSea6), high-resolution river routing model in land surface model are implemented. The characteristics of the current river routing model were investigated and pointed out the weakness. It was found that GloSea6 uses a relatively simple river routing model, and the simulated river storage is overestimated compared to the observation. To simulate accurate river flow and air-land-sea interaction, it is most desirable to coupling a sophisticated and realistic river routing model. As a simple method, it was tried to reduce the amount of freshwater flowing into the ocean by increasing the resolution in same river routing model. The GloSea6, which can replace the existing 1-degree resolution with 0.5-degree and 0.125-degree resolution for river routing model, was newly constructed. Also, as the resolution of the river routing model changed, the setting coefficients for the meandering and river flow velocity were optimized. It is clear that the newly conducted system reduces errors in river flow and discharge compared to existing operational system. The simulation with optimized high-resolution river routing model shows reduced biases in ocean circulation and temperature, especially in the Pacific and Indian Oceans. These results can be useful in improving seasonal prediction due to more accurate air-sea interaction and in applied and policy research on water resources.
How to cite: Ham, S., Lee, J., and Shin, B.: Effects of the high-resolution river routing model on seasonal prediction model, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7497, https://doi.org/10.5194/egusphere-egu25-7497, 2025.
This study investigates a climate dynamics model that incorporates topographical effects and the phase transformation of water vapor. The system comprises the Navier–Stokes equations, the temperature equation, the speciffc humidity equation, and the water content equation, all adhering to principles of energy conservation. Applying energy estimation methods, the Helmholtz–Weyl decomposition theorem, and the Brezis–Wainger inequality, we derive high-order a priori estimates for state functions. Subsequently, based on the initial data assumptions V0 ∈ H4 (Ω), T′0, q0, mw0 ∈H2 (Ω), we can prove that a strong solution to this system exists globally in time and establish the uniqueness of the global strong solution.
How to cite: Lian, R., Ma, J., and Zeng, Q.: Global existence of the strong solution to the climate dynamics model with topography effects and phase transformation of water vapor, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-7966, https://doi.org/10.5194/egusphere-egu25-7966, 2025.
Coupled general circulation models of atmosphere and ocean form the heart of every earth-system model. Currently, the only languages used to write a model at that complexity are Fortran and C, which are limited by traditional programming patterns. However, modern languages like Python and Julia shine with interactivity, accessibility, and extensibility, yet a coupled model of that complexity, written in a modern language, has not yet been attempted. Here, we present a coupled climate model implemented entirely in the Julia programming language. The model integrates an atmospheric component based on the SpeedyWeather.jl library with the ClimaOcean.jl package for ocean and sea-ice dynamics. This approach leverages Julia's strengths combining computational efficiency with a high-level programming interface. The result is a climate modeling framework that maintains computational efficiency without sacrificing flexibility. The dynamical core employs high-order numerical methods, combining Weighted Essentially Non-Oscillatory (WENO) advection schemes in the ocean and spectral methods in the atmosphere, ensuring a robust and accurate representation of transport processes. A high-level interface for coupling the codes is introduced, which is distinctly more flexible than traditional couplers. This interface allows running coupled simulations on laptops as well as high-performance computing (HPC) resources. We will present initial results from coupled idealized simulations, highlighting key features such as boundary currents in the ocean, convective patterns in the atmosphere, and air-sea interactions.
How to cite: Silvestri, S., Wagner, G., Klower, M., Gelbrecht, M., and Constantinou, N.: A composable climate model in pure Julia , EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14018, https://doi.org/10.5194/egusphere-egu25-14018, 2025.
The Amundsen Sea Low (ASL) is a connection between the tropical variabilities and climate changes in the Antarctic. The combination of negative trend of the Interdecadal Pacific Oscillation (IPO) and the positive trend of the Atlantic Multidecadal Oscillation (AMO) is recognized to result in the strengthening of the Amundsen Sea Low (ASL) as well as the change of the Antarctic sea ice in recent decades. In this study, we demonstrate that models in the Coupled Model Intercomparison Project phase 6 (CMIP6) largely underestimate the combined influence of IPO and AMO on the ASL variability. The unrealistic relationship between IPO and AMO among coupled models, which potentially cancels out their effects on the ASL each other, is one potential factor for the underestimated ASL variability. Another factor is the large inconsistent AMO-related teleconnection patterns among models. Further analysis is carried out to explore the origin of the large model spread, and the result suggests that the cool mean SST biases over the northern tropical Atlantic Ocean and the warm mean SST biases over the southeastern Pacific are both found to be the dominant source for the underestimated ASL change. These results emphasize the collaborative effect of the IPO and AMO to the Southern Ocean region and the related model biases among the coupled models.
How to cite: Cai, D.: Underestimated trends of Amundsen Sea Low associated with unrealistic interdecadal tropical variability in CMIP6 models, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-14770, https://doi.org/10.5194/egusphere-egu25-14770, 2025.
In numerical weather prediction and climate modelling, semi-implicit time-stepping methods have long been favoured for their ability to take large time steps without excessively damping slow-moving waves. Fast-Wave Slow-Wave Spectral Deferred Correction (FWSW-SDC) methods offer an attractive alternative, achieving arbitrary-order accuracy by iteratively solving a collocation problem, akin to implicit Runge-Kutta methods. Similar to semi-implicit methods, FWSW-SDC improves stability compared to fully explicit schemes, enabling large time steps while retaining accuracy. Additionally, the rich literature on parallel-in-time SDC methods presents opportunities for both parallelization within the correction process and across time steps.
In this poster, we extend prior work with FWSW-SDC from linear systems to the nonlinear compressible Euler equations, evaluating its potential for numerical weather prediction and climate modelling applications. We apply SDC methods to standard dynamical core test cases, including the non hydrostatic gravity wave, moist rising bubble, and baroclinic wave tests, to assess stability, accuracy, and computational performance. Finally, we explore the implementation of parallelisable SDC preconditioners in the FWSW framework.
How to cite: Shipton, J., Brown, A., Fregin, J., Bendall, T., Melvin, T., Baumann, T., and Ruprecht, D.: Fast-wave slow-wave spectral deferred correction methods applied to the compressible Euler equations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16648, https://doi.org/10.5194/egusphere-egu25-16648, 2025.
Numerical weather prediction directly benefits from advancements in the accuracy, efficiency, and scalability of the atmospheric model. We present the development of a performance-portable high-level Python-based framework for the next-generation ECMWF global atmospheric dynamical core, enabling simulations at unprecedented numerical resolutions. This new model framework, called the Portable Model for Multi-Scale Atmospheric Prediction (PMAP), is an advancement of the Finite-Volume Module (FVM) originally developed in Fortran at ECMWF.
A key feature of the global PMAP is its implementation with the latest version of the GridTools for Python (GT4Py) domain-specific library, named gt4py.next. This library is tailored to the efficient implementation of conservative finite-volume discretization methods that support, among others, ECMWF’s operational octahedral grid. Co-developed with various Swiss partners, the gt4py.next library itself is under continuous extension, optimization, and refinement alongside the PMAP. The model runs distributed across multiple nodes, enabling large-scale simulations on modern supercomputers with accelerators.
We present recent model validation results and provide an analysis of its performance, portability, and scalability on latest European supercomputing platforms.
How to cite: Faghih-Naini, S., Ehrengruber, T., Kühnlein, C., Papritz, L., and Dueben, P.: Developing ECMWF’s portable next-generation atmospheric dynamical core, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-16827, https://doi.org/10.5194/egusphere-egu25-16827, 2025.
GLOBO is an atmospheric general circulation model developed at the Institute of Atmospheric Sciences and Climate of the National Research Council of Italy (CNR-ISAC). It is largely equivalent to the BOLAM atmospheric model, used for synoptic-scale operational numerical weather prediction. The GLOBO model is currently part of the S2S multi-model initiative for prediction at the sub-seasonal to seasonal timescale range.
Here, we present improvements to the performance of GLOBO that were obtained through collaboration with the ESiWACE3 initiative. This service is aimed at supporting the exascale preparations for the weather and climate modelling community in Europe through the establishment of short collaborative projects between Research Software Engineers (RSEs) and model development groups. These collaborations provide guidance, engineering, and advice to improve model efficiency and port models to existing and upcoming computing infrastructures
The main tasks that were carried out throughout the collaboration were aimed at:
- getting a better vectorized instruction set on AMD processors using the Intel compilers
- improving the efficiency of inline functions called inside some loops
- gathering parallel communications before waiting for the data to be exchanged
- reducing the number of unnecessary or redundant communications;
The model was tested at a 78km horizontal resolution with a number of processors ranging between 8 and 240. An improvement in the scalability of the model was observed, leading up to 25-34% speedup (on 240 or 160 processors resp.).
How to cite: Guibert, D., Lucido, L., Raffin, E., Bellucci, A., Davini, P., Fabiano, F., Galizia, A., Lembo, V., and Mastrangelo, D.: Technical improvements with the GLOBO atmospheric model, through collaboration with ESiWACE3, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-17712, https://doi.org/10.5194/egusphere-egu25-17712, 2025.
Numerical modeling frameworks are essential for advancing our understanding of oceanic and atmospheric processes. Traditional ocean models, often written in Fortran, offer high computational efficiency but can be challenging for young scientists due to their complexity and lack of GPU support.
We present the Framework for Idealized Ocean Models (FRIDOM), a Python-based and modular framework for fluid simulations. Inspired by machine learning frameworks like TensorFlow and PyTorch, FRIDOM’s flexible design supports applications beyond ocean-specific contexts or idealized setups. Realistic setups with complex topography and realistic forcings are already possible. Currently, it includes implementations of the 2D shallow water equations and 3D nonhydrostatic Boussinesq equations, compatible with both rectilinear Arakawa C-grids and spectral grids. The framework is designed for seamless integration of new governing equations, grid types, such as unstructured grids used in FESOM, and discretization methods.
FRIDOM also provides advanced flow decomposition tools, including Optimal Balance and Nonlinear Normal Mode Decomposition, for separating balanced and wave components in flow fields. Comprehensive documentation, tutorials, and examples ensure accessibility, making FRIDOM a powerful and user-friendly framework for fluid modeling and analysis, with the capability of performing high-resolution simulations to address important questions in geophysical fluid dynamics.
How to cite: Rosenau, S. and Eden, C.: FRIDOM: A new modular Python based framework for geophysical fluid simulations, EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19121, https://doi.org/10.5194/egusphere-egu25-19121, 2025.
Data assimilation (DA) has made a significant contribution to the increase in forecast skill in recent decades by leveraging real-world observations to improve the states used in the forecasting system. However, the DA stage comprises a large part of the computational work of the system. Therefore, improving DA efficiency is crucial for continued improvements in forecast skill, which in turn requires high productivity software to be available for DA researchers.
4D variational assimilation (4DVar) is a DA method in common use operationally. 4DVar optimises an objective function (agreement with observations and prior forecasts) by updating a control (initial conditions) using the adjoint method. This requires running the linearised system forwards in time, and the system adjoint backwards in time, at every optimisation step. Weak constraint 4DVar is a modification that allows an "inexact model" and splitting the time-series into chunks. This enables reformulating each optimisation step as a saddle point problem, where the forward and adjoint models on each time-chunk can be run in parallel. This time-parallelism can potentially greatly decrease the time required for the optimisation.
The adjoint method is highly effective but requires differentiating every operation in the system. Manually differentiation requires high developer effort for new system components, while automatic numerical differentiation is often computationally inefficient. This is particularly problematic in the development stage, where researchers want to run a variety of methods on a variety of equations. This causes a gap between simple equations (e.g. Lorenz and heat) often used in research, and the fluids models in operational systems.
Symbolic differentiation aims to achieve the efficiency of manual differentiation with the automation of numerical differentiation, and provides a route to closing this gap.
We present a library for constructing the 4DVar system using symbolic automatic differentiation. This is achieved using: Firedrake, a finite element library which provides symbolic differentiation using UFL, a high level DSL; and pyadjoint, a library for symbolically recording code execution and automatically constructing the forward and adjoint models.
To construct the 4DVar system, the user need only run the forward model (i.e. dynamical core) and the observation operators once. Firedrake and pyadjoint record this run and calculate the linearised forward model and the adjoint model, from which all components of 4DVar can be constructed.
The implementation is both space and time parallel, enabling the real performance of the methods to be tested on HPC machines. Because pyadjoint is model agnostic, this library can be used for any equation simply by changing the finite element model. Minimising the code modification required for different models is key for improving researcher productivity. This not only allows a smoother transition up the model hierarchy, but also allows straightforward testing of different dynamical core discretisations. We will demonstrate the software API and discuss the design choices taken, and present preliminary results on different equations.
How to cite: Hope-Collins, J., Tabeart, J., and Ham, D.: Automating the construction of time-parallel 4DVar data assimilation systems for finite element models., EGU General Assembly 2025, Vienna, Austria, 27 Apr–2 May 2025, EGU25-19830, https://doi.org/10.5194/egusphere-egu25-19830, 2025.
