The representation of nonlinear subgrid processes, especially clouds, has been a major source of uncertainty in climate models for decades. Cloud-resolving models better represent many of these processes and can now be run globally but only for short-term simulations of at most a few years because of computational limitations. Here we demonstrate that deep learning can be used to capture many advantages of cloud-resolving modeling at a fraction of the computational cost. We train a deep neural network to represent all atmospheric subgrid processes in a climate model by learning from a multiscale model in which convection is treated explicitly. The trained neural network then replaces the traditional subgrid parameterizations in a global general circulation model in which it freely interacts with the resolved dynamics and the surface-flux scheme. The prognostic multiyear simulations are stable and closely reproduce not only the mean climate of the cloud-resolving simulation but also key aspects of variability, including precipitation extremes and the equatorial wave spectrum. Furthermore, the neural network approximately conserves energy despite not being explicitly instructed to. Finally, we show that the neural network parameterization generalizes to new surface forcing patterns but struggles to cope with temperatures far outside its training manifold. Our results show the feasibility of using deep learning for climate model parameterization. In a broader context, we anticipate that data-driven Earth system model development could play a key role in reducing climate prediction uncertainty in the coming decade.

/pdf/deep-learning-to-represent-subgrid-processes-in-climate-4mcaihu9bx.pdf

Deep learning to represent subgrid processes in climate models.

The two-dimensional turbulent vortex generated by release of buoyant fluid from an instantaneous thermal line source has been simulated by machine numerical solution of a complete set of Eulerian gas equations. The equations included turbulent eddy exchange terms, similar to those used by Smagorinsky, which made possible the generation of computationally stable solutions qualitatively and quantitatively resembling the convective “thermals” studied and described by Scorer and Richards. The results of a number of numerical experiments, performed with varying computational approximations, lead to conclusions as to the importance of various sources of numerical errors and the validity of the eddy exchange formulation. The formulation leads to qualitatively good results with the resolution provided by about 1500 grid points, but it has not yet been possible to exhibit the shape-preserving stage assumed in theoretical treatments and found approximately by laboratory experiments. This is probably due in large part to the neglect of the effects of eddies in the third dimension. DOI: 10.1111/j.2153-3490.1962.tb00128.x

On the numerical simulation of buoyant convection

Global cloud-resolving models (GCRMs) are a new type of atmospheric model which resolve nonhydrostatic accelerations globally with kilometer-scale resolution. This review explains what distinguishes GCRMs from other types of models, the problems they solve, and the questions their more commonplace use is raising. GCRMs require high-resolution discretization over the sphere but can differ in many other respects. They are beginning to be used as a main stream research tool. The first GCRM intercomparison studies are being coordinated, raising new questions as to how best to exploit their advantages. GCRMs are designed to resolve the multiscale nature of moist convection in the global dynamics context, without using cumulus parameterization. Clouds are simulated with cloud microphysical schemes in ways more comparable to observations. Because they do not suffer from ambiguity arising from cumulus parameterization, as computational resources increase, GCRMs are the promise of a new generation of global weather and climate simulations.

/pdf/global-cloud-resolving-models-3lrcisfn6s.pdf

Global Cloud-Resolving Models

ICON-A is the new icosahedral nonhydrostatic (ICON) atmospheric general circulation model in a configuration using the Max Planck Institute physics package, which originates from the ECHAM6 general circulation model, and has been adapted to account for the changed dynamical core framework. The coupling scheme between dynamics and physics employs a sequential updating by dynamics and physics, and a fixed sequence of the physical processes similar to ECHAM6. To allow a meaningful initial comparison between ICON-A and the established ECHAM6-LR model, a setup with similar, low resolution in terms of number of grid points and levels is chosen. The ICON-A model is tuned on the base of the Atmospheric Model Intercomparison Project (AMIP) experiment aiming primarily at a well balanced top-of atmosphere energy budget to make the model suitable for coupled climate and Earth system modeling. The tuning addresses first the moisture and cloud distribution to achieve the top-of-atmosphere energy balance, followed by the tuning of the parameterized dynamic drag aiming at reduced wind errors in the troposphere. The resulting version of ICON-A has overall biases, which are comparable to those of ECHAM6. Problematic specific biases remain in the vertical distribution of clouds and in the stratospheric circulation, where the winter vortices are too weak. Biases in precipitable water and tropospheric temperature are, however, reduced compared to the ECHAM6. ICON-A will serve as the basis of further development and as the atmosphere component to the coupled model, ICON-Earth system model (ESM). Plain Language Summary ICON-A is a new atmospheric model as needed for research on the general circulation of the atmosphere, or as atmospheric component in an Earth system model, as used in climate research. This article describes the construction of the atmospheric model, in particular how two major parts are coupled to each other: “dynamics” and “physics.” Dynamics is the part that solves the equations for the atmospheric motion, temperature, density, and concentrations of water vapor, cloud water, and cloud ice. Physics is the part that computes the changes in these fields related to processes like radiation, cloud condensation, or turbulence. These physical changes depend on the state of the atmosphere as computed by the dynamics, and the changes computed by physics force change in the dynamics. The article documents the details of this construction. Further, the article describes how the physics is tuned to obtain a good representation of the general circulation of the period 1979 to 1988 in comparison to observations. A more detailed evaluation of such simulations is presented in a companion article by Crueger et al. (2018, https://doi.org/10.1029/2017MS001233).

https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1029/2017MS001242

ICON-A, the Atmosphere Component of the ICON Earth System Model: I. Model Description

Radiative-convective equilibrium simulations with a 2 km horizontal resolution are conducted to investigate the impact on convective organization of different parameterizations for horizontal and vertical subgrid turbulence mixing. Three standard approaches for representing horizontal diffusion produce starkly differing mixing rates, particularly for the entrainment mixing into updrafts, which differ by more than an order of magnitude between the schemes. The simulations demonstrate that the horizontal subgrid mixing of water vapor is key, with high mixing rates a necessary condition for organization of convection to occur, since entrainment of dry air into updrafts suppresses convection. It is argued that diabatic budgets, while demonstrating the role of spatially heterogeneous radiative heating rates in driving organization, can overlook the role of physical processes such as updraft entrainment. These results may partially explain previous studies that showed that organization is more likely to occur at coarser resolutions, when entrainment is solely represented by subgrid-scale turbulence schemes, highlighting the need for benchmark simulations of higher horizontal resolution. The recommendation is for the use of larger ensembles to ensure robustness of conclusions to subgrid-scale parameterization assumptions when numerically investigating convective organization, possibly through a coordinated community model intercomparison effort.

https://agupubs.onlinelibrary.wiley.com/doi/pdf/10.1002/2016MS000802

Organization of tropical convection in low vertical wind shears: Role of updraft entrainment

Large-eddy simulations (LES) with the new ICOsahedral Non-hydrostatic atmosphere model (ICON) covering Germany are evaluated for four days in spring 2013 using observational data from various sources Reference simulations with the established Consortium for Small-scale Modelling (COSMO) numerical weather prediction model and further standard LES codes are performed and used as a reference This comprehensive evaluation approach covers multiple parameters and scales, focusing on boundary-layer variables, clouds and precipitation The evaluation points to the need to work on parametrizations influencing the surface energy balance, and possibly on ice cloud microphysics The central purpose for the development and application of ICON in the LES configuration is the use of simulation results to improve the understanding of moist processes, as well as their parametrization in climate models The evaluation thus aims at building confidence in the model's ability to simulate small- to mesoscale variability in turbulence, clouds and precipitation The results are encouraging: the high-resolution model matches the observed variability much better at small- to mesoscales than the coarser resolved reference model In its highest grid resolution, the simulated turbulence profiles are realistic and column water vapour matches the observed temporal variability at short time-scales Despite being somewhat too large and too frequent, small cumulus clouds are well represented in comparison with satellite data, as is the shape of the cloud size spectrum Variability of cloud water matches the satellite observations much better in ICON than in the reference model In this sense, it is concluded that the model is fit for the purpose of using its output for parametrization development, despite the potential to improve further some important aspects of processes that are also parametrized in the high-resolution model

Roeland A. J. Neggers

Papers

Large-eddy simulations over Germany using ICON: a comprehensive evaluation