Search

The new generation of heterogeneous CPU/GPU computer systems offer much greater computational performance but are not yet widely used for climate modeling. One reason for this is that traditional climate models were written before GPUs were available and would require an extensive overhaul to run on these new machines. In addition, even conventional "high–resolution" simulations don't currently provide enough parallel work to keep GPUs busy, so the benefits of such overhaul would be limited for the types of simulations climate scientists are accustomed to. The vision of the Simple Cloud‐Resolving Energy Exascale Earth System (E3SM) Atmosphere Model (SCREAM) project is to create a global atmospheric model with the architecture to efficiently use GPUs and horizontal resolution sufficient to fully take advantage of GPU parallelism. After 5 years of model development, SCREAM is finally ready for use. In this paper, we describe the design of this new code, its performance on both CPU and heterogeneous machines, and its ability to simulate real‐world climate via a set of four 40 day simulations covering all 4 seasons of the year. Plain Language Summary: This paper describes the design and development of a 3 km version of the Energy Exascale Earth System Model (E3SM) atmosphere model, which has been fully rewritten in C++ using the Kokkos library for performance portability. This newly rewritten model is able to take advantage of the state–of–the–science high performance computing systems which use graphical processor units (GPUs) to mitigate much of the computational expense which typically plagues high–resolution global modeling. Taking advantage of this high–performance we are able to run four seasons of simulations at 3 km global resolution. We discuss the biases, including the diurnal cycle, by comparing model results with satellite and Atmospheric Radiation Measurement ground‐based site data. Key Points: Describes the C++/Kokkos implementation of the Simple Cloud–Resolving E3SM Atmosphere Model (SCREAMv1)SCREAMv1 leverages GPUs to surpass one simulated year per compute day at global 3 km resolutionHigh resolution improves some meso‐scale features and the diurnal cycle but large‐scale biases require improvement across all four seasons [ABSTRACT FROM AUTHOR]

Tropical cyclones (TCs) are among the most destructive natural hazards and yet, quantifying their financial impacts remains a significant methodological challenge. It is therefore of high societal value to synthetically simulate TC tracks and winds to assess potential impacts along with their probability distributions for example, land use planning and financial risk management. A common approach to generate TC tracks is to apply storm detection methodologies to climate model output, but such an approach is sensitive to the method and parameterization used and tends to underestimate intense TCs. We present a global TC model (the UQAM‐TCW model thereafter) that melds statistical modeling, to capture historical risk features, with a climate model large ensemble, to generate large samples of physically coherent TC seasons. Integrating statistical and physical methods, the model is probabilistic and consistent with the physics of how TCs develop. The model includes frequency and location of cyclogenesis, full trajectories with maximum sustained winds and the entire wind structure along each track for the six typical cyclogenesis basins from IBTrACS. Being an important driver of TCs globally, we also integrate ENSO effects in key components of the model. The global TC model thus belongs to a recent strand of literature that combines probabilistic and physical approaches to TC track generation. As an application of the model, we show global hazard maps for direct and indirect hits expressed in terms of return periods. The global TC model can be of interest to climate and environmental scientists, economists and financial risk managers. Plain Language Summary: Tropical cyclones (TCs) are among the most destructive natural hazards and yet, quantifying their financial impacts remains a difficult task. Being able to randomly simulate TCs and their features (such as wind speed) with mathematical models is therefore critical to build scenarios (and their corresponding probability) for land use planning and financial risk management. A common approach is to simulate TCs by tracking them directly in climate model outputs but this often underestimates the frequency of intense TCs while being computationally costly overall to generate a large number of events. For these reasons, many authors have looked into alternative approaches that replicate key physical features of TCs but rather using statistical models that are much less computationally demanding. This paper therefore presents a global TC model that leverages the strengths of both statistical and climate models to simulate a large number of TCs whose features are consistent with the physics and observations. As an important global phenomenon that affects TCs globally, we also integrate in our model the effects of El Niño. The paper focuses on the methodology and validation of each model component and concludes with global hazard maps for direct and indirect hits. Key Points: We present a global tropical cyclone (TC) wind model built upon a climate model large ensemble that can be used for risk analysisWe integrate ENSO into our model since it is a strong driver of storm annual frequency, cyclogenesis, trajectories, and intensityWe present global hazard maps consistent with statistical features of TC components and coherent with a global climate model [ABSTRACT FROM AUTHOR]

The atmospheric Green's function method is a technique for modeling the response of the atmosphere to changes in the spatial field of surface temperature. While early studies applied this method to changes in atmospheric circulation, it has also become an important tool to understand changes in radiative feedbacks due to evolving patterns of warming, a phenomenon called the "pattern effect." To better study this method, this paper presents a protocol for creating atmospheric Green's functions to serve as the basis for a model intercomparison project, GFMIP. The protocol has been developed using a series of sensitivity tests performed with the HadAM3 atmosphere‐only general circulation model, along with existing and new simulations from other models. Our preliminary results have uncovered nonlinearities in the response of the atmosphere to surface temperature changes, including an asymmetrical response to warming versus cooling patch perturbations, and a change in the dependence of the response on the magnitude and size of the patches. These nonlinearities suggest that the pattern effect may depend on the heterogeneity of warming as well as its location. These experiments have also revealed tradeoffs in experimental design between patch size, perturbation strength, and the length of control and patch simulations. The protocol chosen on the basis of these experiments balances scientific utility with the simulation time and setup required by the Green's function approach. Running these simulations will further our understanding of many aspects of atmospheric response, from the pattern effect and radiative feedbacks to changes in circulation, cloudiness, and precipitation. Plain Language Summary: Many properties of the atmosphere are affected by the temperature of the ocean surface. Knowing how strong these effects are would help us to better predict global warming. The response to a given surface warming depends on where the warming occurs. To account for this, researchers sometimes simulate the response to individual patches of warming and then assume the response to an arbitrary warming pattern can be summed together from these patch responses. This is sometimes called the atmospheric Green's function method, and it works well at recreating the atmospheric response to historical temperature changes. We are organizing a Green's Function Model Intercomparison Project (GFMIP), in which participants will apply the method consistently for many climate models. This paper presents the GFMIP protocol. In the course of developing this protocol, we found that the atmospheric response to warming is not proportional in all cases: the response to surface warming is not the opposite of the response to surface cooling; warming twice as much doesn't cause twice as much of a response; and making a patch of warming twice as large doesn't cause twice as large a response. GFMIP will help us figure out how to account for this nonlinearity. Key Points: The Green's Function Model Intercomparison Project (GFMIP) explores the atmospheric response to surface temperature patch perturbationsThis paper presents the GFMIP protocol, which was generated using insights from past studies and new sensitivity testsGreen's functions reconstruct the response to historical temperatures, but nonlinearities can affect responses to other warming patterns [ABSTRACT FROM AUTHOR]

The development of any atmosphere or ocean model warrants a suite of test cases (TCs) to verify its spatial and temporal discretizations, order of accuracy, stability, reproducibility, portability, scalability, etc. In this paper, we present a suite of shallow water TCs designed to verify the barotropic solver of atmosphere and ocean models. These include the non‐dispersive coastal Kelvin wave; the dispersive inertia‐gravity wave; the dispersive planetary and topographic Rossby waves; the barotropic tide; and a non‐linear manufactured solution. These TCs check the implementation of the linear pressure gradient term; the linear constant or variable‐coefficient Coriolis and bathymetry terms; and the non‐linear advection terms. Simulation results are presented for a variety of time‐stepping methods as well as two spatial discretizations: a mimetic finite volume method based on the TRiSK scheme, and a high‐order discontinuous Galerkin spectral element method. The experimental procedure for conducting these numerical experiments is detailed. It underscores several key considerations that vary depending on the chosen spatial discretization method. Finally, convergence studies of every TC are conducted with refinement in both space and time, only in space, and only in time. The convergence slopes match the expected theoretical predictions. Plain Language Summary: Before running an atmosphere, ocean, or a coupled climate simulation, every model developer should ensure the correct implementation of each term in the governing equations that drive the models forward in time. This motivates the development of idealized test cases (TCs), each of which verifies a subset of terms in the governing equations with different initial and boundary conditions. Here we present a suite of six TCs for the momentum equation and sea surface height equation for ocean models in a single‐layer configuration. The computed results from the ocean model can be compared to exact solutions. The computed solution always has a small error, but is said to converge to the exact solution with reduction in grid cell size and time step. If the model converges at the expected rate, then we know that it is solving the governing equations correctly. We show results of convergence tests from two models, and share the full specifications of these TCs so that other ocean modelers may reproduce them. Key Points: A suite of test cases is presented for the verification of barotropic dynamics of ocean models, with exact and manufactured solutionsSpecifications are provided for coastal Kelvin wave, inertia‐gravity wave, planetary and topographic Rossby waves, barotropic tide, and non‐linear casesResults are presented for a variety of time‐stepping methods and two types of spatial discretizations: TRiSK and discontinuous Galerkin spectral element method [ABSTRACT FROM AUTHOR]

A reference or "no‐feedback" radiative response to warming is fundamental to understanding how much global warming will occur for a given change in greenhouse gases or solar radiation incident on the Earth. The simplest estimate of this radiative response is given by the Stefan‐Boltzmann law as −4σTe‾3≈−3.8 ${-}4\sigma {\overline{{T}_{e}}}^{3}\approx -3.8$ W m−2 K−1 for Earth's present climate, where Te‾ $\overline{{T}_{e}}$ is a global effective emission temperature. The comparable radiative response in climate models, widely called the "Planck feedback," averages −3.3 W m−2 K−1. This difference of 0.5 W m−2 K−1 is large compared to the uncertainty in the net climate feedback, yet it has not been studied carefully. We use radiative transfer models to analyze these two radiative feedbacks to warming, and find that the difference arises primarily from the lack of stratospheric warming assumed in calculations of the Planck feedback (traditionally justified by differing constraints on and time scales of stratospheric adjustment relative to surface and tropospheric warming). The Planck feedback is thus masked for wavelengths with non‐negligible stratospheric opacity, and this effect implicitly acts to amplify warming in current feedback analysis of climate change. Other differences between Planck and Stefan‐Boltzmann feedbacks arise from temperature‐dependent gas opacities, and several artifacts of nonlinear averaging across wavelengths, heights, and different locations; these effects partly cancel but as a whole slightly destabilize the Planck feedback. Our results point to an important role played by stratospheric opacity in Earth's climate sensitivity, and clarify a long‐overlooked but notable gap in our understanding of Earth's reference radiative response to warming. Plain Language Summary: Earth's climate is stable because a warmer planet loses more energy to space, at infrared wavelengths invisible to the naked eye. The rate of change of this energy loss as the planet warms provides an estimate how Earth's energy balance responds to warming, which is simple enough to write on a small piece of paper. When scientists investigate the warming predicted by climate models, they often start from a similar but not identical calculation of how Earth's energy balance responds to warming. This calculation, based on model output, is about 15% less stabilizing than the simple pencil‐and‐paper estimate. In this paper, we explore the causes of this 15% difference between the pencil‐and‐paper estimate and the calculations using climate models. We show that the difference is primarily caused by the lack of assumed warming in climate models high in Earth's atmosphere, where temperatures are not closely linked to surface warming. This lack of warming acts as a hidden destabilizing feedback in current analysis of climate models. Key Points: Earth's reference radiative response, or "Planck feedback," is ∼0.5 W m−2 K−1 less stabilizing than a Stefan‐Boltzmann estimateWe find this deviation is mostly due to the assumed lack of stratospheric warming in calculations of the Planck feedbackThe lack of stratospheric warming serves as an implicit positive feedback in analysis of climate model warming [ABSTRACT FROM AUTHOR]

When the negatively buoyant air in the cloud downdrafts reaches the surface, it spreads out horizontally, producing cold pools. A cold pool can trigger new convective cells. However, when combined with the ambient vertical wind shear, it can also connect and upscale them into large mesoscale convective systems (MCS). Given the broad spectrum of scales of the atmospheric phenomenon involving the interaction between cold pools and the MCS, a parameterization was designed here. Then, it is coupled with a classical convection parameterization to be applied in an atmospheric model with an insufficient spatial resolution to explicitly resolve convection and the sub‐cloud layer. A new scalar quantity related to the deficit of moist static energy detrained by the downdrafts mass flux is proposed. This quantity is subject to grid‐scale advection, mixing, and a sink term representing dissipation processes. The model is then applied to simulate moist convection development over a large portion of tropical land in the Amazon Basin in a wet and dry‐to‐wet 10‐days period. Our results show that the cold pool edge parameterization improves the organization, longevity, propagation, and severity of simulated MCS over the Amazon and other different continental areas. Plain Language Summary: In nature, cold pools are formed by cold air masses descending from the low to mid‐troposphere in thunderstorms. When these drafts reach the surface, they spread out horizontally. A manifestation of cold pools is the relatively high speed at the gust front, which can lift environmental air producing new convective cells. Moreover, depending on ambient conditions, the cold pools may help organize the new convective cells, increasing their aggregation and forming the so‐called mesoscale convective systems (MCSs). MCSs, which cover hundreds to thousands of km2, significantly impact the global scale circulation, energy budget, hydrological cycle, and population safety. Forecasting MCSs is challenging for global circulation models (GCM) due to the broad spectrum of scales of the involved atmospheric phenomenon. The computational limitations, at present and for some time to come, do not allow running in real‐time GCMs, which explicitly solves all relevant scales of motion. This paper describes a methodology to account for essential interplays between cold pools edges and moist convection to be applied in the GCMs of weather and climate forecasting. We show that the method improves the model simulation of the main types of MCSs over the Amazon Basin and other continental areas. Key Points: A model scheme for including effect of cold pools edges in triggering new convective cells and storm propagation is presentedThe scheme is coupled with a convection parameterization and applied in the modeling of moist convective systemsThe method improves the organization, longevity, propagation, and severity of the simulated mesoscale convective systems [ABSTRACT FROM AUTHOR]

The objective of this study is to evaluate the potential for History Matching (HM) to tune a climate system with multi‐scale dynamics. By considering a toy climate model, namely, the two‐scale Lorenz96 model and producing experiments in perfect‐model setting, we explore in detail how several built‐in choices need to be carefully tested. We also demonstrate the importance of introducing physical expertise in the range of parameters, a priori to running HM. Finally we revisit a classical procedure in climate model tuning, that consists of tuning the slow and fast components separately. By doing so in the Lorenz96 model, we illustrate the non‐uniqueness of plausible parameters and highlight the specificity of metrics emerging from the coupling. This paper contributes also to bridging the communities of uncertainty quantification, machine learning and climate modeling, by making connections between the terms used by each community for the same concept and presenting promising collaboration avenues that would benefit climate modeling research. Plain Language Summary: Climate models are computer simulation codes that incorporate centuries of human knowledge of the physics of planet Earth. They are used to understand the past, the present and make projections about the future of our climate. To validate a climate model, scientists tune a number of its parameters so that it yields a simulated climate resembling real‐life observations as much as possible. The main challenge in this tuning task is the extreme cost of climate models which limits a lot the number of tuning experiments scientists can run. In this paper we are interested in a technique that uses artificial intelligence in order to replace the expensive climate model with a cheaper surrogate. We experiment on a simplified model to assess the strengths and weaknesses of this semi‐automatic technique, and show that it can be more efficient when combined with human expertise. Key Points: The History Matching method is explained in detail then used for tuning a toy coupled model: the Lorenz 96 modelThe importance of several design choices is demonstrated, especially when considering forced experiments such as Atmospheric Model Intercomparison Protocol and Ocean Model Intercomparison ProjectWe argue that this tuning method is semi‐automatic & highlight the importance of human expertise when considering it for real coupled models [ABSTRACT FROM AUTHOR]

In this paper, we introduce a testbed for evaluating and comparing climate modeling systems at cloud resolving scales using hindcasts of the June 2012 North American derecho. To demonstrate its utility for model intercomparison, the testbed is applied to two models: the regionally‐refined Simple Cloud‐Resolving E3SM Atmosphere Model (SCREAM) at 6.5, 3.25 and 1.625 km grid spacing and the Weather Research and Forecasting (WRF) model with 3.2 and 1.6 km grid spacing. We find the simulation results to be highly sensitive to the initial conditions (ICs), initialization time, and model configurations, with ICs from the Rapid Refresh producing the best simulation. Significant improvement is identified in both models as horizontal grid spacing is refined. While a propagation delay of approximately 2 hr is found in both models, SCREAM at 1.625 km simulates the observed bow echo structure of the derecho well and predicts strong surface gusts that exceed 30 m/s. In comparison, WRF has difficulty producing surface wind over 25 m/s, with wind gusts in WRF 42%–46% lower than in SCREAM. However, WRF has a lower bias in simulating cloud top temperature and extent, but overestimates precipitation intensity. Both models reproduce the observed outgoing longwave radiation spatial patterns well (Pearson correlation >0.88), but, compared with NEXRAD observations, simulate generally larger areas of composite radar reflectivity >40 dBZ and underestimate the precipitating area by ∼47%. Plain Language Summary: This paper describes a testbed for evaluating model performance on a particular high‐impact weather event—the June 2012 North American derecho, a storm event associated with extreme winds and precipitation. The testbed is applied to evaluate the Simple Cloud‐Resolving E3SM Atmosphere Model (SCREAM) and the Weather Research and Forecasting (WRF) model at resolutions that resolve cloud systems. The performance of both models is shown to be sensitive to the data set used for model initialization. Finer grid resolution generally leads to better model performance. All simulations show a 2‐hr delay in predicting the evolution of the derecho and produce more intense rainfall. SCREAM generates a more realistic convective front than WRF and produces stronger surface winds. The evaluation protocol can be used to better understand the credibility of model simulations of extreme events and guide model development. Key Points: A testbed of observational products, diagnostics, and metrics is constructed to evaluate hindcasts of the June 2012 North American derechoHindcast results are sensitive to initial conditions, initialization time, horizontal resolutions, and convective and microphysics schemesThe Simple Cloud‐Resolving E3SM Atmosphere Model at 1.625 km successfully reproduces severe surface gusts with wind speeds above 30 m/s [ABSTRACT FROM AUTHOR]

Numerical weather prediction is pushing the envelope of grid resolution at local and global scales alike. Aiming to model topography with higher precision, a handful of articles introduced unstructured vertical grids and tested them for dry atmospheres. The next step toward effective high‐resolution unstructured grids for atmospheric modeling requires that also microphysics is independent of any vertical columns, in contrast to what is ubiquitous across operational and research models. In this paper, we present a non‐column based continuous and discontinuous spectral element implementation of Kessler's microphysics with warm rain. We test the proposed algorithm against standard three‐dimensional benchmarks for precipitating clouds and show that the results are comparable with those presented in the literature across all of the tested effective resolutions. While presented for both continuous and discontinuous spectral elements in this paper, the method that we propose can be adapted to any numerical method used in other codes, as long as the code can already handle vertically unstructured grids. Plain Language Summary: The earth climate is warming faster than ever. While climate models are the tool available to scientists to forecast its future evolution, they are biased by uncertainties that are, arguably, mostly embedded in the modeling of clouds. Thanks to the advent of exascale computing, a reduction of cloud modeling uncertainties can be expected by simulating clouds at higher and higher resolutions. While uniform high resolution across the whole domain is ideal, for computational efficiency reasons scientists are likely to increase the model resolution in some regions more than others not only in the horizontal direction—which is a standard approach—but also along the vertical direction. Grid refinement in the vertical direction, however, may lead to the loss of the vertical structure of the grid columns, affecting the usability of column‐based physics packages that are used to model clouds and precipitation. To overcome this problem, we present an algorithm to solve the equations that model precipitating clouds along arbitrarily shaped grids in any spatial direction. This approach is advantageous from a modeling perspective as well as from a computational one because it allows full flexibility of the domain partitioning algorithms when hundreds of thousands of parallel processors are used. Key Points: A non‐column based spectral element approach to model precipitating clouds is presentedThe approach can be easily applied to any microphysics scheme with and without precipitationThis approach will make full physics simulations of the atmosphere on unstructured grids possible [ABSTRACT FROM AUTHOR]

Single‐model initial‐condition large ensembles are powerful tools to quantify the forced response, internal climate variability, and their evolution under global warming. Here, we present the CMIP6 version of the Max Planck Institute Grand Ensemble (MPI‐GE CMIP6) with currently 30 realizations for the historical period and five emission scenarios. The power of MPI‐GE CMIP6 goes beyond its predecessor ensemble MPI‐GE by providing high‐frequency output, the full range of emission scenarios including the highly policy‐relevant low emission scenarios SSP1‐1.9 and SSP1‐2.6, and the opportunity to compare the ensemble to complementary high‐resolution simulations. First, we describe MPI‐GE CMIP6, evaluate it with observations and reanalyzes and compare it to MPI‐GE. Then, we demonstrate with six application examples how to use the power of the ensemble to better quantify and understand present and future climate extremes, to inform about uncertainty in approaching Paris Agreement global warming limits, and to combine large ensembles and artificial intelligence. For instance, MPI‐GE CMIP6 allows us to show that the recently observed Siberian and Pacific North American heatwaves would only avoid reaching 1–2 years return periods in 2071–2100 with low emission scenarios, that recently observed European precipitation extremes are captured only by complementary high‐resolution simulations, and that 3‐hourly output projects a decreasing activity of storms in mid‐latitude oceans. Further, the ensemble is ideal for estimates of probabilities of crossing global warming limits and the irreducible uncertainty introduced by internal variability, and is sufficiently large to be used for infilling surface temperature observations with artificial intelligence. Plain Language Summary: Climate model simulations that start from different initial states and differ only due to the chaos in the climate system are used extensively to quantify the forced climate response, variability intrinsic to the climate system, and their change under global warming. Here, we present a new version of the Max Planck Institute Grand Ensemble (MPI‐GE CMIP6) that is run as part of the latest generation of climate models. This single‐model ensemble currently consists of 30 realizations for the historical period 1850–2014 and for five scenarios of possible future climates until 2100. The power of MPI‐GE CMIP6 goes beyond its predecessor by not only providing monthly mean but also 3‐hourly to daily model output, the full range of future scenarios including the two highly policy‐relevant scenarios that were designed to match the Paris Agreement global warming limits of 1.5 and 2°C, and the opportunity to compare the low‐resolution ensemble to simulations of the same model version with higher horizontal resolution. In this paper, we describe the new ensemble and demonstrate with application examples how to use its power. Key Points: The Max Planck Institute Grand Ensemble in its CMIP6 version (MPI‐GE CMIP6) is a 30‐member initial‐condition large ensemble with up to 3‐hourly model output and five emission scenariosThe ensemble is specifically suited to investigate climate extremes and Paris Agreement global warming limitsMPI‐GE CMIP6 adequately represents heat extremes, while precipitation extremes are captured by complementary high‐resolution simulations [ABSTRACT FROM AUTHOR]

Clouds, especially low clouds, are crucial for regulating Earth's energy balance and mediating the response of the climate system to changes in greenhouse gas concentrations. Despite their importance for climate, they remain relatively poorly understood and are inaccurately represented in climate models. A principal reason is that the high computational expense of simulating them with large‐eddy simulations (LES) has inhibited broad and systematic numerical experimentation and the generation of large data sets for training parametrization schemes for climate models. Here we demonstrate LES of low clouds on tensor processing units (TPUs), application‐specific integrated circuits that were originally developed for machine learning applications. We show that TPUs in conjunction with tailored software implementations can be used to simulate computationally challenging stratocumulus clouds in conditions observed during the Dynamics and Chemistry of Marine Stratocumulus (DYCOMS) field study. The TPU‐based LES code successfully reproduces clouds during DYCOMS and opens up the large computational resources available on TPUs to cloud simulations. The code enables unprecedented weak and strong scaling of LES, making it possible, for example, to simulate stratocumulus with 10× speedup over real‐time evolution in domains with a 34.7 km × 53.8 km horizontal cross section. The results open up new avenues for computational experiments and for substantially enlarging the sample of LES available to train parameterizations of low clouds. Plain Language Summary: The study of clouds has been impeded by, among other factors, limitations in our ability to simulate them rapidly and on sufficiently large domains. In particular, computational limitations in simulating low clouds are among the reasons for the difficulties of representing them accurately in climate models; this is one of the dominant uncertainties in climate predictions. This paper demonstrates how the large computing power available on tensor processing units (TPUs) (integrated circuits originally designed for machine learning applications) can be harnessed for simulating low clouds. We demonstrate the largest simulations of low clouds to date, with hundreds of billions of variables, and we document their fidelity to aircraft observations. The results open up the large computational resources available on TPUs, hitherto primarily used for machine learning, to the study of clouds in the climate system. Key Points: We introduce a large‐eddy simulation (LES) framework that runs on tensor processing units (TPUs, accelerators designed for machine learning)The fidelity of the LES is established by reproducing aircraft observations of nocturnal stratocumulus clouds over the PacificThe LES exhibit unprecedented scalability on TPUs, enabling the large‐scale generation of training data for cloud parameterizations [ABSTRACT FROM AUTHOR]

An idealized aquaplanet moist global atmospheric model with realistic radiative transfer but no clouds and no convective parameterization is found to possess multiple climate equilibria. When forced symmetrically about the equator, in some cases the Inter Tropical Convergence Zone (ITCZ) migrates to an off‐equatorial equilibrium position. Mechanism denial experiments prescribing relative humidity imply that radiation‐circulation coupling is essential to this instability. The cross‐equatorial asymmetry occurs only when the underlying slab ocean is sufficiently deep and the atmosphere's spectral dynamical core is sufficiently coarse (∼T170 or less with our control parameters). At higher resolutions, initializing with an asymmetric state indicates metastability with very slow (thousands of days) return to hemispheric symmetry. There is some sensitivity to the model timestep, which affects the time required to transition to the asymmetric state, with little effect on the equilibrium climate. The instability is enhanced when the planetary boundary layer scheme favors deeper layers or by a prescribed meridional heat transport away from the equator within the slab. The instability is not present when the model is run with a convective parameterization scheme commonly utilized in idealized moist models. We argue that the instability occurs when the asymmetric heating associated with a spontaneous ITCZ shift drives a circulation that rises poleward of the perturbed ITCZ. These results serve as a warning of the potential for instability and non‐uniqueness of climate that may complicate studies with idealized models of the tropical response to perturbations in forcing. Plain Language Summary: Much of the rainfall in the tropics occurs in the Intertropical Convergence Zones (ITCZs). These regions are associated with the convergence of surface winds, which produces the rising air required to feed the Hadley cells. The ITCZ location is known to be determined by the spatial distribution of atmospheric heating as the Hadley cells act to erase thermal imbalance within the tropics. With hemispherically symmetric heating, a symmetric equatorial ITCZ is expected. However, it is shown in this paper that changes in the ITCZ location affect the distribution of the radiative cooling of the atmosphere through changes in the water vapor distribution, which may make the equatorial ITCZ unstable under some circumstances. In this situation, two alternative climates with the ITCZ shifted into either one of the hemispheres exist. This instability may play an important role for understanding the distribution of tropical precipitation and the potential for the existence of multiple climates for the same set of external conditions. Key Points: The equatorial ITCZ can be destabilized by radiation‐circulation coupling in a hemispherically symmetric aquaplanet modelAs the ITCZ drifts off the equator two different asymmetric equilibria may exist under symmetric forcing and boundary conditionsA theory is proposed based on the energy flux equator sensitivity to the perturbed asymmetric heating when the ITCZ moves off the equator [ABSTRACT FROM AUTHOR]

Parameterization of mesoscale eddies in coarse resolution ocean models is necessary to include the effect of eddies on the large‐scale oceanic circulation. We propose to use a multiple‐scale Quasi‐Geostrophic (MSQG) model to capture the eddy dynamics that develop in response to a prescribed large‐scale flow. The MSQG model consists in extending the traditional quasi geostrophic (QG) dynamics to include the effects of a variable Coriolis parameter and variable background stratification. Solutions to this MSQG equation are computed numerically and compared to a full primitive equation model. The large‐scale flow field permits baroclinically unstable QG waves to grow. These instabilities saturate due to non‐linearities and a filtering method is applied to remove large‐scale structures that develop due to the upscale cascade. The resulting eddy field represents a dynamically consistent response to the prescribed background flow, and can be used to rectify the large‐scale dynamics. Comparisons between Gent‐McWilliams eddy parameterization and the present solutions show large regions of agreement, while also indicating areas where the eddies feed back onto the large scale in a manner that the Gent‐McWilliams parameterization cannot capture. Also of interest is the time variability of the eddy feedback which can be used to build stochastic eddy parameterizations. Plain Language Summary: Climate models are running at horizontal spatial resolutions that capture only the basin‐scale dynamics. This is a strong limitation because the missing mesoscale eddies have a non‐negligible impact on the large‐scale flow. The problem of including the effect of eddies in climate models is a long standing issue in the ocean modeling community. In this work, we decompose the ocean dynamics into stationary large‐scale dynamics and turbulent small‐scale dynamics. One can thus write a system of equations for the large‐scale dynamics—which can be interpreted as the ocean dynamics given by climate models—and another system of equations for the small‐scale dynamics. The numerical implementation of this model reveals that this reduced system of equations faithfully reproduces the dynamics of a high resolution ocean model. This paper is a first step that will eventually guide us to use this model as a parameterization of unresolved scales in low resolution ocean models. Key Points: We implement a numerical version of a multiple‐scale Quasi‐Geostrophic (MSQG) modelMultiple Scale Quasi Geostrophy allows for variable stratification and variable Coriolis parameterThe MSQG model reproduces the dynamics of a high resolution primitive equation model [ABSTRACT FROM AUTHOR]

One of the most important components of an atmospheric radiation scheme is its treatment of gas optical properties, which determines not only the accuracy of its radiative forcing calculations fundamental to climate prediction, but also its computational cost. This paper describes a free software tool "ecCKD" for generating fast gas‐optics models by optimally dividing the spectrum into pseudo‐monochromatic spectral intervals (known as k‐terms) according to a user‐specified error tolerance and the range of greenhouse‐gas concentrations that needs to be simulated. The models generated use the correlated k‐distribution method in user‐specified bands, but can also generate accurate "full‐spectrum correlated‐k" models that operate on the entire longwave or near‐infrared (NIR) parts of the spectrum. In the NIR, the large spectral variation in cloud absorption is represented by partitioning the parts of the spectrum where gases are optically thin into 2–6 sub‐bands, while allowing k‐terms for the optically thicker parts of the spectrum (where clouds and surface reflectance are less important) to span the entire NIR spectrum. Candidate models using only 16 and 32 k‐terms in each of the shortwave and longwave are evaluated against line‐by‐line calculations on clear and cloudy profiles. The 32‐term models are able to accurately capture the radiative forcing of varying greenhouse gases including CO2 concentrations spanning a factor of 12, and heating rates at pressures down to 1 Pa. Plain Language Summary: A crucial component of atmospheric computer models used to make climate projections and weather forecasts is the "gas optics scheme," which represents the interaction of sunlight and infrared radiation with greenhouse gases. This paper describes a free software tool "ecCKD" that uses a number of novel techniques to generate new gas optics schemes that are computationally faster than most existing schemes while still being very accurate. For example, the schemes are able to simulate variations in carbon dioxide concentration spanning a factor of 12 and methane concentration spanning a factor of 10. Users of ecCKD can generate schemes that are optimized for specific applications, such as short‐term weather forecasting or simulating past climates. A special focus has been placed on the near‐infrared part of the solar spectrum to ensure that the schemes work well when computing the interactions of sunlight simultaneously with gases and clouds, important to ensure that the impact of clouds on weather and climate is well simulated. Key Points: We describe a free software tool for generating accurate gas optics models for radiation schemes and test with line‐by‐line calculationsExtra efficiency is achieved via the use of the full‐spectrum correlated‐k method in the thermal‐ and near‐infrared (NIR)The spectral properties of clouds are treated accurately via the use of sub‐bands in the NIR and several other techniques [ABSTRACT FROM AUTHOR]

Process‐scale development, evaluation, and calibration of physically based parameterizations of clouds and radiation are powerful levers for improving weather and climate models. In a series of papers, we propose a strategy for process‐based calibration of climate models that uses machine learning techniques. It relies on systematic comparisons of single‐column versions of climate models with explicit simulations of boundary‐layer dynamics and clouds (Large‐Eddy Simulations [LES]). This paper focuses on the calibration of cloud geometry parameters (vertical overlap, horizontal heterogeneity, and cloud size) that appear in the parameterization of radiation. The solar component of a radiative transfer (RT) scheme that includes a parameterization for 3D radiative effects of clouds (SPARTACUS) is run in offline single‐column mode on an ensemble of input cloud profiles synthesized from LES outputs. The space of cloud geometry parameter values is efficiently explored by sampling a large number of parameter sets (configurations) from which radiative metrics are computed using fast surrogate models that emulate the SPARTACUS solver. The sampled configurations are evaluated by comparing these radiative metrics to reference values provided by a 3D RT Monte Carlo model. The best calibrated configurations yield better predictions of TOA and surface fluxes than the one that uses parameter values computed from the 3D cloud fields: The root‐mean‐square errors averaged over cumulus cloud fields and solar angles are reduced from ∼10 Wm−2 with LES‐derived parameters to ∼5 Wm−2 with adjusted parameters. However, the calibration of cloud geometry fails to reduce the errors on absorption, which remain around 2–4 Wm−2. Plain Language Summary: A way to improve the accuracy of climate models is to improve the physical formulations that represent the effects of small‐scale processes on the evolution of atmospheric state. Processes that involve clouds and radiation are particularly important due to their key role on climate. Choosing values for the parameters inherent to these formulations is a difficult task. This series of papers presents a rigorous strategy for calibrating models. It is based on comparisons between high‐resolution models that accurately represent clouds and single‐column versions of a climate model, on the basis of process‐oriented metrics such as cloud height. A set of acceptable parameters is efficiently found using machine learning techniques. In this third part, the parameters that control the radiative effects of cloud geometry are calibrated. A recent radiation model that includes realistic representation of the radiative effect of cloud heterogeneity, cloud vertical structure and cloud size is evaluated and calibrated using references that are provided by a ray‐tracing algorithm that tracks virtual photons in virtual cloud fields produced by high‐resolution models (Large‐Eddy Simulations [LES]). Calibration improves the model with respect to using parameters diagnosed in the LES. Good agreement is found only when interception of sunlight by cloud sides is represented. Key Points: A novel calibration approach is applied to an offline radiation scheme to disentangle sources of uncertainty in cloud radiative effectsThe SPARTACUS solver is run on cloud profiles derived from Large‐Eddy Simulations (LES) cumulus fields and compared to Monte Carlo 3D radiative transfer computationsAdjusting SPARTACUS cloud geometry parameters provides effective values that improve surface and TOA fluxes compared to LES‐derived values [ABSTRACT FROM AUTHOR]

It is difficult for coarse‐resolution global models of the atmosphere to accurately simulate the observed distribution of low clouds. In particular, it is difficult for moist turbulence closure models to simulate sufficiently bright near‐coastal stratocumulus (Sc) without simulating overly bright marine shallow cumuli (Cu). To parameterize bright Sc, a turbulence parameterization must damp the turbulent fluxes of heat and moisture above cloud top in order to prevent excessive entrainment of dry air into cloud top. To parameterize dim shallow Cu, the subgrid variances of temperature and moisture must remain large, in order to permit partial cloudiness. However, damping the fluxes but not the variances just above cloud top is difficult if a parameterization uses a single "master" time scale to damp both. In nature, the above‐cloud fluxes are damped by pressure fluctuations, whereas scalar variances are damped by a different process, namely, turbulent dissipation. In a stably stratified inversion above cloud, pressure damping is large but turbulent dissipation is small. To avoid this problem, a multitime scale parameterization for damping has been developed. The damping parameterization has been implemented in a global model and evaluated. The parameterization is capable of dimming shallow Cu while producing adequately bright Sc. Plain Language Summary: This paper describes an assumption in turbulence modeling that is known to be questionable but is still sometimes made. To avoid making this assumption, the formulation of a particular turbulence model is generalized. The generalized formulation, when implemented in a global model of the atmosphere, changes the pattern of low‐altitude clouds. Key Points: Some turbulence parameterizations use a single master turbulence length/time scale even though this assumption is an oversimplificationThis paper parameterizes turbulence by use of multiple turbulent time scales, including one for dissipation and another for pressureUse of the new multiscale parameterization in a global model significantly alters the distribution of stratocumulus and shallow cumulus clouds [ABSTRACT FROM AUTHOR]

Semi‐implicit (SI) time‐stepping schemes for atmosphere and ocean models require elliptic solvers that work efficiently on modern supercomputers. This paper reports our study of the potential computational savings when using mixed precision arithmetic in the elliptic solvers. Precision levels as low as half (16 bits) are used and a detailed evaluation of the impact of reduced precision on the solver convergence and the solution quality is performed. This study is conducted in the context of a novel SI shallow‐water model on the sphere, purposely designed to mimic numerical intricacies of modern all‐scale weather and climate (W&C) models. The governing algorithm of the shallow‐water model is based on the non‐oscillatory MPDATA methods for geophysical flows, whereas the resulting elliptic problem employs a strongly preconditioned non‐symmetric Krylov‐subspace Generalized Conjugated‐Residual (GCR) solver, proven in advanced atmospheric applications. The classical longitude/latitude grid is deliberately chosen to retain the stiffness of global W&C models. The analysis of the precision reduction is done on a software level, using an emulator, whereas the performance is measured on actual reduced precision hardware. The reduced‐precision experiments are conducted for established dynamical‐core test‐cases, like the Rossby‐Haurwitz wavenumber 4 and a zonal orographic flow. The study shows that selected key components of the elliptic solver, most prominently the preconditioning and the application of the linear operator, can be performed at the level of half precision. For these components, the use of half precision is found to yield a speed‐up of a factor 4 compared to double precision for a wide range of problem sizes. Plain Language Summary: Numerical models of the Earth system are very important to predict weather and climate (W&C). The models are computationally expensive and run on supercomputers. If the models are enabled to run faster, model simulations can be enhanced—for example, via a higher resolution—and the accuracy of predictions can be improved. One way to enable the models to run faster is via the use of very low numerical precision. For example, via the use of so‐called half precision which is representing all variables of the simulation with only 16 bits, instead of the default value of 64 bits, reducing computational time by up to a factor of four. In this paper, we investigate whether numerical precision can be reduced down to half precision for parts of the linear solvers which are used in some W&C models. Linear solvers help to progress the model state in time and are responsible for a large fraction of the overall computational cost of model simulations. Key Points: A detailed study of mixed‐precision for linear solvers for weather and climate models is performed, including half precision with 16 bitsPrecision can be reduced in important parts of the solver but a naive approach to reduce precision everywhere does not workCompute time can be reduced by a factor four compared to double precision for parts that work in half precision [ABSTRACT FROM AUTHOR]

A new version of the stochastic multiplume Jet Propulsion Laboratory Eddy‐Diffusivity/Mass‐Flux (JPL‐EDMF) parameterization which consistently couples the simplified Khairoutdinov and Kogan (2000), https://doi.org/10.1175/1520-0493(2000)128<0229:ANCPPI>2.0.CO;2, warm phase cloud microphysical parameterization with the parameterization of cloud macrophysical and subgrid scale dynamical processes is described. The new parameterization combines the EDMF approach with an assumed shape of a joint probability density function of thermodynamic and kinematic variables which provide the basis for the computation of all parameterized processes. As far as we are aware this is the first attempt to consistently couple all of these parameterized processes in the EDMF framework. This paper is part one of a two paper series. Here, the JPL‐EDMF parameterization is described and benchmark simulations of precipitating stratocumulus and cumulus convection are performed in a single‐column‐model framework. The parameterization results compare favorably to the reference large‐eddy‐simulation results. In the second part (Smalley et al., 2022, https://doi.org/10.1029/2021MS002729) the JPL‐EDMF parameterization is validated for a wide range of observation‐based scenarios covering the continuous transition from subtropical stratocumulus to cumulus convection derived from global reanalysis, and parameterization uncertainties are studied in detail. Plain Language Summary: Parameterizations are components of atmospheric models that represent the impact of unresolved processes on the resolved scale flow and are crucial for shaping simulated dynamics and thermodynamics. In many atmospheric models, multiple parameterizations are used in parallel and each of them represents subgrid processes only in part, or only for a certain atmospheric regime. Recent parameterization developments focus on unified approaches which represent key unresolved processes in a mathematically consistent framework that allows to capture important interactions among unresolved processes, which is hard to achieve with modularized approaches. We describe a new unified parameterization which represents subgrid‐scale dynamical, cloud and rain processes, thus removing key shortcomings of traditional parameterizations. In this new parameterization, the subgrid‐scale dynamical processes are represented by the stochastic multiplume Eddy‐Diffusivity/Mass‐Flux framework from our previous works. Cloud and rain processes are represented by well‐known parameterizations developed for high‐resolution models. The key features of the new parameterization are a set of consistent assumptions about subgrid variability which is used to evaluate grid mean rates of these three processes, and representation of interactions between them. Our new parameterization is tested against benchmark simulations of marine stratocumulus and cumulus convection and the results compare favorably to the reference large‐eddy‐simulation results. Key Points: We describe a novel unified parameterization that represents interactive subgrid scale dynamical and warm cloud and rain processesThe new parameterization is based on a stochastic multiplume Eddy‐Diffusivity/Mass‐Flux approachThe parameterization successfully reproduces benchmark cases of precipitating marine cumulus and stratocumulus convection [ABSTRACT FROM AUTHOR]

This paper presents the response to anthropogenic forcing in the GISS‐E2.1 climate models for the 21st century Shared Socioeconomic Pathways emission scenarios within the Coupled Model Intercomparison Project Phase 6 (CMIP6). The experiments were performed using an updated and improved version of the NASA Goddard Institute for Space Studies (GISS) coupled general circulation model that includes two different versions for atmospheric composition: A non‐interactive version (NINT) with prescribed composition and a tuned aerosol indirect effect and the One‐Moment Aerosol model (OMA) version with fully interactive aerosols which includes a parameterized first indirect aerosol effect on clouds. The effective climate sensitivities are 3.0°C and 2.9°C for the NINT and OMA models, respectively. Each atmospheric version is coupled to two different ocean general circulation models: The GISS ocean model (E2.1‐G) and HYCOM (E2.1‐H). We describe the global mean responses for all future scenarios and spatial patterns of change for surface air temperature and precipitation for four of the marker scenarios: SSP1‐2.6, SSP2‐4.5, SSP4‐6.0, and SSP5‐8.5. By 2100, global mean warming ranges from 1.5°C to 5.2°C relative to 1,850–1,880 mean temperature. Two high‐mitigation scenarios SSP1‐1.9 and SSP1‐2.6 limit the surface warming to below 2°C by the end of the 21st century, except for the NINT E2.1‐H model that simulates 2.2°C of surface warming. For the high emission scenario SSP5‐8.5, the range is 4.6–5.2°C at 2100. Due to about 15% larger effective climate sensitivity and stronger transient climate response in both NINT and OMA CMIP6 models compared to CMIP5 versions, there is a stronger warming by 2100 in the SSP emission scenarios than in the comparable Representative Concentration Pathway (RCP) scenarios in CMIP5. Changes in sea ice area are highly correlated to global mean surface air temperature anomalies and show steep declines in both hemispheres, with the largest sea ice area decreases occurring during September in the Northern Hemisphere in both E2.1‐G (−1.21 × 106 km2/°C) and E2.1‐H models (−0.94 × 106 km2/°C). Both coupled models project decreases in the Atlantic overturning stream function by 2100. The largest decrease of 56%–65% in the 21st century overturning stream function is produced in the warmest scenario SSP5‐8.5 in the E2.1‐G model, comparable to the reduction in the corresponding CMIP5 GISS‐E2 RCP8.5 simulation. Both low‐end scenarios SSP1‐1.9 and SSP1‐2.6 also simulate substantial reductions of the overturning (9%–37%) with slow recovery of about 10% by the end of the 21st century (relative to the maximum decrease at the middle of the 21st century). Plain Language Summary: The projections of future climate change are uncertain because they are dependent on different possible scenarios of human‐caused emissions and their interaction with natural forcings, internal climate variability, and inter‐model differences. This paper presents the results of the climate model of the NASA Goddard Institute for Space Studies, GISS‐E2.1, for the anthropogenically forced climate response for the twenty first century Shared Socioeconomic Pathways emission scenarios within the Coupled Model Intercomparison Project Phase 6 (CMIP6). The sensitivity of the model response to different magnitudes of anthropogenic forcings in the twenty‐first‐century scenarios were performed using two different versions for the atmospheric composition and two different ocean general circulation models. Compared to CMIP5 GISS‐E2 versions, the CMIP6 GISS‐E2.1 climate model shows a stronger warming by 2100 in comparable scenarios due to larger effective climate sensitivity and transient climate response. Both climate models with two different ocean components project decreases in the Atlantic overturning stream function by 2100. Key Points: GISS E2.1 model with different configurations is used to carry out 134 Shared Socioeconomic Pathway (SSP) experimentsGISS‐E2.1 climate model shows a stronger warming by 2,100 in comparable Representative Concentration Pathway scenarios in CMIP5 due to larger effective climate sensitivity and stronger transient climate responseBoth coupled models, E2.1‐G and E2.1‐H, project decreases in the Atlantic overturning stream function by 2100 with the largest decrease in the warmest scenario SSP5‐8.5 in the E2.1‐G model [ABSTRACT FROM AUTHOR]

We develop a doubly periodic version of the Simple Convection‐Permitting E3SM Atmosphere Model (SCREAM) to provide an efficient configuration for this global convection permitting model (GCPM), akin to a single column model often found in conventional general circulation models. The design details are explained, in addition to the extensive case library associated with the doubly periodic SCREAM (DP‐SCREAM) configuration. We demonstrate that doubly periodic cloud resolving models are useful tools to explore the horizontal resolution sensitivity of GCPMs, in addition to replicating biases seen in the global models. Using DP‐SCREAM, we show that SCREAM exhibits behaviors of a scale aware model as it is able to naturally partition between sub‐grid scale (SGS) and resolved vertical transport across the gray zone of turbulence. We show that SCREAM is reasonably scale insensitive when run at resolutions from 1 to 5 km, but can exhibit sensitivity, particularly for the shallow convective regime, when run at resolutions approaching that of large eddy simulations. We conclude that SGS parameterization improvements are likely needed to reduce this scale sensitivity. Plain Language Summary: Advances in computational resources have allowed climate simulations to be performed with very high resolution, which provides higher quality results. However, these simulations require a lot of time and computer resources to perform, which makes these models difficult to use for the common scientist. In this paper we develop a high‐resolution configuration which focuses on a specific point on the globe, enabling it to run fast and to use minimal computational resources. This allows users and developers to gauge how the model may perform before doing a computationally intensive global simulation. We show that this faster configuration is a useful tool to replicate problems that are found in the global model and is a valuable way to assess sensitivities of the model, particularly pertaining to choices made in its resolution. Key Points: Doubly periodic configurations represent an efficient framework to supplement global cloud resolving modelsShallow convection is sensitive to horizontal resolution in Simple Convection‐Permitting E3SM Atmosphere Model (SCREAM) but stratiform clouds are more robustSCREAM exhibits the behaviors of a scale aware model and can represent a range of cloud regimes [ABSTRACT FROM AUTHOR]

This paper describes an implementation of the combined hybrid‐parallel prediction (CHyPP) approach of Wikner et al. (2020), https://doi.org/10.1063/5.0005541 on a low‐resolution atmospheric global circulation model (AGCM). The CHyPP approach combines a physics‐based numerical model of a dynamical system (e.g., the atmosphere) with a computationally efficient type of machine learning (ML) called reservoir computing to construct a hybrid model. This hybrid atmospheric model produces more accurate forecasts of most atmospheric state variables than the host AGCM for the first 7–8 forecast days, and for even longer times for the temperature and humidity near the earth's surface. It also produces more accurate forecasts than a model based only on ML, or a model that combines linear regression, rather than ML, with the AGCM. The potential of the CHyPP approach for climate research is demonstrated by a 10‐year long hybrid model simulation of the atmospheric general circulation, which shows that the hybrid model can simulate the general circulation with substantially smaller systematic errors and more realistic variability than the host AGCM. Plain Language Summary: This paper presents a computationally efficient novel approach to construct a hybrid model of the atmosphere by combining a physics‐based model of the global atmospheric circulation with a machine learning component. The primary purpose of the hybrid model is to produce quantitative weather forecasts on the same grid as the physics‐based model. It is found that the hybrid model produces more accurate forecasts than the host physics‐based model for the first 7–8 forecast days for most forecast variables, and for even longer times for the temperature and humidity near the Earth's surface. Furthermore, the hybrid model is found to simulate the climate with substantially smaller systematic errors and more realistic temporal variability than the host model. Key Points: A hybrid model incorporating machine learning produces more accurate forecasts and more realistic climate than the host physics‐based modelThe hybrid model states are more realistically balanced and have substantially lower biases than the host modelThe hybrid model produces more realistic atmospheric variability than the host model at time scales shorter than about a week [ABSTRACT FROM AUTHOR]

Quantifying the microphysical process contributions to surface precipitation in numerical simulations can be challenging. This is due to the fact that many microphysical processes contribute to the formation and depletion of rain drops and there is almost always a spatial/temporal mismatch between where/when rain is formed and where/when it strikes the surface. In this work, we develop a tracing method that tracks the sources and sinks of raindrop mass and number as they are advected by the Weather Research and Forecasting model. Applying the method to an idealized squall line confirms that convective precipitation is dominated by warm rain processes (autoconversion and accretion) while stratiform precipitation is dominated by the melting of rimed and unrimed ice crystals. Sensitivity experiments in which the prescribed cloud drop number concentration is increased confirm the conventional wisdom that weakened autoconversion increases the fraction of raindrops originating from cold rain processes. The method also reveals that when applied to deep convection the Khairoutdinov and Kogan autoconversion scheme produces an excessive number of raindrops which are subsequently clipped in P3 microphysics to keep the rain size distribution within prescribed limits. This problem can mostly be mitigated by increasing the assumed radius for raindrops created by autoconversion. Plain Language Summary: Weather and climate models simulate precipitation by accounting for a multitude of processes that form and remove rain drops. Due to spatial and temporal lags between the processes and the consequent surface precipitation, it can be challenging to quantitatively link the former to the latter. This paper remedies the challenge with a tool that tracks processes contributions throughout the rain lifecycle. The tool's utility is demonstrated through an idealized squall line which produces precipitation through various microphysical pathways. Additionally, perturbations to model parameters and their consequent impact on the squall line are interpreted through the lens of the new tool. Key Points: A tool to trace the process contributions to surface precipitation is developedThe tool confirms dominance of warm processes in convective rain and cold processes in stratiform rainThe tool reveals that the autoconversion formulation produces an excessive amount of rain drops that the microphysics scheme clips [ABSTRACT FROM AUTHOR]

Convective aggregation is an important atmospheric phenomenon which frequently occurs in idealized models in radiative‐convective equilibrium (RCE), where the effects of land, rotation, sea surface temperature gradients, and the diurnal cycle are often removed. This aggregation is often triggered and maintained by self‐generated radiatively driven circulations, for which longwave feedbacks are essential. Many questions remain over how important the driving processes of aggregation in idealized models are in the real atmosphere. We approach this question by adding a continentally sized, idealized tropical rainforest island into an RCE model to investigate how land‐sea contrasts impact convective aggregation and its mechanisms. We show that convection preferentially forms over the island persistently in our simulation. This is forced by a large‐scale, thermally driven circulation. First, a sea‐breeze circulation is triggered by the land‐sea thermal contrast, driven by surface sensible heating. This sea‐breeze circulation triggers convection which then generates longwave heating anomalies. Through mechanism denial tests we find that removing the longwave feedbacks reduces the large‐scale effects of aggregation but does not prevent aggregation from occurring, and thus we highlight there must be another process aiding the aggregation of convection. We also show, by varying the island size, that the aggregated convective cluster appears to have a maximum spatial extent of O $\mathcal{O}$(10,000 km). These results highlight that the mechanisms of idealized aggregation remain relevant when land is included in the model, and therefore these mechanisms could help us understand convective organization in the real world. Plain Language Summary: Large tropical storm clouds can cluster together to form organized systems, which are associated with extreme precipitation within the cloudy region, and very dry conditions away from the cloudy region. These systems, called organized deep convection, are often studied in simplified models of the atmosphere where the land, Earth's rotation, and variations in sea‐surface temperature are removed. In this paper we take a step toward reality and look at how including a large, idealized island in a simplified model affects how convection clusters. We show that convective clouds group together over the land throughout our model simulations. We investigate the processes driving this and find that surface flux feedbacks dominate early in the simulations. The feedbacks maintaining the convective aggregation are more complex. We find there is an important role for the longwave feedbacks, but that there must be another necessary process also aiding the clustering process. Our results highlight that the processes in highly idealized modeling studies seem to be relevant in more realistic models, and thus could help us understand how convection organizes in the real world. However, we also conclude it is important that follow‐up work investigates the remaining additional feedbacks to gain a more complete understanding. Key Points: Convection preferentially aggregates over land in a global radiative‐convective equilibrium simulationA global land‐centered circulation drives the aggregation and is triggered through surface fluxes, but maintained through longwave fluxesThe land‐based convective cluster appears to have a maximum spatial scale of O $\mathcal{O}$(10,000 km) [ABSTRACT FROM AUTHOR]

As the climate system warms, megafires have become more frequent with devastating effects. A byproduct of these events is the creation of smoke plumes that can rise into the stratosphere and spread across the globe where they reside for many months. To gain a deeper understanding of the plume dynamics, global climate simulations of a megafire were performed at a wide range of grid spacings from 2.0° down to 7 km, including a 7 km nonhydrostatic experiment. The analysis focuses on how the resolved dynamics affects the specification of the plume characteristics such as injection height and black carbon (BC) mass. Prior studies initialize the smoke plume at one or a few grid points and this is shown here to produce severely dissipative dynamics. In order to validate such simulations with observations, enhancements of the plume characteristics to offset the dissipation is necessary. Using a numerically converged simulation, sensitivity tests show that to approximate the observed stratospheric lifetime, a reduction in BC fraction by 50% is necessary for external mixtures. The vorticity dynamics of the plume is also analyzed with a Lagrangian budget to understand the mechanisms responsible for the evolution of a collocated anticyclonic vortex. The results can be distilled down into a simple conceptual model. As the plume rises, the air diverges at the top of the updraft where the largest concentrations of smoke are found. This divergence induces a dilution of the background cyclonic absolute vorticity producing an anticyclonic vortex. Vortex decay occurs from opposite arguments. Plain Language Summary: Recently, there has been an increase in large and intense wildfires ("megafires") across the Earth in response to global warming. These megafire events produce large amounts of smoke that can rise high up in the atmosphere to a level well above clouds and weather. The smoke can stay at these high levels for long periods of time and spread across much of the Earth, which blocks sunlight from reaching the surface. It is important to understand the properties of these smoke plumes and how to correctly predict their consequences on human life. However, uncertainties in both observations and models make it difficult to achieve these goals. In particular, models contain various sources of uncertainty that can interact in complex ways. In this paper, we show that previous research has used a model grid spacing that does not sample the plume accurately, which leads to errors that affect the interpretation of the smoke properties, evolution of the plume and potential climatic effects. By choosing a model grid spacing that accurately samples the plume structure, the errors in the dynamics component of the model can be minimized, providing a baseline for reducing uncertainty in other parts of the system. Key Points: Simulations of megafire smoke plumes require fully resolved dynamics (∼8 Δx) in order to accurately characterize plume properties (e.g., black carbon [BC] fraction)For the 2017 British Columbia event, smoke injected at ∼12 km height requires a reduction in BC fraction by 50% to match observations for external mixturesAnalysis of the vorticity dynamics shows that smoke plume anti‐cyclonic vortices form (decay) due to the dilution (concentration) of cyclonic absolute vorticity [ABSTRACT FROM AUTHOR]

This paper describes the first implementation of the Δx = 3.25 km version of the Energy Exascale Earth System Model (E3SM) global atmosphere model and its behavior in a 40‐day prescribed‐sea‐surface‐temperature simulation (January 20 through February 28, 2020). This simulation was performed as part of the DYnamics of the Atmospheric general circulation Modeled On Non‐hydrostatic Domains (DYAMOND) Phase 2 model intercomparison. Effective resolution is found to be ∼6× the horizontal dynamics grid resolution despite using a coarser grid for physical parameterizations. Despite this new model being in an immature and untuned state, moving to 3.25 km grid spacing solves several long‐standing problems with the E3SM model. In particular, Amazon precipitation is much more realistic, the frequency of light and heavy precipitation is improved, agreement between the simulated and observed diurnal cycle of tropical precipitation is excellent, and the vertical structure of tropical convection and coastal stratocumulus look good. In addition, the new model is able to capture the frequency and structure of important weather events (e.g., tropical cyclones, extratropical cyclones including atmospheric rivers, and cold air outbreaks). Interestingly, this model does not get rid of the erroneous southern branch of the intertropical convergence zone nor the tendency for strongest convection to occur over the Maritime Continent rather than the West Pacific, both of which are classic climate model biases. Several other problems with the simulation are identified, underscoring the fact that this model is a work in progress. Plain Language Summary: This paper describes the new global 3.25 km version of the Energy Exascale Earth System Model (E3SM) atmosphere model and its behavior in a 40‐day northern‐hemisphere wintertime simulation. In exchange for huge computational expense, this high‐resolution model avoids many but not all biases common in lower‐resolution models. It also captures several types of extreme weather that would simply not be resolved in lower‐resolution models. Several opportunities for further development are identified. Key Points: Describes the Simple Cloud‐Resolving E3SM Atmosphere Model (SCREAM)SCREAM performs well in a 40‐day boreal winter simulation at 3.25 km ΔxResolving deep convection solves many long‐standing climate model biases [ABSTRACT FROM AUTHOR]

Atmospheric inversions provide estimates of carbon dioxide (CO2) fluxes between the surface and atmosphere based on atmospheric CO2 concentration observations. The number of CO2 observations is projected to increase severalfold in the next decades from expanding in situ networks and next‐generation CO2‐observing satellites, providing both an opportunity and a challenge for inversions. This study introduces the TRACE Regional Atmosphere–Carbon Ensemble (TRACE) system, which employ an ensemble‐based simultaneous state and parameter estimation (ESSPE) approach to enable the assimilation of large volumes of observations for constraining CO2 flux parameters. TRACE uses an online full‐physics mesoscale atmospheric model and assimilates observations serially in a coupled atmosphere–carbon ensemble Kalman filter. The data assimilation system was tested in a series of observing system simulation experiments using in situ observations for a regional domain over North America in summer. Under ideal conditions with known prior flux parameter error covariances, TRACE reduced the error in domain‐integrated monthly CO2 fluxes by about 97% relative to the prior flux errors. In a more realistic scenario with unknown prior flux error statistics, the corresponding relative error reductions ranged from 80.6% to 88.5% depending on the specification of prior flux parameter error correlations. For regionally integrated fluxes on a spatial scale of 106 km2, the sum of absolute errors was reduced by 34.5%–50.9% relative to the prior flux errors. Moreover, TRACE produced posterior uncertainty estimates that were consistent with the true errors. These initial experiments show that the ESSPE approach in TRACE provides a promising method for advancing CO2 inversion techniques. Plain Language Summary: To gain a better understanding of the main drivers behind atmospheric carbon dioxide (CO2) variations and trends, it is essential to accurately quantify CO2 exchanges—also known as fluxes—between the atmosphere and other components of the Earth system. It is generally not possible to directly measure surface CO2 fluxes at regional scales; however, fluxes can be inferred from a network of atmospheric CO2 concentration observations combined with atmospheric modeling and inversion methods, which seek to find the most likely fluxes given prior knowledge and observational evidence. This paper presents a new regional inversion framework called TRACE Regional Atmosphere–Carbon Ensemble (TRACE) for deriving CO2 fluxes at high resolution in space and time. One of the major innovations in TRACE is the ensemble‐based dual state and parameter estimation approach, which makes it computationally feasible to ingest large volumes of observations into the system. A series of experiments were carried out in a domain over North America to test the new framework in a controlled setting. In the experiments, we used synthetic tower observations of atmospheric CO2 concentrations to constrain parameters controlling terrestrial biogenic and oceanic CO2 fluxes. The results show that TRACE is capable of accurately estimating both the magnitude and spatial pattern of regional CO2 fluxes. Key Points: An ensemble‐based simultaneous state and parameter estimation approach is introduced for regional CO2 flux inversionsThe new dual‐state TRACE Regional Atmosphere–Carbon Ensemble (TRACE) framework provides a flexible platform for conducting coupled atmosphere–carbon data assimilationControlled experiments show that TRACE is effective at constraining regional CO2 fluxes using in situ CO2 concentration observations [ABSTRACT FROM AUTHOR]

The use of machine learning to build subgrid parametrizations for climate models is receiving growing attention. State‐of‐the‐art strategies address the problem as a supervised learning task and optimize algorithms that predict subgrid fluxes based on information from coarse resolution models. In practice, training data are generated from higher resolution numerical simulations transformed in order to mimic coarse resolution simulations. By essence, these strategies optimize subgrid parametrizations to meet so‐called a priori criteria. But the actual purpose of a subgrid parametrization is to obtain good performance in terms of a posteriori metrics which imply computing entire model trajectories. In this paper, we focus on the representation of energy backscatter in two‐dimensional quasi‐geostrophic turbulence and compare parametrizations obtained with different learning strategies at fixed computational complexity. We show that strategies based on a priori criteria yield parametrizations that tend to be unstable in direct simulations and describe how subgrid parametrizations can alternatively be trained end‐to‐end in order to meet a posteriori criteria. We illustrate that end‐to‐end learning strategies yield parametrizations that outperform known empirical and data‐driven schemes in terms of performance, stability, and ability to apply to different flow configurations. These results support the relevance of differentiable programming paradigms for climate models in the future. Plain Language Summary: Climate projection and weather forecast heavily rely on computer simulations. But, if the physical laws governing the evolution of the climate system are well known, their simulation is still rather challenging. Fluid flows being essentially turbulent, small details at fine scales can have a tremendous impact on larger scales. Still, because of the limitations in computing power, all these interactions across scales cannot be explicitly resolved in computer simulations. Some of these interactions can only be represented approximately, and the design of these approximations is an active research area. Here, we describe a new method which leverages recent advances in machine learning. We propose to train an approximate representation of unresolved scales of motions that optimizes the quality of the climate model over some temporal horizon. This results in more accurate and stable predictions. Our method shows very promising results in toy example flow simulations, but its deployment at scale may seriously challenge the overall design of legacy climate models. Key Points: Subgrid parametrizations can be learned end‐to‐end with a posteriori criteria involving model integration over several time‐stepsA posteriori learning for quasi‐geostrophic turbulent flows can solve numerical stability issues related to energy backscatterLearned parametrizations outperform existing baselines for various evaluation metrics and apply to different flow configurations [ABSTRACT FROM AUTHOR]

Coupled atmosphere‐fire modeling is recognized as a relevant approach for the representation of the interaction between a wildland fire and local meteorology at landscape scales. The atmospheric model component used in the coupled system is based on several approximations, which are adopted for computational efficiency or physical processes representation, including the widely used anelastic approximation. The validity domain of the anelastic approximation may be questioned in the context of high‐resolution wildland fire modeling due to the large fire‐induced heat releases near the surface. This study aims to study this question with the MesoNH anelastic model coupled with the Blaze fire model. A compressible version of the MesoNH‐Blaze coupled model has been developed for comparison with the anelastic system. The FireFlux I experimental fire is used for this comparative study conducted at a 10‐m and a 25‐m horizontal atmospheric resolution. Results show significant anelastic/compressible differences at a 10‐m resolution on the physical processes occurring near the fire with higher horizontal velocities and the presence of gravity waves downstream of the fire. This is in addition to the fire plume with realistic larger vertical velocities. Differences at a 25‐m resolution are much smaller in all evaluated processes. The compressible system only enriches the physics underlying fire‐atmosphere interactions at a very high resolution, which means that the anelastic approximation remains relevant for large‐scale coupled atmosphere‐fire simulations, considering the significant economy concerning numerical costs. Plain Language Summary: Wildfires burn large amounts of forests each year, destroying the living habitat of many species, endangering human settlements and provoking cross‐continent smoke events leading to air quality issues. Wildfires result from complex physical, chemical and biological processes. Understanding the fundamental processes driving wildfire behavior is a key point for the prediction of fire spread across the landscape and the induced atmospheric dynamics. Numerical models coupling the atmospheric dynamics (wind, temperature, air density and pressure, etc.) and fire front evolution are efficient tools in this scientific process. This paper evaluates how such model simulations are affected by the way in which the effects of high air density gradients induced by the very large amount of heat released into the atmosphere are modeled. When the spatial resolution of the atmospheric model is as high as 10 m, an exact – compressible – formulation leads to the formation of gravity waves downstream of the fire, and to stronger fire‐induced wind, fire propagation and larger vertical velocities than an approximate (and computationally faster) approach. At a resolution of 25 m, which is still very high, the approximate approach leads to results, which are as good as the exact one. This can then be used to simulate wildland fire behavior at landscape‐to‐meteorological scales. It also paves the way for future accurate wildfire forecast systems. Key Points: A compressible version of the MesoNH atmospheric model has been developed to evaluate equation system approximations in wildfire simulationHorizontally buoyancy‐driven effects induce turbulent movements, waves ahead of the fire, and a bending of the plumeThe anelastic approximation for the atmosphere model is suitable to represent wildland fire effects, except for resolutions of 10 m or finer [ABSTRACT FROM AUTHOR]

It is very well known that parameter tuning can have a major impact on the performance of Global Climate Models. However, parameter tuning is not the only implementation detail that can drastically affect model performance and the representation of various phenomena in models. "Minor‐looking treatments" often exert a critical control on model performance; they include lower and upper limits of physical variables, thresholds of variables that control the enabling or disabling of a specific process, whether two schemes can work together or only one scheme works exclusively, and numerical methods including the order of calling various physics schemes. The impacts of such minor‐looking treatments are sometimes comparable to or even larger than those obtained by introducing advanced parameterizations based on theory and observation. Because the importance of these treatments is often overlooked and not discussed in the literature, we comprehensively summarize examples of various minor‐looking treatments that can considerably affect model performance. Plain Language Summary: Global climate models (GCMs) are used for climate simulations, including global warming simulations. Global climate models (GCMs) simulate various physical processes such as convection, clouds, radiation, and turbulence. A lot of effort is devoted to improving the representation of these physical processes by introducing more sophisticated or more complicated and detailed processes. The values of parameters in such processes significantly affect the model performance, and many modelers carefully set such parameter values. However, other factors can also affect model performance. The lower or upper limits specified in the detailed implementation of such processes, threshold values for switching on a certain effect, how to calculate each effect, vertical resolutions, and so on can significantly affect the model performance. We call these "minor‐looking treatments" and show a wide range of examples in this paper. Key Points: Minor‐looking treatments crucially determine the performance of global climate modelsThey include lower limits, upper limits, thresholds for enabling physical processes, and numerical methodsMinor‐looking treatments should be documented in detail, shared, and discussed more in the climate community [ABSTRACT FROM AUTHOR]

The development of parameterizations is a major task in the development of weather and climate models. Model improvement has been slow in the past decades, due to the difficulty of encompassing key physical processes into parameterizations, but also of calibrating or "tuning" the many free parameters involved in their formulation. Machine learning techniques have been recently used for speeding up the development process. While some studies propose to replace parameterizations by data‐driven neural networks, we rather advocate that keeping physical parameterizations is key for the reliability of climate projections. In this paper we propose to harness machine learning to improve physical parameterizations. In particular, we use Gaussian process‐based methods from uncertainty quantification to calibrate the model free parameters at a process level. To achieve this, we focus on the comparison of single‐column simulations and reference large‐eddy simulations over multiple boundary‐layer cases. Our method returns all values of the free parameters consistent with the references and any structural uncertainties, allowing a reduced domain of acceptable values to be considered when tuning the three‐dimensional (3D) global model. This tool allows to disentangle deficiencies due to poor parameter calibration from intrinsic limits rooted in the parameterization formulations. This paper describes the tool and the philosophy of tuning in single‐column mode. Part 2 shows how the results from our process‐based tuning can help in the 3D global model tuning. Key Points: We apply uncertainty quantification to single‐column model/large‐eddy simulation comparison to calibrate free parametersWe revisit model development strategy with an emphasis on processes for model calibrationThe proposed tuning tool allows to formalize the complementary use of multicases with various metrics [ABSTRACT FROM AUTHOR]

This paper provides a comprehensive derivation of the total energy equations for the atmospheric components of Earth System Models (ESMs). The assumptions and approximations made in this derivation are motivated and discussed. In particular, it is emphasized that closing the energy budget is conceptually challenging and hard to achieve in practice without resorting to ad hoc fixers. As a concrete example, the energy budget terms are diagnosed in a realistic climate simulation using a global atmosphere model. The largest total energy errors in this example are spurious dynamical core energy dissipation, thermodynamic inconsistencies (e.g., coupling parameterizations with the host model) and missing processes/terms associated with falling precipitation and evaporation (e.g., enthalpy flux between components). The latter two errors are not, in general, reduced by increasing horizontal resolution. They are due to incomplete thermodynamic and dynamic formulations. Future research directions are proposed to reconcile and improve thermodynamics formulations and conservation principles. Plain Language Summary: Earth System Models (ESMs) have numerous total energy budget errors. This article establishes the governing total energy equations for large‐scale ESMs and assesses the energy budget errors in real‐world simulations in a widely used climate model. To move towards a closed energy budget in ESMs, further research on total energy conserving discretizations (in the dynamical core), unified thermodynamics (through thermodynamic potentials/conserved variables) and missing processes is paramount. This research is especially important since some of the energy budget errors will not improve with higher spatial resolution and may even get worse. Key Points: Closing total energy budgets in Earth System Models without ad hoc fixers is a monumental taskLargest errors are from missing processes/terms, thermodynamic inconsistencies and dynamical coreFurther research is needed on conservative discretizations, unified thermodynamics and missing processes [ABSTRACT FROM AUTHOR]

Deep learning can accurately represent sub‐grid‐scale convective processes in climate models, learning from high resolution simulations. However, deep learning methods usually lack interpretability due to large internal dimensionality, resulting in reduced trustworthiness in these methods. Here, we use Variational Encoder Decoder structures (VED), a non‐linear dimensionality reduction technique, to learn and understand convective processes in an aquaplanet superparameterized climate model simulation, where deep convective processes are simulated explicitly. We show that similar to previous deep learning studies based on feed‐forward neural nets, the VED is capable of learning and accurately reproducing convective processes. In contrast to past work, we show this can be achieved by compressing the original information into only five latent nodes. As a result, the VED can be used to understand convective processes and delineate modes of convection through the exploration of its latent dimensions. A close investigation of the latent space enables the identification of different convective regimes: (a) stable conditions are clearly distinguished from deep convection with low outgoing longwave radiation and strong precipitation; (b) high optically thin cirrus‐like clouds are separated from low optically thick cumulus clouds; and (c) shallow convective processes are associated with large‐scale moisture content and surface diabatic heating. Our results demonstrate that VEDs can accurately represent convective processes in climate models, while enabling interpretability and better understanding of sub‐grid‐scale physical processes, paving the way to increasingly interpretable machine learning parameterizations with promising generative properties. Plain Language Summary: Deep neural nets are hard to interpret due to their hundred thousand or million trainable parameters without further postprocessing. We demonstrate in this paper the usefulness of a network type that is designed to drastically reduce this high dimensional information in a lower‐dimensional space to enhance the interpretability of predictions compared to regular deep neural nets. Our approach is, on the one hand, able to reproduce small‐scale cloud related processes in the atmosphere learned from a physical model that simulates these processes skillfully. On the other hand, our network allows us to identify key features of different cloud types in the lower‐dimensional space. Additionally, the lower‐order manifold separates tropical samples from polar ones with a remarkable skill. Overall, our approach has the potential to boost our understanding of various complex processes in Earth System science. Key Points: A Variational Encoder Decoder (VED) can predict sub‐grid‐scale thermodynamics from the coarse‐scale climate stateThe VED's latent space can distinguish convective regimes, including shallow/deep/no convectionThe VED's latent space reveals the main sources of convective predictability at different latitudes [ABSTRACT FROM AUTHOR]

Traditional parameterizations of convection are a large source of error in weather and climate prediction models, and the assumptions behind them become worse as resolution increases. Multifluid modeling is a promising new method of representing subgrid‐scale and near‐grid‐scale convection allowing for net mass transport by convection and nonequilibrium dynamics. The air is partitioned into two or more fluids, which may represent, for example, updrafts and the nonupdraft environment. Each fluid has its own velocity, temperature, and constituents with separate equations of motion. This paper presents two‐fluid Boussinesq equations for representing subgrid‐scale dry convection with sinking and w = 0 air in Fluid 0 and rising air in Fluid 1. Two vertical slice test cases are developed to tune parameters and to evaluate the two‐fluid equations: a buoyant rising bubble and radiative convective equilibrium. These are first simulated at high resolution with a single‐fluid model and conditionally averaged based on the sign of the vertical velocity. The test cases are next simulated with the two‐fluid model in one column. A model for entrainment and detrainment based on divergence leads to excellent representation of the convective area fraction. Previous multifluid modeling of convection has used the same pressure for both fluids. This is shown to be a bad approximation, and a model for the pressure difference between the fluids based on divergence is presented. Plain Language Summary: Clouds and buoyant convection are often smaller than the grid size in weather and climate prediction models, but they are of central importance. Therefore, their effects are parameterized—their interactions with the larger scales are estimated based on properties of the larger scales. Convection parameterizations are a large source of error in models of the atmosphere despite decades of effort to improve them. Some unrealistic assumptions remain such as that convection does not provide a net upward transport of mass, convection is in equilibrium with the surroundings, and that the variations in the horizontal direction are not relevant. These assumptions make it much simpler to incorporate convection parameterizations into existing models. Multifluid modeling does not rely on these assumptions, but the approach means a whole new atmospheric model rather than a stand‐alone routine that can be incorporated into an existing model. This paper provides some closures necessary for multifluid modeling and evaluation of the technique for two‐dimensional dry convection. Key Points: Multifluid modeling enables net mass transport by subgrid‐scale convection and nonequlibrium dynamicsMultifluid modeling means a new dynamical core rather than a parameterization for an existing coreClosures for the pressure difference between fluids and entrainment and detrainment are provided [ABSTRACT FROM AUTHOR]

Tropospheric aerosol radiative forcing has persisted for many years as one of the major causes of uncertainty in global climate model simulations. To sample the range of plausible aerosol and atmospheric states and perform robust statistical analyses of the radiative forcing, it is important to account for the combined effects of many sources of model uncertainty, which is rarely done due to the high computational cost. This paper describes the designs of two ensembles of the Met Office Hadley Centre Global Environment Model‐U.K. Chemistry and Aerosol global climate model and provides the first analyses of the uncertainties in aerosol radiative forcing and their causes. The first ensemble was designed to comprehensively sample uncertainty in the aerosol state, while the other samples additional uncertainties in the physical model related to clouds, humidity, and radiation, thereby allowing an analysis of uncertainty in the aerosol effective radiative forcing. Each ensemble consists of around 200 simulations of the preindustrial and present‐day atmospheres. The uncertainty in aerosol radiative forcing in our ensembles is comparable to the range of estimates from multimodel intercomparison projects. The mean aerosol effective radiative forcing is −1.45 W/m2 (credible interval of −2.07 to −0.81 W/m2), which encompasses but is more negative than the −1.17 W/m2 in the 2013 Atmospheric Chemistry and Climate Model Intercomparison Project and −0.90 W/m2 in the Intergovernmental Panel on Climate Change Fifth Assessment Report. The ensembles can be used to reduce aerosol radiative forcing uncertainty by challenging them with multiple measurements as well as to isolate potential causes of multimodel differences. Plain Language Summary: Atmospheric aerosol particles such as dust, pollutants, and smoke interfere with light from the Sun and modify the properties of clouds and thereby affect Earth's climate. However, the effect that aerosols have on climate is one of the major causes of uncertainty in global climate model simulations. We performed a large number of climate model simulations (called an ensemble), with many parts of the model slightly varied, in order to understand the complex behavior of the model and to explore the causes of uncertainty in model outputs. This paper describes the designs of two climate model ensembles and provides the first analyses of the causes of model uncertainty. The first ensemble was designed to comprehensively understand the behavior of aerosols in the atmosphere, while the other includes more general uncertainties in atmospheric processes that can affect aerosols. Each ensemble consists of around 200 simulations. The ranges of the aerosol climate effect in our ensembles are comparable to the ranges of previous estimates from studies that analyzed multiple climate models. These ensembles can be used to reduce uncertainty in how aerosols affect climate by comparing with satellite and ground‐based measurements. Key Points: Two ensembles of atmospheric simulations were performed perturbing aerosol and physical parameters under different model setupsThousands of Gaussian process emulators sampling parameter spaces enabled statistical analyses of model's parametric uncertainty.Fully explored parametric uncertainty of aerosol radiative forcing in a model was found to be comparable to that in multimodel studies [ABSTRACT FROM AUTHOR]

Model bias is one of the main obstacles to improved accuracy and reliability in numerical weather prediction conducted with state-of-the-art atmospheric models. To deal with model bias, a modification of the standard four-dimensional variational (4D-Var) algorithm, called weak-constraint 4D-Var, has been developed where a forcing term is introduced into the model to correct for the bias that accumulates along the model trajectory. This approach reduced the temperature bias in the stratosphere by up to 50% and is implemented in the European Centre for Medium-Range Weather Forecasts operational forecasting system. Despite different origins and applications, data assimilation (DA) and Deep Learning are both able to learn about the Earth system from observations. In this paper, a deep learning approach for model bias correction is developed using temperature retrievals from radio occultation (RO) measurements. Neural networks (NNs) require a large number of samples to properly capture the relationship between the temperature first-guess trajectory and the model bias. As running the Integrated Forecasting System (IFS) DA system for extended periods of time with a fixed model version and at realistic resolutions is computationally very expensive, we have chosen to train, the initial NNs are trained using the ERA5 reanalysis before using transfer learning on 1 year of the current IFS model. Preliminary results show that convolutional NNs are adequate to estimate model bias from RO temperature retrievals. The different strengths and weaknesses of both deep learning and weak constraint 4D-Var are discussed, highlighting the potential for each method to learn model biases effectively and adaptively. Plain Language Summary In the practice of the numerical weather prediction, the state of the Earth system is estimated via a combination of information from both previous weather predictions and Earth system observations. This complex, mathematical procedure is called data assimilation (DA). Weather predictions could be improved if the error of the numerical models that are used could be reduced. Recent advances in DA at the European Centre for Medium-Range Weather Forecasts indicate that it is possible to estimate and correct for a large fraction of systematic model errors of those models. During DA, the forecast model and Earth system observations are representing the same situation of the global atmosphere. A direct comparison between models and observations during the short time interval of the DA process can be used to diagnose model errors. Deep learning is a comparably new method from machine learning that can be used to learn complex mapping procedures. The question we address in this paper is whether deep learning techniques can be used to predict model errors when they are trained to predict the mapping between the global temperature and the model error that was diagnosed during DA. [ABSTRACT FROM AUTHOR]

Enabling Earth System models to run efficiently on future supercomputers is a serious challenge for model development. Many publications study efficient parallelization to allow better scaling of performance on an increasing number of computing cores. However, one of the most alarming threats for weather and climate predictions on future high performance computing architectures is widely ignored: the presence of hardware faults that will frequently hit large applications as we approach exascale supercomputing. Changes in the structure of weather and climate models that would allow them to be resilient against hardware faults are hardly discussed in the model development community. In this paper, we present an approach to secure the dynamical core of weather and climate models against hardware faults using a backup system that stores coarse resolution copies of prognostic variables. Frequent checks of the model fields on the backup grid allow the detection of severe hardware faults, and prognostic variables that are changed by hardware faults on the model grid can be restored from the backup grid to continue model simulations with no significant delay. To justify the approach, we perform model simulations with a C-grid shallow water model in the presence of frequent hardware faults. As long as the backup system is used, simulations do not crash and a high level of model quality can be maintained. The overhead due to the backup system is reasonable and additional storage requirements are small. Runtime is increased by only 13 % for the shallow water model. [ABSTRACT FROM AUTHOR]

Advances in high‐performance computing have enabled large‐eddy simulations (LES) of turbulence, convection, and clouds. However, their potential to improve parameterizations in global climate models (GCMs) is only beginning to be harnessed, with relatively few canonical LES available so far. The purpose of this paper is to begin creating a public LES library that expands the training data available for calibrating and evaluating GCM parameterizations. To do so, we use an experimental setup in which LES are driven by large‐scale forcings from GCMs, which in principle can be used at any location, any time of year, and in any climate state. We use this setup to create a library of LES of clouds across the tropics and subtropics, in the present and in a warmer climate, with a focus on the transition from stratocumulus to shallow cumulus over the East Pacific. The LES results are relatively insensitive to the choice of host GCM driving the LES. Driven with large‐scale forcing under global warming, the LES simulate a positive but weak shortwave cloud feedback, adding to the accumulating evidence that low clouds amplify global warming. Plain Language Summary: Clouds remain one of the largest uncertainties in our understanding and in predictions of climate change because it is challenging to represent their small‐scale dynamics in climate models. However, high‐resolution simulations can provide faithful simulations of clouds and turbulence in limited areas, and these can be used to calibrate climate models. So far, only a limited set of simulations has been used for calibration of climate models, with focus on a few specific locations. This study presents an experimental setup that allows the high‐resolution simulations to be run anywhere on the globe, in any climate state, driven by output from different climate models. The setup is used to create a library of high‐resolution simulations of clouds across the tropics and subtropics in both the current and a warmer climate. The library substantially expands the data set available for the calibration and evaluation of climate models. Key Points: A library of high‐resolution simulations of clouds is created with large‐eddy simulations (LES) driven by a few global climate models (GCMs)Clouds simulated by LES driven by different GCMs are more similar to one another than the corresponding GCM‐simulated cloudsUnder global warming, LES‐simulated low clouds exhibit a weak but positive feedback on warming [ABSTRACT FROM AUTHOR]

The treatment of cloud structure in numerical weather and climate models is often greatly simplified to make them computationally affordable. Here we propose to correct the European Centre for Medium‐Range Weather Forecasts 1D radiation scheme ecRad for 3D cloud effects using computationally cheap neural networks. 3D cloud effects are learned as the difference between ecRad's fast 1D Tripleclouds solver that neglects them and its 3D SPARTACUS (SPeedy Algorithm for Radiative TrAnsfer through CloUd Sides) solver that includes them but is about five times more computationally expensive. With typical errors between 20% and 30% of the 3D signal, neural networks improve Tripleclouds' accuracy for about 1% increase in runtime. Thus, rather than emulating the whole of SPARTACUS, we keep Tripleclouds unchanged for cloud‐free parts of the atmosphere and 3D‐correct it elsewhere. The focus on the comparably small 3D correction instead of the entire signal allows us to improve predictions significantly if we assume a similar signal‐to‐noise ratio for both. Plain Language Summary: Solar and terrestrial radiation is the primary driver of the Earth's weather and their detailed representation is essential for improving weather predictions and climate projections. Several aspects, however, such as the flow of radiation through the side of clouds and other three‐dimensional effects are often too costly to compute routinely. In this paper we describe how machine learning can help account for these effects cheaply. Key Points: Emulators are used to add 3D cloud effects to the fast 1D radiation solver Tripleclouds cheaplyEmulators are trained on 3D cloud effects from SPARTACUS, which is five times slower than TriplecloudsFor a 1% slowdown in Tripleclouds' runtime, 3D fluxes are emulated with an average error of about 20%–30% [ABSTRACT FROM AUTHOR]

In global atmospheric modeling, the differences between nonhydrostatic (NH) and hydrostatic (H) dynamical cores are negligible in dry simulations when grid spacing is larger than 10 km. However, recent studies suggest that those differences can be significant at far coarser resolution when moisture is included. To better understand how NH and H differences manifest in global fields, we perform and analyze an ensemble of 28 and 13 km seasonal simulations with the NH and H dynamical cores in the Energy Exascale Earth System Model global atmosphere model, where the differences between H and NH configurations are minimized. A set of idealized rising bubble experiments is also conducted to further investigate the differences. Although NH and H differences are not significant in global statistics and zonal averages, significant differences in precipitation amount and patterns are observed in parts of the tropics. The most prominent differences emerge near India and the Western Pacific in the boreal summer, and the central‐southern Indian Ocean and Pacific in the boreal winter. Tropical differences influence surrounding regions through modification of the regional circulation and can propagate to the extratropics, leading to significant temperature and geopotential differences over the middle to high latitudes. While the dry bubble experiments show negligible deviation between H and NH dynamics until grid spacing is below 6.25 km, precipitation amount and vertical velocity are different in the moist case even at 25 km resolution. Plain Language Summary: This paper compares the nonhydrostatic and hydrostatic dynamical cores in the Energy Exascale Earth System Model global atmosphere model. The rainfall difference between the two dynamical cores is noticeable at 28 km resolution, which is much coarser than the conclusions in the traditional studies. The difference maxima are located over the tropical oceans where the moist processes amplify the difference. The difference maxima in summer and winter emerge over different locations indicating season variations. Key Points: Experiments are performed to compare the hydrostatic (H) and nonhydrostatic (NH) dynamic cores in the Simple Cloud‐Resolving Energy Exascale Earth System Model Atmosphere ModelRegions with significant differences between H and NH dynamical cores are identifiedThe physical mechanisms driving these differences are investigated [ABSTRACT FROM AUTHOR]

Accurately representing mesoscale convective systems (MCSs) is crucial to simulating the energy and water cycles in global climate models. Using a novel MCS identification and tracking algorithm applied to observations and model simulations, we evaluate how well the Energy Exascale Earth System Model (E3SM) simulates MCSs over the central US. Simulations performed by E3SM at 25 km grid spacing with and without superparameterization (SP) are compared and evaluated against observations using multiple metrics highlighting important MCS characteristics. Compared to E3SM, the superparameterized model (SP‐E3SM) better simulates MCS number and MCS precipitation amount, diurnal cycle, propagation, and the probability distribution of precipitation rate in both spring and summer. The improvement from SP is partly contributed by improvement in simulating the large‐scale environments, featuring enhanced atmospheric low‐level moisture and larger moisture transport to the central US relative to E3SM. However, SP‐E3SM still underestimates MCS precipitation amount, particularly in summer. This underestimation is closely related to the negative bias in MCS precipitation intensity, although the drier environments simulated during summer also contributes to underestimation of MCS frequency. Without SP, the larger bias in MCS precipitation amount is closely related to the negative bias in MCS frequency, while the substantial dry bias in the large‐scale environments also contributes to underestimation of MCS intensity. Our results suggest that SP improves MCS simulation by improving modeling of the large‐scale environments and convection initiation, which are both major limiting factors in E3SM even at 25 km grid spacing where deep convection is represented by a cumulus parameterization. Plain Language Summary: Large thunderstorms called mesoscale convective systems (MCSs) are an important source of flood‐producing rainfalls. In this paper, we examine how well the Energy Exascale Earth System Model (E3SM) simulate MCSs over the central US in spring and summer by comparing against observations. We run two versions of E3SM: one is the standard E3SM that uses a traditional parameterization to simulate deep convection, the other is E3SM with a superparameterization (SP‐E3SM) that uses a cloud resolving model within each E3SM grid column to explicitly simulate deep convection. By using a novel detection and tracking algorithm to characterize MCS features, we find that SP‐E3SM can better simulate the number, rainfall amount and intensity, and the diurnal cycle of MCSs compared to E3SM. SP improves MCS simulations by improving the representation of the large‐scale circulations and convection initiation. Despite the improvement, SP‐E3SM still underestimates the MCS rainfall amount, which is largely related to the underestimation of MCS rainfall intensity. In contrast, the larger underestimation of rainfall amount in E3SM is attributed in large parts to the lower MCS frequency. This study highlights the importance of large‐scale moisture transport and convection initiation in simulating MCSs. Key Points: Mesoscale convective systems (MCSs) in E3SM at 25 km with and without superparameterization (SP) are evaluated the central USIncluding SP significantly improves MCS simulations in both spring and summer, although large dry biases are still apparent in summerImprovements from SP are contributed by improved simulations of the large‐scale environments and MCS number and precipitation frequency [ABSTRACT FROM AUTHOR]

This paper describes the development and evaluation of a new nonhydrostatic dynamical framework for global and regional atmospheric modeling, with an emphasis on the numerical performance of dry dynamics. The model is formulated in a layer‐averaged manner using a generalized hybrid sigma‐mass vertical coordinate and an unstructured mesh. The mass‐based equations allow a flexible and effective switch between the hydrostatic and nonhydrostatic solvers. The unstructured mesh treats the conventional icosahedral grid and the more general Voronoi polygon in a consistent manner, allowing a flexible switch between quasi‐uniform and variable‐resolution modeling. The horizontal discretization and vertical discretization are formulated in an explicit Eulerian approach, while those terms describing the vertically propagating fast waves are solved implicitly. The model is equipped with physically based Smagorinsky diffusion as a tuning tool. A suite of multiscale test cases from hydrostatic to nonhydrostatic regimes is used to assess the model performance. The general strategies for evaluation focus on two aspects: (i) the nonhydrostatic solver should behave similarly to its hydrostatic counterpart under the hydrostatic regime and (ii) the nonhydrostatic solver should produce unique nonhydrostatic responses under the nonhydrostatic regime. In the context of model evaluation, model sensitivity to numerical configurations is further explored to understand the impact of isolated components, helping to identify appropriate configurations for realistic modeling applications. The present framework is a prototype toward a Global‐Regional Integrated forecast SysTem (GRIST). Plain Language Summary: A dynamical framework describes the evolution of resolved‐scale fluid dynamical processes in a numerical model. It also helps to define model architecture, and its establishment usually initiates the development of a new model. This paper describes the development and evaluation of a new nonhydrostatic dynamical framework, which may be used for global and regional atmospheric modeling applications such as weather forecasting, climate prediction, and basic research. The model is based on a generalized vertical coordinate and an unstructured mesh. It is discretized in a fully explicit Eulerian approach, except for the vertically propagating fast waves. Idealized experimental protocols are used to demonstrate the performance of dry dynamics associated with atmospheric transport and wave‐like phenomena. Sensitivity tests are performed to understand the impact of numerical configurations on the model behavior. Key Points: A layer‐averaged nonhydrostatic dynamical framework on an unstructured mesh is developed for global and regional atmospheric modelingThe baseline numerical performance of dry dynamics is evaluatedThe sensitivity associated with the numerical configuration is explored [ABSTRACT FROM AUTHOR]

A new number format for ensemble simulations for weather and climate predictions is presented. The new format exploits similarities between ensemble members to reduce overall data usage. Data use and movement is the most important performance bottleneck for weather and climate models on modern supercomputers. The new format can be realized on standard hardware and with standard computing languages such as Fortran. The format is tested in simulations with a shallow water model. Advantages and disadvantages are discussed. Plain Language Summary: To generate weather and climate predictions requires a lot of computing power and numerical models for weather and climate predictions run on large supercomputers. Physical information is stored in bits that are either 0 or 1 using different number formats to represent integers and real numbers. In this paper, a new number format is presented that is designed for specific use in weather and climate predictions. Many weather and climate models are run in so‐called ensemble mode. Here several realizations of the same model are run in parallel with small perturbations added to the individual ensemble members. Each ensemble member will then provide a slightly different prediction for a specific variable (e.g., temperature in London), and the spread of the individual predictions can help to estimate uncertainties of forecasts. The new format is using similarities between ensemble members to reduce overall data usage. Data use and movement is one of the most important cost factors in modern supercomputing. Therefore, the new number format is helping to reduce computing cost and savings can be reinvested to improve the model configuration to allow better predictions. The new format can be realized on standard hardware and with standard computing languages. Key Points: A new number format for ensemble simulations for weather and climate predictions is presented and testedThe ensemble format links numerical precision to predictability to optimize the information content per bit to be moved and processedThe format is based on integer and real number arithmetic and can be used on standard hardware with standard computing languages [ABSTRACT FROM AUTHOR]

Limited‐area convection‐permitting climate models (CPMs) with horizontal grid‐spacing less than 4 km and not relying on deep convection parameterisations (CPs) are being used more and more frequently. CPMs represent small‐scale features such as deep convection more realistically than coarser regional climate models (RCMs) with deep CPs. Because of computational costs, CPMs tend to use smaller horizontal domains than RCMs. As all limited‐area models (LAMs), CPMs suffer issues with lateral boundary conditions (LBCs) and nesting. We investigated these issues using idealized Big‐Brother (BB) experiments with the LAM COSMO‐CLM. Grid‐spacing of the reference BB simulation was 2.4 km. Deep convection was triggered by idealized hills with driving data from simulations with different spatial resolutions, with/without deep CP, and with different nesting frequencies and LBC formulations. All our nested idealized 2.4‐km Little‐Brother (LB) experiments performed worse than a coarser CPM simulation (4.9 km) which used a four times larger computational domain and yet spent only half the computational cost. A boundary zone of >100 $ > 100$ grid‐points of the LBs could not be interpreted meteorologically because of spin‐up of convection and boundary inconsistencies. Hosts with grid‐spacing in the so‐called gray zone of convection (ca. 4–20 km) were not advantageous to the LB performance. The LB's performance was insensitive to the applied LBC formulation and updating (if ≤3 $\le 3$‐hourly). Therefore, our idealized experiments suggested to opt for a larger domain instead of a higher resolution even if coarser than usual (∼5 $\sim 5$ km) as a compromise between the harmful boundary problems, computational cost and improved representation of processes by CPMs. Plain Language Summary: Recently, very high resolution (grid‐spacing <4 $< 4$ km) so‐called convection‐permitting climate models (CPMs) were developed, which represent deep convection explicitly. CPMs, however, are computationally very expensive. They need information about the state of the atmosphere at their lateral boundaries from coarser models. This paper investigates the setting of the lateral boundary formulation. We used idealized experiments with grid‐spacing of ≈ 2.4 km, where deep convection was triggered by small hills. We found that a CPM boundary zone > $ > $100 grid points cannot be interpreted reliably. The boundary data should be given to the CPM every 3 hr or more often. Small‐domain CPM simulations all performed worse than a reference simulation on a larger domain with the same resolution or with an even two times lower resolution. We tested different resolutions of the driving data for the CPMs and found that driving data from a model in the "gray zone" of convection (about 4–20 km) is not advantageous for the CPM performance. We concluded that it often might be better to opt for a larger domain with an unusually coarse CPM resolution (ca. 5 km) than for a much smaller domain with grid‐spacing <4 $< 4$ km. Key Points: The nesting challenge of convection‐permitting climate modeling (CPM) is investigated with idealized simulation experimentsNesting the CPM into host simulations with grid‐spacing in the gray zone of convection is not better than into coarser simulationsLarge boundary areas with poor simulation quality suggest using large domains even with grid‐spacing coarser than usually accepted for CPM [ABSTRACT FROM AUTHOR]

Characteristics of, and fundamental differences between, the radiative‐convective equilibrium (RCE) climate states following the Radiative‐Convective Equilibrium Model Intercomparison Project (RCEMIP) protocols in the Community Atmosphere Model version 5 (CAM5) and version 6 (CAM6) are presented. This paper explores the characteristics of clouds, moisture, precipitation and circulation in the RCE state, as well as the tropical response to surface warming, in CAM5 and CAM6 with different parameterizations. Overall, CAM5 simulates higher precipitation rates that result in larger global average precipitation, despite lower outgoing longwave radiation compared to CAM6. Differences in the structure of clouds, particularly the amount and vertical location of cloud liquid, exist between the CAM versions and can, in part, be related to distinct representations of shallow convection and boundary layer processes. Both CAM5 and CAM6 simulate similar peaks in cloud fraction, relative humidity, and cloud ice, linked to the usage of a similar deep convection parameterization. These anvil clouds rise and decrease in extent in response to surface warming. More generally, extreme precipitation, aggregation of convection, and climate sensitivity increase with warming in both CAM5 and CAM6. This analysis provides a benchmark for future studies that explore clouds, convection, and climate in CAM with the RCEMIP protocols now available in the Community Earth System Model. These results are discussed within the context of realistic climate simulations using CAM5 and CAM6, highlighting the usefulness of a hierarchical modeling approach to understanding model and parameterization sensitivities to inform model development efforts. Plain Language Summary: Clouds, circulations and rainfall in the tropics play an important role in Earth's climate. However, global climate models differ in their representation of these features, contributing to uncertainty in future climate projections. One useful tool to better understand model differences and inform efforts to improve models is to analyze idealized configurations. We explore two different numerical representations of an idealized atmosphere relevant to tropical regions to determine the impact on the characteristics of clouds, rainfall and circulations as well as the tropical atmospheric response to warming. We show that our idealized models mirror differences in the low‐cloud structure in the deep tropics of more realistic models. This work also finds similarities between the two numerical representations, such as a decrease in the spatial extent of high altitude clouds with warming. As the surface is warmed, both versions also show increases in the clustering of clouds, the likelihood of extreme precipitation rates, and an estimate of how much warming would occur in response to a doubling of carbon dioxide. Key Points: A new radiative‐convective equilibrium (RCE) configuration is added to the official release of Community Earth System Model (CESM) to facilitate the use of Radiative‐Convective Equilibrium Model Intercomparison Project (RCEMIP) in the CESM hierarchyThe Community Atmosphere Model version 5 and Community Atmosphere Model version 6 (CAM6) RCEMIP climate states and variation across SST experiments are documented in the context of the CESM hierarchyDiffering parameterizations lead to a reduction of low‐level clouds in CAM6, consistent with differences in more realistic simulations [ABSTRACT FROM AUTHOR]

Tropical deep convection can aggregate into large clusters, which can have impacts on the local humidity and precipitation. Sea surface temperature (SST) gradients have been shown to organize convection, yet there has been little work done to investigate the impact of diabatic heating perturbations in the atmosphere on the aggregation of convection. Here we investigate how anomalous diabatic heating of the atmospheric column, through an idealized aerosol plume, affects the existence and mechanisms of convective aggregation in non‐rotating, global radiative‐convective equilibrium simulations. We show that the aerosol forcing has the ability to increase the degree of aggregation, especially at lower SSTs. Detailed investigation shows that the diabatic heating source incites a thermally driven circulation, forced by the shortwave perturbation. The increase in aggregation is caused in part by this circulation, and in part by the longwave heating anomalies occurring due to the surface convergence of moisture and convection. At higher SSTs, longwave feedbacks are crucial for the aggregation of convection, even with the shortwave heating perturbation. At lower SSTs, convection is able to aggregate with the shortwave perturbation in the absence of longwave feedbacks. These perturbations provide a link to studying the effects of absorbing aerosol plumes on convection, for example during the Indian monsoon season. We argue that, as there is aggregation for plumes with realistic aerosol absorption optical depths, this could be an analogue for real‐world organization in regions with high pollution. Plain Language Summary: Tropical clouds grouping together into large clusters in idealized climate models is an interesting topic of research. When this happens, the average temperature increases with more rain where the clouds are and less rain away from them. This paper looks at how we can force clouds to cluster together. We use plumes of aerosols to cause heating perturbations. We show that this causes a large‐scale circulation to develop, with warm air rising near to the plume and cold air sinking further away from it. This helps clouds to cluster together, through changes in the absorption and reflection of longwave and shortwave radiation. We highlight that real‐world pollution plumes or dust events may have similar effects on the clustering of clouds near to them. Key Points: Can increase the degree of aggregation by using a shortwave radiative heating perturbationRadiative heating perturbation incites a global thermally driven circulationShortwave radiative heating perturbation can aggregate convection at lower sea surface temperatures, even without typically crucial longwave feedbacks [ABSTRACT FROM AUTHOR]

Using the depth (z) and density (ϱ) frameworks, we analyze local contributions to AMOC variability in a 900‐year simulation with the AWI climate model. Both frameworks reveal a consistent interdecadal variability; however, the correlation between their maxima deteriorates on year‐to‐year scales. We demonstrate the utility of analyzing the spatial patterns of sinking and diapycnal transformations through depth levels and isopycnals. The success of this analysis relies on the spatial binning of these maps which is especially crucial for the maps of vertical velocities which appear to be too noisy in the main regions of upwelling and downwelling because of stepwise bottom topography. Furthermore, we show that the AMOC responds to fast (annual or faster) fluctuations in atmospheric forcing associated with the NAO. This response is more obvious in the ϱ than in the z framework. In contrast, the link between AMOC and deep water production south of Greenland is found for slower fluctuations and is consistent between the frameworks. Plain Language Summary: The variability and trends in Atlantic meridional overturning circulation (AMOC) are some of the most important characteristics of an ocean model simulation that are relevant for climate. Usually, the AMOC is computed as a meridional transport (from bottom to top) of zonally integrated flow along constant depth levels (z‐AMOC). However, water in the ocean generally flows along the density surfaces. In low latitudes, they are nearly flat and the general picture given by z‐AMOC is highly relevant. In high latitudes, where the most important processes of dense water formation and sinking take place, density surfaces are sloping, and the z‐AMOC gives a distorted image of the real circulation. This is why recent studies underlined the importance of computing the AMOC as a function of density in addition to AMOC as the function of vertical coordinate. In this paper, we analyze the fundamental differences with respect to AMOC variability in both frameworks in a 900‐year run with the AWI climate model. We demonstrate that the latitudinal position of overturning maxima, amplitude, and variability show substantial differences between the frameworks. We suggest that the ϱ‐AMOC and watermass transformation framework should be used routinely in standard analyses, including forthcoming intercomparison projects. Key Points: Patterns of sinking and diapycnal transformation across depth levels and isopycnals, respectively, can be used to study AMOC variabilityThe AMOC subpolar maximum is largely driven by internal transformationsThe density framework illustrates the interplay between surface buoyancy flux and interior‐mixing [ABSTRACT FROM AUTHOR]

Climate predictions using coupled models in different time scales, from intraseasonal to decadal, are usually affected by initial shocks, drifts, and biases, which reduce the prediction skill. These arise from inconsistencies between different components of the coupled models and from the tendency of the model state to evolve from the prescribed initial conditions toward its own climatology over the course of the prediction. Aiming to provide tools and further insight into the mechanisms responsible for initial shocks, drifts, and biases, this paper presents a novel data set developed within the Long Range Forecast Transient Intercomparison Project, LRFTIP. This data set has been constructed by averaging hindcasts over available prediction years and ensemble members to form a hindcast climatology, that is a function of spatial variables and lead time, and thus results in a useful tool for characterizing and assessing the evolution of errors as well as the physical mechanisms responsible for them. A discussion on such errors at the different time scales is provided along with plausible ways forward in the field of climate predictions. Plain Language Summary: The prediction of climate from weeks to years is usually made using complex numerical models that need to have an appropriate representation of the different components of the climate system (atmosphere, oceans, land areas, etc.,) as well as of their interactions. Most of these simulations, however, are affected by errors that originate in part from the numerical methods and approximations used to solve the system of equations of the atmospheric and oceanic systems, and in part by an incorrect representation of the interaction between components. Naturally, these errors lead to a significant decrease in the skill of the predictions, and therefore, must be taken into account. This study presents a new data set made from climate model predictions that have been manipulated in order to be useful for the assessment and further understanding of the physical mechanisms behind the aforementioned errors. The expected increase in our knowledge in the field will help us obtain more skillful climate predictions in the future. Key Points: Initial shocks, drifts, and biases affect climate predictions at all time scalesAs they usually lead to a substantial reduction of the prediction skill, further efforts toward their understanding are neededThe newly developed data set presented in this article is expected to serve as a useful tool for their study [ABSTRACT FROM AUTHOR]

Parameters in climate models are usually calibrated manually, exploiting only small subsets of the available data. This precludes both optimal calibration and quantification of uncertainties. Traditional Bayesian calibration methods that allow uncertainty quantification are too expensive for climate models; they are also not robust in the presence of internal climate variability. For example, Markov chain Monte Carlo (MCMC) methods typically require O(105) model runs and are sensitive to internal variability noise, rendering them infeasible for climate models. Here we demonstrate an approach to model calibration and uncertainty quantification that requires only O(102) model runs and can accommodate internal climate variability. The approach consists of three stages: (a) a calibration stage uses variants of ensemble Kalman inversion to calibrate a model by minimizing mismatches between model and data statistics; (b) an emulation stage emulates the parameter‐to‐data map with Gaussian processes (GP), using the model runs in the calibration stage for training; (c) a sampling stage approximates the Bayesian posterior distributions by sampling the GP emulator with MCMC. We demonstrate the feasibility and computational efficiency of this calibrate‐emulate‐sample (CES) approach in a perfect‐model setting. Using an idealized general circulation model, we estimate parameters in a simple convection scheme from synthetic data generated with the model. The CES approach generates probability distributions of the parameters that are good approximations of the Bayesian posteriors, at a fraction of the computational cost usually required to obtain them. Sampling from this approximate posterior allows the generation of climate predictions with quantified parametric uncertainties. Plain Language Summary: Calibrating climate models with available data and quantifying their uncertainties are essential to make climate predictions accurate and actionable. A primary source of uncertainties in climate models comes from representation of small‐scale processes such as moist convection. Parameters in convection schemes and other parameterizations are usually calibrated by hand, using only a small fraction of data that are available. As a result, the calibration process may miss information about the small‐scale processes in question. This paper presents a proof‐of‐concept, in an idealized setting, of how parameters in climate models can be calibrated using a substantial fraction of the available data, and how uncertainties in the parameters can be quantified. We employ a new algorithm, called calibrate‐emulate‐sample (CES), which makes such calibration and uncertainty quantification feasible for computationally expensive climate models. CES reduces the hundreds of thousands of model runs usually required to quantify uncertainties in computer models to hundreds, thereby achieving about a factor 1,000 speedup. It leads to more robust calibration and uncertainty quantification in the presence of noise arising from chaotic variability of the climate system. We show how uncertainties in climate model parameters can be translated into quantified uncertainties of climate predictions through ensemble integrations. Key Points: We use time averaged climate statistics to calibrate convective parameters and quantify their uncertaintiesWe demonstrate use of the calibrate‐emulate‐sample algorithm to provide efficient calibration and uncertainty quantificationParametric uncertainty in climate predictions is quantified by sampling from the learnt convective parameter distribution [ABSTRACT FROM AUTHOR]

A two‐moment Morrison‐Gettelman bulk cloud microphysics with prognostic precipitation (MG2), together with a mineral dust and temperature‐dependent ice nucleation scheme, have been implemented into the Geophysical Fluid Dynamics Laboratory's Atmosphere Model version 4.0 (AM4.0). We refer to this configuration as AM4‐MG2. This paper describes the configuration of AM4‐MG2, evaluates its performance, and compares it with AM4.0. It is shown that the global simulations with AM4‐MG2 compare favorably with observations and reanalyses. The model skill scores are close to AM4.0. Compared to AM4.0, improvements in AM4‐MG2 include (a) better coastal marine stratocumulus and seasonal cycles, (b) more realistic ice fraction, and (c) dominant accretion over autoconversion. Sensitivity tests indicate that nucleation and sedimentation schemes have significant impacts on cloud liquid and ice water fields, but higher horizontal resolution (about 50 km instead of 100 km) does not. Plain Language Summary: We have implemented a sophisticated cloud microphysical scheme in the Geophysical Fluid Dynamics Laboratory's Atmosphere Model AM4.0. This microphysical scheme predicts both mass and number concentrations of cloud droplets, ice crystals, rain, and snow. Additionally, the nucleation of mineral dust aerosols to ice crystals has been considered as part of an effort to represent aerosol‐cloud‐ice interactions more realistically. We conducted multi‐year global simulations. Simulation results turn out to be in good agreement with observations, and show improvements in coastal stratocumulus, cloud liquid and ice phase partitioning, and the ratio of accretion (collection of cloud drops by raindrops) over autoconversion (collision and coalesce between droplets to form raindrops). Key Points: Two‐moment microphysics with prognostic precipitation and ice nucleation schemes have been implemented in Atmosphere Model version 4.0 (AM4.0), referred to as AM4‐MG2The overall skill scores of AM4‐MG2 are close to AM4.0Improvements include better coastal stratocumulus, more realistic liquid and ice partitioning, dominant accretion over autoconversion [ABSTRACT FROM AUTHOR]

Convection is usually parameterized in global climate models, and there are often large discrepancies between results obtained with different convection schemes. Conventional methods of comparing convection schemes using observational cases or directly in three‐dimensional (3D) models do not always clearly identify parameterization strengths and weaknesses. In this paper we evaluate the response of parameterizations to various perturbations rather than their behavior under particular strong forcing. We use the linear response function method proposed by Kuang (2010) to compare 12 physical packages in five atmospheric models using single‐column model (SCM) simulations under idealized radiative‐convective equilibrium conditions. The models are forced with anomalous temperature and moisture tendencies. The temperature and moisture departures from equilibrium are compared with published results from a cloud‐resolving model (CRM). Results show that the procedure is capable of isolating the behavior of a convection scheme from other physics schemes. We identify areas of agreement but also substantial differences between convection schemes, some of which can be related to scheme design. Some aspects of the model linear responses are related to their RCE profiles (the relative humidity profile in particular), while others constitute independent diagnostics. All the SCMs show irregularities or discontinuities in behavior that are likely related to threshold‐related mechanisms used in the convection schemes, and which do not appear in the CRM. Our results highlight potential flaws in convection schemes and suggest possible new directions to explore for parameterization evaluation. Plain Language Summary: The transport of heat up and down the atmosphere, called atmospheric convection, is a complex process. To simplify the representation of convection in global climate models (GCMs) scientists use "parameterization," which is essentially mathematical equations of physical processes. However, there are many different ways to formulate these equations, and no agreement on which is better. In this work we aim to understand a few popular ways to parameterize convection. We extract one vertical column from five different GCMs and lightly tickle (perturb) it and then observe its responses. We found that different models respond very differently to the same tickling, and this tells us a lot about the model. Importantly, the specific perturbation that we used can single out the responses of convection‐related equations from equations of other processes. All the models in our study have one thing in common: They are quite jumpy when tickled, especially at the top of the boundary layer where clouds start to form. We suspect the culprits are thresholds placed in the models that sometimes lead to sudden changes in their response. Our work highlights potentially problematic behavior that can give clues on how to make climate models better. Key Points: The linear response function method is applied in single‐column model (SCM) simulations and is able to isolate the behavior of convection schemesLinear responses of the models are related to the mean state relative humidity in both shape and magnitudeAll SCMs show discontinuities in their responses which are likely related to thresholds used in convective parameterization [ABSTRACT FROM AUTHOR]