Because longwave (LW) absorption by greenhouse gases and clouds is more significant than the LW scattering effect by clouds, most climate models neglect cloud LW scattering to save computational costs. Ignoring cloud LW scattering directly overestimates outgoing longwave radiation (OLR). This study included ice-cloud LW scattering treatment in the Exascale Energy Earth System Model (E3SM) version 2 and ran fully-coupled simulations, prescribed sea surface temperature simulations, and offline radiative transfer calculations to comprehensively assess the impact of ice-cloud LW scattering on global climate simulation. The instantaneous effect due to ice-cloud LW scattering reduces the OLR by ~1 W/m2 on the global average and 2 W/m2 on the tropical average. Tropospheric warming and high cloud amount reduction act to partially compensate for such instantaneous OLR reduction caused by the inclusion of LW scattering. When the simulation reaches the equilibrium, the surface warms by 0.66 K on average with respect to the simulation without LW scattering, with the Arctic surface temperature differences more than twice as large as that of the global mean. The impact of including LW scattering on the simulated climate change in response to 4×CO2 is also assessed. While including the cloud LW scattering does not significantly modify radiative forcing and total radiative feedback under such a scenario, it results in a 10% more positive cloud feedback.

Parameterizations in Earth System Models (ESMs) are subject to biases and uncertainties arising from subjective empirical assumptions and incomplete understanding of the underlying physical processes. Recently, the growing representational capability of machine learning (ML) in solving complex problems has spawned immense interests in climate science applications. Specifically, ML-based parameterizations have been developed to represent convection, radiation and microphysics processes in ESMs by learning from observations or high-resolution simulations, which have the potential to improve the accuracies and alleviate the uncertainties. Previous works have developed some surrogate models for these processes using ML. These surrogate models need to be coupled with the dynamical core of ESMs to investigate the effectiveness and their performance in a coupled system. In this study, we present a novel Fortran-Python interface designed to seamlessly integrate ML parameterizations into ESMs. This interface showcases high versatility by supporting popular ML frameworks like PyTorch, TensorFlow, and Scikit-learn. We demonstrate the interface’s modularity and reusability through two cases: a ML trigger function for convection parameterization and a ML wildfire model. We conduct a comprehensive evaluation of memory usage and computational overhead resulting from the integration of Python codes into the Fortran ESMs. By leveraging this flexible interface, ML parameterizations can be effectively developed, tested, and integrated into ESMs.

Mesoscale convective systems (MCSs) play an important role in modulating the global hydrological cycle, general circulation, and radiative energy budget. In this study, we evaluate MCS simulations in the second version of U.S. Department of Energy (DOE) Energy Exascale Earth System Model (E3SMv2). E3SMv2 atmosphere model (EAMv2) is run at the uniform 0.25° horizontal resolution. We track MCSs consistently in the model and observations using the PyFLEXTRKR algorithm, which defines MCS based on both cloud-top brightness temperature (Tb) and surface precipitation. Results from using Tb only to define MCS, commonly used in previous studies, are also discussed. Furthermore, sensitivity experiments are performed to examine the impact of new cloud and convection parameterizations developed for EAMv3 on simulated MCSs. Our results show that EAMv2 simulated MCS precipitation is largely underestimated in the tropics and contiguous United States. This is mainly attributed to the underestimated precipitation intensity in EAMv2. In contrast, the simulated MCS frequency becomes more comparable to observations if MCSs are defined only based on cloud-top Tb. The Tb-based MCS tracking method, however, includes many cloud systems with very weak precipitation which conflicts with the MCS definition. This result illustrates the importance of accounting for precipitation in evaluating simulated MCSs. We also find that the new physics parameterizations help increase the relative contribution of convective precipitation to total precipitation in the tropics, but the simulated MCS properties are overall not significantly improved. This suggests that simulating MCSs will remain a challenge for the next version of E3SM.

In recent years, machine learning (ML) models have been used for improving physical parameterizations of general circulation models (GCMs). A significant challenge of integrating ML models into GCMs is the online instability when they are coupled for long-term simulation. In this study, we present a new strategy that demonstrates robust online stability when the entire physical parameterization package of a GCM is replaced by a deep ML algorithm. The method uses a multistep training scheme of the machine learning model with experience replay in which the memory of physical tendencies from the training dataset and the ML algorithm’s own output at the previous time step are used in the training. The physics memory improves the accuracy of the machine learning model, while the experience replay constrains the amplification of cumulative errors in the online coupling. The method is used to train the whole physical parameterization package for the Community Atmosphere Model version 5 (CAM5) with data from its Multi-scale Modeling Framework (MMF) high resolution simulations. Three 6-year online simulations of the CAM5 with the ML physics package at operational spatial resolution with real-world geography are presented. The simulated spatial distributions of precipitation, surface temperature and zonally averaged atmospheric fields demonstrate overall better accuracy than that of the standard CAM5 and benchmark model even without the use of additional physical constraints or tuning. This work is the first to demonstrate a solution to address the online instability problem in climate modeling with ML physics by using experience replay.

Skillful representation of tropical variability and diurnal cycle of precipitation has remained a challenge in global atmosphere models, and often improvements in the variability lead to degradation in the mean-state. Here, we introduce a configuration of the E3SM Atmosphere Model with a new microphysics scheme and several enhancements to the deep convective scheme that improves the variability while keeping the mean-state climatology largely unchanged. The new configuration improves various modes of convectively-coupled equatorial waves, with increased strength of Kelvin waves and more coherent eastward propagation of the Madden Julian Oscillation from the Indian Ocean to the central Pacific Ocean. The same configuration also improves both the phase and amplitude of the diurnal cycle of precipitation, particularly over the continental United States in the boreal summer and over Tropical land regions. Previous studies have shown that, individually taken, some of the deep convective enhancements can improve certain aspects of the variability, and here we show that combining their effects can lead to robust improvements in the variability. This model configuration can form the basis for future studies to examine the response of tropical and diurnal variability under various climate states and their relationships with other modes of variability.

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 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 yrs 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.

This study evaluates high-latitude stratiform mixed-phase clouds (SMPC) in the atmosphere model of the newly released Energy Exascale Earth System Model version 2 (EAMv2) by utilizing one-year-long ground-based remote sensing measurements from the U.S. Department of Energy Atmospheric Radiation and Measurement (ARM) Program. A nudging approach is applied to model simulations for a better comparison with the ARM observations. Observed and modeled SMPCs are collocated to evaluate their macro- and microphysical properties at the ARM North Slope of Alaska (NSA) site in the Arctic and the McMurdo (AWR) site in the Antarctic. We found that EAMv2 overestimates (underestimates) SMPC frequency of occurrence at the NSA (AWR) site nearly all year round. However, the model captures the observed larger cloud frequency of occurrence at the NSA site. For collocated SMPCs, the annual statistics of observed cloud macrophysics are generally reproduced at the NSA site, while at the AWR site, there are larger biases. Compared to the AWR site, the lower cloud boundaries and the warmer cloud top temperature observed at NSA are well simulated. On the other hand, simulated cloud phases are substantially biased at each location. The model largely overestimates liquid water path at NSA, whereas it is frequently underestimated at AWR. Meanwhile, the simulated ice water path is underestimated at NSA, but at AWR, it is comparable to observations. As a result, the observed hemispheric difference in cloud phase partitioning is misrepresented in EAMv2. This study implies that continuous improvement in cloud microphysics is needed for high-latitude mixed-phase clouds.

This work documents version two of the Department of Energy’s Energy Exascale Earth System Model (E3SM). E3SM version 2 (E3SMv2) is a significant evolution from its predecessor E3SMv1, resulting in a model that is nearly twice as fast and with a simulated climate that is improved in many metrics. We describe the physical climate model in its lower horizontal resolution configuration consisting of 110 km atmosphere, 165 km land, 0.5° river routing model, and an ocean and sea ice with mesh spacing varying between 60 km in the mid-latitudes and 30 km at the equator and poles. The model performance is evaluated by means of a standard set of Coupled Model Intercomparison Project Phase 6 (CMIP6) Diagnosis, Evaluation, and Characterization of Klima (DECK) simulations augmented with historical simulations as well as simulations to evaluate impacts of different forcing agents. The simulated climate is generally realistic, with notable improvements in clouds and precipitation compared to E3SMv1. E3SMv1 suffered from an excessively high equilibrium climate sensitivity (ECS) of 5.3 K. In E3SMv2, ECS is reduced to 4.0 K which is now within the plausible range based on a recent World Climate Research Programme (WCRP) assessment. However, E3SMv2 significantly underestimates the global mean surface temperature in the second half of the historical record. An analysis of single-forcing simulations indicates that correcting the historical temperature bias would require a substantial reduction in the magnitude of the aerosol-related forcing.

Mesoscale organization of convection is typically not represented in global circulation models, and hence its influence on the global circulation is not accounted for. A parameterization aiming at representing the dynamical and physical effects of the circulation associated with organized convection, referred to as the multiscale coherent structure parameterization (MCSP), is implemented in the Energy Exascale Earth System Model version 1 (E3SMv1). Simulations are conducted to assess its impact on the simulated climate. Besides E3SMv1 simulations, we performed high-resolution (1 km) simulations using the Weather Research and Forecasting (WRF) Model to determine the temperature tendencies induced by mesoscale convective systems embedded in deep convection. We tuned the free parameters of the MCSP based on the WRF simulations. We found that the MCSP enhances Kevin wave spectra in E3SMv1, improves the representation of the Madden-Julian Oscillation, and reduces model precipitation biases over the tropical Pacific.

Accurate forecast of solar irradiance remains a major challenge, especially under the influences of aerosols, clouds and aerosol-cloud interactions due to their inadequate parameterizations in numerical prediction models. This study focuses on the impacts of cloud microphysics and the indirect aerosol effect on solar irradiance. The state of art Weather Research and Forecasting model specifically designed for simulating and forecasting solar radiation (WRF-Solar) is employed to investigate the sensitivity of the total solar irradiance and its partitioning into direct and diffuse irradiances to aerosol and cloud properties. First, a number of microphysical schemes will be tested against the measurements of shallow cumulus and stratiform clouds at the DOE ARM SGP site. Efforts will be made to quantify the uncertainty spread. The effects of cloud microphysics on surface solar irradiance will be identified. Second, the indirect aerosol effect on cloud formation and thus surface solar irradiance will be investigated by using the Thompson aerosol aware microphysical scheme and different treatment of aerosols. In particular, we will examine the aerosol indirect effects in different cloud regimes. To address the aforementioned problems, we will introduce a new model evaluation framework based on different WRF-Solar setups (nested WRF, WRF-LES, and single column WRF). In addition, different evaluation metrics will be used, including the RMSE, MAPE, and relative Euclidean distance. The results will provide physical insight into the understanding of aerosol-cloud-radiation interactions and into improving solar radiation forecast in cloudy conditions.

Significant changes are found in the modeled phase partitioning of Arctic mixed-phase clouds in the U.S. Department of Energy (DOE) Energy Exascale Earth System Model (E3SM) Atmosphere Model version 1 (EAMv1) compared to its predecessor, the Community Atmosphere Model version 5 (CAM5). In this study, we aim to understand how the changes in modeled mixed-phase cloud properties are attributed to the updates made in the EAMv1 physical parameterizations. Impacts of the Classical Nucleation Theory (CNT) ice nucleation scheme, the Cloud Layer Unified By Binormals (CLUBB) parameterization, and updated Morrison and Gettelman microphysical scheme (MG2) are examined. Sensitivity experiments using the short-term hindcast approach are performed to isolate the impact of these new features on simulated mixed-phase clouds. Results are compared to the DOE’s Atmospheric Radiation Measurement (ARM) Mixed-Phase Arctic Cloud Experiment (M-PACE) observations. We find that mixed-phase clouds simulated in EAMv1 are overly dominated by supercooled liquid and cloud ice water is substantially underestimated. The individual change of physical parameterizations is found to decrease cloud ice water mass mixing ratio in EAMv1 simulated single-layer mixed-phase clouds. A budget analysis of detailed cloud microphysical processes suggests that the lack of ice particles that participate in the mass growth processes strongly inhibits the mass mixing ratio of cloud ice. The insufficient heterogeneous ice nucleation at temperatures warmer than -15C in CNT and the negligible ice processes in CLUBB are primarily responsible for the significant underestimation of cloud ice water content in the Arctic single-layer mixed-phase clouds.

14 15 Key Points: 16 (1) E3SMv1 captures spatial and temporal variability in the observed dust aerosol optical 17 depth, but underestimates long-range transport. 18 (2) The net direct radiative effect of dust simulated by E3SMv1 is-0.42 Wm-2 with a smaller 19 longwave warming than other recent studies. 20 (3) In addition to emission, dry removal of dust are highly sensitive to the increase of 21 horizontal or vertical model resolution. 22 23 24