Search

In climate models, subgrid parameterizations of convection and clouds are one of the main causes of the biases in precipitation and atmospheric circulation simulations. In recent years, due to the rapid development of data science, machine learning (ML) parameterizations for convection and clouds have been demonstrated to have the potential to perform better than conventional parameterizations. Most previous studies were conducted on aqua-planet and idealized models, and the problems of simulation instability and climate drift still exist. Developing an ML parameterization scheme remains a challenging task in realistically configured models. In this paper, a set of residual deep neural networks (ResDNNs) with a strong nonlinear fitting ability is designed to emulate a super-parameterization (SP) with different outputs in a hybrid ML–physical general circulation model (GCM). It can sustain stable simulations for over 10 years under real-world geographical boundary conditions. We explore the relationship between the accuracy and stability by validating multiple deep neural network (DNN) and ResDNN sets in prognostic runs. In addition, there are significant differences in the prognostic results of the stable ResDNN sets. Therefore, trial and error is used to acquire the optimal ResDNN set for both high skill and long-term stability, which we name the neural network (NN) parameterization. In offline validation, the neural network parameterization can emulate the SP in mid- to high-latitude regions with a high accuracy. However, its prediction skill over tropical ocean areas still needs improvement. In the multi-year prognostic test, the hybrid ML–physical GCM simulates the tropical precipitation well over land and significantly improves the frequency of the precipitation extremes, which are vastly underestimated in the Community Atmospheric Model version 5 (CAM5), with a horizontal resolution of 1.9∘ × 2.5∘. Furthermore, the hybrid ML–physical GCM simulates the robust signal of the Madden–Julian oscillation with a more reasonable propagation speed than CAM5. However, there are still substantial biases with the hybrid ML–physical GCM in the mean states, including the temperature field in the tropopause and at high latitudes and the precipitation over tropical oceanic regions, which are larger than those in CAM5. This study is a pioneer in achieving multi-year stable climate simulations using a hybrid ML–physical GCM under actual land–ocean boundary conditions that become sustained over 30 times faster than the target SP. It demonstrates the emerging potential of using ML parameterizations in climate simulations.

International audience; We examine differences among reanalysis high-cloud products in the tropics, assess the impacts of these differences on radiation budgets at the top of the atmosphere and within the tropical upper troposphere and lower stratosphere (UTLS), and discuss their possible origins in the context of the reanalysis models. We focus on the ERA5 (fifth-generation European Centre for Medium-range Weather Forecasts-ECMWF-reanalysis), ERA-Interim (ECMWF Interim Reanalysis), JRA-55 (Japanese 55-year Reanaly-sis), MERRA-2 (Modern-Era Retrospective Analysis for Research and Applications, Version 2), and CFSR/CFSv2 (Cli-mate Forecast System Reanalysis/Climate Forecast System Version 2) reanalyses. As a general rule, JRA-55 produces the smallest tropical high-cloud fractions and cloud water contents among the reanalyses, while MERRA-2 produces the largest. Accordingly, long-wave cloud radiative effects are relatively weak in JRA-55 and relatively strong in MERRA-2. Only MERRA-2 and ERA5 among the reanaly-ses produce tropical-mean values of outgoing long-wave radiation (OLR) close to those observed, but ERA5 tends to underestimate cloud effects, while MERRA-2 tends to overestimate variability. ERA5 also produces distributions of long-wave, shortwave , and total cloud radiative effects at the top of the atmosphere that are very consistent with those observed. The other reanalyses all exhibit substantial biases in at least one of these metrics, although compensation between the long-wave and shortwave effects helps to constrain biases in the total cloud radiative effect for most reanalyses. The vertical distribution of cloud water content emerges as a key difference between ERA-Interim and other reanalyses. Whereas ERA-Interim shows a monotonic decrease of cloud water content with increasing height, the other reanalyses all produce distinct anvil layers. The latter is in better agreement with observations and yields very different profiles of radia-tive heating in the UTLS. For example, whereas the altitude of the level of zero net radiative heating tends to be lower in convective regions than in the rest of the tropics in ERA-Interim, the opposite is true for the other four reanalyses. Differences in cloud water content also help to explain systematic differences in radiative heating in the tropical lower stratosphere among the reanalyses. We discuss several ways in which aspects of the cloud and convection schemes impact the tropical environment. Discrepancies in the vertical profiles of temperature and specific humidity in convective regions are particularly noteworthy, as these variables are directly constrained by data assimilation, are widely used, and Published by Copernicus Publications on behalf of the European Geosciences Union. 8990 J. S. Wright et al.: Tropical high clouds in reanalyses feed back to convective behaviour through their relationships with thermodynamic stability.

Both frequency and intensity of rainfall affect aerosol wet deposition. With a stochastic deep convection scheme implemented into two state-of-the-art global climate models (GCMs), a recent study found that aerosol burdens are increased globally by reduced climatological mean wet removal of aerosols due to suppressed light rain. Motivated by their work, a novel approach is developed in this study to detect what rainfall rates are most efficient for wet removal (scavenging amount mode) of different aerosol species in different sizes in GCMs and applied to the National Center for Atmospheric Research Community Atmosphere Model version 5 (CAM5) with and without the stochastic convection cases. Results show that in the standard CAM5, no obvious differences in the scavenging amount mode are found among different aerosol types. However, the scavenging amount modes differ in the Aitken, accumulation and coarse modes showing around 10-12, 8-9, and 7-8 mm d−1, respectively over the tropics. As latitude increases poleward, the scavenging amount mode in each aerosol mode is decreased substantially. The scavenging amount mode is generally smaller over land than over ocean. With stochastic convection, the scavenging amount mode for all aerosol species in each mode is systematically increased, which is the most prominent along the Intertropical Convergence Zone exceeding 20 mm d−1 for small particles. Regardless of whether the stochastic convection scheme is used, convective precipitation has higher efficiency in removing aerosols than large-scale precipitation over the globe even though convection is infrequent over high-latitudes. The scavenging amount modes in the two cases are both smaller than individual rainfall rates associated with the most accumulated rain (rainfall amount mode), further implying precipitation frequency is more important than precipitation intensity for aerosol wet removal. The notion of the scavenging amount mode can be applied to other GCMs to better understand the relation between rainfall and aerosol wet scavenging, which is important to better simulating aerosols.

A stochastic deep convection parameterization is implemented into the U.S. Department of Energy (DOE) Energy Exascale Earth System Model (E3SM) Atmosphere Model version 1 (EAMv1). This study evaluates its performance on the precipitation simulation. Compared to the default model, the probability distribution function (PDF) of rainfall intensity in the new simulation is greatly improved. Especially, the well-known problem of too much light rain and too little heavy rain is alleviated over the tropics. As a result, the contribution from different rain rates to the total precipitation amount is shifted toward heavier rain. The less frequent occurrence of convection contributes to the suppressed light rain, while both more intense large-scale and convective precipitation contribute to the enhanced heavy total rain. The synoptic and intraseasonal variabilities of precipitation are enhanced as well to be closer to observations. A sensitivity of the rainfall intensity PDF to the model vertical resolution is identified and explained in terms of the relationships between convective precipitation and convective available potential energy (CAPE) and between large-scale precipitation and resolved-scale upward moisture flux. The annual mean precipitation is largely unchanged with the use of the stochastic scheme except over the tropical western Pacific, where a moderate increase in precipitation represents a slight improvement. The responses of precipitation and its extremes to climate warming are similar with or without the stochastic deep convection scheme.

The Energy Exascale Earth System Model (E3SM) developed by the Department of Energy has a goal of addressing challenges in understanding the global water cycle. Success depends on correct simulation of cloud and precipitation elements. However, lack of appropriate evaluation metrics has hindered the accurate representation of these elements in general circulation models. We derive metrics from the three-dimensional data of the ground-based Next generation radar (NEXRAD) network over the U.S. to evaluate both horizontal and vertical structures of precipitation elements. We coarsened the resolution of the radar observations to be consistent with the model resolution and improved the coupling of the Cloud Feedback Model Intercomparison Project Observation Simulator Package (COSP) and E3SM Atmospheric Model Version 1 (EAMv1) to obtain the best possible model output for comparison with the observations. Three warm seasons (2014–2016) of EAMv1 simulations of 3D radar reflectivity features at an hourly scale are evaluated. A general agreement in domain-mean radar reflectivity intensity is found between EAMv1 and NEXRAD below 4 km altitude; however, the model underestimates reflectivity over the central United States, which suggests that the model does not capture the mesoscale convective systems that produce much of precipitation in that region. The shape of the model estimated histogram of subgrid scale reflectivity is improved by correcting the microphysical assumptions in COSP. The model severely underestimates radar reflectivity at upper levels – the simulated echo top height is about 4 km lower than in observations – and this result is not changed by tuning any single physics parameter.

It is an old question whether tropospheric water vapor at different levels changes consistently in response to the enhanced greenhouse gas in the atmosphere. Earlier studies using older versions of climate models and available data revealed a significant difference between models and observations. Water vapor changes in the interior of the tropical troposphere have been found to be more strongly coupled to changes at the surface in climate models than in observations. We reexamine this issue using four leading CMIP5 models (CCSM4, HadGEM2-A, GFDL-CM3 and MPI-ESM-MR) and more updated observational datasets (ERA-Interim and NCEP reanalysis). Focusing on the Tropics, we have calculated the correlations between interannual variation of specific humidity in all levels of the troposphere with that at the surface. It is found that the previously noted biases in the strength of the coupling between water vapor changes in the interior of the troposphere and those at the surface still exist in the updated models—the change in the tropical averaged tropospheric water vapor is more strongly correlated with the change in the surface, especially in the middle troposphere. It is argued that the vertical profile of water vapor correlations in observations is more consistent with the “hot tower” concept for tropical convections. Zonal mean correlation results and those from the moisture regime sorting method are consistent with each other, both of which indicate the role of deep convection as a mechanism to couple the middle tropospheric water vapor and that in the surface and that an inaccurate representation of deep convection as a possible cause for the discrepancies between models and observations in the coupling between middle tropospheric water vapor and those at the surface.