Search

Gravity wave parameterizations contribute to uncertainties in middle atmosphere modeling. To investigate the potential for using machine learning to represent atmospheric gravity waves and the impact of implementing such schemes in a general circulation model (GCM), we train a random forest (RF) emulator on outputs from an existing complex parameterization scheme for orographic gravity wave drag (GWD). The performance of the RF emulator is then evaluated with a focus on stratospheric climatology and variability in climate simulations from the middle atmosphere resolving Beijing Climate Center Atmospheric Circulation Model. In offline tests, the predicted orographic GWD by the RF agrees remarkably well with the target GWD throughout the troposphere and the middle atmosphere. The RF emulator can reproduce the observed climatology of zonal‐mean wind and air temperature in the GCM simulation, as well as its target scheme. Compared to the target orographic GWD parameterization scheme, the RF emulator can reproduce the breakdown of the polar vortex in the Southern Hemisphere stratosphere. This study demonstrates the feasibility of using machine learning to emulate parameterized orographic GWD for modeling the stratosphere with a GCM. Plain Language Summary: Machine learning has been utilized to learn atmospheric physical parameterizations such as moist convection, cloud microphysics, radiation, nonorographic gravity waves, etc. However, it remains unknown whether it is feasible to apply machine learning algorithms to parameterize orographic gravity waves and how such schemes perform when interactively coupled in atmospheric general circulation models. Here we employ the widely used random forest algorithm to learn from an existing complex parameterization scheme for orographic gravity wave drag. Then, we implement the random forest scheme in a high‐top atmospheric general circulation model to test its performance. We find that the random forest scheme acts as well as the target parameterization scheme. This study demonstrates the potential of using machine learning to emulate parameterized orographic gravity waves for modeling the middle atmosphere. In future work, it is recommended that machine learning may be used to develop new parameterization schemes directly from observations or high‐resolution model outputs to overcome critical deficiencies in current orographic gravity wave parameterizations, helping to reduce the common biases in simulating the stratosphere with general circulation models. Key Points: Parameterized gravity wave drag from unresolved orography can be emulated with a random forestThe random forest emulator simulates vertical distributions of the zonal wind and air temperature quite well in a general circulation model (GCM)The GCM configured with the random forest scheme reproduces a reasonable breakdown of the austral polar vortex in the stratosphere [ABSTRACT FROM AUTHOR]

Copyright of Advances in Atmospheric Sciences is the property of Springer Nature and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use. This abstract may be abridged. No warranty is given about the accuracy of the copy. Users should refer to the original published version of the material for the full abstract. (Copyright applies to all Abstracts.)

BCC-CSM2-HR is a high-resolution version of the Beijing Climate Center (BCC) Climate System Model. Its development is on the basis of the medium-resolution version BCC-CSM2-MR which is the baseline for BCC participation to the Coupled Model Intercomparison Project Phase 6 (CMIP6). This study documents the high-resolution model, highlights major improvements in the representation of atmospheric dynamic core and physical processes. BCC-CSM2-HR is evaluated for present-day climate simulations from 1971 to 2000, which are performed under CMIP6-prescribed historical forcing, in comparison with its previous medium-resolution version BCC-CSM2-MR. We focus on basic atmospheric mean states over the globe and variabilities in the tropics including the tropic cyclones (TCs), the El Niño–Southern Oscillation (ENSO), the Madden-Julian Oscillation (MJO), and the quasi-biennial oscillation (QBO) in the stratosphere. It is shown that BCC-CSM2-HR keeps well the global energy balance and can realistically reproduce main patterns of atmosphere temperature and wind, precipitation, land surface air temperature and sea surface temperature. It also improves in the spatial patterns of sea ice and associated seasonal variations in both hemispheres. The bias of double intertropical convergence zone (ITCZ), obvious in BCC-CSM2-MR, is almost disappeared in BCC-CSM2-HR. TC activity in the tropics is increased with resolution enhanced. The cycle of ENSO, the eastward propagative feature and convection intensity of MJO, the downward propagation of QBO in BCC-CSM2-HR are all in a better agreement with observation than their counterparts in BCC-CSM2-MR. We also note some weakness in BCC-CSM2-HR, such as the excessive cloudiness in the eastern basin of the tropical Pacific with cold Sea Surface Temperature (SST) biases and the insufficient number of tropical cyclones in the North Atlantic. [ABSTRACT FROM AUTHOR]

By conducting several sets of hindcast experiments using the Beijing Climate Center Climate System Model, which participates in the Sub-seasonal to Seasonal (S2S) Prediction Project, we systematically evaluate the model's capability in forecasting MJO and its main deficiencies. In the original S2S hindcast set, MJO forecast skill is about 16 days. Such a skill shows significant seasonal-to-interannual variations. It is found that the model-dependent MJO forecast skill is more correlated with the Indian Ocean Dipole (IOD) than with the El Niño-Southern Oscillation. The highest skill is achieved in autumn when the IOD attains its maturity. Extended skill is found when the IOD is in its positive phase. MJO forecast skill's close association with the IOD is partially due to the quickly strengthening relationship between MJO amplitude and IOD intensity as lead time increases to about 15 days, beyond which a rapid weakening of the relationship is shown. This relationship transition may cause the forecast skill to decrease quickly with lead time, and is related to the unrealistic amplitude and phase evolutions of predicted MJO over or near the equatorial Indian Ocean during anomalous IOD phases, suggesting a possible influence of exaggerated IOD variability in the model. The results imply that the upper limit of intraseasonal predictability is modulated by large-scale external forcing background state in the tropical Indian Ocean. Two additional sets of hindcast experiments with improved atmosphere and ocean initial conditions (referred to as S2S_IEXP1 and S2S_IEXP2, respectively) are carried out, and the results show that the overall MJO forecast skill is increased to 21-22 days. It is found that the optimization of initial sea surface temperature condition largely accounts for the increase of the overall MJO forecast skill, even though the improved initial atmosphere conditions also play a role. For the DYNAMO/CINDY field campaign period, the forecast skill increases to 27 days in S2S_IEXP2. Nevertheless, even with improved initialization, it is still difficult for the model to predict MJO propagation across the western hemisphere-western Indian Ocean area and across the eastern Indian Ocean-Maritime Continent area. Especially, MJO prediction is apparently limited by various interrelated deficiencies (e.g., overestimated IOD, shorter-than-observed MJO life cycle, Maritime Continent prediction barrier), due possibly to the model bias in the background moisture field over the eastern Indian Ocean and Maritime Continent. Thus, more efforts are needed to correct the deficiency in model physics in this region, in order to overcome the well-known Maritime Continent predictability barrier. [ABSTRACT FROM AUTHOR]

A Deterministic Time-lagged Ensemble Forecast using a Probabilistic Threshold (DEFPT) method is suggested for improving summer 6–15 day categorical precipitation prediction in China from the Beijing Climate Center Atmospheric General Circulation Model version 2.1 (BCC_AGCM2.1). It is based on a time-lagged ensemble system that consists of 13 ensemble members separated sequentially at 6 hour intervals lagging the last three days. The DEFPT is not intended to predict the probability of rainfall, but rather to forecast rainfall (yes/no) occurrence for different categories of precipitation at any model grid box. A given categorical precipitation is forecasted to occur at one gridbox only when the ensemble probability for that categorical precipitation exceeds a certain threshold. This method is useful for providing an estimate of whether precipitation events will occur to decision-makers based on probabilistic forecasts during days 6–15. A large number of hindcast experiments for 1996–2005 summers reveal that this threshold can be best (and empirically) set as 5/13 and 4/13 respectively for the 6–15 day prediction of 1 + mm (i.e., above 1 mm per day) and 5 + mm rainfall events, using the Relative Operating Characteristic (ROC) curve, the Equitable Threat Score (ETS), the Hanssen and Kuipers (HK) score, and frequency bias (BIA) to achieve best prediction performance. With this set of thresholds, the DEFPT shows skill improvement over the corresponding single deterministic forecast using one initial value and the Time-Lagged Average Forecast (LAF) ensemble method. Similar improvements by the DEFPT are also found for the prediction of several other categories of precipitation between 1 + mm and 10 + mm per day. Application of DEFPT to larger ensemble size and BCC_AGCM version 2.2 with a higher horizontal resolution also demonstrates the effectiveness of the DEFPT for 6–15 day categorical precipitation forecasts. [ABSTRACT FROM AUTHOR]