Advances in Atmospheric Sciences最新文献

In recent years, deep learning methods have gradually been applied to prediction tasks related to Arctic sea ice concentration, but relatively little research has been conducted for larger spatial and temporal scales, mainly due to the limited time coverage of observations and reanalysis data. Meanwhile, deep learning predictions of sea ice thickness (SIT) have yet to receive ample attention. In this study, two data-driven deep learning (DL) models are built based on the ConvLSTM and fully convolutional U-net (FC-Unet) algorithms and trained using CMIP6 historical simulations for transfer learning and fine-tuned using reanalysis/observations. These models enable monthly predictions of Arctic SIT without considering the complex physical processes involved. Through comprehensive assessments of prediction skills by season and region, the results suggest that using a broader set of CMIP6 data for transfer learning, as well as incorporating multiple climate variables as predictors, contribute to better prediction results, although both DL models can effectively predict the spatiotemporal features of SIT anomalies. Regarding the predicted SIT anomalies of the FC-Unet model, the spatial correlations with reanalysis reach an average level of 89% over all months, while the temporal anomaly correlation coefficients are close to unity in most cases. The models also demonstrate robust performances in predicting SIT and SIE during extreme events. The effectiveness and reliability of the proposed deep transfer learning models in predicting Arctic SIT can facilitate more accurate pan-Arctic predictions, aiding climate change research and real-time business applications.

Fires, including wildfires, harm air quality and essential public services like transportation, communication, and utilities. These fires can also influence atmospheric conditions, including temperature and aerosols, potentially affecting severe convective storms. Here, we investigate the remote impacts of fires in the western United States (WUS) on the occurrence of large hail (size: ⩾ 2.54 cm) in the central US (CUS) over the 20-year period of 2001–20 using the machine learning (ML), Random Forest (RF), and Extreme Gradient Boosting (XGB) methods. The developed RF and XGB models demonstrate high accuracy (> 90%) and F1 scores of up to 0.78 in predicting large hail occurrences when WUS fires and CUS hailstorms coincide, particularly in four states (Wyoming, South Dakota, Nebraska, and Kansas). The key contributing variables identified from both ML models include the meteorological variables in the fire region (temperature and moisture), the westerly wind over the plume transport path, and the fire features (i.e., the maximum fire power and burned area). The results confirm a linkage between WUS fires and severe weather in the CUS, corroborating the findings of our previous modeling study conducted on case simulations with a detailed physics model.

Climate models are vital for understanding and projecting global climate change and its associated impacts. However, these models suffer from biases that limit their accuracy in historical simulations and the trustworthiness of future projections. Addressing these challenges requires addressing internal variability, hindering the direct alignment between model simulations and observations, and thwarting conventional supervised learning methods. Here, we employ an unsupervised Cycle-consistent Generative Adversarial Network (CycleGAN), to correct daily Sea Surface Temperature (SST) simulations from the Community Earth System Model 2 (CESM2). Our results reveal that the CycleGAN not only corrects climatological biases but also improves the simulation of major dynamic modes including the El Niño-Southern Oscillation (ENSO) and the Indian Ocean Dipole mode, as well as SST extremes. Notably, it substantially corrects climatological SST biases, decreasing the globally averaged Root-Mean-Square Error (RMSE) by 58%. Intriguingly, the CycleGAN effectively addresses the well-known excessive westward bias in ENSO SST anomalies, a common issue in climate models that traditional methods, like quantile mapping, struggle to rectify. Additionally, it substantially improves the simulation of SST extremes, raising the pattern correlation coefficient (PCC) from 0.56 to 0.88 and lowering the RMSE from 0.5 to 0.32. This enhancement is attributed to better representations of interannual, intraseasonal, and synoptic scales variabilities. Our study offers a novel approach to correct global SST simulations and underscores its effectiveness across different time scales and primary dynamical modes.

El Niño-Southern Oscillation (ENSO) is the strongest interannual climate mode influencing the coupled ocean-atmosphere system in the tropical Pacific, and numerous dynamical and statistical models have been developed to simulate and predict it. In some simplified coupled ocean-atmosphere models, the relationship between sea surface temperature (SST) anomalies and wind stress (τ) anomalies can be constructed by statistical methods, such as singular value decomposition (SVD). In recent years, the applications of artificial intelligence (AI) to climate modeling have shown promising prospects, and the integrations of AI-based models with dynamical models are active areas of research. This study constructs U-Net models for representing the relationship between SSTAs and τ anomalies in the tropical Pacific; the UNet-derived τ model, denoted as τ_UNet, is then used to replace the original SVD-based τ model of an intermediate coupled model (ICM), forming a newly AI-integrated ICM, referred to as ICM-UNet. The simulation results obtained from ICM-UNet demonstrate their ability to represent the spatiotemporal variability of oceanic and atmospheric anomaly fields in the equatorial Pacific. In the ocean-only case study, the τ_UNet-derived wind stress anomaly fields are used to force the ocean component of the ICM, the results of which also indicate reasonable simulations of typical ENSO events. These results demonstrate the feasibility of integrating an AI-derived model with a physics-based dynamical model for ENSO modeling studies. Furthermore, the successful integration of the dynamical ocean models with the AI-based atmospheric wind model provides a novel approach to ocean-atmosphere interaction modeling studies.

Accurate radar quantitative precipitation estimation (QPE) plays an essential role in disaster prevention and mitigation. In this paper, two deep learning-based QPE networks including a single-parameter network and a multi-parameter network are designed. Meanwhile, a self-defined loss function (SLF) is proposed during modeling. The dataset includes Shijiazhuang S-band dual polarimetric radar (CINRAD/SAD) data and rain gauge data within the radar’s 100-km detection range during the flood season of 2021 in North China. Considering that the specific propagation phase shift (K_DP) has a roughly linear relationship with the precipitation intensity, K_DP is set to 0.5° km⁻¹ as a threshold value to divide all the rain data (AR) into a heavy rain (HR) and light rain (LR) dataset. Subsequently, 12 deep learning-based QPE models are trained according to the input radar parameters, the precipitation datasets, and whether an SLF was adopted, respectively. The results suggest that the effects of QPE after distinguishing rainfall intensity are better than those without distinguishing, and the effects of using SLF are better than those that used MSE as a loss function. A Z-R relationship and a Z_H-K_DP-R synthesis method are compared with deep learning-based QPE. The mean relative errors (MRE) of AR models using SLF are improved by 61.90%, 51.21%, and 56.34% compared with the Z-R relational method, and by 38.63%, 42.55%, and 47.49% compared with the synthesis method. Finally, the models are further evaluated in three precipitation processes, which manifest that the deep learning-based models have significant advantages over the traditional empirical formula methods.

As the risks associated with air turbulence are intensified by climate change and the growth of the aviation industry, it has become imperative to monitor and mitigate these threats to ensure civil aviation safety. The eddy dissipation rate (EDR) has been established as the standard metric for quantifying turbulence in civil aviation. This study aims to explore a universally applicable symbolic classification approach based on genetic programming to detect turbulence anomalies using quick access recorder (QAR) data. The detection of atmospheric turbulence is approached as an anomaly detection problem. Comparative evaluations demonstrate that this approach performs on par with direct EDR calculation methods in identifying turbulence events. Moreover, comparisons with alternative machine learning techniques indicate that the proposed technique is the optimal methodology currently available. In summary, the use of symbolic classification via genetic programming enables accurate turbulence detection from QAR data, comparable to that with established EDR approaches and surpassing that achieved with machine learning algorithms. This finding highlights the potential of integrating symbolic classifiers into turbulence monitoring systems to enhance civil aviation safety amidst rising environmental and operational hazards.

Thunderstorm gusts are a common form of severe convective weather in the warm season in North China, and it is of great importance to correctly forecast them. At present, the forecasting of thunderstorm gusts is mainly based on traditional subjective methods, which fails to achieve high-resolution and high-frequency gridded forecasts based on multiple observation sources. In this paper, we propose a deep learning method called Thunderstorm Gusts TransU-net (TG-TransUnet) to forecast thunderstorm gusts in North China based on multi-source gridded product data from the Institute of Urban Meteorology (IUM) with a lead time of 1 to 6 h. To determine the specific range of thunderstorm gusts, we combine three meteorological variables: radar reflectivity factor, lightning location, and 1-h maximum instantaneous wind speed from automatic weather stations (AWSs), and obtain a reasonable ground truth of thunderstorm gusts. Then, we transform the forecasting problem into an image-to-image problem in deep learning under the TG-TransUnet architecture, which is based on convolutional neural networks and a transformer. The analysis and forecast data of the enriched multi-source gridded comprehensive forecasting system for the period 2021–23 are then used as training, validation, and testing datasets. Finally, the performance of TG-TransUnet is compared with other methods. The results show that TG-TransUnet has the best prediction results at 1–6 h. The IUM is currently using this model to support the forecasting of thunderstorm gusts in North China.

To quantify the relative contributions of Arctic sea ice and unforced atmospheric internal variability to the “warm Arctic, cold East Asia” (WACE) teleconnection, this study analyses three sets of large-ensemble simulations carried out by the Norwegian Earth System Model with a coupled atmosphere–land surface model, forced by seasonal sea ice conditions from preindustrial, present-day, and future periods. Each ensemble member within the same set uses the same forcing but with small perturbations to the atmospheric initial state. Hence, the difference between the present-day (or future) ensemble mean and the preindustrial ensemble mean provides the ice-loss-induced response, while the difference of the individual members within the present-day (or future) set is the effect of atmospheric internal variability. Results indicate that both present-day and future sea ice loss can force a negative phase of the Arctic Oscillation with a WACE pattern in winter. The magnitude of ice-induced Arctic warming is over four (ten) times larger than the ice-induced East Asian cooling in the present-day (future) experiment; the latter having a magnitude that is about 30% of the observed cooling. Sea ice loss contributes about 60% (80%) to the Arctic winter warming in the present-day (future) experiment. Atmospheric internal variability can also induce a WACE pattern with comparable magnitudes between the Arctic and East Asia. Ice-loss-induced East Asian cooling can easily be masked by atmospheric internal variability effects because random atmospheric internal variability may induce a larger magnitude warming. The observed WACE pattern occurs as a result of both Arctic sea ice loss and atmospheric internal variability, with the former dominating Arctic warming and the latter dominating East Asian cooling.

Fine particulate matter (PM_2.5) and ozone (O₃) double high pollution (DHP) events have occurred frequently over China in recent years, but their causes are not completely clear. In this study, the spatiotemporal distribution of DHP events in China during 2013–20 is analyzed. The synoptic types affecting DHP events are identified with the Lamb–Jenkinson circulation classification method. The meteorological and chemical causes of DHP events controlled by the main synoptic types are further investigated. Results show that DHP events (1655 in total for China during 2013–20) mainly occur over the North China Plain, Yangtze River Delta, Pearl River Delta, Sichuan Basin, and Central China. The occurrence frequency increases by 5.1% during 2013–15, and then decreases by 56.1% during 2015–20. The main circulation types of DHP events are “cyclone” and “anticyclone”, accounting for over 40% of all DHP events over five main polluted regions in China, followed by southerly or easterly flat airflow types, like “southeast”, “southwest”, and “east”. Compared with non-DHP events, DHP events are characterized by static or weak wind, high temperature (20.9°C versus 23.1°C) and low humidity (70.0% versus 64.9%). The diurnal cycles of meteorological conditions cause PM_2.5 (0300–1200 LST, Local Standard Time= UTC+ 8 hours) and O₃ (1500–2100 LST) to exceed the national standards at different periods of the DHP day. Three pollutant conversion indices further indicate the rapid secondary conversions during DHP events, and thus the concentrations of NO₂, SO₂ and volatile organic compounds decrease by 13.1%, 4.7% and 4.4%, respectively. The results of this study can be informative for future decisions on the management of DHP events.

Spring consecutive rainfall events (CREs) are key triggers of geological hazards in the Three Gorges Reservoir area (TGR), China. However, previous projections of CREs based on the direct outputs of global climate models (GCMs) are subject to considerable uncertainties, largely caused by their coarse resolution. This study applies a triple-nested WRF (Weather Research and Forecasting) model dynamical downscaling, driven by a GCM, MIROC6 (Model for Interdisciplinary Research on Climate, version 6), to improve the historical simulation and reduce the uncertainties in the future projection of CREs in the TGR. Results indicate that WRF has better performances in reproducing the observed rainfall in terms of the daily probability distribution, monthly evolution and duration of rainfall events, demonstrating the ability of WRF in simulating CREs. Thus, the triple-nested WRF is applied to project the future changes of CREs under the middle-of-the-road and fossil-fueled development scenarios. It is indicated that light and moderate rainfall and the duration of continuous rainfall spells will decrease in the TGR, leading to a decrease in the frequency of CREs. Meanwhile, the duration, rainfall amount, and intensity of CREs is projected to regional increase in the central-west TGR. These results are inconsistent with the raw projection of MIROC6. Observational diagnosis implies that CREs are mainly contributed by the vertical moisture advection. Such a synoptic contribution is captured well by WRF, which is not the case in MIROC6, indicating larger uncertainties in the CREs projected by MIROC6.