Correcting the forecast bias of numerical weather prediction models is important for severe weather warnings. The refined grid forecast requires direct correction on gridded forecast products, as opposed to correcting forecast data only at individual weather stations. In this study, a deep learning method called CU-net is proposed to correct the gridded forecasts of four weather variables from the European Centre for Medium-Range Weather Forecast Integrated Forecasting System global model (ECMWF-IFS): 2-m temperature, 2-m relative humidity, 10-m wind speed, and 10-m wind direction, with a forecast lead time of 24 h to 240 h in North China. First, the forecast correction problem is transformed into an image-to-image translation problem in deep learning under the CU-net architecture, which is based on convolutional neural networks. Second, the ECMWF-IFS forecasts and ECMWF reanalysis data (ERA5) from 2005 to 2018 are used as training, validation, and testing datasets. The predictors and labels (ground truth) of the model are created using the ECMWF-IFS and ERA5, respectively. Finally, the correction performance of CU-net is compared with a conventional method, anomaly numerical correction with observations (ANO). Results show that forecasts from CU-net have lower root mean square error, bias, mean absolute error, and higher correlation coefficient than those from ANO for all forecast lead times from 24 h to 240 h. CU-net improves upon the ECMWF-IFS forecast for all four weather variables in terms of the above evaluation metrics, whereas ANO improves upon ECMWF-IFS performance only for 2-m temperature and relative humidity. For the correction of the 10-m wind direction forecast, which is often difficult to achieve, CU-net also improves the correction performance.

The diurnal evolution and distribution of warm-season convective storms in different prevailing wind regimes at the 925 hPa and 500 hPa levels over contiguous North China have been studied using a climatology based on radar reflectivity mosaics, soundings, and reanalysis data from 15 May to 15 September, in 2008–2011. In the wind regimes at the 925 hPa level, the frequency of occurrence of convection is highest for south-southwesterly flows, followed by that for south-southeasterly flows. The low-level southerly warm and moist flow over the foothills and plains favors initiation and intensification of convection resulting from topographic forcing. Among the wind regimes at the 500 hPa level, the frequency of occurrence of convection shows a pronounced maximum for west-southwesterly flows, and most of these storms initiate and develop over the foothills and plains. Convection in the wind regime shows the morphology and slowly east-southeastward moving characteristics of well-organized, longer-lived systems, with environmental conditions conducive to vertical wind shear and instability. At the 500 hPa level, the high frequency of occurrence of convection for west-southwesterly flows is followed by that for west-northwesterly flows. Under west-northwesterly flows, convective storms initiated over mountains located to the northwest in the early afternoon move rapidly to the plains in the southeast of the study region. These storms show the primary formation characteristics of shorter-lived convection and intensify noticeably as they move from the mountains to the foothills and plains.

Beijing, China, is located in a region of complex terrain with high mountain ridges to the northwest and the Bohai Sea to the southeast. The origin of convective storms occurring on the plains can often be traced to the upstream mountains. Under weakly forced conditions, these convective storms most frequently evolve into squall lines (SL) and convective clusters (CC) when reaching the plains. In this study, we analyze 18 SL and 15 CC storm systems and assess their environmental and mesoscale differences between the two phenomena. By analyzing the frequency of convective occurrence for the two types of storms based on composite radar reflectivity, it is found that the high frequencies are located in the south of Beijing for SL and near the center of Beijing for CC. Using storm‐scale reanalysis data produced by a rapid update four‐dimensional variational analysis system that assimilates Doppler radar observations, distinct features of the SL and CC storms are revealed in terms of their convective environments and mesoscale structures, such as cold pool, horizontal wind convergence, and humidity distribution. It is found that low convective inhibition and high low‐level wind speed on the plains are common to both SL and CC, whereas higher vertical shear over the plains and stronger wind speed on both mountains and plains distinguishes SL from CC. We further show that the stronger wind and vertical shear in SL generate stronger and more organized downdrafts, producing a deeper cold pool, strong outflow and convergence, which explains the formation of the high‐frequency center in the south of Beijing. In contrast, the cold pool produced in CC is shallower and weaker, resulting in weaker outflow and convergence and convective activities that are located only in central Beijing.

Weather radar plays an important role in accurate weather monitoring and modern weather forecasting, as it can provide timely and refined weather forecasts for the public and for decision makers. Deep learning has been applied in radar nowcasting tasks and has exhibited a better performance than traditional radar echo extrapolation methods. However, current deep learning-based radar nowcasting models are found to suffer from a spatial “blurry” effect that can be attributed to a deficiency in spatial variability representation. This study proposes a Spatial Variability Representation Enhancement (SVRE) loss function and an effective nowcasting model, named the Attentional Generative Adversarial Network (AGAN), to alleviate this blurry effect by enhancing the spatial variability representation of radar nowcasting. An ablation experiment and a comparison experiment were implemented to assess the effect of the generative adversarial (GA) training strategy and the SVRE loss, as well as to compare the performance of the AGAN and SVRE loss function with the current advanced radar nowcasting models. The performances of the models were validated on the whole test set and inspected in two storm cases. The results showed that both the GA strategy and SVRE loss function could alleviate the blurry effect by enhancing the spatial variability representation, which helps the AGAN to achieve better nowcasting performance than the other competitor models. Our study provides a feasible solution for high-precision radar nowcasting applications.

A case study is presented of convection initiation (CI) resulting from the merger of a cold front with a dryline in southwestern Beijing, China, on the afternoon of 11 June 2011. This process is analyzed with S-band Doppler radar data, surface automatic weather station data, and mesoscale numerical simulation results. The formation of this dryline is analogous to that on the Great Plains of the United States, and it is conducive to CI with mesoscale updrafts generated from the baroclinic frontogenesis, and with favorable instability immediately on the moist side. Prior to the front–dryline merger, as the cold front approached the observed boundary layer convergence line, or the simulated meso- γ-scale secondary dryline, CI occurred ahead of the cold front with little contribution from frontogenetic baroclinity of the dryline. The cold front then merged with the dryline, and the baroclinity of the dryline was enhanced by the associated convergence, to a degree comparable to that caused by frontogenesis of the dryline itself, thus leading to more CI. During the front–dryline merger, meso- γ-scale discrete cold pools associated with the cold front led to a diverse distribution of CI.