In weather forecasting, numerical weather prediction (NWP) that is based on physical models requires proper post-processing before it can be applied to actual operations. Therefore, research on intelligent post-processing algorithms has always been an important topic in this field. This paper proposes a model output deep learning (MODL) method for post-processing, which can improve the forecast effect of numerical weather prediction. MODL is an end-to-end post-processing method based on deep convolutional neural network, which directly learns the mapping relationship between the forecast fields output by numerical model and the observation temperature field in order to obtain more accurate temperature forecasts. MODL modifies the existing deep convolution model according to the post-processing problem’s characteristics, thereby improving the performance of the weather forecast. This paper uses The International Grand Global Ensemble (TIGGE) dataset from European Centre for Medium-Range Weather Forecasts (ECMWF) and the observed air temperature of 2 m obtained from Tianjin meteorological station in order to test the post-processing performance of MODL. The MODL method applied to temperature in post-processing is compared with the ECMWF forecast, Model Output Statistics (MOS) methods, and Model Output Machine Learning (MOML) methods. The Root Mean Square Error (RMSE) of the temperature field predicted by MODL and the observed temperature field is smaller than the other models and the accuracy of the temperature difference of 2 °C (Acc) is higher, especially where the prediction time is in the first three days. The lightweight nature of MODL also makes it suitable for most operations.

An isolated convection was unexpectedly initiated in the evening of 1 August 2019 around the Tianjin urban region (TUR), which happened at some distance from the shear line at lower level and the preexisting convection to the South, analyzed by using ERA5 reanalysis data and observations from surface weather stations, and a S-band radar. The results show that, 42 min before the initiation of the convection, the atmospheric thermodynamic conditions around TUR were favorable for the initiation of the isolated convection, although the southerly and vertical shear of the horizontal wind at the lower level was weak. A sea-breeze front approached the TUR and continued to move West, leading to the triggering of the isolated convection in the context of the urban humid heat island (UHHI) effect. Subsequently, the gust front, which was formed between the cold pool away from the TUR and the warm and humid air of the UHHI, moved northward, approached the convection, and collided with sea breeze front, resulting in five reflectivity centers of isolated convection being merged and the convection’s development. Finally, the isolated convection split into two convections that moved away from the TUR and disappeared at 20:36 Beijing Time. The isolated convection was initiated and developed by the interaction of the sea breeze front and gust front in the context of the UHHI effect. The sea breeze front triggered the isolated convection around TUR in the context of the UHHI effect, and the gust front produced by the early convective storms to the south played a vital role in the development of the isolated convection.

A short-time rainstorm exceeding the extreme historical rainfall occurred in the Jinnan District of Tianjin, China, on 3 July 2022. Due to the concentrated time period of precipitation, it caused serious water accumulation in the Jinnan District. The purpose of this paper is to study the weather mechanism of this extreme rainstorm in the Jinan District of Tianjin. By analyzing the fine observation facts, we can obtain the mesoscale weather characteristics and environmental conditions of the process. The results provide a reference for similar weather forecasting and warning in the future. Based on the 1 min interval precipitation observation data, the ERA5 reanalysis data, the CINRAD-SA radar reflectivity data of Tanggu, the cloud-top brightness temperature data of the Fengyun-4A satellite, and the Variational Doppler Radar Analysis System data, we comprehensively analyzed a record-breaking extreme rainfall process in Tianjin on 3 July 2022. The results show that the extreme rainfall process presents prominent mesoscale weather characteristics, with high precipitation intensity in a short-term period. This process is influenced by multi-scale weather systems, including the 500 hPa upper-level trough and the long-distance water vapor transport by Typhoon Chaba. The rainstorm event is caused by the combined actions of cold pool outflow produced by the upstream precipitation, the easterly disturbance in the boundary layer, the mesoscale temperature front, and the ground convergence line. Specifically, the ground convergence line is formed by the northerly wind of the cold pool outflow and the warm and moist southerly airflow from the ocean, and the temperature front is caused by the horizontal thermal difference of the underlying surface. Both the ground convergence line and temperature front contribute considerably to the triggering of mesoscale convection. The mesoscale secondary circulation is formed in the meridional direction by the meso-γ-scale convergence and its interaction with strong velocity in front of the trough, contributing to the development and maintenance of vertical motion in the Jinnan region of Tianjin and thereby leading to the occurrence and development of this extreme heavy rainfall process.

Using ground-based remote sensing equipment (wind profile radar and microwave radiometer) data and ground automatic station data, a ground-based remote sensing sounding profile system (FAS) is constructed, which aims to make use of its advantages of high resolution, high accuracy, and low cost to make up for the lack of space–time density in existing conventional sounding layer information. The retrieval results of remote sensing sounding profiles in Beijing from May 2021 to September 2022 were tested and evaluated. The results show that the correlation coefficient between FAS and conventional sounding specific humidity is 0.89, the root–mean–square deviation is 1.53 g/kg, and the evolution trends of different data sources of convective available potential energy (CAPE) and vertical wind shear are synchronous. A case study was conducted to evaluate the effectiveness of 40 severe convective processes in the Beijing Plain area. The results show that, due to the minute-level time resolution of FAS, the retrieved convective parameters could track the evolution trend of the atmospheric state with high timeliness, dynamically describe the configuration of thermodynamic parameters, and indicate the time-varying local convective potential and instability level. Therefore, it has certain short-term forecasting significance for the occurrence time, intensity, and convection type.

Based on ECMWF ensemble forecasts, this study examines two warm-sector heavy rainfall (WSHR) events in North China (3 July and 18 August 2022), which were characterized by distinct synoptic patterns, pretrough type, and edge of the western Pacific subtropical high (WPSH) type, yet both featured low-level jets (LLJs). Through ensemble sensitivity analysis and member-based comparisons, we identify key influencing factors and explore their implications for operational forecasting. The main findings are as follows: 1) The critical sensitive factors vary with synoptic pattern. The “7.3 event” exhibits high sensitivity to geopotential height at 500–850 hPa and mid- to low-level temperature fields, with weak cold air intrusion acting as a trigger for convection. In contrast, the “8.18 event” shows strong sensitivity to LLJ intensity and deep-layer thermal structure, where multidirectional moisture convergence serves as the primary convective trigger. 2) In both cases, member-based grouping reveals that consistently stronger jet intensity distinguishes relatively better-performing members from worse-performing members. Members with the LLJ axis located to the left of the ensemble mean yield better precipitation predictions. 3) The rainfall mechanism of the two WSHR events follows a “displaced triggering” pattern, in which convection typically initiates downstream of regions with high moisture and convective available potential energy (CAPE) sensitivity, specifically where LLJ-driven moisture transport overlaps with boundary layer convergence. Key operational indicators for forecasting include identifying high-sensitive regions 36–42 h in advance and focusing on the overlap of the LLJ and boundary layer convergence lines. These findings offer a scientific foundation for enhancing the prediction of WSHR over North China. Significance Statement: Forecasting warm-sector heavy rainfall in North China is notoriously difficult, especially when it occurs under different weather patterns. This study analyzes two such events, both driven by low-level jets but under different large-scale settings—using ensemble forecasts to identify what matters most for prediction. We found that the key factors differ depending on the weather pattern: One event was more sensitive to cold air intrusion and southerly moisture, while the other was highly sensitive to the jets’ position and low-level wind convergence. In the two cases examined, better-performing forecasts were consistently linked to members with stronger jets positioned to the left of the ensemble mean. A key insight is that heavy rain does not fall where areas with high moisture and convective available potential energy (CAPE) sensitivity are located; instead, it develops downstream of them. This “displaced triggering” mechanism means forecasters should focus on the low-level jet, weak cold air, and surface convergence, not just ensemble sensitivity maxima. These findings provide practical guidance for improving the prediction of these high-impact, hard-to-forecast events in North China.