Abstract Accurate short‐term forecasting of thunderstorm gusts remains a major challenge in the Beijing–Tianjin–Hebei region due to limited temporal resolution, the lack of vertical velocity and thermodynamic data, and the difficulty of coupling multi‐scale features. To address these issues, we propose a deep learning model suite called Thunderstorm Gusts Swin Transformer U‐Net (G‐Net), designed to integrate multi‐source meteorological data and capture multi‐scale gust structures. G‐Net includes: (a) a data fusion strategy incorporating 16 key surface and upper‐air variables with 3D vertical profiles; (b) a hybrid Swin Transformer‐U‐Net architecture to extract both global and local features; and (c) a progressive spatiotemporal decoupling architecture, including three functionally distinct model variants: G‐Net1H (1 hr forecast) and G‐Net2H (2 hr forecast) with a Feature Bridging Layer (FBL) designed for minute‐level average wind predictions, and G‐NetG for gust mapping. Results show that G‐Net1H surpasses the baseline models ConvLSTM, TG‐TransUNet, and PredRNN‐V2 in CSI and FAR across all wind speed thresholds within the 0–1 hr forecast. The introduction of spatiotemporal supplementary information and the efficient integration of the FBL enable G‐Net2H to achieve improved 2‐hr forecasts, reducing mean absolute and root mean square errors and enhancing overall predictive performance. G‐NetG further outperforms traditional gust factor methods and deep learning baselines. Evaluation during the 2024 thunderstorm season demonstrates that G‐Net provides the most accurate 0–2 hr forecasts, improving both the timeliness and spatial precision of weather warnings.

The SMART2022 project (Sciences of Meteorology and Artificial Intelligence in Research and Technology for the Beijing 2022 Olympic and Paralympic Winter Games) was initiated to enhance weather forecasting and services for the Beijing 2022 Olympic and Paralympic Winter Games. The focus of the project was to carry out intensive meteorological field experiments and develop advanced high-precision weather forecasting methods to support the games. The field experiments fully considered the advantages and limitations of various types of instruments and provided detailed firsthand information that enabled forecasters to understand the small-scale weather phenomena in the mountain competition venues. Novel methods were developed, enhanced, and used during the games, such as integration-based, 10-min-updating forecasts with 100-m horizontal grid spacing for 0–24-h lead times; large-eddy-simulation-based numerical model forecasts with 67-m horizontal grid spacing for 24–240-h lead times; and artificial intelligence–based, site-specific, postprocessed seamless forecasts with 0–240-h lead times. These innovations increased the precision of the weather forecasting and allowed effective correction of kilometer-scale numerical model forecast errors, with a specific emphasis on ultrahigh-resolution predictions of wind gusts and temperature in the complex mountainous competition areas affected mainly by the continental winter monsoon. Utilization of the SMART2022 achievements had a key role in supporting weather forecasts and services for the official training and competition events, scheduling, competition safety window selection, and other aspects related to both the 2022 Winter Olympics and Paralympics. Ultimately, these achievements contributed to the successful completion of all the snow-based events and the safety of athletes during the 2022 Games.

We investigate the evolution of a supercell storm within a quasi‐linear convective system (QLCS) that occurred in the Beijing area on 12 June 2022. Using high spatiotemporal resolution observations from a C‐band phased array radar (PAR), assimilated into a four‐dimensional variational data assimilation system, we primarily analyze dynamical processes contributing to the development of the supercell storm and its associated mesocyclone. Our study shows that just before the convective cell is triggered, a significant convergence zone develops to the west of the terrain, forming several meso‐γ vortices near the surface. During the merger of the convective cell and the QLCS from upper to lower levels, a strong downdraft generated by the QLCS enhances low‐level horizontal convergence, further producing a stretching effect on the vortices within the storm and significantly increasing vertical vorticity. With the formation of the mesocyclone in the mature stage of supercell storm, the height of the rotational center rises to 4.5 km, and the maximum rotational velocity reaches 20 m/s. Our results indicate that the surface convergence lines and the meso‐γ vortices along them strengthen low‐level convergence and generate strong updrafts, triggering the initial storm. These intense updrafts transform horizontal vorticity into vertical vorticity and transport it upward. Additionally, the process of convective merging leads to strengthen low‐level horizontal convergence, which forcibly stretches the mesovortex, enhancing vertical vorticity and allowing the convective storm to develop in a strong, organized manner and form the supercell storm.

Wind is one of the most important meteorological conditions in previous Winter Olympics,and it is the primary factor that affects the mountain events for Beijing Winter Olympics.Understanding the fine distribu-tion law of wind can provide important theoretical basis for track construction,wind forecast and prevention mea-sures.Using hourly observation data from surface automatic weather stations at different altitudes in Yanqing mountain area of Beijing Winter Olympics from December 2017 to March 2022,this study investigated the char-acteristics of local wind field during winter and early spring(Mar,Paralympics period)under complex terrain,focusing on comparing the frequency of wind speeds and directions,as well as the diurnal and seasonal varia-tions.Firstly,all stations were grouped into four categories using the K-Means clustering algorithm,and Groups 1 to 4 represent the low-elevation Yangqing suburb area,the northeastern foothills transition area,the southwest-ern transition area and the high-elevation mountain top area,respectively.Subsequently,fine-grained characteris-tic analysis was conducted on each group separately.Results show that:(1)The frequency of strong winds is closely related to the altitude,with higher altitudes generally having a higher frequency of strong winds.In Groups 1～2(altitude below 1000 m),the frequency of light winds(≤3.3 m·s-1)exceeds 80%,while the propor-tion of strong winds(≥10.7 m·s-1)is 0%.In Group 3(above 1000 m),the frequency of light winds decreases to below 75%,and strong winds occasionally occur for less than 1%.In Group 4(above 1800 m),there is a signifi-cant shift in the wind speed frequency distribution,with the frequency of strong winds increasing to above 10%,which is much higher during winter compared to early spring.(2)There are significant local variations in the dis-tribution characteristics of wind directions.Group 4 is primarily dominated by large-scale winter monsoonal cir-culation,resulting in a prevailing northwesterly wind,with rare concurrence of other wind directions.Groups 1～3 are influenced by a combination of large-scale circulation,valley wind circulation and underlying surface con-ditions,leading to different frequencies for each wind direction.(3)The diurnal variation exhibits contrasting characteristics between high and low elevations.Groups 1～3 show lower wind speeds at night and higher wind speeds during the day,while Group 4 shows a reserved pattern and an obvious small wind"window period"in midday.Groups 1～3 exhibit distinct daily transitions in wind direction,occurring after sunrise and sunset,where-as Group 4 does not show any diurnal change.(4)From a seasonal perspective,there are significant local differ-ences between early spring and winter.Compared to winter,Group 2 exhibits a daytime wind speed increase in early spring,and Group 3 exhibits a nighttime decrease,while Group 4 exhibits a significant decrease in wind speeds throughout the day.Wind directions in early spring are relatively more variable,with an evident increase in northeasterly winds in Group 1,a delay of about 3 hours in the transition of valley wind circulation in Group 2,and an increase in southwesterly winds in Groups 3～4.This study contributes to a deeper comprehension of the fine-scale spatiotemporal patterns of near-surface local wind fields within complex terrains,and can offer cru-cial background clues for Winter Olympics and small-scale mountainous meteorological monitoring and forecast-ing.

The SMART2022 project (Sciences of Meteorology and Artificial Intelligence in Re-search and Technology for the Beijing 2022 Olympic and Paralympic Winter Games) was initiated to enhance weather forecasting and services for the Beijing 2022 Olympic and Paralympic Winter Games. The focus of the project was to carry out intensive meteorological field experiments and develop advanced high-precision weather forecasting methods to support the games. The field experiments fully considered the advantages and limitations of various types of instruments and provided detailed firsthand information that enabled forecasters to understand the small-scale weather phenomena in the mountain competition venues. Novel methods were developed, enhanced, and used during the games, such as integration-based, 10-min-updating forecasts with 100-m horizontal grid spacing for 0-24-h lead times; large-eddy-simulation-based nu-merical model forecasts with 67-m horizontal grid spacing for 24-240-h lead times; and artificial intelligence-based, site-specific, postprocessed seamless forecasts with 0-240-h lead times. These innovations increased the precision of the weather forecasting and allowed effective correction of kilometer-scale numerical model forecast errors, with a specific emphasis on ultrahigh-resolution predictions of wind gusts and temperature in the complex mountainous competition areas affected mainly by the continental winter monsoon. Utilization of the SMART2022 achievements had a key role in supporting weather forecasts and services for the official training and competition events, scheduling, competition safety window selection, and other aspects related to both the 2022 Winter Olympics and Paralympics. Ultimately, these achievements contributed to the successful completion of all the snow-based events and the safety of athletes during the 2022 Games. SIGNIFICANCE STATEMENT: The successful hosting of the modern Winter Olympics and Paralympics is closely linked to the weather conditions. This paper introduces the SMART2022 project, which aimed to provide essential weather forecasting support for the Beijing 2022 Olympic and Paralympic Winter Games. The project involved meteorological field experiments in two mountainous competition zones over four consecutive winters, using comprehensive arrays of advanced meteorological observation instruments. It also included the development of novel, high-precision weather forecasting systems and methods by utilizing enhanced meteorological observations, advanced numerical weather prediction models, and artificial intelligence techniques. The outcomes of the project in supporting the 2022 Games are evaluated. The project can serve as a demonstration of high-spatiotemporal-resolution field experiments and forecasts for winter mountain weather in the midlatitudes.

During 29th July–1st August in 2023, a persistent heavy rainfall event (“23·7” event) hit North China causing severe floods, enormous infrastructure damage, and large economy loss. Observational analysis shows that the extremely large accumulation of precipitation and long duration of this event are closely related to a slowly moving landfall typhoon “Dusuari” over North China due to the blocking effect of an anomalous high over the mid‐high latitude Asia. The anomalous southeasterly flow induced by the typhoon “Dusuari” and another typhoon “Khanun” over the East China Sea jointly built a highly efficient channel of water vapor supply from southern oceans toward North China. A water vapor budget analysis indicates that precipitation of this event is mainly caused by the dynamic process involving strong ascending motion. Accompanying strong water vapor transportation and convergence over North China, large amount of latent heat is released in the middle and the lower troposphere. The physical mechanisms of heavy rainfall‐induced diabatic heating in maintaining the precipitation over North China is further investigated using statistical analysis and numerical experiments. On one hand, the latent heating released by heavy rainfall induces significant uplifting flows which causes more precipitation. On the other hand, the heavy rainfall‐induced diabatic heating contributes to the enhancement of the westward extension of high‐pressure dam over mid‐high latitude through a regional meridional circulation. This strengthened high‐pressure dam sustained the cyclonic circulation of “Dusuari” over North China, leading to continuous heavy rainfall there.

Study region: Xiangjiang and Hanjiang River basins in the Yangtze River basin, a humid subtropical inland region of central-southern China. Study focus: The general precipitation predictive skills of the mainstay numerical models are still rather limited beyond 10 days, further deteriorating the performance of sub-monthly streamflow prediction. This study proposes a sub-monthly streamflow prediction framework for organically combining stochastic weather generator (SWG) and monthly precipitation prediction to generate ensemble sub-monthly precipitation for streamflow prediction. The SWG-based schemes are then compared with the common numerical hydrometeorology ensemble streamflow prediction over two river basins in China. New hydrological insights for the region: Results show that the numerical streamflow predictions exhibit a less accurate deterministic performance than the SWG-based framework with the climatology scheme over sub-monthly horizon, with leadtime-averaged mean absolute relative error dropping from 19.9 % to 8.3 %, and 21.8-11.1 % for Xiangjiang and Hanjiang river basins, respectively. In addition, the more restrictive parametric adjustment procedure with adjusted precipitation amounts can bring added values and further improve the accuracy of SWG-based daily streamflow prediction for sub-monthly leadtimes. In terms of the probabilistic prediction performance, the SWG-based methods yield approximately equivalent results with the numerical hydrometeorology streamflow prediction, with leadtime-averaged Continuous Ranked Probability Skill Score of the scheme with modified parameters of precipitation amounts being 0.51 and 0.56 for Xiangjiang and Hanjiang River basins, respectively. Furthermore, the SWG-based schemes are more suitable for predicting high-flow events during the flood season, with a significantly smaller Brier Score. However, there are little improvements in probabilistic prediction performance when transition probabilities of precipitation occurrence are progressively modified.

Thunderstorm gusts are a common and hazardous type of severe convective weather, characterized by a small spatial scale, short duration, and significant destructive power. They often lead to severe disasters, highlighting the critical importance of their accurate forecasting. Previous studies have explored the environmental factors and spatiotemporal distribution characteristics of thunderstorm gusts, highlighting the need for improved forecasting methods. In recent years, artificial intelligence techniques have shown promise in enhancing the accuracy of thunderstorm gust forecasting, with various machine learning algorithms and models having been developed. This paper proposes a multiscale feature fusion module called Thunderstorm Gusts Block (TG-Block) and a deep learning model named Thunderstorm Gusts net (TG-net) based on the Attention U-net and TG-TransUnet models, and employs interpretable methods such as Integrated Gradient, Deep Learning Importance Features, and Shapley Additive exPlanations to validate the model’s practical relevance and reliability. The analysis of feature importance underscores the model’s ability to capture key thermodynamic and multiscale weather characteristic information for thunderstorm gust nowcasting. It is, however, worth emphasizing that these conclusions are only based on a limited number of thunderstorm gust examples, and the evaluation results may be affected by specific weather types and sample sizes. Nonetheless, TG-net has been put into real-time operation at the Institute of Urban Meteorology, and we will continue to rigorously validate its performance and make any necessary optimizations and enhancements based on feedback to ensure the robustness and stability of the model.

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.

While previous work on the climatology of Northern China has focused on mean wind speed, wind gusts have received comparatively less attention but are equally important to various users. In this paper, an observed hourly maximum gust wind speeds (HMGS) dataset across North China has been created by using time series from 174 meteoroning from 2015 to 2022. The objective of this study is first to improve our understanding of the spatiotemporal gusts climatology in North China by analyzing the observed gust data. Second, we aim to supplement the observational data by using gust analysis and forecast data with a high spatial-temporal resolution from model simulations. The spatial characteristics of the seasonal cycle of the simulated analysis of mean HMGS and the performance in predicting gusts based on the geographical locations and elevations of the validation stations were investigated by comparing it with the observations. Results indicate the following: 1) Wind direction and intensity are affected by the terrain and climate conditions of different weather stations. Stations situated along the Bohai Bay coastal region and at higher-elevation areas of North China exhibit a higher mean HMGS than those located in the coastal and inland plains. 2) The probability density function curves for wind speed and wind direction exhibit notable variations across different elevation intervals. The contribution of moderate and strong gust wind speeds increases gradually with increasing altitude, while the gust directions in mountainous areas exhibit relatively consistent patterns due to the increased exposure to synoptic-scale forcing at higher elevations. 3) The nowcasting prediction system analysis of mean HMGS provides a higher horizontal resolution that is capable of capturing the contrasts between land and sea, as well as the influence of high HMGS associated with large-scale circulations in

This study investigates the impacts of grid spacing and station network on surface analyses and forecasts in-cluding temperature, humidity, and winds in Beijing Winter Olympic complex terrain. The high-resolution analyses are generated by a rapid-refresh integrated system that includes a topographic downscaling procedure. Results show that sur-face analyses are more accurate with a higher targeted grid spacing. In particular, the average analysis errors of surface temperature, humidity, and winds are all significantly reduced when the grid size is increased. This improvement is mainly attributed to a more realistic simulation of the topographic effects in the integrated system because the topographic down -scaling at higher grid spacing can add more details in a complex mountain region. From 1 km to 100 m, 1-12-h forecasts of temperature and humidity are also largely improved, while the wind only shows a slight improvement for 1-6-h forecasts. The influence of station network on the surface analyses is further examined. Results show that the spatial distributions of temperature and humidity at a 100-m space scale are more realistic and accurate when adding an intensive automatic weather station network, as more observational information can be absorbed. The adding of a station network can also re-duce forecast errors, which can last for about 6 h. However, although surface winds display better analysis skill when more stations are added, the wind at the mountaintop region sometimes encounters a marginally worse effect for both analysis and forecast. The results are helpful to improve the analysis and forecast products in complex terrain and have some impli-cations for downscaling from a coarse grid size to a finer grid.

The rapid development of economy and culture in Beijing-Tianjin-Hebei region has a higher requirement for instantaneous strong wind forecasts. Correctly estimating and predicting instantaneous strong winds on the ground level in winter, especially accurate high-resolution grid-point forecasting of gusts under complex terrain condition, is of great significance for improving the service for major Winter Olympics events, the safe operation in the capital and surrounding cities, and disaster prevention and mitigation capabilities. This study establishes a relationship between the gust coefficient and wind speed, wind direction and terrain height based on long-term series of observation data in Beijing-Tianjin-Hebei. Combined with objective statistical analysis method, gust observation data fusion technology and grid point deviation correction technology, an objective gust forecast method is developed, which not only retains the model physical parameters and local climate characteristics, but also utilizes the grid point deviation correction technology. The results of batch verification and case analysis during the Winter Olympics show that the average absolute errors in the Zhangjiakou competition area and the Yanqing competition area are below 2.3 m/s and 3.0 m/s, respectively. The forecast score of gust wind speed above level 8 in the Yanqing competition area is above 0.5. It solves the bottleneck problem of large gust prediction errors and meets the on-site service requirements of major Winter Olympic activities.

Ensemble forecasting systems have become an important tool for estimating the uncertainties in initial conditions and model formulations and they are receiving increased attention from various applications. The Regional Ensemble Prediction System (REPS), which has operated at the Beijing Meteorological Service (BMS) since 2017, allows for probabilistic forecasts. However, it still suffers from systematic deficiencies during the first couple of forecast hours. This paper presents an integrated probabilistic nowcasting ensemble prediction system (NEPS) that is constructed by applying a mixed dynamic-integrated method. It essentially combines the uncertainty information (i.e., ensemble variance) provided by the REPS with the nowcasting method provided by the rapid-refresh deterministic nowcasting prediction system (NPS) that has operated at the Beijing Meteorological Service (BMS) since 2019. The NEPS provides hourly updated analyses and probabilistic forecasts in the nowcasting and short range (0–6 h) with a spatial grid spacing of 500 m. It covers the three meteorological parameters: temperature, wind, and precipitation. The outcome of an evaluation experiment over the deterministic and probabilistic forecasts indicates that the NEPS outperforms the REPS and NPS in terms of surface weather variables. Analysis of two cases demonstrates the superior reliability of the NEPS and suggests that the NEPS gives more details about the spatial intensity and distribution of the meteorological parameters.

为了进一步提高RISE系统高分辨率网格化预报产品的准确率,同时考虑到深度学习近年来在地学领域的有效应用,采用2019—2021年高分辨率RISE系统数据,设计出卷积神经网络模型Rise-Unet,实现了未来4～12 h地面2 m温度、2 m相对湿度、10 m-U风速以及10 m-V风速预报结果的订正.订正试验结果表明,采用均方根误差和平均绝对误差作为评分标准,与RISE原始预报结果相比,基于Rise-Unet模型可以有效提高温湿风预报结果的准确率.该基于深度学习的Rise-Unet偏差订正技术可应用于RISE系统的后处理模块,对提升RISE系统百米级分辨率或其他高分辨率模式系统格点预报水平具有重要的科学意义和应用价值.

As the mainstream technology of modern weather forecast, numerical weather prediction (NWP) has been developing in the direction of refinement in recent years, yet the prediction error is still unavoidable. Therefore, it is of great significance to improve the accuracy of numerical weather forecast by revising the results. A traditional method of prediction correction, i.e., the Anomaly Numeral-correction with Observations (ANO), is used to correct the forecast based on statistics of historical data. Results indicate that this method has a good effect. As an emerging method, deep learning has been gradually applied to the field of meteorology in recent years, and has achieved significant results in precipitation prediction and cloud image recognition. Domestic scholars in China used CU-Net, a deep learning model to correct the deviations of the model grid point forecast data of 2 m temperature, 2 m relative humidity and 10 m wind respectively from the European Centre for Medium-Range Weather Forecast (ECMWF), which significantly improved the forecast compared with the ANO method. Based on the above tests, this paper uses dense convolutional structure network model to improve the CU-Net model and forms a new deviation correction model for NWP, which is named as Dense-CUnet, and further develops a deviation correction model named Fuse-CUnet to integrates multiple meteorological elements from NWP and topographic features. Deviation correction tests and comparative analysis of these different models have been carried out. Root mean square error (RMSE) and mean absolute error (MAE) are used as the scoring metrics. By comparing with the original prediction results of ECMWF and the results revised by the ANO and CU-Net methods, it is found that the dense-convolution structure network model Dense-CUnet can be used to effectively modify the positive effect. Moreover, the Fuse-CUnet model that integrates multiple elements can greatly improve the revision effect.

数值天气预报作为现代天气预报的主流技术方法,近年来不断朝着精细化方向发展,但预报误差至今仍无法避免.文中在CU-Net模型中引入稠密卷积模块形成数值预报要素偏差订正模型Dense-CUnet,在此基础上进一步融合多种气象要素和地形特征构建了Fuse-CUnet模型,开展不同模型的偏差订正试验和对比分析.以均方根误差(RMSE)和平均绝对误差(MAE)作为评分标准,通过与ECMWF原始预报结果、ANO方法订正结果以及CU-Net方法订正结果进行对比,证明Dense-CUnet模型可有效改进数值预报订正效果,融合多个要素的Fuse-CUnet模型能使订正效果有更大提升.

Precipitation nowcasting on fine scale is of great significance to improve the ability of early warning of flood and waterlogging disasters in modern cities. As a new method, deep learning has more advantages in mining the internal characteristics and physical laws of data. In recent years, the application of deep learning in the field of meteorological radar image has achieved preliminary results. In order to improve the effectiveness of nowcasting on fine scale, a deep convolutional neural network-RainNet is used to propose two ways of rolling approach for precipitation nowcasting. Experiments and comparative analysis are carried out in Beijing-Tianjin-Hebei region on 1 km resolution. Compared with the traditional extrapolation based on Tracking Radar Echoes by Correlation (TREC), the results show that the mean absolute error and correlation coefficient of 1 h nowcasting can be improved. The prediction in 10—50 min in thresholds of 1.04 mm/(10 min) and below is better than that of traditional prediction. Temporal and spatial evolution of precipitation extinction process is better described by deep learning compared with that by traditional extrapolation. The rolling approach with two RainNet models combined outperforms one single model in precipitation nowcasting.

精细尺度降水的临近预报对于提升现代城市内涝和山洪地质灾害预警能力具有重要意义.深度学习作为一种新兴方法,在挖掘数据内部特征及物理规律方面更具优势,近年来在天气雷达图像领域的应用已初见成效.为进一步提升精细尺度降水的临近预报能力,基于深度学习网络模型RainNet,研究建立了两种滚动预报方式,开展了京津冀地区1 km分辨率精细尺度降水滚动式临近预报试验和对比分析.试验结果表明:与传统基于交叉相关的外推预报相比,深度学习网络模型RainNet总体可以明显改进降水1 h临近预报的绝对误差和相关系数;两个RainNet相结合的滚动预报方式对1.04 mm/(10 min)及以下阈值降水,在10-50 min预报性能一致优于传统的交叉相关外推预报.深度学习模型对降水消亡过程的时、空演变趋势刻画更好,尤其更适用于降水消亡过程的临近预报.采用两个RainNet模型相结合的滚动式预报方式优于单一模型滚动预报方式.

几乎所有的数值预报模式都存在系统偏差.虽然目前利用统计订正方法降低个别站点的风速偏差已经取得了一些成功,但基于站点的订正具有空间局限性,仍迫切需要基于格点开展复杂地形下高精度风场的融合预报偏差订正.本研究提出了一种复杂地形下北京冬奥赛区不同海拔高度高精度风场的融合预报订正技术.首先利用冬奥山地赛区及周边133个自动气象站风场实况观测资料与睿图-睿思系统高精度风场预报数据相结合,利用统计偏差订正方法,获取各站点1～12 h的平均系统偏差,然后再将地形降尺度后的中国气象局北京快速更新循环数值预报系统高分辨率风场利用格点偏差订正系数优化后作为背景场融合观测资料,更好地捕捉局地地形对山区风场的影响.结果表明,本方法极大程度降低了风速的系统性偏差,风速预报误差显著降低,12 h风速平均绝对误差和均方根误差降低率最高达40％以上.经过适当的修改,这种方法也可以应用于对其他变量的偏差订正上.

Weather conditions have an important impact on agricultural production, transportation, economic activities, so the improvement of forecast accuracy has been a constant concern of the society. After more than 100 years of continuous development, the accuracy of numerical weather model has been continuously improved, but there are still inevitable forecast errors. Therefore, it is an important issue worthy of study to improve the prediction accuracy by studying various error correction methods and post-processing the results of numerical weather prediction.Machine learning method is applied to revise four meteorological elements forecasted by RMAPS-RISE(rapid-update multi-scale analysis and prediction system-rapid integration and seamless ensemble) system developed by Beijing Institute of Urban Meteorology. First, the data are preprocessed by interpolating the system forecast data and extracting the data of each element site from the grid data. The observations of automatic weather stations and forecast data are processed to establish unified datasets for the application and modeling of machine learning. Second, linear regression method, gradient boosting regression method, XGBoost method and Stacking method are designed to combine various machine learning algorithms to improve the generalization ability of the model. In addition, an error analysis model is constructed according to four correction methods, and the correction technology research and experimental application of the forecast errors of each station's initial time under the complex terrain of Beijing-Tianjin-Hebei are carried out. Finally, the improvement of the revised forecast of different machine learning methods compared with the original RMAPS-RISE system forecast accuracy is compared.In the experimental part, two modeling ideas are proposed, and four machine learning methods are used to conduct correction and comparison experiments. It shows among the modeling ideas based on error analysis, the Stacking method has the best effect, effectively reducing the forecast error of the original system for the next 3-12 hours for 24 initial times. Among the other three single machine learning method, XGBoost method performs the best, followed by the gradient boosting regression method and linear regression method, and all of them have a significant positive effect on the prediction accuracy. Overall, the forecast error correction model based on machine learning methods can effectively reduce the original forecast error of RMAPS-RISE system, and they have broad application prospects in forecast correction. It is helpful to further improve the forecast accuracy of the objective interpretation product of the site under complex terrain.