Forecasting Different Types of Convective Weather: A Deep Learning Approach

A deep learning objective forecasting solution for severe convective weather(SCW) including short-duration heavy rain(HR), hail, convective gusts(CG), and thunderstorms based on numerical weather prediction(NWP) data was developed. We first established the training datasets as follows. Five years of severe weather observations were utilized to label the NCEP final(FNL) analysis data. A large number of labeled samples for each type of weather were then selected for model training. The local temperature, pressure, humidity, and winds from 1000 to 200 h Pa, as well as dozens of convective physical parameters, were taken as predictors in our model. A six-layer convolutional neural network(CNN) model was then built and trained to obtain optimal model weights. After that, the trained model was used to predict SCW based on the Global Forecast System(GFS) forecast data as input. The performances of the CNN model and other traditional methods were compared. The results show that the deep learning algorithm had a higher classification accuracy on HR and hail than support vector machine, random forests, and other traditional machine learning algorithms. The objective forecasts by use of the deep learning algorithm also showed better forecasting skills than the subjective forecasts by the forecasters. The threat scores(TSs) of thunderstorm, HR, hail, and CG were increased by 16.1%, 33.2%, 178%, and 55.7%, respectively. The deep learning forecast model is currently used in the National Meteorological Center of China to provide guidance for the operational SCW forecasting over China.     

Urban scale modelling: The Sydney hailstorm of 14 April 1999

Summary The aim of this study is to point to the very large improvements that are taking place in a range of modelling applications in the urban areas. The particular phenomenon chosen in this study is a supercell, but it could well have been any other aspect of urban modelling. The Sydney hailstorm of 14 April 1999 was a long-lived, high precipitation supercell that produced a massive damage bill of over 2 billion Australian dollars from its hail swath. The Sydney hailstorm was poorly forecast for a number of reasons including: the severe weather season had officially ended so there were no specialist staff on duty when the hailstorm struck Sydney; the storm proved very difficult to predict and it was expected to continue heading out to sea; and the forecast guidance from all available operational numerical models was inadequate at the resolution required for a supercell simulation. Here, our interest is on the last of the problems, namely, the quality of the operational numerical model guidance, especially given the impact it had on a densely populated urban region.In this study, we compare the numerical guidance available at the time with current modeling capability which, although in research mode at present, will soon be available in real-time mode. The operational models were hydrostatic models run at horizontal resolutions of 25km at best, compared with 1km horizontal resolution for the non-hydrostatic research model. The research model also had a high-order differencing scheme and a sophisticated six phase cloud physics scheme compared with the much simpler parameterized convection in the operational models. The operational model produced very little convective precipitation and it was displaced well to the north of Sydney. The research model generated a supercell with a track and a hail size distribution that was encouragingly close to the observed.     

风廓线雷达产品在冰雹天气过程中的应用研究

本文利用威宁县2022年6月2日一次冰雹天气过程风廓线雷达资料和威宁国家基准气候站降水资料,采用数理统计等方法对本次天气过程的风廓线雷达特征进行分析结果表明：冰雹、短时强降水开始前,低空急流的建立和垂直风切变的加强,急流造成的辐合加上近地层风向的切变作用为冰雹和强降水的产生提供了很好的动力条件;冰雹和短时强降水开始发生时,2000米高度以上有很大的正速度,可以作为一个指标判断对流发展的强弱情况;冰雹发生时折射结构常数（Cn2）值在-128~-110dB之间,也能很好的反映冰雹等强对流天气的开始,增强和结束过程。     

基于中尺度数值模式的分类强对流天气预报方法研究

针对雷暴大风、短时强降水、冰雹和龙卷等强对流天气短期预报,采用0.25°×0.2°每天4次日本气象厅(JMA)东亚地区再分析资料计算的百余类对流参数(物理量)及其15 d滑动平均值,根据“邻(临)近”原则对江苏2001—2009年2—9月各类强对流天气进行时间和站点的匹配后,应用相对偏差模糊矩阵评价技术,对上述对流参数进行权重分配和逐次筛选,获得了既体现强对流与气候平均态间明显差异,又体现自身相对稳定的特征对流参数序列。同时,根据历史分类强对流个例中各特征对流参数的频谱分布获得各对流参数的频率分布分段函数,然后基于中尺度数值模式预报的对流参数,综合历史频率分布和权重分配,构建了分类强对流天气预报概率,并以优势概率作为分类判据,做出强对流分类预报。最后建立了业务化系统,以全自动方式提供分类强对流客观预报产品,投入到日常业务和南京青年奥林匹克运动会气象保障服务工作。     

Diagnostic analyses of convective events — The effect of propagating gust fronts

This paper is devoted to the testing experiment of a postprocessing tool aimed at the objective analysis of propagating gust fronts in a given convective environment. The tool is being developed to be applicable in the operational mode by utilizing NWP model outputs. The experiment was carried out on two summer convective cases which occurred in the Czech Republic. The cases were numerically simulated by the limited area NWP model LM COSMO with the horizontal resolution of 2.8 km. They represent different types of convective systems, both accompanied by objectively identifiable gust fronts and causing heavy precipitation. The event from July 2000 was characterised by the development of isolated thunderstorms. The other event from July 1998 was a long-lasting and organised convective system — a squall line. The hypothesis was that the developed postprocessing tool is capable to evaluate the role which downdraft outflows played in the decay and initiation of convective cells by interaction with convective environment and thus in prolongation of convection lifetime.The procedures of the Objective Analysis of Gust Fronts (OAGF) were applied to the thermodynamic outputs of the LM COSMO. The aim was to determine the position of gust fronts within the domain and to assess their speed of movement and the potential to initiate convection according to the properties of ambient vertical shear and stability as well as humidity conditions ahead of the respective downdraft outflows. In addition, the Radar Simulation Model (RSM) was employed to monitor the simulated convective systems in arbitrary PPI and RHI scans and to verify qualitatively the LM COSMO precipitation forecasts.The case studies has confirmed the applicability of the LM COSMO–OAGF chain and RSM, which may represent the potential for improving the operational nowcasting of hazardous convection phenomena. In both simulations, the objective gust fronts moved on into the vertical shear-favourable environment for triggering new convection. In addition to the dynamical organisation, there were also favourable stability and humidity conditions in the area of forced upward motions in the simulation of the event from July 2000.     

Application of the hybrid convection parameterization scheme HYMACS to different meteorological situations

Convection can cause severe precipitation events and is thus of major interest for quantitative precipitation forecasts. Due to its small horizontal scale convection usually has to be parameterized in numerical weather prediction (NWP) models as a subgrid scale phenomenon. Classical mass flux convection schemes assume grid box sizes much larger than the scale of the convective circulation. Thus, the convective mass transport is closed in the local grid column and no net mass transport occurs on the grid scale.In contemporary NWP models with grid sizes of a few kilometers, where convection is already partially resolved, the classical approach leads to a conceptual problem. This can be overcome by a hybrid mass flux convection scheme (HYMACS), in which only the small scale convective updrafts and downdrafts are parameterized, whereas the treatment of the larger scale environmental subsidence is left to the grid scale equations. Different to the classical schemes, HYMACS produces a net convective mass flux exerting pressure gradient forces to the grid scale model.We discuss real cases of different meteorological situations simulated with the operational weather forecast model COSMO of the German Meteorological Service as the hosting model of HYMACS. Precipitation and cloud top pressures simulated with HYMACS are compared with results from classical convection schemes and observational data from the DWD station network and the SEVIRI (MSG) satellite instrument.     

雷暴与强对流临近天气预报技术进展

临近预报指0—6h(0—2h为重点)的高时空分辨率的天气预报,预报对象是该时段内出现明显变化的天气现象,主要包括雷暴、强对流、降水、冬季暴风雪、冻雨、沙尘暴、低能见度(雾)、天空云量等,其中,以雷暴和强对流天气的临近预报最具挑战性。综述了针对雷暴和强对流天气的以主观预报为主、结合客观算法的临近预报技术,同时讨论了高分辨率数值预报模式在临近预报中的应用。主观临近预报技术包括基于多普勒天气雷达观测数据并结合其他资料(常规高空和地面观测、气象卫星云图、快速同化循环的数值预报产品等)对雷暴生成、发展和衰减,特别是对强对流天气(包括强冰雹、龙卷、雷暴大风和对流性暴雨)的临近预报,客观算法包括几种应用最广的雷达回波或云图外推算法和强对流天气识别技术。高分辨率数值预报模式的应用包括与雷达回波外推融合延长临近预报时效,与各种观测资料融合得到快速更新的三维格点资料为雷暴和强对流近风暴环境的判断提供重要参考。     

Climatology of cyclones over the southwest Pacific: 1992–2001

Summary The Australian east coast is subject to a range of synoptic/mesoscale weather systems ranging from low-pressure troughs, small explosive low-pressure systems near the Australian east coast, to synoptic scale maritime low-pressure systems, and both tropical and extra-tropical cyclones. These systems vary dynamically in intensity and structure across a spectrum that includes both weak and intense cold-cored systems, to warm-cored tropical cyclones and “hybrid” systems. A preliminary 10 year climatology, from 1992 to 2001, is presented that concentrates on those cyclones in open waters that threaten life and property as a result of heavy coastal rain or the combined effects of wind, sea state and swell, including both long fetch and storm surges. There are 28 systems in the 10-year climatology. In terms of their development, the most numerous were the 50% of storms that occurred as systems within an easterly trough, followed by 21% that occurred as circulations on decaying Tasman Sea cold fronts. In addition, a numerical weather prediction (NWP) modelling case study of the ‘hybrid’ sub-tropical cyclone system of March 2001 was carried out using archived real-time data. Operational numerical weather prediction (NWP) model forecasts of this system, which was not named in accordance with Australian classification policy, were all poor except for the US MRF model, which provided marginally useful guidance. None of the operational global and regional models available to forecasters predicted an intense cyclone, and only the US MRF model predicted landfall. A high-resolution forecast using all available data produced significantly improved predictions over the operational models up to four days before landfall.     

VIL产品在高山雷达中的应用研究

利用2006—2009年丽江市气象台的冰雹观测记录、短时强降水、民政部门提供的冰雹灾情资料和丽江CINRAD/CC雷达观测资料,对垂直累积液态水含量(Vertically Integrated Liquid Water Con-tent,VIL)进行了统计分析。结果表明,当VIL值>5kg·m-2时将会出现强对流天气,当VIL≥21kg·m-2时预示着发生冰雹的可能性极大,是人工防雹作业启动的关键值,冰雹云中的VIL值远远大于带来短时强降水的普通对流云;在降雹前VIL值都有明显的跃增;距离雷达站较近或较远区域的VIL值会被低估,在37kg·m-2以下也会降雹,通常降雹的VIL值大都在37kg·m-2以上。     

塔里木盆地周边地区冰雹云特征分析

对1998～2001年6～9月在塔里木盆地普查得到105个降雹云团进行分类，按尺度大小分为雷暴云、对流云、中尺度对流系统、冷云核和系统云系云区5类。塔里木盆地2000年降雹云团出现最多，半数以上出现在5月和6月。塔里木盆地降雹云团尺度小、形状不规则、云顶温度较高。     

基于天气形势分型的珠江三角洲强对流潜势预报方法

珠江三角洲出现强对流的机制有多种情况,不同天气形势下引起的强对流,其出现时各项强对流指数必然存在一定的差异。本文对强对流的触发机制进行天气形势分型(冷空气影响触发、西南暖湿气流触发、后汛期副高边缘影响,台风影响等),利用广州热带海洋研究所中尺度模式GRAPES的输出资料,计算出不同天气形势下强对流发生前1h当地各层的涉及水汽条件、不稳定能量、动力条件等强对流指数,对强对流指数阈值进行选取和验证,统计出珠江三角洲不同天气形势下短时强降水和雷雨大风出现时的强对流指数阈值,并建立相应的潜势预报系统,分别通过一次典型的雷雨大风和一次短时强降水天气过程验证,表明该系统提供的强对流分型潜势预报概率产品可以较好的指示出强对流发生的时间和位置以及强对流出现的类型,预报与实况观测比较一致。     

强对流天气监测预报预警技术进展

强对流天气预报业务包括监测、分析、预报、预警和检验等方面。对流初生识别、对流系统强度识别和对流天气类型识别等监测技术取得新进展,综合多源资料的监测技术已应用于中国气象局中央气象台业务。对流系统的触发、发展和维持机制等获得了新认识,我国不同类型强对流天气及其环境条件统计气候特征、分析规范及相应业务产品等为业务预报提供了必要基础和技术支撑。光流法、多尺度追踪技术以及应用模糊逻辑方法的临近预报技术等有明显进展,融合短时预报技术得到广泛应用,对流可分辨高分辨率数值 (集合) 预报及其后处理产品预报试验取得了显著成效,基于数值 (集合) 预报应用模糊逻辑方法的分类强对流天气短期预报技术为业务预报提供了技术支撑。强对流天气综合监测和多尺度自适应临近预报技术、多尺度分析技术以及融合短时预报技术、发展并应用模糊逻辑等方法的、基于高分辨率数值 (集合) 模式的区分不同强度等级和极端性的分类强对流天气精细化 (概率) 预报技术等是未来发展的主要方向。     

Development and behaviour of a radar-based operational tool for hailstorms identification

Hailstorms are a common meteorological phenomenon in Spain which causes substantial economic damage especially in spring and summer. During 2000 and 2001, a radar-based application for convective monitoring was developed at STAP (Forecasting and Analysis Techniques Department); in order to improve this nowcasting tool, it is needed to include an advanced procedure to estimate the presence of hail as a complementary module in the operational application. A preliminary study, carried out during 2001, showed that the Waldvogel technique, VIL (vertical integrated liquid) density and the hail detection algorithm (HDA) are the methodologies which proved more accurate in our latitudes.Throughout the spring and summer of 2001, all available information about hail events (time, place and hail size) and about storms without hail was used for tuning an experimental module to detect severe and non-severe hailstorms, taking into account the selected algorithms. During 2002, further information about storms was gathered in order to assess the behaviour of the developed hail module. Afterwards, a new module calibration was carried out with the information of the complete storms data set of 2001 and 2002. With this new calibration, the hail module became an operational tool during the spring and summer of 2003. Currently, the verification of the operational module using the data from the storms in 2003 is being elaborated. The aim of this work is to put forward the calibration procedure carried out, the verification results of 2002, and the performance of the hail estimate module in selected cases during 2003 in Spain.     

江西局地强对流天气的多普勒天气雷达产品特征

利用江西省2002--2007年局地强对流天气过程的多普勒天气雷达产品资料,通过统计分析多普勒天气雷达最大反射率因子、垂直积分液态含水量及其密度、三体散射和中气旋等产品特征,对江西省突发性局地强对流天气的临近预报预警进行研究。结果认为,江西突发性局地强对流天气主要出现在春季和夏季,夏季突发性局地强对流风暴的伸展高度普遍比春季高。最大反射率因子≥55dBz可以作为局地强对流天气预警的临界指标,≥60dBz可以作为局地冰雹预警的临界指标;VIL密度≥2．8g/m^3可以作为局地强对流天气预警的临界指标,≥3．2g/m^3可以作为局地冰雹预警的临界指标,≥4．0g/m^3可以作为较大冰雹预警的临界指标。三体散射是大冰雹的有效判据。风暴内出现中气旋特征,应立即发布强对流天气预警。很多突发性局地强风暴不一定会出现中气旋特征报警,在这种情况下应通过分析径向速度产品来判断中气旋特征。     

鲁中地区分类强对流天气环境参量特征分析

将山东中部地区16 a暖季(4-9月)106次伴随瞬时风力不低于8级的强对流个例划分为雷暴大风、冰雹雷暴大风和强降水混合型等3种类型,利用常规探空资料和地面观测资料,通过箱须图的形式分别讨论3种类型对应的一系列关键环境参数的分布特征和预报阈值。进一步,又将上述106次个例中的特强对流个例,包括产生25 m/s以上瞬时大风的特强雷暴大风个例、产生不小于20 mm直径冰雹的特强冰雹个例以及50 mm/h或以上强度的特强短时强降水个例提取出来构成一个子集,讨论其关键环境参数分布特征和预报阈值,并与全部对流个例的相应关键环境参数进行比较。最后,对鲁中地区强对流系统的触发机制进行了简要阐述和讨论。结果表明:(1)雷暴大风型、冰雹雷暴大风型和强降水混合型对应的850和500 hPa温差的最低阈值为25℃; 3种类型对应的地面露点最低阈值分别为13、16和24℃; 相应的大气可降水量最低阈值分别为20、24和32 mm; 相应对流有效位能的最低阈值分别为300、900和1300 J/kg; 相应的0-6 km风垂直切变最低阈值分别为12.0、12.5和8.0 m/s。(2)通过地面露点、大气可降水量以及暖云层厚度等关键参数的分布特征可以将上述3种类型的前两种与第3种类型即强降水混合型进行一定程度的区分,但要通过各个关键参数的分布特征区分前两种强对流天气是困难的。(3)对于伴随冰雹的强对流天气,适宜的融化层高度为3.0-3.9 km; (4)特强雷暴大风、特强冰雹和特强短时强降水等3种特强对流类型与全部强对流个例的3种类型相比,其条件不稳定度明显增大,体现为850和500 hPa温差的增大、水汽条件有所加强、对流有效位能明显增大,3种类型特强对流天气对应的对流有效位能最低阈值分别为1000、1100和2000 J/kg; 相应的0-6 km风垂直切变最低阈值分别为16、12和11 m/s,即特强雷暴大风型和特强短时强降水型的风垂直切变阈值明显增大。上述工作构成了山东中部伴随雷暴大风的强对流天气短时预报的一个基础,结合各类强对流天气发生的气候概率,可以通过决策树或模糊逻辑方法制作成适合于地、市气象台的分类强对流天气短时预报系统。      

云可分辨模式中误差的增长和传播对定量降水可预报性的影响

Limitations in the predictability of quantitative precipitation forecasting （QPF） that arise from initial errors of small amplitude and scale are investigated by means of real-case high-resolution （cloud-resolving） numerical weather prediction （NWP） integrations. The case considered is the hail and wind disaster that occurred in Sichuan on 8 April 2005. A total of three distinct perturbation methods are used. The results suggest that a tiny initial error in the temperature field can amplify and influence the weather in a large domain, changing the 12-h forecasted rainfall by as much as one-third of the original magnitude. Furthermore, the comparison of the perturbation methods indicates that all of the methods pinpoint the same region （the heavy rainfall areas in the control experiment） as suffering from limitations in predictability. This result reveals the important role of nonlinearity in severe convective events.     

石河子地区强对流天气统计特征及防雹作业指标

通过148团数字化雷达1994～2001年5～8月8年的回波资料，对系统性对流、局地对流以及地面产生降雹情况进行统计，得出石河子地区对流雹暴的主要路径、时空分布特征。     

中国当代强对流天气研究与业务进展

对当代中国几十年来强对流天气研究和业务进展做了阐述，主要包括强对流系统产生的环境背景和主要组织形态，以及具体强对流天气的有利环境条件、触发机制、卫星云图特征、多普勒天气雷达回波特征以及预报、预警技术等诸方面。总体来看，中国学者对强对流以及不同类型强对流天气（强冰雹、龙卷、雷暴大风）发生、发展的环流背景以及通过雷达和卫星观测到的组织结构及其演变特征都已有了明确认识，研究了对流系统的多种触发机制，深入认识了超级单体、飑线等对流系统的环境条件、组织结构特征和维持机制，了解了中国中尺度对流系统的组织形态和气候分布特征，获得了强冰雹、龙卷、下击暴流和雷暴大风等的雷达、卫星和闪电等的多尺度观测特征、形成机制和现场灾害调查特征，发展了各类强对流天气识别、监测和分析方法以及基于“配料法”和深度学习方法等的预报、预警技术等。因此，强对流天气业务预报水平已得到显著提升。      

恩施山区强冰雹和短时强降水天气落区分析

利用恩施雷达资料,对湖北恩施山区强冰雹和短时强降水天气的落区进行了分析。结果发现:强冰雹的右移特征非常明显,当恩施州境内或边境临近地区有新生的中空回波,且回波移动方向上也同时有对流回波滋生时,则两回波源地之间的区域基本可以确定为此次冰雹过程的主要影响区域,新生中空回波向右偏离初始移动方向大约0~60°之间(两处回波源地之间的连线),一般在1~3h内不断加强形成强冰雹而落下;而短时强降水右移特征不明显,发生短时强降水区域不仅与回波系统的走向和移动有关,还与回波单体移动和传播有关系,其回波系统的移动是对流单体的移动和传播两者的矢量之和,同时根据回波形态及其演变趋势,即可判断短时强降水的落区。      

NWP产品在强对流天气诊断分析中的应用

利用ECMWF、T213、T106数值预报产品及各物理量场,对已知中尺度系统发生发展的大尺度条件,预报中尺度环流出现的统计概率进行探讨:首先将武汉中心气象台过去总结的强对流天气模型与数值预报产品中的环流形势进行对比分析,当所预报的环流形势满足强对流天气模型时,认为大尺度条件将会促进中尺度天气的发生发展。再对数值预报产品中的有关物理要素场和值进行诊断分析,当所诊断的结果反映出有中尺度次级环流出现或有利中尺度对流天气发生时,最后对强对流天气落区、落点及降水性质进行诊断分析。同时,还利用AREM中尺度数值模式对2002年7月21~24日过程进行诊断分析,预报效果较好。,通过诊断分析,得出了强对流天气落区、落点及降水性质与各要素之间预报场和值的关系,同时也为今后精细化预报提供了启示和参考。