基于ECMWF产品的站点气温预报集成学习误差订正

提出一种基于数值模式预报产品的气温预报集成学习误差订正方法,通过人工神经网络、长短期记忆网络和线性回归模型组合出新的集成学习模型（ALS模型）,采用2013—2017年的欧洲中期天气预报中心数值天气预报模式2 m气温预报产品和中国部分气象站点数据,利用气象站点气温、风速、气压、相对湿度4个观测要素,挖掘观测数据的时序特征并结合模式2 m气温预报结果训练机器学习模型,对2018年模式2 m气温6~168 h格点预报产品插值到站点后的预报结果进行偏差订正。结果表明:ALS模型可将站点气温预报整体均方根误差由3.11℃降至2.50℃,降幅达0.61℃（19.6%）,而传统的线性回归模型降幅为0.23℃（8.4%）。ALS模型对站点气温预报误差较大的区域和气温峰值预报的订正效果尤为显著,因此,集成学习方法在数值模式预报结果订正中具有较大的应用潜力。     

基于机器学习的复杂地形下短期数值天气预报误差分析与订正

初步研发了一套基于机器学习方法XGBoost且考虑地形特征影响的数值预报多模式集成技术,并与传统的等权重平均和线性回归方法的集成效果进行了对比分析。利用北京地区快速更新循环数值预报系统每天8次循环预报给出的近地面2 m温度、2 m相对湿度、10 m风速、10 m风向数据产品,分别基于机器学习方法XGBoost、等权重平均方法、线性回归方法构建了3种体现地形因子影响的多模式预报时间滞后集成模型。试验对比分析了暖季、冷季每日不同时刻的模式预报集成订正效果。结果表明:分季节试验中,基于XGBoost模型对2 m温度、10 m风速的集成预报结果相对原始最优预报结果误差明显优于其他两种传统方法。XGBoost对2 m温度集成的误差可降低11.02%—18.09%,10 m风速集成误差可降低31.23%—33.22%,10 m风向集成误差可降低4.1%—8.23%。2 m相对湿度的集成预报误差与传统方法接近。基于XGBoost的多模式集成预报模型可以充分“挖掘”不同模式或不同时刻快速更新循环预报优点,有效降低模式的系统性误差,提供准确性更高的多模式集成确定性预报产品。      

东印度洋天气和风暴潮预报系统的初步评估

东印度洋天气和风暴潮实时预报系统（EPMEF_EIO）由区域大气模式和区域风暴潮模型组成,每天实时运行4次.大气初边场来自美国国家环境预报中心（NCEP）的全球预测系统（GFS）,通过区域嵌套得到印度洋-东印度洋-斯里兰卡区域的3 d预报结果.大气模式的10 m预报风场驱动风暴潮模式,得到东印度洋-斯里兰卡区域的潮汐和风暴潮3 d预报结果.通过与中国科学院南海海洋研究所斯里兰卡站气象塔观测数据、最优台风路径数据和科伦坡水位站数据对比,发现模式预报气温和相对湿度的日变化较观测值偏小,气温总体RMSE为1.26℃,相关系数为0.8,相对湿度的总体RMSE为7.0%,相关系数为0.7;模式预报风速以整体偏大为主,总体RMSE为2.3 m/s,相关系数为0.65;模式预报风向能把握主要的变化趋势,RMSE在20°~32°之间,相关系数约0.65;模式24、48和72 h路径预报平均误差分别为110.5、166.4和181.0 km.此外,模式水位预报的RMSE为0.035 m,占最大振幅约5%,与观测的相关系数达到0.996.这说明了模式可以用于预报潮汐和风暴潮过程.     

对一次温带气旋引发的渤海风暴潮过程的数值模拟

利用三维非静力中尺度大气模式—MM5（Version 3.7）输出的黄渤海海面风场和气压场预报资料,用三维斜压陆架海模式—HAMSOM对2008年8月22日温带气旋造成的渤海风暴潮过程进行了模拟,得到逐时的渤海增水场、渤海风暴潮流场,与验潮站的观测数据进行比较。结果表明：在渤海西部已经转西北风的情况下,塘沽出现了121 cm的高增水,造成这种现象的原因很复杂,其中远距离的气旋作用产生的北黄海海域偏东大风导致北黄海水体大量涌入渤海应该是一个主要原因。这也是今后预报业务中必须特别关注的产生风暴潮的重要因素。数值模拟的塘沽测站的风暴潮增水极值及增水过程都和实测值符合较好,本次过程中数值预报能够很好地模拟出这种特殊的风暴潮。在离岸风的情况下产生风暴潮,这仅靠预报员凭经验主观分析判断是很难的,数值预报可以弥补预报员主观分析的不足。     

基于ConvLSTM的广西短临降水预报

针对传统雷达回波外推算法在快速增长或消散降水过程预报精度较低的问题,利用华南雷达回波拼图资料数据,建立ConvLSTM回波外推模型,对广西区域范围进行短临降水预报研究.采用气象业务中的正确率(POD)、临界成功指数(CSI)及误报率(FAR)评判标准检验预报模型,并将ConvLSTM与光流法的预报结果进行对比分析.结果 表明,ConvLSTM模型的CSI、POD分别比光流法提高0.06和0.059,而FAR下降了0.058.ConvLSTM方法比光流法的回波外推预报准确率高,该方法可为广西短临降水预报提供新的参考.     

支持向量机在大气污染预报中的应用研究

支持向量机是基于统计学习理论的新一代机器学习技术，其非线性回归预测性能优越于传统统计方法。利用前一天该污染物的日均浓度、前一天地面平均风速等7个预报因子建立了基于RBF核函数支持向量回归法的大气污染预报模型，并利用十重交叉验证和网格搜索法寻找模型最优参数。乌鲁木齐大气预报实例表明：支持向量机显示出小样本时预报精度较高和训练速度快的独特优势，为空气质量预报提供一种全新的模式。     

基于SVD与机器学习的华南降水预报订正方法

降水是在多种天气系统和复杂物理过程共同影响下形成的,因此降水预报难度较大。由于数值预报模式的局限性,使得模式预报产品存在一定误差。为探讨更加有效的模式预报产品误差订正方法,基于奇异值分解（SVD）与机器学习（多元线性回归、套索回归、岭回归）构建订正模型,对2007—2019年4月1日—6月30日华南前汛期欧洲中期天气预报中心（EC）模式降水预报产品进行误差订正试验。结果表明:基于SVD与机器学习相结合的订正模型能有效降低EC模式降水预报产品在华南的预报误差,均方根误差最大优化率达4.2%,累计超过69%的站点得到不同程度的优化;SVD与机器学习相结合的订正模型能很好地处理因子间共线性问题,具有更好的鲁棒性;而对多个订正模型加权集成,均方根误差优化率达5.7%,累计超过77%的站点得到优化,显然加权集成方法订正效果不仅优于EC模式预报产品,也优于参与集成的任一订正模型。     

南海海洋气象数值预报系统(Grapes-MAMS)及其业务应用

介绍了中国气象局广州热带海洋气象研究所建立的南海海洋气象数值预报系统(Grapes-MAMS)框架、系统中各模式特点以及针对模式所做的改进工作;并基于广东省沿海浮标资料,分析和研究了该系统中各个专业模式预报性能。得到如下结论:(1)从给出的统计检验和技巧评分来看,区域海浪模式逐年改进,具有一定预报能力;区域海浪模式和欧洲中期天气预报中心全球海浪模式相比,前者预报技巧TS评分大于后者;(2)实测海表温度的同化明显改进了区域海洋环流模式对海表温度的预报,检验结果显示海表温度预报与浮标观测海表温度相关系数在0.7左右,通过了α=0.01显著性检验;(3)将基于广东茂名科学试验基地得到的拖曳系数计算方案引入风暴潮模式,可以有效地改进风暴潮模式对强台风"威马逊"风暴增水预报,2013—2015年6个风暴潮模式预报检验显示,模式可以提前20 h以上给出最大风暴增水预报,而且最大风暴增水预报与实际相差不大。     

统计预报海温场驱动的CAM3.1模式预报试验

基于动力气候模式进行月一季尺度预报的“两步法”思想,提出一种新的预报海温场统计模型,并以该统计模型预报的海温场驱动NCAR CAM3.1模式对1981-2000年月时间尺度的东亚500 hPa高度距平场进行客观回报试验;在此基础上,提出了对预报结果的订正方法。结果表明:统计预报海温模型的预报海温场能够反映出全球海温空间分布的基本特征,并对表征ENSO事件的Ni?o3.4区海温变化的预报能力较强。该统计模型预报的海温场驱动的CAM3.1模式可以较好地预报出东亚500 hPa环流的主要分布特征,试验表明:适当的统计订正方法可以在一定程度上提高CAM3.1模式对东亚夏季500 hPa环流背景的预报技巧。     

机器学习在强对流监测预报中的应用进展

近年来,机器学习理论和方法应用蓬勃发展,已在强对流天气监测和预报中广泛应用。各类机器学习算法,包括传统机器学习算法(如随机森林、决策树、支持向量机、神经网络等)和深度学习方法,已在强对流监测、短时临近预报、短期预报领域发挥了积极的重要作用,其应用效果往往明显优于依靠统计特征或者主观经验积累的传统方法。机器学习方法能够更有效提取高时空分辨率的中小尺度观测数据的强对流特征,为强对流监测提供更全面、更强大的自动识别和追踪能力;能够有效综合应用多源观测数据、分析数据和数值预报模式数据,为强对流临近预报预警提取更多有效信息;能够有效对数值模式预报进行释用和后处理,提升全球数值模式、高分辨率区域数值模式在强对流天气预报上的应用效果。最后,给出了目前机器学习方法应用中存在的问题和未来工作展望。     

Storm surges in the Western Baltic Sea: the present and a possible future

Globally-coupled climate models are generally capable of reproducing the observed trends in the globally averaged atmospheric temperature or mean sea level. However, the global models do not perform as well on regional/local scales. Here, we present results from four 100-year ocean model experiments for the Western Baltic Sea. In order to simulate storm surges in this region, we have used the General Estuarine Transport Model (GETM) as a high-resolution local model (spatial resolution ≈ 1?km), nested into a regional atmospheric and regional oceanic model in a fully baroclinic downscaling approach. The downscaling is based on the global model ECHAM5/MPI-OM. The projections are imbedded into two greenhouse-gas emission scenarios, A1B and B1, for the period 2000–2100, each with two realisations. Two control runs from 1960 to 2000 are used for validation. We use this modelling system to statistically reproduce the present distribution of surge extremes. The usage of the high-resolution local model leads to an improvement in surge heights of at least 10% compared to the driving model. To quantify uncertainties associated with climate projections, we investigate the impact of enhanced wind velocities and changes in mean sea levels. The analysis revealed a linear dependence of surge height and mean sea level, although the slope parameter is spatially varying. Furthermore, the modelling system is used to project possible changes within the next century. The results show that the sea level rise has greater potential to increase surge levels than does increased wind speed. The simulations further indicate that the changes in storm surge height in the scenarios can be consistently explained by the increase in mean sea level and variation in wind speed.     

海面风场订正对风暴潮数值模拟的影响

海洋模式在模拟风暴增水时对于风场的依赖性比较大,准确的风场模拟是正确模拟风暴潮的重要前提.用渤海海域4次典型的风暴潮个例,检验改良的渤海风场推算模式对风暴潮数值模拟的影响,分别用订正前后的风场驱动风暴潮数值模式得到逐时的渤海增水场,并与塘沽实测值比较表明:在4个过程中海面风场订正后风暴潮模式的模拟结果均得到了明显的改进...     

Statistical Characteristics and Modeling of Storm Surges in the North Caspian Sea

The characteristics of storm surges obtained from sea level observations at four hydrometeorological stations in the North Caspian Sea for 2003–2017 are presented. The sea level that by 30 cm exceeds the monthly mean value at the analyzed point of the Caspian Sea was considered as a surge. In total, 370 surges were registered, 83% of them occurred during the cold season (September-April). The maximum surge height was 125 cm, the longest duration was 7 days. The most significant surges on Tyulenii Island were simulated with the operational hydrodynamic model of the sea level and currents of the Caspian Sea using atmospheric forcing from the COSMO model. The mean coefficient of correlation between the simulated and observed sea level is equal to 0.94.     

一个中国沿岸台风风暴潮数值预报系统的建立与应用

依据三维斜压海洋环流模式POM建立了一个中国沿岸台风风暴潮数值预报业务系统.台风风场模型考虑了台风移动和周围环境风场的影响,采用了较合理的强风情况下的风应力计算公式,建立了稳定合理的模式海洋环流气候状态和模式边界条件.大量的数值模拟结果表明,该模式能较好地重现历史台风风暴增水过程,对近2年台风风暴潮个例的预报结果表明,该业务系统对台风风暴增水具有较好的预报能力,文章同时分析了模式存在的一些问题.该业务系统实现了从资料采集、模式运行到预报结果输出的全自动化,显示采用图片和MICAPS两种方式,后者与现有气象业务平台一致.     

Storm Surges in the Strait of Georgia Simulated with a Regional Model

The Strait of Georgia is a large, semi-enclosed body of water between Vancouver Island and the mainland of British Columbia connected to the Pacific Ocean via Juan de Fuca Strait at the south and Johnstone Strait at the north. During the winter months, coastal communities along the Strait of Georgia are at risk of flooding caused by storm surges, a natural hazard that can occur when a strong storm coincides with high tide. This investigation produces storm surge hindcasts using a three-dimensional numerical ocean model for the Strait of Georgia and the surrounding bodies of water (Juan de Fuca Strait, Puget Sound, and Johnstone Strait) collectively known as the Salish Sea. The numerical model employs the Nucleus for European Modelling of the Ocean architecture in a regional configuration. The model is evaluated through comparisons of tidal elevation harmonics and storm surge with observations. Important forcing factors contributing to storm surges are assessed. It is shown that surges entering the domain from the Pacific Ocean make the most significant contribution to surge amplitude within the Strait of Georgia. Comparisons between simulations and high-resolution and low-resolution atmospheric forcing further emphasize that remote forcing is the dominant factor in surge amplitudes in this region. In addition, local wind patterns caused a slight increase in surge amplitude on the mainland side of the Strait of Georgia compared with Vancouver Island coastal areas during a major wind storm on 15 December 2006. Generally, surge amplitudes are found to be greater within the Strait of Georgia than in Juan de Fuca Strait.     

Simulation and Analysis of Sea Floods in the Don River Delta

Storm surges and wind waves in the Taganrog Bay (the Sea of Azov) are simulated with the ADCIRC+SWAN numerical model, and the mechanisms of the Don River delta flooding are analyzed. It is demonstrated that the most intensive flooding of the Don River delta occurs in case of southwestern wind with the speed of not less than 15 m/s. A storm surge leads to the intensification of wind waves in the whole Taganrog Bay due to the general sea level rise. As a result, the significant wave height near the Don River delta increases by 0.5–0.6 m.     

渤海湾风暴潮天气系统及风场结构个例分析

利用常规资料、中尺度模式(MM5)模拟结果和自动站资料分析了2003年10月11日渤海湾风暴潮的天气系统及风场结构特征。分析表明，强冷空气从偏东路径南下与强盛的暖湿急流及地面倒槽的共同作用，形成持续强东北大风是引发渤海风暴潮的主要强迫动力，风应力海水向岸输送(增水)范围约100km持续时间10～12h，风应力增水作用与天文大潮相叠加直接导致风暴潮的发生。1985年以来渤海湾成灾风暴潮发生的几率明显增加。     

Storm surges in Canadian waters

Abstract

Storm surges in various Canadian waters are reviewed. Following a brief discussion of the weather systems that cause storm surges in Canadian coastal and inland waters, the mathematical formulations to describe the development of storm surges are given. In reviewing storm surges in the different Canadian waters, particular attention is given to describe the influence of the presence of sea ice on surge development and the impact of shallow coastal areas, where the coastline configuration is itself changed by the surge, on inland penetration of the storm surge. The Canadian waters that may be affected by storm surges include the east and west coasts, the Beaufort Sea, the Gulf of St. Lawrence and the St. Lawrence estuary, Hudson Bay and the Great Lakes.     

Comparison of three methods for estimating the sea level rise effect on storm surge flooding

Two linear methods, including the simple linear addition and linear addition by expansion, and numerical simulations were employed to estimate storm surges and associated flooding caused by Hurricane Andrew for scenarios of sea level rise (SLR) from 0.15 m to 1.05 m with an interval of 0.15 m. The interaction between storm surge and SLR is almost linear at the open Atlantic Ocean outside Biscayne Bay, with slight reduction in peak storm surge heights as sea level rises. The nonlinear interaction between storm surges and SLR is weak in Biscayne Bay, leading to small differences in peak storm surge heights estimated by three methods. Therefore, it is appropriate to estimate elevated storm surges caused by SLR in these areas by adding the SLR magnitude to storm surge heights. However, the magnitude and extent of inundation at the mainland area by Biscayne Bay estimated by numerical simulations are, respectively, 22–24 % and 16–30 % larger on average than those generated by the linear addition by expansion and the simple linear addition methods, indicating a strong nonlinear interaction between storm surge and SLR. The population and property affected by the storm surge inundation estimated by numerical simulations differ up to 50–140 % from that estimated by two linear addition methods. Therefore, it is inappropriate to estimate the exacerbated magnitude and extent of storm surge flooding and affected population and property caused by SLR by using the linear addition methods. The strong nonlinear interaction between surge flooding and SLR at a specific location occurs at the initial stage of SLR when the water depth under an elevated sea level is less than 0.7 m, while the interaction becomes linear as the depth exceeds 0.7 m.