基于数值预报和随机森林算法的强对流天气分类预报技术

随机森林算法是当前得到较为广泛应用的机器学习方法之一,有着很高的预测精度,训练结果稳定,泛化能力强,解决多分类问题有明显优势。本文将随机森林算法应用于强对流的潜势预测和分类,分短时强降水、雷暴大风、冰雹和无强对流四种类别,基于2005—2016年NCEP 1°×1°再分析资料计算的对流指数和物理量,开展强对流天气的分类训练、0～12 h预报和检验,经2015—2016年独立测试样本检验表明,针对强对流发生站点的点对点检验,整体误判率为21. 9%,85次强对流过程基本无漏报,模型尤其适用于较大范围强对流天气。随机森林算法筛选的因子物理意义较为明确,和主观预报经验基本相符,模型准确率高,可用于日常业务。     

Using a Convection Model to Predict Altitudes of White Stork Migration Over Central Israel

Soaring migrants such as storks, pelicans and large birds of prey rely on thermal convection during migration. The convection model ALPTHERM was designed to predict the onset, strength, duration and depth of thermal convection for varying topographies for glider pilots, based on atmospheric conditions at midnight. We tested ALPTHERM predictions as configured for two topographies of central Israel, the Coastal Plains and the Judean and Samarian Mountains in order to predict altitudes of migrating white storks (Ciconia ciconia). Migrating flocks of white storks were tracked with a motorized glider, to measure maximum altitudes of migration during spring 2000. A significant positive correlation was found between the maximum daily altitudes of migration measured and the predicted upper boundary of thermal convection for the Coastal Plains and Samarian Mountains. Thirty-minute predictions for the Coastal Plains and Samarian Mountains correlated positively with measured maximum migration altitudes per thermal. ALPTHERM forecasts can be used to alter flight altitudes in both civil and especially military aviation and reduce the hazard of serious aircraft collisions with soaring migrants.     

A Numerical Verification of Self-Similar Multiplicative Theory for Small-Scale Atmospheric Turbulent Convection

The self-similar multiplicative theory（SSM theory）, aims to interpret the scaling behavior of the temperature structure function. In the present paper, the author report results from a numerical simulation of atmospheric turbulent convection in order to verify this theory. The simulation was based on a shell model which was deduced from simplified atmospheric convection equations. The numerical results agreed well with the theory prediction of scaling law from the first order to the eighth order. They also showed that the prediction of this theory was better than that given by the Kolmogorov＇s theory in 1941, log-normal, and β model theories.     

Salt transport on islands in the Okavango Delta: Numerical investigations

This study uses a numerical model to investigate the groundwater flow and salt transport mechanisms below islands in the Okavango Delta. Continuous evapotranspiration on the islands results in accumulation of solutes and the formation of a saline boundary layer, which may eventually become unstable. A novel Lagrangian method is employed in this study and compared to other numerical methods. The numerical results support the geophysical observations of density fingering on Thata Island. However, the process is slow and it takes some hundreds of years until density fingering is triggered. The results are sensitive to changes of the hydraulic gradient and the evapotranspiration rate. Small changes may lead to different plume developments. Results further demonstrate that density effects may be entirely overridden by lateral flow on islands embedded in a sufficiently high regional hydraulic gradient.     

An improvement of the mass flux convection parameterization scheme and its sensitivity tests for seasonal prediction over China

A modified cumulus parameterization scheme, suitable for use in a seasonal forecast model, is presented. This parameterization scheme is an improvement of the mass flux convection scheme developed by Gregory and Rowntree (1989; 1990). This convection scheme uses a “bulk” cloud model to present an ensemble of convective clouds, and aims to represent shallow, deep, and mid-level convection. At present,this convection scheme is employed in the NCC T63L20 model (National Climate Center, China Meteorological Administration). Simulation results with this scheme have revealed some deficiencies in the scheme,although to some extent, it improves the accuracy of the simulation. In order to alleviate the deficiencies and reflect the effect of cumulus convection in the actual atmosphere, the scheme is modified and improved.The improvements include (i) the full estimation of the effects of the large-scale convergence in the lower layer upon cumulus convection, (ii) the revision of the initial convective mass flux, and (iii) the regulation of convective-scale downdrafts. A comparison of the results obtained by using the original model and the modified one shows that the improvement and modification of the original convection scheme is successful in simulating the precipitation and general circulation field, because the modified scheme provides a good simulation of the main features of seasonal precipitation in China, and an analysis of the anomaly correlation eoetfieient between the simulation and the observations confirms the improved results.     

冷涡背景下山东省“5·17”极端强对流天气环境条件分析

2020年5月17日,山东省出现大范围强对流天气(简称“5·17”强对流),冰雹范围之广为近10年之首。对流风暴高度组织化,区域性的超级单体群以及一条长度超过500 km的强飑线造成此次极端强对流天气。利用ERA5再分析、加密自动气象观测站、多普勒天气雷达等资料,剖析了此次极端强对流天气的环境条件。结果表明:冷涡位于最有利于山东出现强对流的关键区,大尺度天气系统强迫强,对流层中层异常强的冷空气南下影响前期异常增暖的山东地区,造成“5·17”极端强对流。天气系统的异常程度更能代表动热力强迫的强度,异常程度达到2σ以上有可能造成极端强对流天气。当冷涡南下过程中强度减弱,但异常程度增加时,其东南象限仍能产生极端强对流天气。强的深层垂直风切变有利于对流风暴组织化发展,飑线的长轴走向与0～6 km垂直风切变矢量方向相同,新单体发生、发展、合并的区域位于风矢量差大值中心前沿。低层暖湿平流源源不断地向山东输送暖湿空气,是CAPE重建的机制,是超级单体群和长飑线得以长时间维持的主要能量来源。     

1971—2000年我国东部地区雷暴、冰雹、雷暴大风发生频率的变化

提要：应用1971—2000年华北、华中、华东各省1084个站的地面天气现象观测资料,统计分析了江南（31°N以南）、江淮和黄淮（31°～37°N）及黄河以北（37°N以北）雷暴、冰雹和雷暴大风的年代际变化特征。分析结果表明：雷暴、冰雹、雷暴大风发生频率呈减少趋势。其中江南地区雷暴年代际递减更明显,黄河以北地区冰雹年代际递减最明显,而雷暴大风在三个区域的年代际递减均明显。雷暴大风常与冰雹伴随发生,高原和山地多于平原,高值区位于华北北部和内蒙古中部。对流性天气的减少与水汽和动力条件的减弱有关系。包含热力、动力和水汽条件的综合指数SWEAT无论在空间分布上,还是在30年的演变趋势上,均与对流性天气的分布和演变趋势表现出相似的特征。     

冬、夏季热带及副热带穿透性对流气候特征分析

文中利用热带测雨卫星 (TRMM) 搭载的测雨雷达 (PR) 1998~2007年的探测结果, 就热带及副热带地区穿透性对流的频次、条件降水强度及垂直廓线等特征进行了分析。研究结果表明: 深对流和穿透性对流都主要发生在热带辐合带 ( ITCZ)、南太平洋辐合带 (SPCZ)、亚洲季风区、20°N以南的非洲以及美洲等地区, 它们的空间分布具有明显的地域性和季节变化特征, 而且陆地深对流更容易发展成为穿透性对流, 但绝大部分地区的穿透性对流频次不超过0.2%。对穿透性对流条件降水强度的分析表明, 热带及副热带大部分地区的穿透性对流条件降水强度在10 mm/h以上, 且洋面的条件降水强度要比陆地大, 但由于其频次较小导致其对总降水的贡献并不大。尽管深对流和穿透性对流降水廓线的外形比较相似, 但相同的高度, 深对流的降水强度要比穿透性对流偏小, 而且这种差异随海陆和纬度的不同而有所区别。此外, 热带地区 (15°S~15°N) 冬、夏季深对流和穿透性对流降水廓线都只存在较小差异, 并没有显示出明显的季节变化。     

The Fluxes and Behaviour of Plumes Inferred from Measurements of Coherent Structures within Images of the Bulk Flow

This paper describes how measurements of the movement of identifiable features at the edge of a turbulent plume can be interpreted to determine the properties of the mean flow and consequently, using plume theory, can be used to make estimates of the fluxes of volume (mass), momentum, and buoyancy in a plume. This means that video recordings of smoke rising from a chimney or buoyant material from a source on the sea bed can be used to make accurate estimates of the source conditions for the plume. At best we can estimate the volume flux and buoyancy flux to within about 5% and 15% of the actual values, respectively. Although this is restricted to the case of a plume rising in a stationary and unstratified environment, we show that the results may be of practical use in other more complex situations. In addition, we demonstrate that large-scale (turbulent) coherent structures at the plume edge form on a scale approximately 40% of the local (mean) plume half-width and travel at almost 60% of the average local (mean) velocity in the plume.     

Fluxes and Gradients in the Convective Surface Layer and the Possible Role of Boundary-Layer Depth and Entrainment Flux

We study the relation between fluxes and gradients in the very unstable surface layer by comparing recent proposals in the literature with the well-known Businger–Dyer functions. The recent proposals include results from large-eddy simulation (LES), which account for entrainment effects and effects of the boundary-layer depth. A comparison of the relationships is made with experimental data. The LES-based gradient functions show the impact of entrainment in the surface layer, but the scatter in the field data is too large to confirm this. Therefore this result is preliminary and future tests against new observations are recommended. It appears that the Businger–Dyer relationship behaves differently to the alternatives, and that it deviates from observations for large stability.