【作者】 胡文东；

【导师】 沈桐立；

【作者基本信息】 南京气象学院 ， 气象学， 2003， 硕士

【摘要】 强降水预报是中尺度数值预报的难点之一。为了提高数值预报模式对位于中国西北干旱地区宁夏的强降水预报能力，本文研究了地球同步气象卫星红外资料与宁夏夏季逐时降水的关系。应用地球同步气象卫星红外资料，通过优选人工神经网络和最优拟合后的非线性回归这两种非线性方法，反演了宁夏自地表到300hPa不同层次的相对湿度。使用变分方法将反演值拓展到中尺度数值预报所需的格点上。在进入数值模式之前使用三种方案，对反演的相对湿度进行了质量控制。针对2003年7月21日发生于宁夏北部的一场突发性强对流天气，应用反演温度场，进行了数值模拟试验。试验表明，预报的降水场有了相当程度的改善，预报强降水落区与实况非常接近，spin up问题改进了25％。表明加入卫星反演的湿度场之后，数值预报模式可以对这次强对流过程进行更为合理的描述。主要特点为：高时间分辨率降水与气象卫星资料关系分析；卫星资料非线性反演；使用同化资料进行逐小时降水数值预报试验。具体内容如下： (1)经过地理定位、地图投影后，读取了地球同步气象卫星红外灰度，计算了导出物理量。 (2)对上述卫星资料与宁夏逐时降水关系的分析，表明：a)．宁夏夏季各类降水产生时，卫星资料散布范围相当广泛。对于较强逐时降水，卫星资料在不同地区其范围有不同程度的收缩，具有更好的预报指示意义。b)．宁夏各地逐时降水的极大值均产生于较暖的云团或云系中。强降水产生时云体相对较为稳定。c)．宁夏强降水时大气中水汽含量与南方相比显著偏低。宁夏特殊的地理位置和地形特征造成了水汽输送的阻碍，且冷空气与水汽难以配合，这也是宁夏降水不足、气候干旱的天气学根本原因之一。d)．水汽条件是宁夏夏季降水最为重要的因子，是降水的关键。e)．宁夏夏季各类降水产生时，地球同步气象卫星红外资料的散布较广，使用传统的线性回归方法建立它们之间的对应关系难以取得较为理想的效果。 (3)利用两种非线性方法，建立了气象卫星资料反演大气湿度场的数学模型：a)．通过网络训练，建立了卫星资料反演相对湿度的BP人工神经网络。b)．通过自变量与因变量的最优拟合，找到了地球同步气象卫星红外资料与地表相对湿度间的数学关系。进而完成了使用地球同步气象卫星红外资料及其导出量，对相对湿度的非线性回归。结果分析表明，经过最优拟合变换后，所建立的非线性回归模型与线性回归模型相比，效果有了较大的改善。 (4)变分与质量控制：a)．通过变分方法，利用超松驰叠代，对赫姆霍兹方程进行求解。将反演得到的气象站点相对湿度资料，拓展为模式网格点的湿度资料，从而与数值预报模式相匹配。b)．使用卫星资料之前，首先对卫星探测资料进行了质量控制。利用卫星反演资料的误差、宁夏强降水过程中螺旋度、位涡分布的关系，使用三种方案，对地球同步气象卫星资料反演出的大气湿度场进行了质量控制，以保证反演资料的可靠性，同时也在一定程度上保证了数值预报模式的运行稳定性，可以进入中尺度数值预报模型。c)．试验两种非线性方法反演相对湿度，经过变分及质量控制，对比分析结果表明，反演出的湿度场，能够更加细致地刻画出中尺度各层大气的湿度分布，有利于对中尺度天气系统进行更精细的描述。而且由于质量控制的作用，反演湿度场与原湿度场有较大的一致性，与其它要素分布保持动力与热力匹配，从而有利于模式运行的稳定性。 (5)针对2003年7月21日宁夏北部的一次突发性强降水过程，使用卫星资料反演湿度场，经过变分与质量控制后，利用枷5V3非静力平衡中尺度数值预报模式，进行了数值模拟。结果表明:a).降水预报量级有了明显的改善，具有预报指示意义。b).降水预报区域与实况相比具有较高的一致性。c) .sPin叩问题有所改进，但降水预报时间与实况相比依然滞后。 试验表明，使用本文的反演方法，利用地球同步气象卫星红外资料对各层大气相对湿度进行反演，并进行变分、质量控制，在中国西北宁夏这样一个干旱地区，可以提高中尺度数值预报模式对强降水的预报能力。更多还原

【Abstract】 A Study and Experiment on Severe Rainfallwith Numerical Weather Forecast in NingxiaUsing Geostationary Meteorological Satellite Infrared DataIn order to improve the ability of torrent rainfall forecast in Ningxia, which locates in northwest of China with an arid climate, the relationship between geostationary meteorological satellite infrared data and summer hourly precipitation in Ningxia was analyzed, the geostationary meteorological satellite infrared data were applied to retrieve the relative humidity at medium and low layers of atmosphere. Two different methods to retrieve humidity were developed using optimal artificial neural network and non-linear regression after optimal fitting transmission. The retrieved humidity was put into meso-scale nonhydrostatic numerical weather forecast model MM5V3 after variation analysis, and a severe convection event was analyzed with the retrieved relative humidity. The precipitation forecast was improved quantitively, the forecasted severe convection area was consistent with the observation, and the spin up was reduced as much as 25%. The experiment shows the implement of retrieved humidity enhanced the ability of numerical model to forecast severe convection in Ningxia.(1) Using infrared data of Geostationary Meteorological Satellite, the gray scale and other kinds of element were calculated or derived after map projecting and geographic calibrating.(2) Analysis on the relationship of the element above and hourly precipitation in recent summers in Ningxia was conducted. The summer hourly precipitation in Ningxia, especially those of the heavy rainfall events, had obvious characteristics in the satellite-derived data. a). The satellite data scatter in a pretty large area. b). The scattering areas of the data above contracted for the heavy precipitation of more than 5mm/h. c). The maximum rainfall events in all prefectures in Ningixa occurred in the relative warm and stable clouds, d). Vapor level is much lower compared with those in the heavy rainfall events in south China affected by geographical location and topographic characteristics, e). It is hard to satisfy the conditions of cold air and vapor together for torrent rainfall, in particularly the transportation of vapor in low atmosphere, f). It is one of the main reasons for the arid climate in Ningxia. The vapor, or humidity condition is the key to rainfall in Ningxia, especially to heavy rainfall event.( 3 ) Two kinds of non-linear models to retrieve the relative humidity with geostationary meteorological satellite infrared data were developed. 1.Optimal artificial neural network was set up for the relationship between infrared data of Geostationary Meteorological Satellite and relative humidity in the medium and low layers in atmosphere by error back propagation network training after the experiments with different nodes in each of two hidden layers. 2.non-linear regression wasconducted after optimal fitting to retrieve the relative humidity, a). Experiment showed that the network with 6 nodes in each hidden layer has the best-forecast ability. The BP network was developed by training, b). Successive regression on relative humidity and infrared data of Geostationary Meteorological Satellite was made with Non-linear transitions after optimal fittings taken from 48 functions. To avoid the over-fitting phenomenon, the stable and best-fitted models were adopted, c). The quality of retrieved relative humidity with infrared data of Geostationary Meteorological Satellite was greatly improved compared with the traditional method such as linear regression, d). The humidity at each layer in the low atmosphere in the stations of Ningxia can be retrieved with pretty high accuracy using the non-linear models above.(4) Variation and quality control: a). The retrieved relative humidity at different stations was expanded into the grids with variational method to match the meso-scale numerical weather forecast model MM5v3 for heavy rain forecast, b). The satellite data were quality controlled before re更多还原