【作者】 袁招洪；

【导师】 顾松山；

【作者基本信息】 南京气象学院 ， 大气物理学与大气环境， 2004， 博士

【摘要】 利用建立在长江三角洲地区GPS观测网中GPS资料，针对2002年梅雨期间影响长江三角洲地区的降水过程进行了GPS资料在MM5中尺度数值预报模式中的应用研究。研究表明： 1、GPS是一种连续监测大气水汽的有效手段。与常规探空观测相比，GPS测量的可降水量有很好的代表性。在相距2km时，两种测量手段测量可降水量和总延迟量的平均绝对偏差分别为2.13mm和1.28cm。总延迟的变化主要是由湿延迟的变化而引起，其中静力延迟变化呈现了明显的周期性，而湿延迟的变化与天气过程相关联。通过可降水量、总延迟和湿延迟的时、空变化可以分析天气系统演变。应用MM5模式预报结果计算大气平均温度可提高GPS可降水量的反演精度。 2、MM5模式初始场对静力延迟有较好的描述能力，但对湿延迟和可降水量的描述还存在比静力延迟更大的描述误差。MM5模式连续滚动预报(12h)基本能反映总延迟、静力延迟、湿延迟和可降水量的日际变化趋势，但对这些量逐时预报还存在误差。MM5模式对湿延迟的预报能力明显低于对静力延迟的预报能力，并且模式对总延迟的预报误差主要是由湿延迟预报误差产生的。MM5模式分辨率的提高有利于改善模式对静力延迟、湿延迟和可降水量的描述和预报。 MM5模式对可降水量的预报能力与积云参数化方案的选用有关。在MM5模式24h积分的前10～11h，选用KF、BM和Grell三种积云参数化方案模式对可降水量的预报偏差基本接近，对可降水量具有较好的预报能力，其后三种积云参数化方案对可降水量的预报偏差差异增大，模式积分至20～21h后对可降水量的预报能力明显减小。对模式粗网格，Grell方案总体上表现了最高的可降水量预报能力。对模式细网格，Grell和KF方案比BM方案表现了较高的可降水量预报能力。 3、用GPS测量的可降水量资料调整MM5模式湿度初始场可明显增强模式初始场描述水汽分布的能力，使其对可降水量的描述误差明显减小，有利于模式初始场更好地反映出水汽分布的局地特征，从而有效地控制模式积分初期对可降水量的预报误差。GPS可降水量资料的Nudging同化对可降水量预报改善较小，并且Nudging系数的增加对可降水量预报效果的改善程度影响不大。用GPs资料调整模式湿度初始场对预报6h累积降水量的改善主要是通过改善模式网格尺度降水预报来实现的，GPS可降水量资料Nudging同化主要通过改善次网格尺度降水预报来实现6h累计降水预报能力提高的。总体上，应用GPS可降水量资料调整模式湿度初始场对6h累积降水预报效果的改善优于连续Nudg ing同化对6h累积降水预报效果的改善。 4、背景误差对三维变分同化效果起着关键作用，使用NMC方法重新构建的背景误差更接近实际背景误差。模式变量(。、、、T、p和q)对应的误差水平尺度与NMC方法中预报误差的平均时间长度和模式提供1 Zh和24h预报所选用的积云参数化方案有直接的关系。不同模式变量的误差水平尺度存在差异，并且同一模式变量的误差水平尺度在模式不同的高度层也存在差异。 5、用三维变分技术能有效地同化GPS可降水量资料。GPS可降水量资料的同化不仅能调整MMS模式湿度初始场，而且使模式初始气压、温度和风场也能得到相应的调整.在模式积分的前6小时，Cressman客观分析试验对累计降水量预报优于GPs可降水量资料三维变分同化试验对累积降水量预报，但在模式积分的6一18小时GPS可降水量三维变分同化试验对累计降水的预报却优于cressman客观分析试验。总体上，GPs可降水量资料的同化有利于模式降水预报能力的提高。 6、GPS湿延迟量资料的三维变分同化比GPs可降水量资料同化更容易将GPS对大气的监测信息传递给MMS模式初始场，使得对模式初始湿度、温度、气压和风场的调整程度大于GPs可降水量资料同化对模式初始要素场的调整.尽管湿延迟量资料同化对6h累计降水预报的改善在不同的降水等级表现了不同效果，但总体上湿延迟量资料同化对6h累计降水预报的改善程度要高于可降水量资料变分同化。更多还原

【Abstract】 The GPS data from GPS networks in Yangtze delta is explored to investigate the improvement of MM5 simulation on rainfall event occurred over Meiyu period in 2002 with the aid of initial humidity fields reanalysis and assimulation. The results show:1, GPS is a valuable tool of observing water vapor in atmosphere consecutively. GPS observation which is about 2km far away from radiosonde site is comparable to radiosonde with a absolute bias of 2.13mm on precipitable water (PW) observation and 1.28cm on zenith total delay (ZTD). The variation of ZTD is mainly caused by the zenith wet delay (ZWD). zenith hydrostatic delay (ZHD) and ZWD from different GPS sites exhibit individual features that ZHD has a obvious periodic change and ZWD is related to weather. Accuracy of PW retrieved from GPS observation can be improved by using the average vertical temperature calculated from MM5 outputs.2. The reanalysis fields of MM5 can reveal the distribution of ZHD as a whole and has larger bias on ZWD and PW than ZTD. 12h-simulation of MM5 can show the daily change tendency of ZTD, ZHD, ZWD and PW with a bias in depicting their hourly change. MM5 has a ability of simulating ZWD on the whole with a bias larger than ZHD’s, which manipulates the bias of ZTD simulation. The increase of MM5 resolution can improve the ability of simulating and depicting ZHD, ZWD and PW distribution.KF, BM and Grell parametric schemes have a close ability of simulating PW at the beginning of 10-11h integration of MM5 model, and then the prediction bias of PW increases obviously after 20-21h integration. Grell scheme can simulate PW more accurate than others for coarse grid of MM5 and PW simulation of BM scheme is less accurate than others for fine grid of MM5.3, The initial humidity fields reanalyzed by using GPS PW can obviously improve its capability in revealing the water vapor distribution, which can result in restraining PW prediction bias during the earlier period of model integration so as to improve PW prediction. Nudging assimulation of PW can improve prediction slightly, and increasing nudging gain coefficient play a little role in improving prediction. It’salso found that the reanalysis influences the results of 6h accumulated precipitation through changing the non-convective precipitation prediction mainly, while results improved by the nudging assimulation are substantially associated with convective precipitation change. On the whole, better results are obtained by the reanalysis than by the nudging assimulation.4, Background errors (BE) play a key role in three dimensional variation assimulation (3dvar). The BE calculated by NMC technique reach the true BE more closely than that provided by MM5-3dvar system. The horizontal scalelength of model variables (u. v. T, p and q) is closely related to the average time of NMC technique and convective parameteric scheme of MM5 which affect the 12h and 24h outputs of MM5 integration. The scalelength of model variables is different for each other, which value is associated with the vertical height of the variable on the MM5 level.5, GPS PW data can be assimulated into MM5 by using 3dvar technique. After GPS PW data assimulation, the initial humidity field can be reanalyzed while the initial temperature, pressure and wind fields also being modified. Although MM5 with the Cressman objective analysis predicts the first 6h accumulated precipitation more accurately than 3dvar of GPS PW data, vice versa for 6-18 h accumulated precipitation. On the whole, GPS PW data assimulation will improve precipitation prediction.6, GPS ZWD assimulation ingests atmospheric information observed by GPS into MM5 initial fields more easily than GPS PW assimulation so that the initial fields of humidity, temperature, pressure and wind can be modified in a greater degree. In general, 6h accumulated rainfall prediction of MM5 is improved by 3dvar of GPS ZWD data more significantly than GPS PW data although the former improvement is related to the rainfall amount.更多还原