GPS折射角资料的变分同化试验

摘要

More and more new types of observational data provide many new opportunities for improving numericalweather forecasts. Among these, the GPS (Global Positioning System) bending angle is undoubtedly veryimportant. There are many advantages of the GPS bending angle, such as high resolution, availability inall weather conditions, and global data coverage. Thus it is very valuable to assimilate GPS bending angledata into numerical weather models. This paper introduces how to obtain and assimilate the GPS bendingangle. There are two methods of assimilation: the indirect method and direct method, and they are bothintroduced in this paper. During the minimizing process of variational assimilation, calculation efficiencyis very important and the optimal step size greatly influences the algorithm efficiency. Based on thecharacteristics of the minimizing algorithm, we obtain an adaptive method for calculating the optimizingstep suitable for all kinds of minimization algorithms through mathematical deduction. Finally, a numericalvariational assimilation experiment is performed using the GPS bending angle data of 11 October 1995.The numerical results indicate the validity of the variational assimilation method and the adaptive methodintroduced here.%越来越多的新型观测资料为数值天气预报水平的进一步提高提供了许多新的机会.在各种新型的观测资料中,GPS(全球定位卫星系统)折射角资料无疑是非常重要的.GPS折射角资料具有分辨率高、全天候探测、覆盖全球等优点,实现对GPS折射角资料的变分同化,将具有非常重要的意义.文中介绍了如何获得及同化GPS折射角资料的原理.对GPS折射角资料的变分同化可以分为两种:间接同化和直接同化,文中对这两种方法都作了具体介绍.在变分同化的最小化过程中,计算效率无疑是最重要的,而优化步长的计算又直接关系到算法效率的成败.根据最小化算法的特点,通过数学推导,得出一种适合于各种最小化算法的计算优化步长的自适应方法.最后,还利用1995年10月11日的GPS折射角资料进行了数值试验,结果表明了变分同化方法和计算优化步长方法的有效性.