掩星技术一维变分同化算法的改进

Improvements of One-dimensional Variational Assimilation Algorithm in Occultation Technique

从LEO卫星观测到的掩星数据可以反演地球大气的气压、水汽、温度等廓线;它们对气象和大气科学研究是有价值的数据资源.掩星数据资料的同化技术可以有效地改进这些气象参数廓线,从而改进目前的数值天气预报精度.把掩星观测参数廓线用变分同化方法进入气象业务流程的最大困难是它的计算量太大.通过对一维变分同化价值函数进行改进和对迭代流程进行新的设计,避免了反复计算大维数矩阵的缺点,从而提高了变分同化的计算效率.在适用性讨论中,用背景场向量加上1个高斯白噪声作为真实值来检验CHAMP掩星资料变分同化的结果.

The vertical profiles of water vapor and temperature retrieved from GPS/LEO occultation data are potentially valuable data source for the meteorological and atmospheric science. The assimilation of GPS/LEO occultation data may effectively improve the vertical profiles and hence improve the accuracy of the current numerical weather prediction （NWP）. At present, the largest problem for applying variational assimilation retrieval technique to the current NWP model is the enormous computation. By the improvements of cost function of one-dimensional variational assimilation and the re-design of iteration process, the new algorithm adopted in this paper avoids repeated calculations of large dimensional matrix. Therefore, the computational efficiency of variational assimilation can be improved. In the numerical simulation discussion, the retrieval results from CHAMP occultation refractivity profile is tested when taking ECMWF analytical model added the Gaussian white noise as the true value.