基于Lorenz-96模型的顺序数据同化方法比较研究

顺序数据同化方法在数据同化系统中得到了广泛的应用,其性能各有优缺。选择3种典型的顺序数据同化算法,即集合Kalman滤波,集合转换Kalman滤波和确定性Kalman滤波,使用经典的Lorenz-96模型进行敏感性实验,研究不同的关键参数变化,如集合数目变化、观测数变化、误差放大因子变化和定位半径变化时对同化效果的影响。实验表明：集合数目和观测数目的多少直接影响3种方法的同化效果;协方差放大因子和定位半径的选择会提高同化精度。综合比较,确定性集合Kalman滤波算法是一种具有较强鲁棒性的滤波算法,能够在集合数较小的情况下达到较好的同化效果。

Comparative Studies of Sequential Data Assimilation Methods based on Lorenz-96 Models

Sequential data assimilation methods have been widely applied in many data assimilation systems and each method has its own characteristics.In this paper,we introduce three typical assimilation methods,for example,Ensemble Kalman Filter,Ensemble Transform Kalman Filter and Deterministic Ensemble Kalman Filter.Based on the classical nonlinear model (eg,Lorenz-96 model),the numerical experiments were developed to test the sensitivity of all these methods.Different key parameters were investigated with respect to four aspects,which were the number of ensembles,the number of observations,the inflation factor and the localization radius.The results show:the number of ensembles and observations will directly influence the assimilation results; the optimal selection of the inflation factors and the localization radius will improve the accuracy of the assimilation.According to the final comparative studies,the deterministic EnKF is a method that has a better robust performance.It can achieve a better assimilation effect,especially in the occasion of the small ensemble numbers.