数值模式初值的敏感性程度对四维同化的影响

用著名的Ｌｏｒｅｎｚ系统作了共轭变分同化的数值试验。发现随着模式对初值敏感性程度的增加，用这种方法得到和模式相协调的初始场愈来愈困难，直到某些情况下的完全失败。这表明四维同化和可预报期限是联系在一起的。另一方面，随着方程不精确程度的增加，变分同化的效果愈来愈差，直到所做的预报无任何意义可言。如果在做变分同化的同时对模式参数也进行反演，就可使得基于Ｌｏｒｅｎｚ系统所做的预报效果大大提高。

THE EFFECTS OF THE MODEL SENSITIVITY TO INITIAL CONDITION UPON THE VARIATIONAL FOUR-DIENSIONAL ASSIMILATION -THE STUDY BASED ON LORENZ MODEL

n this paper, some numerical experiments of ad joint variational assimilation has been performed using the famous Lorenz model. With the increase of model sensitivity to initial condition. finding the initial values which harmonize with the Lorenz model become more and more difficult until the scheme is failure on some situation.This shows that the four-dimensional assimilation have some relation with predictability.On the other hand. with the increase of errors in the Lorenz model, the effects of variational assimilation become worse and worse until the forcasting has no meaning. If we Perform variational assimilation and retrieve the model parameters at the same time,The consequene of forcasting can improve greatly based on Lorenz model.