INTRODUCING THE UPDATING OF MULTIREFERENCE STATES INTO  DYNAMICAL ANALOGUE PREDICTION

In order to effectively utilize the analogue information of historical observati ons, some theoretical and technical problems of the analogue correction of error s (ACE) are further explored in this paper based on the previous studies on anal oguedynamical models. Analyses show that the ACE is the redevelopment of metho d s based on the analoguedeviation equation and analoguecorrection equation of e rrors. On the basis of such idea, a new scheme of dynamical analogue prediction allowing for multireferencestate updating (MRSU) has been developed. In this s cheme, when model integration proceeds to the period of analogue updating (PAU), multiple reference states are reselected and optimal forecast vectors are est i mated from multiforecasts produced by the analoguedynamical model using the hy perplane approximation method. Such a “selectionestimation" procedure is peri od ically repeated until the entire forecast is completed. The MRSU experiments of Lorenz model have shown that the MRSU is effective in reducing forecast errors a nd raising forecast skill, thus preliminarily confirming the application prospec t of ACE theory. Furthermore, in the conceptual flow chart of the MRSU, introduc ing the ACE into the complicated numerical model is equivalent to realize the fo recast process of analoguedynamical model without necessity to reestablish t he model equation.