NEW APPROACH TO DYNAMIC DATA MODELING AND ITS APPLICATION TO PRECIPITATION FORECASTING

By use of an observed data series a new dynamic data modeling has been proposed. Taking a nonlinear ordinary differential equation which is retrieved from the data series based on the bilateral difference principle as a dynamic kernel, with the self-memorization principle a forecast model can be established, which is called the DAta-based Mechanistic Self-memory Model (DAMSM). Some computing cases show that the forecasting accuracy of the DAMSM is quite satisfactory. An example of inter-annual precipitation prediction in summer in the Yangtze delta is given.