With the time weighted accumulation index ap(τ), an empirical storm model of the relative deviation of f0F2 from its monthly median is established. It shows good accuracy only at strong geomagnetic disturbed condition in summer and equinoxes, but less accuracy result for winter. The wide used method in weather, ydrography, etc., Kalman filter, is used to update the coefficients of the empirical model and expected to make better improvement. The test with the ionosound data of hourly values from 1986 to 1995 at Changchun illustrates that the root-mean-squre error of the forecasted values in one hour advance is 0.76 MHz in winter, 0.68 MHz in summer and 0.61 MHz in equinoxes, respectively. The root-mean-squre error in different seasons are from 0.87 MHz to 1.43 MHz in the case of ap(τ)>100, i.e. the super strong geomagnetic disturbed condition. Simultaneously the method is compared with IRI2001 including the STORM model for several disturbed events, which shows the ability of updating the model coefficients and prospect of application.