Abstract: The short-term ionospheric forecast mainly denotes a prediction from hours to days in advance on time scale. This task needs a nonlinear recursion between the training data and the target one picked from the measurements, even by using complicated mathematic operations. Recently, an optimized arithmetic in data recursions named as Ensemble Kalman Filter (EnKF) has been widely used in temperature and rainfall predictions and even in ionospheric data assimilations. In this paper an optimizing method for short-term ionospheric f0F2 forecast was provided based on the Ensemble Kalman Filter technique. Firstly, the hourly f0F2 values with 1~24 hour in advance were forecasted by the neural network method. Then the forecasted values by the neural network were adjusted and optimized by introducing the Ensemble Kalman Filter after taking into account of the anterior forecast errors and the trend of f0F2 variations. The forecasted errors are binned with seasons and stations and compared with those of purely neural network and International Reference Ionosphere (IRI) to validate this method. The results show that the forecasting performance by the optimizing model is superior to that by the purely neural network and IRI. This indicated that the Ensemble Kalman Filter technique could be an efficient tool in ionospheric short-term forecast. Furthermore, this optimizing method can also be applied to the short-term forecasting of other ionospheric parameters.