Study On Short-term And Storm-time Forecasting The Critical Frequency Of The Ionospheric F2 Layer

Based on the analysis of the main controlling factors of the ionosphere and its long-term and short-term variation characteristics by using the f0F2 data of spanning nearly 40 years in China, this dissertation mainly focuses on establishing the ionospheric short-term, regional, and storm-time forecasting models at low and middle latitudes. This investigate is important to the ability of the sensation of space weather, radio environmental forecast, alert and effective estimation. The main results are listed as follows:1) A short-term f0F2 forecasting model using the index F10.7Based on solar radio flux F10.7 and hourly f0F2 values that span the period 1958-2006, a short-term predicting technique of the ionosphere f0F2 is introduced by using regression analysis of the observed values f0F2 and F10.7. Eight ionosonde stations used are Manzhouli, Changchun, Wulumuqi, Beijing, Lanzhou, Chongqing, Huangzhou and Haikou stations. The data of the different stations are used to test the forecasting performance respectively, and the results are compared by giving their root-mean-square errors according to different solar activity, season and local time. The results indicate that the method can forecast the f0F2 values effectively one day and three days ahead.2) Short-term forecasting models of the ionospheric f0F2 using the nonlinear networkDue to its non-linear dynamic process associated with the F2 region of the ionosphere with solar photon flux, geomagnetic activity and global thermospheric circulation, a short-term forecasting method of f0F2 at a single station is introduced by using artificial neural network (NN). Their forecasted errors are analyzed, which vary with different season and solar activity. By introducing the Kalman Filter and the Ensemble Kalman Filter, the forecasting values of the neural network were adjusted and optimized after considering the anterior forecast errors and the trend of f0F2 variations. The results show that forecasting performance of the optimizing model is superior to that of the purely neural network and IRI.By using Support Vector Machine (SVM), a different method for forecasting the ionosphere f0F2 at a single station one hour ahead, up to fives and 24 hours ahead is introduced, respectively. The results show that the predicted f0F2 has good agreement with observed data and the performance of the SVM model is superior to that of the autocorrelation and Persistence models. It reflects the potential application of this technique for forecasting f0F2.3) The ionospheric f0F2 regional forecastingTaken into account of the temporal and spatial correlativity of f0F2, a method for the ionospheric f0F2 regional forecasting, up to 5 hours ahead, is introduced by using the neural network. In addition, by introducing the ionospheric distance, latitude factor and longitude factor, the Kriging method has been proposed for the reconstruction of ionospheric foF2 in China region. Based on the measurements of the Chinese stations, the ionospheric reconstruction has been done, which give the estimates of the reconstruction accuracy in Chinese region.4) The storm-time ionospheric f0F2 predictionsAs the variation of f0F2 at storm-time depends obviously on latitude, season and local time, the responses of ionospheric f0F2 to the ionospheric storm at low and middle latitude are studied respectively.Using the geomagnetic indices of Dst and AE and data of ionospheric critical frequency f0F2, an empirical method in predicting storm-time ionospheric f0F2 at a single station an hour in advance is brought out by analyzing the impacts resulted from the equatorward propagation of the composition disturbance zone and the plasma drift induced by the penetration electric field, as well as the local time effect. Ten storms during 2004-2005 are employed to validate the method by studying their predicting errors. They turn out that the empirical model can capture the evolutions of the storm fairly well at low-latitude.As the correlation coefficient of f0F2 and the integrated geomagnetic index ap(τ) is better than that of foF2 and the instantaneous geomagnetic index ap. By analyzing the correlation coefficients of ap(τ) and f0F2, the best fit ofτis determined. Using the support vector machine (SVM), an empirical local ionospheric forecasting model has been developed to predict foF2 during disturbed geomagnetic conditions. The forecasted values foF2 are compared with that by the Persistence model and the STORM model during geomagnetic storm occurring from 2001 to 2006 at Lanzhou, which includes 67 storm events. As for the data sets used in this paper, the result shows the forecasting performance of SVM is better than the latter.