卡尔曼滤波和自相关分析方法短期预报电离层f0F2的比较

电离层参数f₀F₂的预报是电离层研究的一个重要方向.本文选取2011年北京、长春、青岛和苏州四个常规观测站的f₀F₂数据，分别采用卡尔曼滤波、自相关分析法和国际参考电离层模型（IRI）对f₀F₂实施短期预报（提前1h预报），并通过与实际观测数据的对比，对三种方法预报f₀F₂的性能进行了比较.研究结果表明，在磁平静时期，采用卡尔曼滤波方法进行f₀F₂预报的均方误差为0.532MHz，相对误差8.11%，比国际电离层参考模型的均方误差和相对误差分别降低1.47MHz和14.58%；采用自相关分析方法进行f₀F₂预报的均方根误差为0.967MHz，相对误差11.46%，比国际电离层参考模型的均方误差和相对误差分别降低1.035MHz和11.23%.比较结果说明二者对f₀F₂短期预报的精度相对于国际电离层参考模型均有大幅提升.对磁暴期间三种方法的预测性能做了进一步比较，试验结果表明卡尔曼滤波短期预报性能总体上优于自相关分析法，这为f₀F₂短期预报的方法选择提供了一定指导. 

Comparison of Short-time Prediction of f0F2 Using Kalman Filter and Autocorrelation Methodormalsize

f₀F₂ forecast is a significant research aspect in ionospheric study, and much work has been done to improve its prediction performance. In this paper, f₀F₂ data from four ionospheric observation stations (Beijing, Changchun, Qingdao and Suzhou) in 2011 are used to predict f₀F₂ one hour in advance with the method of Kalman filter and autocorrelation analysis. Furthermore, comparisons are carried out between ionosonde observation, the values predicted by International Ionospheric Reference Model (IRI), and the estimated values of Kalman filter and autocorrelation method. The results are described as follows. For the method of Kalman filter, its Root Mean Square Error (RMSE) and Relative Error (RE) are 0.532MHz and 8.11% respectively. The RMSE and RE values are reduced by 1.035MHz and 14.58% compared with the corresponding values obtained by IRI. In terms of autocorrelation analysis, its RMSE and RE are 0.967MHz and 11.46%, and are reduced by 1.035MHz and 11.23% compared with the corresponding values obtained by IRI. It can be concluded that the prediction precisions of above-mentioned two methods have a great promotion compared with the IRI results. Moreover, further comparisons of these three methods are carried out during a geomagnetic storm. Experimental results indicate that Kalman filter method is better than autocorrelation analysis method and IRI model, which might provide suggestions for choosing a method for short-term prediction of f₀F₂.