基于偏最小二乘回归的赤道电离异常区峰值TEC预测

PREDICTION OF PEAK IONOSPHERIC TOTAL ELECTRON CONTENT OF EQUATORIAL IONIZATION ANOMALY BASED ON PARTIAL LEAST SQUARE REGRESSION

基于IGS提供的120°E上空1999—2009年IONEX格式电离层电子浓度总含量(TEC)资料,使用偏最小二乘回归方法进行拟合训练,构建了赤道电离异常南、北驼峰峰值TEC的预测模型,并以2009年的TEC数据进行预测试验。结果表明,该预测模型能较准确地描述南北驼峰峰值TEC的变化特征,但对北驼峰峰值的预测效果要好于对南驼峰的预测;对南、北驼峰峰值TEC预测的平均绝对误差分别为4.441 0、2.915 1 TECU,平均相对误差分别为23.26%、10.78%,预测值与真实值的相关系数分别为0.712 2、0.785 9。

On the basis of the 1999—2009 IONEX format total electron content（TEC） data over 120°E from IGS,the method of Partial Least Square Regression is used to establish the prediction model of EIA（equatorial ionization anomaly） index.The TEC data from 1999 to 2008 were used to fit the prediction model and the all-year TEC data of 2009 was used to test the model.The results of the forecasting experimentations of the model show that this model can accurately describe the variation of EIA index Inc（TEC of north crest） and Isc（TEC of south crest）,but the prediction effect of Inc is better than the Isc.The mean absolute error of Isc and Inc is about 4.4410TECU and 2.915 1 TECU,the mean relative error of Isc and Inc is about 23.26% and 10.78%.The correlation coefficient between the actual values and predicted values of Isc and Inc is 0.712 2 and 0.785 9 respectively.