基于多源数据和深度学习方法的全球实时电离层模型构建

Global Real-time Ionospheric Model Construction Based on Multi-source Data and Deep Learning Method

基于CODE分析中心2017—2018年的全球电离层模型(Global Ionospheric Model,GIM)格网数据,结合地磁Dst指数和太阳活动指数F10.7,提出了预报电离层天顶方向的总电子含量VTEC值的长短期记忆网络LSTM模型,并分别在平静期和磁暴期下对(5°N,120°E)、(30°N,120°E)、(55°N,120°E)三个网格点进行电离层VTEC单步预测。该文将模型预测结果与电离层模型产品值进行对比,绝对误差平均小于2 TECU,并用多个指标对预测结果进行评价,分析了模型在不同纬度和不同电离层环境下的预报精度。

This paper proposes an LSTM network model for predicting ionospheric Vertical Total Electron Content,using the 2017-2018 the Global Ionospheric Model(GIM)grid data from CODE center,as well as the geomagnetic Dst index and solar activity index F10.7 obtained from GSFC/SPDF OMNIWeb.The ionospheric VTEC single-step prediction is performed at three regions(5°N,120°E),(30°N,120°E)and(55°N,120°E)under the calm period and geomagnetic storm period respectively.In this paper,the results of the model are compared with the ionospheric model,and the average absolute error of the model is less than 2 TECU.What’s more,the prediction results are evaluated by several indexes,and the performance of the model in different latitude and longitude and ionospheric environments is analyzed.