导航定位学报

针对台风电离层异常扰动探测参考背景值的精度直接影响探测结果的准确性，而传统方法确定背景值的精度较低的问题，提出一种台风登陆期间电离层TEC扰动探测方法：结合改进的自适应噪声完备集合经验模态分解（ICEEMDAN）和长短期记忆（LSTM）神经网络构建短时电离层总电子含量（TEC）背景值的确定方法；然后将其用于2023年台风“苏拉”登陆期间电离层TEC异常扰动探测。结果表明：1)ICEEMDAN-LSTM模型在地磁活跃期和平静期确定背景值的相对精度分别为91.18%和93.10%，较四分位距法、LSTM模型和CEEMDAN-LSTM模型可提高0.58～60.35个百分点；而均方根误差和偏差则减小数倍；ICEEMDAN-LSTM模型确定背景值的精度显著高于四分位距法和单一的LSTM模型。2）将提出的方法用于2023年台风“苏拉”登陆期间的电离层TEC扰动探测，发现台风登陆点周围的4个格网点电离层TEC在登陆当天和登陆后第1天分别表现出负异常和正负异常现象；结合日地空间环境，台风登陆当天的电离层扰动与“苏拉”存在一定联系，且距离台风路径越近，电离层TEC扰动越显著。

Aiming at the problem that the accuracy of reference background values for detecting anomalous disturbances in typhoon ionosphere directly affects the accuracy of detection results, while the accuracy of determining background values using traditional methods is relatively low, the paper proposed a detection method of ionospheric TEC disturbance during strong typhoon landfall: the improved complete ensemble empirical mode decomposition with adaptive noise(ICEEMDAN) and the long short-term memory(LSTM) neutral networks were combined to construct the method for determining the background values of short-term ionospheric total electron content(TEC); then the method was applied in the detection of ionospheric TEC abnormal disturbances during the landing of Typhoon SAOLA in 2023. Results showed that: 1) The relative accuracy of the ICEEMDANLSTM model in determining background values during geomagnetic active and geomagnetic quiet periods would be 91.18% and 93.10%, respectively, which could be increased by 0.58 to 60.35 percentage points compared with the interquartile range, LSTM model and CEEMDAN-LSTM model; meanwhile, compared to the interquartile range, LSTM model and CEEMDAN-LSTM model, the root mean square error and bias of background values by the ICEEMDAN-LSTM model would be reduced several times; results indicated that the accuracy of the ICEEMDAN-LSTM model in determining background values could be significantly higher than that of interquartile range and a single LSTM model. 2) The proposed method was applied to detect the ionospheric TEC disturbance during the landfall of Typhoon SAOLA in 2023, and it was found that four grid points around the landfall point would exhibit negative and positive/negative anomalies on the day of landfall and the first day after landfall, respectively; combining the spatial environment of solar activity and geomagnetic activity, results indicated that there would be a certain connection between ionospheric disturbances on the day of typhoon landfall and Typhoon SAOLA; moreover, the closer to the typhoon path, the more significant the disturbance of the ionosphere TEC.