Scholarlink

<p>Accurate ionospheric Total Electron Content (TEC) predictions are crucial for enhancing the precision and reliability of global navigation satellite systems (GNSS), especially during space weather events. This study presents a hybrid machine learning model for TEC prediction using multi-source solar, geomagnetic, and space weather variables. The TEC measurements from the IONOLAB of the HYDE station (Hyderabad, India) and the BAKO station (Indonesia) constitute the basis for our validation, which has already been processed through down sampling. In this case, we focus on several selected space weather events: a solar storm on November 3–4, 2021; an enhancement of the solar wind on May 5, 2025; and disturbances from a coronal hole on June 14 and September 11–12, 2025. The solar and geomagnetic parameters were taken from the OMNIWeb database; further information on solar flare activity is obtained from the GOES X-ray flux data. A hybrid prediction model using artificial neural networks (ANN), random forest, and gradient boosting is developed and optimized based on a fitness criterion-based evaluation strategy. The performances of the suggested model have been compared with that of eXtreme Gradient Boosting (XGBoost) and the empirical model IRI-2020 (International Reference Ionosphere) to evaluate its predictive ability at different levels of space weather. Model performance is determined in terms of RMSE (Root Mean Square Error), PCC (Pearson Correlation Coefficient), REL-E (Relative Error), and NSE (Nash–Sutcliffe Efficiency). Moreover, scatter plots are generated to analyse the consistency of predictions made by the model compared to observations, while the Friedman statistical test has been conducted to determine the significance of performance difference between the hybrid model, XGBoost, and IRI-2020 models. Apart from the performance of the TEC prediction, the corresponding range error for the actual TEC, predictions of the hybrid model, and the IRI-2020 model are evaluated in order to assess the impact on Global Positioning System (GPS) positioning accuracy. Across all events and locations, the hybrid model's performance exceeds the values from the IRI-2020 model. For each of the four events studied at Hyde, the hybrid RMSE values (6.1781, 4.2904, 9.8325, and 5.4572 TECU) were all lower than the corresponding IRI-2020 values (13.1224, 8.706, 10.6736, and 9.3336 TECU). At BAKO, all three RMSE values of the hybrid model (11.6017, 8.5181, and 10.0663 TECU) also were lower than corresponding IRI-2020 values (15.4468, 11.6054, and 21.5655 TECU). In summary, the hybrid model outperforms IRI–2020 by providing more accurate results in all comparisons. The evidence above demonstrates a superior capacity for modelling the ionosphere using a hybrid method created as part of this research; therefore, it may provide accurate estimates of TEC and improve GNSS performance during periods of degraded space weather.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Prediction of TEC/estimation of range error in GPS during solar storms, solar wind, and coronal hole activities using a hybrid model

  • 期刊论文

  • 发表日期：

    2026年6月16日

查看全文
  • R. Mukesh,
  • J. Kenisha,
  • N. Deeshika,
  • N. Karthika,
  • M. R. Lakshmika,
  • S. Kiruthiga,
  • Sarat C. Dass,
  • S. Karthick

摘要

Accurate ionospheric Total Electron Content (TEC) predictions are crucial for enhancing the precision and reliability of global navigation satellite systems (GNSS), especially during space weather events. This study presents a hybrid machine learning model for TEC prediction using multi-source solar, geomagnetic, and space weather variables. The TEC measurements from the IONOLAB of the HYDE station (Hyderabad, India) and the BAKO station (Indonesia) constitute the basis for our validation, which has already been processed through down sampling. In this case, we focus on several selected space weather events: a solar storm on November 3–4, 2021; an enhancement of the solar wind on May 5, 2025; and disturbances from a coronal hole on June 14 and September 11–12, 2025. The solar and geomagnetic parameters were taken from the OMNIWeb database; further information on solar flare activity is obtained from the GOES X-ray flux data. A hybrid prediction model using artificial neural networks (ANN), random forest, and gradient boosting is developed and optimized based on a fitness criterion-based evaluation strategy. The performances of the suggested model have been compared with that of eXtreme Gradient Boosting (XGBoost) and the empirical model IRI-2020 (International Reference Ionosphere) to evaluate its predictive ability at different levels of space weather. Model performance is determined in terms of RMSE (Root Mean Square Error), PCC (Pearson Correlation Coefficient), REL-E (Relative Error), and NSE (Nash–Sutcliffe Efficiency). Moreover, scatter plots are generated to analyse the consistency of predictions made by the model compared to observations, while the Friedman statistical test has been conducted to determine the significance of performance difference between the hybrid model, XGBoost, and IRI-2020 models. Apart from the performance of the TEC prediction, the corresponding range error for the actual TEC, predictions of the hybrid model, and the IRI-2020 model are evaluated in order to assess the impact on Global Positioning System (GPS) positioning accuracy. Across all events and locations, the hybrid model's performance exceeds the values from the IRI-2020 model. For each of the four events studied at Hyde, the hybrid RMSE values (6.1781, 4.2904, 9.8325, and 5.4572 TECU) were all lower than the corresponding IRI-2020 values (13.1224, 8.706, 10.6736, and 9.3336 TECU). At BAKO, all three RMSE values of the hybrid model (11.6017, 8.5181, and 10.0663 TECU) also were lower than corresponding IRI-2020 values (15.4468, 11.6054, and 21.5655 TECU). In summary, the hybrid model outperforms IRI–2020 by providing more accurate results in all comparisons. The evidence above demonstrates a superior capacity for modelling the ionosphere using a hybrid method created as part of this research; therefore, it may provide accurate estimates of TEC and improve GNSS performance during periods of degraded space weather.