Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station
Global Positioning System (GPS) services could be improved through prediction of ionospheric delays for satellite-based radio signals. With respect to latitude, longitude, local time, season, solar cycle and geomagnetic activity the Total Electron Content (TEC) have significant variations in both ti...
| Main Authors: | , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
KeAi Communications Co., Ltd.
2020-05-01
|
| Series: | Geodesy and Geodynamics |
| Subjects: | |
| Online Access: | http://www.sciencedirect.com/science/article/pii/S1674984719300242 |
| _version_ | 1818622072210849792 |
|---|---|
| author | G. Sivavaraprasad V.S. Deepika D. SreenivasaRao M. Ravi Kumar M. Sridhar |
| author_facet | G. Sivavaraprasad V.S. Deepika D. SreenivasaRao M. Ravi Kumar M. Sridhar |
| author_sort | G. Sivavaraprasad |
| collection | DOAJ |
| description | Global Positioning System (GPS) services could be improved through prediction of ionospheric delays for satellite-based radio signals. With respect to latitude, longitude, local time, season, solar cycle and geomagnetic activity the Total Electron Content (TEC) have significant variations in both time and space. These temporal and spatial TEC variations driven by interplanetary space weather conditions such as solar and geomagnetic activities can degrade the communication and navigation links of GPS. Hence, in this paper, performance of TEC forecasting models based on Neural Networks (NN) have been evaluated to forecast (1-h ahead) ionospheric TEC over equatorial low latitude Bengaluru (12.97∘N,77.59∘E), Global Navigation Satellite System (GNSS) station, India. The VTEC data is collected for 2009–2016 (8 years) during current 24th solar cycle. The input space for the NN models comprise the solar Extreme UV flux, F10.7 proxy, a geomagnetic planetary A index (AP) index, sunspot number (SSN), disturbance storm time (DST) index, solar wind speed (Vsw), solar wind proton density (Np), Interplanetary Magnetic Field (IMF Bz). The performance of NN based TEC forecast models and International Reference Ionosphere, IRI-2016 global TEC model has evaluated during testing period, 2016. The NN based model driven by all the inputs, which is a NN unified model (NNunq) has shown better accuracy with Mean Absolute Error (MAE) of 3.15 TECU, Mean Square Deviation (MSD) of 16.8 and Mean Absolute Percentage Error (MAPE) of 19.8% and is 1–25% more accurate than the other NN based TEC forecast models (NN1, NN2 and NN3) and IRI-2016 model. NNunq model has less Root Mean Square Error (RMSE) value 3.8 TECU and highest goodness-of-fit (R2) with 0.85. The experimental results imply that NNunq/NN1 model forecasts ionospheric TEC accurately across equatorial low-latitude GNSS station and IRI-2016 model performance is necessarily improved as its forecast accuracy is limited to 69–70%. |
| first_indexed | 2024-12-16T18:19:21Z |
| format | Article |
| id | doaj.art-499cff57e12748d9b353ed20dc5e2a98 |
| institution | Directory Open Access Journal |
| issn | 1674-9847 |
| language | English |
| last_indexed | 2024-12-16T18:19:21Z |
| publishDate | 2020-05-01 |
| publisher | KeAi Communications Co., Ltd. |
| record_format | Article |
| series | Geodesy and Geodynamics |
| spelling | doaj.art-499cff57e12748d9b353ed20dc5e2a982022-12-21T22:21:35ZengKeAi Communications Co., Ltd.Geodesy and Geodynamics1674-98472020-05-01113192201Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS stationG. Sivavaraprasad0V.S. Deepika1D. SreenivasaRao2M. Ravi Kumar3M. Sridhar4Corresponding author.; Department of Electronics and Communication Engineering, Koneru Lakshamaiah Education Foundation, K L Deemed to be University, Vaddeswaram, Guntur District, 522502, Andhra Pradesh, IndiaDepartment of Electronics and Communication Engineering, Koneru Lakshamaiah Education Foundation, K L Deemed to be University, Vaddeswaram, Guntur District, 522502, Andhra Pradesh, IndiaDepartment of Electronics and Communication Engineering, Koneru Lakshamaiah Education Foundation, K L Deemed to be University, Vaddeswaram, Guntur District, 522502, Andhra Pradesh, IndiaDepartment of Electronics and Communication Engineering, Koneru Lakshamaiah Education Foundation, K L Deemed to be University, Vaddeswaram, Guntur District, 522502, Andhra Pradesh, IndiaDepartment of Electronics and Communication Engineering, Koneru Lakshamaiah Education Foundation, K L Deemed to be University, Vaddeswaram, Guntur District, 522502, Andhra Pradesh, IndiaGlobal Positioning System (GPS) services could be improved through prediction of ionospheric delays for satellite-based radio signals. With respect to latitude, longitude, local time, season, solar cycle and geomagnetic activity the Total Electron Content (TEC) have significant variations in both time and space. These temporal and spatial TEC variations driven by interplanetary space weather conditions such as solar and geomagnetic activities can degrade the communication and navigation links of GPS. Hence, in this paper, performance of TEC forecasting models based on Neural Networks (NN) have been evaluated to forecast (1-h ahead) ionospheric TEC over equatorial low latitude Bengaluru (12.97∘N,77.59∘E), Global Navigation Satellite System (GNSS) station, India. The VTEC data is collected for 2009–2016 (8 years) during current 24th solar cycle. The input space for the NN models comprise the solar Extreme UV flux, F10.7 proxy, a geomagnetic planetary A index (AP) index, sunspot number (SSN), disturbance storm time (DST) index, solar wind speed (Vsw), solar wind proton density (Np), Interplanetary Magnetic Field (IMF Bz). The performance of NN based TEC forecast models and International Reference Ionosphere, IRI-2016 global TEC model has evaluated during testing period, 2016. The NN based model driven by all the inputs, which is a NN unified model (NNunq) has shown better accuracy with Mean Absolute Error (MAE) of 3.15 TECU, Mean Square Deviation (MSD) of 16.8 and Mean Absolute Percentage Error (MAPE) of 19.8% and is 1–25% more accurate than the other NN based TEC forecast models (NN1, NN2 and NN3) and IRI-2016 model. NNunq model has less Root Mean Square Error (RMSE) value 3.8 TECU and highest goodness-of-fit (R2) with 0.85. The experimental results imply that NNunq/NN1 model forecasts ionospheric TEC accurately across equatorial low-latitude GNSS station and IRI-2016 model performance is necessarily improved as its forecast accuracy is limited to 69–70%.http://www.sciencedirect.com/science/article/pii/S1674984719300242Global Positioning System (GPS)Global navigation satellite systems (GNSS)Total electron content (TEC)International reference ionosphere (IRI)Neural networks |
| spellingShingle | G. Sivavaraprasad V.S. Deepika D. SreenivasaRao M. Ravi Kumar M. Sridhar Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station Geodesy and Geodynamics Global Positioning System (GPS) Global navigation satellite systems (GNSS) Total electron content (TEC) International reference ionosphere (IRI) Neural networks |
| title | Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station |
| title_full | Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station |
| title_fullStr | Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station |
| title_full_unstemmed | Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station |
| title_short | Performance evaluation of neural network TEC forecasting models over equatorial low-latitude Indian GNSS station |
| title_sort | performance evaluation of neural network tec forecasting models over equatorial low latitude indian gnss station |
| topic | Global Positioning System (GPS) Global navigation satellite systems (GNSS) Total electron content (TEC) International reference ionosphere (IRI) Neural networks |
| url | http://www.sciencedirect.com/science/article/pii/S1674984719300242 |
| work_keys_str_mv | AT gsivavaraprasad performanceevaluationofneuralnetworktecforecastingmodelsoverequatoriallowlatitudeindiangnssstation AT vsdeepika performanceevaluationofneuralnetworktecforecastingmodelsoverequatoriallowlatitudeindiangnssstation AT dsreenivasarao performanceevaluationofneuralnetworktecforecastingmodelsoverequatoriallowlatitudeindiangnssstation AT mravikumar performanceevaluationofneuralnetworktecforecastingmodelsoverequatoriallowlatitudeindiangnssstation AT msridhar performanceevaluationofneuralnetworktecforecastingmodelsoverequatoriallowlatitudeindiangnssstation |