Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network
Unpredictable variability to total electron content (TEC) in the equatorial region and gaps in the TEC database due to Earth infrastructure failures creates a need to develop a TEC estimation model. NN-based approaches are found promising in modelling the ionospheric parameters because they have fle...
Saved in:
| id |
my-usim-ddms-13258 |
|---|---|
| record_format |
uketd_dc |
| institution |
Universiti Sains Islam Malaysia |
| collection |
USIM Institutional Repository |
| language |
en_US |
| topic |
Ionospheric forecasting -- Malaysia Meteorological instruments Electronic instruments Radio meteorology |
| spellingShingle |
Ionospheric forecasting -- Malaysia Meteorological instruments Electronic instruments Radio meteorology Vikneswary Jayapal Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| description |
Unpredictable variability to total electron content (TEC) in the equatorial region and gaps in the TEC database due to Earth infrastructure failures creates a need to develop a TEC estimation model. NN-based approaches are found promising in modelling the ionospheric parameters because they have flexible non-linear function mapping capability, which can estimate any non-linear measurable function with arbitrarily desired accuracy. This work presents, the development of neural network (NN) based model to estimate the ionospheric TEC for a single GPS receiver station (Lat. 1°52’N, Long. 103°48’E, Magnetic dip 14.3°) over Malaysia from February 2005 to December 2006.In NN, the TEC variability is modelled as a function of diurnal variation, seasonal variation, solar and magnetic proxies. The NN1 and NN2 models are used to interpolate and extrapolate the missing hourly TEC data, respectively. The NN’s interpolation capability could be seen more evidently than extrapolation, especially over longer periods of missing data. The NN2 model experienced difficulty in extrapolating the TEC values during the night time than the daytime. NN2 has relative correction (Crel) below than 85% when the missing TEC data are above 60%. For model validation, TEC values from NN2 are compared with the International Reference Ionosphere (IRI) model with respect to GPS TEC. The estimation results for the four seasons in 2006 show that, the NN2 model agrees well GPS TEC during solstice seasons. In terms of average root mean square error (RMSE), NN2 model shows an improvement about 39.9% compared to the IRI model over the four seasons. The predictability of TEC during negative ionospheric storm revealed that the IRI-2007 model tends to yield more accurate estimation results than the NN2 model with Crel is about ̴ 25% higher than NN2 model. Conversely, the NN2 model able to generalize the TEC trend more favourably than the IRI model during positive ionospheric storm effects with Crel is about ̴ 30 to 35% higher than the IRI model. Besides estimation, an ionospheric TEC forecasting model can be highly beneficial as an warning system in oder to lessen the adverse space weather and natural hazard impacts on human life and technologies. Based on the recovery GPS TEC data, a time series forecasting model is developed using a hybrid model that integrates the seasonal autoregressive integrated moving average (SARIMA) and neural networks to forecast the TEC values up to three days ahead. The forecast TEC values from the hybrid model are compared with the individual models SAROMA (FCAST-SARIMA) and NN (FCAST-NN) separately with respect to GPS TEC. Results show that all the three models forecast TEC values fairly well during quiet condition. Meanwhile, the performance of the individual models degraded during moderate contition and the forecasting errors increased for both the single models as the time horizon became larger. Besides, in term of average RMSE the percentage improvements of the hybrib model over SARIMA and NN during disturbed condition are ̴ 13.4% ̴ 26.1%, respectively. The estimation and forecasting models are used in tandem to provide a complete investigation on ionospheric TEC modelling at Parit Raja , Malaysia. |
| format |
Thesis |
| author |
Vikneswary Jayapal |
| author_facet |
Vikneswary Jayapal |
| author_sort |
Vikneswary Jayapal |
| title |
Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| title_short |
Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| title_full |
Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| title_fullStr |
Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| title_full_unstemmed |
Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network |
| title_sort |
estimation and forecasting of ionospheric total electron content based on neural network and hybrid seasonal autoregressive integrated moving average-neural network |
| granting_institution |
Universiti Sains Islam Malaysia |
| url |
https://oarep.usim.edu.my/bitstreams/9b89933e-8290-4e22-83f6-c971f4083488/download https://oarep.usim.edu.my/bitstreams/30a129c8-44bb-4412-8887-769909abd9e4/download https://oarep.usim.edu.my/bitstreams/d994310b-d96d-49da-9981-45a6473508b8/download https://oarep.usim.edu.my/bitstreams/1f6fc33e-123c-4ae6-be25-a4155cca9f24/download https://oarep.usim.edu.my/bitstreams/34aebbad-ab30-40e1-a06f-bc10806b5d1a/download https://oarep.usim.edu.my/bitstreams/f15200be-bc80-4b6b-9f23-24362197b721/download https://oarep.usim.edu.my/bitstreams/a933a2ad-540d-4387-8959-db7f39463110/download https://oarep.usim.edu.my/bitstreams/c07fa133-2bf4-481f-b0a2-70c29ac78282/download |
| _version_ |
1812444749626343424 |
| spelling |
my-usim-ddms-132582024-05-29T18:54:10Z Estimation and Forecasting of Ionospheric Total Electron Content Based on Neural Network and Hybrid Seasonal Autoregressive Integrated Moving Average-Neural Network Vikneswary Jayapal Unpredictable variability to total electron content (TEC) in the equatorial region and gaps in the TEC database due to Earth infrastructure failures creates a need to develop a TEC estimation model. NN-based approaches are found promising in modelling the ionospheric parameters because they have flexible non-linear function mapping capability, which can estimate any non-linear measurable function with arbitrarily desired accuracy. This work presents, the development of neural network (NN) based model to estimate the ionospheric TEC for a single GPS receiver station (Lat. 1°52’N, Long. 103°48’E, Magnetic dip 14.3°) over Malaysia from February 2005 to December 2006.In NN, the TEC variability is modelled as a function of diurnal variation, seasonal variation, solar and magnetic proxies. The NN1 and NN2 models are used to interpolate and extrapolate the missing hourly TEC data, respectively. The NN’s interpolation capability could be seen more evidently than extrapolation, especially over longer periods of missing data. The NN2 model experienced difficulty in extrapolating the TEC values during the night time than the daytime. NN2 has relative correction (Crel) below than 85% when the missing TEC data are above 60%. For model validation, TEC values from NN2 are compared with the International Reference Ionosphere (IRI) model with respect to GPS TEC. The estimation results for the four seasons in 2006 show that, the NN2 model agrees well GPS TEC during solstice seasons. In terms of average root mean square error (RMSE), NN2 model shows an improvement about 39.9% compared to the IRI model over the four seasons. The predictability of TEC during negative ionospheric storm revealed that the IRI-2007 model tends to yield more accurate estimation results than the NN2 model with Crel is about ̴ 25% higher than NN2 model. Conversely, the NN2 model able to generalize the TEC trend more favourably than the IRI model during positive ionospheric storm effects with Crel is about ̴ 30 to 35% higher than the IRI model. Besides estimation, an ionospheric TEC forecasting model can be highly beneficial as an warning system in oder to lessen the adverse space weather and natural hazard impacts on human life and technologies. Based on the recovery GPS TEC data, a time series forecasting model is developed using a hybrid model that integrates the seasonal autoregressive integrated moving average (SARIMA) and neural networks to forecast the TEC values up to three days ahead. The forecast TEC values from the hybrid model are compared with the individual models SAROMA (FCAST-SARIMA) and NN (FCAST-NN) separately with respect to GPS TEC. Results show that all the three models forecast TEC values fairly well during quiet condition. Meanwhile, the performance of the individual models degraded during moderate contition and the forecasting errors increased for both the single models as the time horizon became larger. Besides, in term of average RMSE the percentage improvements of the hybrib model over SARIMA and NN during disturbed condition are ̴ 13.4% ̴ 26.1%, respectively. The estimation and forecasting models are used in tandem to provide a complete investigation on ionospheric TEC modelling at Parit Raja , Malaysia. Universiti Sains Islam Malaysia 2016-07 Thesis en_US https://oarep.usim.edu.my/handle/123456789/13258 https://oarep.usim.edu.my/bitstreams/c205f32d-661f-4ad4-b600-b93c78a3c22d/download 8a4605be74aa9ea9d79846c1fba20a33 https://oarep.usim.edu.my/bitstreams/9b89933e-8290-4e22-83f6-c971f4083488/download 4fd10a7c31ca80ff2ab1c7479f16285c https://oarep.usim.edu.my/bitstreams/30a129c8-44bb-4412-8887-769909abd9e4/download 3e7baa9c1bc0e7f659fcee39b3c0b63d https://oarep.usim.edu.my/bitstreams/d994310b-d96d-49da-9981-45a6473508b8/download 6779e17e8752e99e23214480a7480d45 https://oarep.usim.edu.my/bitstreams/1f6fc33e-123c-4ae6-be25-a4155cca9f24/download 0543da739ac741da25ce1a2fd02ad285 https://oarep.usim.edu.my/bitstreams/34aebbad-ab30-40e1-a06f-bc10806b5d1a/download 7e83316c5019dbbd0019e9a9407d3613 https://oarep.usim.edu.my/bitstreams/f15200be-bc80-4b6b-9f23-24362197b721/download 44585283115a0f69724f855adf62c7c0 https://oarep.usim.edu.my/bitstreams/a933a2ad-540d-4387-8959-db7f39463110/download 806fa22463aafea53f14e7546aa4377b https://oarep.usim.edu.my/bitstreams/c07fa133-2bf4-481f-b0a2-70c29ac78282/download a35ff5b1105c7f6c523a7da6f4bbe7ae https://oarep.usim.edu.my/bitstreams/0d31a0ed-e748-4056-97e9-6dc3b39c02de/download ee3c3fb9c25a56dab6df5518ef4ff8d3 https://oarep.usim.edu.my/bitstreams/b8eff402-1d63-47f2-9636-293812b4a01a/download 63072055e0f7d8e52413128f8fd9eba9 https://oarep.usim.edu.my/bitstreams/4b176f52-1ea2-4b19-b16f-f68687b9141c/download 1d4a9479bb588ada7d5affa65843a44d https://oarep.usim.edu.my/bitstreams/c3f4c851-fe5b-46ae-a5ff-9b8204f95d20/download 3b26ffcafac2e4921ff4035cdd5d35ab https://oarep.usim.edu.my/bitstreams/98ab8a19-8977-488e-b30b-cda2f79d1604/download defd578aa7c4db412444e8fd2adf7a11 https://oarep.usim.edu.my/bitstreams/15e0184c-4e05-4ba9-b4cf-05c41b7e0011/download 68b329da9893e34099c7d8ad5cb9c940 https://oarep.usim.edu.my/bitstreams/9d5cbc29-e129-470d-83f6-a76e3bd58f7f/download acab011f91988593d75287f18e864ca7 https://oarep.usim.edu.my/bitstreams/f7d4539d-1437-4c9f-b534-7f5021a0caba/download ee2cc85936e86722c4948df716d15369 Ionospheric forecasting -- Malaysia Meteorological instruments Electronic instruments Radio meteorology |