Modelling storm-time TEC changes using linear and non-linear techniques
- Authors: Uwamahoro, Jean Claude
- Date: 2019
- Subjects: Magnetic storms , Astronomy -- Computer programs , Imaging systems in astronomy , Ionospheric storms , Electrons -- Measurement , Magnetosphere -- Observations
- Language: English
- Type: text , Thesis , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/92908 , vital:30762
- Description: Statistical models based on empirical orthogonal functions (EOF) analysis and non-linear regression analysis (NLRA) were developed for the purpose of estimating the ionospheric total electron content (TEC) during geomagnetic storms. The well-known least squares method (LSM) and Metropolis-Hastings algorithm (MHA) were used as optimization techniques to determine the unknown coefficients of the developed analytical expressions. Artificial Neural Networks (ANNs), the International Reference Ionosphere (IRI) model, and the Multi-Instrument Data Analysis System (MIDAS) tomographic inversion algorithm were also applied to storm-time TEC modelling/reconstruction for various latitudes of the African sector and surrounding areas. This work presents some of the first statistical modeling of the mid-latitude and low-latitude ionosphere during geomagnetic storms that includes solar, geomagnetic and neutral wind drivers.Development and validation of the empirical models were based on storm-time TEC data derived from the global positioning system (GPS) measurements over ground receivers within Africa and surrounding areas. The storm criterion applied was Dst 6 −50 nT and/or Kp > 4. The performance evaluation of MIDAS compared with ANNs to reconstruct storm-time TEC over the African low- and mid-latitude regions showed that MIDAS and ANNs provide comparable results. Their respective mean absolute error (MAE) values were 4.81 and 4.18 TECU. The ANN model was, however, found to perform 24.37 % better than MIDAS at estimating storm-time TEC for low latitudes, while MIDAS is 13.44 % more accurate than ANN for the mid-latitudes. When their performances are compared with the IRI model, both MIDAS and ANN model were found to provide more accurate storm-time TEC reconstructions for the African low- and mid-latitude regions. A comparative study of the performances of EOF, NLRA, ANN, and IRI models to estimate TEC during geomagnetic storm conditions over various latitudes showed that the ANN model is about 10 %, 26 %, and 58 % more accurate than EOF, NLRA, and IRI models, respectively, while EOF was found to perform 15 %, and 44 % better than NLRA and IRI, respectively. It was further found that the NLRA model is 25 % more accurate than the IRI model. We have also investigated for the first time, the role of meridional neutral winds (from the Horizontal Wind Model) to storm-time TEC modelling in the low latitude, northern and southern hemisphere mid-latitude regions of the African sector, based on ANN models. Statistics have shown that the inclusion of the meridional wind velocity in TEC modelling during geomagnetic storms leads to percentage improvements of about 5 % for the low latitude, 10 % and 5 % for the northern and southern hemisphere mid-latitude regions, respectively. High-latitude storm-induced winds and the inter-hemispheric blows of the meridional winds from summer to winter hemisphere have been suggested to be associated with these improvements.
- Full Text:
- Date Issued: 2019
- Authors: Uwamahoro, Jean Claude
- Date: 2019
- Subjects: Magnetic storms , Astronomy -- Computer programs , Imaging systems in astronomy , Ionospheric storms , Electrons -- Measurement , Magnetosphere -- Observations
- Language: English
- Type: text , Thesis , Doctoral , PhD
- Identifier: http://hdl.handle.net/10962/92908 , vital:30762
- Description: Statistical models based on empirical orthogonal functions (EOF) analysis and non-linear regression analysis (NLRA) were developed for the purpose of estimating the ionospheric total electron content (TEC) during geomagnetic storms. The well-known least squares method (LSM) and Metropolis-Hastings algorithm (MHA) were used as optimization techniques to determine the unknown coefficients of the developed analytical expressions. Artificial Neural Networks (ANNs), the International Reference Ionosphere (IRI) model, and the Multi-Instrument Data Analysis System (MIDAS) tomographic inversion algorithm were also applied to storm-time TEC modelling/reconstruction for various latitudes of the African sector and surrounding areas. This work presents some of the first statistical modeling of the mid-latitude and low-latitude ionosphere during geomagnetic storms that includes solar, geomagnetic and neutral wind drivers.Development and validation of the empirical models were based on storm-time TEC data derived from the global positioning system (GPS) measurements over ground receivers within Africa and surrounding areas. The storm criterion applied was Dst 6 −50 nT and/or Kp > 4. The performance evaluation of MIDAS compared with ANNs to reconstruct storm-time TEC over the African low- and mid-latitude regions showed that MIDAS and ANNs provide comparable results. Their respective mean absolute error (MAE) values were 4.81 and 4.18 TECU. The ANN model was, however, found to perform 24.37 % better than MIDAS at estimating storm-time TEC for low latitudes, while MIDAS is 13.44 % more accurate than ANN for the mid-latitudes. When their performances are compared with the IRI model, both MIDAS and ANN model were found to provide more accurate storm-time TEC reconstructions for the African low- and mid-latitude regions. A comparative study of the performances of EOF, NLRA, ANN, and IRI models to estimate TEC during geomagnetic storm conditions over various latitudes showed that the ANN model is about 10 %, 26 %, and 58 % more accurate than EOF, NLRA, and IRI models, respectively, while EOF was found to perform 15 %, and 44 % better than NLRA and IRI, respectively. It was further found that the NLRA model is 25 % more accurate than the IRI model. We have also investigated for the first time, the role of meridional neutral winds (from the Horizontal Wind Model) to storm-time TEC modelling in the low latitude, northern and southern hemisphere mid-latitude regions of the African sector, based on ANN models. Statistics have shown that the inclusion of the meridional wind velocity in TEC modelling during geomagnetic storms leads to percentage improvements of about 5 % for the low latitude, 10 % and 5 % for the northern and southern hemisphere mid-latitude regions, respectively. High-latitude storm-induced winds and the inter-hemispheric blows of the meridional winds from summer to winter hemisphere have been suggested to be associated with these improvements.
- Full Text:
- Date Issued: 2019
Single station TEC modelling during storm conditions
- Authors: Uwamahoro, Jean Claude
- Date: 2016
- Language: English
- Type: Thesis , Masters , MSc
- Identifier: http://hdl.handle.net/10962/3812 , vital:20545
- Description: It has been shown in ionospheric research that modelling total electron content (TEC) during storm conditions is a big challenge. In this study, mathematical equations were developed to estimate TEC over Sutherland (32.38oS, 20.81oE), during storm conditions, using the Empirical Orthogonal Function (EOF) analysis, combined with regression analysis. TEC was derived from GPS observations and a geomagnetic storm was defined for Dst ≤ -50 nT. The inputs for the model were chosen based on the factors that influence TEC variation, such as diurnal, seasonal, solar and geomagnetic activity variation, and these were represented by hour of the day, day number of the year, F10.7 and A index respectively. The EOF model was developed using GPS TEC data from 1999 to 2013 and tested on different storms. For the model validation (interpolation), three storms were chosen in 2000 (solar maximum period) and three others in 2006 (solar minimum period), while for extrapolation six storms including three in 2014 and three in 2015 were chosen. Before building the model, TEC values for the selected 2000 and 2006 storms were removed from the dataset used to construct the model in order to make the model validation independent on data. A comparison of the observed and modelled TEC showed that the EOF model works well for storms with non-significant ionospheric TEC response and storms that occurred during periods of low solar activity. High correlation coefficients between the observed and modelled TEC were obtained showing that the model covers most of the information contained in the observed TEC. Furthermore, it has been shown that the EOF model developed for a specific station may be used to estimate TEC over other locations within a latitudinal and longitudinal coverage of 8.7o and 10.6o respectively. This is an important result as it reduces the data dimensionality problem for computational purposes. It may therefore not be necessary for regional storm-time TEC modelling to compute TEC data for all the closest GPS receiver stations since most of the needed information can be extracted from measurements at one location.
- Full Text:
- Date Issued: 2016
- Authors: Uwamahoro, Jean Claude
- Date: 2016
- Language: English
- Type: Thesis , Masters , MSc
- Identifier: http://hdl.handle.net/10962/3812 , vital:20545
- Description: It has been shown in ionospheric research that modelling total electron content (TEC) during storm conditions is a big challenge. In this study, mathematical equations were developed to estimate TEC over Sutherland (32.38oS, 20.81oE), during storm conditions, using the Empirical Orthogonal Function (EOF) analysis, combined with regression analysis. TEC was derived from GPS observations and a geomagnetic storm was defined for Dst ≤ -50 nT. The inputs for the model were chosen based on the factors that influence TEC variation, such as diurnal, seasonal, solar and geomagnetic activity variation, and these were represented by hour of the day, day number of the year, F10.7 and A index respectively. The EOF model was developed using GPS TEC data from 1999 to 2013 and tested on different storms. For the model validation (interpolation), three storms were chosen in 2000 (solar maximum period) and three others in 2006 (solar minimum period), while for extrapolation six storms including three in 2014 and three in 2015 were chosen. Before building the model, TEC values for the selected 2000 and 2006 storms were removed from the dataset used to construct the model in order to make the model validation independent on data. A comparison of the observed and modelled TEC showed that the EOF model works well for storms with non-significant ionospheric TEC response and storms that occurred during periods of low solar activity. High correlation coefficients between the observed and modelled TEC were obtained showing that the model covers most of the information contained in the observed TEC. Furthermore, it has been shown that the EOF model developed for a specific station may be used to estimate TEC over other locations within a latitudinal and longitudinal coverage of 8.7o and 10.6o respectively. This is an important result as it reduces the data dimensionality problem for computational purposes. It may therefore not be necessary for regional storm-time TEC modelling to compute TEC data for all the closest GPS receiver stations since most of the needed information can be extracted from measurements at one location.
- Full Text:
- Date Issued: 2016
- «
- ‹
- 1
- ›
- »