Figure 4 - uploaded by John Bosco Habarulema
Content may be subject to copyright.
3: A South African map showing the four dual frequency GPS receiver stations and the location of the ionosonde used in this study.
Citations
... Layers in feed-forward networks are connected in one direction to allow data to flow from the input to the output layer and do not allow closed paths inside the network connections; examples include Perceptron and Adaline (Haykin, 2001). In a feed-back connection, some or all layers have connections that allow them to move backwards to the preceding layer (Habarulema, 2010;Kröse and van der Smagt, 1996) (Figure 2). ...
... Three layers of a feed forward network (Habarulema, 2010) On the other hand, in recurrent networks, there are no feedback connections. The network's dynamical features are crucial in this type of network. ...
The results of investigations from a complete analysis of ANN application on Total Electron Content (TEC) prediction are presented in this paper. TEC is important in defining the ionosphere and has many everyday applications, for example, satellite navigation, time delay and range error corrections for single frequency Global Positioning System (GPS) satellite signal receivers. The total electron content (TEC) in the ionosphere has been measured using GPS. GPS are not installed in every point on the earth to make global TEC measurements possible. As a result, it is crucial to have certain models that can aid to get data from places where there is not any in order to comprehend the global behavior of TEC. Neural Network (NN) models have been shown to accurately anticipate data patterns, including TEC. The capacity of neural networks to represent both linear and nonlinear relationships directly from the data being modeled is what makes them so powerful. The survey from literature reveals that, Levenberg-Marquardt algorithm is preferred and used mostly because of its speed and efficiency during learning process, and that ANN showed a good prediction of TEC compared to the IRI model. As a result, NNs are suitable for forecasting GPS TEC values at various locations if the model's input parameters are well specified.
... Neural networks can determine the solar activity derived from statistical and non-linear algorithms that evaluate the complex relationship between input and output to determine the data pattern. Several investigators [12] [13] have established that neural networks are effective for modelling ionospheric changes that depend mainly on sun activities.The solar parameters, like SSN have been providedisparities in the Solar cycle (SC) properties [14] [15]. Neural networks can be integrated with other techniquesto improve their accuracies [16]. ...
Nowadays, solar spectral irradiances are modelled by solar activity indices, which are used to identify the solar energy absorbed in the environment. This paper devises the Deep LSTM model for predicting solar activity using Sunspot Number (SSN) and Solar Radio Flux (SRF). The processing steps involved in the solar activity prediction are technical indicator extraction and solar activity prediction. In this paper, the solar indices are considered as an input of solar activity prediction, which is acquired from the solar cycle progression dataset. The technical indicators, like Simple Moving Average (SMA), Average True Range (ATR), Relative Strength Index (RSI), William’s %R, Stochastic %D, and Commodity Channel Index (CCI)are extracted for attaining a better prediction performance. In addition, the solar activity prediction is carried out using Deep LSTM based on SSN and SRF. The Deep LSTM is an effective deep learning technique that is widely utilized for prediction purposes since it has a better prediction ability. Moreover, the experimental result demonstrates that the devised Deep LSTM attained the minimum Mean Absolute Error (MAE), Mean Squared Error (MSE) and Root Mean Squared Error (RMSE) of 1.186, 2.869 and 1.693, correspondingly.
... Ionospheric storms, which alter depending on the solar action, the Earth's turning, and spatial, regular, monthly, and seasonal circumstances, have different impacts in the ionosphere [17,18]. e TEC values, which change over time and ought to be evaluated along with their location in space, are the principal factors for solar activity and ionosphere-magnetosphere-Sun interaction [3,[19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36]. is essay predicts the TEC values through an artificial neural network model (ANNm) [7,8,[37][38][39][40][41][42][43][44][45] over the superstorms of November 20, 2003 (Dst � -422 nT) and November 08, 2004 (Dst � -374 nT). e predicted TEC values are criss cross checked with the values attained from the IRI-2012 and 2016 [31,[46][47][48] model. ...
... Figure 5 We consider that our consequences merit the readers' evaluation. e relatively small RMSE (TECU) rates, with the margin of error of the prediction reliability of the TEC data, are exhibited in Figure 5 One may discover R correlation constant of some forecasts in Tulunay et al. [35] with 99.0%, Ansari et al. [19] with 91.6%, Inyurt and Sekertekin [42] with 88.0%, Razin and Voosoghi [45] e TEC values ANN estimation model outcomes of the super GSs seem comparable. e model not only exhibits the suitability of the IRI (2007)-estimated TEC outputs but also shows the reliability of the consequences. ...
Modeling and forecasting of Total Electron Content (TEC) values by an Artificial Neural Network model (ANNm) have high agreement on November 2003, 2004 superstorms. The work discusses Solar Wind Parameters (SWp) from OMNI (Operating Missions as a Node on the Internet) and TEC (TECU) data (International Reference Ionosphere) IRI-2012, IRI-2016 on November 20, 2003 (Dst = –422 nT) and on November 08, 2004 (Dst = –374 nT) Geomagnetic Storms (GSs). The paper commences with a 120-hour GS exhibition of SWp and proceeds with the correlation data of the variables, their hierarchical tracks, and inner dispersions. The ANNm with SWp as the input and TEC data as the output are introduced. The performance of the ANNm for 2003 and 2004 superstorms is adequate. The Correlation Coefficient (R) and Root Mean Square Error (RMSE) of the ANNm are 97.5%, 1.17 TECU (IRI-2012), and 97.9%, 1.09 TECU (IRI-2016) for the 2003 GS and 97.0%, 0.89 TECU (IRI-2012), and 98.0%, 1.61 TECU (IRI-2016) for 2004 GS. Parameters effect of the R constant of TEC data points out to the dynamic pressure (nPa), the magnetic field Bz component (nT), the flow speed (km/s), and the proton density (1/cm³). Besides, the absolute total error and the variance of the predicted TEC data for November 2003 and November 2004 GSs are 0.06 (0.30%) with 0.013 variance (IRI-2012), 0.09 (0.49%) with 0.016 variance (IRI-2016) for 2003 storm and 0.13 (0.73%) with 0.033 variance (IRI-2012), and 0.11 (1.06%) with 0.035 variance (IRI-2016) for 2004. It means that the paper models TEC data with considerable consistency over the SWp.
... In Africa, a good number of research have also been conducted to investigate ionospheric variation during geomagnetic storms (e.g., Adebesin, 2008;Adebiyi et al., 2014;Habarulema, 2010;Joshua et al., 2018;Matamba & Habarulema, 2018;Olawepo & Adeniyi, 2012;Rabiu et al., 2007;Uwamahoro et al., 2018; and many more). ...
... In the present work, a version of the storm-time model for the AfriTEC is developed by using artificial neural networks to model TEC measurements obtained across Africa during storm periods (|Dst| ≥ 50 nT and Kp ≥ 4). Using data for all (both quiet and storm) periods does allow the networks to learn storm-time variations, so long as storm indices are included as inputs (e.g., Habarulema, 2010;Okoh et al., 2016). However, Okoh et al. (2019) observed that including data from all periods does not allow the networks to adequately learn responses to TEC during storm conditions. ...
... For the first time, we have developed a storm-time TEC model over the African sector taking into account data from 2000 to 2018 from GPS observations. Previous modeling efforts were restricted by lack of observations (Habarulema, 2010;Habyarimana et al., 2020;Matamba & Habarulema, 2018;Okoh et al., 2016) especially in the Northern Hemispheric part of the African continent and ocean areas. To address this specific issue, we made use of the TEC derived from radio occultation (COSMIC) data, scaled it to GPS TEC based on the neural network approach which led to improved coverage. ...
This paper presents the development of a storm‐time total electron content (TEC) model over the African sector for the first time. The storm criterion used was |Dst| ≥ 50 nT and Kp ≥ 4. We have utilized Global Positioning System (GPS) observations from 2000 to 2018 from about 252 receivers over the African continent and surroundings within spatial coverage of 40°S–40°N latitude and 25°W–60°E longitude. To increase data coverage in areas devoid of ground‐based instrumentation including oceans, we used the available radio occultation Constellation Observing System for Meteorology, Ionosphere, and Climate (COSMIC) TEC from 2008 to 2018. The model is based on artificial neural networks which are used to learn the relationship between TEC and the corresponding physical/geophysical input parameters representing factors which influence ionospheric variability. An important result from this effort was the inclusion of the time history of the geomagnetic activity indicators dKpdtanddDstdt which improved TEC modeling by about 5% and 12% in middle and low latitudes, respectively. Overall, the model performs comparatively well with, and sometimes better than, the earlier single station modeling efforts even during quiet conditions. Given that this is a storm‐time model, this result is encouraging since it is challenging to model ionospheric parameters during geomagnetically disturbed conditions. Statistically, the average root‐mean‐square error (RMSE) between modeled and GPS TEC is 5.5 TECU (percentage error = 30.3%) and 5.0 TECU (percentage error = 30.4%) for the Southern and Northern Hemisphere midlatitudes respectively compared to 7.5 TECU (percentage error = 22.0%) in low latitudes.
... It drives the magnetosphere, thermosphere and ionosphere of the earth by its energy source [1]. The solar windmagnetosphere-ionosphere form a single system that is driven by the energy and momentum released by the solar wind to the ionosphere through the magnetosphere, a concept that is known as "solar wind -magnetosphere -ionosphere coupling" [2]. Here solar wind is referred to as a continuous flow of plasma which comes out of the sun to the Earth's atmosphere at a speed in a range of ~ 300 -700 km/s and consists of electrons and protons which are equal in proportion [3]. ...
... Computer neural networks (also referred to as artificial neural networks) have been demonstrated by several authors (e.g., Habarulema, 2010;Habarulema et al., 2009;Okoh et al., 2016) as very efficient tools for ionospheric modeling in parts of the continent. The strengths and advantages of neural networks derive from their ability to represent both linear and nonlinear relationships directly from the data being modeled (Baboo & Shereef, 2010). ...
... We also introduced the Julian number of days from the start of the years as inputs to enable the neural networks learn seasonal variations in the TEC data. To allow for a numerically continuous trend in the data, the diurnal and seasonal variations, represented by day of the year (DOY) and hour of the day (HH), were each split into two cyclical components as defined by equations (2) and (3) (Habarulema, 2010;Poole & McKinnell, 2000;Williscroft & Poole, 1996). ...
The first regional total electron content (TEC) model over the entire African region (known as AfriTEC model) using empirical observations is developed and presented. Artificial neural networks were used to train TEC observations obtained from Global Positioning System receivers, both on ground and onboard the Constellation Observing System for Meteorology, Ionosphere, and Climate satellites for the African region from years 2000 to 2017. The neural network training was implemented using inputs that enabled the networks to learn diurnal variations, seasonal variations, spatial variations, and variations that are connected with the level of solar activity, for quiet geomagnetic conditions (−20 nT ≤ Dst ≤ 20 nT). The effectiveness of three solar activity indices (sunspot number, solar radio flux at 10.7‐cm wavelength [F10.7], and solar ultraviolet [UV] flux at 1 AU) for the neural network trainings was tested. The F10.7 and UV were more effective, and the F10.7 was used as it gave the least errors on the validation data set used. Equatorial anomaly simulations show a reduced occurrence during the June solstice season. The distance of separation between the anomaly crests is typically in the range from about 11.5 ± 1.0° to 16.0 ± 1.0°. The separation is observed to widen as solar activity levels increase. During the December solstice, the anomaly region shifts southwards of the equinox locations; in year 2012, the trough shifted by about 1.5° and the southern crest shifted by over 2.5°.
... For this reason, most often, global models cannot be used in nowcasting and forecasting of the regional ionosphere with high precision in demand. Over the South African region, different researchers/groups have attempted to develop regional models that can more accurately define the local ionosphere (e.g., Habarulema, 2010;McKinnell, 2002;Okoh et al., 2010;Opperman, 2007;Ssessanga et al., 2014;Ssessanga et al., 2015). However, most of these models have until now focused on the simplistic 2-D horizontal structure of the ionosphere, in which important information about ionospheric vertical dynamical changes is lost. ...
... Moreover, because IRI is empirical in nature, it does not accurately describe ionospheric dynamics in areas (e.g., Southern Hemisphere/within the African sector) or during times that suffered data paucity during early stages of model development (e.g. Habarulema, 2010;Habarulema & Ssessanga, 2017;McKinnell & Poole, 2004;Okoh et al., 2012;Ssessanga et al., 2014). Figure 4b are distributed around the 1:1 correlation line (optimal line) especially in the left panel (quiet period), and the horizontal stratification observed in IRI VTEC values is no more. ...
One of the major research areas in the space weather community is the ability to understand, characterize, and model a time‐space variant ionosphere through which transionospheric signals propagate. In this paper a strong constraint four‐dimensional variational data assimilation (4D‐var) technique was used to more accurately estimate the South African regional ionosphere (bound latitude 20–35°S, longitude 20–40°E, and altitude 100–1,336 km). The altitude was capped to the JASON‐1 satellite orbital altitude for the purpose of eliminating the plasmasphere contribution hence reducing the computation expense. Background densities were obtained from an empirical internationally recognized ionosphere model (IRI‐2016) and propagated in time using a Gauss‐Markov filter. Ingested data were STECs (slant total electron content) obtained from the South African Global Navigation Satellite System receiver network (TrigNet). The vertically integrated electron content was validated using Global ionosphere Maps and JASON‐3 data over the continent and ocean areas, respectively. Further, vertical profiles after assimilation were compared with data from a network of ground‐based regional ionosondes Hermanus (34.25°S, 19.13°E), Grahamstown (33.3°S, 26.5°E), Louisvale (21.2°S, 28.5°E), and Madimbo (22.4°S, 30.9°E). Results show that assimilation of STEC data has a profound improvement on the estimation of both the horizontal and vertical structures during quiet and storm periods. Accuracy of the horizontal structure decreases from the continent toward the ocean area where GPS receivers are less abundant. Superiority of assimilating STEC is best pronounced during daytime especially when estimating maximum electron density of the F2 layer (NmF2), with a 60% root‐mean‐square error improvement over the background values.
... Many researchers have investigated the performance of a number of versions of NeQuick and International Reference Ionosphere (IRI) models in estimating or predicting TEC at different locations including Nigeria (MigoyaOrue´ et al. 2008;Coisson et al. 2008;Bidaine and Warnant, 2010;Habarulema, 2010;Adewale et al. 2011Adewale et al. , 2012Okoh et al. 2012Okoh et al. , 2015Okoh et al. , 2016Rabiu, et al. 2014) Okoh et al. (2012 observed that the IRI TEC values compared well with the GPS observed TEC values at Nsukka, Nigeria (6.87° N, 7.38°E; dip latitude 2.97) with 0.9 correlation coefficients and root-mean square deviations generally around 20-50% for diurnal comparisons. Adewale et al. (2011) reported that IRI-2007 (NeQuick option) gave a relatively poor TEC prediction at Lagos (6.5N, 3.4E; dip latitude 3.03S), between 0200 and 0600 h LT, the TEC percentage deviation having values greater than 50% during all seasons considered in year 2009. ...
... Arsitektur recurrent hampir sama dengan feedforward back propagation, namun ditambah dengan layer konteks untuk menampung hasil output dari hidden layer [17]. Feedback dapat menyebabkan proses iterasi akan jauh lebih cepat, sehingga membuat kecepatan update parameter dan konvergensinya menjadi lebih cepat [18]. ...
Bali is one of the favorite tourist attractions in Indonesia, where the number of foreign tourists visiting Bali is around 4 million over 2015 (Dispar Bali). The number of tourists visiting is spread in various regions and tourist attractions that are located in Bali. Although tourist visits to Bali can be said to be large, the visit was not evenly distributed, there were significant fluctuations in tourist visits. Forecasting or forecasting techniques can find out the pattern of tourist visits. Forecasting technique aims to predict the previous data pattern so that the next data pattern can be known. In this study using the technique of recurrent neural network in predicting the level of tourist visits. One of the techniques for a recurrent neural network (RNN) used in this study is Long Short-Term Memory (LSTM). This model is better than a simple RNN model. In this study predicting the level of tourist visits using the LSTM algorithm, the data used is data on tourist visits to one of the attractions in Bali. The results obtained using the LSTM model amounted to 15,962. The measured value is an error value, with the MAPE technique. The LSTM architecture used consists of 16 units of neuron units in the hidden layer, a learning rate of 0.01, windows size of 3, and the number of hidden layers is 1.
... Neural network forecasts are derived from nonlinear, statistical algorithms that determine and model complex relationships between inputs and outputs to find patterns in the data that can be extrapolated. Some researches (e.g., Habarulema, 2010;Okoh et al., 2016) have demonstrated that neural networks are efficient for modeling ionospheric variations that depend mostly on the Sun's activities. For parameters like SSN which have shown variations in the solar cycle properties (e.g., the value of SSN at the cycle peaks and the cycle durations), neural networks can be guided in techniques that combine other methods to increase their accuracies. ...
The Sun is the major driver of space weather events, and as a result, most applications requiring modeling/forecasting of space weather phenomena depend largely on the activities of Sun. Accurate modeling of solar activity parameters like the sunspot number (SSN) is therefore considered significant for the quantitative modeling of space weather phenomena. Sunspot number forecasts are applied in ionospheric models like the International Reference Ionosphere model and in several other projects requiring prediction of space weather phenomena. A method called Hybrid Regression-Neural Network that combines regression analysis and neural network learning is used for forecasting the SSN. Considering the geomagnetic Ap index during the end of the previous cycle (known as the precursor Ap index) as a reliable measurement, we predict the end of solar cycle 24 to be in March 2020 (±7 months), with monthly SSN 5.4 (±5.5). Using an estimated value of precursor Ap index as 5.6 nT for solar cycle 25, we predict the maximum SSN to be 122.1 (±18.2) in January 2025 (±6 months) and the minimum to be 6.0 (±5.5) in April 2031 (±5 months). We found from the model that on changing the assumed value of precursor Ap index (5.6 nT) by ±1 nT, the predicted peak of solar cycle 25 changes by about 11 sunspots for every 1-nT change in the assumed precursor Ap index.
