Comparison of Ionospheric Anomalies over African Equatorial/Low-latitude Region with IRI-2016 Model Predictions during the Maximum Phase of Solar Cycle 24 Paul Amaechi, Elijah Oyeyemi, Andrew Akala, Mohamed Kaab, Waqar Younas, Zouhair Benkhaldoun, Majid Khan, Christine Mazaudier

To cite this version:

Paul Amaechi, Elijah Oyeyemi, Andrew Akala, Mohamed Kaab, Waqar Younas, et al.. Comparison of Ionospheric Anomalies over African Equatorial/Low-latitude Region with IRI-2016 Model Predic- tions during the Maximum Phase of Solar Cycle 24. Advances in Space Research, Elsevier, 2021, ￿10.1016/j.asr.2021.03.040￿. ￿hal-03213004￿

HAL Id: hal-03213004 https://hal.archives-ouvertes.fr/hal-03213004 Submitted on 30 Apr 2021

HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Advances in Sp ace Research Comparison of Ionospheric Anomalies over African Equatorial/Low-latitude Region with IRI-2016 Model Predictions during the Maximum Phase of Solar Cycle 24 --Manuscript Draft--

Manuscript Number: AISR-D-21-00077R1 Article Type: EM -Earth Magnetosphere/Upper Atmosphere Keywords: IRI-2016, Equatorial Ionization Anomaly, Hemispheric asymmetry, Winter anomaly, Semiannual anomaly Corresponding Author: Paul Obiakara Amaechi, Ph.D Chrisland University Lagos, NIGERIA First Author: Paul Obiakara Amaechi, Ph.D Order of Authors: Paul Obiakara Amaechi, Ph.D Elijah O. Oyeyemi, Ph.D Andrew Akala, Ph.D Mohamed Kaab, Ph.D Waqar Younas, MSc Zouhair Benkhaldoun, Ph.D Majid Khan, Ph.D Christine Amory Mazaudier, Ph.D Abstract: The capability of IRI-2016 in reproducing the hemispheric asymmetry, the winter and semiannual anomalies has been assessed over the equatorial ionization anomaly (EIA) during quiet periods of years 2013-2014. The EIA reconstructed using Total Electron Content (TEC) derived from Global Navigation Satellite System was compared with that computed using IRI-2016 along longitude 25o - 40oE. These were analyzed along with hemispheric changes in the neutral wind derived from the horizontal wind model and the TIMED GUVI columnar O/N2 data. IRI-2016 clearly captured the hemispheric asymmetry of the anomaly during all seasons albeit with some discrepancies in the magnitude and location of the crests. The winter anomaly in TEC which corresponded with greater O/N2 in the winter hemisphere was also predicted by IRI-2016 during December solstice. The model also captured the semiannual anomaly with stronger crests in the northern hemisphere. Furthermore, IRI-2016 reproduced the variation trend of the asymmetry index (A) in December solstice and equinox during noon. However, in June solstice the model failed to capture the winter anomaly and misrepresented the variation of A. This was linked with its inability to accurately predict the pattern of the neutral wind, the maximum height of the F2 layer and the changes in O/N2 in both hemispheres. The difference between variations of EUV and F10.7 fluxes was also a potential source of errors in IRI-2016. The results highlight the significance of the inclusion of wind data in IRI-2016 in order to enhance its performance over East Africa. Response to Reviewers:

Powered by Editorial Manager® and ProduXion Manager® from Aries Systems Corporation Highlights

Highlights

1. IRI-2016 captured the hemispheric asymmetry over the African EIA with discrepancies in crests’ magnitude/ location

2. The winter anomaly with greater O/N2 in the winter hemisphere was predicted by IRI-2016 during December solstice only

3. The semiannual anomaly was predicted with stronger northern crests while the asymmetry was missed in June solstice Manuscript Click here to view linked References

1 2 3 4 Comparison of Ionospheric Anomalies over African Equatorial/Low-latitude Region 5 with IRI-2016 Model Predictions during the Maximum Phase of Solar Cycle 24 6 7 *Paul O. Amaechi1, Elijah O. Oyeyemi2, Andrew O. Akala2,3, Mohamed Kaab4,5, Waqar 8 6 4 6 7,8 9 Younas , Zouhair Benkhaldoun , Majid Khan , Christine-Amory Mazaudier 10 11 1Department of Physical Sciences, Chrisland University, Abeokuta, Nigeria 12 2Department of Physics, University of Lagos, Akoka, Yaba, Lagos, Nigeria 13 3Maritime Institute, University of Lagos, Akoka, Yaba, Lagos, Nigeria 14 4 15 Oukaimeden Observatory, LPHEA, FSSM, Cadi Ayyad University, Marrakech, 16 Morrocco 17 5National School of Applied Sciences of Beni Mellal, Sultan Moulay Sliman University, 18 Beni Mellal, Morocco 19 6 20 Department of Physics, Quaid-i-Azam University, Islamabad 45320, Pakistan 7 21 LPP, CNRS/Ecole Polytechnique/Sorbonne Université/Université Paris- 22 Sud/Observatoire de Paris, 75006 Paris, France 23 8 T/ICT4D Abdus Salam ICTP, Italy 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 *Corresponding author: Paul O. Amaechi (email: [email protected]) 56 57 Tel: +2348032050324 58 59 60 61 62 1 63 64 65 1 2 3 4 Abstract 5 6 7 The capability of IRI-2016 in reproducing the hemispheric asymmetry, the winter and 8 semiannual anomalies has been assessed over the equatorial ionization anomaly (EIA) 9 10 during quiet periods of years 2013-2014. The EIA reconstructed using Total Electron 11 Content (TEC) derived from Global Navigation Satellite System was compared with that 12 computed using IRI-2016 along longitude 25o - 40oE. These were analyzed along with 13 14 hemispheric changes in the neutral wind derived from the horizontal wind model and the 15 TIMED GUVI columnar O/N2 data. IRI-2016 clearly captured the hemispheric 16 17 asymmetry of the anomaly during all seasons albeit with some discrepancies in the 18 magnitude and location of the crests. The winter anomaly in TEC which corresponded 19 with greater O/N in the winter hemisphere was also predicted by IRI-2016 during 20 2 21 December solstice. The model also captured the semiannual anomaly with stronger crests 22 in the northern hemisphere. Furthermore, IRI-2016 reproduced the variation trend of the 23 24 asymmetry index (A) in December solstice and equinox during noon. However, in June 25 solstice the model failed to capture the winter anomaly and misrepresented the variation 26 of A. This was linked with its inability to accurately predict the pattern of the neutral 27 28 wind, the maximum height of the F2 layer and the changes in O/N2 in both hemispheres. 29 The difference between variations of EUV and F10.7 fluxes was also a potential source 30 31 of errors in IRI-2016. The results highlight the significance of the inclusion of wind data 32 in IRI-2016 in order to enhance its performance over East Africa. 33 34 35 Keywords: IRI-2016, Equatorial Ionization Anomaly, Hemispheric asymmetry, Winter 36 37 anomaly, Semiannual anomaly. 38 39 40 41 1. Introduction 42 43 The ionosphere is the ionized part of the atmosphere which extends from about 60 44 45 to 1000 km (Hargreaves, 1995) and comprises of free ions and electrons in sufficient 46 47 number as to affecting radio signals. Ionospheric anomalies which were first reported in 48 the works of earlier scientists (Appleton, 1938; Berkner et al., 1936) have been the 49 50 subject of various investigations over the years. Ionospheric anomaly originally referred 51 52 to departures from the solar controlled behavior in which the critical frequency foF2 53 54 varies regularly with the solar zenith angle ( ) as it does in the well-known Chapman 55 56 layer (Rishbeth, 1998). According to Rishbeth휒 (1998) the term anomaly originally meant 57 58 ‘‘any departure from solar controlled behavior in which the critical frequency foF2 varies 59 regularly with the solar zenith angle ( ) as it does in the well-known Chapman layer 60 ’’. 61 휒 62 2 63 64 65 1 2 3 4 Salient ionospheric anomalies are the seasonal anomaly otherwise known as the winter 5 6 anomaly, the annual anomaly or non-seasonal anomaly and the semiannual anomaly as 7 8 well as the equatorial ionization anomaly (EIA). 9 10 The winter anomaly is characterized by greater peak electron density of the F2 11 12 layer (NmF2) in the winter hemisphere than its conjugate summer hemisphere during 13 14 solstices (Torr and Torr, 1973). For the annual anomaly, the NmF2 is significantly greater 15 16 in December than June solstice globally (Gowtam and Tulasi Ram, 2017) while for the 17 18 semiannual anomaly it is higher in equinox than solstice (Rishbeth, 1998; Yasyukevich et 19 20 al., 2018). The EIA also known as the Appleton Anomaly (Appleton, 1946), is an 21 essentially feature of the low-latitude ionosphere. It is discernable by the reduction in 22 23 ionization at the magnetic equator (the trough) as against two peaks on both sides of it at 24 25 about ±15o magnetic latitude (the crests) (Namba and Maeda, 1939). 26 27 28 Several mechanisms have been put forth to explain the ionospheric anomalies. It 29 30 has been suggested that the winter anomaly is caused by change in neutral composition, 31 32 driven by global thermospheric circulation, which is greater in winter than in summer 33 (Burns et al., 2014; Rishbeth and Setty, 1961; Zou et al., 2000). The annual anomaly is 34 35 driven by the higher December solstice solar flux which is responsible for greater 36 37 dissociation of molecular oxygen than June solstice (Yonezawa and Arima, 1959) as well 38 39 as by changes in the Sun-Earth distance, difference between the geographic and magnetic 40 41 equator and the tilt of the geomagnetic dipole (Zeng et al., 2008). The semiannual 42 43 anomaly is easily explained in terms of stronger (weaker) vertical drifts and ionization 44 due to the fact that photoionization is more effective in equinox (solstice) (Fejer et al., 45 46 1995) as well as greater O/N2 ratio in equinox as compared to solstice (Rishbeth et al., 47 48 2000; Zhao et al., 2007). The EIA is brought about by the fountain effect (Mitra, 1946) 49 50 which is the upward transport of plasma by an ExB force above the magnetic equator, 51 52 and it subsequent diffusion along the magnetic field lines under the effects of gravity and 53 54 pressure gradient (Karia et al., 2018; Sterling et al., 1969). 55 56 The EIA is however, far from being symmetric (Oyedokun et al., 2020). This is because 57 58 of the presence of transequatorial neutral winds which affect the equatorial plasma 59 60 diffusion resulting in significant asymmetry through the transport of plasma from one 61 62 3 63 64 65 1 2 3 4 hemisphere to the other (e.g., Heelis and Hanson 1980; Khadka et al., 2018). This in 5 6 conjunction with the highest Total Electron Content (TEC) values at the EIA crests are 7 8 sources of additional errors in critical navigation and positioning applications. The delay 9 suffered by a radio wave propagating through the ionosphere is proportional to TEC and 10 11 could be significantly high at the EIA crests. This in turn will translate to increase 12 13 positioning errors, especially in single-frequency Global position System (GPS) 14 15 receivers. More so, the increased in ionospheric electrodynamical and chemical processes 16 17 at the EIA will undoubtedly affect the performance of models, especially over least 18 19 studied region such as Africa. This poses serious concern to the modeling community and 20 underscores the significance of our study. 21 22 23 The International Reference Ionosphere (IRI) is an empirical model aimed at 24 25 establishing an international standard for specifying the climatology of ionospheric 26 27 parameters (Bilitza and Reinisch, 2008). It is widely accepted as a standard model to 28 29 describe ionospheric features in various applications such as communication, aviation, 30 31 navigation. IRI is constantly being ameliorated using a wide range of ground and space 32 data, Global Navigation Satellite System (GNSS) networks, ionosondes and satellite data. 33 34 This has led to the existence of several versions of the model of which IRI-2016 is the 35 36 latest (Bilitza et al., 2017). 37 38 39 IRI has been extensively utilized in various longitudinal sectors to study the 40 41 dynamics of the ionosphere. However, studies carried out over Africa have shown that 42 43 even the latest version of IRI (IRI-2016) is susceptible to prediction errors (Endeshaw, 44 2020; Melaku and Tsidu, 2019; Mengistu et al., 2019; Mengistu and Melaku, 2020). 45 46 Unfortunately, due to the limited number of simultaneous studies in both hemispheres, 47 48 there is little information on the performance of IRI-2016 over the African EIA. 49 50 Particularly, there are no studies assessing the capability of the model to reproduce the 51 52 interhemispheric asymmetry as well as the winter and semiannual anomalies. In addition, 53 54 the global success of IRI and its application in critical GNSS systems require that its level 55 of accuracy over all regions of the globe be ascertained. This is more pertinent over the 56 57 Africa longitude which has long been poorly represented in the existing global database 58 59 (Bilitza et al., 2014). 60 61 62 4 63 64 65 1 2 3 4 The aim of this work is to investigate ionospheric anomalies over the African 5 6 equatorial/low-latitude region and then examine how IRI-2016 reproduces the observed 7 8 hemispheric asymmetry in the EIA, winter and semiannual anomalies during the 9 maximum phase of solar cycle 24 (SC-24). This falls well within the scope of the 10 11 Committee on Space Research / International Union of Radio Science (COSPAR /URSI) 12 13 IRI Working Group session (at the 43rd COSPAR Scientific Assembly), which is geared 14 15 towards improving the description of hemispheric differences in IRI. 16 17 18 2. Data and method of analysis 19 20 2.1 Data sets 21 22 23 In this study archived daily GNSS observation data in the Receiver Independent 24 25 Exchange (RINEX) format with 30 seconds resolution have been utilized. The 26 corresponding TEC data predicted by the recent version of the IRI-2016 model was also 27 28 employed. The geographic location of the GNSS stations is shown in Fig. 1. We have 29 30 also exploited the O/N2 ratio data obtained from the Global Ultraviolet Imager (GUVI) 31 32 on board the Thermosphere, Ionosphere, Mesosphere, Energetics and Dynamics 33 34 (TIMED) satellite. Furthermore, the direction and magnitude of the meridional winds 35 were derived from the recent version of the Horizontal Wind Model (HWM14) (Drob et 36 37 al., 2015). Finally, the daily Extreme Ultraviolet (EUV) flux (26–34 nm) and solar flux 38 39 (F10.7) measurements as well as the planetary K index (Kp) were utilized. A summary 40 41 of the sources of data and links to access them is given in Table 1. 42 ‐ 43 44 All these data sets were obtained during quiet days of years 2013 and 2014. Both 45 46 years fall within the maximum phase of SC-24 with mean observed annual solar flux of 47 122.76 and 146.54 solar flux unit (s.f.u), respectively. A day was deemed quiet if its Kp 48 49 was lesser or equal to 3 (Amaechi et al., 2020; Fejer et al., 2008). 50 51 52 2.2 Method of analysis 53 54 55 2.2.1 TEC derived from GNSS measurements (GNSS-TEC) 56 57 58 For the estimation of GNSS-TEC, we first subjected the GNSS observables to 59 quality check using the Translating Editing and Quality Checking (TEQC) software 60 61 62 5 63 64 65 1 2 3 4 (Estey and Meertens, 1999). Then, we analyzed them following the procedure of Seemala 5 6 and Delay (2010). This entailed estimating slant TEC (STEC) which is also known as 7 8 relative TEC using the so called carrier phase to code leveling technique described in 9 Hansen et al. (2000). The algorithm of Blewitt (Blewitt, 1990) was used to detect and 10 11 correct eventual cycle slips in phase measurements. 12 13 14 The STEC obtained was thereafter, calibrated by removing satellite and receiver 15 16 biases (Sardon et al., 1994). This calibrated STEC was then converted to vertical TEC 17 18 (VTEC) using the thin shell ionospheric model described by Mannucci et al. (1993). The 19 o 20 effect of multipath was minimized by using an elevation cut-off mask of 30 . The EIA 21 was reconstructed using hourly averages of VTEC values at ionospheric pierce point 22 23 (IPP) using GNSS station within 25o - 40oE with a latitudinal extent of ± 30o geographic 24 25 latitude. The temporal and spatial resolution of the EIA maps obtained was 1 hour x 1 26 27 degree. We recall that there is a displacement of about 9o between the magnetic equator 28 o o 29 and geographic equator within longitude 25 - 40 E (Amaechi et al., 2020). 30 31 32 2.2.2 TEC derived from IRI model (IRI-TEC) 33 34 35 Hourly values of the TEC were derived from IRI-2016 with the NeQuick option 36 for the topside electron density profile (Ne) and the latest bottomside thickness option 37 38 (ABT-2009) selected. In addition, the Consultative Committee International Radio 39 40 (CCIR) for F-peak option was used since it is the recommended option for the continent 41 42 (Tariku, 2020). Also, the F-peak storm model option was set to off because we are 43 44 considering geomagnetically quiet days. Finally, the required modeled TEC were 45 obtained by integrating the Ne profile from an altitude of 90 - 2000 km (upper boundary 46 47 for IRI model). The IRI-TEC so derived was utilized to reconstruct the EIA using the 48 49 same spatio-temporal resolution with GNSS-TEC. Because of the difference in the upper 50 51 boundary of integration between IRI (2000 km) and GNSS (20,200 km), we excluded the 52 53 contribution of the plasmaspheric electron contribution (PEC) to GNSS-TEC data. 54 55 Details about the estimation technique can be found in Akala et al. (2015), Cherniak et al. 56 (2012) and Karia et al. (2015; 2018). 57 58 59 The interhemispheric asymmetry was quantified using the asymmetry index (A) 60 61 computed with both GNSS-TEC and IRI-TEC data. A is a good measure of the 62 6 63 64 65 1 2 3 4 asymmetry of the anomaly (Paul and DasGupta, 2010). It was computed in the local noon 5 6 (1200 – 1700 LT) and post sunset (1900 – 2200 LT) using equation 1. Both time intervals 7 8 correspond to periods when the anomaly is fully developed in Africa (Amaechi et al., 9 2018). 10 11 12 13 Eq. (1) 14 A1− A2 15 whereA = A1S and A2 are the areas computed under the TEC- latitude plot on both 16 17 sides of the magnetic equator up to the crest in both hemispheres and S is the strength of 18 19 the anomaly defined by and A is the asymmetry index. 20 A1+ A2 21 S = 2 22 In the computation of A1 and A2, we first calculated the areas of small polygons 23 24 whose vertices were defined by the TEC and the corresponding magnetic latitude. Then, 25 26 we integrated the estimated areas of the polygons to get total area under the curve. 27 Positive (negative) value of A is an indication of stronger winter crest in in December 28 29 (June) solstice. 30 31 32 2.2.3 Thermospheric composition (O/N2) ratio and meridional winds (U) 33 34 To take into account thermospheric composition changes, we computed the O/N2 35 36 ratio during quiet days of year 2013 and 2014 for longitude 25o - 40oE. This data set is 37 38 available in IDLsave format at guvitimed.jhuapl.edu and comprises of O/N2 39 40 measurements as function of latitude, longitude and UT time as the spectrometer takes 41 42 the measurements at a specific location. We have averaged daily measurements 43 corresponding to the Northern hemisphere (0:30oN) and Southern hemisphere (0:30oS), 44 45 o o separately along longitude 25 - 40 E. Thereafter seasonal averages of O/N2 were 46 47 calculated by taking arithmetic mean of all available data points belonging to respective 48 49 geographic location. 50 51 52 The meridional wind velocity was calculated using the technique described in 53 54 Amaechi et al. (2020). The estimation technique which takes into account the 55 56 recommendation of Chartier et al. (2015) is described in details in Kaab et al. (2017). It 57 entails estimating the airglow-weighted meridional winds, U, between 200 and 300 km 58 59 using equation 2. 60 61 62 7 63 64 65 1 2 3 4 = Eq. (2) 5 ∑ uzaz 6 az 7 where U and are the meridional winds from HWM14 and the calculated redline 8 9 volume emissionz z rate at altitude z, respectively (Link and Cogger, 1988). 10 u a 11 12 The seasonal profiles of wind over longitude 25o - 45oE with a latitudinal extent 13 o 14 of ±30 were reconstructed using data from all the stations in Fig. 1. This was done by 15 16 averaging the wind data binned over 1 hours and 1 degree. 17 18 The seasonal variations of all parameters of interest (TEC, O/N , U, EUV and 19 2 20 F10.7 fluxes) were examined. We took monthly average of November, December and 21 22 January (May, June and July) to represent December (June) solstice and the average of 23 24 February, March and April (August, September and October) for March (September) 25 26 equinox. 27 28 29 3. Results 30 31 3.1 The winter anomaly 32 33 34 In Fig. 2 the seasonal variations of the EIA reconstructed using GNSS-TEC and 35 36 IRI-TEC are shown during solstices of years 2013 and 2014. In this figure, the summer 37 38 and winter hemispheres are respectively indicated for June/December solstice. It could be 39 seen that the anomaly reconstructed using GNSS-TEC showed a conspicuous 40 41 hemispheric asymmetry with stronger crest in the winter than summer hemisphere in 42 43 December solstice. This was reproduced by IRI-2016 with the model showing higher 44 45 crests magnitude with farther location in December solstice 2013. In 2014 nevertheless, 46 47 the predicted crests magnitude were lower than the observed one while the crests location 48 49 remained farther. In June solstices the observed and predicted anomalies were both 50 asymmetric. However, the anomaly reconstructed using GNSS (IRI) measurements 51 52 showed a stronger crest in the winter (summer) hemisphere. In addition, the predicted 53 54 crests magnitude was conspicuously higher than the observed one. 55 56 57 The variations of the asymmetry index (A) during noon and post sunset are shown 58 59 in Fig. 3(a-b). In the noon of both solstices, IRI-2016 underestimated the magnitude of A 60 61 (Fig. 3a). The underestimation was nevertheless pronounced in June solstice. The model 62 8 63 64 65 1 2 3 4 further rightly (wrongly) predicted the sign of the asymmetry in December (June) 5 6 solstice. The same trend of prediction of the sign of A was observed in the post sunset 7 8 (Fig. 3b). However, the model underestimated the magnitude of A in June solstice while 9 there was no significant difference in the observed and predicted magnitude and sign of A 10 11 in December solstice. Fig. 3(c-d) depicts a typical example of variation of the peak height 12 13 of the F2 layer (hmF2) obtained from IRI-2016 during solstices. In reconstructing this 14 15 figure, we took the average of the hmF2 in both hemispheres during the local noon (1200 16 17 – 1700 LT) and post sunset (1900 – 2200 LT). Generally, it could be seen that hmF2 was 18 19 higher in the northern than southern hemisphere. Furthermore, it was higher in December 20 solstice than June solstice. The values obtained in the northern (southern) hemisphere 21 22 during noon were 379 (326 km) and 401 (391 km) for June and December solstice, 23 24 respectively (Fig. 3c). For the post sunset, they were 369 (327 km) and 452 (421 km) 25 26 (Fig. 3d). 27 28 29 3.2 The semiannual anomaly 30 31 32 Fig. 4 presents the variations of GNSS-TEC and IRI-TEC in equinox and solstices 33 in the years 2013 and 2014. The black solid lines indicate the location of the magnetic 34 35 equator. We have combined TEC data for both equinoxes in order to clearly observe the 36 37 semiannual anomaly. In addition to the hemispheric asymmetry observed during 38 39 solstices, the equinoctial crests had different amplitudes with the stronger crests and 40 41 larger latitudinal extent in the northern hemisphere. IRI-2016 clearly reproduced this 42 43 asymmetry but showed relatively stronger (weaker) crests magnitude in the northern 44 (southern) hemisphere than GNSS especially, in 2013. In 2014 nevertheless, IRI-2016 45 46 predicted weaker magnitude of the crests in both hemispheres. During all seasons, the 47 48 model also clearly reproduced the solar activity dependence of TEC over the EIA with 49 50 stronger (weaker) crests magnitude in 2014 (2013). The respective observed annual solar 51 52 flux was 122.72 (146.54 sfu). 53 54 55 Fig. 5 shows the variations of A during equinox and solstices of years 2013 and 56 2014. It could be seen that the smallest magnitude of A computed using GNSS-TEC was 57 58 observed in equinox during noon (Fig. 5a) and post sunset (Fig. 5b). Contrastingly, the 59 60 model showed that the smallest magnitude of A occurred in June solstice during noon. 61 62 9 63 64 65 1 2 3 4 During post sunset however, it rightly predicted its occurrence in equinox. Also, the 5 6 model rightly (wrongly) captured the sign of A and overestimated (underestimated) its 7 8 magnitude during noon (post sunset). We finally noted a significant reduction in the 9 observed and predicted magnitudes of A during equinox. We recall that the model 10 11 predicted the asymmetry of the EIA but failed to capture its direction in June solstice 12 13 (Fig. 5a-b). This was related to IRI’s inability to correctly represent variations of hmF2 in 14 15 both hemispheres during June solstice (Fig. 3c-d) as shown earlier. This also underpinned 16 17 the inability of IRI-2016 to represent the direction of interhemispheric winds as pointed 18 19 earlier. 20 21 3.3 Change in thermospheric composition 22 23 24 The seasonal variations in O/N2 ratio in both hemispheres from 2013 to 2014 are 25 26 shown in Fig. 6. From this figure, O/N2 was higher in the southern/winter hemisphere 27 28 than northern /summer hemisphere in June solstice. In December solstice, the reverse was 29 30 the case with O/N2 being higher in the northern/winter than southern/summer 31 32 hemisphere. Also, O/N2 was the highest in both hemispheres in equinox than solstices. It 33 was equally found that March equinox had the higher O/N ratio than September equinox. 34 2 35 However, the difference in the equinoctial O/N2 ratio between both hemispheres was not 36 37 significantly different. 38 39 40 3.4 Thermospheric neutral wind 41 42 43 The seasonal variations of thermospheric neutral wind velocity in the years 2013 44 45 and 2014 are shown in Fig. 7. The broken horizontal lines represent the magnetic equator. 46 47 Positive (negative) value of the velocity indicates a northward (southward) wind while 48 the level of the contour on the color bar gives the wind’s magnitude. From this figure 49 50 December solstice was essentially marked by northward wind with higher velocities in 51 52 the northern than southern hemisphere. A southward wind was however observed from 53 o o 54 05:00 – 09:00 UT within latitude 10 S to 30 S (Fig. 7a). In the post sunset to post 55 56 midnight period, the northward wind reached its highest velocity of about 120 m/s away 57 from both crests. In the early morning of June solstice, the neutral wind was mostly 58 59 northward with a velocity lesser than 20 m/s. From 05:00 – 09:00 UT, it was 60 61 predominantly northward (southward) in the northern (southern) hemisphere with a 62 10 63 64 65 1 2 3 4 higher velocity in the southern hemisphere. It turned southward from 10:00 – 20:00 UT 5 6 with a velocity higher than 40 m/s in the southern hemisphere. During equinoxes, the 7 8 wind velocity was generally lesser than 60 m/s. In March equinox, from 0500 – 1100 UT, 9 the wind was northward (southward) in the northern (southern) hemisphere. It later 10 11 turned northward (southward) within 15oS - 30oN (15oS – 30oS) from 12:00 – 17:00 UT. 12 13 Thereafter, it was directed northward within 8oS to 30oN and 10oS – 30oS. In September 14 o o 15 equinox, it was mostly northward in the northern hemisphere while from 5 S - 30 S, it 16 17 turned southward within 04:00 – 16:30 UT. From 17:00 – 24:00 UT, the wind was 18 19 northward with a velocity that decreased from southern hemisphere towards the equator 20 and increased towards the northern hemisphere thereafter. From Fig. 7b, the seasonal 21 22 pattern of neutral wind in 2014 was similar to that in 2013 (Fig. 7a) except for the fact 23 24 that its velocity was relatively smaller in 2014 than 2013. 25 26 27 Fig. 8 presents the seasonal correlation between EUV (26 – 34 nm) and F10.7 solar 28 2 29 fluxes in the years 2013 and 2014. We computed the coefficient of determination (R ) 30 31 which indicates the proportion of the variation in the dependent variable (F10.7 flux data) 32 explained by the independent variables (EUV flux data) in the linear regression model. 33 34 The MATLAB code used for the calculation of R2 is based on comparing the variability 35 36 of the estimation errors with that of the original values. From the computed values of R2, 37 38 it could be seen that there was a good agreement between EUV and F10.7 with about 84 39 40 – 90% of the earlier data set being related linearly to the earlier in December solstice 41 42 2013 and June solstice 2014. On the other hand, there was a slight reduction in the 43 relation between both quantities in March equinox 2013 to December solstice 2014. 44 45 During these seasons, it was found that about 62 – 74% of EUV flux data was still related 46 47 to solar flux data. Nevertheless, in March and September equinox 2014, there was a weak 48 49 relation between both solar parameters with only 27 – 39% of data being linearly related. 50 51 52 53 4. Discussion 54 55 In this section, we have discussed the mechanisms that modulated the hemispheric 56 57 asymmetry of the EIA, winter and semiannual anomalies. These included the meridional 58 59 wind, change in thermospheric composition and the fountain effect. Thereafter, the 60 capability of IRI-2016 in reproducing these anomalies was assessed while plausible 61 62 11 63 64 65 1 2 3 4 reasons for the model’s eventual misrepresentation of observed hemispheric features 5 6 were proffered. 7 8 9 4.1 The winter and semiannual anomalies 10 11 o 12 It was observed that over longitude 25 - 40 E, the transequatorial neutral winds 13 blew from the summer to the winter hemisphere during solstices (Fig. 7) while the EIA 14 15 was stronger in the winter than summer hemisphere (Fig. 2). Ordinarily, summer-to- 16 17 winter winds drag ionization along the field lines, uplifting (lowering) the F-layer in the 18 19 summer (winter) hemisphere (Balan et al. 1995; Kwak et al., 2019; Titheridge 1995). 20 21 However, the summer-to-winter winds affect the fountain effect by reducing (enhancing) 22 the diffusion of plasma into the EIA in the summer (winter) hemisphere (Gowtam and 23 24 Tulasi Ram, 2017; Tulasi Ram et al. 2009). As such, the pattern of winds which drove the 25 26 observed asymmetry of the anomaly (Fig. 3a-b) was favorable to plasma diffusion in the 27 28 winter hemisphere and contributed to the enhancement of ionization over the crest in that 29 30 hemisphere. 31 32 33 On the other hand, the O/N2 ratio which was found to be higher in the winter than 34 35 summer hemisphere (Fig. 6) also played a crucial role in driving the winter anomaly over 36 Africa. The larger (smaller) O/N2 in the winter (summer) hemisphere results from the 37 38 summer-to-winter difference in solar radiation and the consequent the global 39 40 thermospheric circulation (Burns et al., 2014; Qian et al., 2016; Qian and Yue, 2017; 41 42 Yasyukevich et al., 2018). The changes in O/N2 ratio in turn lead to the changes in the 43 44 production/loss rate of electrons with greater changes in the winter hemisphere resulting 45 in higher ionization (Millward et al., 1996). The greater density ratio of O/N in winter 46 2 47 than in summer was thus, responsible for the winter anomaly as pointed out by Rishbeth 48 49 (1998). 50 51 52 The observed semiannual anomaly with higher TEC over the crests of the 53 54 anomaly in both hemispheres during equinox (Fig. 4) followed the seasonal pattern of the 55 56 change in thermospheric composition (higher O/N2 in equinox than solstices) (Fig. 6). In 57 addition, the semiannual anomaly was dominant in the northern hemisphere. This was in 58 59 line with the higher O/N2 observed in that hemisphere. This feature was also obvious in 60 61 parameter A which showed that the anomaly was deflected northward in the noon during 62 12 63 64 65 1 2 3 4 equinox (Fig. 5). These results are in line with those of Rishbeth et al. (2000) who found 5 6 that the O/N2 ratio was greater in equinox than solstice at low-latitude. Zhao et al. (2007) 7 8 had suggested that the O/N2 variation contributes to some good extent to the semiannual 9 annual anomaly of TEC possibly because of the dynamics associated with thermospheric 10 11 neutral wind and electric fields at low-latitude. Results of the seasonal variation of the 12 13 asymmetry index (A) revealed that the EIA was less asymmetric in equinox than solstice. 14 15 This was consistent with the reduction in the neutral wind velocity in equinox (Fig. 7) 16 17 and re-emphasized the role of neutral wind in driving seasonal variation over the EIA. 18 19 20 The semiannual anomaly of TEC at the EIA could also be explained in terms of 21 solar EUV and change in electrodynamics. It is well known that the intensity of solar 22 23 radiation which depends on the Sun’s elevation is responsible for ionization in the 24 25 ionosphere. During equinox, the Sun is overhead at the equator and more ionization is 26 27 produced. Also, the vertical plasma drift velocity (Vz) which results from the interaction 28 29 of the eastward electric field and horizontal magnetic field play a significant role in the 30 31 transport and redistribution of plasma at low-latitude. Vz is stronger in equinox than 32 solstice over the African longitude (Amaechi et al., 2018; Fejer et al., 1995). The 33 34 implication is that the fountain effect will be stronger in equinox and more plasma will be 35 36 lifted up and diffused towards the crests. Consequently, the anomaly will appear stronger 37 38 and well-developed in equinox than solstice. 39 40 41 4.2 IRI prediction of the winter and semiannual anomalies 42 43 44 IRI-2016 reproduced the hemispheric asymmetry as well as the winter anomaly in 45 December solstice (Fig. 2). The model also predicted the direction of the asymmetry as it 46 47 rightly showed a stronger anomaly in the northern (winter) hemisphere. However, the 48 49 model underestimated the magnitude of the asymmetry index especially during noon 50 51 (Fig. 3a-b). This underestimation could have been due to the model’s inability to capture 52 53 the magnitude of the neutral wind in December solstice. In June solstices, IRI-2016 still 54 55 reproduced the hemispheric asymmetry but failed to capture the winter anomaly. The 56 model further misrepresented the direction of the asymmetry (showing a stronger 57 58 anomaly in the summer instead of the winter hemisphere) and underestimated its 59 60 magnitude. To have an insight into the model’s misrepresentation of the winter anomaly 61 62 13 63 64 65 1 2 3 4 and the asymmetry of the anomaly in June solstice, we examined the hmF2 values 5 6 obtained from IRI-2016 for both hemispheres in June and December solstices (Fig. 3c-d). 7 8 It was found that the model did not show higher hmF2 in the southern (winter) 9 hemisphere in June solstice. We thus, inferred that the misrepresentation of hmF2 in both 10 11 hemispheres during June solstice could be a reason for the model’s failure to capture the 12 13 winter anomaly. Given the difference in the pattern of wind during both solstices (Fig. 7), 14 15 we further deduce that IRI-2016 could not capture the direction of the interhemispheric 16 17 wind and by extension the change in thermospheric composition in both hemispheres. 18 19 This finding therefore calls for the inclusion of neutral wind data from the African 20 longitude into IRI-2016 in order to ameliorate its predictive capability. Our result is in 21 22 line with Kumar (2020) who reported greater departure of IRI-2016 from observed in-situ 23 24 measurements of CHAMP and GRACE during June solstice over longitude 79oE. In the 25 26 same vein, Karia et al. (2018) showed that IRI-2016 had more discrepancies in the 27 o 28 southern hemisphere of the EIA over longitude 73 E in June solstice of 2012. 29 30 31 It was also found that IRI-2016 correctly predicted the semiannual anomaly and rightly 32 showed that equinox had the stronger crests in both hemispheres (Fig. 4). Furthermore, 33 34 the model captured the stronger crests in the northern hemisphere along with the 35 36 reduction in the magnitude of the asymmetry especially in the post sunset. It further 37 38 represented the solar activity dependence of TEC over the EIA with stronger and well- 39 40 developed crests in 2014 (which had a higher solar flux than 2013). IRI-2016 41 42 nevertheless, overestimated the magnitude of the asymmetry during noon and 43 misrepresented its direction in the post sunset of equinox. It equally overestimation 44 45 /underestimated the magnitude of the northern/southern crest in 2013 while it 46 47 underestimated it in 2014. The discrepancies in the model’s predictive capability of the 48 49 semiannual anomaly could be related to its inability to accurately represent the variations 50 51 in electric field and consequently the fountain effect. Any misrepresentation of the 52 equatorial electrojet (EEJ) which is a proxy of Vz will translate to differences between 53 54 the observed and modeled TEC at the EIA. For example, Silva et al. (2020) showed that 55 56 over the Brazilian region, the vertical drift could be responsible for deviations in the 57 58 spatio-temporal features of the EIA crests in 2014/2015. In the same vein, Sousasantos et 59 60 al. (2020) established that the Scherliess and Fejer model (Scherliess and Fejer, 1999) 61 62 14 63 64 65 1 2 3 4 which drives Vz in IRI had intrinsic trend of underestimation that appears to be 5 6 independent of latitude and season. Recently, Mengistu et al. (2018) and Kumar (2020) 7 8 advocated the inclusion of the EEJ measurements in the IRI model in order to increase its 9 predictive capability over the African and Indian low-latitude, respectively. We however, 10 11 note that given the extreme variability of the EEJ (Doumouya and Cohen, 2004; 12 13 Venkatesh et al., 2015), long term measurements will be needed in Africa in order to 14 15 clearly understand its variability before inclusion in IRI. Unfortunately, instruments such 16 17 as incoherent scatter radar (ISR) which are capable of providing direct measurements of 18 19 EEJ are quite scanty over the African longitude. 20 21 In addition, the difference between the variability of F10.7 and EUV fluxes could 22 23 also be a major source of discrepancies in IRI-2016. F10.7 flux which is often taken as an 24 25 input in the IRI model is just a proxy of EUV flux. EUV flux is the main solar flux 26 27 parameter which exerts a greater control on ionization in the ionosphere (Kumar, 2016). 28 29 If the variability of EUV is not well-reproduced by F10.7, then the output IRI parameter 30 31 (e.g., IRI-TEC in our case) will not matched the observed one (GNSS-TEC). The analysis 32 of the relation between EUV and F10.7 fluxes in 2013 and 2014 revealed that the nature 33 34 of this relation varies significantly from one season to another (Fig. 8). This clearly 35 36 implies that F10.7 flux does not always follows the trend of EUV flux. Emmert et al. 37 38 (2010) had stressed on the difference between EUV and F10.7 during the current solar 39 40 cycle while Kumar (2020) had emphasized on how such difference could reduce the 41 42 performance of IRI-2016. 43 44 5. Conclusion 45 46 47 The capability of the IRI-2016 model in reproducing the hemispheric asymmetry, 48 49 winter and semiannual anomalies has been assessed over Africa. The data were obtained 50 o 51 from a chain of GNSS receivers within 25 - 40° E during quiet period of the years 2013 52 and 2014. We equally analyzed meridional neutral wind and O/N2 ratio measurements in 53 54 both hemispheres. The results showed that: 55 56 57 (i) The hemispheric asymmetry of the anomaly was clearly depicted by IRI-2016 with the 58 59 model predicting farther crests location with respect to the magnetic equator in all 60 61 seasons. 62 15 63 64 65 1 2 3 4 (ii) IRI-2016 overestimated (underestimated) the magnitude of the crests in both 5 6 hemispheres in December solstice of 2013 (2014) while in June solstice it overestimated 7 8 it during both years. 9 10 (iii) The model predicted the winter anomaly in December solstice but failed to do so in 11 12 June solstice. It also clearly represented the semiannual anomaly with stronger crests in 13 14 the northern hemisphere. 15 16 17 (iv) The IRI model reproduced the seasonal trend of variation of the asymmetry index (A) 18 19 in December solstice and equinox (in the noon). In June solstice, the predicted variation 20 21 of A was the opposite of what was observed with GNSS-TEC. In addition, the model 22 underestimated (overestimated) the magnitude of A during solstices (equinox) in the 23 24 noon. In the post sunset, it showed a better agreement with the observed magnitude of A 25 26 during all seasons except June solstice. 27 28 29 (v) The IRI model misrepresented the hmF2 in both hemispheres during June solstice. 30 31 This was a likely reason for its inability to predict the winter anomaly. This finding also 32 33 implied that IRI-2016 failed to capture the variability of the vertical drift and by 34 35 extension the contribution of meridional wind. 36 37 (vi) There was a difference in the seasonal variability of F10.7 and EUV fluxes with the 38 39 correlation between both parameters varying from 0.27 – 0.90. This implied that F10.7 40 41 solar flux which is an input for IRI-2016 did not always capture the changes in EUV flux. 42 43 F10.7 was thus, a potential source of discrepancies in IRI-2016 in 2013 – 2014. 44 45 46 This is the first study dedicated to assessing the capability of IRI-2016 in 47 48 reproducing the hemispheric features of the anomaly in Africa during the maximum 49 phase of solar cycle 24. However, it is to be noted that past studies have reported some 50 51 discrepancies in IRI-2016 over this longitude but with stations located mostly in the 52 53 northern hemisphere. The present study placed emphasis on the model’s performance 54 55 over both hemispheres using a chain of GNSS receivers and further stressed on the 56 57 significance of the inclusion of meridional neutral wind and EEJ data for Africa in IRI- 58 59 2016 in order to improve its performance. The study finally underpinned the contribution 60 61 62 16 63 64 65 1 2 3 4 of solar input parameters (e.g., F10.7) as a source of eventual misrepresentation of the 5 6 observed ionospheric parameters (e.g., TEC). 7 8 9 Acknowledgements 10 11 The authors wish to express their appreciation to the following organizations: the 12 13 UNAVCO (for the GNSS data), the Dominion Radio Astrophysical Observatory 14 15 (DRAO), Penticton, B.C. (for the solar flux data) and the Space Sciences Center of the 16 17 University of Southern California (for the SOHO/SEM EUV data). We are also grateful 18 to the IRI development community and GSFC, NASA for the online version of the IRI- 19 20 2016 model and to IZMIRAN for making available the IRI-Plas model at 21 22 http://ftp.izmiran.ru/pub/izmiran/SPIM/. Special thanks to the PI (A.B. Christensen) and 23 24 Project Scientist (L. Paxton) of the GUVI team which provides the O/N2 data. The wind 25 26 data were derived from the pyglow package, which is an open source software available 27 at https://github.com/timduly4/pyglow/. 28 ‐ 29 30 References 31 32 33 Akala, A. O., Somoye, E.O., Adewale, A.O., Ojutalayo, E.W., Karia, S.P., Idolor, R.O., 34 35 Okoh, D., Doherty, P.H., 2015. Comparison of GPS-TEC observations over Addis 36 37 Ababa with IRI-2012 model predictions during 2010–2013, Advances in Space 38 Research, 56, 1686–1698 39 40 Amaechi, P. O., Oyeyemi, E. O., Akala, A. O., Falayi, E. O., Kaab, M., Benkhaldoun, Z., 41 42 et al., 2020. Quiet time ionopheric irregularities over the African Equatorial 43 44 Ionization Anomaly Region. Radio Science, 55, e2020RS007077. 45 46 https://doi.org/10.1029/2020RS0077. 47 Amaechi, P. O., Oyeyemi, E. O., Akala, A. O., 2018. Variability of the African equatorial 48 49 ionization anomaly (EIA) crests during the year 2013. Canadian Journal of 50 51 Physics, 97(2), 155–165. 52 53 Appleton, E. V., 1946. Two anomalies in the ionosphere. Nature, 157, 691 54 55 Appleton, E. V., 1938. Radio transmission and solar activity. Nature, 3594, 142, 499-501. 56 57 Balan, N., Bailey, G. J., 1995. Equatorial plasma fountain and its effects: Possibility of an 58 additional layer. Journal of Geophysical Research: Space Physics, 100(A11), 59 60 21421-21432. 61 62 17 63 64 65 1 2 3 4 Berkner, L. V., Wells, H. W., Seaton, S. L., 1936. Characteristics of the upper region of 5 6 the ionosphere. Terr. Magn. Atmos. Electr., 41(2), 173–184. 7 8 Bilitza, D., Altadill, D., Truhlik, V., Shubin, V., Galkin, I., Reinisch, B., Huang, X., 9 2017. International Reference Ionosphere 2016: from ionospheric climate to real- 10 11 time weather predictions. Space Weather. 15 (2), 418–429. 12 13 Bilitza, D., Reinisch, B. W., 2008. International Reference Ionosphere 2007: 14 15 Improvement and new parameters, Adv. Space Res., 42, 599–609. 16 17 Blewitt, G., 1990. An automatic editing algorithm for GPS data. Geophysical Research 18 19 Letters, 17(3), 199–202. https://doi.org/10.1029/GL017i003p00199 20 Burns, A. G., Wang, W., Qian, L., Solomon, S. C., Zhang, Y., Paxton, L. J., Yue, X., 21 22 2014. On the solar cycle variation of the winter anomaly, J. Geophys. Res. Space 23 24 Physics, 119, 4938–4949, doi:10.1002/2013JA019552. 25 26 Chartier, A. T., Makela, J. J., Liu, H., Bust, G. S., Noto, J., 2015. Modeled and observed 27 28 equatorial thermospheric winds and temperatures. Journal of Geophysical 29 Research: Space Physics, 120, 5832 5844. https://doi.org/10.1002/2014JA020921 30 – 31 Cherniak I.V., Zakharenkova, I.E.A. Krankowski , Shagimuratova, I.I., 2012. 32 33 Plasmaspheric Electron content derived from GPS TEC and FORMOSAT- 34 35 3/COSMIC measurements: Solar minimum condition. Advances in Space 36 37 Research 50, 427–440. 38 39 Doumouya, V., Cohen, Y., 2004. Improving and testing the empirical equatorial 40 electrojet model with CHAMP satellite data, Ann. Geophys., 22, 3323 3333. 41 – 42 Drob, D. P., Emmert, J. T., Meriwether, J. W., Makela, J. J., Doornbos, E., Conde, M., ... 43 44 Klenzing, J. H. (2015). An update to the Horizontal Wind Model (HWM): The 45 46 quiet time thermosphere. Earth and Space Science, 2(7), 301-319. 47 48 Emmert, J., Lean, J., Picone, J., 2010. Record-low thermospheric density during the 2008 49 50 solar minimum. Geophys. Res. Lett. 37, L12102. 51 https://doi.org/10.1029/2010GL043671. 52 53 Endeshaw, L., 2020. Testing and validating IRI-2016 model over Ethiopian ionosphere. 54 55 Astro-physics and Space Science, 365(3), 1-13. 56 57 Estey, L. H., Meertens, C. M., 1999. TEQC: The multi purpose toolkit for 58 GPS/GLONASS data. GPS Solutions, 3(1), 42 49. 59 – ‐ 60 61 62 18 63 64 65 1 2 3 4 Fejer, B. G., Jensen, J. W., Su, S. Y., 2008. Quiet time equatorial F region vertical plasma 5 6 drift model derived from ROCSAT 1 observations.Journal of Geophysical 7 Research, 113, A05304. https://doi.org/10.1029/2007JA012801 8 ‐ 9 Fejer, B.G., de Paula, E.R., Heelis, R.A., Hanson, W. B., 1995. Global equatorial 10 11 ionospheric vertical plasma drifts measured by the AE E satellite. J. Geophys. 12 13 Res., 100(A4), 5769– 5776. 14 ‐ 15 Gowtam, V. S., Tulasi Ram, S., 2017. Ionospheric winter anomaly and annual anomaly 16 17 observed from Formosat-3/COSMIC Radio Occultation observations during the 18 19 ascending phase of solar cycle 24. Advances in Space Research, 60(8), 1585- 20 1593. 21 22 Hansen, A., Blanch, J., Walter, T., 2000. Ionospheric correction analysis for WAAS quiet 23 24 and stormy (pp. 634–642). ION GPS, Salt Lake City, Utah, September 19–22. 25 26 Hargreaves, J. K., 1995. The Solar-Terrestrial Environment: An Introduction to Geospace 27 28 -The Science of the Terrestrial Upper Atmosphere, Ionosphere, and 29 Magnetosphere, Cambridge Atmos. Space Sci. Ser., vol. 5, Cambridge Univ. 30 31 Press., New York 32 33 Heelis, R. A., Hanson, W. B., 1980. Interhemispheric transport induced by neutral zonal 34 35 winds in the F region. Journal of Geophysical Research: Space Physics, 85(A6), 36 37 3045-3047. 38 39 Kaab, M., Benkhaldoun, Z., Fisher, D. J., Harding, B., Bounhir, A., Makela, J. J., et al., 40 2017. Climatology of thermospheric neutral winds over Oukaïmeden Observatory 41 42 in Morocco. Annales Geophysicae, 35(1), 161–170. ttps://doi.org/10.5194/angeo- 43 44 35-161-2017 45 46 Karia, S.P., Patel, N.C., Pathak, K.N., 2018. On the performance of IRI-2016 to predict 47 48 the North-South Asymmetry of the Equatorial Ionization Anomaly around 73°E 49 50 longitude. Advances in Space Research, 63(6), 1937-1948. 51 https://doi.org/10.1016/j.asr.2018.09.033 52 53 Karia, S.P., Patel, N.C., Pathak, K.N., 2015. Comparison of GPS based TEC 54 55 measurements with the IRI-2012 model for the period of low to moderate solar 56 57 activity (2009–2012) at the crest of equatorial anomaly in the Indian region. Adv. 58 59 Space Res. 55 (8), 1965–1975. 60 61 62 19 63 64 65 1 2 3 4 Khadka, S. M., Valladares, C. E., Sheehan, R., Gerrard, A. J., 2018. Effects of electric 5 6 field and neutral wind on the asymmetry of equatorial ionization anomaly. Radio 7 8 Science, 53(5), 683-697. 9 Kumar, S., 2020. North-South asymmetry of equatorial ionospheric anomaly computed 10 11 from the IRI model. Ann. Geophys., 63, 3, doi:10.4401/ag-8324 12 13 Kumar, S., 2016. Performance of IRI-2016 model during a deep solar minimum and a 14 15 maximum year over global equatorial regions. J. Geophys. Res. Space Physics, 16 17 121, 5664-5674. 18 19 Kwak, Y. S., Kil, H., Lee, W. K., Yang, T. Y., 2019. Variation of the Hemispheric 20 Asymmetry of the Equatorial Ionization Anomaly with Solar Cycle. Journal of 21 22 Astronomy and Space Sciences, 36(3), 159-168. 23 24 Link, R., Cogger, L., 1988. A reexamination of the OI 6300 Å nightglow. J. Geophys. 25 Res., 93(A9), 9883 9892. https://doi.org/10.1029/JA093iA09p09883 26 – ‐ 27 28 Mannucci, A. J., Wilson, B. D., Edwards, C. D., 1993. A new method for monitoring the 29 ionosphere total electron content using the GPS global network. 30 Earth’s 31 Proceedings of ION GPS 93, Institute of Navigation (pp. 1323–1332). 32 33 Melaku, M., Tsidu, G. M., 2019. Comparison of quite time ionospheric total electron 34 ‐ 35 content from IRI-2016 model and GPS observations. Annales Geophysicae 36 37 https://doi. org/10.5194/angeo-2019-44. 38 39 Mengistu T. G., Melaku, Z. M., 2020. Comparison of quiet-time ionospheric total 40 electron content from the IRI-2016 model and from gridded and station-level GPS 41 42 observations. Annales Geophysicae, 38(3), 725-748. 43 44 Mengistu, E., Moldwin, M. B., Damtie, B., Nigussie, M., 2019. The performance of IRI- 45 46 2016 in the African sector of equatorial ionosphere for different geomagnetic 47 48 conditions and time scales. Journal of Atmospheric and Solar-Terrestrial 49 50 Physics, 186, 116-138. 51 Mengistu, E., Damtie, B., Moldwin, M.B., Nigussie, M., 2018. Comparison of GPS-TEC 52 53 measurements with NeQuick2 and IRI model predictions in the low latitude East 54 55 African region during varying solar activity period (1998 and 2008–2015). 56 57 Advances in Space Research, 61, 1456–1475. 58 59 60 61 62 20 63 64 65 1 2 3 4 Millward, G. H., Rishbeth, H., Fuller-Rowell, T. J., Aylward, A. D., Quegan, S., Moffett, 5 6 R. J., 1996. Ionospheric F2 layer seasonal and semiannual variations, J. Geophys. 7 8 Res., 101(A3), 5149–5156. 9 Mitra, S.K., 1946. Geomagnetic control region F2 of the ionosphere. Nature 158, 668- 10 11 669. 12 13 Namba, S., Maeda, K.-I., 1939. Radio Wave Propagation, 86 pp., Corona, Tokyo 14 15 Oyedokun, O. J., Akala, A. O., Oyeyemi, E. O., 2020. Characterization of African 16 17 Equatorial Ionization Anomaly (EIA) during the maximum phase of solar cycle 24. 18 19 Journal of Geophysical Research: Space Physics. doi:10.1029/2019ja027066. 20 Paul, A., DasGupta, A., 2010. Characteristics of the equatorial ionization anomaly in 21 22 relation to the day to day variability of ionospheric irregularities around the 23 24 post sunset period. Radio Science, 45, RS6001. 25 ‐ ‐ 26 Qian, L., Burns, A. G., Wang, W., Solomon, S. C., Zhang, Y., Hsu, V., 2016. Effects of 27 28 the equatorial ionosphere anomaly on the interhemispheric circulation in the 29 thermosphere, J. Geophys. Res. Space Physics, 121, doi:10.1002/2015JA022169. 30 31 Qian, L., Yue, J., 2017. Impact of the lower thermospheric winter-to-summer residual 32 33 circulation on thermospheric composition, Geophys. Res. Lett., 44, 3971–3979, 34 35 doi:10.1002/2017GL073361. 36 37 Rishbeth, H., Muller-Wodarg, I. C. F., Zou, L., Fuller-Rowell, T. J., Millward, G. H., 38 39 Moffett, R. J., Idenden, D. W., Aylward, A. D., 2000. Annual and semiannual 40 variations in the ionospheric F2-layer: II. Physical discussion, Ann. Geophys., 18, 41 42 945–956. http://www.ann-geophys.net/18/945/2000/ 43 44 Rishbeth H. 1998. How the thermospheric circulation affects the ionosphere. J. Atm Sol- 45 46 Terr Phys 60: 1385–1402. DOI: 10.1016/S1364-6826(98)00062-5. 47 48 Rishbeth, H., Setty, C.S.G.K., 1961. The F-layer at sunrise. Journal of Atmospheric and 49 50 Terrestrial Physics, 20(4), 263-276. 51 Sardon, E., Rius, A., Zarraoa, N., 1994. Estimation of the transmitter and receiver 52 53 differential biases and the ionospheric total electron content from GPS 54 55 observations. Radio Science, 29(3), 577–586. https://doi.org/10.1029/94RS00449 56 57 Scherliess, L., Fejer, B. G., 1999. Radar and satellite global equatorial F region vertical 58 59 drift model. J Geophys Res 104(A4): 6829–6842. 60 61 Seemala, G., Delay, S. B., 2010. GNSS TEC data processing. 2nd Workshop on Satellite 62 21 63 64 65 1 2 3 4 Navigation Science and Technology for Africa, Trieste, 6–24 April 2010. 5 6 Silva, A. L. A., Sousasantos, J., Marini-Pereira, L., Lourenço, L. F. D., Moraes, A. O., 7 8 Abdu, M. A., 2020. Evaluation of the dusk and early nighttime Total Electron 9 Content modeling over the eastern Brazilian region during a solar maximum 10 11 period. Advances in Space Research. doi.org/10.1016/j.asr.2020.12.015 12 13 Sousasantos, J., Abdu, M. A., Santos, A., Batista, I., Silva, A., Lourdes, L. E., 2020. 14 15 Further complexities on the pre-reversal vertical drift modeling over the Brazilian 16 17 region: A comparison between long-term observations and model results. J. Space 18 19 Weather Space Clim. 10, 20. ttps://doi.org/10.1051/swsc/2020022 20 Sterling, D. L., Hanson, W. B., Moffett, R. J., Baxter, R. G., 1969. Influence of 21 22 electromagnetic drifts and neutral air winds on some features of the F 2 region. 23 24 Radio Sci.4, 1005–1023. 25 26 Tariku, Y. A., 2020. Comparison of performance of the IRI 2016, IRI Plas 2017, and 27 28 NeQuick 2 models during different solar activity (2013–2018) years over South 29 American sector. Radio Science, 55(8), 1-17. 30 31 Titheridge, J. E., 1995. Winds in the ionosphere - A review. Journal of Atmospheric and 32 33 Terrestrial Physics, 57(14), 1681-1714. 34 35 Torr, M.R., Torr, D.G., 1973. The seasonal behaviour of the F2-layer of the ionosphere. J 36 37 Atm. Terr. Phys., 35: 2237–2251. DOI:10.1016/0021- 9169(73)90140-2. 38 39 Tulasi Ram, S., Su, S. Y., Liu, C. H., 2009. FORMOSAT 3/COSMIC observations of 40 seasonal and longitudinal variations of equatorial ionization anomaly and its 41 ‐ 42 interhemispheric asymmetry during the solar minimum period. Journal of 43 44 Geophysical Research: Space Physics, 114(A6). 45 46 Yasyukevich, Y.V., Yasyukevich, A.S., Ratovsky, K.G., Klimenko, M.V., Klimenko, V. 47 48 V., Chirik, N.V., 2018. Winter anomaly in NmF2 and TEC: when and where it 49 50 can occur. Journal of Space Weather and Space Climate, 8, A45. 51 Venkatesh, K., Fagundes, P. R., Prasad, D. S. V. V. D., Denardini, C. M., de Abreu, A.J., 52 53 de Jesus, R., Gende, M., 2015. Day-today variability of equatorial electrojet and 54 55 its role on the day-to-day characteristics of the equatorial ionization anomaly over 56 57 the Indian and Brazilian sectors, J. Geophys. Res. Space Physics, 120, 58 59 doi:10.1002/2015JA021307. 60 61 Yonezawa, T., Arima, Y., 1959. On the seasonal and non seasonal annual variations and 62 22 63 ‐ 64 65 1 2 3 4 the semi annual variation in the noon and midnight electron densities of the F2 5 layer in middle latitudes. J. Radio Res. Labs., 6, 293 309. 6 ‐ – 7 8 Zhao, B., Wan, W., Liu, L., Mao, T., Ren, Z., Wang, M., Christensen, A. B., 2007. 9 Features of annual and semiannual variations derived from the global ionospheric 10 11 maps of total electron content. Ann. Geophys., 25, 2513–2527 12 13 Zeng, Z., Burns, A. G., Wang, W., Lei, J., Solomon, S.C., Syndergaard, S., Qian, L., 14 15 Kuo, Y. H., 2008. Ionospheric annual asymmetry observed by the COSMIC 16 radio occultation measurements and simulated by the TIEGCM. J. Geophys. 17 ‐ 18 19 Res., 113, A07305, doi:10.1029/2007JA012897 20 Zou, L., Rishbeth, H., Müller-Wodarg, I.C.F., Aylward, A.D., Millward, G.H., Fuller- 21 22 Rowell, T.J., Idenden, D.W., Moffett, R.J., 2000. Annual and semiannual 23 24 variations in the ionospheric F2-layer. I. Modelling. Ann Geophys 18: 927–944. 25 26 DOI: 10.1007/s00585-000-0927-8. 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 23 63 64 65 Table Click here to access/download;Table;Table 1.docx

S/N Data Organization Data link /Instrument 1 GNSS-TEC UNAVCO www.unavco.com 2 IRI-TEC COSPAR/URSI https://ccmc.gsfc.nasa.gov/modelweb/models/iri2016_vitmo.php 3 O/N2 ratio GUVI guvitimed.jhuapl.edu 4 EUV SEM/SOHO https://soho.nascom.nasa.gov/data/data.html 5 F10.7 DRAO http://www.spaceweather.gc.ca/solarflux/sx-5-en.php 6 Kp ISGI isgi.unistra.fr

UNAVCO- University Navstar Consortium SEM - Solar EUV Monitor; SOHO - Solar Heliophysical Observatory (SOHO) DRAO-Dominion Radio Astrophysical Observatory ISGI- International Service of Geomagnetic Indices

Figure Click here to access/download;Figure;List of Figures.docx

List of Figures

Fig. 1. Geogaphic locations of the GNSS stations used: the magenta solid curve indicates the magnetic equator (dip = 0) while the horizontal green curves indicate ±20 degree dip latitudes.

Fig. 2. Seasonal variations of the EIA reconstructed using GNSS -TEC and IRI-TEC during solstices of years 2013 - 2014. The horizontal black lines represent the magnetic equator.

Fig. 3. Noon and post sunset variations of the: (a-b) asymmetry index (A) in solstices of years 2013 -2014 and (c-d) maximum height of the F2 layer (hmF2) in year 2013. A was computed with TEC derived from GNSS and predicted by IRI, respectively.

Fig. 4. Variations of GNSS-TEC and IRI-TEC in equinox and solstices in the years 2013 and 2014. The horizontal black lines represent the magnetic equator.

Fig. 5. Changes in the asymmetry index during equinox and solstices of 2013 and 2014.

Fig. 6. Seasonal variations in O/N2 ratio in both hemispheres from 2013 to 2014.

Fig. 7. Seasonal variations of thermospheric neutral wind velocity in the years 2013 and 2014. The broken horizontal lines represent the magnetic equator.

Fig. 8. Seasonal correlation between EUV flux (26–34 nm) and F10.7 solar flux in year 2013 and 2014.