Variability of TEC and improvement of performance 1 of the IRI model over Ethiopia during the high solar 2 activity phase 3

10 This paper discusses the monthly and seasonal variation of the total electron content (TEC) and 11 the improvement of performance of the IRI model in estimating TEC over Ethiopia during the 12 solar maximum (2013-2016) phase employing GPS TEC data inferred from the GPS receivers 13 installed at different regions of Ethiopia. The results reveal that both the measured and modeled 14 seasonal diurnal VTEC values start increasing at 03:00 UT (06:00 LT) and attain their peak 15 values (mostly in the time interval of 09:00-13:00 UT or 12:00-16:00 LT). Moreover, both the 16 arithmetic mean measured and modeled VTEC values, generally, show maximum and minimum 17 values in the equinoctial and June solstice months, respectively. The results also show that, even 18 though overestimation of the modeled VTEC has been observed on most of the hours, the model 19 is generally good to estimate both the monthly and seasonal diurnal hourly VTEC values, 20 especially in the early morning hours (00:00-03:00 UT or 03:00-06:00 LT). Moreover, the overall 21 Ann. Geophys. Discuss., https://doi.org/10.5194/angeo-2018-48 Manuscript under review for journal Ann. Geophys. Discussion started: 5 June 2018 c © Author(s) 2018. CC BY 4.0 License.


Introduction
The energy transferred from the sun causes atoms and molecules existing in the atmosphere to undergo chemical reactions and become ionized (Kelley, 2009).This ionized and conductive region of Earth's atmosphere, extending from about 50 to 1000 km and possessing free electrons and positive ions generally in equal numbers in a medium that is electrically neutral, is termed as ionospher.The existence of these ions (plasma) in the ionosphere results in the possibility of radio communications over large distance by making use of one or more ionospheric reflections (Hunsucker and Hargreaves, 2003).
On the other hand, the ionosphere affects the electromagnetic waves that pass through it by inducing additional transmission time delay (Gao and Liu, 2002).Because of its dispersive character, electromagnetic signals (such as GPS signals) experience time delay (modulated codes) and advance (carrier phase) as they propagate through the ionosphere.This delay is directly proportional to the integral number of electrons in a unit cross-sectional area (usually referred to as total electron content, TEC) along the signal path extending from the satellite to the receiver on the ground, and inversely proportional to the square of the frequency of propagation (Hofmann-Wellenhof et al., 1992;Misra and Enge, 2006).The dispersive ionosphere introduces GPS satellites orbiting at 20,200 km (Hansen et al., 2000).The relative ionospheric delay of the two signals is proportional to the TEC.Time delay measurements of L1 and L2 frequencies can, therefore, be converted to TEC along the ray path from the receiver to the satellite (Lanyi and Roth, 1988).The GPS signals traverses the ionosphere carrying signatures of the dynamic medium and thus offers opportunities for ionospheric research.As a result, global and regional maps of ionospheric TEC can be produced using data from the worldwide network of the International GPS Service (Lanyi and Roth, 1988).The availability of TEC measurements is also important to the development of ionospheric models such as the International Reference Ionosphere, IRI (Bilitza, 2001).The International Reference Ionosphere (IRI) is an international project sponsored by the Committee on Space Research (COSPAR) and the International Union on Radio Science (URSI).
Using the GPS satellites and the IRI model, there have been so far several researches conducted globally in connection with the TEC variability and performance of the model over equatorial and low latitude regions.Based on the findings, the researchers have put their own views concerning the TEC variation and the prediction performance of different versions of the model.Ezquer et al. (2014), for instance, noted that IRI 2012 predictions show significant deviations from experimental values during the period of 2008-2009 for a station placed at the southern crest of the equatorial anomaly in the American region.Olwendo et al. (2012a) also noted that seasonal average IRI 2007 TEC values were higher than the GPS-TEC data for the period of 2009-2011 over different regions in Kenya.In addition, Olwendo et al. (2012b) reported that the IRI 2007 TEC is too high for all seasons except for the March equinox (where there seems to be good agreement between observation and model) during the lowest solar activity phase (2009)(2010).The report of Kumar (2016) on the validation of the IRI 2012 models for the global equatorial region also showed that the IRI 2012 model generally overestimated the observed VTEC over equatorial regions during the solar minimum year (2009) and solar maximum (2012) phases.Abdu et al. (1996); Kakinami et al. (2012); Kumar et al. (2015) attempted to describe the model's capacity to estimate the TEC using different versions of the model.Asmare et al. (2014) and Tariku, 2015a andTariku, 2015b also attempted to see patterns in both the measured and modeled VTEC variations during the low and high solar activity phases employing different GPS stations and IRI 2012 model at various regions of Ethiopia.Asmare et al. (2014), for instance, showed that the model entirely overestimated both monthly and seasonal VTEC values during phases of low solar activity.In addition, the model performance in estimating diurnal VTEC variations was found to be better during low solar activity phases than during high solar activity phases.In addition, the highest and the lowest values of the VTEC are observed in the equinoctial and the June solstice months, respectively during both the low and high solar activity phases.
Thus, this study is mainly important to observe the TEC variation and the improvement of performance of the IRI model in estimating the TEC variation over low latitude African regions during the high solar activity phase (2013)(2014)(2015)(2016) employing the GPS VTEC data inferred from different regions of Ethiopia.This is because the TEC is the major parameter that can largely affect radio wave propagation in the ionosphere.Consequently, for a better radio wave propagation which can foster Earth-to-space communication, the TEC in the ionosphere has to be studied and its effects on the signal propagated through the ionosphere must be identified.In addition, for a better improvement of the IRI model in estimating the variation of TEC, its performance has to be continuously tested, especially over the equatorial and low latitude regions, where the dynamics of the ionosphere is very complex.To observe the TEC variation and improvement of performance of the IRI model in estimating the TEC variation the latest versions (IRI 2007, IRI 2012and IRI 2016) with NeQuick option for the topside electron density during the solar maximum phase have been considered.This is conducted to choose and use the best version of the model in estimating the TEC variability in some occasions when the GPS TEC data are scarcely available in the receiver.The prediction performance of the model has been tested by comparing the modeled TEC values with the GPS-TEC values recorded in the receivers.

TEC from dual frequency GPS receiver
As different studies (e.g.Ciraolo et al., 2007;Mannucci et al., 1998) show that the GPS measurements are used to estimate the TEC along a ray path between a GPS satellite and receiver on the ground.These GPS measurements can be recorded using either single or dual frequency GPS receivers.However, to eliminate ionospheric errors in the estimation of TEC dual frequency receivers are better (Klobuchar, 1996).Moreover, by computing the differential phases of the code and carrier phase measurements, dual frequency GPS receivers can provide integral information about the ionosphere and plasma sphere (Ciraolo et al., 2007;Nahavandchi and Soltanpour, 2008).Hence, in this paper, the GPS-TEC data have been obtained from dual frequency receiver using pseudo-range and carrier phase measurements.The TEC inferred from the pseudo-range (P) measurement is given by:  1  2  2  2   2  1   2  2   2  1   3 .40

 
Similarly, the TEC from carrier phase measurement (Ф) is given as where 1 f and 2 f can be related with the fundamental frequency, As shown above, by cross correlating the f 1 and f 2 modulated carrier signals which are generally assumed to travel along the same path through the ionosphere, the GPS receiver obtains the time delay of the code and the carrier phase difference.As the TEC obtained from code pseudo-range measurements is free of ambiguity, but with relatively much noise; and the TEC obtained from carrier phase measurements has relatively less noise, but is ambiguous, linearly combining both code pseudo-range and carrier phase measurements for the same satellite pass is supposed to increase the accuracy of TEC (Klobuchar et al., 1996;Gao and Liu, 2002).
To better characterize the TEC over a given receiver position and see the overall ionization of the Earth's ionosphere, the slant TEC (STEC) must be converted into equivalent vertical TEC (VTEC) at the mean ionospheric height, h m =350 km (Mannucci et al., 1998;Norsuzila et al., 2008Norsuzila et al., , 2009)).Hence, the relationship between STEC and VTEC in terms of the zenith angle '

 at
the Ionospheric Piercing Point (IPP) and the zenith angle  at the receiver position can be given by: (5) Substituting equation ( 5) into equation ( 4) and rearranging, we get Here, e R is the radius of the Earth in kilometers.

TEC from the International Reference Ionosphere (IRI) model
The International Reference Ionosphere (IRI) is an international empirical standard model used for the specification of ionospheric parameters.The model provides average values of electron density, electron content, electron and ion temperature, and ion composition as a function of height, location, local time, and sunspot number for magnetically quiet conditions (Bilitza, 2001;Bilitza et al., 2014;Bilitza et al., 2017)

Data sources and method of analysis
The data required for both the experimental and model were obtained from Ethiopian regions shown in Figure 1 during the solar maximum (2013-2016) phase.Table 1 also shows the GPS receiver locations used for the study.The raw GPS data for the described station were obtained from the University NAVSTAR Consortium (UNAVCO web site, http://www.unavco.org/).The data gained from this web site have two forms: observation and navigation data in which both of them are zipped.To use the data for the desired purpose, the GG software (GPS-TEC calibrating software) was used to process the required data in five minutes interval and an elevation cut-off 10 o (see Ciraolo, et al., 2007).
To get the required results, the corresponding modeled VTEC values were inferred from the latest versions of the model (IRI 2007, IRI 2012and IRI 2016) that include some latest input the absolute differences between the IRI VTEC and GPS VTEC in all the monthly and seasonal variations were determined.In addition, the percentage differences between the IRI VTEC and GPS VTEC for the arithmetic monthly and seasonal VTEC variations have also determined.

Diurnal monthly and seasonal variation of VTEC and performance of the IRI model
The interval of 09:00-13:00 UT or 12:00-16:00 LT) due to enhancement of ionization during the described time; while their values start decreasing in the nighttime hours and become minimum after midnight hours (on average at 03:00 UT or 06:00 LT) as shown in figures 2-7.Moreover, in some hours, the modeled VTEC values are found to be in a good agreement with the measured (GPS VTEC) values, especially in the nighttime hours (00:00-03:00 UT or 03:00-06:00 LT).
Moreover, the model is found to underestimate the VTEC values during the daytime hours (09:00-13:00 UT or 12:00-16:00 LT).The mismodelings observed in both cases may be due to the difference in the model and experimental slab-thickness as noted by different findings (e.g.Nigussie et al., 2013;Rios et al., 2007).It is known that, in general, electron population in the ionosphere is mainly controlled by solar photo-ionization and recombination processes (Wu et al., 2004).Thus, for the equinoctial months, as the subsolar point is around the equator where the east ward electrojet associated electric field is often largest, it would be speculated that the peak photoelectron abundance and intense eastward electric field will be set up in the described region.On the contrary, for solstice months photoelectrons at the equator decrease as the sub solar point moves to higher latitudes.
Moreover, the change of direction of neutral wind may account for the highest VTEC values in the equinoctial months and lowest values in the June solstice months.A meridional component of neutral wind blows from the summer to the winter hemisphere that is able to reduce the ionization crest value during summer solstice as it blows in an opposite direction to the plasma diffusion process originating from the magnetic equator.Thus, in equinoxes meridional winds blowing from the equator to polar regions may attribute to a high ionization crest value.Hence, a seasonal effect on the crest should be expected with the crest maximum at the equinoxes and minimum in the summer season or June solstice (Bhuyan and Borah, 2007;Wu et al., 2004), which is consistent with the result of this work.

Arithmetic mean of monthly and seasonal variation of VTEC and performance of the IRI model
To visualize the monthly and seasonal VTEC variations, the arithmetic mean hourly

Author contribution
All the required issues for the manuscript are prepared by the corresponding author, Yekoye    the storm option is "on" and "off," respectively.
parameters which are supposed to improve the capacity of the model in estimating ionospheric parameters.The online IRI versions of the model were obtained from http://omniweb.gsfc.nasa.vitmo.html.To get the VTEC values, the year, date, month, location, Ann.Geophys.Discuss., https://doi.org/10.5194/angeo-2018-48Manuscript under review for journal Ann.Geophys.Discussion started: 5 June 2018 c Author(s) 2018.CC BY 4.0 License. the hour profile, the upper boundary altitude (2000 km), daily sunspot number and F10.7 radio flux, topside electron density options (NeQuick, IRI01-corr, IRI2001), CCIR for F peak model, and ABT-2009 for bottomside thickness option were used as the input parameters,.In order to observe the pattern of the hour-to-hour variability of VTEC, the mean monthly and seasonal hourly GPS TEC and the corresponding IRI TEC data have been used during the period of 2013-2016.To see the monthly and seasonal arithmetic mean VTEC variation and the model performance, the hour-to-hour measured and modeled VTEC values have been correspondingly added and averaged for the whole days in each month and season.The seasons could be classified as December solstice (November, December and January), March equinox (February, March and April), June solstice (May, June and July) and September equinox (August, September and October).For a better understanding on the performance of the model, the absolute differences between the monthly and seasonal GPS VTEC and the corresponding IRI VTEC values have been determined.The differences have been calculated by subtracting the experimental VTEC values from the model.In order to clearly see the validation of the model, diurnal monthly and seasonal VTEC variation results are displayed in Figs 2-7.The results reveal that, almost both the monthly diurnal GPS VTEC and IRI VTEC values start increasing at 03:00 UT (06:00 LT) and attain their peak values in the daytime hours (especially in the time Ann.Geophys.Discuss., https://doi.org/10.5194/angeo-2018-48Manuscript under review for journal Ann.Geophys.Discussion started: 5 June 2018 c Author(s) 2018.CC BY 4.0 License.
For instance,Rios et al. (2007) using the IRI 2001 model, showed that IRI predicted slab thickness is higher than the measured values except between (10:00-14:00 LT) which can attribute to VTEC fluctuations in similar trend.This is almost consistent with the result determined in this work.Using IRI 2007 model,Nigussie et al. (2013) also suggested similar possible reason for the discrepancy between the model and the experimental VTEC values.It could also be resulting from poor estimation of the hmF2 and foF2 from the coefficients, which in turn may result in poor estimation of VTEC by the IRI model (e.g.Chakraborty et al., 2014;Kumar et al, 2015).The underestimation of the IRI VTEC values by the GPS VTEC values may also attribute to the enhancement of the plasmaspheric electron content above 2000 kms during the daytime hours.Moreover, the maximum peak of both the measured and modeled VTEC values are generally observed in the equinoctial months; while, the minimum peak values are observed in the June solstice months (see Fig.2-7).For instance, over Arba Minch station (see Figs. 2 and 3), the highest and lowest peak measured monthly VTEC values of about 80 and 40 TECU are observed in March and July, respectively.Similarly, the highest and lowest peak modeled Ann.Geophys.Discuss., https://doi.org/10.5194/angeo-2018-48Manuscript under review for journal Ann.Geophys.Discussion started: 5 June 2018 c Author(s) 2018.CC BY 4.0 License.seasonal VTEC values of about 55 and 41 TECU are observed in April and July, respectively in using IRI 2007 model with NeQuick option for the topside electron density.The seasonal diurnal VTEC values generally follow the pattern of the diurnal monthly VTEC values with the lowest and highest values being observed at about 03:00 UT (06:00LT) and in the time interval of about 09:00-13:00 UT (12:00-16:00 LT), respectively (see Fig. 4-7).In addition, the highest and lowest peak measured seasonal VTEC values of about 80 and 50 TECU are observed in the March equinox and June solstice, respectively.The highest and lowest peak modeled seasonal VTEC values of about 54 and 43 TECU are observed in the March equinox and June solstice, respectively when using IRI 2007 model with NeQuick option for the topside electron density over Arba Minch station (see Fig. 6).In addition, in using IRI-2012 model with IRI-2001 option for the topside electron density, the highest and lowest peak measured seasonal VTEC values of about 70 and 40 TECU are observed in the March equinox and June solstice, respectively over Ambo station in 2014.Similarly, the highest and lowest peak modeled seasonal VTEC values of about 74 and 60 TECU are observed in the March equinox and June solstice, respectively in 2014 when using the same version of the model (IRI 2012) with IRI-2001 option (see Fig. 5).
Fig.10).In general, using the IRI 2016 model shows highest overestimation of the VTEC as compared to others(IRI 2007 and IRI 2012).For instance, the highest monthly and seasonal deviations of about 25% and 20% are observed between the modeled and corresponding measured values in September and the June solstice, respectively in using IRI 2016 model (see the bottom right panels of Fig.10).

Figure 1 : 449 Figure 2 : 453 Figure 3 :
Figure 1: Location of GPS receivers used for the study

Figure 7 :
Figure 7: A graph to illustrate diurnal seasonal VTEC variation and performance of the IRI model over Asosa station during the period of 2016

Figure 9 :
Figure 9: A graph to illustrate the arithmetic mean monthly and seasonal VTEC variation and performance of the IRI-2012 model over Ambo station during the period of 2014

Figure 11 :Figure 12 :
Figure 11: A graph to illustrate the arithmetic mean monthly and seasonal VTEC variation and performance of the IRI model over Nazret station during the period of 2015 Bilitza et al., 2017).For a given location, time and date, like the previous versions of the model, IRI-2016 model provides the monthly averages of ionospheric parameters (such as TEC) in the altitude range from about 50-2000 km(Bilitza et al., 2017; . To enhance the capacity of the model, improvements have been made through the ingestion of all worldwide available data from ground-based as well as satellite observations.As a result, a new version of the model (IRI 2016) relationship to the propagation factor M(3000)F2.Thus, the new model options enable the IRI 2016 model to predict evening peaks that was not possible in the old versions.In addition, the improvement of the ion composition model in the topside ionosphere can lower the transition Ann.Geophys.Discuss., https://doi.org/10.5194/angeo-2018-48Manuscript under review for journal Ann.Geophys.Discussion started: 5 June 2018 c Author(s) 2018.CC BY 4.0 License.height from close to 1000 km down to almost 600 km in the new version of the model.A number of smaller changes have also been made concerning the use of solar indices and the speed-up of the computer program ((http://omniweb.gsfc.nasa.gov/vitmo/iri-vitmo.html) that was accessed for the period of 25-30/01/2018.