Total electron content variation over HALY (Al-Jouf), Saudi Arabia and comparison with IRI-2012 and IRI- 2016 models

Ionospheric perdition studies are very few over Saudi Arabia due to less availability of data measurement. Although such kind of studies have been carried out all over the world, there still remains scope to ascertain prediction error in this country. Hence, in the current study, the ionospheric variation from April 2016 to February 2018 (almost 22 months) was studied over a GPS site HALY (29.140N; 36.10 0E), Al-Jouf, Saudi Arabia. Diurnal, monthly and seasonal ionospheric variations were investigated and compared with the existing global IRI (IRI 2012 and IRI 2016) models. Percentage deviation between observed and modeled TEC variation values indicated largescale deviation around 200% during the time of storm. Results showed that the IRI 2012 model had the lowest Root Mean Square Error (RMSE) value (2.7437) during the September Equinox while IRI2016 showed the highest RMSE magnitudes (3.0166) during the December Solstice. In some seasons, the RMSE values were observed to be better for IRI 2012 while on other occasions, it emerged that IRI 2016 yielded more accurate results. Such variations indicate that even the most updated version of the IRI 2016 model is unable to provide perfect estimation and the requirement of further research and improvement in this field cannot be denied.


Introduction
Global Navigation Satellite Systems (GNSS) signals are the combination of other global navigation systems such as GPS from United States, GLONASS from Russia, Galileo from Europe and local systems like COMPASS from China, IRNSS from India and QZSS from Japan. The signals from these networks are used for the purpose of navigation and play an important role in atmospheric studies by using forecasting methods. Extracting the atmospheric delay of the wet part of the troposphere is also possible nowadays. Ground-based GNSS receivers can estimate Integrated Precipitable Water Vapor (IPWV) and Precipitable Water Vapor in all weather conditions [1][2][3][4][5][6]. GNSS signals also consider the impact of rainfall, theoretical results pertaining to which have been discussed by Solheim et al [7]. The sensitivity of the total delay in the presence of severe precipitation has been analyzed by several authors [e.g. 8,9]. Several studies have been carried out by using highprecision GNSS based on crustal deformation and Strain analysis [10][11][12][13]. The space weather plays a significant role in our day-to-day life, including in the advancement of the high-tech military systems based upon the highest strata of the Earth's atmosphere and ionosphere. The study of ionosphere contributes significantly to the information system, health, satellite communications, air traffic control, navigation and positioning system and space science. The Earth's upper atmosphere is ionized by the extreme ultraviolet radiation generated in the ionosphere. The geomagnetic storms are mainly responsible for the increased/decreased electron density, total electron content, and thickness of the ionosphere.
To study the GNSS and ionospheric variations the most significant parameter to be evaluated is the Total Electron Contents (TEC). Recent advancement in the GNSS technology has enabled the measurement of the Radio Occultation (RO) of the Earth' atmosphere with the help of the GNSS satellite constellation of Low Earth Orbit (LEO) satellites. Various models have been proposed to increase enhanced applications of ionospheric data and GNSS in several studies. Several empirical models have been developed for the estimation TEC, e.g. electron density profiler, NeQuick and International Reference Ionosphere (IRI) etc. The IRI model estimates various parameters like electron temperature, ion temperature, and monthly mean values for electron ion density [14][15][16]. The availability of GNSS data plays a highly significant role in the validation of the estimated values of the empirical models in the given region. In the Saudi Arabian region, the availability of the experimental data measurement is very low, due to the slightly low prediction ability of the ionospheric model. Though the IRI model versions are updated frequently to obtain a better estimation, there still remains scope for improvement with regard to validation with the GNSS ionosonde and coherent scatter radar. Extensive research has been conducted so far with the use of IRI models for different equilateral and low-latitude locations [17][18][19][20][21][22][23][24][25][26][27][28].The lack of coherence is observed especially during geomagnetic disturbance. Recently, Sharma et al [29] studied the feasibility of the NeQuick-2 model to explore TEC variation in the ionosphere over the Manama, Bahrain region and observed that the data was in good agreement during a specific period; however, the need of further experimentation and more scientific efforts was stressed, for improving the empirical model and ensuring better representation of a realistic model. Hence, in the current study, the validity of IRI models (IRI 2012 and IRI 2016) has been checked in one of the UNAVCO GPS sites named HALY, which is located at Al-Jouf, Saudi Arabia.

Data and Methods:
The present study attempted to assess the reliability of IRI-2012 and IRI-2016 models by comparing the variability with GPS-derived TEC from a UNAVCO GPS site named HALY located at over Al-Jouf, Saudi Arabia (Fig. 1).The IRI-2012 and IRI-2016 models can be accessed (https://ccmc.gsfc.nasa.gov/modelweb/models/iri2012_vitmo.php) and https://ccmc.gsfc.nasa.gov/modelweb/models/iri2016_vitmo.php), respectively. The aim of this study is to compare the IRI-2012 and IRI-2016 models by using the GPS-TEC values obtained over Al-Jouf, Saudi Arabia, using almost two years GPS data set from March 2016 to February 2018. The GPS data in RINEX format have been accessed from UNAVCO website and processed with GPS-TEC analysis program provide by GOPI-Seemala [30]. The program first calculates slant TEC (STEC) value and then converts them into vertical TEC (VTEC) by using the single layer model (SLM) mapping function, which assumes the ionospheric electron density to be concentrated in a thin shell at 350 km above the Earth. BRDC file was downloaded from the CDDIS server (ftp://ftp.aiub.unibe.ch/CODE) automatically. The following formula was used to convert VTEC from STEC: Where RE stands for radius of the Earth (6378 km), α is the elevation angle and hmax(350 km) is the ionospheric thin shell above the Earth. The elevation angle of 20 0 was selected to avoid the multipath error. Percentage deviation from GPS-TEC values and IRI models were estimated for comparison purpose by using the following equation [31] (%) 100 Root mean square error was calculated to check the comparison of seasonal variation by using the formula given below: Where GPSTEC stands for GPS-TEC values and ModelTEC stands for IRI models TEC values. N denotes the number of observations.

Diurnal VTEC variation
The earth rotates on its axis and takes around twenty-four hours for one rotation; such type of motion of earth is known as rotational motion. VTEC activity is due to the rotational motion of the earth and it keeps changing with time at a particular location. Activity change of VTEC variation is called diurnal VTEC variation. In the current study, the VTEC variation over HALY GPS site located in Saudi Arabia has been checked for the year of April 2016 to February 2018 and compared with IRI models. The contour plot of these variations has been plotted in Fig. 2. It may be observed that the VTEC of the year 2017 varied from around 1 TECU to 20 TECU (Fig. 2a). At the starting of the day, from 0.00 UT till 4.00 UT, the VTEC variation was found to be low, and such was observed round the year. A small variation was observed around the day 350, which may be attributed to a storm day or to an error in VTEC measurement. However, after 4.00 UT the VTEC variation became high till 10.00 UT. It may be noticed the VTEC values were rather high between 7.00 UT to 13.00 UT. The variation of VTEC was not constant between these periods and it kept on varying all over the year. It was very low, around 4 TECU, on the first day of the year and started to increase around the 40 th day of the year. The VTEC value became around 12 TECU on the 50 th day of the year and sometimes touched 18 TECU between the 100-150 th days. The value of VTEC was very low again between the 200-250 th day (around 3 TECU) and began to increase between the 250-325 th day of the year 2017. The VTEC value again became low from the 325 th day till the end of the year. It may be noticed that the VTEC variation after 13.00 UT was high during the 50-300 th day, but declined with time and became very low after 19.00 UT. By the end of the day, the variation is same all over the year, as it was in the starting of the day. Such variations have been discussed by several authors [32][33][34]. While the diurnal variation of VTEC depends on the rotational motion of the earth, the seasonal variation is caused by the revolutionary motion.
The diurnal VTEC variation of IRI models (IRI-2012 and IRI-2016) for the year 2017 at the HALY site have been plotted in Fig. 2b and Fig. 2c. Both IRI models contour plots show low VTEC values at the start and end of the day but touch higher values between the hours of 6:00 UT to 14:00 UT; same pattern was observed at the HALY GPS site. The notable difference is the discrepancy of VTEC values. The GPS VTEC discrepancies were rather high, but the IRI-2012 models showed a pattern of constant variation. Remarkably, smaller discrepancies were observed in the IRI-2016 model compared to the IRI-2012 model. IRI-2016 has the newest upgradation in the model after IRI-2012, having included two new F2-peak height hmF2 modeling selections with their data sources from ionosonde quantities and COSMIC radio occultation. Most importantly, the IRI Real-Time Assimilative Modeling (IRTAM) can even cope with the times of disturbed ionospheric situations, having inserted a number of digitonide limitations. However, the IRI-2016 model still unable to reach and perfect model the VTEC variation.    [34]. The period from July 2017 to September 2017 showed low VTEC values (about 5 TECU), and after that it started to increase again during November 2017. The VTEC value has been noticed to be low during January 2018 to February 2018, again reaching about 5 TECU. The corresponding IRI plots showed almost the same pattern, but they showed higher values than the GPS ones. Overestimation by IRI models, as compared to GPS data, has been reported by many studies [35][36][37][38].  Monthly percentage deviation between GPS-VTEC verses IRI models was calculated by using equation (2) and plotted in Figs. 4a & 4b. Positive percentage deviation means the GPS VTEC are overestimated, and vice versa. It is clear from the figures that the GPS VTEC values were overestimated during the late night (18:00 UT-24.00 UT) and morning period (00:00 UT-03:00 UT), by up to 50%. This estimation started to decrease and became 0 during the period 03:00 UT to 05:00 UT at the start of the day and 16:00 UT-17:00 UT at the end. The percentage deviation was negative between the hours of 06:00 UT to 16:00 UT and reached up to more than 100%. In January 2018, the deviation suddenly became as high as 200% (negative), which was enormous. The relatively large deviation of VTEC corresponded to the storm effect, which prompted the penetration of electric field, followed by enhanced Fountain effects [29].

Seasonal VTEC variation
The incidence of seasons is due to the Earth rotation and its tilted axis by about 23.5 degrees toward, or away from the Sun through the annual orbit of Earth around the Sun. The seasonal average of VTEC values over the selected study period has been plotted in Fig .5. Clearly, the VTEC value started from 5 TECU and reached up to about 15 TECU during the March equinox during the daytime, while during the September equinox, the VTEC touched 11 TECU. Several studies across several regions have come up with similar findings, i.e., higher VTEC values during the March equinox as compared to the September equinox [34,39]. Within the solstice seasons, the June solstice showed the highest TEC values (about 13 TECU) while the December solstice reached up to only 9 TECU. The June solstice is greater which means disappearance of winter anomalies which has been discussed in many studies [38,40]