A model of the deviation between IRI-2016 and ionospheric TEC observation based on GISTM at low latitude Indonesia region

The diurnal variation pattern of total electron content(TEC) from the International Recurrence Ionosphere (IRI-2016) model is generally is good agreement with observational data at all latitudes. However, at low latitudes the IRI TEC is not as accurate as in midlatitudes. Some reports showed that the IRI-2016 TEC model at the maximum and minimum solar activities at low latitudes tends to be underestimate. In this paper, TEC from GISTM (GPS Ionospheric Scintillation and TEC Monitoring) that is installed at Pontianak station (0.03° S;109.33°E geomagnetic latitude 9.7°S ) used for validation and evaluation of IRI-2016 TEC model over the low latitude Indonesian sector. Data from the 2012 - 2014 period representing solar maximum and data from 2018 representing solar minimum, were used in this study. We also developed VTEC deviation (ΔTEC) model by using multiple linear regression and tested the model by using data 2019. Results show that IRI-2016 generally reproduces the diurnal variations pattern but underestimates the observation data for many hours each day especially during maximum solar activity 2012-20114. The highest deviation in the equinoctial months and lowest during the June and December solstice. The deviation (ΔTEC) model showed good agreement with observation data on January and December but not for other months especially on equinox months. Testing of the VTEC deviation model using 2019 data did not show significant improvement compared to the IRI-2016 itself, which generally produces good TEC prediction in most of the months except for equinox months.


Introduction
The upper atmosphere contains electrons and ions at altitudes above 50 km up to 1000 km known as the ionosphere layer. The density of electron and ions reaches maximum level at an altitude of about 300 to 400 Km. These electrons and ions are generated by photo-ionization process of constituent molecules and atoms such as nitrogen and oxygen. Changes in electron and ion density are influenced by solar activity and have been reported to have an impact the propogation of electromagnetic waves especially radio waves. Signals from communication and navigation satellites are affected by the induced ionspheric delays [1]. The time delay of trans ionospheric satellite signal is proportional to the total electron content (TEC) [2] [3]. Therefore, TEC is an important parameter that determines the success rate of radio wave propagation in satellite communication and accurate position solutions in satellite navigation applications. The time delay of trans ionospheric satellites is a function of TEC 2 unit (TECU) and operating frequency. In the L band GPS/ GNSS frequency satellite, 1 TECU will introduce a pseudorange delay of 0.163 meters and 0.267 meters in L1 and L2 frequencies respectively.
The IRI-2016, is an empirical global model for ionospheric parameters which is a refinement of previous years' models. The output of the model includes the electron density in the ionosphere layer, the ionospheric plasma frequency starting from D layers, E, F and F2 (foF2), the height of the ionospheric layer, the total electron content (TEC) and other parameters.  [9]. From those reports, the comparison or validation of the IRI model against TEC observation data from middle to low latitudes shows that the pattern of TEC variation agrees with the observational data at all latitudes. The level of TEC values is quite good in middle latitudes but not for low latitudes. However, the latest report by [10] and [11] regarding the response of the IRI-2016 TEC model at the maximum and minimum solar activities showed the TEC level at low latitudes tends to be lower than the observation data. In this paper, the TEC from GISTM (GPS Ionospheric Scintillation and TEC Monitoring) installed at Pontianak (0.03 o S;109.33 o E) was used for validation and evaluation IRI-2016 TEC model. The operation of GISTM at Pontianak often suffers from interruption sothere are a lot of missing data. Later, a model will be developed to describe a relationship between IRI-2016 TEC model and observation. TEC derived from code (pseudorange) measurements in equation (1) are noisier. This noise must be removed. The absolute TEC value can be expressed as,

Material and Method
where is the absolute TEC value, and are the differential (inter-frequency) biases within the satellite and the receiver. These biases typically vary by satellite and can change over time. The calculated satellites biases can be downloaded from http://aiuws.unibe.ch/ ionosphere/p1c1.dcb.). and are errors caused by the presence of multipath and background noise. The GPStation-6 hardware provides (ISMRAWTEC log) and carrier phase (ISMREDTEC log) TEC measurements. The use of ultra-stable oven-controlled crystal oscillator (OCXO) and the narrow delay-locked loop (DLL) bandwidths (BW) by the hardware greatly reduces the noise contribution in raw TEC measurements. TEC calibration procedure of GSV4004B is done in post-processing mode by taking the minimum TEC value during pre-dawn. This allows us to determine the receiver bias offset to be subtracted corrected minus about 7 TECU. In this study, we used TEC measurement at 60 seconds resolution (TOW) convert to vertical TEC (VTEC) at the Ionospheric Pierce Point (IPP). The slant to vertical conversion factor in terms of the zenith angle at the IPP, χ', and the zenith angle at the receiver position, χ [12], as follows: where R e is the radius of the Earth, and h m is the height of the ionospheric layer, assumed here to be 350 km. In this work, we used the GISTM TEC an hourly and daily basis. To obtain the daily profile, an hourly average was applied to TEC data with elevation mask 30 o . A median TEC from 8 to 10 satellites channel every 60 seconds was used as the representative 1-minute TEC value, then the hourly mean was obtained by averaging the 1-minute data every 60 minutes.

Result and Discussion
Data Data used in the analysis are monthly mean diurnal VTEC. Figure 1 In this plot, the two-peak maximum of VTEC during the equinox, month March-April and September-October, can be clearly seen. These two peaks are not symmetrical, as the March equinox peak is higher than the other. [13] [14] reported that these equinoctial asymmetries present at low latitude are influenced by the thermosphere and ionospheric dynamics processes and meridional wind control as well. Also, we can see that  Unfortunately, there are no observation data during September and December in 2012. As we can see the VTEC generally follows the diurnal variations but underestimates the observation data in absolute value for many hours of the day. The hourly VTEC are enhanced during the equinoctial months, with the March equinox higher than September equinox, and depleted in the solstice months, with the December solstice higher than the June solstice. Also, the VTEC attains its maximum values between 9:00-15:00 UT (12:00-18:00 LT), and minimum between 21:00-24:00 UT (00:00-3:00 LT). In the right panel in figure 3 we present the deviation between the GISTM and IRI-2016 predicted values in the monthly mean VTEC. The highest deviation occurred variation in the equinoctial months (Mar-Sept), about 50%, and lowest during the solstices, although the deviation is slightly higher in Previous reports by [15] and [10] showed IRI-2016 model tends to underestimate VTEC at low latitude. due to the presence of equatorial ionospheric anomaly. The deviation (ΔTEC) model by using multi-linear regression in equation 7, obtained a set of regression coefficient values for each month, from January to December as shown in table 1. These coefficients are used to predict the deviation of VTEC between IRI-2016 and GISTM data, then those values were added to IRI-2016 TEC as correction. Testing ΔTEC model using observation data from 2019 is shown in figure 5. As shown in figure 5, ΔTEC model good agreement with observation data from January and December but not for other months especially on equinox month. Generally, the ΔTEC model did not perform much better than the original IRI-2016 model output, as shown in figure 6. As we can see in figure 6, IRI-2016 good agreement in most of the month except for equinox month. From these results. the attempt to build a model TEC deviation using linear regression method has not been able to improve, since even the IRI-2016 model can provide better prediction as shown in figure 6.

Concluding Remarks
In this article, we have validation VTEC IRI-2016 model using the GISTM data at a low latitude station in Pontianak, Indonesia (0.03 o S;109.33 o E geomagnetic latitude 9.7 o S ) from 2012-2014 and 2018. The comparison was investigated in terms of monthly mean diurnal VTEC variations. Results show that predicted VTEC from the IRI-2016 model generally follow the diurnal variations but are underestimation in most of the hours in the observation data. The highest deviation (ΔTEC) in the equinoctial months (Mar-Sept) and lowest in June and December solstices. We also developed a VTEC deviation (ΔTEC) model by using multiple linear regression to improve the IRI-2016 output. The model good agreement with observation data in January and December but not for other months