Ionospheric Behavior of fo F2 over Chinese EIA Region and Its Comparison with IRI-2016

: The ionograms, which were recorded by the ionosonde located at Pu’er station (PUR, 22.7°N, 101.05°E, Dip Latitude 12.9°N) in the Southwest of China in the year of 2016, were used to study the ionospheric behavior of the ordinary critical frequency of the F2 layer ( fo F2) in the region of the northern equatorial ionization anomaly. To verify the performance of the International Reference Ionosphere (IRI) over the Southwest of China, a comparative study of the observed fo F2 and the latest version of the International Reference Ionosphere (IRI-2016) was carried out. We found that the fo F2 in equinox months is greater than summer and winter. Moreover, a higher frequency of the observed bite-out of fo F2 in January and April than other months and the IRI-2016 cannot represent the bite-out of fo F2 in diurnal variations. Compared to the observations at Pu’er Station, the IRI-2016 underestimated fo F2 for most time of the year. The IRI with the International Radio Consultative Committee (CCIR) option overestimated fo F2 is higher than that with the International Union of Radio Science (URSI) option. Furthermore, the normalized root mean square error of fo F2 from the IRI-2016 with the CCIR option is less than that with the URSI.


Introduction
The ionosphere has a significant effect on the amplitude and phase of high frequency radio and satellite signals. To overcome the ionospheric effects on radio wave communication, studies of the ionosphere have become a hot topic. As the maximum of the electron density in the ionosphere appears in the F2 layer, this region is the most important for radio wave propagation. Therefore, the two most important parameters of the F2 layer, the ordinary critical frequency of the F2 layer (foF2), and the height of the maximum electron density in F2 layer (hmF2) are employed to study radio wave propagation in the ionosphere. Therefore, these two parameters are used to test the performance of the International Reference Ionosphere (IRI) over the Southwest of China and to verify the performance of the IRI in this region in this study.
As one of the most widely used global empirical models, the International Reference Ionosphere (IRI) was developed to represent global climatological ionospheric variations by the Committee on Space Research (COSPAR) and the International Union of Radio Science (URSI). The latest version of IRI is IRI-2016 (it can be accessed from https://ccmc.gsfc.nasa.gov/pub/modelweb/ionospheric/iri/iri2016/) [1]. The foF2 and hmF2 of the ionosphere can be calculated by the IRI with appropriate inputs. Therefore, many studies [2][3][4][5][6][7][8] carried out the comparison between them to test the performance of the IRI, which will help to improve the accuracy of the IRI in future work. Among those studies, Rao et al. [7] summarized comparative studies of observations and the IRI model from different sectors of the globe.
This study focuses on the diurnal and seasonal variations of the foF2 over the Chinese Equatorial Ionization Anomaly (EIA) region (Pu'er station, 22.7°N, 101.05°E, Dip Latitude 12.9°N) in the year of 2016. Since Pu'er station is a newly installed ionosonde station in the Chinese EIA region along the longitude of approximately 100°E, it can provide more data to carry out statistical studies of the ionosphere [9]. Therefore, a comparative study between the observations and the IRI model was performed to test the performance of the IRI model over the Southwest sector of China.

Data Set
Recently, a multifunctional High-Frequency radar, which is called Wuhan Ionospheric Sounding System (WISS) [10], was installed to monitor characteristics of the ionosphere at Pu'er in the Southwest of China. The ionosonde system implements vertical incidence ionospheric sounding continuously. In this study, the foF2 data were scaled manually by the ionoScaler [11] from ionograms recorded at Pu'er station in the year of 2016 to study ionospheric variations. The data resolution we used in this study is 30 min. For solar activity, the annual average value of F10.7 is 88.8 during the year of 2016. Therefore, this year (descending phase of solar cycle) is close to the low solar activity.

Magnetically Quiet Time Behavior
There are two options (URSI [12] and CCIR [13]) for evaluating foF2 in the IRI model and both of them are compared with the observed foF2 at Pu'er station. Figures 1a-c show the foF2 observed and calculated by the IRI with different options, respectively, as a function of day of year and local time (LT), in the year of 2016 at the Pu'er station. The blank spaces represent no data as the parameters have not been scaled by the ionoScaler. For the foF2 observed, it is shown that there is an evident semi-annual variation of the max of the foF2 every day. The foF2 in equinox months is greater than that in solstice months. Bite-outs in foF2 are a well-known ionospheric behavior and it is usually recognized that the maximum depression of foF2 is greater than 0.5 MHz [14]. For the diurnal variation of the foF2, the observed bite-out of foF2 appeared in January and April more frequently than other months. Figure 2 shows the bite-outs in foF2 that occurred on some days. It can be seen that the IRI cannot give the phenomenon. To give an explanation to the bite-out, Hirono and Maeda [15] and Rao [16] reported that the noontime bite-out occurred mainly located in two periods in Asia, one is in the forenoon and the other is in the afternoon. A forenoon bite-out is also reported by some studies [17,18] besides the noontime bite-out. Venkatesh et al. [19] studied noontime bite-outs of the Total Electron Content (TEC) and foF2 over India and Brazil at low latitudes (anomaly crest regions) and concluded that the mechanism of bite-outs around the crest related to some downward force that is different from that at the equator. However, Venkatesh et al. [19] did not specify what the downward force is. It needs further study to investigate the physical mechanism of the bite-out at low latitudes. with CCIR option (c) foF2 calculated by IRI with International Union of Radio Science (URSI) option (d) the normalized deviation of the observed foF2 to that derived by IRI with CCIR (e) the normalized deviation of the observed foF2 to that derived by IRI with URSI (f) monthly average (MA) of the observed foF2 (g) monthly average of foF2 calculated by IRI with CCIR option (h) monthly average of foF2 calculated by IRI with URSI option (i) normalized deviation of monthly average observation foF2 to that derived by IRI with CCIR (j) normalized deviation of monthly average observation foF2 to that derived by IRI with URSI.  To further reveal the difference between the observations and IRI-2016, the monthly average of foF2 was employed to make a comparative study. Figure 1f shows the contour plot of the monthly average (MA) of foF2 in the year 2016, as a function of local time and month. In summer, the diurnal peak value of foF2 mostly occurred at around 16:00 LT. It can also be seen from Figure 1f that the maximum value of foF2 in diurnal variation occurred more and more early from summer to winter. Figure 1g,h show the monthly average of foF2 calculated by IRI with both options. They can also represent the semi-annual variation of foF2. For the diurnal variation of foF2, the normalized differences of the monthly averages of foF2 are calculated by As shown in Figures 1 i,j, they illustrate that the IRI-2016 underestimated the foF2 from the afternoon to mid-night period. The different value is much less during the pre-sunrise and morning period than the afternoon period. For the seasonal variation of foF2, the different value is much greater in equinox months than the solstice months. Figure 1 i,h also shows that the minimum difference between the observations and IRI-2016 occurred in summer.
The diurnal variation of the monthly average of the foF2 from the observation and IRI-2016 along with their discrepancies, are presented in Figure 3.  the IRI with CCIR overestimated foF2 is higher than that with URSI in December. However, it needs more validation since the valid data in December are less than the other months. The Normalized Root Mean Square Errors (NRMSEs) of the URSI and CCIR in foF2 are 0.2698 and 0.1147, respectively. The results show that the performance of the IRI-2016 with the URSI is better than that with the CCIR in our study. We also computed the mean of residuals, standard deviations and correlation coefficients between the observations and IRI-2016. As shown in Table 1, the monthly average with both options are almost the same for the correlation coefficients. However, IRI with URSI is better than that with CCIR for 1-h resolution. For both data resolution, the IRI-2016 with the CCIR option has a smaller standard deviation than that with the URSI. So as the mean of residuals, IRI-2016 with URSI is better than that with CCIR. Thus, the IRI-2016 with the URSI performed a little better than the CCIR in prediction of the foF2 since the URSI option is a little better than the CCIR.

Magnetically Disturbed Period
Geomagnetic storms are disturbances in Earth's magnetic field which occur when a coronal mass ejection or a high-speed solar wind sweeps past the Earth. It can result in many hours of vibrant auroras at high, middle, and lower latitudes. When the ionosphere is affected by the geomagnetic storm, an ionospheric storm occurs. So it is well known that the ionosphere can be affected by a geomagnetic storm through the thermosphere [20][21][22][23][24][25]. The variation of the foF2 can be used to analyse the ionospheric storm. Thus, in order to study variations of the foF2 during geomagnetic storms, a magnetic storm that occurred on 11th October, 2016 was considered in this study. The measure of the disturbance was computed by where the oF2 refers to the monthly median of the foF2. In this study, we consider that it is a quiet ionosphere when the variability of 2 foF ∆ established within±20%, and it indicates the positive and negative storm effect with that 2 foF ∆ ranges above 20% and below -20%, respectively, in the ionosphere [26,27]. is also calculated by the IRI with the foF2 STORM model [28,29]. However, neither of the IRI-2016 values with both the CCIR and URSI can represent the corresponding changes, as shown in Figure 4. Recently, Rao et al. [7] carried out a comparative study between the IRI-2016 with ionospheric observations over another Chinese EIA station (Guangzhou, Geog. Lat. 23.10°N, Geog. Long. 113.40°E, dip, Lat. 13.49°N). In this study, Rao et al. [7] found that the IRI-2016 underestimated the foF2 in winter and equinoxes, and overestimated the foF2 in summer. However, the IRI-2016 almost underestimated foF2 values in the whole year of 2016 at Pu'er station in our study. For the temporal variation of the foF2, Rao et al. [7] found that the discrepancies were greater during the forenoon hours and smaller during the afternoon hours. However, it can be seen from Figure 1i,j that the greater discrepancies occurred in the afternoon hours at Pu'er station. All these results show that longitudinal variations of the ionosphere also have significant effects on the discrepancy between ionospheric observations and the IRI models.
Although a good agreement between ionospheric observations and the IRI models has been reported by many studies [7,30], its discrepancies cannot be ignored [7,31]. Especially, that electrical field and neutral winds play significant roles in the ionosphere over the equatorial and low latitudes. As a result, even the latest IRI model (e.g., the IRI-2016) also cannot remove these discrepancies with the ionospheric observations. To overcome this issue, a higher coverage of the ionosondes around the globe and the proposal of more practical physical models may be necessary for the IRI model.

Conclusions
In this study, we investigated the ionospheric behavior of foF2 at the northern equatorial ionization anomaly along the longitude of about 100°E and carried out a comparative study in foF2 between the observations and IRI-2016. The major findings are summarized as follows: (1) For seasonal variation, the foF2 in equinox months is greater than in solstice months. The average value of the foF2 in summer is the lowest value. However, the monthly average of the observed foF2 is the lowest in November, while it is in December for the IRI-2016.
(2) Mostly, the IRI-2016 underestimated the foF2 compared with the observations in the present study. The IRI-2016 with the URSI performed a little better than the CCIR in prediction of the foF2.
(3) The observed bite-out of foF2 in January and April occurred more frequently than other months and the IRI-2016 cannot represent the bite-out of foF2 in diurnal variations.