*Article* **Assessment of Polar Ionospheric Observations by VIPIR/Dynasonde at Jang Bogo Station, Antarctica: Part 1—Ionospheric Densities**

**Eunsol Kim 1, Geonhwa Jee 1,2,\*, Young-Bae Ham 1,2, Nikolay Zabotin 3, Changsup Lee 1,2, Hyuck-Jin Kwon 1, Junseok Hong 4, Jeong-Han Kim <sup>1</sup> and Terence Bullett <sup>5</sup>**


**Abstract:** Vertical incidence pulsed ionospheric radar (VIPIR) has been operated to observe the polar ionosphere with Dynasonde analysis software at Jang Bogo Station (JBS), Antarctica, since 2017. The JBS-VIPIR-Dynasonde (JVD) provides ionospheric parameters such as the height profile of electron density with NmF2 and hmF2, the ion drift, and the ionospheric tilt in the bottomside ionosphere. The JBS (74.6◦S, 164.2◦E) is located in the polar cap, cusp, or auroral region depending on the geomagnetic activity and local time. In the present study, an initial assessment of JVD ionospheric densities is attempted by the comparison with GPS TEC measurements which are simultaneously obtained from the GPS receiver at JBS during the solar minimum period from 2017 to 2019. It is found that the JVD NmF2 and bottomside TEC (bTEC) show a generally good correlation with GPS TEC for geomagnetically quiet conditions. However, the bTEC seems to be less correlated with the GPS TEC with slightly larger spreads especially during the daytime and in summer, which seems to be associated with the characteristics of the polar ionosphere such as energetic particle precipitations and large density irregularities. It is also found that the Dynasonde analysis seems to show some limitations to handle these characteristics of the polar ionosphere and needs to be improved to produce more accurate ionospheric density profiles especially during disturbed conditions.

**Keywords:** polar ionosphere; VIPIR; Dynasonde; Jang Bogo Station (JBS); Antarctica

#### **1. Introduction**

The ionospheric density is principally governed by solar EUV radiation, but the polar ionospheric density exhibits various characteristic features due to the additional magnetospheric forcings such as electric fields and energetic particles as well as the unique geometry of nearly vertical geomagnetic field line (e.g., [1,2]). The energetic particles precipitate into the polar upper atmosphere and produce additional ionization mainly in the auroral region but also in the polar cap region. The soft electron precipitation also produces the F-region ionization in the cusp region. The polar ionospheric density is further redistributed by the plasma convection induced by the magnetospheric electric field, which transports the dayside plasma to the night side to produce the characteristic features of the polar cap ionosphere, such as the tongue of ionization (TOI) and the polar cap patch (e.g., [3,4]). The ionospheric density distributions in the polar region are closely associated with coupling processes between the ionosphere and the magnetosphere, being strongly

**Citation:** Kim, E.; Jee, G.; Ham, Y.-B.; Zabotin, N.; Lee, C.; Kwon, H.-J.; Hong, J.; Kim, J.-H.; Bullett, T. Assessment of Polar Ionospheric Observations by VIPIR/Dynasonde at Jang Bogo Station, Antarctica: Part 1—Ionospheric Densities. *Remote Sens.* **2022**, *14*, 2785. https://doi.org/ 10.3390/rs14122785

Academic Editor: Fabio Giannattasio

Received: 6 May 2022 Accepted: 7 June 2022 Published: 10 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

affected by solar wind conditions, which makes it difficult to understand and requires routine monitoring of the states of the polar ionosphere. Another important factor of the polar ionosphere's dynamics is its atmospheric wave activity.

The world-wide network of ionosondes has a relatively good coverage over the globe, but it is sparse at high latitudes, particularly in the southern hemisphere. Other ground-based observations for the polar ionospheric density are also mostly located in the Arctic: for example, incoherent scatter radars (ISRs) at Poker Flat (Alaska), Resolute Bay (Canada), Sondrestrom (Greenland), Kiruna (Sweden), Tromsø and Svalbard (Norway), and Sodankylä (Finland). On the other hand, there are only a few observation sites for the ionospheric density in the southern polar region. Only one ISR has been operational in Syowa station, and a few digisondes are operational, for example, at Zhongshan station and Casey station. Recently, an ionospheric sounding system was installed at Jang Bogo Station (JBS), Antarctica, and started operating in 2017 to collect ionospheric parameters in the southern polar region. The sounding system is called the Vertical Incidence Pulsed Ionospheric Radar (VIPIR), and it utilizes the Dynasonde mode of operation and the Dynasonde analysis software to conduct echo recognitions and ionogram inversions to produce ionospheric parameters such as bottomside ionospheric electron density profiles with error bars, the F-region peak density (NmF2) and the peak height (hmF2), estimates for the ion drifts, and ionospheric tilts [5–8]. The JBS-VIPIR-Dynasonde (JVD) is distinguished from a conventional digital ionosonde, for example, the digisonde series from Lowell Digisonde International, which is one of the most widely operated digital ionospheric sounding systems around the globe (e.g., see https://www.digisonde.com/index.html, accessed on 3 March 2022). The data acquisition and analysis procedures in the Dynasonde are performed with minimized assumptions and no data pre-processing such as Fourier transform is applied to reduce the loss of precision in the phase-based physical parameters of the radio echoes such as the line-of-sight Doppler, range resolution, and angles of arrival of received signals. This approach also allows the application of sophisticated upper-level analysis techniques producing parameters of the Traveling Ionospheric Disturbances (TIDs), of kilometer-scale irregularities, and vector velocities of the isodensity ionization contours, all from a single standard ionogram mode. All of this data processing is performed autonomously and in real time. Historically, there have been a few Dynasondes in the polar regions such as at the EISCAT Tromsø and Svalbard observatories, at the IRF Kiruna and Lycksele stations in the northern polar region, and at the Halley base in Antarctica. However, those at Lycksele and at the Halley base are no longer in operation. The JVD is currently the only available Dynasonde for the ionospheric observation in the southern polar region. In order to evaluate the overall quality of the ionospheric data obtained from JVD, we compare the JVD bottomside ionospheric densities with the independent measurements of total electron content (TEC) from a Global Positioning System (GPS) receiver simultaneously operated at JBS. We fully understand that parameters measured by the two instruments are not exactly the same. However, the GPS TEC is the only available measurement related to the ionospheric density to be compared with JVD data at the moment. There have been a few comparison studies between ionosonde and GPS TEC measurements. However, these are mostly conducted at low and middle latitudes [9–11], and no comparison has been performed in the polar region. As for the other ionospheric parameter such as ion drift and irregularity, there will be a separate study on the comparison with the SuperDARN radar observations around JBS. In the following sections, we briefly introduce the observations of JVD densities, as well as GPS TEC, and the results of the comparison will be presented.

#### **2. Data**

#### *2.1. The Observation of Ionospheric Density from JVD*

The VIPIR was installed at JBS (74.62◦ S, 164.23◦ E geographic coordinates and 79.87◦ S geomagnetic latitude), Antarctica, in 2015, but the ionospheric data with a high temporal resolution of 2 min were not available until 2017, when the installation and subsequent test operations were complete. The Dynasonde data analysis procedure estimates the ionospheric quantities by using the echo information, including two angles of arrival, phasebased group range, line-of-sight Doppler, polarization, and amplitude. Every sounding session produces a long list (up to a few thousand) of detected radio echoes. The echo information is inverted to the height profiles of ionospheric density, of the vertical Doppler, and of the two tilt components via the NeXtYZ inversion procedure within the Dynasonde analysis software [8] to produce the ionospheric parameters. More details of the VIPIR hardware and Dynasonde software can be found in Ham et al. [7].

In principle, the ground-based ionospheric sounding technique is optimized to measure the height profiles of the bottomside ionosphere when the density monotonically increases to the F-region peak with the increase in height. Note that it is not intended for TEC measurements since no direct information of the topside ionosphere is available from the ground sounding technique. It is also important to notice that the ionospheric density profiles can be severely disturbed, especially in the polar region during magnetic storm/substorm or sporadic E. For example, when the E-region peak density (NmE) exceeds the NmF2 during auroral events, the ionosonde can theoretically observe the density only up to the E-region peak height (hmE) around 110–120 km and is unable to observe the F-region ionosphere. The ionospheric density with so-called E-layer-dominated ionosphere (ELDI) has been frequently observed in the auroral region, for example, in Tromsø and Svalbard in the winter for the solar minimum period [12]. During the ELDI events, the radar signals from the ionosonde are blocked and cannot reach above the E-layer peak heights, and the F-region density profiles can only be observed by other observation techniques such as ISRs [2,12] and GPS radio occultation from COSMIC or CHAMP satellites [13,14]. Since the JBS is mostly located in the polar cap during the nighttime, at the vicinity of the poleward boundary of the auroral oval at dawn and dusk, and in the cusp region near magnetic local noon [15], the ionospheric density profiles can be affected by geomagnetic activities. It is found that the JVD observations can be erroneous when the ionosphere is severely disturbed, as will be shown later in the paper. In order to minimize the effects of the severely disturbed ionosphere as well as the ELDI events from JVD observations, we perform the assessment study only for geomagnetically quiet times (Kp < 2), during which about 80% of the data were utilized. The ionospheric density parameters including NmF2 and bottomside TEC (bTEC) from the JVD observations are compared with the GPS TEC measurements during the solar minimum years of 2017 to 2019. The mean F10.7 index was about 72 solar flux units (sfu) for the three-year period. The JVD observations for disturbed times will be briefly discussed in Section 4.

#### *2.2. The Observation of GPS TEC at JBS*

The National Geographic Information Institute (NGII) has been operating a dualfrequency GPS receiver at JBS for geodetic survey, providing a slant TEC (STEC) along a ray path between the GPS satellites and the receiver. The STECs were derived by the Space Geodesy group at the Korea Astronomy and Space Science Institute (KASI) with a 30 s time interval using a geometry-free combination as below:

$$\text{STEC}\_{\text{P}} = \frac{1}{40.3} \frac{f\_1^2 f\_2^2}{f\_1^2 - f\_2^2} (P\_2 - P\_1), \tag{1}$$

$$\text{STEC}\_{\text{L}} = \frac{1}{40.3} \frac{f\_1^2 f\_2^2}{f\_1^2 - f\_2^2} (\phi\_1 \lambda\_1 - \phi\_2 \lambda\_2), \tag{2}$$

where *f* and *λ* are the signal frequency and wavelength, respectively, and *P* and *φ* are pseudorange measurements and carrier phase measurements, respectively. The unit of TEC is TECU (1 TECU = 1016 electrons/m2). The STECP is absolute but noisy, while STECL is more precise but includes ambiguities. Therefore, the final STECs are derived through a phase leveling from both code and carrier-phase-based STECs within an arc of each satellite [16]. Then, the STECs can be converted into vertical TECs (VTECs) using a single layer approximation as follows:

$$\text{VTEC} = \sqrt{1 - \left(\frac{R\_E}{R\_E + h\_{ipp}} \cos(\epsilon l)\right)^2} \times (\text{STEC} - b\_r - b\_s), \tag{3}$$

where *RE* and *el* are the Earth's radius and the elevation angle of the satellite, respectively, and *hipp* is the height of the ionospheric pierce point (IPP), which is set to be 350 km in altitude in this study. The *bs* and *br* are the differential code biases (DCBs) of the satellites and the receiver, respectively. The satellite DCBs are provided by the Center for Orbit Determination in Europe of Astronomical Institute University of Bern, and the receiver DCBs are calculated from a single-receiver method [17].

The STEC to VTEC conversion is inherently ambiguous. As noted above, it invokes a thin-shell model of the ionosphere. This model is easy to handle, but it is very far from being realistic, as it compresses electron content from the altitude range extending up to about 20,000 km into the single point at about 350 km (the effective altitude of the shell). The unrealistic character of the thin-shell model may introduce substantial uncertainty into the VTEC estimates. Even though the phase leveling is applied for more precise TEC and both satellite and receiver DCBs are corrected, it is still challenging to estimate TEC in Antarctica. Extremely low electron density in high latitude regions can sensitize the TEC estimation because even a small amount of TEC changes in the phase-leveling process could be critical. Furthermore, strong ionospheric density irregularities causing scintillations or loss of signal lock are frequent at JBS during the December solstice [18], and they can also affect the phase-leveling accuracy, owing to the changes in ambiguities. This means that using VTEC data as a truth for assessing an ionospheric sounding data has a certain limitation.

A lack of the TEC measurements with high elevation angles greater than about 70◦ due to the inclination of GPS satellites causes large displacements among the IPPs within a specific time window. At the same time, there are various ionospheric density structures in the polar region, causing large density variations. For example, the density levels of TOI or the polar cap patch in the polar ionosphere can be at least two times greater than the ambient electron density [19]. Hence, the resulting large density gradients with large displacements of IPPs in the GPS TEC measurements should be considered in the comparison with the JVD densities. Even though a cut-off elevation angle of 15◦~30◦ is typically applied to avoid multipath effects [20,21], it may not be appropriate for representing the VTECs over the JBS. Therefore, we used the larger elevation cut-off angle of 50◦ to increase the accuracy of the averaged VTECs at the zenith of JBS during the times when the JVD ionospheric density measurements were available. Figure 1a shows the spatial distribution of IPPs around the JBS for the GPS TEC measurements with the elevation angles greater than 50◦ for a day (DOY 291, 2018) as an example. Note that the spatial distributions of IPPs are mostly located at the lower latitudes of the JBS, considering the 55◦ inclination of GPS satellites. They are mostly located within less than 2.5◦ around the JBS (i.e., within about 250 km of the JBS). This spatial difference from the JBS may still be large and may not be negligible for the comparison between JVD electron densities and GPS TEC measurements, so it needs to be considered to interpret the results of the comparison. There are about 1~4 GPS satellites located around the JBS for a 10 min time interval, during which the JVD densities and GPS TEC measurements are averaged for the comparison. Figure 1b shows the traces of the GPS satellite paths for 10 min in the azimuth-elevation coordinate.

**Figure 1.** Spatial distribution of the ionospheric pierce points (IPPs) (black dots) for the TEC measurements over the JBS (blue star) from the GPS satellites with the elevation angles greater than 50◦ for DOY 291, 2018 (**a**) and four GPS satellite paths for 10 min from 23:50 to 23:59 UT in azimuth-elevation coordinate (**b**) over JBS.

#### **3. Comparisons between JVD-Observed Densities and GPS TEC Measurements**

Both JVD and GPS measurements are valuable additions to the sparse ionospheric observations in the southern polar region. Bearing in mind all limitations of such approach, we initially compared the JVD NmF2 and bTEC with VTEC measurements from a colocated GPS receiver at JBS for geomagnetically quiet times during solar minimum years. Figure 2 shows the scatter plot of 10 min averaged GPS TEC vs. JVD NmF2 (left) and bTEC (right) from 2017 to 2019 from the top to bottom panels. The linear Pearson correlation coefficients for each case are shown at the upper-right corner of each panel. Note that both NmF2 and bTEC are supposed to be somewhat correlated to the GPS TEC, considering how much the F-region peak density contributes to the GPS TEC (e.g., [9,22]). It was found in Figure 2 that the JVD NmF2 is highly correlated with GPS TEC, but the JVD bTEC shows slightly lower correlations, which seems to imply that the JVD measurement of the F-region peak density is more accurate than the measurement of bTEC calculated from the density profile of the bottomside ionosphere. However, we should remember that bTEC has the more complex nature compared to NmF2: it is computed from the density profiles that are estimated by the NeXtYZ inversion procedure using not only the observed information but also the physics-based and empirical models for the daytime D-region (or nighttime E-region) ionization and for the E-F valley region [8], which may deviate from the true ionosphere. The bottomside TEC involves an effective thickness *Hb* of the bottomside ionosphere (bTEC = NmF2 × *Hb*), and the VTEC involves an effective thickness *Ht* of the entire ionosphere (VTEC = NmF2 × *Ht*). These two TECs may not correlate well with each other, especially in the polar region. Nonetheless, the correlations between the two independent measurements are fairly strong.

**Figure 2.** The scatter plot of JVD NmF2 (**left**) and bottom bTEC (**right**) vs. GPS TEC averaged for 10 min during quiet time from 2017 (top) to 2019 (bottom). The linear correlation coefficients for each case are shown at the upper right-corner of the panels.

The next comparisons with the GPS TEC measurements were performed for the international quiet days (IQDs) during the study period. Figure 3 shows the diurnal variations in 10 min. averaged JVD bTEC and NmF2 and GPS TEC for eight IQD cases (mean Kp~0.4). We chose the cases when the geomagnetically quiet condition persists for at least three consecutive days from the lists of the five quietest days for each month provided by the World Data Center for Geomagnetism, Kyoto. The diurnal variations in JVD NmF2 very closely follow the variations in GPS TEC, although the IPPs of the GPS satellites mostly exist at lower latitudes. The JVD bTECs also show similar diurnal variations from the JVD NmF2 and GPS TEC but with a somewhat larger spread. The larger spread in the JVD measurements may be associated with the ionospheric structures causing spread F on the ionogram. Since the spread F ionograms show multiple refractive scatterings, it may complicate the analysis of the ionogram data [23]. The TID activity caused by atmospheric gravity waves is the most common mechanism of the spread F ionogram. Shimazaki [24] reported that the spread F can also be caused by charged particles precipitating into the F-region ionosphere at high latitudes with lower energy than auroral particles. According to their study, the high-latitude spread F appears even in the sunlit conditions, while it mainly occurs during nighttime at low and middle latitudes. When the spread echoes appear on the ionogram, it may be a challenge for automatic scaling to produce the realistic ionospheric density profiles. In addition, the ionospheric irregularities have been reported to frequently occur especially in summer in the southern polar ionosphere [18]. Figure 3 also shows a tendency of a larger spread (particularly in bTEC) during the daytime in the summer season (see Figure 3f). As will be discussed in the next section, the Dynasonde analysis procedure seems to show some limitations to estimate the ionospheric density profiles, in particular when the characteristic features of the polar ionosphere exist: for example, energetic particle precipitations and ionospheric density irregularities.

**Figure 3.** The diurnal variations of 10 min averaged JVD bTEC (green), NmF2 (blue), and GPS TEC (yellow) for eight IQDs. Note that LT = UT + 11 at JBS. The eight IQD cases include DOY 302–304, 2017 (**a**); DOY 308–310, 2017 (**b**); DOY 139–141, 2018 (**c**); DOY 291–293, 2018 (**d**); DOY 319–321, 2018 (**e**); DOY 347–350, 2018 (**f**); DOY 54–57, 2019 (**g**); DOY 107–111, 2019 (**h**).

When the solar-production is negligibly small near the polar winter (see Figure 3c), the magnitudes of JVD bTECs are very close to the GPS VTECs for DOY 139–141 (early winter), 2018. In summer season, however, the differences between the two TECs become large, especially during daytime (see Figure 3f). Typically, the bottomside ionospheric TEC is known to contribute to GPS TEC by about 10~40% in the low and middle latitude ionospheres [10]. Figure 4 shows the annual variations in the ratios of 10 min averaged JVD bTECs to GPS VTECs for quiet conditions during the study period of 2017 to 2019. Each pixel in the figure indicates the total amount of data for three years within a bin of a day and a 0.1 ratio interval. The ratios are mostly less than about 0.5 in the austral summer season, peaking at about 0.35, but they tend to be slightly enhanced up to about 0.7 in the austral winter season. The daily medians, as depicted by the yellow line, are the average values per each DOY for the three-year period. However, the ratios are sometimes greater than 1, indicating that the JVD bTECs occasionally exceed the GPS VTECs, particularly in the winter season. Note that the enhanced contribution of bTEC to GPS TEC may be related with the E-region density enhancements by stronger energetic particle precipitations in winter season [25–27]. Moreover, the solar production mainly responsible for the F-region density is nearly absent in polar winter, which reduces the contribution of the F-region density to GPS TEC. This aspect of the ratio is a unique characteristic of the polar ionosphere. Nonetheless, occasions of the large ratio (greater than 1), while statistically insignificant, clearly imply that there might be some quality issues in measurements from both GPS receiver and JVD.

**Figure 4.** The annual variation in the ratios of JVD bTECs to GPS VTECs averaged for 10 min during the period of 2017 to 2019. The yellow line indicates daily median values, and the bluish colors indicate the numbers of data in each bin.

Figure 5 shows the diurnal variations in the hourly mean GPS VTEC (yellow) and JVD bTEC (green) for each month from 10 min averaged VTEC measurements with standard deviations (error bars) during quiet times for the period from 2017 (top) to 2019 (bottom). The GPS TEC data were not available in April and May in 2017. Note that LT = UT + 11 at JBS. The mean values of the bottomside ionospheric TEC are mostly well below the GPS TEC measurements. However, the differences between GPS TEC and JVD bTEC become smaller at night and especially in winter when the solar production is nearly absent but the additional production by auroral precipitation exists in the polar nighttime E-region ionosphere. As will be discussed later in the next section, however, the JVD seems to have some issues with regard to the estimation of the ionospheric density profiles during auroral events, which may be the reason for slightly larger standard deviations in JVD bTECs. Nonetheless, the climatological characteristics of JVD bTEC are generally consistent

with the GPS TEC measurements in the polar region: the ionospheric densities are greater during daytime than nighttime and in summer than in winter during solar minimum years. Both measurements also show the variations with the solar activity, decreasing toward the solar minimum from 2017 to 2019.

**Figure 5.** The diurnal variations in mean GPS TEC (yellow) and JVD bTEC (green) from 10 min– averaged values for each month are presented for a three-year period. Note that LT = UT + 11 at JBS.

#### **4. Common Type of Misestimated Ionospheric Density Profiles from the Dynasonde**

One of the characteristic features of the polar ionosphere is that the E-region peak density can be equivalent or even greater than the F-region peak density, due to the additional production by energetic particle precipitations. The so-called E-layer dominated ionosphere (ELDI) frequently occurs at high latitude in winter for solar minimum and geomagnetically disturbed times when the F-region density is minimized due to the reduced solar production, but the E-region density is increased by the enhanced energetic particle precipitations (e.g., [2,12]). Since the ionospheric sounding technique cannot observe the ionosphere above the density peak height whether it occurs in the F-region or in the E-region, the observed ionospheric densities from the JVD should be carefully examined, in particular when the energetic particle precipitations exist. During disturbed times, it is well known that the signals can be blocked by enhanced E-region density (e.g., ELDI), attenuated by increased D-region densities (e.g., polar cap absorption), and experience scintillations by density irregularities [18,28,29]. When these happen, the measurements from the ionospheric sounding must be affected by them, and the resulting ionospheric density profiles may not represent the state of the ionosphere well. Figure 6 shows an example of the ionogram produced by JVD at around 21.5 LT on 6 May 2018 (F10.7 = 68.4 sfu, Kp = 3), at which there was an auroral event over JBS. The ionogram and resulting density profile in Figure 6 imply that the echoes are reflected by the enhanced E-region density caused by auroral particle precipitation at around 100–120 km altitude: that is, the F-region could not be observed by JVD since the F-region density is probably smaller than the E-region density, which is indicated by there being no reflected signal above the E-region.

**Figure 6.** An example of the ionogram and resulting density profile obtained by JVD in winter at night on 6 May 2018 when the F10.7 was 68.4 sfu and the Kp index was 3.

Note that this is an example of absolutely correct processing of available ionogram data. The Dynasonde analysis detected 998 radio echoes, calculated their physical parameters and classified them into 6 traces (5 of which were ordinary polarization, reflecting complex structure of the disturbed E region, and 1, #6, was extraordinary). The autonomous analysis has chosen trace #1 for the profile inversion and successfully obtained the E-region profile with reasonable error bars (The topside part of this profile is not an actual inversion, it is just a Chapman-model-based extrapolation, which, in this case, has little to do with the real upper ionosphere).

The described procedure is completely based on objective physical parameters contained in the list of detected radio echoes. Note that it does not use the poorly defined notion of "the leading edge" of the ionospheric reflections. Most of the time, it works very well. Sometimes, however, when the ionospheric structure is particularly complex, the autonomous analysis makes mistakes in trace selection. Figure 7 shows an example of the erroneous density profile estimated at around 03 LT on 15 August 2018 with an auroral event over JBS during low solar and moderately disturbed geomagnetic conditions (F10.7 = 70.6 sfu, Kp = 3). The autonomous analysis software successfully identified 4130 radio echoes and classified them into 22 traces. It should have selected traces #1 and #15 for further analysis. However, wrong traces #4, #7, and #14 were chosen instead of the trace #1 for the profile inversion and this resulted in an unusually thin density profile peaking around 250 km in height, which does not represent the actual ionospheric density profile. This erroneous density profile belongs to the type of the misestimated density profiles by Dynasonde analysis, mostly occurring during the auroral events. This example indicates that the trace selection within the autonomous Dynasonde analysis software does not work dependably when the aurora occurs. This is not to say that such problems are unique to JVD, or to Dynasondes in general. This kind of misestimation has been observed at other high-latitude Dynasonde locations and with other amplitude-based ionospheric sounding techniques. The phase-based approach, with multiple physical parameters of the radio echoes readily available for the analysis, definitely has more diverse tools for improving results of this sort, and this must be one of the directions of future work. The current version of the software includes the expert reprocessing capability when an experienced operator is able to correct the trace selection mistakes.

During such disturbed conditions as auroral events in the polar ionosphere, the ionospheric density profiles often deviate from typical mid-latitude ionospheric density profiles with an F-region peak at around 300 km and a smaller E-region density peak at around 120 km or no E-region at night. The disturbed density profiles in the polar ionosphere may still be a challenge in the ionospheric sounding techniques such as the Dynasonde, as well as the more conventional digisonde. In conclusion, the electron density profiles estimated by the Dynasonde are mostly in reasonable agreement with the independent GPS TEC measurements during undisturbed conditions, but caution is required when using some of the analysis products during disturbed conditions in the polar ionosphere.

**Figure 7.** An ionogram and the resulting density profile obtained by autonomous Dynasonde analysis on 15 August 2018 (F10.7 = 70.6 sfu, Kp = 3).

#### **5. Summary and Conclusions**

The ionospheric densities obtained from the JVD are assessed using GPS VTEC data, the only available independent measurement of the ionospheric densities at JBS, Antarctica. We fully understand that the two instruments do not exactly measure the same parameters. Moreover, the VTEC measurements at high latitudes are very difficult by themselves, and their results cannot be considered as the ground truth. This study was performed mostly for geomagnetically quiet times during solar minimum years from 2017 to 2019. The JVD NmF2 is well correlated with GPS VTEC measurements but the JVD bTEC shows slightly less correlations. The more detailed comparisons between JVD densities and GPS VTEC have been performed for international quiet days. Both JVD NmF2 and bTEC are generally in a good agreement with GPS TEC but the JVD bTECs show relatively larger spread, which may be associated with the diverse characteristic features of the polar ionosphere such as energetic particle precipitations and large density irregularities. Those same features create additional difficulties for estimates of proper density profiles. The median ratios of JVD

bTEC to GPS VTEC are about 0.5 on average but tend to be larger in winter. It indicates that the solar production is nearly absent in polar winter, but the effect of particle precipitation enhances the contribution of bottomside ionosphere to the total electron content in the polar region. Finally, it was found that the autonomous Dynasonde estimation of the ionospheric density profiles seems to have issues with trace selection during disturbed conditions, which need to be addressed in future versions of the Dynasonde analysis software.

**Author Contributions:** Conceptualization, G.J. and E.K.; methodology, E.K. and G.J.; software, E.K.; validation, G.J. and N.Z.; formal analysis, E.K.; investigation, E.K., G.J., N.Z., Y.-B.H., C.L. and H.-J.K.; resources, Y.-B.H., C.L., J.H., H.-J.K., J.-H.K., N.Z. and T.B.; data curation, E.K., Y.-B.H., C.L., J.H. and H.-J.K.; writing—original draft preparation, E.K.; writing—review and editing, E.K. and G.J.; visualization, E.K.; supervision, G.J. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the grant PE22020 from Korea Polar Research Institute (KOPRI).

**Acknowledgments:** Junseok Hong was supported by a basic research fund from the Korea Astronomy and Space Science Institute (KASI) (2022185010).

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


## *Article* **Ionospheric Behavior during the 10 June 2021 Annular Solar Eclipse and Its Impact on GNSS Precise Point Positioning**

**Juan Carlos Valdés-Abreu 1,2,†, Marcos A. Díaz 1,2,\*,†, Manuel Bravo 3,4,†, Juan Carlos Báez 5,† and Yohadne Stable-Sánchez 1,†**


**Abstract:** The main effects of the 10 June 2021 annular solar eclipse on GNSS position estimation accuracy are presented. The analysis is based on TEC measurements made by 2337 GNSS stations around the world. TEC perturbations were obtained by comparing results 2 days prior to and after the day of the event. For the analysis, global TEC maps were created using ordinary Kriging interpolation. From TEC changes, the apparent position variation was obtained using the postprocessing kinematic precise point positioning with ambiguity resolution (PPP-AR) mode. We validated the TEC measurements by contrasting them with data from the Swarm-A satellite and four digiosondes in Central/South America. The TEC maps show a noticeable TEC depletion (<−60%) under the moon's shadow. Important variations of TEC were also observed in both crests of the Equatorial Ionization Anomaly (EIA) region over the Caribbean and South America. The effects on GNSS precision were perceived not only close to the area of the eclipse but also as far as the west coast of South America (Chile) and North America (California). The number of stations with positioning errors of over 10 cm almost doubled during the event in these regions. The effects were sustained longer (∼10 h) than usually assumed.

**Keywords:** solar eclipse; ionosphere; precise point positioning; GNSS; total electron content; rate of total electron content index; Swarm satellite measurements; ionosonde; electron density

### **1. Introduction**

A solar eclipse is a natural phenomenon that occurs when the Moon moves in the way between the Sun and Earth, totally or partially blocking the Sun, casting a shadow over the Earth. Since the Sun is one of the major drivers of atmospheric effects, such as its ionization at high altitudes, its blocking produces several disturbances. The atmospheric effects of a solar eclipse have been the subject of extensive research, mainly in meteorological parameters, total column ozone, photochemistry, gravity waves, and ionospheric parameters [1]. Despite the large number of studies concerning eclipses, the event of a solar eclipse is still unique since it happens at different seasons, different times of the day, different locations, and under different synoptic and geomagnetic conditions [1–3]. In addition, with every new eclipse, the scientific community gains larger numbers and a variety of instruments, which allow us to revisit the proposed conclusions from previous eclipses.

**Citation:** Valdés-Abreu, J.C.; Diaz, M.A.; Bravo, M.; Báez, J.C.; Stable-Sánchez, Y. Ionospheric Behavior during the 10 June 2021 Annular Solar Eclipse and Its Impact on GNSS Precise Point Positioning. *Remote Sens.* **2022**, *14*, 3119. https://doi.org/10.3390/rs14133119

Academic Editor: Fabio Giannattasio

Received: 21 April 2022 Accepted: 13 June 2022 Published: 29 June 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

The ionosphere is directly affected since this atmospheric layer is produced by solar radiation. The total electron content (TEC) is a measure of the electron density in the ionosphere integrated along the line of sight, thus, an indication of its ionization. TEC can be obtained using a radio link between a satellite and the ground. Nowadays, the most common system delivering TEC measurements is the Global Navigation Satellite Systems (GNSS), which requires TEC measurements to improve the precision of position estimation. TEC is expressed in TEC units (TECu), where 1 TECu = 10<sup>16</sup> e/m2. The perturbation of the ionosphere can be analyzed through the variations of TEC. The main parameters of the TEC variations during the eclipses are the delay value (*τ*) relative to the maximum phase of the eclipse; its amplitude (A), which generally is a decrease; and the duration (Δ*T*) of the perturbation [4]. Since the Moon's shadow moves rapidly from west to east across the Earth at supersonic speed, the total eclipse lasts just a few minutes anywhere [5–7]. Previous works have reported the depletion of TEC after the onset of the partial solar eclipse and have presented values of A in percent (A[%]) that can reach up to −64% with *τ* from −30 to 180 min [8–11]. This delay has been interpreted as an indicator of the combined effect of the photochemical processes and plasma dynamics [1,12]. Some works have reported Δ*T* from 50 to 240 min [4,13]. However, some studies have reported even longer effects [14,15]. A historical summary of ionospheric responses to solar eclipses since 1920 can be found in the Appendix in Bravo et al. [16].

Recent studies have shown that the effects of a solar eclipse on the ionosphere are not only local but can affect other geographic regions outside the umbra/penumbra of the eclipse [15,17–20]. These effects may be due to transport between hemispheric magnetic conjugates, alteration of the equatorial fountain effect, generation of a disturbed dynamo, and/or Atmospheric Gravity Waves (AGWs) that generate Traveling Atmospheric Disturbances (TADs) and/or Traveling Ionospheric Disturbances (TIDs).

An annular solar eclipse took place on 10 June 2021. The first external contact (P1 time) and the last external contact (P4 time) of the solar eclipse were at 08:12:22 UT and 13:11:22 UT, respectively. The partial solar eclipse was seen from the following geographic regions: in parts of the eastern United States and northern Alaska, Canada and parts of the Caribbean, Europe, Asia, and northern Africa. The annular eclipse was visible from parts of northeastern Canada, Greenland, and the Arctic Ocean, passing through the North Pole, and ending in Russian territory. It maximum magnitude was 0.944: this is the fraction of the angular diameter of a celestial body being eclipsed. This magnitude value was reached at geographic coordinates 80.815◦N, and 66.78◦W, at 10:41:57 UT (Greatest Eclipse time, GE time, http://xjubier.free.fr/en/site\_pages/solar\_eclipses/ASE\_2021\_GoogleMapFull.html, last accessed on 15 June 2022). The paths at ground level and at 350 km of altitude of the annular eclipse are shown in Figure 1 (see Supplementary Materials, Video S1). Due to the specific geometry of each eclipse, the paths differ both geographically and temporally according to the height considered, which could be significant when analyzing them [21].

Solar eclipses are rare events and, particularly, the 10 June 2021 event is an excellent opportunity to study the eclipse-induced effects on the polar ionosphere. Since the ionospheric variations can perturb GNSS, the eclipse can be used to study the positioning errors in these regions. There are some studies on the effects of the ionosphere during solar eclipses over the northern polar region. One of the first reported ones was the total solar eclipse that occurred on 9 March 1997 over Kazakhstan, Mongolia, eastern Siberia, and the Arctic Ocean (*τ* = −26 to 180 min and A = −5 TECu) [4,22]. Another reported one is the total solar eclipse that occurred on 1 August 2008 over Canada, northern Greenland, the Arctic Ocean, central Russia, Mongolia, and China (*τ* = −27 to 44 min, and A[%] = −40 to −11%) [10]. The most recent one is the eclipse that occurred on 20 March 2015 that covered the North Atlantic, Faroe Islands, and Svalbard (A[%] = −50 to −10%) [5,23]. This last one happened during the recovery phase of the most intense geomagnetic storm during Solar Cycle 24, the so-called St. Patrick's Day Storm. Due to the limited availability of GNSS stations around the globe at the time of these previous studies, they were focused on a regional scale. The increasing number of accessible GNSS stations around the world allows a study

on a global scale, facilitating the search of potential interactions between regions. This can show how spreadable GNSS disturbances are. In particular, the poles are of interest since several ionospheric disturbances can start from there during geomagnetic storms.

**Figure 1.** Instruments used in present work: 2337 GNSS stations (red dots), 4 digisondes (blue rhombuses), and 24 selected GNSS stations (green triangles). The magnetic equator (black line) and the annular eclipse path at ground level (blue line) are shown. Eclipse obscuration mask from P1–P4 time (shaded region), the annular solar eclipse path (magenta line), the maximum obscuration (magenta dot) at 350 km altitude are presented. The magenta dashed line starts from the maximum obscuration of the solar eclipse to its conjugate location in the Southern Hemisphere.

In GNSS receivers, TEC is estimated simultaneously from several satellites of the network, which serves to study the ionosphere. (e.g., [4,8,24], among many others). During the eclipses, the ionization decreases, producing a depletion in TEC. Although a decrease in electron concentration during a solar eclipse could produce an improvement in the positioning precision, it actually generates positional errors [25,26]. Few authors have analyzed the GNSS positioning errors caused by the influence of solar eclipses. The eclipses that occurred over Croatia on 11 August 1999 [27], over China on 22 July 2009 [28], and over the United States on 21 August 2017 [26] are some of the studies that analyzed GNSS positioning errors.

For the 1999 solar eclipse, Filjar, et al. [27] used a single frequency receiver located in the north of Croatia with ∼95% of maximum percentage of obscuration (MPO). In this work the authors did not relate ionospheric disturbances with positioning variations. The authors collected the horizontal positioning at the eclipse's maximum obscuration time (MOT). They calculated an average positioning error of ∼34 m on horizontal Global Positioning System (GPS) accuracy for that time. These horizontal values could be due to the use of a single-frequency, the number of receiver channels, and the possible influence of Selective Availability (until May 2000).

Jia-Chun et al. [28] used eight GPS stations to study the TEC changes and their effect on the positioning during the 22 July 2009 solar eclipse. They possessed a real-time point positioning and real-time precision of single baselines. The measurements were affected by a geomagnetic storm (Dst peak = <sup>−</sup>80 nT and Kp-index = 5+), which made it difficult to separate the influence of the eclipse from the storm one.

In the case of the 2017 eclipse, Park et al. [26] computed and compared the rate of change of the TEC (ROT) with respect to the day before and the day after the eclipse; and with a time window of 3 h, from 16 UT. They determined the means of positioning errors at four GNSS stations (localized in Oregon, United States) within the path of the total solar eclipse which reached ∼32 cm. However, on reference days, the means of positioning errors were between 7–14 cm. The authors used the average length of the eight baselines, which was ∼270 km. On the eclipse day, the means of positioning results were −4 to 324% over the day before and the day after the eclipse. Yuan et al. [25] established the ionospheric eclipse factor method (IEFM) to model the ionospheric delay searching for the improvement of the GNSS positioning estimation. In this contest, the paper introduces the concept of the ionospheric eclipse factor method for the IPP for relatively precise separation of daytime from nighttime for the ionosphere. Although the ionospheric eclipse factor is not related to a solar eclipse as an astronomical phenomenon that occurs when the Moon obscures the Sun from Earth, this method could be used in future studies related to the impact of solar eclipses on GNSS positioning.

In our case, we obtain the apparent position variation using the post-processing kinematic precise point positioning (PPP) with ambiguity resolution (PPP-AR) mode. We chose this method because PPP demonstrates a high ability to improve position estimation. PPP is used for calculating the coordinates of a single receiver without the need for a reference station nearby as a control station. In addition, we can find some free PPP services available online [29]. PPP-AR is an enhanced version of the PPP technique that resolves the carrier phase ambiguities, improving the PPP accuracy [29–31]. Katsigianni et al. [30] recently presented a comparison between PPP and PPP-AR. In order to offer the community the possibility of evaluating our analysis, we used an online service. Thus, we selected the CSRS-PPP service for this work because it is one of the most commonly used PPP online services in the field. We also applied the common noise filter to more than 2300 GNSS stations, to correct the time series of the North, East, and Up components of the GNSS receivers, as described in [29].

The regular ionospheric effects of solar eclipses are not yet fully understood. Studies of the eclipse-induced effects on the ionosphere are important because they provide a better understanding of the processes that control the ionosphere and that can cause GNSS positioning errors. In the present paper, we present the impact of the 10 June 2021 annular solar eclipse on ionospheric variations that also cause errors in GNSS positioning. Therefore, we first analyze the ionospheric behavior at a global scale based on 2337 dual-frequency (DF) GNSS stations, Swarm-A satellite, and four ionospheric stations. We used GNSS stations distributed around the world since they will allow us to evaluate the effects beyond the northern polar region with a higher spatial resolution than ever before. Unlike previous studies about the GNSS positioning errors caused by the influence of solar eclipses, our study is focused on a global scale. This allowed us to find other locations in the world that could be affected by a perturbation in the north pole and how that perturbation propagates to those potential locations.

#### **2. Materials and Methods**

The methodology used in the 10 June 2021 annular eclipse is mainly based on the one described in Valdés-Abreu et al. [29]. However, in this work we incorporate the processing of ionospheric data from a Low Earth Orbit (LEO) satellite. The procedure of this work also includes the analysis of geophysical and geomagnetic conditions close to the date of the eclipse (10 June 2021). How we use this new set of data is detailed below.

#### *2.1. Estimation of the Ionospheric Total Electron Content*

The inherent space-time variability of the ionosphere can be observed through TEC that can be obtained using GNSS stations [32]. Then, GNSS measurements based on dualfrequency signals *f*<sup>1</sup> and *f*2, were used to obtain the vertical TEC (VTEC) data. The groundbased dual-frequency GNSS (DF-GNSS) receiver continually records two types of delay: the pseudoranges and the carrier phases of the two signals. The obtained data was used to estimate the slant TEC (STEC) and to calculate the VTEC. STEC and VTEC were calculated from Receiver Independent Exchange Format (RINEX) files by using the GPS-TEC analysis software (GPS-TEC program Ver 2.9.5, developed by Dr. Gopi Seemala, https://seemala. blogspot.com/2017/09/gps-tec-program-ver-295.html, last accessed on 17 April 2022) [24]. VTEC values were estimated with a satellite cut-off elevation angle of 30◦ at an altitude of 350 km to reduce possible errors. The TEC values were released every 30 s and were corrected for the satellite and receiver bias using the data obtained from the AIUB Data Center of Bern University in Switzerland (ftp://ftp.aiub.unibe.ch/CODE/, last accessed on 17 January 2022).

For the final selection of the RINEX files of each GNSS station, we took into account several aspects. First, we considered the quality of the files during the selected period of days (DoYs 159–163). Second, we verified that there were no errors or data-gap after TEC estimation and post-processing PPP-AR. This aspect is critical to relating TEC and/or ROTI with positioning variations. We used data from all available stations in the polar regions. We also tried to cover regions such as Africa, Australia, and Asia. The International GNSS Service (IGS) stations (http://www.igs.org, last accessed on 17 January 2022) [33]; the Chilean network of GNSS receivers operated by the National Seismological Center at University of Chile (CSN in Spanish); University NAVSTAR Consortium (UNAVCO); the Argentine Continuous Satellite Monitoring Network (RAMSAC in Spanish) [34]; the Brazilian Network for Continuous Monitoring of the Institute of Brazilian Geography and Statistics (IBGE in Portuguese); the Geoscience Australia; the Low-Latitude Ionospheric Sensor Network (LISN, http://lisn.igp.gob.pe/, accessed on 26 July 2021); and the African Geodetic Reference Frame (AFREF) provided RINEX files of 2337 GNSS stations that met the requirements we imposed (see Figure 1).

Additionally, the differential VTEC (DVTEC) in TECu and the percentage changes of DVTEC (DVTEC[%]) were used. These parameters are studied in the analysis of ionospheric irregularities, defined as the relative variation of VTEC, epoch by epoch, with respect to the mean value (in time) of *VTEC* as shown in Equations (1) and (2) [35].

$$DVTEC\_t = VTEC\_t - VTEC\_t \tag{1}$$

$$DVVTEC[\%\_{t}]\_{t} = \frac{DVTEC\_{t}}{\overline{VTEC\_{t}}} \cdot 100\tag{2}$$

where *t* represents the epoch, and *VTECt* is calculated by averaging the values of VTEC at the same time of the day, *t*, for the reference DoYs 159, 160, 162 and 163 which correspond to 2 days before and 2 days after the day of the eclipse (DoY 161).

According to the methodology [29], we used the ordinary Kriging interpolation method to produce the TEC maps at each ionospheric pierce point (IPP). With this method, we filled in the spatial gaps of the global ionosphere TEC maps, minimizing the effects of the inhomogeneous distribution of GNSS receivers. Before interpolating, we selected a spatial resolution of 2.5◦ × 2.5◦. Then, we employed the Kriging package implemented in Python (https://github.com/ERSSLE/ordinary\_kriging, last accessed on 17 January 2022).

#### *2.2. ROT and ROTI*

In order to detect, investigate and characterize the occurrence of ionospheric irregularities, we have used the Rate of change of the TEC Index (ROTI). The ROT and ROTI values are usually expressed in TECu/min. ROTI is defined as the standard deviation of the rate of TEC (ROT), and it is estimated by dual-frequency GNSS data with the time interval of 5 min by using Equation (3) [29]:

$$ROTI = \sqrt{\langle ROT^2 \rangle - \langle ROT \rangle^2} \tag{3}$$

where · represents the temporal average. ROT is defined as the TEC variation rate of two successive epochs as stated in Equation (4) [29]:

$$ROT = \frac{STEC\_t^i - STEC\_{t-1}^i}{k\_t - k\_{t-1}} \tag{4}$$

where *i* indicates the observed satellite and *t* denotes the time of epoch. Hence, *kt* − *kt*−<sup>1</sup> is the time interval between the subsequent epochs.

Depending on the ROTI value, the activity level can be classified in ranges such as: weak (if 0.25 ROTI < 0.5); moderate (if 0.5 ROTI < 1); and strong (if ROTI 1), according to Liu et al. [36].

#### *2.3. Low Earth Orbit Satellite Measurements and Ionospheric Data*

Additionally, we analyze ionospheric measurements provided by a LEO satellite, the European Space Agency's Swarm mission. This mission is a constellation of three LEO satellites that were successfully launched on 22 November 2013, and are still operating. This constellation is designed to provide measurements of the Earth's magnetosphere and ionosphere, studying the impact of the solar wind on the dynamics of the upper atmosphere [37,38]. The Swarm-Alpha (A), Bravo (B), and Charlie (C) are three identical satellites that share the same design and payloads.

All three satellites were put into a circular near-polar orbit with a low eccentricity. Swarm-A/C pair have the same orbit configuration (inclination of 87.35◦, altitude of ∼450 km, east–west separation of about 1–1.5◦ in longitude), while Swarm-B has a different one (inclination of 87.75◦, altitude of ∼510 km). These satellites fly above the F-layer peak (the peak altitude of the ionospheric electron density). In addition, Swarm-A/C fly in tandem, while Swarm-B moves away from the couple Swarm-A/C by covering different local times [37,38].

The Swarm spacecraft were equipped with different payloads, including GPS receivers and Langmuir Probes (LP), among others. We considered the ionospheric VTEC values associated with the point where the link path between GPS and Swarm-A satellite pierces the spherical thin shell located 400 km above the Swarm-A orbit. We also used in-situ electron density (*Ne*) measurements by LP at ∼450 km [38,39] for each of the five selected DoYs in June 2021 (https://Earth.esa.int/web/guest/swarm/data-access, last accessed on 17 January 2022).

We used the Swarm Level 2 (L2) TEC (TECxTMS\_2F) data product, which contains time series of slant and vertical (absolute and relative) TEC for each GPS satellite in view (at most eight due to instrumentation design). The cadence of the ionospheric TEC data is 1 Hz since it was changed from 10 s (0.1 Hz) to 1 s (1 Hz) on 14 July 2014 [38,40].

We also used the Swarm LP data, which is part of the EFI package (EFIX\_LP\_1B plasma data). LP provides measurements of in situ *Ne* and electron temperature with a 2 Hz sampling rate [38,39].

#### *2.4. Apparent Position Variation Using Kinematic Precise Point Positioning*

The RINEX files of 2337 GNSS stations were processed using the Canadian Spatial Reference System (CSRS-PPP) online service (https://webapp.geod.nrcan.gc.ca/geod/ tools-outils/ppp.php, last accessed on 6 January 2022) [41] with ambiguity resolution (PPP-AR) mode. The CSRS-PPP provides centimeter-level estimations with converged float solutions [41–44].

Usually, the CSRS-PPP report can provide a different reference start value for different days. To facilitate the evaluation of the apparent position variation time series, we process the data to have an equal reference for all the used data. At each of the 2337 stations, we applied the common noise filter to correct the time series of the North, East, and Up components, using the equation [29]:

$$CAPday\_t = APday\_t - RP\_t \tag{5}$$

where *t* is the epoch, *CAPdoyt* is the corrected apparent position, *APdoyt* is the apparent uncorrected position, and *RPt* is the reference position. We use the average of AP, *AP*, from the same reference days mentioned in Equation (1) to calculate *RPt*.

At each of the 2337 stations, the maximum error was obtained within the selected five days. Subsequently, the error of each station per component was classified by intervals, counting the percentage of the total number of stations that fell into each interval. In addition, a 3D position error was calculated as:

$$3D\_t = \sqrt{East\_t^2 + North\_t^2 + \mathcal{U}p\_t^2} \tag{6}$$

We use two threshold values. First, we selected threshold values for maximum 3D positioning error greater than or equal to 10 cm (3D 10 cm) since according to data on quiet days, over 90% of the GNSS stations had 3D errors of less than 10 cm, while during ionospheric disturbances, only ∼40% kept this level of accuracy [29]. Second, we applied the Equation (6) to the horizontal and the vertical components presented in [29,30], obtaining the threshold of the 3D positioning error root mean square (3D-RMS) greater than or equal to 3 cm (3D-RMS 3 cm).

#### *2.5. Geomagnetic and Geophysical Conditions*

The geomagnetic data downloaded from OMNIWeb Plus Data (https://omniweb. gsfc.nasa.gov, last accessed on 11 May 2022) for the 10 June 2021 annular solar eclipse indicates a period of low activity. Except in DoY 163, where it was 4− between 3 and 6 UT, the estimated 3-hour planetary index (Kp) was 3−. The disturbance storm time index (Dst) peak was >−17 nT, except after DoY 162 where a minimum of −37 nT was reached at 11 UT. The interplanetary magnetic field (IMF) Bz component in GSM coordinate peak was >−7.4 nT after DoY 162 and the solar wind speed (Vsw) was 330–520 km/s during 8–13 June 2021 (see Figure 2).

**Figure 2.** Variations of 3-hourly Kp, Dst, IMF-Bz, AE, and Vsw indices that characterize the geomagnetic conditions on 8–13 June 2021. P1–P4 time (light grey bar) and GE time (red dashed line) are also represented.

The Auroral Electrojet index (AE) is a good proxy of the geomagnetic activity level at mid/high latitudes [29,45]. Following De Michelis et al. [46], we selected two distinct datasets corresponding, respectively, to geomagnetically quiet (AE < 50 nT) and active (AE > 300 nT) periods. Figure 2 also illustrates that the AE index was over 500 nT between 5 and 15 UT on DoY 162, and between 1 and 5 UT on DoY 163; so these time periods showed some activity in the auroral regions. Since these days are used by comparison with the day of the eclipse, these periods of time were treated with care to avoid interfering with the eclipse analysis.

Therefore, the geomagnetic conditions were generally quiet, except on DoY 162 where a weak geomagnetic storm took place between 8 and 16 UT. DoY 162 did not cause problems in the ionospheric TEC background to our results for the eclipse day. However, the geomagnetic activity the day after the eclipse had significant effects on GNSS positioning errors comparable to the positioning errors caused by the annular solar eclipse. These effects will be presented in more detail in the coming section.

#### Earthquake Occurrence

We also reviewed the occurrence of earthquakes (EQs) around the world, with a moment magnitude greater than 5 Mw and a depth of over 70 km on 8–13 June 2021. This review is important because EQs are sources of TEC disturbances and thus positioning errors. In the period analyzed, 15 moderate EQs of less than 5.7 Mw occurred (https://earthquake.usgs.gov, accessed on 17 January 2022). However, none of them produced noticeable effects on TEC or on the position estimation on the GNSS receivers during the analyzed period of days.

#### **3. Results**

In this section, we present the main results obtained after applying the methodology described in the previous section. The results obtained in this work can be divided into two main parts: (1) the analysis of the TEC maps that present the effects on the ionosphere at a global scale; and (2) the calculation of the positioning errors that these ionospheric effects generate.

#### *3.1. Ionospheric Behavior and TEC Maps*

From the data of each station, we can estimate the VTEC for each station during the selected period of days. By using the ordinary Kriging interpolation, as described in Section 2.1, it is possible to obtain VTEC maps. Figure 3 shows a summary of the TEC maps by contrasting the eclipse (VTECe) and control (*VTEC*) days. We present some particular hours: 09.15 UT, 10.70 UT (GE time), 12.00 UT, 13.19 UT (P4 time), 13.72 UT (P4 time + ∼0.5 h), and 17.66 UT. Figure 3 also shows the eclipse masks from 20% obscuration and with intervals of 20%, at 350 km altitude (white line). From these figures, it is possible to notice that the Greenland and South American sectors are two of the most affected in terms of VTEC depletion. VTECe in IPP and DVTEC[TECu] in Figure 3 use the Kriging interpolation method and are shown in equidistant cylindrical projection and Northern Hemisphere polar plots (see Supplementary Materials, Figure S1).

The 09.15 UT, 12.00 UT, and 13.72 UT maps were chosen in particular because they show the greatest apparent position variations during the eclipse time window. The 17.66 UT map was chosen because the ionosphere was roughly recovered by that time. For a better visualization of the eclipse effects, a third column has been incorporated where the VTEC differential in TEC units (DVTEC[TECu]) is shown.

**Figure 3.** Ionospheric TEC maps during the 10 June 2021 Annular Solar eclipse using the Kriging interpolation method. From left to right panels: VTECe, *VTEC*, and DVTEC[TECu]. From top to bottom panels: 09.15 UT, 10.70 UT (GE time), 12.00 UT, 13.19 UT (P4 time), 13.72 UT, and 17.66 UT. Eclipse obscuration masks from 20% obscuration and with intervals of 20%, at 350 km altitude (white line) are shown. The white dashed line starts from the maximum obscuration of the solar eclipse to its conjugate location in the Southern Hemisphere at 350 km altitude.

DVTEC had values of around −20% (−2 TECu) over the oceanic sectors when the eclipse began (P1-time). However, at these locations, the GNSS receivers are scarce, which can cause less reliable interpolation. This value could be considered as part of the nonsignificant variations in DVTEC. A similar problem is identified over Central Africa, where there is a value of 50% (7 TECu), possibly also due to the few receivers in this area (see Figure 1). The anomalies in these areas were observed more than 5 h before the eclipse. We will focus mainly on changes generated over the continental areas of America and Europe, while the other areas will not be considered for this analysis.

When the ionospheric TEC effects due to eclipse have already begun, the 09.15 UT maps show a slight depletion of −30% (−1 TECu) across eastern Canada under the shadow of the eclipse. At 10.70 UT (GE-time), these changes expand beyond the shadow area of the eclipse (see second row of Figure 3). The 10.70 UT maps show that ionospheric TEC depletion did not only occur across the obscuration region over the Northern Hemisphere. DVTEC[%] had values of around −60% (−4 to −2.5 TECu) over the South and East coasts of Greenland, −50% (−3 TECu) eastern Canada, and around −50% (∼−3 TECu) over the Lesser Antilles.

The 12.00 UT, 13.19 UT, and 13.72 UT maps show DVTEC[%] had values of around −30% (−3 to −1.5 TECu) over Russia after GE time. Figure 3) also illustrates how the TEC disturbance moved from West to East over the Northern Hemisphere, following the path of the annular solar eclipse. At 12 UT, there is a recovery of the ionospheric TEC over Canada and Greenland regions, but DVTEC[%] had values of less than −50% in East coast of Greenland. Ionospheric TEC depletion had values of around −60% (∼−5 TECu) over the Lesser Antilles near the north crest of Equatorial Ionization Anomaly (EIA) and less than −30% (∼−5 TECu) appeared over South America near the south crest of EIA. The 13.19 UT and 13.72 UT maps show another slight recovery of the ionospheric TEC in the North Atlantic and Greenland, as well as a TEC depletion over Russia. Moreover, TEC depletion was accentuated in the EIA crests over South America, where DVTEC[%] had values of less than −60% (<−11 TECu). It is shown that the effects lasted beyond the end of the eclipse.

The 17.66 UT maps present the global recovery of the ionosphere a few hours after the end of the eclipse. These maps show a slight DTEC[%] enhancement in the center of the EIA form ∼−16% to ∼10%. But the TEC depletion was ∼−20% (−5 to −3 TECu) in the EIA crest over South America. The TEC behavior in the EIA crests was maintained until after 19.66 UT.

On the other hand, DoY 162 had geomagnetic activity (see Section 2.5). Therefore, we checked if the DVTEC changes that we observed for the day of the eclipse were due to using DoY 162 as one of the reference days. We compute a new *DVTECt* (*DVTECnewt*, Equation (1)), and a new *VTECt* by averaging the VTEC values at the same time of the day, *t*, for the reference days, DoYs 159, 160, and 163. For each map, we used the map algebra (*DVTECt* − *DVTECnewt*). The mean value ranged between −0.5 and 0.1 TECu, with a standard deviation of less than 0.7 TECu (see Supplementary Materials, Figure S2, Table S1). Therefore, geomagnetic activity during the DoY 162 did not cause problems in the background to our ionospheric TEC results for the eclipse day.

#### Ionospheric Behavior Using Swarm Satellite Measurements

We also present ionospheric behavior using Swarm-A measurements (see Figure 4). We illustrate VTEC at 850 km altitude on DoY 161 compared to DoY 159 during three ascending passes of the Swarm-A satellite (45◦S, see Figure 4 (upper panels)). We selected the three Swarm-A satellite passes that best fit the eclipse region and eclipse time window. The first satellite pass (∼8.50–9.20 UT) occurred after P1 time. The greatest ionospheric TEC degradation was −30◦S–30◦N (∼−1.7 TECu, −35%). The second pass (∼10.10–10.80 UT) was close to the GE time. As latitude increases, TEC decrease. The third satellite pass (∼11.70–12.40 UT) was performed prior to P4 time (∼−1.9 TECu, −37%). It is possible to see that TEC decrement was concentrated between 10 and 30◦N (VTEC was close to ∼−1 TECu, and ∼−30%). Ionospheric TEC depletion was greatest in the 15–75◦N region (∼−2 TECu, −45%).

Figure 4 (bottom panels) depict in situ *Ne* measurements made by Swarm-A Langmuir probe. Figure 4 (upper panels) show that VTEC behaves similarly to *Ne*. We observed that *Ne* decrease was −46% (−0.26 × 105 e/cm3) at ∼8.86 UT in ∼1◦N; −55% (−0.39 × <sup>10</sup><sup>5</sup> e/cm3) at ∼10.62 UT in ∼50◦N; and −55% (−0.72 × 105 e/cm3) at ∼12.05 UT in ∼20◦N.

**Figure 4.** Ionospheric behavior using Swarm-A measurements. (**Upper panels**) present ionospheric TEC data taken by Swarm-A satellite at 850 km (at 400 km above the Swarm-A). The TEC gathered through the satellite orbit is presented over an Earth map (left panel) and as a profile with data obtained in one of the comparison days (write panel). The three satellite passes are from 45◦S to ∼90◦N, from left to right, between ∼11.70–12.40, ∼10.10–10.80, and ∼8.50–9.20 UT during 2 days before eclipse day (DoY 159), and eclipse day (DoY 161). The annular eclipse path at 450 km (grey line) and 850 km altitude (magenta line) are also shown on DoY 161. VTEC on DoY 161 (red dots) compared to DoY 159 (green dots) between ∼11.70–12.40 UT, ∼10.10–10.80 UT, and ∼8.50–9.20 UT. (**Bottom panels**) depict Swarm-A in situ electron density (*Ne*) presented in the same way than the VTEC data.

#### *3.2. ROTI and GNSS Precise Point Positioning Accuracy Maps*

We estimated the ionospheric TEC, ROTI, and positioning for the full 5 days but only show 6 hours per day. On the eclipse day, we observe the largest positioning variations during this time window (around P1–P4 time, see Figure 5). This study focuses on the positioning accuracy of the stations during the 10 June 2021 annular solar eclipse, during the time between 8 and 14 UT. We estimated the PPP-AR using the CSRS-PPP service, as described in Section 2.4. Comparing the eclipse day with respect to the DoYs 159 and 160, we can see that the percentage of GNSS stations that exceeded maximum 3D positioning error 10 cm and 3D-RMS 3 cm (positioning thresholds) jumped from ∼180 (∼8%) to 333 (∼14%) and from ∼170 (∼7%) to 210 (∼9%), respectively. In addition, the ROTI threshold 0.25 TECu/min was taken according to Liu et al. [36], and used in the methodology [29]. Figure 6 shows maximum 3D positioning errors, 3D-RMS of the apparent position, and ROTI maps, for each of the five selected DoYs.

**Figure 5.** Time variations of the percentage of stations with 3D positioning error greater than 10 cm on DoY 161. The (**left panel**) shows the 36 GNSS stations localized in Greenland. The (**right panel**) presents the 335 GNSS stations that are situated in South America.

ROTI was calculated as described in Section 2.2 to study the relationship between the variation of TEC and the positioning error for the eclipse. The images in the Figure 6 (right panel) show five maps of ROTI, each representing the stations that had ROTI greater than 0.25 TECu/min. Each map represents a different day, but at the same time as that of the solar eclipse, DoYs 160–163 between 8 and 14 UT. We do not show the maps for DoY 159 because they are similar to those for DoY 160.

**Figure 6.** Maximum 3D positioning errors 10 cm (**left panels**); and ROTI 0.25 TECu/min (**right panels**) between 8 and 14 UT. From top to bottom shows DoYs 160–163. Annular eclipse path at ground level (blue line), and at 350 km of altitude (magenta line) are also shown.

The TEC data (e.g., Figure 3) and the derived PPP-AR (see Figure 6 (left panels)), and Figure 7) show that there are two regions where the errors are more severe during the solar eclipse (DoY 161, between 8 and 14 UT). These regions are Greenland and South America. For this reason, we focused this study on these sectors.

**Figure 7.** Behavior of the maximum 3D positioning error, 3D-RMS, and ROTI in Greenland and South America sectors, where 36 GNSS stations are localized in Greenland (**upper panels**) and 335 GNSS stations are situated in South America (**bottom panels**). Percentage of GNSS stations where (**a**,**b**,**e**,**f**) 3D-RMS 3 cm; (**c**,**d**,**g**,**h**) maximum 3D position 10 cm; (**a**,**c**,**e**,**g**) ROTI 0.25 TECu/min; and (**b**,**d**,**f**,**h**) ROTI 0.50 TECu/min. Percentage of GNSS stations meeting 3D-RMS (blue bars); maximum 3D positioning error (light blue bars); ROTI (orange bars); 3D-RMS and ROTI (green bars); and maximum 3D positioning error and ROTI (light green bars) values.

In the Greenland and South America regions, we can find 36 (∼2%) and 335 (∼14%) of the 2337 total available GNSS stations, respectively. We determined the percentage of stations localized in both regions that had errors that fell at certain intervals during the time period of the annular eclipse (between 8 and 14 UT). Figure 7 shows the percentages of stations that meet the thresholds of maximum 3D positioning error, 3D-RMS, and ROTI. We determined the number of GNSS stations based on ROTI activity and positioning values (see Supplementary Materials, Tables S2–S4, where each column in these tables represents the percentage of stations with maximum 3D position error, 3D-RMS, and ROTI at certain intervals in the selected period).

Figure 5 shows in more detail the percentage of stations with maximum 3D errors 10 cm on eclipse day. In Greenland, during the inicial period of the eclipse (∼P1 time), the percentage of stations with maximum 3D positioning error rises to 60%. Subsequently, the value remains at ∼28% until it increases to ∼33% between 10.5 and 11.25 UT (around GE time). The value then returns to ∼28% until 14 UT (after P4 time), when it drops to ∼11% of stations. In South America, the percentage of stations with maximum 3D positioning error had a maximum of ∼34% between 9.75 and 11.5 UT (around GE time). We also were able to observe a decrease in stations that exceeded the threshold maximum 3D positioning error from ∼28% to ∼22% between 16 and 17 UT.

#### *3.3. Ionospheric Behavior and GNSS Positioning Errors by Region*

To study the effects of the eclipse, we selected 24 stations from among the 2337 GNSS stations (see Figure 1). We chose five GNSS stations close to the annular solar eclipse (KMOR, KAGZ, MARG, IQAL, PICL). There were six GNSS stations located in the partial eclipse region (CN00, TRO1, SVTL, TIXI, MAG0, YAKT). Furthermore, we used five stations located in the Caribbean and South America (LMMF, BOAV, PIFL, MSBL). The sunrise (in PICL and CN00 stations) and the sunset (in MAG0 and YAKT stations) happened during the eclipse time window at ground level but did not take place at the ionospheric height of 350 km. More details about the GNSS stations and eclipse conditions (with respect to the ionospheric height of 350 km) can be found in Table 1.

**Table 1.** Detail about the GNSS stations analyzed, their location, eclipse condition, magnitude and change in VTEC at each station. Eclipse characteristics (time and maximum obscuration) for the 24 selected GNSS stations. Further, we have also estimated eclipse conditions at the ionospheric height of 350 km by the method suggested by Verhulst et al. [21]. Ionosphere, 3D eclipse, and 3D non-eclipse values are between 8 and 14 UT. In 3D non-eclipse: MAX refers to the maximum 3D positioning error of the reference days, and we calculate RMS by taking all the values of the reference days between 8 and 14 UT.


<sup>1</sup> *τ* refers to GE time, not to MOT.

Figure 8 presents the results of ionospheric TEC of 12 stations from among the 24 selected GNSS stations for the eclipse day (VTECe), the reference days (*VTEC*), and the final results of DVTEC [%]. The vertical blue shaded region between P1 time and P4 time, with GE time (brown dotted line). The vertical yellow shaded region between C1 time and C4 time, with MOT (black dotted line). Each plot is shown between 5 and 23 UT. The maximum reduction values of TEC for each station are indicated in Table 1 (*τ* = 1 to 288 min, and A[%]= −65 to −27%). This eclipse occurred during the morning at most of the selected stations but took place in the afternoon at five stations (NYA1, TRO1, SVTL, TIXI, MAG0, YAKT). The Sun's activity became stronger around noon and the clear TEC reduction during the eclipse can be observed. The stations that are near the path of the annular eclipse at 350 km altitude and the east coast of Greenland reached lower values of DVTEC [%] ∼−55% than the stations with the annular eclipse at the surface level DVTEC [%] ∼−40%.

Figure 8 shows the ionospheric TEC changes for 12 of the 24 GNSS stations presented in Table 1. The TEC disturbance lasted longer at the GNSS stations located in South America and the Lesser Antilles (CN00, LMMF, CN57, BOAV, PIFL, MSBL). In these GNSS stations, the ionospheric effect caused by the eclipse started at ∼8.5–9 UT and ended ∼18–21 UT (Δ*T* > 10 h). The ionospheric response is similar in BOAV and MSBL stations where A[%] ∼−30%. In MAG0, TIXI, and YAKT stations, we observe a TEC depletion during the eclipse time window, but it is not as noticeable as in the other cases.

**Figure 8.** The behavior of the ionospheric TEC during the 10 June 2021 Annular Solar Eclipse in 12 of the 24 selected GNSS stations. DVTEC[%] (blue line), VTECe (red dashed line) and *VTEC* (green dashed line). The GNSS stations are ordered by latitude and then by longitude. P1–P4 time is represented by the light blue bar, C1–C4 time by the yellow bar, MOT by black dotted line, and GE time by the red dotted line.

Additionally, the number of GNSS stations according to ROTI activity was: 5 strong (BLAS, PIFL, LEFN, NYA1, and MARG stations), 2 moderate (KMOR and KAGZ stations) and 17 without activity (see Table 1).

In the same way, we presented the results of PPP-AR of 24 DF-GNSS stations during eclipse day. The time series were corrected for the common noise filter of the East, North, and Up components. The stations had variations in position within the time window of the eclipse (between 8 and 14 UT). The station with the highest positioning errors in the East, North, and Up components was PIFL stations. KAGA, GLS2, SENU, and MSBL stations also showed position variations between 5 and 8 UT.

Equation (6) is used to obtain the 3D results. Then, the maximum 3D positioning error and 3D-RMS values (between 8 and 14 UT) for each station are indicated in Table 1. We note that the GNSS stations can be separated according to the percentage of maximum 3D positioning error and 3D-RMS, with respect to the maximum values of reference days.

In the case of maximum 3D positioning error, four stations presented values below 0% (MARG, ALGO, CN00, and LMMF); eight 0–25% (KAGZ, IQAL, BLAS, BOAV, NYA1, SVTL, MAG0, and YAKT); seven GNSS stations were between 25 and 50%; three stations were 60–70% (GLS2, KUAQ, and SENU); and two stations > 100% (PIFL and TRO1).

On the other hand, for 3D-RMS in a percentage, 1 station was < 0% (CN00); 11 stations were 0–25% (KAGZ, MARG, LEFN, BLAS, KAGA, GLS2, KUAQ, NYA1, SVTL, TIXI, and YAKT); 7 stations were 25–50% (KMOR, IQAL, ALGO, LMMF, CN57, PIFL, and MSBL); 3 stations were 75–100% (SENU, PICL, and BOAV); and 2 stations > 100% (TRO1 and MAG0).

#### A Case Study

We will describe in more detail the results obtained with PIFL GNSS station (6.79◦S, 43.04◦W). PILF had the largest ionospheric disturbances and GNSS positioning errors (see Table 1). TEC depletion had values around −65% (−11.8 TECu) at 110 min after GE time (see Figure 8). Figure 9 presents the ionospheric behavior (ROT, ROTI) and kinematic

DF-GNSS PPP-AR mode during DoYs 160–163 between 5 and 23 UT. We do not show DoY 159 because it does not differ significantly from DoY 160.

**Figure 9.** Ionospheric behavior and apparent position variation of the PIFL GNSS station. From top to bottom show DoYs 160–163. ROT (TECu/m) (**left panels**), ROTI (TECu/m) (**middle panels**), DF-GNSS PPP-AR (cm): East [cm] (blue line), North [cm] (red line), and Up [cm] (green line) (**right panels**). P1–P4 time on non-eclipse day (blue dashed lines), P1–P4 time on eclipse day (light blue bar), GE time (red dotted line).

Figure 9 (left, middle panels) illustrates examples of GPS ROT and GPS ROTI variations along with all visible GPS satellites. On eclipse day, we can observe a |ROT| > 1.5 TEC/min in eight Pseudo Random Noises (PRN-4, 7, 8, 9, 14, 27, 28, 30). The |ROT| value was exceeded by 3–4 PRNs during the reference DoYs 159, 160, and 163. In contrast, the |ROT| was exceeded by seven PRNs on DoY 162 (see Figure 9 (left panels)). On DoY 161 between 10.66 and 17.39 UT (6.73 h), we could note 22 and 8 ROTI values >0.5 and >1 TECu/min, respectively (see Figure 9 (middle panels)). The ROTI peak was 1.9 TECu/min at 12.75 UT, estimated from the PRN-4. On this day, nine PRNs (PRN-4, 7, 8, 9, 10, 14, 27, 28, 30) presented a moderate and/or strong ROTI activity. Regarding DoYs 159, 160, 162 and 163, we observed 12, 8, 20 and 6 values with moderate and/or strong ROTI activity. Then, this

station showed strong TEC activity during each of the five DoYs. The ROTI value was higher on eclipse day 1.9 TECu/min at 12.76 UT (∼25 min after P4 time).

Figure 9 (right panels) show the apparent position variation of kinematic DF-GNSS PPP-AR mode in the East, North, and Up components for the PIFL GNSS station. These time series has been corrected for the common noise filter. On DoY 161, the apparent peak ground displacement in the East, North, and Up components were 18, 40, and 119.8 cm, respectively. Moreover, the maximum 3D positioning error 10 cm ∼9.80 UT by ∼3.30 h. The Up, North, and East components are ordered from highest to lowest errors. Then, positioning errors in the three components and their results were clear during the eclipse time window (after GE time), relative to the reference days.

#### **4. Discussion**

In this section, we discuss the main findings regarding the 10 June 2021 annular solar eclipse. The main goal is to study the positioning errors of GNSS receivers caused by this solar eclipse. In order to verify our findings, we compare our ionospheric values with results presented for other solar eclipses over the northern polar region (9 March 1997 [4,22]; 1 August 2008 [10]; and 20 March 2015 [5,23]).

There are several free-to-use software available for single-station TEC estimation methods [47,48]. We selected GPS-TEC software because it is a widely used method by the scientific community to study phenomena such as geomagnetic storms [29] and solar eclipses [49], among others. GPS-TEC software is fundamentally based on the assumption that ionospheric density depends on altitude to determine VTEC from STEC.

#### *4.1. Ionospheric Behavior*

The present analysis aims to show, as best as currently possible, the effects that the solar eclipse generates both in the ionosphere under the moon's shadow as well as in the global ionosphere. The relevance of this event is that there are few of them that occur in polar regions, in this case, in the Arctic.

We have used interpolated global maps from TEC and we have calculated the difference between eclipse and reference days. The results of the TEC maps show a significant reduction under the moon's shadow, except at the CN00 station that has similar behavior to the LMMF, CN57, BOAV, and MSBL stations (see Figures 3 and 8, Table 1); the GNSS stations located in the region of the eclipse reaching a maximum of *τ* = 1 to 288 min, A peak ∼−5 TECu, A[%] = −61 to −27%. Table 1 details these parameters for GNSS stations of some selected regions (see Figure 1). The values of these parameters agree with those obtained for the solar eclipses of 9 March 1997 solar eclipse [4,22], 1 August 2008 [10], and 20 March 2015 [5,23].

TEC depletion was not as pronounced in the MAG0, TIXI, and YAKT stations (A[%] = −38 to −30%) compared to others GNSS stations (A[%] = −61 to −40%) with a similar percentage of obscuration (∼88%). This could be due to the fact that the sunset in MAG0, TIXI, and YAKT stations happened during the time-window of the eclipse at ground level. Moreover, the other stations were closer to the greatest eclipse (see Figures 1 and 8, Table 1).

In addition to the decrease of TEC in the ionosphere under the Moon's shadow, we have observed interesting and significant effects far from that region. This is the case of a significant decrease in TEC that seems to move southward from the shadow, passing through the North Atlantic, and remaining stationary for several hours over the Caribbean and the north of Brazil, at the stations CN00, LMMF, CN57, BOAV, PIFL, MSBL (see Figures 3 and 8, and Table 1). The delay value relative to GE time was between ∼30 and ∼168 min, A peak ∼−11 TECu, A[%] = −65 to −28%, ΔT > 10 h. This area coincides with the location of the crests of EIA. The TEC variations were more intense north and south of the magnetic equator, where they were similar to those obtained at the GNSS stations located in the eclipse region.

On the other hand, applying *DVTECt* − *DVTECnewt*, we observed that the mean was between −0.5 and 0.1 TECu; and standard deviation was less than 0.7 TECu. Therefore, TEC changes due to the weak geomagnetic activity during the DoY 162, did not cause problems in the ionospheric TEC background to our presented results for the eclipse day (see Supplementary Materials, Figure S2, Table S1).

In order to verify the negative disturbance in TEC on EIA crests, we have compared them with ionosonde observations of the sector involved (Figure 10). The Ramey (RA, 18.5◦N, 67.1◦W) station on the Caribbean side, and Sao Luis (SL, 2.6◦S, 44.2◦W), Fortaleza (FZ, 3.9◦S, 38◦W) and Cachoeira Paulista (CP, 22.7◦S, 45.0◦W) stations on the Brazil side were selected. The geographic locations of these stations are indicated with blue rhombuses in Figure 1. The data is obtained from the Digital Ionogram Data Base (http://giro.uml. edu/didbase/scaled.php, accessed on 4 August 2021) [50].

**Figure 10.** Comparison between TEC differences and differences in the critical frequency of the plasma (foF2) and its height (hmF2). Observations (red circles), reference variation (black line), and P1–P4 time (shaded interval) are shown.

As a result, Figure 10 shows coherence between TEC and the critical frequency of the plasma (foF2) of each station. That is, the electron concentration after the eclipse maximum (∼11 UT) decreases (red circles) with respect to the reference curve (black line) calculated as indicated in Section 2.1. These same changes can be seen in the height of maximum electron concentration (hmF2). The decrease in foF2 and hmF2 is notorious at stations near the anomaly's crest (RA, FZ, CP); however, it is not very significant in the stations at the magnetic equator (SL). Moreover, similar ionospheric effects were seen in distant regions in the moon's shadow [16–20,51–56]. Differences between foF2 and TEC may be due to the fact that foF2 was the result of the original autoscaled records, and also that TEC was calculated from a spatial average.

A possible explanation for this phenomenon is that the eclipse could alter the thermospheric neutral wind regime and thus generate a ionospheric disturbance dynamo, which could be observed at the equator as a counter-electrojet. This counter-electrojet could be observed in the vertical drift of the plasma, for instance, the one measured by the Jicamarca incoherent radar. However, there are no measurements at Jicamarca for this period. Another way to observe is to calculate the difference in the horizontal component between an equatorial magnetometer and another in low latitude [57], or in the temporal variation of the same horizontal component of an equatorial magnetometer. In this case, neither the difference between Jicamarca (12.0◦S, 76.8◦W, I = 1◦)—Piura (5.2◦S, 80.6◦W, I = 11◦; available at http://lisn.igp.gob.pe/, last accessed on 22 May 2022), in the west coast of South America, nor the variation of the magnetometer of Kourou (5.2◦N, 52.7◦W; I = 13◦; available at https://intermagnet.github.io, last accessed on 22 May 2022), in the east coast of South America shows significant variations during the eclipse day with respect to the other days (figure not shown), which rejects this hypothesis. Another possible explanation could be that due to the fact that the partial eclipse begins at low latitudes (see Supplementary Materials, Video S1) the electron concentration never reaches normal values again. An eclipse also can cause effects on a global scale. Because the eclipse-induced abrupt cooling of the atmosphere can result in an instantaneous temperature shift and pressure differential, triggering AGWs, and associated TADs and/or TIDs. However, a detailed investigation of these causes is out of the scope of the current paper [17].

On the other hand, ionospheric effects in the magnetic conjugate of the eclipse (end of the white line in Figure 3, at 10.70 UT) are not possible to observe due to the lack of receivers in this region (see Figure 1).

On DoY 161, there was low ROTI activity in the western region of the United States of America, compared to the reference DoYs. The decrease in the percentage of GNSS stations in South America with weak ROTI activities caused the increase of stations without activity up to 89%. Then, the number of stations with strong ROTI activity only increased from 1% to 3% in this sector. However, the behavior of the ROTI activity in Greenland was less than the reference days (see Figures 6 and 7; Supplementary Materials, Table S4).

The behavior of ionospheric TEC and ROTI shows that electrons were less active in the ionosphere during the solar eclipse (see Table 1, and Figures 3, 4, 6–8 and 10). The behavior of the ROTI in the eclipse region was consistent with that indicated by Park et al. [26]. They found a significant reduction in the ROT during the eclipse. Furthermore, eclipse day was the least ROTI active in Greenland because we were able to observe a clear reduction in ROTI values compared to the other four DoYs (see Figure 6 (left panels), Figure 7 (upper panels)).

On the other hand, as a consequence of the geomagnetic activity (AE-index > 500 nT) in the polar regions from 5–15 UT on DoY 162; we can see an increase in ROTI activity (ROTI 0.25 TECu/min) that starts at the northern polar region, propagating later the increment toward the equator (∼50◦N), which agrees with previous studies [29,58,59] (see Figure 6 (left panels)). In South America, the percentage of stations with ROTI activity increased from ∼12% to 24% (see Figure 6 (left panels), Figure 7 (upper panels), and Supplementary Materials, Table S4).

The results obtained with the GNSS stations at 350 km (see Figures 3 and 8) were consistent with the ionospheric TEC behavior at 400 km above the Swarm-A (see Figure 4). At P1 time, we observe a TEC depletion (∼−1.7 TECu, −35%) in the central Atlantic region, where the eclipse started and its conjugate. The greatest TEC reduction (∼−2 TECu, −45%) occurred at GE time (see Figure 4 (upper panels)). This TEC value was similar to that reported by Cherniak and Zakharenkova (−2 to −1.5 TECu) [60]. From Figure 4 (buttom panels), we were also able to show that the disturbance remained in the North and South American regions (TEC ∼ −30%) even though the eclipse was already over the northern European and Asian regions. Moreover, we can observe a close similarity in the behavior of in situ Ne and VTEC (see Figure 4). Furthermore, the results of ionospheric plasma depletion using Swarm-A LP were consistent with the findings presented in [60].

#### *4.2. Ionospheric Impacts on GNSS Positioning Errors*

The manner in which we present the positioning errors in this work was through the statistics of perturbed stations around the world and, in particular, in the Greenland and South American sectors.

On the eclipse day, we could see a slight increase in the percentage of GNSS stations around the world that exceeded both positioning thresholds compared to previous days. The main increment suffered by the maximum 3D positioning error goes from ∼8% to ∼14%. Then, Greenland and the southern sector of America were within the regions that presented GNSS stations with the highest positioning errors during the eclipse time window. This positioning behavior in both regions was consistent with the global ionospheric TEC changes.

Contrary to what happens with the activity of electrons, the percentage of stations that exceed both positioning-error thresholds was greater on DoY 161 compared to DoYs 159, 160, and 163 (see Figures 6 and 7 (upper panels)). We could see similar behavior in both cases of the 3D-RMS and ROTI activity relationships (RMS&ROTI). The eclipse day was the second DoY with the highest percentage of stations that exceeded the positioning-error thresholds. The effects of the eclipse day were only exceeded by DoY 162 due to weak geomagnetic activity (AE-index >500 nT) in the polar regions from 5 UT to 15 UT. In South America, the behavior of maximum 3D positioning error 10 cm (34%) and 3D-RMS 3 cm (22%) on day 161 was similar to day 162. In Greenland, these parameters were also similar on days 161 and 162, where maximum 3D positioning error 10 cm was >55% and 3D-RMS 3 cm was ∼17%. However, the effects on positioning on DoY 162 were slightly higher (see Figures 6 and 7 (bottom panels)).

Our RMS position values for the quiet days were in accordance with those from previous results [29,30]. They showed that the precision of the post-processing kinematic PPP-AR method was 0.8 and 2 cm for the horizontal and vertical components, respectively. Moreover, our 3D-RMS results in percentage (3 cm = −4 to 225%) are consistent with the −4 to 324% presented by Park et al. [26].

Unlike previous studies [61–64], the results presented by Valdés-Abreu et al. [29], suggested that positioning errors also occur, regardless of whether the ROTI has rapid variations, with or without ROTI activity, in this type of DF-GNSS stations with the use of PPP-AR. Moreover, our results confirm that ionospheric disturbance sources can cause degradation of the GNSS accuracy (maximum 3D positioning error 10 cm and 3D-RMS 3 cm) when ROTI 0.25 TECu/min, ROTI 0.5 TECu/min, and without ROTI activity (see Table 1, and Figures 6–9). In addition, not all GNSS stations that had ROTI activity presented position errors.

Further, the ROTI activity–positioning variation relationship would have been met if two necessary conditions had been observed on each day in Figure 7. First, the positioning bars (RMS and MAX) had to be greater than or equal to the ROTI activity bar. This condition ensures that any ROTI activity causes variations in GNSS positioning. Second, the positioning ROTI bars (RMS&ROTI and MAX&ROTI) had to be the same or similar to the ROTI activity bar.

In most GNSS stations, we can observe the positioning errors were around the beginning of the TEC reduction (∼P1 time), the TEC peak (∼MOT and GE time), and/or in the final phase of the TEC recovery (see Figure 8). Then, we can see from one to more than three time slots with positioning errors. In general, after P1 time, the behavior of the stations with maximum 3D positioning error 10 cm is similar to the DVTEC [%] of the stations located in Greenland (see Figure 8 (left, center left panels), Figure 5 (left panels)) and South America (see Figure 8 (center right panels), Figure 5 (right panel)). For example, in the Greenland region, the ionospheric TEC depletion was significant until ∼14 UT, and the recovery also could be observed in the rapid decrease from 28% to 11% of stations that exceeded the threshold of maximum 3D positioning error ∼14 UT, where the persistence of the positioning errors provoked by the 10 June 2021 annular eclipse lasted ∼6 h. Although the TEC depletion in sectors of South America could be observed until

∼19–21 UT, a ionospheric TEC enhancement was observed around 16–17 UT, similar to the behavior of the GNSS stations with a maximum 3D positioning error greater than the threshold of 10 cm (from 28% to 22% between 16 and 17 UT). Therefore, the persistence of the positioning errors provoked by the 10 June 2021 annular eclipse lasted ∼10 h.

The annular eclipse in Greenland caused significant TEC changes (∼−60%), although with low ROTI activity. However, the GNSS positioning errors are similar to those caused during a weak geomagnetic storm with high auroral activity.

From Table 1, Figures 8 and 9, we see that the stations (PIFL, TRO1, GLS2, KUAQ, SENU) presented maximum 3D position errors >60%, also had A[%] −49%, but without ROTI activity (<0.25 TECu/min). PIFL station was the only one with ROTI activity over the 5-day period under consideration. Additionally, not all the stations that had A[%] −49% got maximum 3D erros >60% (BLAS, LEFN, IQAL, SVTL, LMMF, and CN57 stations). The LMMF station presented A[%] = −61%, but maximum 3D position errors = −8%, and ROTI = 0.2 TECu/min. Although the BLAS and LEFN stations had strong ROTI activity and the IQAL station had weak ROTI activity, the percentage of maximum 3D positioning error in these stations was between 24% and 31%.

Thus, the results suggest that when maximum 3D errors >60%, with respect to the maximum of the reference days, we can find A[%] −49%, but not the opposite. The results also reinforce the idea that ROTI activity is not a necessary condition to affect GNSS accuracy. We were not able to estimate the ionospheric effects on GNSS positioning in the magnetic conjugate region of MPO of the solar eclipse, due to the lack of GNSS stations in this region.

Our study showed that the ionospheric TEC disturbances due to the solar eclipse in the polar regions can produce disturbances in low and medium latitudes. Ionospheric changes can cause GNSS positioning errors. The estimation of these errors is critical in teleoperated and autonomous (ground, maritime, and aerial) applications and other highprecision activities. For example, mining, agriculture, and fishing are all key economic activities in Chile that are considering the use of more teleoperated or autonomous systems. If the positioning error in the GNSS receivers spikes in vehicles in these industries, it could impose a serious risk to people and infrastructure. For open-pit mines, a high error can generate a failure in the estimation of the terrace on which a vehicle is located, with the consequent risk of falling. Halting autonomous operations during some events such as eclipses can reduce potential risks, but they can be complex for these industries. Stopping the operation for even a short period of time, such as an hour, could be prohibitively expensive. Therefore, forecasting the impact should be precise in location and duration.

#### **5. Conclusions**

In this work, we analyzed the ionospheric behavior during the 10 June 2021 annular solar eclipse and its impact on DF-GNSS PPP-AR accuracy. We use a large global GNSS network located around the planet to estimate the effects on positioning. This solar eclipse had a trajectory over the northern polar region. We used global ionospheric TEC maps with data gathered by ground-based GNSS stations.

The TEC maps show a noticeable depletion under the moon's shadow, reaching A[%] < −60%. Furthermore, a significant TEC decrease (A[%] < −60%) can also be observed far from the ionosphere under the moon's shadow in regions close to the crests of the EIA over the Caribbean and South America, with a duration or ΔT over 10 h. Then, percentages of the ionospheric TEC over the Caribbean and South America were similar to those obtained for GNSS stations located in the region of the eclipse. Our study also confirms that there are cases and places where the disturbance can last much longer than previously expected.

We show that TEC enhancement caused by geomagnetic activity on the day after the eclipse did not cause problems in the ionospheric TEC background to our presented results for the eclipse day. We also validated the ionospheric variations estimated with GNSS receivers through measurements from other instruments such as the Swarm-A satellite (VTEC and in situ Ne), and four ionosondes (TEC, foF2, and hmF2). The ionospheric behavior clearly demonstrates that electrons are less active in that layer during the solar eclipse. Furthermore, our results are consistent with ionospheric effects reported in similar previous solar eclipses.

This study not only analyzes the eclipse's day but also compares the effects of the ionosphere and its impact on the positioning precision with those over 2 days previous and 2 days after the day of the eclipse. The day of the eclipse was the day with the second highest percentage of stations that exceeded the selected positioning thresholds (maximum 3D positioning error 10 cm, 3D-RMS 3 cm), only surpassed by the day after, which had geomagnetic activity. The data analysis shows that the eclipse had a significant effect on GNSS precision for a long time (∼10 h). The Greenland and South America sectors are within the regions that presented GNSS stations with the highest positioning errors during the eclipse time window. Moreover, both regions had the greatest ionospheric TEC decrease (∼−60%).

The ROTI variations were not relevant. Thus, the results reinforce the idea that ROTI activity is not a necessary condition to affect DF-GNSS PPP-AR accuracy. Additionally, the results suggest that when maximum 3D errors are larger than 60%, the A[%] is much less than −49%. However, the opposite is not necessarily true.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/article/ 10.3390/rs14133119/s1, Video S1: Eclipse obscuration mask from P1 to P4 time at 350 km altitude. Figure S1: Ionospheric TEC maps during the 10 June 2021 Annular Solar Eclipse: the world and Northern Hemisphere polar plots. Figure S2: DVTEC[TECu] maps using the Kriging interpolation method to the eclipse day. Table S1: Mean and standard deviation for each map of Figure S2. Table S2: 3D-RMS in Greenland and South America. Table S3: Maximum 3D positioning error in Greenland and South America. Table S4: ROTI in Greenland and South America.

**Author Contributions:** Conceptualization, J.C.V.-A., M.A.D. and M.B.; methodology, J.C.V.-A. and M.A.D.; software, J.C.V.-A., M.B. and Y.S.-S.; validation, J.C.V.-A., M.A.D., M.B., J.C.B. and Y.S.-S.; formal analysis, J.C.V.-A., M.B. and Y.S.-S.; investigation, J.C.V.-A., M.A.D., M.B., J.C.B. and Y.S.-S.; resources, J.C.V.-A., M.A.D., M.B., J.C.B. and Y.S.-S.; data curation, J.C.V.-A., M.B. and Y.S.-S.; writing original draft preparation, J.C.V.-A., M.B. and Y.S.-S.; writing—review and editing, M.A.D. and J.C.B.; visualization, J.C.V.-A., M.A.D. and Y.S.-S.; supervision, M.A.D. and J.C.B.; project administration, M.A.D.; funding acquisition, M.A.D. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Air Force Office of Scientific Research under award numbers FA9550-18-1-0249. This work was also partially funded by the ANID-FONDECYT 1211144 and the National Agency for Research and Development (ANID)/Scholarship Program/Doctorado Nacional/2018–21181599 (CONICYT Doctoral Grant Number 21181599). M.B. thanks the support of ANID-FONDECYT Regular 1211144 and FONDECYT Postdoctorado 3180742. J.C.B. was also supported by ANID PIA (ACT192169) and supported by Fondecyt project (Nº1200779 ANID, Chile).

**Data Availability Statement:** The RINEX files were obtained from: IGS stations; the Chilean network of GNSS receivers operated by CSN; UNAVCO; RAMSAC; Brazilian Network for Continuous Monitoring of IBGE; the Geoscience Australia; LISN; and AFREF. The satellite and receiver bias the data were obtained from AIUB Data Center of Bern University in Switzerland. The geomagnetic data were downloaded from: WDC for Geomagnetism, Kyoto; and OMNIWeb Plus Data Documentation. EQ data were available in the USGS Comprehensive Catalog of Earthquakes. The ionospheric TEC and LP measurements of in situ electron density data provided by ESA Swarm mission.

**Acknowledgments:** The authors would sincerely thank Miguel Martínez-Ledesma from Universidad de Concepción for his advice on plotting the eclipse obscuration mask.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **References**


## *Article* **Advanced Classification of Ionospheric Troughs in the Morning and Evening Conditions**

**Alexander Karpachev**

Pushkov Institute of Terrestrial Magnetism, Ionosphere, and Radiowave Propagation (IZMIRAN), 4, Kaluzhskoe Hwy, Troitsk, 108840 Moscow, Russia; karp@izmiran.ru; Tel.: +7-9164069331

#### **Highlights:**


**Abstract:** The separation and classification of ionospheric troughs in the winter evening and morning ionospheres of the southern hemisphere were performed using CHAMP satellite data for high solar activity (2000–2002). In the high-latitude ionosphere, the main ionospheric trough (MIT) was separated from the high-latitude trough (HLT). The separation was carried out using a thorough analysis of all the characteristic structures of the ionosphere in the framework of the auroral diffuse particle precipitation model. Two types of high-latitude troughs were identified: (1) a wide trough associated with zone II of diffuse precipitation on the poleward edge of the auroral oval and (2) a narrow trough of ionization, which is presumably associated with an electric field action. The poleward wall of MIT is as ever formed by diffuse precipitation in zone I on the equatorward edge of the auroral oval. The HLT and MIT separation is most difficult at the longitudes of the eastern hemisphere, where all structures are located at the highest latitudes and partially overlap. In the mid-latitude ionosphere, all the characteristic structures of the ionosphere were also identified and considered. MIT was separated from the ring ionospheric trough (RIT), which is formed by the decay processes of the magnetospheric ring current. The separation of MIT and RIT was performed based on an analysis of the prehistory of all geomagnetic disturbances during the period under study. In addition to the RIT, a decrease in the electron density equatorward of the MIT was found to be often formed at the America–Atlantic longitudes, which masks the MIT minimum. For completeness, all cases of a clearly defined polar cavity are also presented.

**Keywords:** main ionospheric trough; high latitude trough; ring ionospheric trough; low latitude trough; auroral diffuse precipitation

#### **1. Introduction**

The ionization trough was discovered by Muldrew from the Alouette 1 data [1]. Muldrew identified it as the main ionospheric trough (MIT). MIT is located equatorward of the auroral oval. The results of numerous MIT studies are summarized in reviews [2–5]. Inside the auroral oval, another (high–latitude, HLT) trough was detected and studied in detail using the OGO 6 satellite data [6]. The so-called ring ionospheric trough (RIT) was later discovered equatorward at MIT [7]. RIT is formed by the decay process of the magnetospheric ring current and is typically observed in the recovery phase of a storm/substorm [8,9]. The multitudes of MIT and HLT, as well as MIT and RIT, partially

**Citation:** Karpachev, A. Advanced Classification of Ionospheric Troughs in the Morning and Evening Conditions. *Remote Sens.* **2022**, *14*, 4072. https://doi.org/10.3390/ rs14164072

Academic Editor: Fabio Giannattasio

Received: 31 July 2022 Accepted: 18 August 2022 Published: 20 August 2022

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2022 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

overlap; therefore, the problem of trough separation arises in the region of the intersection. This problem was posed in prior studies [10,11] and elaborated in detail in a study by Karpachev [12]. However, this problem was not completely solved, so Liu and Xiong [13], processing a large dataset of CHAMP satellites, wrote: "There is no agreed quantitative definition of an MIT [14,15], as sometimes it is difficult to identify the individual midlatitude trough from kinds of depletion structures extending along latitude, such as highlatitude trough and wave-like disturbances". An almost unambiguous solution to the problem of trough separation in the midnight ionosphere was obtained in a recent study by Karpachev [16]. Significant progress in the separation and classification of various structures of the midnight high-latitude and mid-latitude ionosphere has been achieved based on a thorough analysis of a large dataset of CHAMP satellites, which allows us to consider the phenomenon from different perspectives. As a result, two troughs were identified in the auroral ionosphere. A wide trough (HLT1) was identified in the framework of a simple visual model of diffuse auroral precipitation [17]. This model describes zone I of diffuse precipitation on the equatorward edge of the auroral oval and zone II on its poleward edge. It was determined that the precipitation in zone II forms the poleward wall of HLT1 and the precipitation in zone I forms the MIT poleward wall. This is the key point in separating MIT and HLT1.

The analysis is most effectively conducted in the framework of the longitudinal effect because the boundaries of both zones change with longitude by 2.5◦ in latitude [18], similar to the longitudinal variations in the MIT position. Analyzing the longitudinal variations of these structures, it was found that the problem of MIT and HLT1 separation is radically different in the western and eastern hemispheres. In the western hemisphere, MIT is located at lower latitudes than in the eastern hemisphere, and it is further removed from the auroral oval; therefore, it is quite simple to separate it from the HLT. In the eastern hemisphere, MIT shifts to high latitudes so that the region of its existence begins to overlap strongly with the statistical position of zone I of precipitation and the region of existence of HLT. Therefore, the separation of troughs in the eastern hemisphere was carried out according to the correspondence of the MIT poleward wall to the position of zone I and the HLT1 poleward wall to the position of zone II of precipitation. In addition, the trough minimum position relative to zone I was also monitored. The second trough (HLT2), described in [6,19], is associated with the action of local electric fields [19]; therefore, it is narrow in latitude (3–5◦) and can be observed in any region of the auroral oval.

Because MIT is observed equatorward of the auroral oval, it is, by definition, a subauroral trough. The RIT is formed equatorward of MIT; therefore, it can be defined as a mid-latitude trough. RIT is formed even after a slight increase in geomagnetic activity and can be observed for a long time (sometimes two days) at latitudes of 53–57◦ GMLat [8,9]. Therefore, the frequency of RIT occurrence is determined by the degree of perturbation of the ionosphere, which is higher under high solar activity. The separation of MIT from RIT is also a difficult problem, but we can use the separation method carefully developed earlier [8,9]. It consists of the following: If MIT and RIT are observed simultaneously, then the equatorward trough is defined as the RIT. If there is one trough, its dynamics in prehistory must be considered. If the variations of the trough position correspond to the MIT model constructed in terms of the Kp index [20], then it is defined as MIT. If the trough is much more equatorward than the model position, then this is the RIT. However, there may be controversial cases that must be considered carefully. This will be shown below.

At longitudes of America and the Atlantic, a special structure is often recorded: a weak electron density minimum at the base of the MIT poleward wall and a deep minimum located much more equatorward. This structure is recorded in the evening more often than in the morning. The first minimum is located poleward of the average MIT position, whereas the second is, correspondingly, equatorward. Consequently, there is the problem of accurately determining MIT position, which is also discussed below.

Finally, to complete the pattern, the clearest cases of a polar cavity are also highlighted. However, a thorough analysis has not been conducted because, at most longitudes, the satellite did not reach the geomagnetic latitudes necessary for registration.

This work is an extension of the study of the irregular structure of the ionosphere in the early evening and early morning hours, based on the experience gained from the analysis of the midnight ionosphere [16]. In both studies, only the pattern in the southern hemisphere was considered. The southern hemisphere is much better suited for practicing this technique because there are a number of interesting effects in this hemisphere. This particularly applies to the longitudes of America.

#### **2. Materials and Methods**

The CHAMP satellite carried out in situ measurements of electron density *Ne* [21]. Variations in *Ne* are presented below in terms of plasma frequency *fp* (*Ne*[cm<sup>−</sup>3] = 1.24·104*fp*<sup>2</sup> [MHz]). The CHAMP altitude has changed from ~450 km to ~300 km, which is close to the height of the F2 layer maximum. It revolved on a nearly polar orbit with an inclination of 87◦. The CHAMP data time resolution of 15 s is less than 1◦ of latitude, which allows accurate determination of the minimum trough position. The CHAMP data are available on the website http://op.gfz-potsdam.de/champ (accessed on 12 January 2015).

In this study, CHAMP data for local winter conditions in the southern hemisphere were used. The data only for high solar activity with F10.7~180 sfu for the period of 2000–2002 and the evening (17–19 LT) and morning (05–06 LT) conditions were considered. About 1500 CHAMP passes in the winter high- and mid-latitude ionosphere for Kp ≤ 4 were examined. The MIT is usually defined as a sufficiently deep decrease in electron density of at least ~30% relative to the top of the steep equatorward wall [3]. However, in the morning sector, the electron density usually monotonously falls to high latitudes without noticeable inflection on the equatorward wall, and the trough definition becomes uncertain. Therefore, we have not estimated the level of electron density decrease at the MIT minimum. The position of MIT was determined by the electron density minimum at several degrees equatorward of the base of the poleward wall [3]. In the morning, the poleward wall was almost always clearly defined. If the trough was poorly expressed or masked by ionospheric plasma irregularities on some satellite path, the position of its minimum was determined through coordination with well-expressed troughs on neighboring paths.

Stricter criteria were imposed on the selection of the HLT. The HLT is observed in the auroral oval, where the electron density is highly irregular and several density minima can be observed. Therefore, the HLT was recorded only in obvious cases wherein it was clearly structured and when its poleward wall did not extend beyond the poleward diffuse precipitation zone. Similarly, the polar hole was defined only as a broad minimum of the electron density at latitudes above the poleward precipitation zone. Finally, only pronounced troughs were recorded equatorward of MIT.

#### **3. Structure of the Evening Ionosphere**

Figure 1 shows the positions of the different structures in the winter evening (17–19 LT) ionosphere of the southern hemisphere in terms of geomagnetic latitude—geographical longitude. The following structures are presented in Figure 1: polar hole, HLT, MIT, RIT and specific equatorward minimum of electron density. To eliminate the dependence on geomagnetic activity, the positions of the MIT, RIT and HLT were reduced to Kp = 2 according to Λcorr = Λc − *a*(Kp(τ) − 2), where Λc is the current position of the structure and the *a* factor is 2.0◦ for the MIT [20], 1.5◦ for RIT [9], and ~1.5◦ for HLT [6]. The Kp(τ) index was used because it considers the prehistory of geomagnetic activity development [22]. In the evening sector, the dependence of MIT position on local time is quite strong; it was also considered according to the model [20].

**Figure 1.** On the top: Longitudinal variations in the magnitude of the MIT poleward wall (dots and approximations). On the bottom: Longitudinal variations in the positions of main structures in the evening winter ionosphere of the southern hemisphere: polar hole (triangles), HLT1 (empty squares), HLT2 (filled squares), MIT (dots), RIT (red dots), equatorward density minima (purple circles), and low-latitude structures (green asterisks). The shaded latitude belts show diffuse auroral precipitation in zones I and II. The upper curve represents the CHAMP inclination equal to 87◦.

An analysis of the structures of the high latitude ionosphere was conducted using a model of auroral particle precipitation constructed from DMSP satellite data recorded in both hemispheres [17]. The model agrees well with other statistical models of auroral electron precipitation in the nighttime conditions [23–25]. There are no data in the literature regarding the dependence of the position of the diffuse precipitation zones on longitude in the evening sector. However, because the position of MIT showed a pronounced longitudinal effect [12], we preserved the shape and amplitude of the longitudinal effect in the auroral precipitation in the same form as revealed in the interval of 21–03 MLT [18] and used it to analyze the structure of the midnight high-latitude ionosphere [16]. For the southern hemisphere, these boundaries are presented in Figure S2 (in Supporting Information) according to [18]. The equatorward and poleward boundaries of the auroral oval experience synchronous longitudinal variations with an amplitude of ~2.5◦. In Figure 1, zones I and II of the diffuse precipitation taken from Figure S1 are shaded. The average (for all longitudes) position of the equatorward boundary of the auroral precipitation oval corresponds to model [17] for Kp = 2. The upper thin curve in Figure 1 depicts the CHAMP satellite inclination. The satellite inclination of 87◦ does not limit the observations of the discussed structures, except for the polar hole. However, polar hole cases are shown in Figure 1 solely for completeness of the pattern; only unambiguous cases were selected.

The longitudinal variations in the MIT position are quite confidently determined from the CHAMP data in the evening (correlation coefficient *r* = 0.72, standard deviation σ = 1.8◦). The amplitude of the longitudinal effect reaches ~6◦, which is greater than at midnight [16]. At longitudes of the eastern hemisphere between 0◦ and 120◦E, the MIT is located farthest from the pole; thus, this region overlaps with zone I of diffuse precipitation, even when considering the assumed longitudinal effect in the position of this zone. Even more important is that the clearly expressed MITs in this region are superimposed on well-defined HLTs. Regardless, the problem of separating MIT and HLT in the evening sector is more complicated than in the midnight sector and requires detailed consideration. Figure 2 shows several examples of the structure of the evening ionosphere, which require careful analysis. The shaded bars in Figure 2 show zones I and II of diffuse precipitation, which positions correspond to the current values of the longitude and Kp index. In some cases, the zones are slightly shifted to better compliance of the structures on the grounds that CHAMP provided the current data, and the precipitation model is statistical.

**Figure 2.** Characteristic examples of ionospheric troughs at different longitudes in the evening ionosphere. The shaded bars in (**a**) (zones I and II of precipitation) correspond to *fp* profile 1, i.e., Kp = 2+; the bar in Figure 2h corresponds to *fp* profile 1, i.e., Kp = 2.

The most difficult problem with the separation of MIT and HLT is observed in the eastern hemisphere at longitudes of 30–90◦E, where the locations of MIT and HLT partially overlap. The weakly pronounced poleward wall of MIT at these longitudes (upper panel of Figure 1) complicated this problem. The data on the poleward wall were obtained during the quiet period from 10–17 July 2001.

The situation at problematic longitudes is considered below using individual characteristic examples. In Figure 2a, curve 1 depicts an example of a conventional MIT. Its minimum is located at quite a high latitude (−70◦), but its poleward wall is definitely formed by diffuse precipitation in zone I. The MIT poleward wall is low, and the electron density minimum at −77◦ is apparently a sign of a high-latitude trough (HLT). Curve 2 depicts an example of a well-defined HLT with a minimum latitude of −76◦ (under Kp = 3+), i.e., deep inside the auroral oval. Its poleward wall is certainly associated with zone II of precipitation. This type of high-latitude trough has been defined as HLT1 [16]. The MIT on latitudinal *fp* profile 2 is completely filled, and only ionospheric plasma irregularities are observed at latitudes of its assumed minimum. This latitudinal *fp* profile is likely a consequence of the previous disturbance with Kp = 4+. Thus, the main criterion for the separation of MIT and HLT1 is the correspondence of the poleward wall of the trough to precipitation in zones I and II, respectively. The position of the trough minimum was also taken into consideration.

Figure 2b shows examples of the simultaneous existence of two electron density minima. Then, a lower-latitude minimum (<70◦) corresponds to the MIT, and a higherlatitude one (>70◦) refers to the HLT. The poleward walls of both troughs correspond to the related precipitation zones, although the MIT poleward wall is not pronounced in either case.

Curve 1 in Figure 2c represents a trough with a minimum at a very high latitude of −72◦ for Kp = 2. However, its poleward wall is clearly formed by precipitation in zone I, i.e., MIT. Of note that, in this case, MIT is located deep in the region of HLT existence. The latitudinal profile 2, as well as in Figure 2b, shows two *fp* minima. The equatorward minimum at latitude of −71.7◦ (under Kp = 2+) coincides with the MIT minimum on curve 1. However, its weak poleward wall is formed by only one *fp* value, i.e., it is not reliable. Therefore, this structure was attributed to the HLT, in which the poleward wall is located at latitudes of (−77–79◦), i.e., at latitudes of zone II of diffuse precipitation. Thus, as the above examples show, the identification and separation of troughs in the challenge region at longitudes of 30–90◦E was possible, although it required a thorough analysis of complex, controversial cases.

In Figure 2d, MIT and HLT are observed at the same time. However, in this case, the high-latitude trough is very narrow and deep. Such troughs were recorded onboard the OGO 6 satellite [6]. The authors of this study associated these troughs with the action of electric fields in the region of the high-latitude convection of ionospheric plasma. This trough can be defined as HLT2. In Figure 1, the approximating curve is determined for both HLT types. This curve shows the longitudinal effect in the HLT position, but it is distorted because, at most longitudes, the inclination of the satellite's orbit does not allow recording of the highest-latitude cases of HLT. (The HLT shown in Figure 1 sometimes goes beyond satellite inclination, which is associated with the reduction of the position of the troughs to Kp = 2).

In Figure 1, in the longitude range from 150◦ to 240◦E, several HLTs are located at very low latitudes, equatorward of the polar oval. Such a case is presented in Figure 2e for HLT1. In this case, the MIT minimum and its poleward wall are much more equatorward than zone I of precipitation. We can deduce that zone I is located more equatorward at most longitudes than in Figure 1. However, this assumption cannot be verified without data on precipitation at all longitudes in the evening sector.

In Figure 2f, three troughs are observed simultaneously: HLT1, MIT and RIT. A wellexpressed RIT was observed after a substorm with Kp = 5. The RITs in Figure 1 are indicated by red (filled) circles. The RIT is formed mainly in the western hemisphere. This is because the geomagnetic field in this hemisphere is weak, the precipitation of hot ions of the magnetospheric ring current is intense, and the trough is formed more often.

In Figure 2g, curve 1 shows the polar cavity at latitude of −82◦, MIT at latitude of −68◦, and the *fp* minimum at latitude of −61.5◦. The polar cavity is observed at latitudes >78◦; therefore it was recorded only at longitudes of the eastern hemisphere, where the satellite inclination allowed. The *fp* minimum is depicted by the green asterisk in Figure 1 at a longitude of 110◦E. It clearly stands out from among other troughs, and the reason for its formation is unknown, although it is very clearly expressed.

Curve 1 in Figure 2h displays the structure often observed at the longitudes of America and the Atlantic, as discussed in the Introduction. This structure is characterized by a small *fp* minimum at latitude of −62◦ at the base of the MIT poleward wall and a deeper and wider minimum at latitude of −55◦. This minimum makes it difficult to accurately determine the position of the MIT minimum. These equatorward electron density minima are marked in Figure 1 by purple circles. Figure 1 shows that both the RIT and the equatorward minimum are located in the same region, mainly in the western hemisphere. However, the RIT is formed after geomagnetic disturbance, and a structure with two electron density minima is formed under long-term quiet conditions. In addition, this structure has a specific shape.

Curve 2 in Figure 2h depicts a structure at a longitude of 352◦E, formally similar to the structure represented by curve 1. However, the *fp* peak in this case is so pronounced that it represents a special independent structure. In other words, a question arises regarding the reasons for the formation of this peak. The structure in question was formed in a weakly disturbed ionosphere at Kp = 4−, but it is often observed at American longitudes under completely quiet geomagnetic conditions. The *fp* peak is located on the MIT equatorward wall so that the MIT minimum appears at a latitude of −65.3◦. This peak forms a pronounced equatorward minimum at latitude of −57◦, which can easily be mistaken for MIT with automatic and even manual data processing. However, even at Kp = 4−, this minimum is much more equatorward than the "normal" MIT, as shown in Figure 1.

The top panel in Figure 1 shows the longitudinal variations in the electron density on the top of the poleward wall derived with correlation coefficient *r* = 0.68, standard deviation σ = 0.43◦, and amplitude *A*~1.5◦. At small and especially large longitudes, the MIT poleward wall is surprisingly low. Another specific feature of the evening ionosphere at these longitudes is the slow decrease in electron density from middle to high latitudes, i.e., the small latitudinal gradient at the latitudes of MIT. This demonstrates curve 1 in Figure 2i. Latitudinal *fp* profile 1 was obtained on 3 August 2002 in the longitudinal sector of 340◦E. The electron density slowly and almost monotonously decreases to high latitudes, showing neither a noticeable minimum nor a poleward wall. Such *fp* profiles are quite often observed at the discussed longitudes. Thus, in the evening, at small and large longitudes, the MIT as a structure is often not formed at all, neither its minimum nor the poleward wall. This is an unexpected fact that requires comprehension.

Latitudinal *fp* profile 2 in Figure 2i was obtained on 25 July 2002 under almost the same conditions in the same longitude sector. However, the latitudinal *fp* profile differs sharply from the usual profile at these longitudes. It is characterized by a high electron density at high latitudes, which quickly drops toward the equator. As a result, MIT is not observed, but a well-defined minimum is formed at an extremely low latitude of −50◦. The whole structure of the ionosphere on this path looks strange. This structure has been recorded in the evening sector only once; a low latitude minimum is also marked by a green asterisk in Figure 1 at a longitude of 340◦E and latitude of −50◦. A possible explanation is associated with a strong particle precipitation equatorward of the "normal" boundary of diffuse precipitation, that is, at the latitudes of the MIT minimum, which is consequently filled with ionization [26]. A minimum latitude of −50◦ is probably a consequence of the formation of a subauroral polarization steam (SAPS) [27]. The study [27] provides an example of the formation of a trough at a latitude of 53◦, according to the Millstone Hill radar and the DMSP F13 satellite measurements on 12 April 2001. The deep trough was formed under the action of the western component of the plasma drift, driven by a northward electric field. Thus, Figure 2h,i demonstrates the presence of problems in the identification of MIT at large longitudes in the western hemisphere.

As shown above, the main problem in the eastern hemisphere is the separation of high-latitude cases of the MIT from low-latitude cases of the HLT. The main criterion for separation is the correspondence of the MIT poleward wall to the precipitation in zone I and the HLT1 poleward wall to the precipitation in zone II. The solution to this problem was presented in sufficient detail above in the analysis of Figure 1. However, it is worth demonstrating this solution once again based on a very illustrative example obtained on 1 August 2001. Figure 3 on the left shows the longitudinal variations in the position of the troughs and both zones of diffuse precipitation. The Kp index varied in the period under review from 2− to 4−. To eliminate dependence on geomagnetic activity, the data were reduced to Kp = 2. Figure 3 shows that, in the western hemisphere, the trough minimum is located equatorward of the auroral oval of the precipitation and, by definition, is the MIT.

**Figure 3.** (**Left**): Longitudinal variations in the trough position on 1 August 2001. Precipitation zones I and II are hatched. (**Right**): Latitudinal *fp* profiles for the paths marked on the left.

Examples of the latitudinal cross-sections of the MIT are shown in Figure 3 on the right for paths 31, 1 and 3. The trough on these paths is clearly expressed; its poleward wall exactly corresponds to zone I of precipitation at a latitude of approximately −70◦. In the eastern hemisphere, the troughs on paths 13–23 are located inside the auroral oval and by definition belong to the HLT. Examples of well-formed HLTs are shown on the right for paths 13, 17 and 19. Their poleward walls are located at latitudes of 76–80◦, i.e., they are formed by precipitation in zone II. Path 7 is transitional; the poleward wall of the trough corresponds to zone II of precipitation, and the electron density minima at its wide bottom corresponds to both MIT and HLT. On paths 17 and 19, weak variations in ionospheric plasma are observed in the latitudinal region at approximately −66◦, which can be assessed as a sign of MIT. On the other paths in the eastern hemisphere, there are no even weak manifestations, i.e., MIT has not formed in the eastern hemisphere. If the MITs in the western hemisphere combine with the HLTs in the eastern hemisphere into a single branch, there will be strong longitudinal variations in the trough position with an amplitude of ~11◦. Note, however, the artificial character of the longitudinal effect in this case.

In Figure 2h, curve 2 depicts a specific structure, which is characterized by a shallow electron density minimum at the base of the poleward wall and a well-expressed minimum, which is much more equatorward than the "normal" minimum of the MIT. Because this structure is quite often formed at longitudes of America and the Atlantic (purple circles in Figure 1), it is worth considering its formation in detail. Figure 4 on the left shows the longitudinal variations in the trough position for quiet conditions (Kp varied from 1− to 2+) for the period of 4 July 2001. Both zones of diffuse precipitation are plotted in Figure 4. The thick curve shows the average position of MIT, which is highlighted in Figure 1. The numbers of characteristic satellite paths are indicated. The classic MIT is observed on path 14, and its poleward wall is formed by precipitation in zone I. In the eastern hemisphere, MIT is located slightly above the average position. MIT gradually enters zone I of precipitation, and on path 20, it transforms into HLT1, the poleward wall of which is already formed by precipitation in zone II. The MIT on path 20 did not manifest in any way, as in the previous event, at the same longitude (Figure 3).

**Figure 4.** On the **left**: longitudinal variations on 4 July 2001 in the MIT position (black circles) and equatorward density minimum (blue squares). The average position of MIT (from Figure 1) is shown by a thick curve, and the polar circle is a dashed curve. Hatching shows the position of both zones of precipitation and the plasma density peak. On the **right**: *fp* latitudinal cross-sections for the paths indicated on the left.

A remarkable structure was formed in the western hemisphere. On paths 2, 4 and 6, a well-expressed ionospheric plasma peak is observed at latitudes of 62–65◦, which is much more equatorward of the zone I precipitation. This peak defines the steep poleward wall of the trough. The appearance of plasma peaks distant from the auroral oval may be due to two reasons. Sufficiently strong particle precipitation sometimes occurs in the evening sector equatorward of the stable boundary of diffuse precipitation, i.e., already inside the trough [26]. The resulting peak of ionization can be amplified and shifted further to the equator by the zonal drift created by the SAPS [27], as the SAPS flows developed nearby. A small minimum *fp* is observed at the base of the poleward wall. Its position is plotted in Figure 4 as the MIT minimum because we have no other way to determine the position of the trough; it is masked by the well-defined equatorward minimum of *fp*. On path 2, this minimum is located at a latitude of −55◦, which does not correspond to the minimum of the "normal" MIT at all. Then, this minimum shifts to high latitudes up to −61◦ on path 8. As a result, an ordinary MIT with a poleward wall formed by zone I precipitation was observed on path 8.

The dashed curve in Figure 4 shows the position of the polar circle at a geographic latitude of −66◦. Thus, the observed minimum *fp* is obviously associated with the decay of electron density during the polar night. Automatic data processing of the MIT position has been frequently used recently. It is clear that any program will take the equatorward minimum as the minimum of MIT in Figure 4. Then, the longitudinal variations in the trough position, considering paths 20 and 22, will reach 18◦, and the average position of MIT in the western hemisphere will be underestimated. There is a serious problem in determining the position of MIT because such a situation, as in Figure 4, is observed quite often. In the study by [2] on a large set of ISIS 1 and Injun 5 satellite data, it was determined that the minimum MIT was located 2–5◦ away from the auroral oval. This is true for the average position of the trough (the bold curve in Figure 4). However, in this case, even the base of the poleward wall is 6–7◦ away from the auroral oval. Thus, we were forced to take the minimum *fp* at the base of the poleward wall as the minimum of MIT.

#### **4. Structure of the Morning Ionosphere**

Figure 5 shows the longitudinal variations in the position of the main structures of the morning ionosphere. The designations are the same as in Figure 1. The CHAMP data for 05–06 LT were used. They were selected in the interval of Kp < 4 and were again reduced to Kp = 2. The longitudinal variations in the MIT position from the CHAMP data are less confidently determined in the morning (*r* = 0.55, σ = 2.1◦) than in the evening because of the small amplitude *A*~4◦. Longitudinal variations in the electron density on the MIT poleward wall (upper panel) were obtained during the quiet period from 3 July to 12 July 2001. These variations in the morning sector, as well as in the evening, are reliably revealed (*r* = 0.65, σ = 0.50◦, *A*~1.5◦).

**Figure 5.** The same as in Figure 1, but for 05–06 LT.

Although in the position of the morning MIT, a weaker longitudinal effect was observed than in the evening; nevertheless, we retained the shape and amplitude of the longitudinal effect in precipitation in the same form as they were revealed for 21–03 MLT in [18]. The sharp difference between the morning and evening ionospheres is that MIT and HLT are separated by a large gap in the morning. Considering the results of the midnight ionosphere study [16], a trend can be identified: MIT is most closely located to the auroral oval in the evening, farther away from the oval at midnight, and even farther in the morning. Accordingly, the problem of separating MIT and HLT is most acute in the evening and simplest in the morning.

Examples of the most characteristic structures in the morning ionosphere are shown in Figure 6. Figure 6a shows an example of HLT1 recorded on 8 August 2002 at a longitude of 196◦E. The poleward wall of the HLT is formed by precipitation at latitudes near −76◦, which corresponds to zone II of precipitation. Only a slight decrease in electron density is observed at the latitude of MIT.

**Figure 6.** Characteristic examples of ionospheric troughs at different longitudes in the morning ionosphere.

At longitudes from 150◦ to 200◦E, several MIT cases are located so high in latitude that their identification can be questioned. In Figure 6b, the minimum trough is located at a very high latitude of −64.6◦ for 5.0 LT and Kp = 2−. This is because the sharp minimum electron density corresponds to the base of a very steep poleward wall. However, the poleward wall is certainly formed by the precipitation in zone I, and this is MIT. Moreover, on this latitudinal *fp* profile, there is another high-latitude trough with the poleward wall located at a latitude of −78◦, which corresponds to the precipitation in zone II.

Figure 6c clearly identifies three structures recorded on 9 August 2002 in the longitudinal sector of 182◦E at Kp = 4−: the polar cavity, HLT and MIT. The high-latitude trough is quite narrow; therefore, it was assigned to HLT2. In Figure 5, it is located at the lowest possible latitude of −66◦ (reduced to Kp = 2), but it cannot be confused with MIT. Thus, even in the challenge longitudinal interval of 150–200◦E, the MIT and HLT can be separated based on a thorough analysis. However, observing a trough path by path at longitudes of 120–210◦E, the MIT can transfer into HLT and vice versa, and with a cursory analysis, it may not be noticed. This situation is demonstrated in Figure 3 for evening hours. The low poleward wall aggravates the problem, as well as in the evening sector.

The structures of the morning ionosphere are located at much lower latitudes than in the evening. As a result, the satellite inclination makes it possible to record the HLT at almost all longitudes. Therefore, the approximating curve for the HLT in Figure 5 reproduces the longitudinal variations in the HLT position more adequately than in Figure 1 for evening hours. The polar cavity in the morning is also much more equatorward than in the evening and is often superimposed on the precipitation in zone II. However, a detailed analysis of the polar cavity was not included in the objectives of this study.

On 17 July 2001, under Kp = 4, two troughs were also recorded: MIT and equatorward RIT (Figure 6d). The RIT is formed much more often in the morning than in the evening [9], which confirms the comparison of Figures 1 and 5. The morning RIT is, as in the evening, observed more often at longitudes with a weak geomagnetic field. Figure 6d shows the case in which MIT and RIT are separated very clearly. However, this is not always observed; therefore, we must separate them by analyzing the prehistory of the dynamics of both troughs during geomagnetic disturbance, starting with the main phase of the storm/substorm, as mentioned in the Introduction.

Figure 6e shows two examples of a structure that is most often recorded in the evening (Figure 2h, curve 1), but it is sometimes observed in the morning at longitudes in America and the Atlantic. It is characterized by a high poleward wall (compare with Figure 6b), shallow minimum or inflection of *fp* at its base and a deep minimum of *fp* far from the poleward wall. This structure is discussed in detail in Figure 4. In Figure 5, the equatorward minima of electron density is marked with purple circles. There are much fewer of these minima than in the evening. They are usually located far equatorward from the MIT average position. In any case, this minimum masks the MIT minimum and makes it difficult to determine the true MIT position.

The structure in Figure 6f looks like the structure in Figure 6e. However, in this case, the *fp* minimum at the latitude of −51.7◦ is more than 10◦ away from the base of the poleward wall and is accompanied by a sharp increase in *fp* to the equator. As a result, the structure in Figure 6f is similar to the structure that was observed in the evening in Figure 2i (curve 2) but is even more pronounced. In addition, in the evening, this structure is rarely observed; however, because the event repeats, a question arises about the mechanism of its formation.

In Figure 6g, two *fp* minima are observed in the longitudinal sector 269◦E. The *fp* minimum at a latitude of −58.2◦ corresponds exactly to the average position of MIT for Kp = 2 in Figure 5.

The minimum latitude of −62.8◦ is located on the base of the steep poleward wall. The observed *fp* profile can be interpreted as an irregular structure at the bottom of the trough. However, the question remains as to why the deep minimum of the electron density is so often formed on the base of the sharp poleward wall of MIT.

The latitudinal *fp* profile in Figure 6h shows the deep and narrow low-latitude trough (LLT) on the equatorward wall of MIT at a latitude of −45.5◦. Another such LLT was observed at a longitude of 72◦E and latitude −45◦ (see Figure 5). Troughs at such low latitudes are observed quite rarely and probably belong to troughs associated with hot particle precipitation from the inner radiation belt [28] or with the penetration of electric fields deep into the plasmasphere [29].

The structure in Figure 6i also defies unambiguous identification. A deep *fp* minimum is observed at a latitude of −47.5◦. This is 4–5◦ equatorward than observed on previous days at the same Kp and LT values. This minimum was formed after a weak disturbance with Kp = 3+. Even after such a weak disturbance, the RIT is often formed. However, in this case, it cannot be stated unambiguously. Nevertheless, in Figure 5, this trough is defined as the RIT. Equatorward of RIT at latitude of −47.5◦ another LLT is observed.

#### **5. Discussion**

The analysis of the localization of different structures of the high-latitude and midlatitude ionospheres in terms of the longitudinal effect proved to be effective. This is because the positions of all structures, including the auroral oval of particle precipitation, experience quite strong variations with longitude. The MIT position changes with longitude by 6◦ in the evening and by 4◦ in the morning. The position of the boundaries of the auroral diffuse precipitation changes by 2.5◦, according to the results, which was obtained so far only in

one interval 21–03 MLT [18]. When variations in the position of the precipitation boundaries in the morning and evening sectors are revealed, the mutual location of the auroral oval and MIT can be clarified. However, we are confident that this will not fundamentally change the results of the analysis performed in this study.

In [2], the average distance of 2–5◦ between the auroral oval and MIT was determined. Now, we can refine this estimate. In the evening, MIT's average position is approximately 3◦ away from the equatorward border of diffuse precipitation, 4.5◦ at midnight [16], and as much as 6◦ in the morning.

Because MIT in the eastern hemisphere is located at a higher latitude than in the western hemisphere, the region of its existence in the eastern hemisphere is partly superimposed on zone I of precipitation and the location of the HLT. Therefore, the problem of MIT and HLT separation is most difficult at longitudes of 0–120◦E in the evening and at midnight and 150–200◦E in the morning. The problem is complicated by the fact that, at these longitudes, the poleward wall of the MIT is poorly expressed. Therefore, the main criterion for separating troughs is the correspondence of the MIT poleward wall to the precipitation in zone I and the HLT poleward wall to the precipitation in zone II. The position of the trough minimum also matters. All of this makes it possible to almost unambiguously separate MIT and HLT 1. However, to do this, we should carefully analyze each case because when observing path after path, situations arise when the MIT shifting to the pole transforms into HLT1 and vice versa. At the same time, only weak traces of MIT can be observed on the equatorward wall of HLT1, which is difficult to identify as a trough.

As for HLT2, the narrow trough described in [6] is most often observed at midnight, less often in the evening and even less often in the morning.

The polar cavity was recorded only where the inclination of the satellite allowed, i.e., at longitudes of 60–210◦E. Analysis of the polar cavity was not a priority task; therefore, only obvious cases were recorded. The polar cavity was observed at latitudes above −76◦ in the morning and at those above −78◦ in the evening.

The RIT was also separated from MIT. For this purpose, the prehistory of the development of all, even weak disturbances, was considered. RIT is more often observed in the morning than in the evening. In the evening, it localizes exclusively at the longitudes of the western hemisphere, where the geomagnetic field is weak. In the morning, the RIT is observed at all longitudes, including in the eastern hemisphere.

In the evening, an extremely specific region of the ionosphere forms at the longitudes of America. First, in addition to the RIT, a special structure is often formed here (and partly over the Atlantic): a weak minimum of electron density is observed at the base of the poleward wall of the MIT, and a deep minimum is formed equatorward. This equatorward minimum is located at a much lower latitude than the statistical minimum of MIT. This seems to be associated with the decay of the electron density beyond the polar circle. This minimum masks the minimum of MIT, and with careless data processing, the true position of MIT can be greatly underestimated. Second, the poleward wall of MIT, probably driven by a horizontal drift, sometimes shifts to the equator, and then the entire structure of the high-latitude ionosphere is located much more equatorward than usual (see Figure 4). In this case, the amplitude of the longitudinal variations in the trough position can reach an incredible 18◦. Third, the poleward wall of MIT at the longitudes of America is very low. Moreover, the electron density often monotonically falls to high latitudes without showing a noticeable minimum. As a result, the trough does not form at all at these longitudes.

#### **6. Conclusions**

The main result of the study is an almost unambiguous solution to the problem of separating different ionization troughs in the morning and evening ionosphere. The problem of the MIT and HLT separating is more complicated in the eastern hemisphere, the problem of the MIT and RIT separating is more complicated in the western hemisphere. Considering the results of [16], it can be argued that the problem of separation and classification of troughs in the nighttime ionosphere of the southern hemisphere is solved. This led to a

decrease in the data scatter in the MIT position, the standard deviations are 1.8◦ in the evening, 1.85◦ at midnight, and 2.1◦ in the morning, which are less than in other statistical studies (see, for example, [10,30,31]). This is important for creating an accurate MIT model.

**Supplementary Materials:** The following supporting information can be downloaded at: https: //www.mdpi.com/article/10.3390/rs14164072/s1. Figure S1: Model of auroral particle precipitation: diffuse auroral zone I equatorward of aurora (blue), structured auroral oval precipitation (auroral lights region or aurora, green), and soft diffuse precipitation zone II (orange) poleward of aurora. Figure S2: Longitudinal variations in the averaged auroral precipitation energy flux at 21–03 MLT under Kp = 2 for the June solstice (Jun.) in southern hemisphere [18].

**Funding:** This research received no external funding. The author would like to thank sponsors and operators of the CHAMP mission; Deutsches GeoForschungsZentrum (GFZ) Potsdam and German Aerospace Center (DLR).

**Data Availability Statement:** The CHAMP data are available on the website: http://op.gfz-potsdam. de/champ (accessed on 12 January 2015).

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Remote Sensing* Editorial Office E-mail: remotesensing@mdpi.com www.mdpi.com/journal/remotesensing

MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel: +41 61 683 77 34

www.mdpi.com

ISBN 978-3-0365-5904-9