In this work, we used ionospheric parameters of the vertical electron density profiles obtained from Digisonde, called ionograms. This equipment is a high-frequency (HF) radar that transmits radio waves continuously into the ionosphere ranging from 1 to 30 MHz (Reinisch et al., 2009). We used the Digisonde data from CXP and CG over the Brazilian sector provided by the Brazilian Studies and Monitoring of Space Weather (EMBRACE) program (available at http://www2.inpe.br/climaespacial/portal/en/, last access: 2 October 2021).

We evaluated the F region behavior using height parameters, such as the virtual height (h′F) and peak height (hmF2), which is important to investigate the changes in this region. Also, we analyzed the frequency parameters of the F and E regions and Es layer (foF2, foE, and fbEs, respectively), which are related to the electronic density at the layer peak. The fbEs is the frequency at which reflection from a layer at superior heights starts to be visible in ionograms. The time resolution is 10 min from the ionograms in stations considered here.

The GNSS receiver data were used to obtain the total number of electrons (TEC) in a given ionospheric path. TEC is a measurement of the electrons in a column of unitary cross-sectional area between the satellite and the receiver. Due to the high number of stations over the Brazilian sector, it is possible to construct the two-dimensional maps of the absolute vertical TEC values ranging from 50 to ∼500 km of spatial resolution in latitude and longitude, every 10 min (Otsuka et al., 2002; Takahashi et al., 2016). This current work analyzes the EIA during the eclipse occurrence using these maps, available online on the EMBRACE website.

We have used a theoretical model, called MIRE, which provides the E region and Es layer electron densities as follows (Carrasco et al., 2007; Resende et al., 2017a, b, 2020, 2021):

1ne=O2++NO++[O+]+N2++Fe++Mg+. This model solves a set of partial differential equations of the continuity and momentum between 00:00 and 24:00 UT in the height range from 86 to 120 km for the main molecular or atomic ions in the E region (NO+, O2+, N2+, O+), as well as metallic ions (Fe+, Mg+).

The MIRE continuity equation of each constituent Ni,

2∂[Ni]∂t=P-L-∂VizNi∂z, is used to calculate the ion density by taking into account the production (P), loss (L), and transport ∂VizNi∂z. The transport term depends on the wind and electric field parameters (Resende et al., 2020). This analysis only considers the E region chemistry, thus neglecting the transport terms and metallic ions. More details about the MIRE model can be found in Carrasco et al. (2007) and Resende et al. (2017a).

We divided this study into two ionospheric responses of the solar eclipse that occurred on 14 December 2020: (1) the ionospheric changes in the F region, and consequently EIA behavior, and (2) the E region and Es layer behavior during the solar eclipse hours. Figure 1 shows the solar eclipse evolution between 15:45 and 18:00 UT for every 15 min at 250 km height. The colors mean the obscuration varying from 10 % (purple) to 100 % (yellow). Notice that only two Digisonde stations (CXP and CG) had a solar eclipse obscuration over the Brazilian sector with available data. In CXP, the solar eclipse influence starts about 16:15 until 18:00 UT, while in CG it was between around 16:00 and 17:15 UT. Therefore, we analyzed these regions since the solar eclipse provides a great opportunity to study the responses of the ionospheric regions to the rapid solar radiation variation.

The interplanetary medium parameters are measured from the Proton and Alpha Monitor (SWEPAM) and Magnetic Field Experiment (MAG) instruments aboard the Advanced Composition Explorer (ACE) spacecraft (Stone et al., 1998). Figure 2 shows the solar wind speed, Vp (a), proton density, Np (b), and Bz component of the Interplanetary Magnetic Field (IMF) (c) measured at the L1 Lagrangian point. Also, we show the Dst index in panel (d). The data for a solar eclipse that occurred on 14 December 2020 are shown with the orange line, and we used a reference period on 4 December 2020 that is shown with the blue line.

The Vp observed in the eclipse day and the quiet period is concentrated below 400 km s-1, which is considerably slow wind (Tsurutani et al., 2011; Isaacs et al., 2015). The proton density fluctuates around 2 particles cm-3 during almost the entire eclipse day, except in a short period between 16:48 and 19:12 UT, reaching ∼5 particles cm-3. This short period can be associated with the solar sector boundary crossing (figure not shown here). However, these values are still considered low. Also, the Bz component fluctuates around zero during almost the entire eclipse day, reaching the maximum negative value of ∼-2.5 nT during a short time and presenting few positive incursions (maximum of ∼2 nT). To confirm that the event occurred on a geomagnetically quiet day, we show the Dst index in Fig. 2d. Although this parameter showed a minor enhancement during the solar eclipse event starting at 14:00 UT, the values remained very low, oscillating around zero. Finally, all the interplanetary parameters showed that this day is not geomagnetically disturbed in terms of the ionosphere influence. Therefore, this event is an excellent opportunity to analyze the ionospheric influences during a solar eclipse occurrence.

In Fig. 3, we investigate the minimum F layer virtual height h′F (top) and F layer peak height hmF2 (bottom) to both analyzed regions in panels (a) CG and (b) CXP between 14:00 and 19:00 UT. The blue lines are the height parameters for the quietest day of the month (4 December 2020), whereas the red lines refer to the solar eclipse day on 14 December 2020. The grey line represents the solar eclipse obscuration for each region. At CG, the maximum obscuration occurred at 16:45 UT, reaching 20 %. On the other hand, the maximum obscuration occurred at 17:15 UT in CXP, with the most substantial value of 29 %. Unfortunately, we do not have data over Santa Maria station (29.7∘ S, 53.8∘ W; dip ∼37∘ S) during the solar eclipse hour occurrences, where the maximum obscuration reached 53 %.

In both regions, we observe a strong fluctuation with high values of the h′F before the solar eclipse onset. This behavior happens because a C4.0 class solar flare occurred between 14:09:00 and 14:56:00 UT, causing a radio blackout of the E and Es layer regions and partially the F region (Nogueira et al., 2015). The h′F has low values concerning the quiet reference value after 16:15 and 17:00 UT for CG and CXP, respectively. At the same hours, a constant decrease was observed in the hmF2 parameter for these regions.

To better observe this scenario, we show the ionograms for both regions. Figure 4 refers to CG on 14 December 2020 for 15:00, 17:30, and 17:50 UT. Notice that in hours before the solar eclipse, the F1 region is present as the red arrow indicates. However, we observe that the F1 region has completely disappeared at 17:30 UT, returning at 17:50 UT. Here, we believe that the F1 region can suffer from the lost ionization, as discussed in Fargues et al. (2001). In fact, as the peak obscuration for the CG station occurred around 16:20 UT, the absence of the F1 region can be associated with the recombination processes during the solar eclipse. The same behavior seems to occur in CXP, as is shown in Fig. 5. In this case, we observe the F1 layer at 16:50 UT. Around 17:10 UT, this layer disappears completely. However, a strong Es layer of type c (Esc) caused by tidal winds (Resende et al., 2017a) appears also and can block the F region. This absence of the F1 region lasted until 18:10 UT when this layer occurred again (as seen in the ionogram at 18:20 UT). Thus, in both regions, the F1 absence lasts around 1 h, and we consider that it is due to the solar eclipse event and the Es layer presence, which can block the F region. All these characteristics make the height profile of these regions decrease significantly, as we saw in the h′F and hmF2 parameters in Fig. 3.

Chandra et al. (2007) studied the ionospheric effects of the total solar eclipse of 11 August 1999 over the Ahmedabad region (23∘ N, 73∘ E). The authors did not find any decrease in the critical frequency of the F1 layer. However, Minnis (1955) analyzed the E and F1 layers during the solar eclipse of 25 February 1952. The author rewrote Chapman's equation by considering the fraction of the ionizing radiation lost during the unobscured times. The theoretical results showed significant weakness in the F1 layer due to the loss of electrons caused by an effective recombination process. More recently, Adeniyi et al. (2007) showed the solar eclipse effect on the ionosphere over an equatorial station (8.53∘ N, 4.57∘ E; dip 4.1∘ S) in the African region. This event was on 29 March 2006, and the maximum obscuration was 99 % in this station. One of their results was the evident absence of the F1 regions in ionograms over the station analyzed. An explanation was that the electron density in the layers became so thin during the solar eclipse events that the ionosonde could not detect it. However, unlike our result, the E region also disappears in Adeniyi et al. (2007). In Figs. 4 and 5, the E and Es layers are evident, leading to uncertainties about the solar eclipse effect in the ionosphere on 14 December 2020. We believe here that the electron density decreases due to the recombination factor during the solar eclipse. The Esb layer simultaneous appearance resulted in a significant weakening of this layer, making detection by the Digisonde difficult.

We do not observe significant differences in the F2 layer densities in the ionograms data from these stations. Figure 6 shows the foF2 parameter for CG (a) and CXP (b) during the reference day (blue line, 4 December 2020) and during the solar eclipse event (red line, 14 December 2020). The grey line represents the solar eclipse obscuration for each region. Notice that over CG, the F region electron density (related to the foF2) was smaller than the reference day since the previous hours of the solar eclipse event. Over CXP, the foF2 values are practically the same on the 2 d analyzed. We credit this behavior to the low solar eclipse obscuration (20 %–30 %) over the ionospheric stations. The loss processes were insufficient to weaken the electron density at the F region, as seen in other events (Adeniyi et al., 2007).

The short ionization interruption could affect the EIA over the South American sector, as shown in Fig. 7. This figure shows the TEC behavior through the maps (Takahashi et al., 2016), in which we have two high-density areas between 20–30∘ S and 40–60∘ W that characterizes the EIA. The red line refers to the magnetic equator, the circles (Fig. 7c) refer to the eclipse area at 17:00 UT, and the color scale in Fig. 7a and b indicate the TEC intensity from 0 to 50 TECU (1 TECU =1016 electrons m-2). The EIA results from the equatorial plasma downward flows along magnetic field lines because of the diffusion and gravity. Therefore, two plasma crests are seen over the off-equatorial region, in the Northern Hemisphere and Southern Hemisphere (Nogueira et al., 2011). This behavior is evident in Fig. 7a, when it is possible to observe the EIA peak around ±15∘ in TEC maps at 17:00 UT, being stronger in the southern regions. Notice that, on 14 December 2020 (Fig. 7b), we observed a clear weakening of the EIA crests compared with the typical behavior of the ionosphere plasma (Fig. 7a).

To better observe this difference, Fig. 7c shows the relative difference (RD) parameter over the TEC maps computed by

3RD(%)=(TECSE-TECRef)/TECRef×100. The RD is calculated through the TEC maps for the solar eclipse day (TECSE) concerning the typical day (TECRef). This result shows that the TEC is between 30 % and 50 % smaller during the eclipse occurrence over the Brazilian sector.

Vyas and Sunda (2012) analyzed the TEC changes during an annular solar eclipse over the Indian sector on 15 January 2010. They showed a TEC reduction in the EIA localization that was named as inhibited EIA region. They attributed this behavior to the combined effects of the solar eclipse, which induce attenuation of extreme ultraviolet (EUV) solar irradiation and the inhibited equatorial electrodynamics, affecting the EIA. The negative deviation was 20 %–40 % in the inhibited EIA region. C. H. Chen et al. (2019) modeled the EIA dynamic variations during a solar eclipse that occurred on 21 August 2017, around North America and South America. They also obtained the TEC difference, and their results showed a significant reduction around EIA regions at solar eclipse times. Huang et al. (2020) studied the ionospheric responses at low latitudes in a solar eclipse on 21 June 2020. The authors also found that the EIA decreases significantly in the solar eclipse hours. In summary, Table 1 presents some recent studies that observed the EIA decrease during the solar eclipse events. Hence, we show the work reference, the solar eclipse date, and the percentage of the EIA decrease concerning the typical periods.

All the studies cited before concluded that the ionization loss caused variation in the dynamic processes during the obscuration times, affecting the EIA behavior. The main hypothesis is that if the solar eclipse goes through the equatorial regions, the fountain effect will change, and consequently, less plasma density reaches the low latitudes. As predicted by Martínez-Ledesma et al. (2020), in our work, we believed that the minor TEC reduction around the magnetic equator locations was enough to cause a plasma density decrease in both EIA crests. Another hypothesis here is that the partial absence of the radiation over the Peruvian equatorial sector changes the conductivity in this locality and, maybe, reaches the entire equatorial area. Thus, we reduce the equatorial density, and, consequently, the fountain effect can be affected. Therefore, although this solar eclipse event almost did not reach the equatorial regions, we suppose it was enough to influence the fountain effect. Also, Huang et al. (2020) mentioned that during the eclipse events the transequatorial northward or southward neutral wind can weaken, causing a reduction of the EIA crests.

Le et al. (2009) showed an analysis of the ionosphere in the conjugate hemisphere during the solar eclipse on 3 October 2005. Their main result is a decrease in the electron temperature in both conjugate points, which is associated with a reduction in the photoelectrons traveling along the magnetic field lines from the eclipse region to the conjugate region. Thus, the authors proposed that solar eclipse events can cause a disturbance in the ionospheric regions in the conjugate hemisphere. In such an analysis, the TEC decreases around 32 % in the 300 km. Recently, Zhang et al. (2021) analyzed the TEC perturbations in the south/north EIA crests of the solar eclipse on 21 August 2017. They found that in the southern crest of the anomaly the TEC reduced significantly, while in the northern crest it stayed almost undisturbed. They mentioned that there is a northward motion tendency for plasma within the flux tubes that can inhibit the typical diffusion of the equatorial fountain effect.

We already analyzed the TEC around the days of the solar eclipse (13 and 15 December 2021). Here, we found the same behavior for the southern crest of the EIA, characterized by a decrease of around 30 %. For the northern EIA crest on 13 December we also observed a reduction in the TEC, and on 15 December, we did not see any significant modification (not shown here). Thus, this work establishes that it is possible that the plasma movement on the flux tube modified the equatorial fountain effect. However, to validate all these hypotheses, it is necessary to consider other equipment, which will be carried out in future work.

The E region is dominated by the production and loss process in the ionosphere. Therefore, it is expected that the electron density of this layer suffers an influence during eclipses. Figure 8 shows the variation of the E region critical frequency parameter (foE) in CG and (b) CXP between 14:00 and 19:00 UT. The blue lines refer to the quietest day of the month (4 December 2020), and the red lines refer to the solar eclipse day on 14 December 2020. The grey line represents the solar eclipse obscuration for each region. We note that the ionization starts to decrease around 15:40 and 15:50 UT for CG and CXP, respectively. Moreover, the values remain low until 18:00 UT with respect to the quiet reference day.

We observe a decrease of around 15 % in the ionization in the E region during the solar eclipse obscuration (decrease from 4 to ∼3.3 MHz). Chernogor et al. (2019) analyzed the solar eclipse effects on 20 March 2015, on the mid-latitude daytime ionospheric plasma using observations from Kharkiv incoherent scatter radar (49.60∘ N, 36.30∘ E). They show that in heights less than 210 km, it is common to see a decrease in the electron density, mainly in the maximum phase of the solar eclipse. They found that the electron density reduced by 18.5 % in this event. Also, they believe that the explanation is related to the E region chemistry since the loss for recombination in these heights is quadratic. Thus, as the ionization radiation from the Sun has been removed in this region, the loss processes become effective quickly, and the E region density is affected directly. Nonetheless, Rishbeth (1968) had already reported that during partial or total eclipse events, the recombination process is not enough for the electron density to decay drastically. In fact, the authors mentioned that the duration time of the eclipse is not sufficient to affect the E region chemistry, making it disappear or diminish substantially. This fact can explain the reason for the E region electron density decrease of 15 % in our data.

Figure 9 shows the E region electron density simulated by MIRE in height–time–intensity (HTI) maps over CG (left panel) and CXP (right panel). The profile background shows the regular E region, which is described by the significant electron density values in the daytime and low values during the nighttime. The metallic ions and transport terms in Eqs. (1) and (2) were negligible since they are not important to analyze the E region. We have the E region profile considering the usual conditions in panel (a) and in panel (b) we reduced the E region ionization by 15 %.

As we expected, the results show an E region ionization reduction for both regions. We do not observe any differences in the E region behavior concerning CG and CXP. In these two sites, the electron density maximum in the model decreased by 4.92 electrons cm-3 and 4.85 electrons cm-3 (in log scale) at around 15:00 UT (as shown in the contour in the Fig. 9). Therefore, we show a significant reduction in the E region electron density, which is mainly driven by chemical processes, which is directly affected by events such as eclipses, even if they are not total.

Le et al. (2008) studied the changes on the E and F region parameters during a total solar eclipse that occurred on 11 August 1999. In such an analysis, they used the ionosonde network, over the European sector, and simulations. Their results show a high agreement between the solar eclipse responses in the E region density (NmE) between the ionosonde data and the Theoretical Ionospheric Model of the Earth in the Institute of Geology and Geophysics, Chinese Academy of Sciences (TIME-IGGCAS). The authors mentioned that the E and F1 regions are mainly dominated by the photochemical process, and for this reason, a clear electronic density occurring synchronously with the solar eclipse beginning is observed. The modeled and observational results in Le et al. (2008) are in according with our analysis here.

Some studies showed an intensification in the Es layers during eclipses. The main hypothesis is that atmospheric gravity waves can be induced during the total/partial solar eclipse and affect the vertical wind shear, strengthening the Es layer (Chen et al., 2010a, b; Yadav et al., 2013). However, other studies such as Pezzopane et al. (2015) showed that the solar eclipse did not affect the Es layer in terms of its intensity. In fact, the authors analyzed a partial effect of the solar eclipse that occurred on 20 March 2015 in mid-latitudes. Although their results did not show any Es layer intensity modification, they observed an evident influence in its time duration. The Es layer lasted longer, and they attributed this effect to the traveling ionospheric disturbances (TIDs) likely caused by gravity wave propagation.

Therefore, we evaluated the fbEs parameter over (a) CG and (b) CXP, as shown in Fig. 10. The blue line refers to the quiet period reference and the red line to the solar eclipse day. In CG, the Es layer did not occur at almost any time of the day, which can be related to the weak wind that could not form the Es layers in this period (Resende et al., 2017a). Thus, it is not possible to analyze the Es layer in this region. On the other hand, over CXP, we noted an interesting behavior: a peak in the fbEs at around 17:10 and 17:30 UT. To better analyze this fact, we show the ionogram (c) at 17:20 UT, indicating that this parameter reached values higher than 6 MHz. After these hours, it returns to the typical values around 4 MHz (not shown here).

The Es layer seen over CXP is of type c, which is very common for this region and related to the zonal component of the wind. As we observe a clear increase of this layer electronic density, the hypothesis that the gravity waves influenced the winds becomes plausible. To verify if the gravity waves occurred, we use the Digisonde data as described in Abdu et al. (2009). The upper panel of Fig. 11 shows the temporal variations in the true height of the ionogram fixed frequencies of 5, 6, and 7 MHz over CXP. Notice that an oscillation between 13:00 and 19:00 UT is apparent, indicating the presence of gravity waves. In the bottom panel of the same figure, we plotted the d(hF) using a band-pass filter (30 min–3 h) to remove considerable F region height gradients. We can observe a well-noticed in-phase oscillation during the eclipse hours (∼ 16:00 and 18:00 UT). This fact means that there was gravity wave propagation in the E region that reached the F region. Also, by the dashed lines during the eclipse hours, we observe clear gravity waves propagating upward. It is important to mention here that we observe oscillations before the solar eclipse. However, during the solar eclipse hours, it seems that there was an intensification of the velocity drift. This fact explains the atypical Es layer in Cachoeira Paulista in these hours. In this work, we only show the gravity waves presence to be a possible cause of the Es layer density enhancement. Thus, the gravity waves may have intensified the Es layer over CXP, as proposed by Chen et al. (2010a, b) and G. Chen et al., 2019.