On the Sources of the Ionospheric Variability in the South American Magnetic Anomaly During Solar Minimum

We investigate for the first time the variability of the F2 layer critical frequency (foF2), its peak height (hmF2), the thickness parameter B0, and the E region critical frequency (foE) over Santa Maria (29.7° S, 53.7° W, dip angle = −37°), a station located in the central region of the South American Magnetic Anomaly. The selected ionospheric parameters were obtained from ionograms recorded by a recent Digisonde Portable Sounder 4‐D. The time period covers 309 days from 1 September 2017 to 30 August 2018. The diurnal analyses revealed a large day‐to‐day ionospheric variability, with some peculiarities as a strong semiannual pattern superimposed to expected ionospheric behavior. Furthermore, the results show significant differences between the averaged foF2 in December and June solstices, revealing a possible presence of the annual asymmetry. The coefficient of variation (CV) is used as a quantitative description of the variability of each parameter versus time and season. Considering low solar flux and geomagnetically quiet days only, we note that CV is smaller during the daytime and larger during nighttime for all parameters. The least variable ionospheric parameter in our study is foE, while the most variable one is B0. Regarding the F2 layer parameters, we observe that foF2 is much more variable than hmF2. We attribute the observed CV to the neutral atmosphere source over Santa Maria. The ionospheric variability is in general enhanced during geomagnetically disturbed periods. The estimated CV is higher over Santa Maria than Wuhan, China (30.5° N, 114.4° E, dip angle = 46°), a station with no influence of the South American Magnetic Anomaly.


Introduction
It is well established that production, loss, and transport processes that occur in the ionosphere is subject to many influences from the Sun, magnetosphere, and lower and middle atmosphere, which deviate it from the climatological mean. This deviation is broadly known as ionospheric variability and can occur on hourly, daily, seasonally, and solar cycle time scales (Araujo-Pradere et al., 2005;Forbes et al., 2000;Rishbeth & Mendillo, 2001;Zhang & Holt, 2008). Rishbeth and Mendillo (2001) describe the possible causes of ionospheric variability into four categories: (i) solar ionizing radiation (solar flares, 27-day solar rotation, formation and decay of active regions, 11-year solar cycle, and solar zenith angle dependence), (ii) solar wind-geomagnetic activity (substorms, geomagnetic storms, energetic particle precipitation, and Joule heating), (iii) neutral atmosphere-meteorological (through the upward propagating waves like planetary, tidal, gravity, and infrasonic waves and their modulations), and (iv) electrodynamics (prompt penetration electric field, plasma convection at high latitude, and electric fields from lightning sprites). All these sources of ionospheric variabilities impose certain challenges for studying and/or modeling the ionosphere that may improve existing models. Once ionospheric measurements are usually limited in space and time, any quantitative measurements of the ionospheric variability using some of the most important parameters of the E and F regions have been an important part of space weather research and modeling. Such measurements are especially important when the region under study encompasses certain regional peculiarities like the South American Magnetic Anomaly (SAMA; Abdu et al., 2005).
Several authors have studied the ionospheric deviation using different methods and parameters for determining the variability. Forbes et al. (2000) obtained some estimates of how ionospheric variability is divided between solar ionizing flux, solar wind conditions, and neutral atmosphere influences. The authors found that the daily fluctuations in the peak electron density of the F 2 layer (N m F 2 ) under quiet geomagnetic conditions range from 25% to 35% at periods from few hours to 1-2 days and vary between 15% and 20% at periods from 2 to 30 days. They assumed that these fluctuations are due the neutral atmosphere sources since the deviation of solar flux variation is small within these time scales. The authors analyzed separated zones of latitude without considering fluctuations by daytime/nighttime periods or by seasons. Such diurnal/nocturnal period and seasonal ionospheric variability study was made by Rishbeth and Mendillo (2001). They found that the N m F 2 variability is greater by night than at day and greater at equinox than at solstice months. It is also higher at subauroral and equatorial latitudes than at midlatitudes. Some other major efforts to study the E and F regions variabilities were conducted over several regions, most of them in the Northern Hemisphere (Abe et al., 2013;Akala et al., 2010;Bilitza et al., 2004;Medvedeva & Ratovsky, 2015;Somoye & Akala, 2011;Tsagouri et al., 2018;Zhang et al., 2004;Zhang & Holt, 2008). In the South American sector, the short-term variabilities (few days in 2002) of the F 2 peak parameters were discussed based on data collected in three Conjugate Point Equatorial Experiment (COPEX) station, Boa Vista (2.8°N, 60.7°W, dip angle = 22°), Cachimbo (9.5°S, 54.8°W, dip angle = −4°), and Campo Grande (20.5°S, 55°W, dip angle = −22°), by Abdu et al. (2009) and McNamara et al. (2008). Moreover, Batista and Abdu (2004) analyzed the parameters f o F 2 , h m F 2 , and B0 obtained by digital ionosondes over two stations in Brazil, São Luís (2. 6°S , 44.2°W, dip angle = −0.5°) at the magnetic equator and Cachoeira Paulista (22.7°S, 45°W, dip angle = −28°), close to the southern crest of the Equatorial Ionization Anomaly (EIA). The authors compared their observational data with the modeled results by International Reference Ionosphere, in which it showed some discrepancies in specific hours. Recently, Romero-Hernandez et al. (2018) studied the day-to-day variation of the ionosphere in the Northern and Southern Hemispheres over the Latin American sector but using daytime total electron content variation. However, the previous mentioned works in the South America sector did not explore or discussed the dispersion of ionospheric data from their mean values qualitatively.
To the best of author's knowledge, it seems that the prior studies about ionospheric variabilities using ionosonde data have focused in F region parameters collected in the Northern Hemisphere, equatorial region in Africa, or fall within borders to one of the crests of the EIA. Moreover, no previous work appears to have examined simultaneously the variabilities in the E and F region parameters. It is noted that the ionospheric variability would be understood more when the E region is studied simultaneously with the F region parameters. Therefore, there exists a gap in proper understanding the ionospheric variability in South America, especially in the Brazilian sector where there are some unique geophysical characteristics. One of them is the high magnetic declination angle that reaches approximately −20°in the equatorial region due to the dip equator secular displacement effect causing considerable changes in the Equatorial Electrojet and consequently in sporadic E (Es) layers Moro et al., 2016aMoro et al., , 2016bMoro et al., 2017;Resende et al., 2016). There is also a more pronounced energetic particle precipitation in the SAMA region (Moro et al., 2013(Moro et al., , 2012 when compared with other regions having the same latitudinal range, due to the global minimum in the intensity of the geomagnetic field. Thus, this paper uses, for the first time, data acquired by a Digisonde Portable Sounder 4-D (DPS-4D) recently installed at Santa Maria (29.7°S, 53.7°W, dip angle = −37°), a low latitude station located close to the center of the SAMA. The period of study covers 309 days from 1 September 2017 to 30 August 2018, which is characterized by low solar activity with few geomagnetically disturbed days. We use the coefficient of variation (CV) as an index of variability of f o F 2 , f o E, h m F 2 , and B0, which will be defined in next section, to study their detailed behaviors in the SAMA region. The results obtained over Santa Maria is compared with the ionospheric data acquired by a DPS-4D installed at the Wuhan Ionospheric Observatory, China (30.5°N, 114.4°E, dip angle = 46°), a station with no influence of the SAMA. The positions of Santa Maria and Wuhan are marked by the black stars in the map of Figure 1. The solid black line across the continents is the geomagnetic equator. In addition, we included the locations of several other stations that will be mentioned in the discussions. Therefore, the main objectives of this work are (1) to discuss the characteristics of the data and the system recently installed in Santa Maria, (2) to analyze the diurnal and seasonal patterns of the ionospheric variability through the selected parameters for the whole year of data, (3) to analyze the sources of ionospheric variability over Santa Maria, (4) to compare the results over Santa Maria with Wuhan during two geomagnetic storms, and (5) to discuss the physical reasons of the obtained results.

Data and Method of Analysis
Monitoring Meridian Project of China (Wang, 2010), the Brazilian Study and Monitoring of Space Weather (Embrace) Program from the Brazilian National Institute for Space Research (INPE/MCTIC), and Federal University of Santa Maria. The system fills a gap of ionospheric sounders operating in real time between the Cachoeira Paulista station and the Port Stanley station (51.6°S, 57.9°W, dip angle = −49.8º), located in Argentina, as can be seen in Figure 1. The Digisonde in Santa Maria is set to transmit radio waves continuously into ionosphere from 1 MHz and increases the frequency up to 20 MHz with the sweep rate of 25 kHz for each round. The train of echoes to form an ionogram is transmitted/received with a 5-min time resolution.
An example of the daytime ionogram collected at Santa Maria station at 18:00 universal time (UT) on 12 July 2017 is shown in Figure 2. The local time (LT) is LT = UT − 3 hr. The modern ionosondes like the DPS-4D measure the polarization of the received signals identifying the O and X echoes by the red and green pixels as labeled in the figure. The critical frequencies obtained from ionograms provide direct measurements of electron density at the layer peak. In the example ionogram, the lower penetration frequency for the F 2 region, f o F 2 is 6.10 MHz. Thus, the electron density at the F 2 layer peak can directly be obtained by N m F 2 (el m 3 ) = 1.24 × 10 10 (f o F 2 /MHz) 2 . On the other hand, the altitude of the density peak of F 2 layer, h m F 2 , is not measured directly (Huang & Reinisch, 2001), but it can be obtained after processing analysis. In the example ionogram, its value is 222.8 km.
The lower penetration frequency for the E region, f o E, is 2.9 MHz for the case shown in Figure 2. The physics involved in this region is simpler than for the F region, and its main features are well described by the Chapman theory (Hargreaves, 1995). The f o E presents a well-known variation with the solar zenith angle and satisfies f o E = A (cos χ) 0.25 , where A is determined by electron production and loss rate (Titheridge, 2000).
In addition to the E and F layers, Figure 2 also shows the calculated vertical electron density profile. The α-Chapman estimate of the F 2 layer topside is shown as the dashed black line. We also choose to analyze the parameter B0 since it provides a measure of the thickness of the bottomside profile and it is used in the empirical models like the International Reference Ionosphere. B0 is calculated as the height difference between h m F 2 and the height where the electron density has dropped to 0.24 × N m F 2 (Bilitza et al., 2000).
Other features usually observed in the ionograms collected in Santa Maria are the skywave signals sent by the Digisonde installed at Campo Grande station. The traces between 10 and 13.5 MHz are the oblique incidences from Es, F 1 , and F 2 layers, identified as 1-hop E Es, 1-hop F1, and 1-hop F2, respectively. Our ionograms can therefore simultaneously register the vertical incidence and oblique incidence traces. However, the system is limited to automatically identify signals arriving within zenith angles of less than 15°as vertical incidence and those with zenith angles larger than 15°as oblique incidence. The automatic program simply disregards any signal flagged as oblique incidence. All recorded ionograms used in this work are initially auto-scaled by the Automatic Real Time Ionogram Scaler with True Height integrated into the Digisonde Ionogram Data Visualization/Editing Tool (SAO Explorer) developed by the Center for Atmospheric Research, University of Lowell Massachusetts. Nevertheless, in order to obtain reliable f o F 2 , f o E, h m F 2 , and B0 parameters, each ionogram was checked and manually edited. The reason is that the high magnetic declination angle, which is a peculiarity in the Brazilian sector, affects the physics of the E and F layers development (Moro et al., 2017), controlling the F layer dynamo development in evening hours (Abdu et al., 2005). Therefore, manual scaling is a wide technique used in ionograms collected in Brazil to achieve a good quality of ionospheric parameters.
In order to quantify the ionospheric variability over Santa Maria, which is the focus of this paper, we use two methods for analyzing the data: (1) using the mean values approach and (2) using the CV, also known as relative standard deviation, as an index of variability. The CV has been used for several authors to express the variability in the ionospheric parameters (Akala et al., 2010;Forbes et al., 2000;Rishbeth & Mendillo, 2001;Somoye & Akala, 2011;Tsagouri et al., 2018;Zhang & Holt, 2008, among others). The CV is statistically defined in Equation (1) as In this equation, the standard deviations (σ) and the mean (μ) values of the parameters f o F 2 , f o E, h m F 2 , and B0 calculated for certain time intervals are used to obtain the respective percentage relative standard deviation: The seasonal effects are studied in this work by using the mean and standard deviation data from 3 months that make up each season as follow: December solstice (73 days of November, December, and January), March equinox (65 days of February, March, and April), June solstice (80 days of May, June, and July), and September equinox (91 days of August, September, and October). The solar emission at the wavelength of 10.7 cm (F 10.7 , measured in 10 −22 Wm −2 Hz −1 ) and geomagnetic activity index Kp, for looking at their influence in the ionospheric variability, if any, are provided online (https://omniweb.gsfc.nasa.gov/form/ dx1.html).
In the next section, we discuss the diurnal and seasonal patterns of f o F 2 , f o E, h m F 2 , and B0 observed in the central region of the SAMA during the last solar minimum (2017)(2018). In the following, we quantify the ionospheric variability through the CV during geomagnetically quiet days. At last, we show two case studies considering the CV calculated for geomagnetically disturbed days. The results obtained for Santa Maria are compared with the CV estimated from ionospheric data acquired in Wuhan, a station with no influence of the both SAMA and EIA.

Diurnal and Seasonal Variations of the Ionospheric Parameters Over Santa Maria
An overview of the f o F 2 , f o E, h m F 2 , and B0 observed over Santa Maria is depicted in Figure 3. The mass plots in the left panels display 309 single days variations from 1 September 2017 to 30 August 2018 of the f o F 2 and f o E parameters in megahertz and h m F 2 and B0 parameters in kilometer from top to bottom. The diurnal variation of these parameters in the seasonal scales are presented in the right panels of Figure 3. The UT runs along the horizontal axes, and the LT scale is at the top of the Figures 3a and 3e.
The sunrise varied from 9:15 UT to 10:30 UT from 1 September 2017 to 30 August 2018 over Santa Maria, while the sunset took place between 20:43 UT and 23:39 UT. Therefore, we consider daytime hours from 9:00 UT (06:00 LT) to 21:00 UT (18:00 LT). The number of data available during nighttime hours is small as compared to the daytime hours. This is due to the occurrence of spread F at night, which manifest as diffused echoes on the ionograms after sunset. The spread F may be associated to the plasma bubble occurrences, regions in the ionospheric plasma where the electron density is reduced due to the Rayleigh-Taylor instability process (Abdu, 2001).
It can be seen from Figure 3 that all parameters show diurnal and a large day-to-day variability during all seasons. We observe some peculiarities in the ionospheric parameters superimposed to expected ionospheric behavior already described in preview works. One of the peculiarities seen in the diurnal variations of f o F 2 is the strong semiannual pattern, which is the larger f o F 2 values for equinoxes than for solstices. From the left panels, we observe an exceptionally high afternoon variation with few observations from 6 to 9 MHz between 14:00 UT and 22:00 UT. The high afternoon variability between 250 and 300 km in h m F 2 is slightly seems in a shorter period, from 14:00 UT to 19:00 UT. The diurnal variation patterns show that f o F 2 , f o E, and B0 are higher during daytime hours, while the opposite holds for nighttime hours. Regarding the h m F 2 values, it presents, in general, opposite behavior. The nighttime values are higher than those of the daytime values except for periods around 9:00 UT in June solstice and around 14:30 UT in December solstice. It is also remarkable the mean differences between December (red dots) and June (black dots) solstices, which are presented in all parameters in the right panels, reveal a possible presence of the annual asymmetry (sometimes called by annual or nonseasonal anomaly). The annual asymmetry, in the world as a whole, is an annual variation of the maximum electron density approximately 20% greater in December than in June (Rishbeth & Müller-Wodarg, 2006;Zou et al., 2000). A subtle difference between March (green dots) and September (blue dots) equinoxes during the afternoon has also been detected.
The general behavior of f o F 2 presented in Figures 3a and 3e shows clearly that the diurnal trend depends upon the rotation of the Earth through the photoionization process of the neutral atmosphere by solar radiation. The plasma production is small at night if there is no energetic particle precipitation in the SAMA region (Moro et al., 2013). The lowest predawn values of f o F 2 occur between 8:00 UT and 9:00 UT, while its daytime local maxima are observed in the afternoon with time depending the season. The post-sunrise increase is observed in all the seasons until pronounced peaks of 10

Journal of Geophysical Research: Space Physics
and day-to-day variability in f o F 2 is caused basically by the EIA influence and not by the presence of the SAMA. In other words, the effects of the EIA in the ionospheric parameters are more intense than the ones caused by the SAMA.
As can be noted from Figures 3b and 3f, the f o E parameter presents a regular and smooth diurnal variation, increasing almost linear from sunrise to the peak in the afternoon and falling until sunset. In the nighttime hours, the detection of the E region is difficult under the condition of low electron density, and the Digisonde is not able to detect this layer. During geomagnetically disturbed days, which is not the case presented in Figure 3, it is possible to detect the E layer due to the presence of the SAMA. The diurnal variation of f o E responds positively to the intensity of the Sun, which is a well-known characteristic of the E layer (Abe et al., 2013;Ouattara et al., 2009 Gong et al. (2012). In addition to these factors, the nighttime increase in h m F 2 values may be also explained by the action of equatorward meridional winds over Santa Maria, which pushes the electron density up along the magnetic field lines. Unfortunately, our study has been limited by uncertainties in the knowledge of the neutral wind in the subtropical region of Santa Maria. On the other hand, Kalita et al. (2015) studied the characteristics of the F 2 layer parameters over Dibrugarh (27.5°N, 95°E, dip angle = 43°), a midlatitude station in India. The authors pointed out that winds could be a dominant factor in the h m F 2 behavior over Dibrugarh because the winds produce the maximum effects in height at this magnetic dip angle. As Santa Maria is located at a magnetic dip angle of −37°, there is a high probability that the effects of winds in the ionospheric F 2 height be a dominant mechanism for h m F 2 behavior observed in our results. After midnight, the h m F 2 values fall to lower altitudes where the neutral density and the chemical loss are greater and f o F 2 and B0 also decreases. A precipitous decrease between 8:00 UT and 9:00 UT (when the solar zenith angle is greater) is observed in h m F 2 , except in June solstice because it occurs 1 hr earlier.
The behavior of rapid ascent/descent may be explained by the increase of electron concentration near F 2 peak by solar radiation. Thus, the height of the F 2 layer shifts downward, showing a dip in the h m F 2 due to the fast plasma production. This is also observed at Jicamarca by Lee and Reinisch (2012). We also observe a small increase in h m F 2 soon after sunrise hours until noon especially in December solstice and equinoxes, while this increase is not evident in June solstice. The peaks may be explained in terms of the increase in the temperature due to daylight that increases the concentration of atomic oxygen, as explained by Ameen et al. (2018)

Journal of Geophysical Research: Space Physics
Since N m F 2 is proportional to (f o F 2 ) 2 , we note in Figure 3e a big difference in the F 2 layer ionization between December solstice months (November, December, and January) and June solstice months (May, June, and July). The difference between the mean f o F 2 values in these months may be the annual asymmetry, but it can only be separated from the seasonal anomaly by combining data from opposite seasons in the two hemispheres, which it is not the focus of this work. Regarding the annual asymmetry that might be the case we are observing here, it is characterized by greater average N m F 2 in December than June, both by day and by night (Zou et al., 2000). Rishbeth and Müller-Wodarg (2006) explained that a possible cause is the December/July variation of 3.5% in Sun-Earth distance and consequent 7% variation in the flux of ionizing radiation, associated with the asymmetries of the geomagnetic field and winds. We remind the reader that Santa Maria is located near the central region of the SAMA, that is, a region with big asymmetries in the geomagnetic field as compared with the Northern Hemisphere. In addition, Santa Maria is in the boundary of low and middle latitudes, and then the thermospheric meridional neutral wind affects the ionosphere in this latitude thought transport processes of the electric charges controlling the h m F 2 (Balan et al., 1997;Liu et al., 2009). An upward (downward) plasma movement along the magnetic field lines is caused by an equatorward (poleward) wind (Batista et al., 1997;Romero-Hernandez et al., 2018;Shim et al., 2010). Therefore, changes in h m F 2 will produce variations in N m F 2 (due to changes in f o F 2 ) as observed in our results from December solstice to June solstice in Figure 3e. We also note an equinoctial asymmetry after~18:00 UT, with the ionization in March equinox being stronger than that in September equinox. Overall, we note from Figure 3 that the day-to-day variability is a permanent feature of the ionospheric E and F regions over Santa Maria.

Ionospheric Variabilities During Geomagnetically Quiet Days
In our attempt to study quantitatively the ionospheric variabilities and its main causes over Santa Maria, we may reduce the four possible sources described before to three broad categories: fluctuations of the solar ionizing radiation, geomagnetic activity, and by an internal process in the neutral atmosphere. Therefore, we discuss the relative contribution of these sources by using rough arguments based on observed solar emission and geomagnetic index data.
Regarding the first possible source of ionospheric variability, that is, the solar ionization radiation, we have an important advantage. The 1-year data used are characterized by a very low level of solar and magnetic activity. The average of the F 10.7 index from 1 September 2017 to 30 August 2018 is only 71.6 ± 3.5 × 10 −22 Wm −2 Hz −1 . With the 27-day averaged values of F 10.7 (sunspot number) in the range 67.6-80.1 (1-18) in the period, the ionospheric data used in this work represent low solar activity. The variabilities that may be caused by variations in the F 10.7 are largely excluded in this work.
In order to exclude the geomagnetic activity impacts on the ionospheric variabilities and to estimate the variabilities due to the neutral atmosphere source, we evaluated the CV using equation (1) considering data acquired during geomagnetically quiet days (Kp ≤ 3) only. The CV is calculated on an hourly basis, and the results are presented in Figure 4. The f o F 2 CV, f o E CV, h m F 2 CV, and B0 CV on a seasonal scale presented from top to bottom should be mainly due to the meteorological sources. The diurnal variations of CV are not straightforward due to specific characteristics in each parameter, time, and season. In order to summarize its general characteristics, we present in Table 1 the minimum, maximum, average (μ), and corresponding standard deviations (σ) of f o F 2 CV, f o E CV, h m F 2 CV, and B0 CV for each season.
The diurnal variations of f o F 2 CV values presented in Figure 4a show that CV is over 10% in the predawn hours, decreases down during daytime hours, and then begins to increase in the equinox months toward the dusk. In December and June solstices, the f o F 2 CV remains below 20% during nighttime hours. The minimum f o F 2 CV is 5.6 ± 2.26% observed at 9:00 UT in June solstice, as shown in Table 1. In order to determine if the value 5.6 ± 2.26% is statistically significant different from the other minimum f o F 2 CV values (7.0 ± 0.24%, 9.4 ± 0.95%, and 19.6 ± 0.65%), we performed the Kruskal-Wallis (Kruskal & Wallis, 1952)  Many studies are concerned with the f o F 2 CV, and most of them were conducted in the equatorial region and Northern Hemisphere. For instance, Akala et al. (2010) presented the diurnal, seasonal, and solar activity impacts on the f o F 2 CV over the Southeast Asian sector. The authors revealed that f o F 2 is more susceptible to variability during the nighttime than the daytime and is characterized by a post-sunset peak (15-50%) and a post-midnight peak (  Regarding their results during low solar activity only, which is the case here, the authors found that the CV is the lowest during daytime (5-15%) and increases during nighttime (15-30 %) for both stations. As the authors compared the data acquired at Korhogo (9.3°N, 5.4°W, dip angle = 0.67°) located in the equator and Ouagadougou (12°N, 1.8°W, dip angle = 15.9°) situated in the EIA crest, significant differences were observed in the f o F 2 variability, being greater in the former station than in the second. Zhang et al. (2004) found that the variability in f o F 2 is much lower during daytime hours than in nighttime hours, with the maximum variability occurring at presunrise hours at Hainan, China (19.4°N, 109.0°E, dip angle = 22.8°). We do not include the specific values here because Zhang et al. (2004) obtained the variability in terms of percentage interquartile values. The lower f o F 2 CV values during daytime hours than nighttime hours are attributed by the authors to the steep electron density gradients caused by the onset and turn-off of solar ionization and superimposition of ionospheric F region irregularities (spread F) on the background electron density. At night, when the photochemical control is weaker and the ionospheric electron density is dependent on the recombination rate, the electron density is more sensitive to the meteorological activity as meridional winds. We have no reason to disagree with the preview conclusions, which is consistent, for example, with the mass plot shown in Figure 3a.
Inspection of Figure 4b shows that the diurnal variation of f o E CV is characterized by a maximum at the sunrise. It decreases from the dawn values to a plateau of around 5 ± 1% between 12:00 UT and 22:00 UT in all seasons. Table 1 shows that f o E CV spans from 3.2 ± 0.25% at 14:00 UT in June solstice to 10.2 ± 1.79% at 10:00 UT in September equinox. March equinox is the most variable season with an average of 6.9 ± 1.3%, and June solstice is the least variable season with an average of 4.6 ± 1.4%. These results reveal that the f o E is the least variable parameter in our study, that is, the variability of the E region is less than that of the F region. The diurnal variation pattern of f o E CV observed is due to the predominant control of photoionization and loss process that take place below the F 2 peak. It is expected a balance between the formation and recombination processes during daytime, which sets the E region into a quasi-stationary equilibrium leading to small f o E CV. Both the daytime and nighttime seasonal patterns of the f o E CV values agree closely with the results reported by Abe et al. (2013) at Ouagadougou during low solar activity. The authors found that the magnitude of the CV ranges from 3.0% observed at local noon to 7.8% observed during sunrise hours, higher magnitude in June solstice and the lowest magnitude in December solstice. Comparing their results with ours, we see that the variability over Santa Maria is slightly higher in magnitude than Ouagadougou.
Since a permanent feature of the equatorial F region is the E × B dynamics caused by the equatorial electrojet and prereversal enhancement, we would expect larger day-to-day variability in Ouagadougou than in Santa Maria.
The diurnal variation pattern of h m F 2 CV and B0 CV presented in Figures 4c and 4d, respectively, is more complicated when compared with f o F 2 CV and f o E CV. We note in general a presunrise, local midday, and a post-sunset peak for March equinox, June solstice, and September equinox. The presunrise and post- 10.1029/2019JA026780

Journal of Geophysical Research: Space Physics
sunset peaks are not observed in December solstice for h m F 2 CV. The occurrence time of the presunrise peak is shifted to later (9:00 UT) for B0 CV. From Table 1, we see that h m F 2 CV ranges from 4.3 ± 0.28% at 5:00 UT in December solstice to 17.4 ± 1.23% at 14:00 UT in September equinox. The highest seasonal average of h m F 2 CV equal to 11.6 ± 2.8% is detected in September equinox, while the least seasonal average of 7.7 ± 2.2% is detected in December solstice. Our findings reveal that the variability of h m F 2 is much less than that observed in the f o F 2 parameter. This is in line with the observations of Tsagouri et al. (2018) over seven ionospheric locations in the Northern Hemisphere in the Year 2002. The authors found that on average, the CV is about 14% and 7% for f o F 2 and h m F 2 , respectively. There is no expressive difference between the variabilities detected here in h m F 2 and those reported by Altadill (2007) using ionosonde data acquired from January 1995 to December 2005 at the Ebro Observatory (40.8°N, 0.5°E). The author found that h m F 2 CV ranges from 10% to 20%, larger by night than by day with an average of 13.3%.
Concerning B0 CV shown in Table 1, the minimum observed value is 18.4 ± 1.82% at 3:00 UT in December solstice. We are not able to determine the season with the maximum variability since the 39.2 ± 4.08%, 42.1 ± 2.81%, 42.2 ± 4.92%, and 38.9 ± 3.25% values are not significantly different at the 0.05 level from the Kruskal-Wallis test, as mentioned before. The highest seasonal average of B0 CV is 34.2 ± 4.8% detected in June solstice, and the least is 25.5 ± 5.8% observed in December solstice. Therefore, we find that the most variable parameter in our work is B0. For the B0 CV obtained in Hainan, China, Zhang et al. (2004) found that it has a minimum value at hours after sunrise and sunset in winter and equinox seasons, while it has maximum values around port-midnight hours.

Ionospheric Variabilities During Geomagnetically Disturbed Days
In order to further investigate the geomagnetic activity impacts on the ionospheric variabilities in the central region of the SAMA, we evaluated the CV during geomagnetically disturbed days and compare it with the The disturbed CV calculated from data acquired during the geomagnetically disturbed days (7, 8, 15, 27, and 28) of September 2017 is compared with the CV obtained from the geomagnetically quiet days (9, 22, 23, 25, and 26) of the month. The variabilities evaluated for Santa Maria are compared with the CV obtained from ionospheric data acquired at the Wuhan Ionospheric Observatory for the same days. The estimated hourly CV is presented in the left panels of Figure 5 for Santa Maria and in the right panels for Wuhan. The resulted diurnal variation of CV considering the active days is shown by the red dots, while the CV averaged from parameters measured in quiet days is shown by black dots.
The enhancement of the variability in all parameters considering geomagnetically disturbed days is significant over Santa Maria when compared with the CV evaluated with geomagnetically quiet days. In average, f o F 2 CV is 20.9 ± 7.3% during disturbed conditions and 11.7 ± 4.1% during geomagnetic quiet days. Regarding the f o E CV, the variability increased to 5.0 ± 1.8% from 3.4 ± 1.0%. The h m F 2 CV is on average 11.5 ± 4% in disturbed conditions while only 6.5 ± 1.8% in quiet condition. With respect to the most variable parameter in our study, B0 CV achieved 29.0 ± 8.5% during geomagnetically disturbed days, which is higher than the 20.4 ± 4.1% obtained from geomagnetically quiet days. These results clearly indicate that during geomagnetically disturbed periods, the ionosphere over Santa Maria is subject of this important source of variability in addition to the meteorological source present in the geomagnetically quiet period.
Regarding the CV estimated over the Wuhan station, in general, CV also shows an enhancement in the disturbed days when compared with quiet days. In average, the f o F 2 CV is 13.8 ± 5.4% during disturbed conditions and 8.0 ± 2.3% during geomagnetic quiet days. The f o E CV increased to 8.7 ± 1.4% from 4.8 ± 3.3%. The h m F 2 CV is on average 7.7 ± 4.3% in disturbed conditions while 5.8 ± 2.3% in the quiet condition. We do not observe a large enhancement of B0 CV over Wuhan during the geomagnetically disturbed days. It increased to only 22.8 ± 12.2% as compared with 21.7 ± 10.9% during geomagnetically quiet days.

Journal of Geophysical Research: Space Physics
In order to confirm the same trend observed for the September 2017 geomagnetic storm, we also present another case study in Figure 6. The Dst index recorded in 26 August 2018 geomagnetic storm indicates a sharp drop to a minimum of −174 nT between 7 and 8 UT on 26 August. Following the same strategy of the first case study, we calculated the CV during the geomagnetically disturbed days (15, 17, 20, 26, and 27) of August 2018 and compare it with the CV obtained from the geomagnetically quiet days (6, 10, 13, 14, and 23) of the same month. The results of both cases studies are placed in Table 2. Repeatedly, we note that in general, the degree of variabilities also increases in all parameters in both stations, that is, Santa Maria and Wuhan. The only exception is for f o E CV at Wuhan, which presents higher value during the quiet period as compared with the disturbed period. However, this difference is not significant since it lies within the standard deviation of the mean values.
Another interesting point that comes from the analysis of Figures 5 and 6 and Table 2 is that the variabilities during the geomagnetic storms are higher over Santa Maria than Wuhan. It can be explained due to the presence of the SAMA in the former station. The increase in the ionization production (rising the f o F 2 ) over Santa Maria is explained by the trapped and azimuthally drifting of energetic particles coming deeper down into the dense neutral atmosphere through the low magnetic field intensity (conserving the second adiabatic invariant) while bouncing between hemispheres. Even though the September 2017 geomagnetic storm was less intense than the August 2018 as indicated by the Dst index, the higher variability detected in the ionospheric parameters during the former storm can be explain by the fact that it consisted of two consecutive magnetic storms separated in time by~13 hr, as well explained by Blagoveshchensky et al. (2019).
Some of the aforementioned works studied the influence of geomagnetic activity on the ionospheric variability. For example, Forbes et al. (2000) found that the f o F 2 CV increases with the geomagnetic activity, considering data of two solar cycles. However, Araujo-Pradere et al. (2005) argued that the variability tended to increase with geomagnetic activity at low and middle latitudes in June solstice and equinoxes but remained fairly constant in December solstice.
The geometry of the geomagnetic field in the SAMA region associated with the plasma transport along the field and perpendicular to it by winds makes the interpretation of the large amount of variability in the observed ionospheric parameters difficult. The meridional winds may transport the plasma in altitude and latitude, given the geometry of the geomagnetic field. The results of the CV provided in Figure 3 is attributed to the neutral atmosphere as the main source of ionospheric variability. The detailed explanation of the "meteorological" as a main source of the variability remains a challenge due to the lack of measurements in the south of Brazil, but the average values of the ionospheric parameters shown in this work are what models driven by climatological parameters should attempt to reproduce.

Conclusions
We have presented the first study of the diurnal, seasonal, and ionospheric variability in the f o F 2 , h m F 2 , B0, and f o E parameters over Santa Maria (29.7°S, 53.7°W, dip angle = −37°), a station located near the central region of the SAMA. The ionospheric parameters were recorded with a time resolution of 5 min from 1 values present opposite behavior, that is, the nighttime values are higher than those of the daytime values except for short periods around 9:00 UT in June solstice and around 14:30 UT in December solstice. It also observed a remarkable mean difference in all parameters between December and June solstices, revealing a possible presence of the annual asymmetry, and a lesser extent difference between March and September equinoxes. Comparing our results with the finds from other stations located in the EIA crests and outside the SAMA's control, we found that the effects of the EIA are more intense in the ionospheric parameters than the ones caused by the SAMA.
The CV is used as an index to quantitatively describe the ionospheric variability of each parameter. The whole year analyzed is characterized by low solar and geomagnetic activity, which is an important advantage to estimate the relative contribution of different sources under enough rigid restrictions on the threshold level of the geomagnetic activity. Considering the low F 10.7 and sunspot number during the period analyzed and excluding the geomagnetically disturbed data, the f o F 2 CV, f o E CV, h m F 2 CV, and B0 CV obtained should be mainly due to the meteorological sources. The diurnal analysis reveals that the ionosphere in the central region of the SAMA is more susceptible to variability during nighttime than daytime hours. The f o F 2 CV ranges from 5.6 ± 2.26% detected in June solstice to 49.4 ± 1.90% observed in September equinox. In average, f o F 2 is more variable during September equinox (31.3 ± 6.8%) and less variable in June solstice (13.4 ± 2.8%).
The f o E CV is characterized by a maximum at the sunrise, spanning from 3.2 ± 0.25% in June solstice to 10.2 ± 1.79% in September equinox. The highest variabilities of f o E occur in March equinox (6.9 ± 1.3%) and the lowest in June solstice (4. 6 ± 1.4%). These results reveal that the f o E is the least variable parameter in our study, that is, the variability of the E region is less than that of the F region. Regarding the h m F 2 CV, it ranges from 4.3 ± 0.28% in December solstice to 17.4 ± 1.23% in September equinox. The highest seasonal average of h m F 2 is detected in September equinox (11.6 ± 2.8%), while the least seasonal average is detected in December solstice (7.7 ± 2.2%). Our findings reveal that the variability of h m F 2 is much less than that observed in the f o F 2 parameter. Finally, the minimum B0 CV is 18.4 ± 1.82 % observed in December solstice. We are not able to determine the season with the highest B0 variability since the maximum B0 CV values are not significantly different among them. In average, the highest variability is detected in June solstice (34.2 ± 4.8%) and the least one in December solstice (25.5 ± 5.8%). Therefore, B0 is the most analyzed variable parameter in our work.
The geomagnetic activity impacts on the ionospheric variabilities in Santa Maria, Brazil, are evaluated during two geomagnetic storms, and the results are compared with Wuhan, China. The estimated CV is enhanced during geomagnetically disturbed days in both stations and reveals that the variability in the former stations is generally higher on average. This fact may be explained by the presence of the SAMA in the Brazilian lowlatitude station, which leads to an increase of the ionization production in the ionosphere. It, in turn, is caused by the trapped and azimuthally drifts of energetic particles that precipitate into the dense neutral atmosphere through the low magnetic field intensity (conserving the second adiabatic invariant) while bouncing between hemispheres. Therefore, our results clearly indicate that the ionosphere over Santa Maria is subject of this important source of variability during geomagnetically disturbed periods in addition to the meteorological source present in the geomagnetically quiet periods.