Primary Joint Statistical Seismic Influence on Ionospheric Parameters Recorded by the CSES and DEMETER Satellites

Oppositely to a previous statistical work using a single time resolution of the total ion density measured onboard the Detection of Electro‐Magnetic Emissions Transmitted from Earthquake Regions (DEMETER) satellite, this work deals with statistical seismo‐ionospheric influences by comparing different parameters and various time resolutions. The O+ density and electron density recorded by the China Seismo‐Electromagnetic Satellite (CSES) for more than one year and by the DEMETER satellite for about 6.5 years have been utilized to globally search ionospheric perturbations with different time resolutions. A comparison is automatically done by software between the occurrence of these ionospheric perturbations determined by different data sets, and the occurrence of earthquakes under the conditions that these perturbations occur at less than 1,500 km and up to 15 days before the earthquakes. Combined with statistical results given by both satellites, it is shown that the detection rate r of earthquakes increases as the data time resolution and the earthquake magnitude increase and as the focal depth decreases. On average, the number of perturbations is higher the day of the earthquake, and then smoothly decreases the days before, which is independent of either ionospheric parameters or time resolutions. The number of right alarms is high near the South Atlantic Magnetic Anomaly area but its relationship with seismic activities is weak. The ion density tends to be more sensitive to seismic activities than the electron density but this needs further investigations. This study shows that the CSES could effectively register ionospheric perturbations due to strong EQs as the DEMETER satellite does.


Introduction
With the development of Earth observation satellites, their onboard experiments have gradually shown their potential application in the field of earthquake (EQ) monitoring and investigation. This is due to their advantages of fast-speed, large-scale, and high-resolution results, especially for areas with harsh natural conditions. On one hand, scientific data from satellites have been utilized to distinguish precursors prior to strong EQs. As one result of researches during the last ten years, it has been shown that the ionosphere is unexpectedly sensitive to the seismic activity (Hayakawa & Molchanov, 2002). On the other hand, these satellite data have also been combined with ground-based observations to study the lithosphereatmosphere-ionosphere coupling (Hayakawa & Molchanov, 2002;Li et al., 2016Li et al., , 2019Molchanov et al., 2004;Molchanov & Hayakawa, 2008;Pulinets et al., 1994Pulinets et al., , 1997Pulinets et al., , 2000Pulinets & Ouzounov, 2011;Sorokin et al., 2015, and references therein).
There has been many statistical works on ionospheric variations associated with strong seismic events using satellite measurements. Seismo-ionospheric disturbances within a few days before EQs have been registered in 73% of EQs with magnitude 5.0, and in 100% of EQs with magnitude 6.0 (Pulinets et al., 2003). Liu, Chen, et al. (2009) have performed a statistical analysis on GPS TEC and found that seismo-ionospheric variations above the epicentral area occurred 5 days prior to 16 EQs out of 20 events with M ≥ 6.0 in the Taiwan area from September 1999 to December 2002. Statistical analyses have also been performed using DEMETER data sets. A statistically significant decrease of wave intensity at 1.7 kHz during nighttime four hours before the occurrence of EQs has been reported respectively by Němec et al. (2008Němec et al. ( , 2009) and by Píša et al. (2012Píša et al. ( , 2013. Statistical analyses have been performed using a total ion density (the sum of H + , He + , and O + ) data set during the DEMETER lifetime (6.5 years) in the epicenter areas of earthquakes as well as in their magnetically conjugate point areas and the results have shown a significant statistical correlation between ionospheric anomalies and large events within a few days before the events (Li & Parrot, 2012Parrot, 2011Parrot, , 2012Parrot & Li, 2017;Yan et al., 2017). Zhang et al. (2013) have found that there are increases in the number of electron bursts prior to strong EQs with a magnitude over 7.0 during the entire operation period of the DEMETER satellite. However, Akhoondzadeh et al. (2010) statistically analyzed the ionospheric variations prior to four large EQs simultaneously using, on one hand, several satellite parameters of DEMETER: IAP ion density and ion temperature, ISL electron density and electron temperature, and on the other hand, GPS TEC. Their results show that there is a very good agreement between the different parameters and that, most positive and negative anomalies appeared 1 to 5 days before all studied EQs during quiet geomagnetic conditions, although their amplitude depends on the magnitude of the EQs involved.
The CSES (China Seismo-Electromagnetic Satellite) has been launched for more than one year and effective payload data have been recorded. It is possible to check its response to seismic activities during this period. So, in this paper, a primary statistical analysis on seismo-ionospheric influence of different parameters of ion density and electron density recorded by CSES and DEMETER satellites will be comparatively shown. The CSES and DEMETER satellites are briefly described in section 2. In section 3, the data processing method is retrospectively introduced. In section 4, automatic statistical results for different parameters recorded by CSES will be compared with that of DEMETER and confirmed. Discussion and conclusions are provided in section 5.

The CSES and the DEMETER Satellite
The CSES was launched successfully on 2 February 2018. The CSES is a Sun-synchronous satellite orbiting at a height of 500 km with a circular orbit and the descending node happens at 14:00 local time (LT). There are eight scientific payloads onboard, including a search-coil magnetometer, an electric field detector, a high precision magnetometer, a plasma analyzer package, a Langmuir probe, an energetic particle detector, a GNSS occupation receiver, and a three-frequency beacon. Of these, the Langmuir probe (LAP) and plasma analyzer package (PAP) are the space plasma in situ detection payloads. Their scientific design parameters can be referred to Shen et al., 2018;Yan et al., 2018). The LAP allows access to the electron density and temperature. The operational modes of the LAP include survey mode and burst mode. The survey mode is used mainly to detect global electron density and electron temperature with sweeping period of 3 s (second), while the burst mode primarily allows detection of key areas, over China and within global main seismic belts with sweeping period of 1.5 s. The PAP measures ion density, composition, temperature, and flow velocity. PAP has also the same operational modes as that of LAP but with a little higher resolution of 1 s for survey mode and 0.5 s for burst. DEMETER was launched in June 2004 onto a polar and circular orbit which measures electromagnetic waves and plasma parameters all around the globe except in the auroral zones (Parrot et al., 2006). DEMETER is a low-altitude satellite with an altitude of 710 km, which was decreased to 660 km in December 2005. The orbit of DEMETER is nearly sun-synchronous and the upgoing half-orbits correspond to night time (22.30 LT) whereas the downgoing half-orbits correspond to day time (10.30 LT). The onboard experiments of DEMETER includes six scientific payloads, the IAP-plasma analyzer instrument, the ICEelectric field instrument, the ISL-Langmuir probe, the IMSC-search-coil magnetometer instrument, the IDP-particle detector instrument, and the BANT-an electronic unit. The variations of the ion density and the electron density are measured by the instrument IAP and ISL, respectively. These two experiments have different operational modes. The ISL is with an experimental resolution of 1 s for all data, while the IAP has two experimental data resolutions of 4 s in survey mode and 2 s in burst mode. Details of the IAP and ISL experiments can be found in Berthelier et al. (2006) and Lebreton et al. (2006). The satellite's science mission has come to an end in December 2010.

Data
The O + is the main ion among the ions H + , He + , and O + detected by the satellites although it depends on some factors, such as, local time, altitude, and so on. So, the used data of CSES include PAP O + density and LAP electron density from 1 August 2018 to 30 November 2019. During this period, 4,317 strong EQs with magnitudes M W equal to or more than 4.8 occurred (USGS: http://www.usgs.gov).
The DEMETER data used here include the completed data for the parameter IAP O + density and ISL electron density covering its lifetime from the mid-2004 to the end 2010. During this period (6.5 years in total) there are 21,863 strong EQs with magnitude M W ≥ 4.8 which took place (USGS: http://www.usgs.gov).
The Kp index (http://isgi.unistra.fr) is also checked in order to avoid the effect from the solar activities during all the periods considered in this paper.

Data Processing Method
The data processing method here is similar to the one used before by Parrot (2012, 2013). First of all, a software is used to automatically search global ionospheric perturbations in several data sets (issued from CSES or DEMETER). Only the perturbations which comply with the duration time between 20 and 120 s are kept in the perturbation database but without a limit on perturbation amplitudes. Here, the minimum duration time 20 s is calculated automatically by the improved software according to data sample rate instead of previously used~23 s, five data points for survey mode in IAP DEMETER (Li & Parrot, 2012. The information for each perturbation in the perturbation database includes peak appearing time, orbit number, location (latitude and longitude), background value, amplitude, change trend (increase or decrease; if the amplitude is larger than the background value, it is increase, if not, it is decrease), increase or decrease percentage, duration time, and extension distance (km).
In order to examine the capability of recording seismic influence on the ionosphere with different data time resolutions using our software, the raw data are sampled at different time resolutions: 1 and 3 s for PAP O + density and 3 s for LAP electron density of CSES; 4 s for IAP O + density, and 3 and 4 s for ISL electron density of DEMETER. Thus, six data sets have been established: PAP-1 s, PAP-3 s, and LAP-3 s for CSES; IAP-4 s, ISL-3 s, and ISL-4 s for DEMETER. At the same time, the SAVGOL method is employed to smooth the data before searching for perturbations. The SAVGOL function returns the coefficients of a Savitzky-Golay smoothing filter (Savitzky & Golay, 1964). So, at this stage, two O + perturbation databases (1 and 3 s data are used respectively, including 48,529 and 24,438 perturbations in total) and one electron database After, a second software is used to check whether the ionospheric perturbations are corresponding to an EQ or not under the following three limits, (i) the Kp index is kept to be less than 3 in order to reduce the effect of the geomagnetic activity on the ionosphere. This geomagnetic activity is induced by solar magnetic storms, (ii) the distance (D) between the location of the perturbation on the orbit and the epicenter is equal to or less than 1,500 km, and (iii) the delay time (T) before an EQ is equal to or less than 15 days. If an earthquake is corresponding to one or to more than one perturbation, we consider it is a good detection; if not, it is a bad detection. If a perturbation corresponds to an earthquake, it is a right alarm; if not, it is a false alarm. More details can also be found in Parrot and Li (2017) and Li and Parrot (2018).
An example of ionospheric perturbation detected by the software and corresponding to an EQ is shown in Figure 1. It corresponds to an EQ occurring on 21 August 2018 at 21:31:47 UT with a magnitude equal to M W 7.3 and a depth equal to 147 km. Its position was 10.77°N, 62.90°W. Figure 1a shows variations of PAP parameters. From the top to the bottom, the panels show the densities of the H + , He + , and O + ion. The X axis represents UTC (Universal Time Coordinated)/BJT (Beijing time), latitude, longitude, and altitude of CSES. At 6:00 UT, about 15 h before this EQ, the orbit 03040 flew up this epicentral area and an increase of O + labeled by a black arrow in the bottom panel with red curve is observed in Figure 1a (this variation seems not obvious at this period because it is formed automatically by the CSES data processing system). Correspondingly, the payload LAP also recorded clear increase of electron density on the same orbit as it is shown in the bottom panel with a black arrow in Figure 1b.
These two ionospheric perturbations of PAP and LAP have been successfully detected by the software and their information is shown in Table 1. From Table 1, one can see that these two ionospheric perturbations

EQ Detection Rate
Here, a parameter r is defined as the EQ detection rate, which is the ratio of the number of EQs detected corresponding to one or more than one ionospheric perturbations and the number of EQs which comply with the limit conditions.
EQs are still divided into different groups during this work in order to gain an easy comparison as properties of EQs vary. The 4,317 M W ≥ 4.8 EQs occurring between August 2018 and November 2019 as CSES flies have been divided into three groups according to their magnitudes: 4.8 ≤ M W ≤ 5.0 2524 EQs, 5.0 < M W ≤ 6.0 1624 EQs, and M W > 6.0 169 EQs. Then, in light of previous statistical seismo-ionospheric influences for EQs located in different areas of the world (see, e.g., the effect of the South Atlantic Magnetic Anomaly (SAMA) on EQ detection reported by Parrot (2012, 2013)), two specific rectangular zones have been selected (see Figure 6 in section 5), (i) a Zone1 with latmin = −70°, latmax = −45°, longmax = 150°W, longmin = 20°W, which lies in the SAMA area in south hemisphere and includes 190 EQs, and (ii) a Zone2 with latmin = 0°, latmax = 30°, longmin = 90°E, longmax = 150°E, which lies in low-middle latitude in north hemisphere and includes 774 EQs. For each group of EQs, they are detected under the conditions of D = 1,500 km, T = 15 days, and the focal depth d = 0-1,000 km with different ionospheric databases confirmed by data sets of PAP-1 s and PAP-3 s data and LAP-3 s from CSES. To check effects of the focal depth (d) of earthquakes on the detection rate, "crust" earthquakes with d ≤ 20 km have been comparatively detected for each group of EQs. The Kp index is kept to be less than 3 during this period in order to eliminate the effect from solar activities. Their corresponding detection rates are listed in Table 2.
From Table 2, except M > 6.0 with contrary result due to its less examples (169 in total and 77 with the depth d ≤ 20 km), the detection rate r for "crust" EQs with d = 0-20 km is thoroughly higher than that of EQs with d = 0-1,000 km for each group of EQs in all data sets. These results tend to verify a definite conclusion that ionospheric influence can be affected by the epicenter depth: The "crust" EQs can be easily detected than "deep" ones, which has been already reported by Li and Parrot (2012) and Silina et al. (2001).
For a better comparison, the data in Table 2 with d = 0-1,000 km have been shown in Figure 2 under the form of histograms.
From Table 2 and Figure 2, it is clear that the detection rate r increases as the magnitude of EQs increases. Additionally, for the same parameter PAP O + density with different data time resolutions of 1 and 3 s, the detection rates of PAP-1 s for six group EQs of 4.8 ≤ M W ≤ 5.0, 5.0 < M W ≤ 6.0, M W > 6.0, Zone1, and Zone2 are higher than that of PAP-3 s, which means that the detection rate r increases as the time resolution of data increases; for different parameters of PAP O + density and LAP electron density with the same time   Table 3.
For a better comparison, the data of Table 3 is shown in Figure 3 under the form of histograms.
From Table 3 and Figure 3, one thing we can reconfirm is that the detection rate r increases as the time resolution of the same ionospheric parameter ISL electron density increases.
Another point is that the detection rates for IAP density are clearly higher than that of ISL density for each group of EQs when the time resolution is the same, which probably gives a conclusion that the ion density is more sensitive to seismic activities than the electron density although this claim is not completely verified by the CSES data.

Temporal Evolution of Seismo-ionospheric Influences
To check the occurrence frequency of ionospheric perturbations during EQ preparation, a study has been conducted on the different cases, and the corresponding results are shown in Figure 4. Figure 4 displays the number of detected perturbations corresponding to an EQ (right alarms) as a function of days before the good detections considering the following data sets recorded by CSES: All EQs (4,317 EQs), 4.8 ≤ M W ≤ 5.0 EQs, Zone1, and Zone2. In each panel of Figure 4 the results are expressed as a percentage relative to the total number of right alarms within 15 days detected by PAP-1 s (red line), PAP-3 s (blue line), and LAP-3 s (orange line) data sets. In a similar way, Figure 5 is related to DEMETER and displays the relative  It can be seen that, for all cases in Figures 4 and 5, the number of perturbations is maximum for days close to the EQ day and smoothly decreases when the time before the EQ is increasing. This is a variation that is intuitively expected. On one hand, this variation seems not mainly affected by the data time resolution of a given parameter whatever this parameter is. On the other hand, this trend becomes more obvious when the number of samples is large enough. Thus, three lines decay smoothly as day goes for All EQs (see the top left panel in Figures 4 and 5) but with a few fluctuations for other cases (see other panels also in Figures 4 and 5). However, the percentages increase more obviously in Zone2 (see the bottom right panel in Figures 4 and 5) than in Zone1 especially one week before the EQs. This means that the perturbations have a little relationship with seismic activity in the Zone1 area. In this case, false alarms become more important due to outer disturbances of the active E × B drift occurring in the South Atlantic Magnetic Anomaly (Abdu & Batista, 1977;Abdu et al., 2003Abdu et al., , 2005.

Discussion and Conclusions
As CSES has run more than 1 year since it was launched on 2 February 2018 in China, this increases the probability to examine the effectiveness of seismo-ionospheric influences recorded by the scientific payloads of PAP O + density and LAP electron density. For comparison, two corresponding DEMETER parameters of IAP and ISL for about 6.5 years have also been checked to gain some similar results.
Oppositely to previous statistical works with DEMETER using a single time resolution of one parameter of IAP total ion density (Li & Parrot, 2012, this work is associated with statistical seismo-ionospheric influences using two parameters recorded by the DEMETER satellite and CSES, respectively. At the same time, various time resolutions for a given parameter are also employed to survey the effects of seismic activities on the ionosphere and also to check the efficiency of our two software. Numerous investigations have shown that ionospheric variations generally appeared several days to two weeks prior to EQs. In this work, ionospheric perturbations are automatically searched with a delay time before an earthquake equal to or less than 15 days. The statistical results have displayed that totally the occurrence rate of these perturbations is the highest at the day of the earthquake and then gradually decreases at the day before. While this situation shows a different art of styles for groups of EQs due to regional influences, such as SAMA in Zone1. Relatively, Zone2 displays a more reliable result with averaged occurrence rates within one week (7 days) being 79.6% for all data sets. This result is highly coincident with some investigations that ionospheric variations mainly occurred several days prior to EQs (Akhoondzadeh et al., 2010;Liu, Chen, et al., 2009) and with the statistical work conducted by Li et al. (2019), who found that ionospheric variations mainly happened 6 days and gained a high climax 3 days before the Wenchuan main event on 12 May 2008 in the light of different authors (can be referred to Figure 4 in Li et al., 2019).
However, the abnormal range in the ionosphere arise from seismic activities has not been well established so far. The software in this paper is designed to only accept ionospheric perturbations with a duration of 20-120 s (about 160-840 km if the speed of the satellite 7.0 km/s is considered) and positive results have been given. In order to further examine this design is effective for the investigations, variations with a duration of 200-300 s using PAP-1s data set from August 1 2018 to November 30 2019 have been accepted this time and 509 ionospheric perturbations have been attained in total. These perturbations are located in the map of the world (see Figure 6 with purple dots). As a comparison, a global ionospheric plasma variation has been given by O + ion density recorded by CSES during this period and the main seismic zones in the world are basically determined by 4,317 EQs occurred during this period, which cover especially plate-boundary interfaces, Circum-Pacific seismic belt, and Chile seismic zone (black dots in Figure 6). Zone1 and Zone2 employed above in section 4 are also added to Figure 6 by two black empty rectangles.
From Figure 6, it is clear that these perturbations collect mainly around the equator. This distribution is not coincident with the main seismic belts of the world, but keeps the similar shape as the background distribution of ion density. These large-scale ionospheric dynamical variations near the equator have been formed arise from some complex origins, such as Equatorial plasma bubbles, which are extremely dynamical phenomena with density drop out more than an order of magnitude over distances of a few kilometers perpendicular to the magnetic field leading to large-scale ionospheric variations . At higher latitudes, more intense waves are characterized by auroral emissions duo to strong sources of ELF/VLF emissions (Lefeuvre et al., 1992).
In the equatorial and low midlatitude ionospheric regions, the distribution of plasma is controlled by the coupled processes of plasma diffusion, E × B drifts, thermospheric neutral winds and chemical processes (Horvath & Lovell, 2009). The daytime (nighttime) F region plasma is transported by a vertical upward (downward) E × B drift, created by interaction between the ionospheric E field and the geomagnetic B field, over the dip equator, and by field-aligned diffusions on both sides of the dip equator (Balan & Bailey, 1995;Balan et al., 1997;Hairston et al., 1997). These processes have a tendency to create a plasma distribution symmetric to the dip equator. However, this tendency is interrupted by the meridional and transequatorial neutral winds, which move the plasma along the magnetic field lines and produce hemispheric and interhemispheric plasma flows, respectively, and by the accompanying chemical processes Kil et al., 2006;Titheridge, 1995).
The results show that the detection rate r could be affected by the data time resolutions because high time resolution data can record small scale ionospheric variations. Thus, the detection rate tends to increase as the time resolution increases. For the same time resolution, the detection rate r determined by IAP O + density for all cases are always higher than that of ISL electron density on DEMETER (see Table 3 and Figure 3), but these results gained partly on CSES (see Table 2 and Figure 2) probably owing to a less number of samples. However, there are still a high false alarm number and a high bad detection number due to the fact that a single satellite cannot continuously survey a given area, and then the natural disturbances of the ionosphere. Thus, it seems that EQs in SAMA zone have a higher detection rate but the right alarms have a weak relationship to seismic activities (see bottom left panels in Figures 4 and 5).
However, over the South Atlantic, the total B field intensity is anomalously low~22.8×10 3 nT from Trivedi et al. (2005), a phenomenon known as the South Atlantic Magnetic Anomaly (SAMA), that makes the E × B drift unusually strong, since its magnitude is E × B/B 2 (Kendall & Pickering, 1967). Furthermore, there are special electrodynamic effects in the SAMA region that can further increase the magnitude of the E × B drift by increasing the E field. Because of these plasma dynamics, the plasma density is highly variable over the SAMA and can be anomalously low (Abdu et al., 2005). Due to energetic particle precipitations, there is an enhanced E layer ionization over the SAMA (Abdu & Batista, 1977). The E layer conductivity is a maximum, where the magnetic field is a minimum, at the center of the SAMA (310°E, 10°S in geographic coordinates), over south Brazil, and decreases with increasing distance away from that center, toward the African continent. This can result in a westward conductivity gradient over the SAMA (indicated as DS by Abdu et al., 2003) that can add to the background conductivity gradient, which is also westward directed during the postsunset hours. Thus, this can create a locally high (or modified) conductivity distribution that is a regular feature of the ionosphere over the SAMA (Abdu et al., 2005). According to their model simulations, this increased conductivity will create a significantly stronger vertical E × B at the prereversal enhancement over Brazil (east coast of South America) than over Jicamarca (west coast of South America). This combined to the fact that Zone1 is the seismic zone which is at the highest geomagnetic latitude make the number of ionospheric perturbations higher than in another seismic region.
Overall, the results given by CSES and DEMETER have shown the following: 1. The CSES ionospheric data can effectively respond to strong EQs. 2. The detection rate r increases as the time resolution of the satellite data, and the magnitude of EQs involved increase and decreases as the epicentral depth of seismic events increases. 3. On average, the occurring frequency of perturbations is higher the day of the EQ and then gradually decreases before the event. This is independent of the data sets we use, either the ion density or the electron density, and whatever are their time resolutions. 4. The ion density seems to be more sensitive to the seismic activities than the electron density, but it needs further investigations with more data.
However, there still leaves a large of number of false alarms although an earthquake is searched around 1,500 km. Taken PAP-1 s as example, the total number of perturbations detected is 48,529; perturbations due to the solar activity with Kp ≥ 3 stands for 14.5% (7,020/48,529) and ones corresponding to earthquakes for 35.3% (17,112/48,529), which indicates that the number of false alarm stands for 50.2% (24,397/48,529). So the natural ionospheric variations are so important that we have to consider eliminating them in the future work. Additionally, the CSES and DEMETER satellite are at different height orbits. Different heights correspond to different satellite velocity, whose influence on the detection of earthquakes and the duration time of perturbations will also be considered.

Data Availability Statement
The data utilized in this paper are available through the website (http://www.leos.ac.cn/).