Pre-Earthquake Ionospheric Anomalies Associated with the 2015 Gorkha Earthquake of Nepal Detected by GPS-TEC Measurements

The present study analyses the variations in the ionospheric total electron content (TEC) prior to and during the 2015 Gorkha Earthquake in Nepal (Mw = 7.8) on 25 April 2015, utilising data from the widely distributed Global Positioning System (GPS) network. This study aimed to determine the association between ionospheric TEC anomalies and the occurrence of earthquakes. The finding shows that anomalous TEC changes occurred several days to a few hours prior to the major impending events. The results reveal that deviations in vertical total electron content (VTEC) at distant locations from the epicentre are less than those observed at the epicentre, implying that variation in ionospheric VTEC is nearly inversely proportional to the distance of GPS stations from the epicentre. In view of the solar - terrestrial environment, the pre - earthquake ionospheric anomalies could be associated with the 2015 Gorkha Earthquake. The VTEC anomaly was identified when it crosses the upper bound (UB) or lower bound (LB). The outcomes additionally show that TEC variation was dominant in the vicinity of the earthquake epicentre. We also observed contrast in TEC throughout the globe using global ionospheric maps at regular 2 - hour UT intervals, the day before, during and after the earthquake. As a result, we observed that areas heavily influenced by TEC were found to be transposed from eastern sectors to western sectors through the equatorial plane. TEC Maps indicate that most of the Indian regions, Northern China, Nepal, Bhutan, were heavily affected, indicating the earthquake's onset influence on the day of the event. Furthermore, we examined the cross - correlation of the SGOC station's TEC with the rest of the stations and discovered that the correlation increased gradually with epicentral distance from the surrounding stations, which was an intriguing result.


Introduction
The ionosphere is a complex layer in the upper atmosphere that undergoes spatial and temporal changes, mainly due to geomagnetic activity and ionisation from solar extreme ultraviolet (EUV) radiation. Usually, all energy input for ionosphere variations is predominantly attributed to solar and geomagnetic influences, while other influences have been essentially ignored (Pulinets et al., 1998).
However, recent studies have shown the importance to study ionospheric disturbances for earthquake monitoring (Pulinets, 2007;Sharma et al., 2017a;Sha et al., 2020b), and significant studies have focused on ionospheric perturbations due to earthquakes (Liu et al., 2000;Pulinets, 2002;Pulinets et al., 2005;Sharma et al., 2017a;Sha et al., 2020aSha et al., , 2020b. Major earthquakes are destructive and unforeseen natural phenomena where the unexpected fault rupture results in a sudden release of energy initially stored in the Earth's crust (Ohnaka, 2013). Numerous attempts have been made to predict earthquakes, and numerous measurements have been tested, such as those from Ionosonde, seismicity (Wang et al. 2010), radon and hydrogen gases (Planini et al., 2004) and energetic electron precipitation (Kudela et al., 1992). Fortunately, variations in the ionosphere have been suggested being the key parameter to detect pre-earthquakes. For instance, Jiang (2017) investigated the ionospheric total electron content (TEC) before the 2014 Mw 8.2 Chile earthquake and observed positive anomalies 4 days before the main event. TEC is the total number of electrons present along a path between two points, with units of electrons per square meter, where 10 16 electrons/m 2 =1 TEC unit (TECU) (Klobuchar, 1991;Adhikari et al., 2019;Silwal et al., 2021). Based on TEC observations, Global Positioning System (GPS) stations have become an important data source to study ionospheric TEC before major earthquakes. In addition, other data sources to detect TEC perturbations include, e.g., COSMOS Satellite measurements, DEMETER Satellite measurements (Pulinets, 2002;Haase, 2011;Pisa et al., 2012). The lithosphere-atmosphere-ionosphere coupling mechanism can eventually regulate the connection between ionospheric TEC anomalies and earthquake occurrences . The mechanism is thought to be reliant on changes in global electricity caused by ions accumulated during the development of stress in the crustal area prior to earthquakes Friedemann and Kulahci, 2009;Sharma et al., 2017a). Pullinets (2004) suggested a detailed description of the process via ion cluster formation in the near-ground layer around the earthquake preparation zone. Although several studies have reported the TEC anomalies from 0 to 8 days before major seismic events (Pulinets, 2009;Liu et al., 2011;Sharma et al., 2017a;Sha et al., 2020a, 2020b, Parashar et al., 2018, TEC disturbances due to earthquakes are difficult to detect, primarily due to lack adequate observations and other drivers of TEC variations such as geomagnetic storms (Pullinets, 2007(Pullinets, , 2009  . The main shock was followed by two large aftershocks of Mw 6.7 (on 26 April 2015 at 09:10 UTC) and Mw 7.3 (on 12 May 2015 at 07:05 UTC). This paper reports the TEC disturbances observed by GPS stations near and relatively distant from the epicentre during the April 2015 Gorkha earthquake.

Data and Method
The GPS-derived TEC data of nine stations (see Table 1) used in this work was obtained from UNAVCO, which is freely available on the website of UNAVCO (https://www.unavco.org/dai). The available data was in the RINEX (Receiver Independent Exchange) format v2.1, a standard ASCII format, which was further processed using the calibration technique of Ciraolo et al. (2007). The calibration technique of Ciraolo et al. (2007) is based on a thin shell model. The software converted the RINEX files into ASCII files, taking the TEC bias error calculations into account, eventually giving the values of VTEC throughout a day in 30-second intervals in an ASCII format. This approach assumes that the ionosphere is concentrated within a thin shell located at around 357-457 km of altitude (Oikonomou et al., 2016). First, the slant range TEC (STEC) was obtained from the RINEX dual-frequency carrier-phase observations L1 and L2 (Chen et al., 2017; In this equation, R is the Earth radius; α is the elevation angle of the satellite; and H is the height of the ionosphere, which has been taken as an altitude of 400 km. The original GPS data from SGOC, HYDE and IISC stations are sampled at a 30s interval, and that from the other six stations (CHLM, NPGJ, KKNI, DNSG, JMSM, DNDG) at a 15s interval. We resampled all the data to 1h resolution through a mean average running filter for this study. In order to identify the seismo-ionospheric disturbances, we employ the upper and lower bounds (Liu et al., 2004;Sharma et al., 2017a) from the median (µ) and standard deviation (σ) of a 15-day running-filter: Upper Bound (UB) = µ+1.34 σ Lower Bound (LB) = µ-1.34 σ (4) Here, positive and negative anomalies are considered when the VTEC values exceed the upper and lower bounds, respectively (Liu et al., 2009;Yao and Chen et al., 2012). The abnormal signal is considered at a confidence interval of about 82% (Kotz and Johnson, 1982;Oikonomou et al., 2016). We identify the possible effects by analysing the disturbance storm time Dst and the planetary Ap indices to exclude perturbations due to geomagnetic activity. Then, we study the crosscorrelation between the different stations in the time domain (Usoro, 2015;Adhikari et al., 2017Adhikari et al., , 2018Silwal et al., 2021). Cross-correlation is the standard method to estimate the degree to which two different series are correlated. This technique compares and evaluates the information between two time series of the included parameters as a function of a time lag (Finch & Lockwood, 2007;Mannucci et al., 2008). Figure 1 shows a network of the GPS receiver stations utilised in this study, and Table 1 provides each GPS station's location and the distance to the earthquake epicentre. Table 2 provides the reference stations' information to study the similarity pattern between the distant station (outside Nepal) through the cross-correlation technique.   , and negative anomalies on the days 11 th and 29 th of April at the CHLM GPS station (located at 57.1 km distance from the epicentre). The low TEC value observed was attributed to the reversal of electric field direction as postulated by Pulinets (1998) and Sharma et al. (2017a). In order to check the consistency of TEC anomalies, similar anomalies were expected to be observed from other stations. The analysis shows identical patterns in TEC anomalies from all six observations stations. Additionally, at other observatories GPS stations, the investigation revealed high TEC anomalies, crossing UB on the same anomaly days that we observed at CHLM station except for 29

Results and Discussion
April, including 4 April 2015 prior to the main event at DNGD located at a distance of 409.7 km from the epicentre. Observations at KKNI and NPGJ GPS stations located at a distance of 71.9 km and 307.61 km, respectively, indicate a high TEC anomaly on 5 th and 23 rd April before the earthquake. TEC derived from DNSG and JMSM located at a distance of 95.6 km and 115.8 km, respectively, showed similar anomalies like other stations. In addition to the mainshock on 25 April 2015, it was followed by big aftershocks on the same day as well as on 26 April with a magnitude of 6.7 at around 12:54:08 NST (07:08 UTC), with an epicentre located about 17 km south of Kodari, Nepal (USGS earthquake catalogue). These anomalies were also checked with other geomagnetic indices (Dst, Ap) and solar flux, F10.7, to detect the effect of a geomagnetic storm, as shown in Figure 3. It was observed that these anomalies were not affected by any geomagnetic phenomenon except for 16 th and 17 th of April. The change in TEC depends on solar activity, geomagnetic storms and receiving GPS station location and also varies with time and space. The energy and currents released by a geomagnetic storm increase the total height-integrated number of ionospheric TEC (Sharma et al., 2017a). An increase in ionospheric TEC results in the spatial variability of the ionosphere and cause ionospheric delays in the GPS signals (Pulinets, 2009).
Overall, the TEC values corresponding to the anomaly time from all six stations was observed that the earthquakes had at least one high TEC value crossing UBL and low TEC value crossing LBL, which was also observed in earlier studies (Sharma et al., 2017a(Sharma et al., , 2019(Sharma et al., , 2020.       Figure 4 describes a high deviation in TEC right after the mainshock of earthquake event, which was observed at 6:11 UT (~11:56 LT), and the almost constant increment remained for about 7 hours after the mainshock, i.e., up to 14:00 UT. As observed from the DNGD GPS station, the maximum VTEC difference between the day of the earthquake and the quiet day reached 17 TECU whereas 25 TECU from DNSG and JMSM stations, around 32 TECU while observed from KKNI, 30 TECU from CHLM and 22 TECU from NPGJ GPS stations. The high deviation of VTEC on event day from the quiet day was observed during the time interval of 6:00 UT to 14:00 UT. However, there was a relatively small deviation <10% before the mainshock and after 7 hours of the mainshock, i.e., after 14:00 UT in all stations, as witnessed from Figure 4. It is noteworthy that, throughout the earthquake event day, the positive deviation of VTEC was recorded.
It was also noticed that the deviation decreased with an increase in distance from the epicentre. The DNGD station, which is ~409 km from the epicentre, showed a maximum deviation of only 17 TECU, whereas a deviation up to 32 TECU was spotted from the KKNI GPS station, which is only 71.9 km away from the epicentre. TEC variation is found to increase as the epicentre distance decreases, which was also supported in earlier studies (Sharma et al., 2017a). The deviation of VTEC from quiet days fairly relates to the distance (see table 1) of the station from the earthquake's epicentre. Since earthquake event day was also the geomagnetically quiet day, increment in VTEC during and after the earthquake event compared to other quiet days disclosed a momentous relation between seismic activity like earthquake and the ionospheric TEC.    In Figure 6, no noticeable lag has been found between stations nearby the epicentre. On earthquake event day, the GPS TEC response to the seismic activity is relatively rapid, as reflected by a zerotime lag value. All the stations affected by earthquake show a similar change in the global electric circuit, produced by gathering ions in the atmosphere originating from the stress in the crustal region prior to an earthquake Friedemann and Kulahci, 2009). On the other hand, the decreased cross-correlation coefficient with a low time lag value between the stations, which are not affected by the earthquake, accounts for the fewer perturbations on the ionospheric TEC during the event day. This indicates that the findings obtained are encouraging because, in future days, the space weather associated with the earthquake's impact can easily be predicted by precise measurements of lead or lag in the response of the GPS TEC linked with seismic events.

Conclusions
We have attempted to investigate TEC variations during the 2015 Gorkha earthquake, observed by widely distributed GPS receivers in Nepal. Following conclusions are made from the present work: 1) The 2015 Gorkha Earthquake with magnitude 7.8 was preceded by TEC anomalies as observed at a station closer to the epicentre (57 km) and three GPS stations at a distance of (1000-2570 km) from the epicentre. The most apparent anomaly was observed on 14 th and 24 th April 2015 (one day before mainshock), although the earliest abnormality was observed on 11 April 2015, which was also observed in earlier studies (Sharma et al., 2017a(Sharma et al., , 2019(Sharma et al., , 2020. 2) By analysing the data of geomagnetic indices for the same observation period, we observed no geomagnetic storm before and after the earthquake. Thus, we can conclude that the TEC variations as observed were attributed due to seismogenic causes.
3) The analysis of deviation in VTEC on earthquake event day (also a geomagnetically quiet day) from the mean VTEC of the top four quietest days showed that there was an abrupt rise in the value of GPS VTEC after the earthquake event relative to that of the quiet geomagnetic days and the high deviation persisted approximately 7 hours after the earthquake occurred.

4)
The less deviation of VTEC observed on the GPS station far from the epicentre than that of stations nearby suggests a high increment in VTEC near the epicentre, and variation in ionospheric VTEC almost inversely depends upon the distance of GPS stations from the epicentre. As the intensity of the earthquake decreases with distance from the epicentre, this spatial variation of VTEC can also be summarised as the outgrowth of ionospheric TEC perturbations with the intensity of the earthquake.

5)
The cross-correlation between GPS derived VTEC on earthquake event day within stations of earthquake affected region (of Nepal) showed a strong correlation of VTEC with coefficient +1 at a zero-time lag. Similarly, the cross-correlation study of VTEC on a quiet day among the stations of mid-latitude and lower latitudes (includes earthquake affected and unaffected stations) also showed a strong positive correlation with a peak at +1 at phase. On the other hand, during the earthquake event day, instead of taking the same GPS stations for cross-correlation study, we only observed a strong positive association between VTEC derived from stations that were not affected by the earthquake (SGOC, IISC, HYDE) with a correlation coefficient of +1 at phase. In contrast, all cross-correlation curves between VTEC derived from earthquake-affected and unaffected GPS stations displayed a peak at 0.8 with a certain time-lag indicating less correlation between VTEC from earthquake-affected and unaffected regions. Overall, cross-correlation findings have concluded that the increase in VTEC value in the earthquake event day relative to the quiet day is due to the seismic waves produced by the earthquake.
6) TEC Maps evince that most of the Indian regions, Southern China, Nepal regions were affected heavily, explaining the effect of the earthquake. There were no significant fluctuations a day before and after the earthquake. All the Nepalese stations attain their highest TEC values during the event. When observed at Global Ionospheric Maps, TEC prolongs towards lower or mid-latitudinal region (30 to -30 degrees) through the equatorial plane during the event.
As these naturally occurring phenomena overwhelm the Earth, we have observed abrupt modifications in ionospheric TEC data detected in different monitoring stations. By integrating measurements based on the global impact of such natural events, we can monitor the response of TEC towards the seismic events, and it could be used as a suitable tool to forecast the possible seismic activity. Pulinets (1998) also concluded ionosphere-seismic consequences are an ongoing process and are a key cause of ionosphere variability, and such consequences may be seen at least five days before an earthquake and can be used to predict a near-future earthquake in a region.