Derry Holding 1. July 29, 2017

Transcription

1 Multi-GNSS Vertical Total Electron Content Estimates: Data Analysis and Machine Learning with Python to Evaluate Ionospheric Perturbations from Earthquakes Derry Holding 1 1 Independent Researcher, Holding Engineering Consultancy, Frankfurt, Germany July 29, 2017 Abstract Today, Global Navigation Satellite System (GNSS) observations are routinely used to study the physical processes that occur within the Earth s upper atmosphere. Due to the experienced satellite signal propagation effects the total electron content (TEC) in the ionosphere can be estimated and the derived Global Ionosphere Maps (GIMs) provide an important contribution to monitoring space weather. While large TEC variations are mainly associated with solar activity, small ionospheric perturbations can also be induced by physical processes such as acoustic, gravity and Rayleigh waves, often generated by large earthquakes. In this study Ionospheric perturbations caused by four earthquake events have been observed and are subsequently used as case studies in order to validate an in-house software developed using the Python programming language. The Python libraries primarily utlised are Pandas, Scikit-Learn, Matplotlib, SciPy, NumPy, Basemap, and ObsPy. A combination of Machine Learning and Data Analysis techniques have been applied. This in-house software can parse both receiver independent exchange format (RINEX) versions 2 and 3 raw data, with particular emphasis on multi-gnss observables from GPS, GLONASS and Galileo. BDS compatibility is to be added in the near future. Several case studies focus on four recent earthquakes measuring above a moment magnitude (M W) of 7.0 and include: the 11 March 2011 M W 9.1 Tohoku, Japan, earthquake that also generated a tsunami; the 17 November 2013 M W 7.8 South Scotia Ridge Transform (SSRT), Scotia Sea earthquake; the 19 August 2016 M W 7.4 North Scotia Ridge Transform (NSRT) earthquake; and the 13 November 2016 M W 7.8 Kaikoura, New Zealand, earthquake. Ionospheric disturbances generated by all four earthquakes have been observed by looking at the estimated vertical TEC (VTEC) and residual VTEC values. The results generated from these case studies are similar to those of published studies and validate the integrity of the in-house software. 1 Introduction Natural hazards such as earthquakes, tsunamis and volcanic eruptions pose grave threats to humans. Global Navigation Satellite Systems (GNSS) satellites can act as a primary sensor to identify characteristic signatures caused by natural hazards. These signatures come in the form of crustal deformation measurements and atmosphere-ionosphere coupling. At ionospheric altitudes the presence of neutral particles and ions enables electron density fluctuations [17]. Empirical data retrieved from a high-end GNSS 1

2 receiver can be utilized in such a way to manipulate the observations and so producing geometry-free observables that ultimately detect ionospheric disturbances caused by seismic activity. In recent years several severe tsunamis and earthquakes have highlighted the necessity for effective and robust method modeling as well as providing a stable observational system that identifies the characteristic of an impending tsunami or TEC perturbations in the ionosphere caused by seismic activity. Measurement data obtained from GNSS receivers, configured to collate observables from BeiDou, Galileo, GLONASS and GPS, have proven to be a resourceful method for remote sensing the ionosphere [4]. Many delays contribute to errors in GNSS positioning, such as the receiver and clock offsets, and multipath. However one of the largest errors in GNSS positioning is attributable to the delay in the atmosphere. The GNSS signal from the satellite traverses through the vacuum of space until it reaches the Earth s atmosphere. Upon reaching the Earth s atmosphere, in particular the ionosphere, the GNSS signal experiences refraction, diffraction, and a decrease in the apparent speed. Subsequently this induces an apparent delay of the transit time of the signal from the satellite to the receiver. Computation of the differential phase of the code and carrier measurements retrieved from a GNSS receiver subsequently enables the calculation of the ionospheric delay, in particular the TEC. Numerous studies have already been carried out investigating the ionospheric TEC perturbations following a major seismic event by a dense network of groundbased GNSS receivers. Throughout history mankind has regularly suffered from natural disasters such as the devastation caused be earthquakes and tsunamis. The Sumatra tsunami of 2004 claimed the lives of a staggering 228,000 people, and more recently the death toll following the Tohoku earthquake of 2011 was approximately 16,000 [10]. On 11 March 2011 at 5:46:23 Universal Time (UT) a magnitude M W 9.0 earthquake struck off the coast of Japan causing widespread devastation affecting the northern region of Tohoku as well as triggering off a major tsunami [11]. Several observations have already been published demonstrating that earthquakes and tsunamis produce gravity and acoustic waves that transcend up to the ionosphere influencing fluctuations in electron density in the F region. Previous studies have shown that atmospheric gravity waves generated by tsunamis may be detected as traveling ionospheric disturbances (TIDs), which in turn can also be identified via TEC measurements retrieved from GNSS observations [11]. The ionospheric signature generated by tsunamis has frequently been demonstrated whereby remote observations are made. There are also sources of TIDs that are not attributed to tsunamis or earthquakes, but to significant tropospheric weather, geomagnetic and auroral activity. This makes detection of tsunami-driven TIDs increasingly more challenging [10]. In order to determine the effects of seismic activity and augment previous similar studies this research focuses on observations on the day of a seismic event. By doing so the geomagnetic activity along with TEC perturbations can be observed. Differentiation between regular TIDs and those induced by seismic shocks can also to be determined. When a GNSS satellite is positioned near to the local horizon, its signal will first propagate through the ionosphere, then through the troposphere, before arriving at a receiver. A GNSS receiver configured to collect measurements with a 10 elevation cut-off, and assuming a 350 km ionospheric shell height would have a distance between the ionospheric pierce point (IPP) and the receiver of 1900 km. For 15 elevation cut-off, this distance is approximately 1300 km. Consequently the geometry involved in producing the GNSS TEC measurements is such that a tsunami signature from a coastline can be observed despite being hundreds of kilometers away from land. Current tsunami monitoring systems comprise tidle gauges, Deep-ocean Assessment and Recording of Tsunamis (DART) systems, and seafloor pressure recording sensors, however GPS TEC observations can also assist significantly to these tsunami monitoring systems. Several networks of ground-based GNSS receivers exist that can produce TEC observations. The GEONET network in the Japan region consists of a highly dense network of GNSS receivers. This 2

3 is in contrast to the sparse GNSS receiver coverage currently established in the South Atlantic. For this report, seven sites have been selected for ionospheric behavior analysis: the MIZU site in Japan; the KEPA and KRSA sites in the South Atlantic; the HOIK, KAIK, WEST, and LKTA sites in New Zealand. This report focuses on TEC analysis with the dual-frequency technique, with emphasis on the analysis on the acoustic as opposed to gravity waves. Using the obspy [5] [19] package in Python, Figure 1 has been generated for the four earthquakes outlined in Table 2, which are focused on in this case study. Focal mechanism (beachball) maps are frequently employed by seismologists to demonstrate the magnitude, characteristic, and location of an earthquake. 2 Methodology Determining the absolute VTEC values are useful in order to understand the background ionospheric conditions when looking at the TEC perturbations, however small-scale variations in electron density are of primary interest. Quality checking processed GNSS data, applying carrier phase leveling to the measurements, and comparing the TEC perturbations with a polynomial fit creating residual plots are discussed in this section. Finally this section ends with the types of waves and fault motions involved during and following seismic activity. 2.1 Data Collection and Data Manipulation Time delay and phase advance observables can be measured from dual-frequency GNSS receivers to produce TEC data. There exists a globally distributed network of GNSS receivers, which among other functions, accurately measure the ionosphere delay. The International GNSS Service (IGS) provides high-quality GNSS data products in order to facilitate scientific, educational, and commercial application orientated towards the geodetic and space community. The IGS federation is a non-profit organization comprising more than 200 agencies, universities, and research institutions, globally. The primary aim of the IGS is to provide a highly precise GPS data set in the form of an array of IGS products, that enable scientific advancement and public benefit. The IGS comprises a global network of GNSS ground stations, data centers, and data analysis centers. The available data and derived data products are utilized in Earth science research; positioning, navigation, and timing (PNT) applications; and education. The present IGS network comprises a global network of 350 GNSS receivers. The coverage of the network is widespread but sparse [10]. Other major regional GNSS station networks also exist, such as the GEONET network in Japan, which consists of over 1200 GPS stations. The Southern California s Plate Boundary Observatory (PBO) network consists of over 850 GPS stations in a very dense and local coverage [10]. For the purpose of this report, RINEX (Receiver Independent Exchange Format) files were collated from seven different sites: MIZU at Mizusawa in Japan; and KEPA and KRSA at the King Edward Point Geodetic Observatory; HOIK, KAIK, WEST, LKTA all based in New Zealand. The MIZU data were retrieved from the IGS Network and the KEPA and KRSA data were retrieved via the University of Luxembourg network. Data retrieved from HOIK, KAIK, WEST, LKTA were made available courtesy of the GeoHazard Monitoring Department, GeoNet, New Zealand. The GNSS data comprises both observation and ephemeris (navigation) data. The observation and ephemeris data used were of the RINEX V2 and RINEX V3 formats. The data sampling rate of the RINEX files was 1 Hz, with the exception of the data retrieved from the New Zeland, which has a data sampling rate of 30 Hz The format of how the observations are given in the RINEX V2 and RINEX V3 versions are different. Each different format is to be handled differently, and so a unique method of parsing the data for each format is required. Data processing of the RINEX V2 data was performed with the use of a Python RINEX parsing and VTEC calculation program. Two RINEX parsers support both the RINEX V2 and RINEX V3 formats. These parsers con- 3

4 Figure 1: Global map focusing on four recent earthquakes measuring above a moment magnitude (M W ) of 7.0 for this case study. For clarity, the tectonic plate boundaries are shown here in red. Information given in the grey boxes denote the location, magnitude, and date of the earthquake. SSRT refers to South Scotia Ridge Transform; NSRT is the North Scotia Ridge Transform. Note the difference in colors between the beachballs, where orange represents between M W 6.0 and M W 7.9, and red represents greater than or equal to a magnitude of M W 8.0 solidate the data blocks into multidimensional arrays, then mathematically manipulate the subsequent multidimensional arrays, in order to plot vertical TEC values (VTEC) against a timeline. Typically a 12 MB file will take 1 min to process. A precondition for the RINEX V2 parsing was first necessary. This pre-step involves retaining the GPS (GLONASS, or Galileo) data and filtering out the remaining GNSS and SBAS systems from the RINEX V2 file. To achieve this, UNAVCOs TEQC tool was employed [9]. In the case of manipulating Galileo data, TEQC was employed to include the Galileo observables on a signal-by-signal basis into a RINEX file, thus preparing the RINEX data in order to be processed by the Python RINEX parsing tool. Fermat s principle states that the path the wave traverses is the path that can be traversed in the quickest time. This results in the GNSS signal s path being arched, see Figure Total Electron Content Electromagnetic waves are influenced by refractive mediums in terms of the distance the wave traverses. 4

5 D = S ρ = c t = = s r s r n ds s r (n 1) dρ } {{ } delay dρ s + r s n ds r n dρ } {{ } bending effect (3) As the GNSS signal propagates through the ionosphere, this delay is known as the ionosphere delay or ionospheric refraction. To first order, the refractive index of a frequency, f, can be approximated to Figure 2: Wave bending in a refractive medium The refractive index, n, is defined as the ratio of the speed of light, c, in a vacuum and the velocity of the wave, v, in the medium: where: n = c v with v = 1 ɛµ (1) ɛ permittivity of the medium µ permeability of the medium c is 299,792,458 ms 1 For an increment of time, dt, and distance, ds, the velocity can be shown as v = ds/dt. Using v = ds/dt with (1), then the traversed time t of a signal from the GNSS satellite s to receiver r is given by: t = s r dt = s r 1 v ds = 1 s n ds (2) c r The difference between the measured range S and the geometrical range ρ gives the delay (or advance) D, and using (2) can be shown as [15]: n ph = N e f 2 and D ph = 40.3 f 2 where: n ph n gr D ph D gr N e n gr = N e f 2 (4) N e dρ D gr = 40.3 f 2 phase refractivity index, group refractivity index, phase delay, group delay, electron density. N e dρ (5) The ionized gas disperses electromagnetic waves within the ionosphere, and so the refractive index is dependent on the frequency and electron density. Dispersion in terms of pseudorange measurements of group and phase are equal in magnitude but opposite in sign. Since n gr > n ph, it logically follows that v gr < v ph, so that the different velocities induce the carrier phase advance and modulated code delay. In terms of GNSS, the observed code measurements that propagate with the group velocity are delayed and the observed carrier phase measurements that propagate with phase velocity are advanced. Comparing these measurements with the geometric distance between the satellite and the receiver, the code 5

6 pseudoranges are measured too long, while the phase pseudoranges are measured too short [1]. The number of free electrons along the path of a signal from the satellite to receiver influences a radio signal s propagation speed. The number of free electrons within a tube, with cross section 1 m 2 extending from the satellite to the receiver can be defined as the Total Electron Content (TEC) or Slant Total Electron Content (STEC) ST EC = R S N e dρ (6) where N e is the variable electron density along the path of the signal, and integration is along the signal path from the satellite to receiver, dρ [21]. Substituting STEC into (5) gives D ph = 40.3 f 2 ST EC D gr = 40.3 ST EC (7) f 2 as a final representation of ionospheric refraction (ionospheric delay) to first order given in terms of TEC, in metres. TEC is often denoted in TEC units (TECU) where [14] 1 T ECU = el m 2 (8) Navigation systems are vulnerable to TEC fluctuation events, where one TECU corresponds to a range delay of approximately 0.16 m and 0.54 ns on the GPS L1 frequency [6]. The nominal range is between to electrons per m 2 with minimum density occurring at local midnight and maximum density at local afternoon [27]. 2.3 Determining TEC from Dual- Frequency Obervables Depending on the geographic location of the earthquake and tsunami the station of interest can be selected. Using the Python tool, RINEX files are parsed and an STEC time series can be generated. The timeframe selected from the RINEX data is chosen encompassing the period just before and after the earthquake event. Quantifying TEC is challenging simply because TEC depends on the electron density, which in turn depends on temporal and spatial (observation site) variations, line of sight (LOS) of the signal, and sunspot activity. This makes it difficult to find an appropriate formula to model the effects of the ionosphere. The most efficient method to eliminate the effect of the ionosphere is by using signals whose frequencies differ. Primarily for this reason space geodetic techniques require GNSS satellites to transmit in at least two different frequencies [1]. A dualfrequency GNSS receiver can estimate ionospheric group delay and phase advance from the measurements. The dispersive nature of the ionosphere acting on GNSS signals enables elimination of the ionospheric refraction of the observables. This assumes however that the second and higher order terms are neglected. The elimination of the ionospheric refraction is performed by taking the ionosphere-free linear combination of pseudorange and carrier phase measurements. GPS dual-frequency carrier phase and code delay observations are combined to obtain observables that relate to the TEC values along the GNSS signal s LOS. The procedure for determining first order ionospheric effects in GNSS measurements is performed by the geometry-free linear combination. The geometric distance between the receiver and satellite, and introducing an error term gives [24] P = ρ+c(δ R δ S )+c(b R b S )+I+T +M P R +ɛ (9) where Ideally the true range of the satellite ρ, is to be determined. Instead it is the biased and noisy measurement of ρ, that is, the pseudorange PR S that is obtained. Elaborating the terms in (9) a little further: ρ: the true geometric range is considered to be the geometric distance between the satellite s antenna electrical phase center and the receiver s electrical phase center. The actual location of the antenna s electrical phase center is dependent on antenna hardware, the carrier 6

7 ρ δ S δ R B R b S I T M P R ɛ true geometric range between the satellite and receiver; satellite clock error due to the difference in the arbitrary timescale; receiver clock error due to the difference in the arbitrary timescale; code biases induced by receiver hardware delay; code biases induced by satellite hardware delay; signal delay of the ionosphere; signal delay of the troposphere; code multipath error caused by reflected signals; all unmodelled effects and receiver noise. frequency, the azimuth and elevation of the signal. The true geometric range also compensates for all relevant displacements, including but not limited to ocean tide loading, earth body tides, atmospheric loading, and plate tectonics [24], B R and b S : GNSS satellites modulate different codes and navigation messages onto different carrier frequencies. The dispersive property of the ionosphere means that these codes, navigation messages and carrier frequencies are also dispersed differently. The GNSS signal is imperfectly synchronized as a result of the different digital and analog signal paths that correspond to each individual signal. This leads to the satellite biases. The receiver biases are induced by the analog hardware and by-pass filters, I: the electromagnetic GNSS signal is dispersive when traversing through the ionosphere. This ultimately leads to the delay of the signal. Ionosphere delay principles are discussed in great detail throughout this research, T : the tropospheric delay comprises a dry and hydrostatic component that is dependent of the local atmosphere. Tropospheric refraction unlike that of the ionospheric signal delay does not depend on the frequency of the carrier for frequencies used in the GNSS frequency range, M: The path of a GNSS signal not only takes the route of direct line of sight, but may also reflect from objects in the surrounding environment, such as trees or within urban canyons. The GNSS signal arrives at the GNSS receiver antenna through different paths, resulting in multipath. The effect of multipath is much larger for pseudorange than for carrier phase. The range between the satellite and the receiver can also be determined through the carrier phase of the signal. The total number of carrier cycles plus the fractional cycle both at the receiver and the satellite, then multiplied by the carrier wavelength, results in the range. Ranges determined by the carrier are more accurate compared to those obtained from the pseudoranges, since the wavelength of the carrier phase is much smaller than that of the codes. Carrier phase measurement can be expressed as [27]: L L = ρ + c(δ R δ S ) + c(d R d S ) where D R d S P M CP λ N ɛ I + T + P + M CP + λn + ɛ (10) phase measurement in unit of lengths; phase biases induced by receiver hardware delay for the carrier wave; phase biases induced by satellite hardware delay for the carrier wave; phase wind-up effect; phase multipath error caused by reflected signals; carrier phase wavelength; initial integer ambiguity; all unmodelled effects and receiver noise. Note that the negative sign in the above equation for the ionospheric delay. This is because for the carrier phase traversing the ionosphere, the signal is advanced, not delayed as is the case for the code measurements. The phase and pseudorange measurements, like that in (10) and in (9) can be expressed in terms of a frequency, i, giving 7

8 P i = ρ + c(δ R δ S ) + c(b R b S ) L i = ρ + c(δ R δ S ) + c(d R d S ) + I i + T + M P Ri + ɛ (11) I i + T + M CP i + λ i N i + ɛ (12) where i is equal to 1, 2, and denotes the different frequencies. Corrections related to relativistic effects, phase wind-up effect, and antenna phase center deviations are omitted from (11) and (12). The Linear Combination (LC) enables dualfrequency receivers to measure the ionospheric delay via its dual-frequency observations. The pseudocode ranges are determined from the measurements of the signals P 1 and P 2 that are modulated on the two carrier frequencies L 1 and L 2. The ionospheric terms I 1 and I 2 are thus identical to (7). By combining the measurements, L 1 and L 2, the ionospheric delay can be removed from the measurements and provide the LC solution. The following observable equations comprise the dimension of length; multipath terms and noise are not explicitly expressed; higher-order ionospheric terms are neglected: L 1 = ρ I 1 + λ 1 N 1 ( ) f 2 L 2 = ρ 1 I 1 + λ 2 N 2 f 2 2 P 1 = ρ + I 1 ( f 2 P 2 = ρ + 1 f 2 2 ) I 1 (13) Where ρ is non dispersive delay and consists of LOS, clocks and tropospheric biases; I 1 is the dispersive delay on the first frequency; and N 1, N 2 are integer ambiguities on L 1 and L 2. For GPS the satellites operate on two different frequencies, f 1 and f 2, which are derived from the fundamental frequency f 0, of MHz f 1 = 154 f 0 = MHz f 2 = 120 f 0 = MHz (14) Since it assumed that the L1 and L2 observables traverse along the same path through the ionosphere, then the difference in ionosphere delays between these frequencies can be measured with use of a dualfrequency GPS receiver. This leads to the obtained group delay as being ( 1 P 2 P 1 = 40.3 T EC f2 2 1 ) f1 2 (15) Where P 1 and P 2 are group path lengths corresponding to the f 1 and f 2 frequencies, respectively. Using (15) results in the TEC to be derived as T EC = 1 ( ) f1 f 2 (P 2 P 1 ) (16) 40.3 f 1 f 2 where P 1 and P 2 are the pseudoranges measured in L1 and L2, respectively [27] [22]. In terms of representing (16) for the GPS L1 and L2 carriers, this becomes T EC = (P 2 P 1 ) (17) For GLONASS, determining (17) is dependent on the frequency and so is dependent on the satellite number. GLONASS satellites use Frequency Division Multiple Access (FDMA) modulation techniques unlike that of GPS and Galileo, which uses Code Division Multiple Access (CDMA). 2.4 Phase and Group Delay Since the ionosphere is dispersive, the time delay attributed to the ionosphere is dependent on the signal s frequency. The dispersive nature of the ionosphere means that the codes which are the modulations on the carrier wave are dispersed differently compared to the carrier phase along the signal s path through the ionosphere. In terms of GPS, this means that the P, C/A codes and navigation message are slowed down and affected by the group delay. However the carrier wave itself appears to speed up in the ionosphere and so affected by the phase delay. Calling this increase in speed a delay is counter intuitive and so this term is also referred to as the phase advancement. 8

9 Using data retrieved from the Center of Orbit Determination in Europe (CODE) site (ftp://ftp. unibe.ch/aiub/code), the differential code biases are subtracted from the ionospheric observables [18] The geometry-free linear combination is employed to determine the ionosphere delay in terms of electron density (STEC) and instrumental biases DCB C1 C2. Currently the DCB C1 C2 formats are not available as a final product and so for this report the DCB C1 C2 has been determined using CODE s products as outlined in (18) of the ionospheric pierce point (IPP) as in Figure 3, and for example in Figure 11. Since the GNSS signal s path length depends on the satellite s position in the sky, this also has an effect on the TEC: the lower the satellite elevation, the longer the path of the signal and so the higher the TEC value [27] [3] [7] [23] [2] [16]. Assuming no lateral electron gradients exist, TEC is calculated by mapping the slant of the path signal to its corresponding vertical path. So STEC is defined as the sum of electron density Ne, along the ray path, as defined in (6). SDCB theo C1C2 = SDCB P 1P 2 + SDCB P 1C1 SDCB P 2C2 (18) where SDCBC1C2 theo is the theoretical value for SDCB C1-C2. SDCB P 1P 2, SDCB P 1C1, and SDCB P 2C2 are all instrumental biases retrieved from CODE for SDCB P1-P2, SDCB P1-C1, and SDCB P2-C2, respectively [18]. 2.5 Determining VTEC: Thin Shell Mapping Function Accurate monitoring of the ionosphere has profound implications in most areas of the GNSS community. The thin shell mapping function is one of the first assumptions to consider when ionospheric corrections are estimated and applied from GNSS data. In many GNSS applications the typical assumption is to consider a fixed mapping function constant, that is projected onto a 2D distribution of electron content at a given effective height. The main contributor to TEC variations occurs at around the height whereby the ionization within the ionosphere is at a maximum. This maximum occurs in the F 2 layer, and so this enables a simplistic thin shell encompassing the Earth when modeling the ionosphere. The thin shell mapping function assumes that all electrons are compressed into this spherical shell of infinitesimal thickness surrounding the Earth. The point of intersection along the line of sight is referred to as the ionospheric pierce point (IPP). The propagation of TIDs are represented by subionospheric points (SIP), which are a projection Figure 3: Propagation path through the ionosphere The slant or obliquity factor (OF) as is known in the Klobuchar model is applied to the TEC value to obtain the Vertical TEC, or VTEC. The VTEC value however does not correspond to the location of the GNSS receiver, and due to the thin shell assumption it instead refers to the SIP, as shown in Figure 3. The mapping process depends on the elevation angle, e, and its complement zenith angle, z, of the satellite. Calculating the elevation angle is performed by the parameters given in the Broadcast Ephemeris and depends on the type of GNSS under analysis. The method employed for the satellite orbit determination, which in turn is used to determine the elevation and azimuth for GPS, Galileo, 9

10 and BeiDou follows a very similar process with one another. For GLONASS the process employed uses the Runge Kutta 4th order technique [8] [24]. The lower the elevation angle between a particular receiver-satellite pair, the higher the absolute STEC values as well as increased amplitude variations in STEC. This is because the integration of TEC perturbation is over a longer signal ray path at a lower elevation angle than that of a shorter ray path at a higher elevation angle, with the highest elevation being when the satellite is located at zenith. Therefore exaggerated fluctuations of STEC are observed at lower elevation angles than at higher ones. This does not however imply that seismic-driven TEC perturbations are deemed invalid when considering STEC observations made at lower elevation angles, however caution should be taken since the amplitudes of these TEC perturbations may be artificially amplified [10]. The ionospheric shell height, H, used in ionosphere modeling has been open to debate for many years and typically ranges from km, which corresponds to the maximum electron density within the ionosphere. The mapping function compensates for the increased path length traversed by the signal within the ionosphere. Figure 4 demonstrates the impact of varying the IPP height on the TEC values. At zenith angle, z, VTEC can be translated to STEC respective of path lengths through the ionospheric shell by Figure 4: Impact on TEC values from varying IPP heights. The height of the thin shell, H, is increased in 50 km increments from 300 to 500 km. MF(z) thin shell mapping function; z, z respective geocentric zenith angles at the height of the GNSS receiver and the ionospheric pierce point at the thin shell; R, R O respective radius to the GNSS receiver and the mean radius to the Earth (approximately 6,371 km); H height of the thin shell above the Earth s mean surface. ST EC MF (z) = V T EC = 1 cosz = 1 1 sin2 z with sinz R = R O + H sinz (19) where: The mapping function increases with increasing zenith angle z to the satellite. For small elevation angles TEC can reach up to three times the value of VTEC, as can be seen by Table 1. Figure 4 shows how increasing the IPP value from 300 km in increments of 50 km impacts the calculated VTEC via the mapping function. This variation can be seen to differ whereby the resulting VTEC difference is not linear, but is also dependent on the elevation angle. Note the relatively consistent VTEC differences (between adjacent IPP heights) till 6.5 UT, thereafter the VTEC differences between each IPP altitude until 7.9 UT become greater. Figure 5 shows the elevation angle for the corresponding satellite (G21) for Figure 4. The greater variation observed after 6.5 UT can be explained by the decreasing elevation angle from 31 to 2 towards the end of the recorded measurements. At low elevations the GNSS signal has a greater path length to traverse through the ionosphere. By rearranging (19) it can be derived that an in- 10

11 low-latitudes where electron density is greater, then obvious errors become more apparent [27]. Figure 5: Elevation angle on 11 March 2011 for G21 at MIZU. crease in H decreases sinz which entails an increase in cosz. Hence an increase in the height of the thin shell in the mapping function results in an increase in the VTEC value. Table 1: Mapping Function, MF, comparisons. First three columns units in degrees (Source [24]). Elevation angle z z MF Distance (km) Values of mapping functions range from 1 in the zenith direction to approximately 3 at an elevation angle of 5. It should be noted that the thin layer model currently deployed in the GPS domain consists of discrepancies, resulting from STEC to VTEC conversions. Within the mid-latitude regions the electron density is small and so in these geographic locations the conversion induces small errors. At 2.6 Carrier Phase Leveling Single-frequency GPS users can utilise models of the ionosphere such as the Klobuchar Model, also known as the GPS Broadcast Model. The Klobuchar Model is constructed using ionospheric parameters given in the GPS broadcast message and is represented by a third degree polynomial. The coefficients of this third degree polynomial are transmitted as part of the broadcast message header [22]. Using this method enables ionospheric parameters such as TEC to be extracted. For dual-frequency GNSS users however TEC values can be retrieved with the use of dual-frequency measurements by applying calculations like that in (16). Calculation of TEC by applying (16) for pseudorange measurements in practice produces a noisy outcome and so the relative phase delay between two carrier frequencies which produces a more precise representation of TEC fluctuations is preferred. Since the actual number of phase cycles is unknown, obtaining the differential carrier phase yields a precise measure for only the relative TEC; the absolute TEC values are thus only determined if the pseudorange measurements are also used. Therefore the pseudorange measurements are used to obtain the absolute scale of the TEC values and the carrier phase measurements are then used to provide increased precision of the TEC values [22]. As previously mentioned the noise level of the pseudorange measurements is high and so produces TEC values with greater uncertainty whereas the noise level of the carrier phase measurements is significantly lower than that of the pseudorange. To circumvent the effect of pseudorange noise on TEC data, GNSS pseudorange measurements can be smoothed by carrier phase measurements, with the use of the carrier phase smoothing technique, which is often referred to as carrier phase leveling [22]. The carrier phase smoothing technique is essentially a combination of the noisy code pseudorange measurements with the comparatively smooth carrier phase measurements. These comparatively smaller measure- 11

12 ment errors seen in the carrier phase measurements as opposed to the pseudorange measurement errors enable the determination of STEC through smoothing the pseudoranges [22]. The steps applied within the carrier phase leveling technique are outlined below: 3. The smoothed differential delay was then converted to absolute STEC by multiplying with a constant (as in Equation 21) then using the mapping function as in Equation 19 the conversion from STEC to VTEC was finally made [22]. 1. The phase measurements given in units of phase cycles are multiplied by the wavelength of the carrier to obtain values in units of length. The geometry free linear combination is given by [25]: L 4 = L 1 L 2 ( = 1 f 1 2 ) f2 2 I 1 + (λ 1 N 1 λ 2 N 2 ) (20) The resulting phase derived slant delay calculated from the geometry free linear combination, L 4 (20), is scaled to zero for the relative range at the first epoch. This step enables the elimination of the integer ambiguity, N, if no cycle slips are present [22]. This results in L 4 = ( 1 f 1 2 ) f2 2 I 1 (21) 2. The code multipath effect typically seen at low elevation angles or at either end of the path of the signal was eliminated by fitting the code differential delay at the higher elevation angles. The parser tool used to plot ionospheric delay from the RINEX data has been developed in such a way to seek out the highest elevation angle for each individual satellite, then at this particular epoch identifying the corresponding differential delay value and defining this value as the shift value. This shift value was subsequently added to the relative phase values and so fitting the data to the code differential delay. Consequently the absolute differential delay was determined and any remaining noise was regarded as negligible [22]. This procedure is shown in Figure 6. Figure 6: Phase smoothed code differential delay This report is primarily although not exclusively focused on only the observations of small-scale variations in the electron density, which implies just the changes in TEC values as opposed to the absolute TEC values [10]. The slant TEC (STEC) values are mapped to the vertical using the mapping function and so producing estimates for VTEC [10]. Nevertheless for completeness the plots shown in the Results section are given in absolute TEC values, as described in the above carrier phase leveling procedure. Longer periodic variations in the TEC time series, such as diurnal and seasonal trends, are avoided by selecting time series data sets typically of the 3 4 hour range. The IPP geographic location is defined as the geographic location where the LOS between the GNSS satellite and the GNSS receiver penetrates the F region of the ionosphere at its peak electron density. For the Tohoku event, Galven et al [10] determined the F region peak density altitude by using JPL s Global Assimilative Ionosphere Model (GAIM). The GAIM process ingests ground-based 12

13 GPS as well space-based radio-occulation TEC measurements from the COSMIC constellation. This subsequently generates an electron density profile as a function of altitude. This particular profile was determined for the location above the Tohoku earthquake epicenter. Using the GAIM profiles over Japan resulted in an F region peak of 300 km altitude. The IPP altitude in this report based on these findings has applied an F region peak of 300 km altitude. 2.7 Residual Determination For the purpose of this study the monitoring of small-scale variations in ionospheric electron density from the ionospheric observables are of particular interest. Longer period variations can be associated with diurnal alterations, and changes in the receiversatellite elevation angles. In order to remove these longer period variations in the TEC time series as well as to monitor more closely the small-scale variations in ionospheric electron density, a higher-order polynomial is fitted to the TEC time series. This higher-order polynomial fit is then subtracted from the observed TEC values resulting in the residuals. The variation of TEC due to the TID perturbation are thus represented by the residuals. For this report the polynomial order applied was typically greater than 4, and was chosen to emulate the nature of the arc for that particular time series. The order number selected is dependent on the nature of arcs displayed upon calculating the VTEC values after an initial inspection of the VTEC plots [10]. 2.8 Waves and Fault Motion Earthquakes are generally categorized by their fault motion and are commonly consisting of one or more types of a fault motion. The reverse motion is predominantly caused by the uplifting at the surface whereby vertical co-seismic crustal displacement is a standard feature. Normal motion (also known as thrust motion) is caused by the subsidence at the surface. The third type of motion, the strike-slip motion (SSM) is predominantly characterised by horizontal motion and so generates little vertical displacement [4]. Following an earthquake are atmospheric waves, which appear as one of three types: (1) acoustic waves occurring in close proximity to the epicenter; (2) gravity waves generated by a tsunami; (3) secondary acoustic waves caused by the Rayleigh surface wave originating from the epicenter [17]. These coseismic TIDs have been identified as having propagation speeds of 1000 ms -1, 3.4 km s -1, and up to 200 ms -1 (triggered by a fault slip earthquake followed by an impending tsunami with a depth of 5000 m) for acoustic, Rayleigh, and gravity waves, respectively [28] Acoustic Waves Co-seismic vertical motion of the ground perturbs the ionosphere and is referred to as co-seismic ionospheric disturbances (CID). The sudden impulsive force of the ground or sea induces atmospheric pressure waves that propagate skyward into the troposphere and ionosphere. This affect is detectable using remote sensing techniques that detect ionospheric disturbances. These atmospheric waves can occur as an acoustic wave or as a Rayleigh wave, both of which demonstrating different propagation velocities. The vertical component of the co-seismic crustal deformation is analogous to a piston-like movement and is attributable to the source of the primary acoustic waves [4]. Acoustic waves are described as being sound waves induced by longitudinal compression in the propagation s direction. These acoustic waves are typically generated by earthquakes and propagate through the atmosphere at the speed of sound. In a region where the fault line is located underneath a large body of water such as an ocean, this displacement of the seafloor and ultimately the overlying body of water is responsible for the generation of a tsunami [4]. Depending on the region of the atmosphere traversed by the acoustic wave, the speed can vary between several hundreds of ms -1 at sea level to 1000 ms -1 at 400 km altitude. Therefore a duration of between 10 to 15 min is typical for a wave to affect 13

14 the F region and to observe TEC perturbations. The acoustic wave generated by a seismic event at the epicenter traverses isotropically in the atmosphere, whereby the horizontal and vertical components of the propagation velocity are identical [10] Rayleigh Waves Earthquakes also generate Rayleigh waves. These waves transverse solid Earth waves propagating along the surface with a horizontal velocity of about 3.4 km s -1. These solid Earth (Rayleigh) waves also produce acoustic waves within the immediate atmosphere. These newly generated acoustic waves induce a corresponding electron density fluctuation traversing in the ionosphere that is also consistent with the moving horizontal 3.4 kms -1 Earth wave. Not all submarine earthquakes produce tsunamis, and so tsunami-driven ionospheric disturbances may not necessarily be observed. Earthquake-driven ionospheric disturbances would exist regardless, and so care needs to be taken when differentiating the source of the ionospheric disturbance [10] Gravity Waves As a tsunami begins to traverse the ocean a gravity wave is generated, propagating obliquely into the atmosphere with both horizontal and vertical components. In the case of gravity waves, the Earth s surface rises and falls generating vertical transverse oscillations of pockets of air. These oscillations then induce buoyancy waves and are observed as the internal gravity waves. These waves propagate vertically with velocities in the region of 40 to 50 ms -1, reaching the F region in over 2 hours [10]. As the gravity wave propagates towards higher altitudes, the amplitude of the wave also increases as a result of the exponential decline of atmospheric density with increasing height. The wave amplitude continues to increase as the propagation altitude through the ionosphere also increases. The end effect is waves with much larger amplitudes are observed in the F as opposed to the E region. Gravity waves could be powerful enough to perturb the E region and subsequently be detected in TEC observations, however these perturbations would pale into insignificance compared to the ensuing F region TEC perturbations. This further advocates the reasoning behind the assumption that tsunami-driven waves will be observed as significant TEC perturbations at an altitude of 300 to 400 km where peak electron density occurs [10]. The TIDs caused by tsunamis can be easily differentiated from the TIDs caused by other natural sources (such as geomagnetic activity, ionospheric storms, tropospheric weather) by the period, velocity and propagation direction of the tsunamis that generated these TIDs [17]. 3 Results For this analysis four earthquake events have been selected with either an IGS or private GNSS station with near proximity to the epicenter. This section describes the seismological characteristics of each event, as well as the ionospheric perturbation associated with each of these events. Table 2: Earthquake main parameters. Seismological parameters were taken from the Earthquake Hazards Program of the U.S. Geological Survey (USGS) website ( Date Name M w Time (UT) 11/03/11 Tohoku 9.0 5:46:24 17/11/13 19/08/16 Scotia Sea South Georgia 7.7 9:04: :32:22 13/11/16 Kaikoura :02: Tohoku Earthquake Location (lat; lon) 38.30; ; ; ; Depth (km) The sampled data was retrieved from the IGS station, MIZU, located at Mizusawa, Japan. The MIZU site is 39N and 141E The 14

15 location of the data collection site, MIZU, and the earthquake epicenter can be seen in Figure 7. Figure 8 displays the ionospheric delay in terms of vertical TEC (VTEC), in units of TECU (1 TECU = el m -2 ). The plot is split into two smaller subplots, the upper section displaying the ionospheric delay (VTEC) in units of TECU, the lower displaying the residuals. The vertical grey-dashed lined corresponds to the epoch of the earthquake at 05:46:23 UT (2:46:23 PM local time) on March In the upper section of the plot, the blue line corresponds to the absolute VTEC value calculated from the observations, in this case L1 and L2 on GPS, whereby the carrier phase leveling technique was applied to the data set. The VTEC values are mapped from the STEC values which are calculated from the LOS between MIZU and the GPS satellite PRN 18 (on Figure 8 denoted G18). For this particular data set as seen in Figure 8, a polynomial fit of five degrees was applied, which corresponds to the red-dashed line. As an alternative to polynomial fitting, bandpass filtering can be employed when TEC perturbations are desired. However for the scope of this report polynomial fitting to the time series of TEC data was the only method used [11]. In the lower section of Figure 8 the residuals are plotted. The residuals are simply the phase smoothed delay values (the blue line) minus the polynomial fit line (the red-dashed line). All ionosphere delay plots follow the same layout pattern and all time data is represented in UT (UT = GPS - 15 leap seconds, whereby 15 leap seconds correspond to the amount of leap seconds at the time of the seismic event). The time series shown for the ionosphere delay plots are given in terms of decimal of the hour, so that the format follows hh.hh. Figure 7: MIZU IGS station and Tohoku earthquake epicenter. Figure 8: VTEC and residual plot for G18 at MIZU on March In Figure 8 it can be seen that the resulting residuals resemble a quasi damped sine wave signature with 11 min periods commencing 12 min after the earthquake event. Acoustic waves in the atmosphere propagate omnidirectionally at the speed of sound, which at 300 km altitude (F peak region) is 1000 ms -1. The VTEC perturbations seen in the plots are consistent with a propagation velocity of an acoustic wave generated by the earthquake [11]. Compared to the velocity of the tsunami traveling horizontally this velocity would be ms -1 away from the epicenter. A 15

16 (a) (b) (c) (d) Figure 9: VTEC and residual plots at MIZU on March The blue curve in the upper portion of the plot is the observed VTEC value and the red-dashed curve is the polynomial fitting applied to the curve. The vertical grey-dashed line signifies the earthquake epoch. The lower portion of the plot demonstrates the residuals, which have been calculated from the observed VTEC values minus the polynomial fit values. The plots are from perspective of the GNSS receiver at MIZU for four GPS satellites (a) G05 (b) G09; (c) G15; (d) G18. The y-axes for absolute VTEC values have been adjusted and so are not consistent between each plot. The residuals y-axes for each plot however are consistent and are limited to ±1.8 TECU. The time scales (x-axes) are dependent on when the satellite was tracked between 5.00 UT and 8.00 UT, and so may also vary. 16

17 (a) (b) (c) (d) Figure 10: VTEC and residual plots at MIZU on March The plots are from the perspective of the GNSS receiver at MIZU for four GPS satellites (a) G21 (b) G26; (c) G27; (d) G28. 17

18 gravity wave would subsequently be generated propagating obliquely upward from the tsunami wavefront [11]. The vertical speed of the gravity wave is 50 ms -1 and is significantly lower than that of the acoustic wave. Consequently, this implies the initial TEC perturbations seen from Figure 8 are highly likely to have been induced by the acoustic waves generated from the earthquake. The gravity waves however are expected to be seen at a later stage [11]. In order to establish a higher degree of confidence whether a TEC perturbation has been caused by a surface disturbance, such as from seismic activity, not one but multiple satellite-receiver pair time series are to be analysed. The initial TEC fluctuations seen in the residual plots for G05, G09, G15, G18, G21, G26, G27, and G28 are congruent with a propagation velocity of an acoustic wave ( 1000 ms -1 ) and so therefore likely to have been caused from such acoustic waves generated from the earthquake. In order to diminish any ambiguity regarding the origin of TEC perturbations such as those caused at the surface (like that of an earthquake or tsunami) or another source such as from a geomagnetic disturbance not just one but several receiver-satellite pair time series were assessed [11]. Figure 11: Geometry of the satellites, ground track and IPP. GNSS satellites orbit the Earth at km altitude. The IPP is set to 300 km altitude. If a GNSS satellite were imagined to be orbiting instead of at km, at the IPP altitude of 300 km along the satellite-receiver LOS, then its projected ground track on the surface of the Earth at IPP is represented by the blue line. The focal mechanism (beach ball) corresponds to the epicenter of the earthquake; the red stars (located on the blue lines) represent the location of this imaginary GNSS satellite upon the IPP at the exact time an earthquake strikes. The orange lines represent tectonic plate boundaries. The residual plots in Figure 9 and Figure 10 show significant variations with amplitudes ranging from 0.6 TECU (G09, G15, G21, G26, G27, G28) to 1.2 TECU (G05). The amplitude with the least variation (0.6 TECU) with greatest background absolute TEC (21.2 TECU) is observed for G21. This corresponds to an observed variation of 2.8%. Conversely the greatest amplitude variation of 1.2 TECU and background absolute TEC of 14 TECU is seen for G05. In this case the observed variation is 8.6%. The initial TEC perturbations in each of the residual plots occurs at 15 min after the earthquake event, with a disturbance still present even after 1.5 h 18

19 (G18). If these TEC perturbations at 15 min after the earthquake were caused by gravity waves generated by the tsunami then this would imply a velocity much faster than the typical 50 ms -1 consistent with the velocity of gravity waves. At 50 ms -1 with an F region peak of 300 km, the resulting time taken would be 1.7 h from sea level. Following on from these estimates the TEC perturbation signatures observed were very unlikely to have been generated from a tsunami-driven gravity wave. TEC perturbation was observed 15 min after the earthquake occurred. Consequently it is much more probable that the TEC perturbation observed is the result of acoustic waves generated by the earthquake itself. Such acoustic waves would exist regardless of the generated tsunami being present or not. 3.2 Scotia Sea Earthquake Two large seismic events in the South Scotia region are under analysis in this report. The first event occurred on the 17 November 2013 with a magnitude of 7.7 M W at 9:04:55 UT with a depth of 10 km. A seismic event struck in the South Georgia Island Region with a magnitude of 7.4 M W at 7:32:22 UT on 19 August 2016, at a depth of 10.0 km. The University of Luxembourg, along with the United Kingdom National Oceanography Centre, British Antarctic Survey, and UNAVCO collaborated in February 2013 to construct the King Edward Point Geodetic Observatory on South Georgia, in the South Atlantic Ocean [26]. The remote location of the South Georgia Island within the South Atlantic as well as the proximity to the Scotia tectonic plate amplifies the importance of this region for numerous global monitoring networks. The seismic, geomagnetic and oceanic domains in particular continue to benefit from the presence of monitoring networks in the South Atlantic. The King Edward Point Geodetic Observatory comprises of two autonomous GNSS stations (character ID: KEPA and KRSA) that employs a Trimble NetR9 GNSS receiver and Trimble choke ring GNSS antenna. The ionospheric TEC perturbations were captured by a GNSS receiver at KEPA for GPS satellites G20, G25, and G31, as well as for GLONASS satellite R20, see Figure 12. It is generally assumed that the background ionospheric parameters such as the absolute VTEC in the registration zone (in this case the GNSS receiver site) are one of the main factors that may influence the behavior of the TEC perturbation [4]. The perturbation is transferred using the neutral waves via charged particles. Therefore for a successful comparison of TEC residuals from different earthquakes should also consider the geometry of the satellitereceiver pair. From all four plots, the R20 satellite was the closest satellite-receiver pair that captured the most profound ionospheric disturbance with a recorded TEC response occurring at 11 min after the earthquake with a largest amplitude of 1.3 TECU registered. A significant variation of TEC values was observed with a period of 9 min. The amplitude of this variation is 1.2 TECU with a background of absolute TEC being TECU. Consequently the observed variation has a range of 2.6% of the background TEC. GPS satellite G20 recorded a largest amplitude of 0.6 TECU at 19 min after the earthquake whereas G25 registered an even smaller amplitude of 0.3 TECU at 20 min later. For G20 this corresponds to an amplitude variation of 0.6 TECU with a background of absolute TEC being 37 TECU, leading to 1.6% of the background TEC. For G31 due to the wave-like nature of the residual curve before and after the earthquake, no disturbance is observed although there does appear to be a more apparent perturbation at 60 min; however since the amplitude in the residual plot does not appear more significant than the background TECU variation this perturbation is consequently deemed negligible. 19

20 (a) (b) (c) (d) Figure 12: VTEC and residual plots at KEPA on 17 November The blue curve in the upper portion of the plot is the observed VTEC value and the red-dashed curve is the polynomial fitting applied to the curve. The vertical grey-dashed line signifies the earthquake epoch. The lower portion of the plot demonstrates the residuals, which have been calculated from the observed VTEC minus the polynomial fit. The plots are from the perspective of the GNSS receiver at KEPA, for one GLONASS satellite (a) R20 and three GPS satellites (b) G20; (c) G25; (d) G31. 20

21 Figure 13: Geometry of the satellites projected ground track whereby the IPP is set to 300 km altitude. The orange lines correspond to tectonic plate boundaries. 3.3 South Georgia Earthquake In the South Georgia Island region located in the North Scotia Ridge Transform (NSRT) plate boundary between the South American and Scotia plates [29] on 19 August 2016, a magnitude of 7.4 M W earthquake struck at 7:32:22 UT. This subsection analyses the data retrieved from KEPA and KRSA. As well as computing the GPS and GLONASS TEC values, four Galileo satellites (E08, E14, E26, E28) are also analysed. Taking a look at the Galileo satellites first, Figure 14 demonstrates the TEC perturbations as computed for the Galileo L1 and L5 carrier frequencies. It should be noted that these computed values do not take into consideration the DCB values and are therefore not included in the final VTEC values. The L1 and L5 carrier frequencies are outlined here first because of the phase difference between L1 (frequency band E1, MHz) and L5 (frequency band E5a, MHz). From all five Galileo frequency bands (E1, E5a, E5b, E5, E6) these two combinations in particular create the largest phase difference, and when considering the dispersive nature of the ionosphere, this phase difference facilitates an optimised condition for analysis for TEC perturbation. Nevertheless later on in this report several other signal combinations are presented and discussed to augment the case for wide separation of signal carriers for TEC observation, as opposed to a narrow separation of signal carriers. On Galileo satellite E08 and applying a five-order polynomial fitting, it can be seen that the residual plot after the earthquake event shows little or no striking perturbation. Interestingly though is the disturbance seen min prior to the earthquake. The debate of preseismic TEC changes has injected excitement into this research area ( [12] [20] [13]) and remains a contested topic, however this particular avenue of discussion is beyond the scope of this report. The observations for E14 show almost no noticeable TEC perturbation and so are deemed as negligible. For plot E24, the reader should be aware that the vertical grey-dashed lined representing the earthquake epoch is located almost on the y-axis itself. This is because for this particular data set, the first tracked epochs coincide with the minutes just before the actual earthquake itself. At 4 min following the earthquake a very slight magnitude of 0.3 TECU can be seen, which compared to a background absolute TEC of 29 TECU gives 1.0% observed variation. Lastly, for satellite E26 a slight TEC perturbation is seen at 15 min following the earthquake. A magnitude of 0.3 TECU can be seen, and compared to a background absolute TEC of 20 TECU gives 1.5% observed variation. 21

22 (a) (b) (c) (d) Figure 14: VTEC and residual plots at KRSA on 19 August The plots are from the perspective of the GNSS receiver at KRSA, for four Galileo satellites (a) E08; (b) E14; (c) E24; (d) E26. The y-axes and x-axes in all plots do not conform with one another but are adjusted to fit the data. The y-axes for the residual section of each plot is consistent with one another. 22

23 combination (wider phase difference) as opposed to when the TECU is lower. This can be seen albeit slightly from the residual plots in the lower section of each plot. For satellite E08 in Figure 14 the phase difference for the E1-E5 signal combination is MHz and consequently the residuals display a more significant TEC disturbance. Figure 15: Geometry of the Galileo (E08, E14, E24 and E26) satellites projected ground track whereby the IPP is set to 300 km altitude. The orrange lines correspond to tectonic plate boundaries. For Galileo satellite E08, Figure 16 demonstrates a comparison of two further signal combinations used for analysing TEC perturbations. The left plot in Figure 16 is for the frequency band combination E5a- E5b, and the plot on the right for frequency band combination E5b-E5. The phase difference between E5a-E5b equates to 30.7 MHz, whereas the phase difference between E5b-E5 is 15.3 MHz. Therefore in order to facilitate an analysis of TEC perturbation, when comparing these two signal combinations, the E5a-E5b signal combination is favored over the E5b- E5 combination due to a greater phase difference. This is further augmented by Figure 16 whereby the background absolute TEC range for the E5a- E5b signal combination is 5.1 TECU compared to 3.4 TECU for the E5b-E5 combination. This implies that any perturbation in the TEC value seen in the residuals is more profound for the E5a-E5b 23

24 (a) (b) Figure 16: VTEC and residual plots at KRSA on 19 August Both plots are from the perspective of the GNSS receiver at KRSA, for the same Galileo E08 satellite, however for the signal combinations (a) L5-L7 (frequency band E5a-E5b); (b) L7-L8 (frequency band E5b-E5). Figure 17: VTEC and residual plots for KRSA on 19th August 2016 for GLONASS R03. 24

25 Further TEC perturbations can be seen for GLONASS R03 at 12 min (see Figure 17) following the earthquake. Note the zigzag-like effect on the observations for up to 14 min after the initial disturbance, comprising what appears to be five very small waves, each with 3 min in wavelength. However these disturbances are so slight that the resulting residuals appear almost uninterrupted, despite the TEC perturbations being more apparent in the absolute TEC observations (VTEC). This indicates that even when the amplitude of the residuals is minimal with the resulting plot indicating a negligible effect, sometimes the nature of the absolute TEC observations may indicate a TEC perturbation characteristic of a seismic active signature. Like that of GLONASS R03, the GPS G10 satellite observations shows TEC perturbations at 10 min after the earthquake for both KEPA and KRSA (see Figure 18), however like that of R03 (Figure 17) the disturbance on the ionosphere is more apparent on the absolute TEC observations as opposed to the residual plots. The plots shown in Figure 18 are shown on the same scale. The elevation angles for both pairwise combinations (KEPA-G10 and KRSA- G10) are identical because of the close proximity of the two stations with one another. Therefore when considering the possible culprits for the difference in maximim VTEC ( 3 TECU) between the two plots in Figure 18, the satellite-receiver geometry can be excluded. A more likely culprit for this discrepancy could be the receiver DCBs, which for this report has not been individually assessed for each station. Since this observation in TEC perturbation is seen at two independent GNSS receiver sites for the same satellite, this augments the notion that the earthquake in the South Georgia Island Region on 19 August 2016 was detected in the ionosphere via the TEC perturbations, albeit not very strikingly. 3.4 Kaikoura Earthquake The recent Kaikoura earthquake that struck 13 November 2016 at 11:02 UT was a magnitude of 7.8 M W. The epicenter was approximately 15 km northeast of Culverden and 60 km south-west of Kaikoura. Several sites were used to analyse the effects of (a) (b) Figure 18: VTEC and residual plots at KEPA and KRSA on 19 August The plots are for one GPS satellite (a) G10 as seen from KEPA; (b) G10 as seen from KRSA. Both the y and x axes are identically scaled. 25

26 the ionospheric disturbances caused as a result of this earthquake and shall be discussed within this subsection. The Galileo satellites were recorded at 30 second samples at several sites in the New Zealand region. The absolute VTEC values as seen in Figure 20 have been calculated without compensating for the Galileo DCBs. In any case it is the residuals that are of greater importance for this analysis and so for now the DCBs can be neglected. At HOIK a TEC perturbation can be seen for both E08 and E22. The more profound of the two satellites, E22, shows a variation of 1.8% at 20 minutes after the earthquake. For KAIK it is E08 that shows a significant TEC disturbance at 20 minutes after the earthquake, with a residual of 0.6 TECU. However for this particular plot (plot (c)), the residuals were already indicating an unstable behavior beginning even before the earthquake event.consequently it is E22 (plot (d)) that provides a clearer impression of TEC perturbation. The residuals of E22 are very stable throughout the two hours with the exception of 20 minutes after the earthquake where a slight TEC disturbance of 0.3 TECU can be seen. For WEST (plot (e)) no significant TEc disturbance can be seen. Lastly LKTA (plot (f)) shows a lot of similarity with that of plot (d), whereby a TEC disturbance for E22 can be seen 20 minutes after the earthquake. This plot shows a marginally more significant TEC disturbance than that of plot (d). 26

27 Figure 19: Location of the sites under analysis and the epicenter of the earthquake, as given by the beachball (focal mechanism). The orange line represents the tectonic plate boundary between the Australian Plate (above plate) and the Pacific Plate (below plate). The blue lines denote the geometry of the Galileo (E08 and E22) satellites projected ground track whereby the IPP is set to 300 km altitude. The red stars along these blue lines is the projected location of the satellite at the moment the earthquake occurs. 27