The Influence of Geomagnetic Storms on Calculating Magnetotelluric Impedance


 Magnetotelluric (MT) field data contain natural electromagnetic signals and artificial noise sources (instrumental, anthropogenic, etc.). Not all available time-series data contain usable information about the electrical conductivity distribution at depth, particularly when the signal-to-noise ratio (SNR) is low. Geomagnetic storms represent temporary disturbances of the Earth's magnetosphere caused by solar wind-shock wave interacts with Earth's magnetic field. The variation of the electromagnetic signal increases dramatically in the presence of a strong geomagnetic storm. Using the data observed during a strong geomagnetic storm may overcome the locale noise and bring a reliable MT impedance at contaminated sites. Three case studies are presented to show the positive effect of geomagnetic storms on MT field data. A more reliable and interpretable impedance calculated from a survey line contaminated by strong noise is obtained using the data observed during a strong geomagnetic storm.

shows three case studies influenced by geomagnetic storms. 78 In the practical MT surveys, we may meet the noisy sites occasionally that can't obtain a 79 reliable impedance by the current method. When we redo the MT surveys at noisy sites, we may 80 acquire MT data during strong geomagnetic storms. Although strong geomagnetic storms do not 81 occur frequently, we could predict strong geomagnetic storms using space weather forecasts and 82 acquire MT data during intense geomagnetic storms. Using the data observed during the intense 83 storm period may bring a reliable result from the site contaminated with continuous noise.  Fig. 1 The geomagnetic intensities along the N-S direction during a storm day and a non-storm 89 day. The black lines denote the non-storm day's data, and the red lines denote the storm day's data. 90 The left is a profile in the time domain, and the right is a profile in the frequency domain. The geomagnetic storm is a temporary disturbance of the Earth's magnetosphere caused by a solar 93 wind shock wave interacts with the Earth's magnetic field. Geomagnetic storms start when the 94 enhanced energy of the solar wind transfer into the magnetosphere. A magnetic storm is seen as a 95 rapid drop in the magnetic field strength at the Earth's surface. Fig.1 shows the X (N-S) 96 component of the geomagnetic field during a storm day and a non-storm day at the Kakioka 97 (KAK) station in Japan. In 1973, the KAK Magnetic Observatory was designated as one of four 98 facilities to calculate the disturbance storm time (Dst) index, representing the strength of the 99 equatorial ring current encircling the Earth. The intensity of the magnetic field observed during 100 the storm day can be almost two orders stronger than during the non-storm day. 101 The disturbance storm time (Dst) index is a negative index of geomagnetic activity used to 102 estimates the averaged change of the horizontal component of the Earth's magnetic field based on 103 measurements from a few magnetometer stations. It is derived from hourly scalings of low-104 latitude horizontal magnetic variation and expressed in nanoteslas. When the Dst index is less 105 than -50 nT, it is categorized as a geomagnetic storm. When the Dst index is less than -100 nT, it 106 is categorized as a strong geomagnetic storm. In this section, we analyzed the geomagnetic storm 107 event statistically by the Dst index. 108 It shows that about 46% of strong geomagnetic events lasted more than 4 hours, and 8% of 117 the strong geomagnetic event lasted more than one day. The longest strong geomagnetic event 118 lasted 55 hours. There are 688 strong geomagnetic storms from 1957 to 2020; one year had about 119 ten strong geomagnetic events, and about five events lasted more than 4 hours on average. 120 Fig. 4 shows the monthly count of strong geomagnetic storms. One hour was recorded as one 121 count in this figure. For example, a 3-hour storm is counted as three storms. The high probability 122 of a strong geomagnetic storm occurred around April and October. 123 Fig. 5 shows the yearly count of geomagnetic storms that occurred in each year. Fig.6 shows 124 the FFT result of the yearly count of storms from 1957 to 2020. There is a 10.7-year peak, which 125 corresponds to the 11-year solar cycle. 126 This section concludes that the geomagnetic storm has a seasonal and 11-year solar cycle. The 127 strong geomagnetic storm doesn't happen frequently and causes significant EM field variations 128 observed on the Earth's surface. More than four channels are observed simultaneously in MT fieldwork; the time series of each 178 channel is divided into N segments, and N spectra can be obtained from these N segments by 179 applying the Fourier transform to each channel. 180 In polar coordinates, the cross-power spectra are expressed as follows: 181 where j denotes the imaginary number unit, i=1,2,…, N; A i and B i are the spectra calculated from 183 the i th segment from the different channel; and φ A i and φ B i denote the phases of A i and B i , 184 respectively. The overline denotes the complex conjugate. 185 The amplitude of the cross-power spectra equal the product of |A i | and |B i |, and the phase equals 186 the phase difference (PD) between A i and B i . 187 The auto-power spectra are calculated as follows: 188 (3) 189 The PD is calculated as follows: 190 where θ i denotes the angle of the PD between the two spectra at a specific frequency. 192 The linear coherency is proposed as the cosine of the PD as follows: 193 where Lcoh denotes the linear coherency and Re denotes the real part of the complex number. 195 The value of Lcoh lies in the range of (-1,1). When the PD is close to 0°, the Lcoh value is high 196 and close to 1. According to Euler's formula, Lcoh is also equal to the real part of e j(φ A i −φ B i ) .
If there is a remote site available, for the north-south direction, the linear coherency between 198 the remote and local magnetic fields (RLcoh) is defined as follows: 199 where H x_i and H xr_i are the local and remote magnetic field spectra at a specific frequency 201 calculated from the i th segment. 202 The field MT data include natural EM signals and noise coming from the local environment. 203 We can rewrite the magnetic field H as follows: 204 When the signal-to-noise ratio (SNR) is high at both local and remote sites, the PD between 212 the local and remote magnetic fields should be close to 0°, and the RLcoh value should be close 213 to one. The amplitude ratio (AR) between the local and remote magnetic fields (R_AR) is 214 calculated as follows: 215 In contrast, in the presence of strong noise, the PD between the local and remote magnetic 218 fields will be scattered; therefore, the RLcoh will be unstable; and the R_AR value will deviate 219 from one. 220 RLcoh and R_AR are parameters to measure the similarity between the remote and local magnetic fields. If there is a quiet remote reference site, we could use RLcoh and R_AR to 222 evaluate the variation of SNR change with time at the local site. 223 224

Polarization Directions 225
Weckmann et al. (2005) showed the effectiveness of using the polarization directions to 226 estimate the background noise. The polarization directions for the electric field (α E ) and magnetic 227 field (α H ) (Fowler et al., 1967) at a specific frequency are defined as: 228 (10) 230 We can rewrite the polarization directions as follows: 231 where A i and B i are H x_i and H y_i or E x_i and E y_i , respectively. The polarization direction is 233 related to the PD and amplitude ratio (AR) between the two orthogonal fields. A variety of 234 sources generate natural magnetic signals. These sources generate magnetic fields that vary in 235 their incident directions. The PD and amplitude ratio between the two orthogonal magnetic fields 236 vary with time; thus, there is no preferred polarization direction for the magnetic field. However, 237 according to a given conductivity distribution in the subsurface, a preferred polarization direction 238 may exist for the induced electric field (Weckmann et al., 2005).

Ordinary Coherency 241
The coherency is a quantitative measure of the phase difference (PD) consistency between the 242 two channels. If two channels are coherent, their phases must be either the same or have a 243 constant difference (Marple and Marino, 2004). Coherency is defined as the ratio between cross-244 power spectra density and the root of auto powers spectra density. For A and B spectrum at a 245 specific frequency, it is defined as: 246 where the brackets represent the averages of N individual auto power spectra and cross-power 248 spectra. For instance, 249 (13) 250 251

CASE STUDIES 252
Three case studies are shown to evaluate the influence of geomagnetic storms on the MT data. 253  storm days in the three case studies. We used the moving median filter to smooth the spectra. The 258 magnetic coils are used to observe the magnetic field at Sawauchi station, and we need to 259 calibrate to the spectrum. The fluxgate magnetometer is used in the USArray and KAP03 project, 260 and the calibration factor is 1. Because we have not calibrated the spectrum observed at the 261 Sawauchi station, its intensity is smaller than that observed in the USArray and KAP03 projects. 262 During the storm day, the intensity is approximately five times stronger than that measured during 263 the non-storm days between 10 and 1000 seconds at Sawauchi and USArray project. Moreover, 264 the intensity is approximately 50 times stronger than that during non-storm days between 10 and 265 1000 seconds in KAP03. 266 Table 1 shows the name of each result and the corresponding data used to calculate the 267 impedance in studies 2 and 3. The Quiet parameter was calculated using the data observed during 268 the non-storm period, and QuietRR was calculated using the data observed during the non-storm 269 period and using the remote reference technique. The Storm parameter was calculated using the 270 data observed during the storm. StormRR was calculated using the data observed during the 271 storm period and using the remote reference technique. The period shows the month and day of 272 the data. For example, 06.20-06.22 means the time from June 20 00:00:00 to June 22 00:00:00.  Sawauchi station, Japan. The geomagnetic storm occurred on August 26. The MT time-series data 291 were stored in three files. Two files sampled the high-and middle-frequency bands (2,400 and 292 150 Hz) intermittently; the other files continuously sampled the low-frequency data (15 Hz). The 293 high-frequency band (2,400 Hz) was sampled for 1 second at intervals of 4 minutes from the 294 beginning of the minute, and the middle-frequency band (150 Hz) was sampled for 16 seconds at 295 intervals of 4 minutes from the beginning of the minute. 296 First, we analyzed the spectrum variation along with the Dst index. To obtain precise spectral 297 information from these datasets, we first applied a set of Slepian tapers and then used the fast 298 does not change correlated with the geomagnetic storm. 306 We calculated the impedance using each day's data. Fig. 10 shows typical MT sounding curves 307 and the coherency distribution using the data observed during the storm day (August 26) and non-308 storm day (August 23). The sounding curves calculated using the storm data was more stable than 309 the result using the non-storm data between 300 and 1,000 seconds in the Z xy and Z yy 310 components. The sounding curves of Z xx and Z yx are almost the same. In this result, the phases of there was no increase correlated with the storm. This result agrees that the interaction between the 341 solar wind and the magnetosphere does not contribute to the MT high-frequency signal. The signal strength at periods larger than 4 seconds increased dramatically along with the 343 geomagnetic storm. Because the natural EM signal strength between the dead band (0.1-10 344 seconds) is low, and local noise can easily influence it. The enhancement of the natural EM signal 345 may produce a more reliable impedance result. Next, we will investigate the change in impedance 346 value during storm and non-storm days at 10 seconds. 347 In the second case study, long-period 5-component MT time-series data observed at two sites 393 (ALW48 and TNV48) were used. The data sets were recorded with a 1-second sampling period 394 for around two weeks in 2015 from the USArray project. The geomagnetic storm occurred 395 between June 22 and June 24. 396 Fig.15 shows the distribution of coherency in different periods and cross-power spectra at 16-second during the storm and non-storm days. The ordinary coherency increased from 4 to 40-398 second and 400 to 2,000-second during the geomagnetic storm. The low coherency during the 399 non-storm day may be attributed to the local random noise. We can see the signal strength 400 increased dramatically from the distribution of cross-power spectra. The preferred direction of PD 401 between the orthogonal electric and magnetic field becomes more obvious at 16-second. 402 Fig. 15 compared four results calculated using the data observed at site TNV48 and using 403 ALW48 as the remote reference site. The apparent resistivity of Quiet in the period from 8 to 30-404 second is severely down-biased. And the phase of Quiet is scattered from 8 to 30-second and 400 405 to 2,000-second. The result calculated using the storm data is much stable than the result 406 calculated using the non-storm data. After comparing all results, the StormRR is the most reliable, 407 and we regard it as the true model here. The Storm result is closer to the true model than the 408 Quiet result between 4 to 30-second. We can see from the case study that the signal strength 409 increased during the geomagnetic storms, and a more reliable impedance is obtained using the 410 storm data.  Coh(E x , H y ) and Coh(E y , H x ), increased and were close to one across all periods. The preferred 446 direction of the phase difference between the orthogonal electric and magnetic fields is almost the 447 same at 84 seconds. 448  phase calculated by non-storm data is close to 0°, and the apparent resistivity increases as a line 500 on the log scale. That is the phenomenon of local noise (Zonge and Hughes, 1987). 180° or 0° 501 would correspond to a dipole electric source, which could be the train line. The impedance 502 changed using geomagnetic storm data. This result coincides with the preferred direction of PD 503 changed at 84 seconds in Fig. 22. The QuietRR result calculated using seven days of data (see Table 1) coincides with the Storm result but is slightly different in the XY component between 20 505 and 40 seconds. The remote reference technique can only reduce the influence of local noise. 506 From Fig. 8, the signal strength during this storm is almost 50 times stronger than that during the 507 non-storm days. The noise can be neglected in this condition. We believe that the Storm result is 508 more reliable. In this section, three parameters (polarization direction, RLcoh and R_AR) are used to analyze 518 the data observed at site 133. Fig. 24 shows the variation in the polarization direction at 84 519 seconds from October 26 to October 31. The magnetic field polarization has a preferred direction 520 at approximately -30° during non-storm days (October 26 to October 29) and becomes scattered 521 during geomagnetic storm days (October 29 to October 31). On the other hand, the electric field 522 polarization direction is scattered during non-storm days and has a preferred direction of 523 approximately 60° during geomagnetic storms. The polarization direction is a function of the 524 amplitude ratio and PD. The local EM noise source usually has a constant location; the incident 525 direction and the energy exhibit similar properties over time. Contrary to the natural EM signal, 526 the incident direction and power change with time. If there is a preferred polarization direction for 527 the magnetic field, we can consider that the local environment contaminates the data in that 528 period. That coincides with the high Coh(E x , H y ) and Coh(E y , H x ) and the preferred direction of 529 PD is close to 0° and -180° during the non-storm period. The data are dominated by coherent 530 noise during non-storm days. 531 When the portion of the natural magnetic signal (H MT and H r MT ) is high in the local and remote 537 sites; the PD will be close to 0°; therefore, RLcoh should be close to 1, and R_AR should be 538 stable and close to 1. Because the natural signal is weak and easily influenced by local noise 539 during non-storm days, RLcoh is scattered and low; R_AR is scattered and high during non-storm 540 days. The natural magnetic signal portion increased drastically during the geomagnetic storm, the 541 variation in RLcoh and R_AR became stable. This result indicates that the SNR is low during 542 non-storm days and becomes high during storm days. non-storm data are close to 0°, and the apparent resistivity increases as a line on the log scale 557 between 10 and 200 seconds. A similar situation occurs at site 133. We consider that the data are 558 dominated by strong coherent noise during non-storm days. The Coh(E x , H x ) value is high while 559 the Coh(E x , H y ) value is low during storm days; this can be interpreted as the phenomenon of 560 PROQ. The QuietRR result using four-day data (see Table 1) coincides with the Storm result; 561 moreover, the Storm result is smoother, and the error bar is smaller than that of the QuietRR 562 result.  Fig. 27 shows the MT sounding curve and coherency distribution using the data observed 572 during the storm and non-storm days at site 136. Coh(E x , H y ) is relatively high between 10 and 573 1000 seconds during the non-storm data; Fig. 28 shows the distribution of cross-power spectra of 574 the ExHx and ExHy components at 168 seconds during the storm and non-storm days. The 575 preferred direction of PD between Ex and Hy is close to 0°. We consider that the strong coherent 576 noise caused this phenomenon. 577 On the other hand, Coh(E x , H x ) is high, while Coh(E x , H y ) is low during the storm day. That can 578 be explained as the phenomenon of PROQ. The QuietRR result using four days of data (see Table  579 1) partially coincides with the Storm result. Moreover, the Storm result is smoother, and the error 580 bar is smaller.  Fig. 29 and Fig. 30 show the MT sounding curve and coherency distribution using the data 592 observed during the storm and non-storm days at sites 139 and 145, respectively. Both the 593 coherency between the orthogonal electric and magnetic fields increased during the storm days. The result calculated by the data observed on the storm day is smoother; the XY component of 595 the QuietRR result has a similar trend to the Storm result. However, the YX component is very 596 different between the QuietRR and Storm results at both sites. It is difficult to distinguish which 597 represents the real conditions. From the perspective of SNR and based on the analysis in the 598 previous case study, the storm has a positive effect on the MT data quality; we believe that the 599 Storm result is more reliable.

DISCUSSION 610
In this section, we discuss how to use multiple parameters to estimate the data quality. Coherency 611 is an important parameter to discuss the data quality. However, the characteristic of coherency is 612 different in different situations. At first, we discuss the relationship between impedance and 613 coherency. According to least-squares theory (Sims et al., 1971); Z xy can be calculated as follows: 614 Because various sources generate natural magnetic signals, they generate magnetic fields that 624 vary in their incident directions, which means Hx and Hy are not coherent, and Coh(H x , H y ) is a 625 small value. In the condition that the Coh(E x , H y ) is relatively high while the Coh(E x , H x ) is small. 626 The numerator of Eq. 14 will be dominant by the C part. The denominator is dominant by the E 627

part. 628
The Z xy can be rewritten as follows: 629 In this situation, Z xy is determined by the orthogonal component of the electric and magnetic field. 631 A similar analysis to Z xx , Z xx is undeterminable. When Coh(E x , H y ) is relatively high while 632 Coh(E x , H x ) is small; the field data can be explained as the 1-D and 2-D cases. Here we also need 633 to quantify the coherency value in the different geological environments by doing some 634 simulation. For example, rotate the observation axes in the 2-D case by the step of 5°, how high 635 the coherency will be. We can see the example at TNV 48 from USArray, site 142 from KAP03. 636 The coherency between the orthogonal magnetic and electric components is relatively low during 637 the non-storm day and increased dramatically during the strong storm. The low coherency can be 638 attributed to the incoherent noise in this case. 639 On the contrary that the coherency between the orthogonal component Coh(E x , H y ) is relatively 640 low while the Coh( E x , H x ) is high. The Z xy is undeterminable and Z xx is determinable. The 641 phenomenon of PROQ appears. In this situation, we cannot explain the data by the 1-D or 2-D 642 case. We can see the example at Sawauchi station, sites 130 and 136 from KAP03. Both site 130 643 and 136 is contaminated by coherent noise, and the Coh(E x , H y ) become low while the Coh(E x , H x ) 644 become relatively high during the storm day. 645 The coherent noise may have a high coherency value and appear as the spike, or convex-like, 646 or other kinds of noise in the time domain at the different channel simultaneously. And the phase 647 difference between the two-channel tends to 0° or 180°. It is better to check the phase by plot the 648 distribution of the cross-power spectra. To estimate the data quality precisely, we would better 649 combine other parameters to discuss the situation. 650 The polarization direction is a function of PD and AR between the two orthogonal fields. The 651 local EM noise source usually has a constant location; the incident direction and the energy have 652 a similar property along with time. Contrary to the natural EM signal, the incident direction and 653 power are changed with time. If there is a preferred polarization direction for the magnetic field, 654 we can consider that the data is contaminated by coherent noise in that period. This situation can 655 be seen at site 133. But sometimes, the data is contaminated by incoherent noise. There is no 656 preferred polarization direction for the magnetic field. This situation appears in site 142 but is not 657 shown in this paper. 658 Suppose there is a quiet remote reference site. We also could use the RLcoh and R_AR to 659 measure the similarity between the local and remote sites to evaluate the influence of noise. This 660 example is shown in the data analysis at site 133. 661 Finally, the most important parameter to discuss the data quilty is the result impedance. The 662 sounding curve should be smooth according to the forwarding modeling. On the other hand, in 663 the influence of strong locale noise, the phase will be close to 0º or 180°, and the apparent 664 resistivity increases as a line in the log scale (Zonge and Hughes, 1987) Finally, we will discuss the source effect and nonstationarity of the data observed during the 671 storm day. At mid-latitudes, geomagnetic pulsations (Pc's) in the Pc3-4 band (~10 -100 s) associated with field-line resonances can violate the fundamental assumption of the MT method 673 over the resistive regions; where skin depths are large (Murphy and Egbert, 2018). In this case, 674 the source effect is inevitable and is place-dependent. In this paper, from the perspective of SNR, 675 we demonstrate the positive effect of a geomagnetic storm on the MT data quality, the impedance 676 calculated using the data observed during the geomagnetic storm and the non-storm day at the 677 quiet site 163 and Sawauchi station coincide well. It shows that the signal holds the plane-wave 678 assumption, and the nonstationarity is not a problem for the method based on the FFT in this area. 679 Otherwise, the result calculated by the storm period data should be biased. The souces effect may 680 be considered near the auroral or equatorial electrojets. But the plane wave assumption is 681 generally acceptable at midlatitudes. 682 683

CONCLUSIONS 684
It is well known that the signal strength will increase during a geomagnetic storm in the MT 685 community. Still, the demonstration that shows the positive effects on the MT impedance by the 686 field data is rare. This paper showed the positive influence of the geomagnetic storm on MT data 687 quality by three case studies in mid-latitude. Using the data observed during a strong geomagnetic 688 storm may overcome the influence of the local noise, depending on the strength of the 689 geomagnetic storm and local noise. We obtained a more reliable and interpretable impedance 690 using the data observed during the strong geomagnetic storm to calculate the impedance in the 691 survey line from Kap03, which is contaminated by the strong noise. 692 MT field data include natural signal sources and noise. Along with urban constructions, 693 artificial disturbances to EM observations are becoming more and more serious. The observation 694 occasionally contains continuous noise, which is difficult to get a reliable result from the current 695 technique. When we redo the MT campaign in the noisy site, we may get a reliable result using 696 the data observed during geomagnetic storms. Sometimes, the variation during storm periods can 697 be 100 times greater than in the non-storm period data. In that condition, the noise can be neglected. However, a strong geomagnetic storm doesn't occur frequently. It is possible to predict 699 the geomagnetic storm by the space weather forecast information. The Space Weather Prediction 700 Center (SWPC; see the website in references) provides information about space weather in the 701 coming three days. Utilizing the data observed during the strong geomagnetic storm may bring a 702 reliable result despite the site contaminated by continuous noise. 703 To get the accurate complex coefficient from the time series, we suggest that it is better to 704 contain at least four times longer than the expected period. For 1,000-second, a time-series 705 segment with 4,000 seconds is needed to get accurate spectra. The overlay rate is 50% to keep 706 each data's independence and get more sample data. By the continuous 4-hour time-series data, 707 we may get about eight samples to do the impedance estimation in the frequency domain by FFT. 708 If there is continuous 4-hour geomagnetic storm data, we may get a relatively reliable tensor until 709 1,000 seconds, depending on the geomagnetic storm's length. The longer the geomagnetic storm 710 last. A more stable result can be obtained. By the statistical analysis of the geomagnetic storm, 711 one year had about ten strong geomagnetic events, and about five events lasted more than 4 hours 712 on average. That is practical and meaningful for MT exploration. The geomagnetic intensities along the N-S direction during a storm day and a non-storm day. The black lines denote the non-storm day's data, and the red lines denote the storm day's data. The left is a pro le in the time domain, and the right is a pro le in the frequency domain.   The monthly count of strong geomagnetic storms based on the Dst index. The yearly count of geomagnetic storms based on the Dst index from 1957 to 2020. The calculated periods by Fourier analysis using the yearly count of geomagnetic storms from 1957 to 2020.

Figure 7
The location map in the three case studies (KAP03, USArray, Sawauchi). The left map shows the detailed site location used in USArray, and the right map shows the survey line of KAP03. Note: The designations employed and the presentation of the material on this map do not imply the expression of any opinion whatsoever on the part of Research Square concerning the legal status of any country, territory, city or area or of its authorities, or concerning the delimitation of its frontiers or boundaries. This map has been provided by the authors. Comparison of the spectrum calculated by the Hx component observed during the storm and non-storm days. The black lines denote the non-storm day's data, and the red lines denote the storm day's data. The horizontal axis denotes the period. The vertical axis denotes the intensity.     The XY component of the impedance curve was calculated by each day's data at a period of 10 seconds. The horizontal axis denotes the date. The upper gures show the apparent resistivity, and the lower gures show the impedance phase. The red lines show the apparent resistivity and phase calculated by the data from August 20 to August 28.

Figure 13
The time variation of the impedance curves calculated using each hour's time-series data at a period of 10 seconds. The horizontal axis denotes the time. One result was calculated using one-hour data. The unit of Hx is nT.

Figure 14
The time-frequency distribution against the Dst index. The sampling rate is 150 Hz. The content is the same as Fig. 9.

Figure 15
The distribution of coherency in different periods and cross-power spectra at 16-second during the storm and non-storm days. The black color denotes the result using the non-storm data, and the red color denotes the result using the storm data.

Figure 16
The MT sounding curves using the data observed during storm day and non-storm day. The Quiet result is drawn in black; the QuietRR result is drawn in blue; the Storm result is drawn in red; the StormRR result is drawn in purple.  The distribution of coherency in different periods and cross-power spectra at 84 seconds during the storm and non-storm days at site 163.

Figure 19
The MT sounding curves calculated using the data observed during the storm and non-storm days at site 163. The triangles denote results calculated by the EMT code; the circles denote the results calculated by the BIRRP.

Figure 20
The distribution of coherency in different periods and cross-power spectra at 84 seconds during the storm and non-storm days at site 142. The red color denotes the result during storm days. The black color denotes the result during non-storm days.

Figure 21
MT sounding curves using the data observed during the storm and non-storm days at site 142. The Storm result is in red. The Quiet result is shown in black. The QuietRR result is shown in blue.

Figure 22
The distribution of coherency across different periods and cross-power spectra at 84 seconds during the storm and non-storm days at site 133. The contents have the same meaning as those in Fig. 20.

Figure 23
MT sounding curves using the data observed during the storm and non-storm days at site 133. The contents have the same meaning as those in Fig. 21.

Figure 24
The variation in polarization direction at 84 seconds using the data observed at site 133 from October 26 to October 31. The upper gure shows the polarization directions for the electric eld, and the lower gure shows the polarization directions for the magnetic eld.

Figure 25
The variation in RLcoh versus R_AR at 84 seconds using the data observed at site 133 from October 26 to October 31. The blue and the red line denotes the RLcoh. Blue indicates a negative value, and red indicates a positive value. The black curve denotes the log value of R_AR.

Figure 26
The MT sounding curves and coherency distributions obtained using the data observed during storm days and non-storm days at site 130. The storm result is shown in red. The quiet result is shown in black.
The QuietRR result is shown in blue. For coherency, the red color denotes the result during storm days. The black color denotes the results obtained during non-storm days.

Figure 27
The MT sounding curves and coherency distribution using the data observed during the storm and nonstorm days at site 136. The colors have the same meanings as those in Fig. 26.

Figure 28
Distribution of cross-power spectra of ExHx and ExHy components at 168 seconds between the storm day and non-storm day at site 136. The colors have the same meanings as those in Fig. 20.

Figure 29
MT sounding curves and coherency distribution using the data observed during the storm and non-storm days at site 139. The colors have the same meanings as those in Fig. 26.

Figure 30
MT sounding curves and coherency distribution using the data observed during the storm and non-storm days at site 145. The colors have the same meanings as those in Fig. 26.