ACCEPTED ON ANNALS OF GEOPHYSICS, 62, 2019; Monitoring the largest North Korean nuclear explosion 2017, through Indian Seismological Network

Seismological characteristics of the North Korean largest nuclear test of September 2017 have been examined using the data ofthe Indian Seismological Network. Full waveform modelling of the ground motion data of Indian stations for this nuclear test shows16% isotropic component, 47.5 % DC and 35.8 % CLVD components. The Indian stations being located about 3500 to 5000 km away from the source, gave lesser isotropic component as compared to that from the nearby stations around the North Korean test site. This is attributed to the rapid attenuation of the high frequencies emitted from the source. Its average body wave magnitude, mb from the Indian stations broadly agrees with that reported by worldwide data. It was found that the surface wave magnitude of this test in North Korea was large as compared to those from the Kazakistan and Nevada nuclear tests for almost similar mb. It is hypothesized that more powerful fusion process in the nuclear test could result in larger tectonic slip.

been discussed by Prakash et al [2018]. The decomposition in ISOLA for the inversion process namely, volumetric or isotropic (ISO); compensated linear vector dipole (CLVD) and Double Couple (DC) is based on the methods given by Vavrycuk [2001] and Benetatos et al [2013]. Thus the principal component in the resultant moment tensor indicates the source property namely explosion, an implosion or a double couple.
In the present analysis, digital data from the broadband Indian stations [ Figure 3] was used to model the source parameters of the recent event using ak135 velocity model. The signal to noise ratio was checked to define the proper frequency band for the inversion. The details of constraining the depth of the source below the USGS epicentral location of this event are similar to those described by Prakash et al [2018] with the filter frequency band of 0.03 to 0.1 Hz and cosine tapering. Moment tensor inversion was then carried out using the data of selected Indian seismological stations as shown in

DISCRIMINATION TECHNIQUES
Methods to discriminate the nuclear explosions from earthquakes have been generally based on the mb-Ms differences. Of late some other methods like complexity versus spectral ratio, and the differences in the third moment frequency contents (TMF) have also been employed for this purpose.
Methods that will be used in this paper are discussed below.

mb VERSUS Ms
Underground nuclear explosions generally release tectonic stress near the site of detonation.
Love waves are generated by the tectonic component alone while Rayleigh waves are generated in both the cases whether earthquake or explosion. Thus, the surface wave magnitude becomes an important discriminatory criterion for earthquakes and nuclear explosions. Since Himalayan tectonics due to the fractured lithosphere caused by multiple collisions of the Indian and Eurasian plates lies in the path of seismic waves from the North Korean test site to the Indian subcontinent, it is of interest to work out the relations between mb and Ms for the earthquakes of magnitude 6 or more as well as all the earthquakes recorded by the USGS catalogue for the period 1961 to 2017 in the Himalayan region [24 -40 0 N and 70 -98 0 E] and then compare with those of the nuclear tests. Using screening relationship (1), the earthquakes of magnitude 6 and above [ Figure 6a] shows that these earthquakes clearly fall above the screening line.
However if all the earthquake data that extend to lower magnitudes was taken [ Figure 6b], then some earthquakes would found to cross the screening line and located in the nuclear explosion zone.
This suggests that the screening line cannot be extended to the lower magnitudes. Also the mb versus Ms plot of the 2017 largest nuclear test lies close to screening line.
The mb and Ms values for the 2013 nuclear test ranged from 4.6 to 5.3 and 3.14 to 4.2 with their mean values as 4.9 and 3.7 at the Keskin SP array (BRTR) station respectively [Semin et al., 2013] and were below the provisional screening line (2) at each array site given by Selby et al. 2012.
The magnitude mb given by USGS was slightly larger as 5.1 for North Korea event.
The variation of mb versus Ms for the largest nuclear test, 2017 at the Indian stations are shown in Fig. 6a and 6b. The mean values of mb and Ms from the Indian data were 6.37 and 5.2 respectively.
The moment magnitude of this test from the Indian stations was found as 5.8 [ Table 4], while much lower value of 5.24 was reported from the close by stations [Wang et al., 2018].

COMPLEXITY (C) TEST
Seismograms of nuclear explosions are much simpler as compared to earthquakes due to generation of mainly compressional waves while they appear more complex due to P as well as S waves in the later. Based on the larger fraction of the total energy in the initial part of the seismograms in the case of nuclear explosions and larger energy centred in the later portion of the seismograms in the case of earthquakes [Gaber et al., 2017] discrimination methods were developed using spectral complexity, waveforms and their amplitudes.
Complexity (C) is defined as the reverse ratio between the energy content within the first five seconds (t1) of the P-waves to the energy content in the following thirty seconds(t2) [Kelly 1968;Gaber et al., 2017]. C parameter was computed as follows [Kelly,1968]. The following equation of Kelly (1968) was used in this study to calculate the C parameter which resamples complexity Where s (t) refers to the signal amplitude as function of time (t) and C is known as the ratio of integrated powers of the vertical component of the velocity seismogram S 2 (t) in the selected time windows length [t0, t1 and t2] where t0 is the onset time of P-wave, [t0-t1] and [t1-t2] are the first-and second-time windows. C value was estimated in a time window [t0-t1: 2~5 sec and t1-t2:25-35 sec].

SPECTRAL RATIO (SR)
The complexity in the frequency domain is expressed by Spectral Ratio (SR) parameter. It is defined as the ratio of integrated spectral amplitudes a(f) of the seismogram in the chosen frequency bands (high-frequency band h1, h2 and low-frequency bands l1 and l2) [Gaber et al., 2017]. It is computed from the following relation [Gitterman and Shapira, 1993] Where h1and h2 represent the high-frequency band while l1 and l2 are the low-frequency bands. The best discriminating bands for integration limits are based on the spectra of explosion and earthquake after testing a number of frequency bands. In the present study, we used eight stations which were in the epicentral distances of less than 3500 km for both explosion and earthquake for computing the complexity and spectral ratio parameters for each station. For the calculation of SR, we selected the values for the filters [l1-l2]: 0.7-1.3 Hz, [h1-h2]: 1.5-2.1 Hz which performed well. The results of our analysis for both C and SR parameters are given in Table 5. Complexity versus the spectral ratio of each station was plotted for the same station in case of 2017 North Korean explosion and the 2014 Bay of Bengal earthquake [ Figure 7].

THIRD MOMENT OF FREQUENCY (TMF)
Third Moment of Frequency (TMF) is defined as [Weichert 1971], Where f is expressed in Hz. Since higher weightage is given on the high frequency components in this method, explosions usually give large TMF values.
Various discrimination techniques as given above were used to distinguish the earthquake of May 21, 2014 Bay of Bengal earthquake from the nuclear explosion of September 9, 2017 in North Korea by the help of data collected from the Indian stations. The broadband data was converted into short period using SAC tool and then the instrument response was removed from each station. P-wave Spectra was also calculated from the signals recorded at northeast Indian seismic stations located at distances ranging from about 3500 to 5500 for both North Korea explosion and Bay of Bengal earthquake. The complexity (C), spectral ratio (SR) and Third Moment of frequency (TMF) method was applied to the seismic signals recorded from both the earthquake and the explosion. The relations for the complexity (C) and the spectral ratio (SR) of the 2017 explosion in North Korea and the 2014 earthquake in the Bay of Bengal are shown in Figure 7 and 8 and Table 5. The results of different methods of discrimination showed that the September 9, 2017 event in North Korea is an underground nuclear test.

YIELD OF NORTH KOREAN 2017 TEST
As is well known, the correspondence between the seismic magnitude and explosive yield of an underground nuclear test is associated with a very large uncertainty. This is because of the lack of detailed knowledge about basement rock structure, the depth of the test and attenuation characteristics of the medium between the test site and the recording station of the yield.
Some empirical relations between mb and the yield of the nuclear test are as follows Ringdahl et al. 1992 for North America and Eurasia: where Y is in kilotons Murphy (1996) determined the above relation based on the spectra of teleseismic P-waves from the Soviet tests and relative estimates of the attenuation parameters for the western U.S and Central Asia.
It gives 0.5 magnitude more in Semipalatinsk as compared to NTS shots of the same yield.
Murphy (1981) for Nevada region: For shots in hard rocks or below water table, it was reported that the test in dry alluvium at NTS may have magnitude 1 unit lower. Vergino and Mensing (1990) gave the relation: Where Ar varies between 3.76 and 3.87 for different areas within NTS. This result was suggested for any type of rock including dry alluvium. Adushkin and Vadim (1993) reported that the measurements from Borovoye station in Kazakhistan can be used to estimate the yields of US explosions to about 20% uncertainty. Khalturin et al. (1998) suggested that the magnitude relation between the teleseismic magnitude and yield is given by Where b=1 and Y is greater than 1  The above relations are discussed later from the Indian stations data.

DISCUSSION
The data of close by seismological stations around the 2017 largest nuclear test was used to study its source characteristics by several workers. Liu et al (2018) Wang et al. (2018) found its moment as 9.5 x 10 16 Nm giving Mw as 5.24 and 50 to 90 % positive isotropic component and relatively small CLVD or double couple contributions [ Table 3]. In contrast, the full wave form modelling results from the Indian stations located about 3500 to 5500 km away showed the isotropic component as 16.7 % while the DC and CLVD components were 47.5 and 35.8 % respectively. This difference is attributed to the attenuation of the higher frequencies from the source to the Indian stations.
The result of the double couple source mechanism is however similar to that of IRIS report as shown in  [Sykes and Cifuents 1984]. Although the values of mb for the Russian nuclear explosions during 1978 to 1982 ranged from 5.576 to 6.242, their Ms varied from 3.637 to 4.106. Adushkin and Vadim [1993] analysed Nevada nuclear tests at a well calibrated station at Borovoye seismic station and found slight difference between mb [International Seismological Centre, U.K] and that determined from this station. The difference of the results of Kazakhistan and Nevada nuclear test was attributed to differential attenuation of P-waves. The values for the Cannikin [1971], Milrow [1965] and Longshot [1965] in Amchitka islands were 6.8, 6.5 and 6.1 for mb and 5.7, 5.0 and 4.6 for Ms respectively.

Comparison of the smaller North Korean nuclear test [2013] with those of India and Pakistan
[1998] of comparable body wave magnitude close to 5 may be interesting due to the different tectonics and path effects. Roy et al. [1998] found the surface wave magnitude as 3.56 for the Pokhran event of 1998 while the mb and Ms of 4.9 and 3.7 were reported for 2013 North Korean nuclear test. Gupta et al [1999] compared the spectral characteristics of the 1998 Pokhran (India) and Chaghai (Pakistan) nuclear tests and found distinct difference in the energy contents in various frequency ranges. The energy from the Pokhran event was in the frequency range of 3.5 to 6.0 compared to a range of 1 to 3 Hz for the Chaghai explosion showing the influence of the local tectonics. However, this aspect could not be studied for North Korea tests as the Indian stations are located far away. Baruah et al [2016] found that similar magnitude earthquake (about 100 km west of the test site) of 2009 and 1998 Indian nuclear test showed that Pn/Lg and Pn/Sn amplitude ratios of the explosion and the earthquake had distinct differences in the higher frequency window. It was found that the nuclear tests in the North Korea, India and Pakistan and an earthquake [2009]  NORSAR estimated the explosive yield at 120 kilotons TNT corresponding to a magnitude of 5.8. On the other hand, Sykes and Cifuentes [1984] reported the yields of seven nearly identical Soviet nuclear tests close to 150 kilotons which were within the limit set by Threshold Test Ban Treaty. The highest yielding test series by the USA and USSR gave yields of 50 megatons. The largest underground nuclear explosion of magnitude 6.8 by the USA called Cannikin gave yield of 5 megatons. On the other hand, the yield estimates of 58+_10 kilotons were estimated for the Indian nuclear explosions in 1998 while the first event in 1974 gave a yield of 12 kt to 13 kt only [Roy et al., 1999]. The yield of the Indian nuclear explosion was estimated as 50 kt from the surface wave magnitude while it was reported earlier as 10 to 50 kt from body waves [Baruah et al., 2016]. Douglas et al.