Seismic Signature of the Super Cyclone Amphan in Bay of Bengal Using Coastal Observatories Operating Under National Seismological Network of India

We examined the seismic noise data collected from coastal and inland observatories in India, affected by the super cyclonic storm Amphan in the Indian Ocean, to understand the storm dynamics. Prominent disturbances in the 0.05–0.50 Hz frequency range were observed at the seismic stations, arising due to ocean‐continent interactions. The coastal stations displayed more pronounced ground motions contrary to the inland stations, with spindle‐shaped seismic wave envelopes intensifying as Amphan approached. The maximum ground displacements and energy occurred hours after the cyclone's eye, with maximum wind speed, moved away from the stations and not when it was close to the station. We observed significant variations in primary (0.05–0.10 Hz) and secondary microseism (0.10–0.50 Hz) energy during Amphan's directional changes. Secondary microseisms in short and long periods were found at 0.20–0.50 Hz and 0.10–0.20 Hz, respectively. Primary microseisms exhibited a simple pattern and were the weakest among the three energy bands. The CAL seismic station's seismic wave envelope showed an en‐echelon feature with increasing amplitude as Amphan approached, indicating the influence of ocean resonance and coastal wave reflection. This study demonstrates monitoring of the tropical cyclone paths based on seismic signatures obtained using microseisms recorded at seismic stations, a cost‐effective tool. Integrating these seismic signals with atmospheric observations in near real‐time would probably enable an effective monitoring of cyclones and timely issuance of their alerts.


Introduction
The super cyclonic storm (SuCS) Amphan (pronounced as UM-PUN) was a destructive and severely damaging tropical cyclone that originated in the Bay of Bengal in May 2020 (Ahammed & Pandey, 2021;Chatterjee & Roy, 2021;Haldar et al., 2023;IMD, 2020;Mishra & Vanganuru, 2020).The entire track length of the cyclone was ∼1,765 km, which persisted for more than five and a half days (IMD, 2021).This cyclone caused severe widespread damage along the eastern Indian coast, resulting in a massive loss of ∼$13 billion (Chatterjee & Roy, 2021;Haldar et al., 2023).It may be noted that the last super cyclone that hit the Odisha coastal region occurred in the Bay of Bengal about two decades ago, in 1999, which killed more than 10,000 people and caused huge economic devastation in the area (Thomalla & Schmuck, 2004).A large-scale dynamic of the oceanic phenomena usually drives the energy into the earth in the form of seismic waves and modifies the seismic background vibrations significantly in the near coast regions (e.g., Adimah & Padhy, 2020;Chi et al., 2010;Gaebler & Ceranna, 2021;Gerstoft et al., 2006;Gualtieri et al., 2018;Pandey et al., 2020;Retailleau & Gualtieri, 2021).
The Indian Ocean experienced the devastating impact of SuCS Amphan from 16 May to 22 May 2020.The Amphan marked a historical event, becoming the first super cyclone with extremely high wind speeds exceeding 200 km/hr in the Bay of Bengal (Ahmed et al., 2021).However, in the northern Indian Ocean, the Amphan is the third SuCS after the Kyarr in 2019 and the Gonu in 2007, after the 1999 SuCS (NASA's Earth Observatory, 2020;Down To Earth, 2020).On 20 May 2020, at 12.00 UTC, the Amphan made landfall near the Sundarbans, between the Digha coast (India) and Hatiya Island (Bangladesh).It caused the deaths of 84 human lives (72 persons in India and 12 Bangladesh), widespread devastation, and huge economic loss (Ahmed et al., 2021;Kumar et al., 2021).Amphan's immense power ripped roofs from houses, uprooted trees, and generated storm surges of up to 4.6 m in areas like Digha, West Bengal (Bloch, 2020).Within Kolkata in west Bengal, about 15,600 trees were snapped or uprooted due to gusting with continuous strong winds, blocking numerous roads and causing structural damage to properties.Electricity and water supplies were disrupted across eastern India and Bangladesh for days to weeks.The SuCS left a trail of destruction affecting different areas, including agricultural land (78.2%), inland water bodies (9.3%), forests (7.0%), mangroves (3.8%), and built-up areas (1.4%) across the eastern India and Bangladesh (Devi et al., 2021;Kumar et al., 2021).While the severely affected zone (within 2 km of the cyclonic path) was relatively smaller (2,033 km 2 ) as compared to the highly affected (5-10 km; 10,268 km 2 ) and moderately affected (10-20 km; 21,160 km 2 ) regions, the devastation was most concentrated in areas close to the Amphan's path.This SuCS also impacted the Sundarbans, the world's largest mangrove forest.Although the Amphan cyclone affected ∼36.7 km 2 of mangrove forests and suffered severe to very high levels of damage, the strong structure of these plants helped to reduce the cyclonic and tsunami effects and weaken storm surges/tsunami during landfall (Jaiswal et al., 2009;Patel et al., 2016).A significant decrease (around 30%) in aerosol optical depth observed in April-May 2020 compared to 2019 is believed to have formed a warm pool and subsequent cyclogenesis in the Bay of Bengal.The post-cyclone period (29 May 2020) also noticed a concerning 70% increase in COVID-19 cases compared to the pre-cyclone period (19 May 2020), highlighting the added burden such disasters place (Devi et al., 2021).To mitigate such enormous losses, a timely understating of such a disastreaous phenomenon and its accurate tracking, well in advance, before the landfall has become an essential requirement.The fast-moving seismic signals (5-8 km/sec) caused by such a strong system (SuCS) in the midocean are quickly recorded at the coastal seismic stations.Conspicuously, in addition to monitoring the meteorological parameters and sea surface temperature, etc., the analysis of seismic background noise recorded at the coastal seismic stations may significantly mitigate the disaster due to such super cyclonic storms.
The ambient seismic ground vibration, also called seismic background noise (SBN) of the solid earth is originated mainly due to oceanic-driven winds such as tropical cyclones, extratropical cyclones, and other stormy phenomena (e.g., Brwomirski, 2001;Chi et al., 2010;Díaz et al., 2023;Gualtieri al., 2018;Hua et al., 2023;Retailleau & Gualtieri, 2021;Schulte-Pelkum et al., 2004;Webb, 1992) and it can be recorded globally using seismographs.The accepted mechanisms responsible for the generation of SBNs are (a) direct coupling between oceanic waves in shallower water and that near the seafloor in the 10-20 s period (Webb, 2002) and (b) nonlinear interaction of two sets of opposite moving oceanic gravity waves having similar frequency content that generate <10 s microseism (Ekström & Ekström, 2005;Longuet-Higgins, 1950;Rhie & Romanowicz, 2004).A tropical cyclone is found to be one of the most powerful storms capable of producing generally seismic noise of secondary microseisms type, whose sources are much localized (Gualtieri et al., 2018;Retailleau & Gualteri, 2021;Singh et al., 2020).
Several researchers have carried out different seismological analyses of SBN to characterize cyclone generated oceanic as well as seismic waves for their effective monitoring (e.g., Borzi et al., 2022;Chi et al., 2010;Diaz et al., 2023;Fan et al., 2019;Gerstoft et al., 2006;Gualteri et al., 2018;Hua et al., 2023;Retailleau & Gualtieri, 2019;Sufri et al., 2014;Tabulevich, 1971;Zhang et al., 2010).They applied a wide range of methods to monitor oceanic storms with seismological waveforms, such as single station (Sufri et al., 2014), array analysis (Borzi et al., 2022;Hua et al., 2023), and seismic interferometry (Retailleau et al., 2017).However, no one has yet to monitor the tropical cyclones formed in the Indian Ocean seismically.Spectral characteristics of seismic noise indicate significant amplitude signals in a narrow specific period associated with a moving cyclone in the ocean while passing close to the onshore seismic station.Decays in cyclone energy after landfall and during its movement over the land area have been inferred from the seismic signals (Chi et al., 2010;Gualtieri et al., 2018;Sufri et al., 2014).Gualtieri et al. (2018) suggested an adaptative statistical model, which demonstrated the relationship between the spectral amplitude of the short-period secondary microseism and the tropical cyclone intensity in the northwest Pacific Ocean.In addition to background noise, the body waves or multi-phase seismic sources were also found to be associated with cyclones, and it was successfully demonstrated by a few researchers (Gualtieri et al., 2018;Retailleau & Gualtieri, 2021).It is now believed that the seismological tools could be better used to track the path of the cyclone juxtaposed oceanic wave models with several limitations, including a resolution that is found to be too coarse for tropical cyclones producing aliasing, underdetermining the wind speeds and oceanic wave heights, and computing facilities etc (Chi et al., 2010;Fan et al., 2019;Retailleau & Gualtieri, 2019;Sufri et al., 2014;Zhang et al., 2010).However, the relationship between the seismic signals and characteristics of a cyclonic storm is still under debate due to its complexity in terms of frequency-dependent energy transfer during the process of ocean and earth interaction (Ardhuin et al., 2011;Hasselmann, 1963;Janssen, 2004;Ochi, 2003).In this study, an effort has been made to demonstrate the signature of Amphan SuCS in seismic data, and further, it is used to monitor the movement of Amphan in the Bay of Bengal.It is mentioned that the study of cyclonic characteristics using ambient seismic noise becomes significant in exploring the untapped historical seismic data acquired since the early 20th century.

Earth and Space Science
10.1029/2023EA003191

Data Analysis and Methodology
We considered SBN data from seismic stations of the National Seismological Network (NSN) of India, which is maintained and operated by National Centre for Seismology, Ministry of Earth Sciences, New Delhi (Figure 1).These seismic stations are equipped with a 24-bit DM-24 digitizer coupled with three components120s triaxial broadband seismometers and are operating at different sampling rates say 40 samples per second (sps) or 100 sps (Bansal et al., 2021).The installation particulars of seismic stations and the categorization of seismometers are outlined in the research paper by Bansal et al. (2021).The longest period in the waveforms was conspicuously found to be limited to ∼100 s.We mention that out of the currently operated 154 seismic stations by NCS (MoES), many are located in the coastal areas of India (Figure 1).In the present study, we used NSN data of seven seismic stations located close to the coastal regions (MDR, VJD, VIS, RAG, BWN, CAL, and PBA) to understand the seismic signatures of progression of the Amphan SuCS as a function of distance to the coast (Figure 1).In addition, two seismic stations located inland (SHL and BHP), away from the coast, were also considered to understand the impact of the cyclone at farther distance and to compare with the results of the coastal stations.
In the first processing stage, the instrumental response is removed following standard procedures, and the data are prepared for further analysis.In order to understand the energy levels recorded at the different frequency bands, we conducted research on the Power Spectral Density (PSD) for the vertical component recorded at the SHL, VIS, and MDR broadband seismic station within the NCS seismic network (Figure 2).The data are divided into 30 min windows with 50% of overlap.The PSD of each window is computed using the Welch (1967) method.The PSDs, expressed in decibels (dB) relative to ground acceleration, are explicitly showcased for the vertical component.A comparison is made between these observed PSDs and the recently proposed models by Peterson (1993), namely the New High Noise Model (NHNM) and New Low Noise Model (NLNM).Despite recording all seismic ground motion before and after the occurrence of SuSC within both NHNM and NLNM, discernible PSD variations exist.This investigation holds significant importance in verifying the authenticity of recorded seismic ground motion, which is crucial for subsequent studies.The same analysis procedure was also applied to all stations under NSN to characterize the performance of the seismic stations before and during the lockdown period (Pandey et al., 2020).We found that the SBN in long periods (>20 s), primary microseism band (11-20 s), and secondary microseisms (1-10 s) performed well.The noise levels found within the NLNM and NHNM model to determine the strength of the seismic noise energy and authenticity of our data, which may help to address the concern about the potential impact on peak positions, a thorough analysis.Each statistical technique was systematically applied, and we have quantified the variations in peak positions resulting from these methods.In order to provide a comprehensive understanding, we have included a comparison with the raw data in our analysis.The raw data serves as a baseline, and we have similar positions of peaks before and after applying each statistical technique.This allows for a clear assessment of the influence of these methods on peak positioning.The comparative analysis is detailed in (Figures 2-6), where changes in peak positions are presented qualitatively.This approach enhances the transparency of our methodology and facilitates an evaluation of the robustness and reliability of the statistical techniques employed.To capture SuCS generated seismic signals but avoided anthropogenic signals, mainly focusing on oceanic microseisms (which typically manifest below 1 Hz), we use the frequency band of 0.05-0.5 Hz to investigate seismic noise during the storms.
The continuous stream of seismic ground motion data of about 1 month covered the entire duration of pre-, co-, and post-Amphan cyclonic storms were analyzed.The seismic noise data used were ensured free from mass centering pulse, earthquakes, calibration pulse, and anthropogenic disturbances.Usually, the surface waves (Love and Rayleigh) in the 5-120 s bands dominate the seismic noise.Many researchers have suggested short-period and long-period seismic noise to be well recorded on the vertical and horizontal components of the seismic records, respectively (Jana et al., 2017;Webb, 2002).Due to the gravity effect, the seismic noises recorded on the horizontal components are noisier than the vertical components (De Angelis, 2008).We used seismic noise wavefield data recorded on the vertical components in the analysis.Spectrograms were computed at SHL, CAL, BWN, RAG, VIS, and MDR seismic stations, covering the entire track of the SuCS (Figures 1 and 3).The effect of Amphan on seismic noise recorded at CAL station is found to be the strongest (Figure 1) and statistically significant at all periods.Therefore, we considered CAL station for a detailed investigation.
Following the methodology of Lecocq et al. (2020), we computed the PPSD from 30 min windows with 50% overlap so that a single value is gained for each window, calculated using Welch's method (Welch, 1967)  we opted not to employ the default parameters of McNamara and Boaz (2010).Instead, we utilized less smoothing to achieve a finer frequency resolution and more dynamic spectra (Lecocq et al., 2020).The PPSD was computed using the ObsPy implementation (Beyreuther et al., 2010;Krischer et al., 2015;Megies et al., 2011), equivalent to McNamara & Boaz, 2010 methods.SBN data of May 01-31, 2020 (Julian Day 121-152), including the duration of Amphan, was used in the Fourier-transform analysis.We mention that the PPSDs improved the resolution of spectral characteristics, and no artifacts were observed in the spectrum owing to energy leakage among different frequencies.Also, spectral filtering (0.05-0.5 Hz) was applied to enhance the microseisms' signal-to-noise before the spectrograms' estimation.While this method reduces numerical noise in the power spectra, it reduces frequency resolution due to frequency binning.However, this effect is mitigated through robust smoothing parametrization.Subsequently, the windowed segments are transformed into a periodogram using the squared magnitude of the discrete Fourier transform (Lecocq et al. (2020).We followed this procedure for the whole data segment, and spectrograms were generated at each recording site, computing the average Fourier transform's square value using the Power Spectral Density (PSD) approach (Figure 3).
We analyzed diurnal variations in ground displacements nanometer (nm) during pre-, co-, and post-cyclonic periods at seismic stations in the frequency bands 0.05-0.5 Hz (Figure 6).We also presented diurnal  Peterson (1993).The circles on the graph represent SPDF (short period double frequency), LPDF (long period double frequency), and SFM (single frequency microseism or Primary microseism).The peak PSDs in bB are observed during the Amphan Cyclonic storms.The SPDF, LPDF, and SFM fall within the 2-6s, 6-10s, and 10-20s ranges, respectively.The graphs show different colors corresponding to different dates.variations of hourly ground displacements for a better understanding of the impact of the cyclone on the ground motion (Figures 7 and 8).In the analysis, a stable time window of 30 min width was considered to collect the noise samples from the continuous seismic noise waveforms.About 200 noise samples from the pre-, co-, and postcyclonic segments were examined to estimate the average noise spectrum at every station.The PPSD approach was used to evaluate SBN quantitatively based on energy level distribution with frequency at the seismic stations.The noise power density acceleration spectrum, usually known as the spectrum of seismic noise and calculated in dB in terms of 1 (m/s2)2/Hz, was determined at each recording site.PSD is a measure of the strength of the signal as a function of frequency that is utilized to understand the seismic noise pattern at various time intervals and is commonly used to describe broadband random variables.These were computed for each station in a period ranging between 0.1 and 100 s (Figure 2).Further, to calculate the level of seismic noise from the ground displacement, the computed PPSDs in acceleration (dB) were converted into displacement power spectral (Disp pow ) using the relationship proposed by Lecocq et al. (2020).Subsequently, root mean square displacement (d rms ) was computed in the time domain using a bandpass filter to Disp pow in the chosen frequency band by Parseval's identity theorem (Figures 6-8).The median of the d rms curve represents the spectral of SBN displacement variation for each calendar day.We also illustrated day-wise hourly variations in grid and clock maps to understand the differences in ground displacement (nm) during pre-, co-, and post-cyclonic storms (Figures 7 and 8).

Results and Discussion
Seismological stations located near the coastal regions, which are equipped with highly sensitive broadband seismometers, have recorded the signature of the movement of Amphan SuCS.Based on the analysis of the microseism data recorded by nine seismic stations located in the coastal and inland area, a life cycle for the Amphan SuCS was established that persisted for about a week from 13 May to 20 May 2020.However, the cyclone warning system of the India Meteorological Department (IMD) continuously monitored the low-pressure circulation developed in the Bay of Bengal since 23 April 2020 and its intensification with time, about three weeks before the formation on 13 May 2020 (IMD, 2021).In the initial stage, between 16 and 18 May 2020, Amphan moved slower than the average speed; however, a substantial increase was observed subsequently that eventually reached a maximum speed of ∼29 km/hr on 20 May 2021, immediately before the landfall (IMD, 2020).The analysis of the recorded seismic signals demonstrated that the intensification in terms of power, amplitude, and displacement was noticed when SuCS intensified into the depression over the southeast Bay of Bengal in the early morning of 16 May 2020 (Ahmed et al., 2021;IMD, 2020).The conversion depression of SuCS with intensity on 18 May 2020, is illustrated in the displacement shown in Figure 7.A high signal-to-noise ratio is observed at the seismic stations while Amphan moves closely.The CAL station is found to be closest to the SuCS track among all the seismic stations, which is supported by both the recorded waveform and inferred characteristics derived from seismic ground motions (Figure 1).A few stations, namely SHL and BHP that are located away from the coastal region, further inland also recorded the ground motion perturbations significantly around 0.2 Hz due to the movement of SuCS (Figures 2 and 3).Stations SHL and BHP are installed approximately 320 and 720 km from  the coastline, respectively, corresponding to the path of the cyclone's propagation.The seismic data from these stations further aids in comprehending the movement of SuCS over land (Figures 1-3), and the result elucidates the correlation between seismic signals and the energy decay of cyclone upon landfall (Tanimoto & Lamontagne, 2014;Tanimoto & Valovcin, 2015).Lower perturbations of ground vibration at the SHL station may probably be attributed to the diminishing intensity of the SuCS, which is weakening.The coastal stations (MDR, VJD, RAG, VIS, BWN, and CAL) have recorded stronger cyclone signals in comparison to the farther inland stations (SHL and BHP), corroborating well with the observations by Gualtieri et al. (2018).We observed multiple characteristics at different frequency bands in the spectrograms, akin to the results from the global observations (Bromirski et al., 2005;Gualtieri et al., 2018).The energy bands from 0.05 to 0.10 Hz are clearly recognized in the spectrograms.PSD analysis as single-frequency microseism bands, or primary energy from Oceanic waves (Hasselmann, 1963), and 0.10-0.50Hz energy spectrum is considered secondary microseisms (Cessaro, 1994;Longuet-Higgins, 1950;Retailleau & Gualtieri, 2021).Bromirski et al. (2005) further classified double-frequency microseism as short-period (0.20-0.45 Hz) double-frequency and long-period (0.085-0.2Hz) double-frequency energy bands.This study mainly analyzed primary and double-frequency energy bands in short and long periods.The seismic stations located in coastal areas clearly demonstrate the occurrence of both short and long-period double-frequency energy bands.At the same time, SHL and BHP stations slightly away from the coastal belts demonstrate feeble energy between 0.1 and 0.5 Hz (Figures 3-5).We also observed that the maximum PSD of ambient noise propagated.At the same time, the cyclone arrived near the station (Figure 4).Diurnal variations of hourly ground displacement (d rms ) in terms of the nanometer (nm) during pre-co-and post-SuCS is depicted in the frequency between 0.1 and 0.50 Hz (Figure 6).The increase in the median d rms values during cyclonic periods might be representative of short and long period double frequency.
The envelopes of the waves indicate spindle shapes with increased amplitudes when Amphan approaches the seismic stations.A large spectral peak was observed simultaneously at the corresponding frequency (0.05-0.10 Hz) bands (Figure 3), suggesting the microseism energy might have been generated from the interaction of the storm with the coastline.The more intense microseisms during May 17-21, 2021 (Figures 6-8), around a period of ∼5 s (i.e., ∼0.2 Hz), indicate the approaching SuCS toward the respective seismic station.Typically, the maximum displacement amplitude was not associated with the time when the eye of the cyclone was closest to the station and had maximum wind speed; instead, it was recorded several hours after the cyclone eye moved away from the stations (Figures 6 and 7).The seismic noise levels in the frequency range of 0.1-0.5 Hz at stations along the coast can capture the cyclone track.
The SuCS Amphan track was monitored successfully by IMD using a 3-D INSAT geostationary satellite, 3-DR polar-orbiting satellites including SCATSAT (Scatterometer Satellite) and ASCAT (Advanced Scatterometer), and observations by ships and buoys deployed in the ocean.On 20 May 2020, the buoys data indicated a maximum wave height of 3.41 m, a minimum air pressure of 973 hPa, and a minimum wind speed of 3 m/s (Figure 9).Interestingly, these observations align with comparable variations in seismic ground motion attributes across all seismic stations, as illustrated in Figures 3-8.A clockwise re-curve in the track was observed; initially, it moved toward NNW till 17 May 2020 and subsequently re-curved to the NNE direction (Figure 1).Such track change of the cyclone is found to be well captured by coastal seismic stations.The seismic track of the cyclone exhibits that from 18 May midnight to 20 May 2020, Amphan gradually moved from Madras (MDR) in the south to Kolkata (CAL) and to Shillong (SHL) in the north, which collaborated well with the IMD observations (IMD, 2020).The system attained its maximum intensity on 19 May 2020 (IMD, 2020), which is evident from the observed peak of ground motion spectra in the frequency band 0.10-0.50Hz (Figures 2, 6, and 7).The largest amplitude in the spectrogram also coincided with the sharp change in the direction of Amphan (Figure 3).The SuCS started weakening after 19 May 2020 under unfavorable conditions and low oceanic thermal energy, which is well reflected in the seismic ground motion data.Arrivals of the secondary microseisms also witnessed lower power during the period 20-21 May 2020 (Figure 3).However, the single frequency peaks at 0.2 Hz remained visible almost for the entire period (i.e., 01-31 May 2020) and often showed a dispersive gliding.We observed a straightforward pattern with low amplitude and narrow band energy at single frequency bands (0.05-0.10 Hz).
Diurnal variations of hourly ground displacement in the spectral frequency band 0.10-0.50Hz (Figures 7 and 8) consistently indicate a maximum induced vertical displacement when the eye of Amphan moved away from the nearest seismic station, analogous to the spectrogram observations.Such typical observation may be attributed to wave behavior and its interaction or cyclone movement (Tabulevich, 1971).Cyclonic storms that become less powerful at inland stations may be associated with two possible factors: a considerable distance of seismic stations from the coastline and a relatively large attenuation of seismic energy in the eastern Indian Ocean (Singh, 1990).
Seismic records at SHL and BHP stations, which are located on the continent away from the coast, are weakly influenced by Amphan.The oceanic-continental boundary is believed to facilitate reduced energy in short-period ambient seismic noise (Gualtieri et al., 2015).A continuous characteristic change in ground motion due to Amphan is observed at seismic stations close to the coast and inland in the primary microseisms range (i.e., 0.05-0.10Hz).Such continuous features may be associated with generating similar seismic sources of primary microseisms on the continental shelf or due to propagation effects (Ardhuin & Herbers, 2013;Hasselmann, 1963).Bromirski et al. (2005) suggested the short period (0.10-0.50 Hz) secondary microseism frequency band usually originated locally on the seafloor by wind sea waves; however, the long period double-frequency signal (0.10-0.20 Hz) is excited by the ocean waves impacting distantly located coastline.In the present study, seismic ground motion data was significantly influenced by the distant cyclonic storm and its movement approaching seismic recording stations.The near-source wavefield indicated a clear signal with two similar energy bands in the secondary microseisms frequency range.We observed that the seismic ground motion was strongly excited in the high-frequency range (short period double frequency) due to the shallow water region close to the coast.However, the amplitudes of this excitation decayed with distance and were observed feebly at far inland stations SHL and BHP.Such decayed in the amplitudes of the seismic waves may be attributed to both ground motion generated in shallow coastal waters and the attenuation properties of the medium.
As Amphan approached the nearby seismic stations, the frequency bands with maximum energy content shifted to a relatively lower frequency (≤0.2 Hz), which later attained a higher frequency range.At the same time, the SuCS moved away from the respective station and formed a spindle-shaped spectrogram (Figures 2-5).Further, the front wall of the eye of the SuCS in the ocean resulted in decreased frequency at coastal stations, and the level of energy was subsequently found to reduce while the eye moved toward the seismic station.The cyclone provided typical ground motion features when it approached the coast and during a change of direction due to a re-curve (Figures 3-8).The inland far-field station RAG recorded energy in a high-frequency band till the SuCS re-curved NNE ward, probably due to directivity effects.The envelopes of the single-frequency bands have echelon features with increasing amplitude while Amphan moved toward CAL seismic station, wherein maximum amplitude signals increased from the lower frequency (0.04 Hz) to the higher frequency (0.08 Hz) and again moved to the lower frequency band (Figure 3).These significant changes in frequencies band were noticed while moving cyclones in the narrow region.The large spectral amplitude at the CAL station may be associated with the high wind speed (185-232 km/hr) of the SuCS and resonance due to the presence of the continent on either side of the cyclonic wave.The relationship between significant changes in direction and the amplitude in the spectrograms is supported by clear temporal correlations between the observed spectral amplitudes of microseisms during Cyclone Amphan's approach and the nearby seafloor patterns.For instance, Figures 3 and 4 show that the spectral amplitudes of both primary and secondary microseisms increased after 18 May 2020, coinciding with Cyclone Amphan's approach toward CAL.The peak spectral amplitudes occurred on 20 May 2020, followed by a Earth and Space Science 10.1029/2023EA003191 SINGH ET AL. subsequent decrease.These heightened spectral amplitudes align with the increased intensity and proximity of the cyclone to the CAL station.The proximity resulted in the superposition of ocean waves with waves reflected from shorelines, as shown in the satellite images in Figure 10.As a cyclone moves toward the CAL station on the east bank of the Hooghly River, the waves it generates continuously interact with the seafloor, producing microseisms continuously.Additionally, changes in the cyclone's direction could alter the characteristics of the waves and their interaction with the seafloor, causing variations in the intensity of the microseisms.The geological features of the seafloor and coastlines near the CAL seismic stations could influence the resonance and amplification of the microseisms.We also hypothesize that the CAL station is located near specific geological structures that enhance seismic energy transmission, contributing to microseisms' delectability.
We utilized Moored Buoy National Institute of Ocean Technology (NIOT) data set to examine the correlation between our seismic data.The wave height (m), air pressure (hPa), and wind current (m/s) in the buoy data revealed comparable oceanic conditions in the Bay of Bengal during the Amphan Cyclone period.As Amphan traversed the regions, the wave height at the CB06 station surged by 3.41 m on 20 May 2020.Simultaneously, the air pressure dropped to 995 hPa at both stations, and the wind speed decreased on the same day.Our seismic stations recorded maximum activity when Amphan neared the CB06 station.Notably, both the dominant frequencies and the timing of frequency shifts in spectral displacement and PSD changes in the seismic band correlated with ocean wave periods derived from buoy data and the nearby BD11 deep-sea buoy (see Figure 9).We compared the moment of shifting of frequencies with the track of Amphan and found such a shift in the frequency band, both while the cyclone was moving along the eastern coast and when it re-curved.At the time of Amphan's landfall at Digha (West Bengal) and its crossing over CAL (West Bengal) station, the intensified microseisms signals disappeared.Once Amphan approached the coastal belt, intensified microseisms appeared at the CAL site.On 17 May 2020, the cyclone started, and the peak frequency changed from 0.07 to 0.01 Hz (Figures 2 and 4).Beginning on 21 May 2020, new microseisms started developing after Amphan turned and circulated gradually to the northeast (Figures 1, 4, and 7).During this cyclonic period, several oceanic disturbances could occur more than thousands of kilometers away from the affected area; therefore, detecting their effects quickly without analyzing specified data is impossible.For instance, fewer or no microseism motions were observed at the MDR station while the Amphan cyclone was propagating toward CAL, indicating the energy level from other remote cyclonic storms was weak compared to the signals generated by the Amphan cyclone.Additional supportive meteorological, geomorphological, and geophysical data is needed to understand the causes of diminished strong singles while propagating cyclones.The seismic signature of Amphan using coastal seismic stations is validated by IMD Scatterometer Satellite and Indian National Satellite System (INSAT) data (Figure 10).It is also to be mentioned that the current standard methods, such as satellites, numerical modeling, and buoys, are far better for precisely monitoring cyclones.This paper emphasizes that seismic waveform may be a useful component in addition to other tools when studying or monitoring cyclones.
The study clearly reveals that broadband seismometers operating in the coastal region can record the ground motion changes due to the progression of cyclones in the ocean and are effectively used for the analysis.However, the augmentation of more seismic stations along the coastal region and strategically located ocean-bottom seismometers would enhance the monitoring capability of such atmospheric phenomena.Here are recommendations for establishing a seismological network dedicated to cyclone monitoring.It is crucial to explore advanced seismological technologies, such as broadband seismometers capable of capturing a broader range of frequencies, including those associated with cyclone-induced seismic activity.Additionally, integrating cutting-edge sensors with high sensitivity and low noise is essential to enhance the network's detection capabilities.The distributed sensor network should be optimized to ensure comprehensive coverage in cyclone-prone regions, guaranteeing adequate density in vulnerable areas.This involves strategically placing ground-based and ocean-bottom seismometers to capture seismic activity on land and beneath the ocean surface.Additionally, implementing real-time data processing and analysis is imperative for swift detection and response to cyclone-related seismic signals.We also believe that incorporating machine learning algorithms for automated event detection and classification reduces reliance on manual analysis, thereby improving the overall efficiency of the monitoring system.Implement regular maintenance schedules and continuous monitoring to ensure the optimal performance of the seismological network.Establish protocols for the timely replacement or repair of equipment to prevent downtime during critical periods.By considering these technological and distribution-related recommendations, developing a seismological network for monitoring cyclones can be more effective, reliable, and responsive to the dynamic nature of cyclonic events.

Conclusions
The seismic signature of Amphan SuCS track was analyzed using background noise waveforms recorded at seismic stations operating in the coastal region under the National Seismological Network of India.This study analyzed ground motion records, considering the seismic energy source to be a supercyclone developed in the Bay of Bengal.We found significant primary (0.05-0.10 Hz) and secondary microseisms (0.10-0.50 Hz) generated due to Amphan SuCS.The coastal stations recorded significant seismic ground motions compared to inland stations due to ground motion generated in shallow coastal waters and the attenuation properties of the medium.The envelopes of the seismic waves were found to be in spindle shape with increasing amplitude while Amphan approached the stations.However, the maximum displacement amplitude is typically observed a few hours after the eye of the cyclone moves away from the station.The highest wind speed associated with Amphan barely increased the displacement amplitude, while the super cyclone's eye was close to the station.The secondary microseisms frequency has been reported, and they are classified as short-period (0.20-0.50 Hz) and long-period (0.10-0.20 Hz) double microseisms frequency bands.The envelopes in the single frequency band have an echelon shape with increasing amplitude as Amphan approaches the CAL station.Signals increased from the lower frequency to the higher frequency and jumped back to the lower frequency band.These significant changes in frequency bands were noticed many times at the CAL station.The progression and characteristics of the SuCS Amphan along the eastern coast of India are well corroborated with the IMD observations, INSAT, and buoys data.The spindle shape of seismic waves and an echelon shape in single frequency with increasing amplitude conspicuously indicate the SuCS approaching the seismic station near the coast.It is emphasized that analyzing seismic data recorded at the coastal seismic stations vis-à-vis monitoring of meteorological parameters, sea surface temperature, etc. would lead to a holistic understanding of cyclone genesis and its movement until the landfall.

Figure 1 .
Figure 1.Map shows the bathymetry and topography data with seismic stations and the Aphan Cyclonic storm track.Triangles represent the seismic stations utilized in this study with station codes as follows: SHL (Shillong), CAL (Kolkata), BWN (Bhubaneswar), RAG (Rayagada), VIS (Visakhapatnam), VJD (Vijayawada) MDR (Madras), BHP (Bhopal), and PBA (Port Blair).The Amphan traveled along the eastern coast of India, motivating seismic ground motion analysis.The red curve line shows the track of the cyclonic storm with the date.The seismic waveform recorded during a cyclonic storm at seismic stations is also shown on the map.The bottom of the map includes an elevation scale for topography and bathymetry.

Figure 2 .
Figure 2. Illustrates the Power Spectral Density (PSD) of the vertical component recorded at the SHL, VISK, and MDR broadband stations of the NCS seismic network.The PSDs, expressed in decibels (dB) relative to ground acceleration, are presented for the vertical component.These observed PSDs are compared with the recently proposed High Noise Model (HNM) and Low Noise Model (LNM) byPeterson (1993).The circles on the graph represent SPDF (short period double frequency), LPDF (long period double frequency), and SFM (single frequency microseism or Primary microseism).The peak PSDs in bB are observed during the Amphan Cyclonic storms.The SPDF, LPDF, and SFM fall within the 2-6s, 6-10s, and 10-20s ranges, respectively.The graphs show different colors corresponding to different dates.

Figure 3 .
Figure 3. Analyzed spectrograms for the pre-, co-, and post-Amphan super cyclonic storm ground motion from 30 April 2020, to 31 May 2020, at the seismic stations MDR, VIS, RAG, BWN, CAL, and SHL.Arrows on the spectrograms indicate the arrival time of the Amphan's closest approach to each station.Cyclonic Energies are recorded in multiple frequency bands.SFM, LPDF, and SPDF microseisms were detected in the frequency bands of 0.05-0.10Hz, 0.10-0.20 Hz, and 0.20-0.50Hz, respectively.The light orange, black, and white dotted lines indicate respective SFM, LPDF, and SPDF ranges.LPDF microseisms are mainly generated at shallow water depths near the coast, while local winds mainly cause SPDF microseisms.

Figure 5 .
Figure 5.The PSD [(1 (m/s 2 ) 2 /Hz)] of ambient seismic noise during cyclonic storms.The maximum PSD was shifted when the cyclone arrived near the station.

Figure 6 .
Figure 6.Daily fluctuations in spectral displacement within the frequency range of 0.2-0.5 Hz in terms of d rms curves pro-, co-, and post-super cyclonic storms from 1 May 2020, to 31 May 2020.The red solid line indicates the start of cyclonic storms, while the dotted line represents the ending of the cyclone.

Figure 7 .
Figure 7. Grid maps show the daily changes in hourly ground displacement (measured in nanometers) before, during, and after cyclonic storms at SHL, CAL, BWN, RGD, VIS, MDR, BHP, and PBA within the frequency bands of 0.2-0.5 Hz.The analysis of seismic waveform data covers the period from 1 May 2020, to 31 May 2020.

Figure 8 .
Figure 8.The clock map shows the daily fluctuations in hourly found displacement (measured in nanometers) before, during, and after cyclonic storms at seismic stations CAL, BWN, SHL, RAG, VIS, MDR, BHP, and PBA within the frequency bands of 0.2-0.5 Hz stations.The seismic waveforms from 1 May 2020, to 31 May 2020, are used in this study.The displacement scale is shown.

Figure 9 .
Figure 9. Data on wave height (m), air pressure (hPa), and wind speed (m/s) at the CB06 station are collected.Air pressure (hPa) data is also obtained from the BD11 station.The information is sourced from INCOIS.The CB06 station is situated close to the MDR seismic station.