Ambient noise levels in Gujarat State (India) seismic network

Abstract

The Gujarat state seismic network (GSNet), comprising of 50 broadband seismograph (BBS) stations and 40 strong motion accelerographs (SMAs), has been operated and maintained by the Institute of Seismological Research (ISR) since 2006. Nineteen permanent BBS stations are connected through VSAT and the rest are kept in an offline mode. The local geology beneath seismic stations varies from Mesozoic to Quaternary formations. The seismic background noise (SBN) at these stations was calculated and found that it varies widely as a function of period, time and geographic location. We have studied the SBN variation along these three parameters for 14 online BBS stations of the GSNet. It was found that the stations located on the Deccan trap and Mesozoic formations are good sites with low SBN while stations on Quaternary, Tertiary and soft soil are noisy. The comparison between day and night noise spectra shows that SBN increases during the daytime at most of the stations. Three typical noisy peaks at three different periods were recorded at all of the stations, which is a global phenomenon. The results of this study will be helpful in selecting sites for future earthquake observations.

1. Introduction

The Gujarat state, situated in the western part of India, is one of the most active intraplate regions of the world. The region witnessed two large earthquakes, the 1819 Kachchh earthquake (Mw 7.8) and the 2001 Bhuj earthquake (Mw 7.6), over a span of 183 years as well as a few moderate earthquakes (Rajendran and Rajendran 2001). The Gujarat (India) Seismic Network (GSNet) was established by the Government of Gujarat with funding support from the Asian Development Bank (ADB) after the disastrous 2001 Bhuj earthquake. The network covers an area of approximately 196 000 km². The network comprises 50 broadband seismographs (BBSs) and 40 strong motion accelerographs (SMAs) along with a central data receiving centre. The data of 19 BBS stations are brought to the central station through VSAT (Ashara et al.2006, Chopra et al. 2008) and the remaining stations are kept in an offline mode. Themain objective is to record and locate any earthquake of magnitude ≥2.5 within Gujarat state and to report it to the authorities of state concerned within 10 minutes of the arrival of seismic waves.

The seismic stations of GSNet are located on varied geological formations including sandstone, basalt, limestone and hard soil of Mesozoic to Quaternary era. The high sensitivity of BBS made it possible to study the seismic noise of the Earth's surface. Peterson (1993) determined the New Low Noise Model (NLNM) and the New High Noise Model (NHNM) curves representing the lower and upper bounds of a cumulative compilation of representative power spectral densities (PSD) determined for noisy and quiet periods at 75 digital seismic stations worldwide. These so-called Peterson (1993) curves have become the standard by which the levels of noise at seismic sites are evaluated. It has been observed that the noise varies widely as a function of period, time and subsurface geology. A precise evaluation at a particular time, place, and period can only be obtained by field measurements. We estimated the level of seismic background noise (SBN) along these parameters. Many researchers have studied the SBN and its sources in different parts of the world for permanent as well as temporary recording stations. The significance and usefulness of seismic data is increased as the noise level is decreased. In view of this, studying and understanding the seismic noise is important for reducing the noise level.

McNamara and Buland (2004) have carried out noise studies for continental stations of the United States and found that very few stations reach the new low noise level (NLNM) proposed by Peterson (1993) and also observed strong geographical variations for periods less than 1 s. Further it was discovered by them that the stations surrounded by urban areas show high noise levels. Wilson (2002) observed that high frequency (0.3–8.5 Hz) median noise levels are controlled by their proximity to local cultural noise sources. The lowest noise levels are observed at culturally remote sites. De la Torre and Sheehan (2005) determined the influence ofcultural noise in Nepal as well as in Tibet and found that the SBN during the daytime was around 4 dB noisier in horizontal components than the vertical components of the Nepal array but there was hardly any day/night variations noticed for the horizontal components of the Tibet array.

Leon (2001), while studying the RISTRA array of broadband stations extending from west Texas to southwest Utah, observed strong variations in SBN for higher frequencies (1.0–10 Hz) and longer periods (0.003–0.03 Hz) at different stations. The least amount of variation is seen in the 0.07–0.2 Hz frequency range, which contains the microseismic frequencies. Leon (2001) also observed that the sensors placed on loose, unconsolidated filled material, when combined with the high amount of cultural activity, contribute higher than average noise levels at higher frequencies. Powell (1992) found that traffic raises SBN between 0.8–5.0 Hz but no distinct peak could be associated with traffic in north central North Carolina and further observed that for frequencies greater than 4 Hz, a significant portion of noise was produced byLove and Rayleigh waves. De Angelis (2008) suggested that one of the factors contributing to more noise in horizontal components in comparison to the vertical is the effect of local tilt induced by temperature and barometric pressure variations. The SBN study by Vila (1998) in the eastern Pyrenees shows that the noise level of horizontal components is of the same order as the vertical, showing good site conditions as well as good coupling of the sensors. Wilson (2002) carried out this type of study for the temporary seismic stations in the southwestern United States. He observed that during long periods (0.01–0.06 Hz) typical noise levels for shallow broadband sites are controlled by local site conditions and are 10–12 dB (vertical) and 25–35 dB (horizontal) higher than the highest quality site.

De Angelis (2008) observed that in the area around Soufriere Hills Volcano, noise levels at frequencies greater than 1 Hz are low, reflecting the absence of cultural noise. It was also observed by him that the background seismic noise is affected bymeteorological conditions, causing PSD levels to increase significantly in the 2.0–8.0 Hz frequency range. A good seismic record is also affected by the recording system noise and may contribute to the background noise (Rodgers et al.1987, Powell 1992). Although some observed high frequency spikes are shown due to unknown vibrations of the equipment or intrinsic seismometer noise, system noise levels are typically well below other sources of noise. However system noise could be a matter of concern at very quiet sites (Leon 2001).

The diurnal variations in SBN were also studied by many researchers. Given (1990) observed that the noise levels for the IRIS/IDA array in USSR during the daytime is 1–14 dB higher than during night hours for frequencies above 1.0 Hz. Monitoring of ambient noise conditions at a site is used to determine the utility of data recorded at seismic stations (Given 1990, Li et al.1994). Many studies of the Earth's internal structure can be carried out by decreasing the noise level. Better understanding of the seismic background noise is the first step in reducing the noise level of seismic data (Stutzmann et al.2000). A good site for seismic stations requires low SBN and depends upon to what extent we can minimize various noises (Young et al.1996). In the present study SBN analysis at 14 sites of GSNet is carried out for estimating the noise variations with period, time, geographical location and local site conditions.

2. Geology and seismotectonics of Gujarat

The state of Gujarat is divided into three distinct divisions, namely, Kachchh, Saurashtra and Mainland Gujarat.

2.1. Kachchh

The Kachchh region lies in the northwestern part of Gujarat. It comprises two distinct physiographic units, viz, (a) Rann and (b) Mainland Kachchh. The Rann is a dry bed remnant of a sea and remains a saline desert for most of the year. It mainly consists of fine silt and clay. The Mainland Kachchh is an isolated and detached landmass flanked by the Rann in the north. The central part of the Kachchh forms a tableland sloping on all sides having a crescent shape. The Kachchh region is characterized by uplifted highlands and islands surrounded by plains of the Great Rann, Banni and Little Rann. The northern margins of the uplifts are demarcated by major faults, namely the Kachchh Mainland Fault (KMF), Katrol Hill Fault (KHF) and Island Belt Fault (IBF). The sedimentary rocks of the Mesozoic era are the principal rock types. The Mesozoic sediments ‘were uplifted, folded, intruded and covered by Deccan Trap basaltic flows at some places in Late Cretaceous and Early Paleocene time’ (Biswas 1987).

2.2. Saurashtra

The Saurashtra is a peninsula and a horst, surrounded by the fractures related to thethree intersecting rift trends, viz, Delhi (NE–SW), Narmada (ENE–WSW) and Dharwar (NNW–SSE). This region comprises mostly of Mesozoic and Cenozoic rocks. The Mesozoic sediments occupy an area of approximately 5000 km² in the northeastern part of Saurashtra and are mainly of lower cretaceous. The basaltic rocks of Deccan trap cover a large area and are prominently exposed. They are fringed by the Tertiary and Quaternary rocks along the coastal plains (Merh 1995).

2.3. Mainland Gujarat

The Mainland Gujarat is divisible into two well defined sub zones, the eastern rockyhighlands and the western alluvial plains. The eastern rocky highlands are between 300 m and 1100 m high and are believed to be the extensions of the major mountains of western India – the Sahyadari, Satpura and the Aravalli. The western alluvial plains comprise a thick pile of unconsolidated sediments deposited by combinations of fluvial and aeolian agencies mainly during the Quaternary period (Merh 1995).

The seismic stations of GSNet are spread over these three distinct regions and located on sandstone, basalt, limestone and hard soil of Mesozoic to Quaternary age. The locations of the seismic stations with their underlying geology are given in table1 and shown in figure 1.

Table 1. Online broadband seismograph (BBS) seismic stations of GSNet and local geology.

Serial no.	Station	Code	Instrument	Geology	Latitude (°)	Longitude (°)	Region
1	Naliya	NAL	Guralp CMG-3T	Limestone	23 19 63 N	68 49 66 E	Kachchh/Bhuj
2	Badargarh	BAD	Guralp CMG-3T	Sandstone	23 27 96 N	70 37 11 E	Kachchh/Bhuj
3	Chobari	CHO	Guralp CMG-3T	Quaternary	23 30 78 N	70 20 61 E	Kachchh/Bhuj
4	Radhanpur	RAD	Guralp CMG-3T	Quaternary	23 49 19 N	71 37 03 E	North Gujarat
5	Gandhinagar	GND	Guralp CMG-3T	Quaternary	23 12 12 N	72 38.92 E	South Gujarat
6	Jhagadia	JAG	Guralp CMG-3T	Tertiary	21 43 55 N	73 09.08 E	South Gujarat
7	Valsad	VAL	Guralp CMG-3T	Basalt	20 13 02 N	73 26 87 E	South Gujarat
8	Ukai	UKE	Guralp CMG-3T	Basalt	21 13 32 N	73 35 06 E	South Gujarat
9	Kadana	KAD	Guralp CMG-3T	Lunavada (Quartzite)	23 17 22 N	73 51 23 E	South Gujarat
10	Morbi	MOR	Guralp CMG-3T	Basalt	22 50 33 N	70 53 62 E	Saurashtra
11	Rajkot	RAJ	Guralp CMG-3T	Basalt	22 21 41 N	70 45 83 E	Saurashtra
12	Una	UNA	Guralp CMG-3T	Basalt	20 58 65 N	70 55 58 E	Saurashtra
13	Bhavnagar	BHV	Guralp CMG-3T	Basalt	21 41 61 N	70 58 80 E	Saurashtra
14	Surendranagar	SUR	Guralp CMG-3T	Sandstone	22 43 09 N	71 35 09 E	Saurashtra

Figure 1. Geology of the Gujarat showing the permanent online seismic stations used for the study. Available in colour online.

3. Instrumentation and installation

GSNet comprises of 50 seismic stations equipped with CMG 3T 120 s Guralp sensors. Out of 50, 19 are permanent stations equipped with 24 bit DM-24 Guralp digitizer operating at sampling interval of 50 samples per seconds (sps) and the remaining 31 are offline temporary stations equipped with 24 bit Reftek digitizers operating at 100 sps. The data from the permanent stations are brought to the centre stations via VSAT. In order to minimize the noise due to temperature variations, aspecial vault room has been constructed at these permanent stations (figure 2). The vault room is double walled with a false ceiling to control the temperature variations. At stations on loose soils, the upper soil cover has been removed from 5 to 15 feet and a concrete pier has been constructed for placing the sensor. Some of the stations are located in remote places while some are near populated areas prone to cultural noises. The network started functioning fully from January 2007; however some stations started functioning as early as August 2006.

Figure 2. Block diagram showing the design of the vault room in which the installation of CMG-3T has been carried out and the recording room at the Permanent Broadband Seismograph Station. Available in colour online.

4. Data selection and methodology

The continuous data stream of 24 h duration was selected in such a way that there should not be any corruption by local and teleseismic earthquakes, mass centring pulse, calibration pulse and other disturbances (figure 3). Successive 30 minute time windows were selected from 24-h data, making a total of 48 samples (24 for day and 24 for night) for calculating single noise spectra. At times, when continuous data streams free from disturbances were not available, the data window was substituted from next day or previous day data. To study the background noise level, the power spectrum density (PSD) was calculated by the noise module of SEISAN software. In seismic noise studies, the seismic background noise is often displayed as acceleration power spectral density in dB relative to ((1 m/s²)²)/Hz. When estimating noise spectra, no attenuation correction is done and the normalization of the spectrum is

where, P is the Peterson Power spectrum, F^DFT is the discrete Fourier transform, Δtis the sampling interval and T is the length of the time window. The factor 2 comes from the fact that only the positive frequencies are used, so only half the energy is accounted for. The total power is proportional to the length of the time window since the noise is considered stationary, so normalizing by T, the length of the time window should not influence the results. This noise option is a handy method of checking the noise characteristics of a given seismic station and its comparison with global standards (SEISAN version 8.2.1 www.uib.no/rg/geodyn/artikler/2010/02/software – cached).

Figure 3. A sample record of the 1 hour of seismic background noise (SBN) at the permanent seismic station Naliya on the limestone. Available in colour online.

A median for each station is determined that will be representative noise spectra for the particular station. In addition to that, the PSD was separately estimated for day and night hours to see the variations. The daytime was selected as 8:00 am to 8:00 pm local time when transportation and cultural noise is maximum and night-time was selected from 8:00 pm to 8:00 am local time when the transportation and cultural noise is supposed to be minimum.

5. Results and discussion

The noise spectra are determined between 0.1 and 10 Hz and interpreted at three different frequency bands (figure 4): long period (0.01–0.1 Hz), microseism (0.1–1.0 Hz) and high frequency (short period) (1.0–10.0 Hz). The wind turbulence and human activity are the major sources for short period noise whereas oceans and atmospheric perturbations are major sources for microseisms and longer period noise, respectively. Two dominant peaks of microseisms, namely, single frequency and double frequency peaks are found at all the sites at 0.07 Hz and 0.14 Hz, respectively. The double and single peak levels, at all the stations are below NHNM. The level of double frequency peak is around 120 dB at all stations except at Kadana and Surendernagar where the level is around 10 dB lower. The level of single frequency peak is around 150 dB at all the stations except at Valsad and Naliya where the level is around 10 dB lower. The noise level at longer periods (0.01–0.1 Hz) is within the NLNM and NHNM limits at all the stations but is around 10 to 20 dB lower than NLNM and NHNM at Kadana and Rajkot. It has been observed that at long period and microseism bands, all stations perform well and noise levels are within NLNM and NHNM, but the behaviour of stations at a frequency greater than 1 Hz is different and is more influenced by cultural activities and predominant frequencies of the sites. The background noise levels are more at Chobari, Gandhinagar, Jhagdia, Naliya and Radhanpur than other sites. These sites are located either on Quaternary or Tertiary formations. The noise levels at these sites cross the NHNM level between 2 and 5 Hz, due to the fact that the predominant frequency of these sites is between 2 and 5 Hz (Chopra 2010, Chopra et al. 2011). This implies that data from these sites are influenced by cultural activities and predominant frequencies. Powell (1992) and Stutzmann et al. (2000) made similar observations and found that the cultural noise raises spectral levels. The noise levels at sites located on Deccan basalt are comparatively lower than others. Among all the sites Kadana and Una are the quietest.

Figure 4. Comparison of the vertical component median seismic noise of 14 stations with the US Geological Survey high and low noise models of Peterson (1993). The power spectral densities (PSDs) are in units of dB with respect to acceleration (reference level 1(m/s²)²/Hz). The grey lines represent the noise spectra of 30-min time windows and the dark lines represent the medians of these. The station codes are at the top of each figure.

Figure 4. (Continued).

The cultural noise levels are also studied by comparing day and night background noise levels (figure 5). The stations Chobari, Gandhinagar, Jhagdia and Naliya show different day and night noise levels in short period ranges (f > 1 Hz). These stations are located close to towns/villages. In view of this, the effect of human activity is observed at all these stations during the daytime. The maximum difference is at Chobari (∼15 dB) at a frequency range of 2–5 Hz. The stations Chobari and Gandhinagar are on Quaternary deposits whereas Jhagdia and Naliya are on Tertiary formation. It has been observed that the noise levels of Quaternary and Tertiary stations are nearly the same. The station Radhanpur, located on thick Quaternary deposits, does not show a difference in noise levels during day and night hours though it is very near to the town and highway. This may be due to the fact that vehicular traffic noise from the highway is constant all along the day/night. Thestations Morbi, Rajkot, Surendernagar, Ukai, and Valsad are quiet sites and do not have day/night noise level variations. These sites are located far away from the village and highways and the broadband sensors are installed on fresh Deccan trap rocks in a vault. Similarly Kadana site installed on Quartzite is also quiet and does not have day/night noise level variations. Similar diurnal variations in SBN resulting from human activity patterns have been observed by Stuzmann et al. (2000), Given(1990) and Li et al. (1994). The effect is maximum (∼5 dB) at Chobari, Gandhinagar, Morbi and Valsad. Almost all the stations show different day and night noise levels in long period range.

Figure 5. (a) Day and night comparison of the vertical component median seismic noise of 14 stations. The power spectral densities (PSDs) are in units of dB with respect to acceleration (reference level 1 (m/s²)²/Hz. The station codes are at the top of each figure. (b) Day and night comparison of the vertical component median seismic noise of 14 stations. The PSD is in units of dB with respect to acceleration (reference level 1 (m/s²)²/Hz. The station codes are at the top of each figure. Available in colour online. Available in colour online.

Figure 5. (Continued).

The station Jhagdia is near the town, surrounded by trees. It shows high levels of noise, both during the daytime and at night-time. However, it has been observed that, at frequencies greater than 2 Hz, the level of noise is decreased by 10 to 12 dB during the night hours. The trees may act as interlocutors for noise due to wind energy that may be transmitted to the ground, and may be the reason for the presence of high noise levels around the clock. However, the decrease of 10 to 12 dB at frequencies greater than 2 Hz at night may be due to the fact that cultural activities are absent in the night. A strong correlation has been observed between wind and SBN levels by Young et al. (1996) and Vila (1998). The absence of objects that couple wind energy to the ground (trees, manmade structures) is an important factor for selecting a quiet site. In these cases, placing a seismometer at a deeper depth will greatly reduce the wind effects (Withers et al.1996, Young et al.1996).

6. Conclusion

This paper presents the seismic background noise levels of 14 broadband stations of GSNet established in the Gujarat state. These sites have been chosen by considering the local geology, topography and nearby sources of noise, such as traffic and vegetation. The PSDs of SBN are determined in the frequency range 0.01–10 Hz using seismic data free from earthquakes and instrument calibration/mass centring pulses. It has been observed that the stations established on soft soils such as Chobari, Gandhinagar and Radhanpur are noisier. Also the stations Naliya and Jhagdia, which are on Tertiary formation, have the same noise levels as those of stations on soft soil. The day and night noise levels are also studied for these stations. It has been observed that Chobari, Gandhinagar, Jhagdia and Naliya are showing variations in noise levels during day and night hours in the long period (0.01–0.1 Hz) and short period (1–10 Hz) ranges. The day and night noise level variation in the short period range is maximum at Chobari (∼15 dB). The effect of long period noise sources from atmospheric perturbations can been seen in Chobari, Gandhinagar, Morbi, Rajkot and Valsad, where the noise levels in the long period range are 2 to 5dB lower at night. The study will be useful for selecting sites for future seismic stations elsewhere.

Acknowledgment

Authors are thankful to the Director General, ISR for giving permission to publish the study. We are also thankful to the technical team of ISR for helping to run the GSNet network and for providing the good quality data for this study. The authors are grateful to the reviewers for their comments/suggestions that have helped to improve the earlier version of the manuscript.