beta

You're viewing our new website - find out more

Publication - Research Publication

Unconventional Oil and Gas: Understanding and Monitoring Induced Seismic Activity

Published: 8 Nov 2016
Part of:
Business, industry and innovation
ISBN:
9781786523952

An independent research project into understanding and monitoring induced seismic activity.

93 page PDF

6.3MB

93 page PDF

6.3MB

Contents
Unconventional Oil and Gas: Understanding and Monitoring Induced Seismic Activity
5 Lessons and Experiences from Seismic Monitoring and Mitigation

93 page PDF

6.3MB

5 Lessons and Experiences from Seismic Monitoring and Mitigation

5.1 Overview

A dense network of monitoring stations is essential for reliable detection and discrimination of induced seismic events, and to allay public concern. Models of the detection capability of past and present networks of sensors for detecting earthquakes confirm that existing earthquake catalogues are likely to be incomplete at magnitudes of less than 2.0 ML. A comparison of modelled ground motions for a range of earthquake magnitudes with existing regulatory limits for ground vibrations for quarrying and blasting suggests that earthquakes with magnitudes of 3.0 or less are unlikely to exceed the limits above cosmetic damage may occur except at distances of less than a few kilometres. Smaller earthquakes may also exceed acceptable levels, but again only at small distances of less than a few kilometres.

5.2 Reliable Determination of Earthquake Locations

The proximity of the epicentre of the magnitude 2.3 Blackpool earthquake in 2011 to the site of ongoing hydraulic fracturing operations led to immediate speculation that the earthquake was linked to this. However, the closest seismometer to the site was approximately 75 km away, and although the calculated epicentre was less than 2 km northwest of the drill site, uncertainties in the location, particularly the depth, were large, making it difficult to conclusively link the earthquake with operations at the Preese Hall drill site. Figure 5.1 shows the earthquake location calculated using the NonLinLoc non-linear earthquake location algorithm (Lomax et al., 2009) with 36 phase arrivals from 25 stations. The scatter in the location probability distribution function (red dots), extends approximately 4 km in the horizontal plane and 5 km in the vertical plane.

Figure 5.1. location calculated using NonLinLoc (Lomax et al., 2009) for the earthquake on 1st April 2011. The location was calculated using 36 phase arrivals from 25 stations. The blue stars show the maximum likelihood location. Red dots show the density-scatter in the location probability distribution function. The Preese Hall site is at (x, y) = (0, 0). Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016.

Figure 5.1. location calculated using NonLinLoc (Lomax et al., 2009) for the earthquake on 1st April 2011. The location was calculated using 36 phase arrivals from 25 stations. The blue stars show the maximum likelihood location. Red dots show the density-scatter in the location probability distribution function. The Preese Hall site is at (x, y) = (0, 0). Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016.

In May 2011, the British Geological Survey installed two seismometers close to the Preese Hall site. A further earthquake with a magnitude of 1.5 ML occurred on 27 May 2011, and again this was felt locally. Data from the nearby stations helped to reduce the uncertainty in location estimates (Figure 5.2) providing more conclusive evidence that the earthquakes were linked to the hydraulic fracturing. In addition, a number of other smaller earthquakes were also detected on 26 and 27 May while hydraulic fracturing was ongoing.

Figure 5.2. Location calculated using NonLinLoc (Lomax et al., 2009) for the earthquake on 27 th May 2011. The location was calculated using 18 phase arrivals from 12 stations. The blue star shows the maximum likelihood location. Red dots show the density-scatter in the location probability distribution function. Open triangles show the locations of the two seismometers installed close to the Preese Hall. Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016.

Figure 5.2. Location calculated using NonLinLoc (Lomax et al., 2009) for the earthquake on 27th May 2011. The location was calculated using 18 phase arrivals from 12 stations. The blue star shows the maximum likelihood location. Red dots show the density-scatter in the location probability distribution function. Open triangles show the locations of the two seismometers installed close to the Preese Hall. Earthquake data from the British Geological Survey UK Earthquake Catalogue © NERC 2016.

Figure 3.3 also clearly shows the improvement in detection capability following the installation of additional stations close to the Preese Hall site, with the detection of a number of earthquakes with lower magnitudes that could only be observed on these stations.

This example highlights the importance of an appropriate monitoring network for reliable detection and location of any seismic events before, during and after any operations that may induce seismic activity. It also shows how local monitoring stations are essential to reduce uncertainty and contribute to a robust scientific dataset that may be used to develop strategies that mitigate the incidence of induced seismicity.

5.3 Determination of Background Activity Rates

Given a region of homogeneous seismicity, the value of the rate parameter in any sub-region will scale with relative size of the region. For example, if a region where seismicity is homogeneous and has 1000 earthquakes above a given magnitude each year, a sub-region, whose area is ten times smaller, will have 100 earthquakes above the same magnitude each year. Applying the seismicity rate for Scotland, which suggests there should be eight earthquakes with a magnitude of 2 or above each year, to an area the size of the Midland Valley, there will only be one earthquake with magnitude 2 or above each year. Scaling this to an even smaller area of 100 km 2 then there will be a magnitude 2 earthquake every 150 years. This has important implications for baseline monitoring in small regions, particularly where activity rates are low, since the number of earthquakes above the detection threshold of the network in that region in a given period of time may be very low. Either a longer period of observation or a reduction in the detection threshold will be required to reliably determine seismicity rates.

5.4 Detection Capability

The detection capability of any network of seismic sensors is a complex function of many factors including the distribution, density and characteristics of individual sensors, their local site and noise conditions, as well as processing software and processing strategies. The amplitude of the ground motions caused by any earthquake is a function of both the magnitude of the earthquake and the distance of the earthquake from the recording position. An event may be undetected because it is too small or too distant, so its signal is indistinguishable from the background noise on the sensors. Also, many detection algorithms require the signal from an event to exceed the background noise level by a certain ratio on a number of sensors for an event to be detected. If the density of the sensor network is low, this will only happen for larger events. The detection of small earthquakes thus requires relatively high sensor densities.

We use a simple model for the amplitude of seismic waves as a function of magnitude and distance, combined with a given sensor density and estimates of seismic noise at each sensor location, to calculate the theoretical detection capability of the BGS seismic monitoring network at different points in time. Figure 5.3 shows the variation in the magnitude of earthquakes that would be detected by the network in Central Scotland in: (a) 1970; (b) 1980; (c) 1990; and (d) 2016. A signal in excess of three times the noise level needs to be recorded on at least three sensors for an earthquake to be detected. We assume uniform noise levels at each station based on a UK average model.

There are clear differences in detection capability with time as a result of changing numbers and densities of sensors. In 1970, the network was centred on Edinburgh, then expanded over the next two decades. Detection capability is observed to increase with time from 1970 to 1990 across the area of the scoping study in the Midland Valley as more sensors were installed. Post-1990 the network becomes more uniform but with a wider sensor spacing, which results in some local reduction in detection capability, e.g. around Glasgow. It is important to note that at no time was the network able to reliably detect earthquakes with a magnitude of 0.5 or below, even in areas of relatively high station densities.

Reliable location and magnitude measurement places additional constraint on network design, since measurements at more stations are needed than for detection alone. In addition, location errors depend on the distribution and density of the recording stations. These errors may be large if the station density is insufficient, or if the closest stations are far from the earthquake source. As we show above, large errors can limit the ability to discriminate between induced and natural earthquakes.

In general, the results support our use of a magnitude of completeness of 2.0 for the Scottish earthquake catalogue and for calculation of activity rates and recurrence parameters. However, completeness will also vary as a result of failure of instrumentation and variations on background noise levels at individual sites.

Figure 5.3. Modelled detection capability for the BGS seismic monitoring network (black triangles) in central Scotland at four points in time: (a) 1970; (b) 1980; (c) 1990; and (d) 2016. The contours show the spatial variation in magnitudes that can be detected. Detection requires a signal in excess of three times the background noise to be recorded at three or more stations. The scoping study area is delineated by the grey shaded area.

Figure 5.3. Modelled detection capability for the BGS seismic monitoring network (black triangles) in central Scotland at four points in time: (a) 1970; (b) 1980; (c) 1990; and (d) 2016. The contours show the spatial variation in magnitudes that can be detected. Detection requires a signal in excess of three times the background noise to be recorded at three or more stations. The scoping study area is delineated by the grey shaded area.

5.5 Possible Ground Motions for Small and Moderate Earthquakes

Seismic hazard is often expressed as the probability of a particular level of ground motion being exceeded within a certain period of time. Accurate assessment of seismic hazard requires knowledge of how ground motion relates to the characteristics of an earthquake, how it attenuates with distance, and how it might be affected by the geological conditions at the site of interest. In the assessment of seismic hazard, strong ground motions are commonly estimated using empirical ground motion prediction equations ( GMPEs). Abramhamson et al., (2008) provide a comparison of the recent Next Generation Attenuation ( NGA) models. However, the choice of an appropriate model is often difficult, since in most parts of the world there are insufficient data to produce well-constrained empirical models. In such cases it is now becoming generally accepted that it is more appropriate to use a robust ground motion model derived from a large international data set with the widest possible sampling of the magnitude-distance domain than a local model that may be less well constrained (Douglas, 2007; Bommer et al., 2007). For example, the SHARE [13] project used a number of GMPEs including Chiou and Youngs (2008) and Akkar & Bommer (2010).

An alternative approach is to simulate ground motion using stochastic modelling based on the earthquake source parameters as well as parameters to characterise path and site effects (Boore,1983, 2003). Here, we use this approach and the SMSIM software (Boore, 2005) to explore possible ground motions for small to moderate earthquakes that might occur in Ireland, and compare these with some existing regulations for vibrations from blasting in the UK.

The earthquake source is parameterised by the seismic moment, source spectrum shape and stress drop. The former can be calculated directly from earthquake magnitude (Hanks and Kanamori, 1979). Here, we assume that for small magnitudes local magnitude, M L, is approximately equal to moment magnitude, M W. We use a single corner frequency model for the shape of the source spectrum (Brune, 1970). Stress drop is an important parameter in the dynamics of the rupture process and can have a strong effect on recorded ground motions. However, most earthquakes have stress drops in the range of a few MPa to a few tens of MPa. Here, a fixed stress drop of 3 MPa has been assumed.

Path effects are incorporated using geometrical spreading and anelastic attenuation terms. At short hypocentral distances geometrical spreading is dominated by the body wave term and we use the path attenuation quality factor determined for the UK by Sargeant and Ottemoller (2008). We do not consider either site-specific attenuation or amplification. Note that hypocentral distance is the distance between the earthquakes focus and the observer that includes the effect of the earthquake focal depth. As a result, the greater the focal depth, the greater the hypocentral distance. Earthquakes induced by anthropogenic activities often occur at very shallow depth. For example, earthquakes related to coal mining in the UK often have depths of around 1 km, corresponding to the depth of the mining activity. Similarly, the earthquakes induced by hydraulic fracturing operations at Preese Hall, Blackpool, UK, occurred at depths of approximately 3km, close to the point of fluid injection. This means that despite having generally small magnitudes, such induced earthquakes can often be felt as a result of their proximity to the surface.

Figure 5.4 shows curves (coloured lines) of ground velocity as a function of hypocentral distance calculated for earthquakes with magnitudes of 2.0, 3.0, 4.0 and 5.0. The dashed lines show the limits for acceptable levels of ground vibrations caused by blasting from BS 6472-2 and also the limits for vibrations caused by blasting, above which cosmetic damage could take place ( BS 7385-2). Blasting occurs on a regular basis throughout the British Isles and maximum magnitudes for quarry blasts recorded in the BGS catalogue are around 2.5 M L. The limits specified by BS 6472-2 are 6-10 mm/s during the working day, 2mm/s at night time and 4.5mm/s at other times. BS 7385-2 gives limits of 15 mm/s at 4 Hz, increasing to 20 mm/s at 15 Hz and 50 mm/s at 50 Hz. The limits increase with the frequency of the vibration since high frequency vibrations are less likely to cause damage. In simple terms, the observed frequencies for earthquake ground motions are largely controlled by the magnitude of the earthquakes and the stress drop, although anelastic attenuation and site conditions can also play an important role.

Earthquakes with magnitudes of 4.0 or above may approach the limits above which cosmetic damage may be observed, as specified in BS 7385-2, but generally only at hypocentral distances of less than 10 km. Smaller earthquakes, with magnitudes of 3.0 may also exceed the limits recommended in BS 6472-2, though at even smaller distances of less than a few kilometres. This seems reasonably consistent with observations that the largest mining-induced earthquakes, with magnitudes of around 3.0 M L, caused some superficial damage (Westbrook et al., 1980; Redmayne, 1998) including, minor cracks in plaster and harling.

Given the strong variability of observed ground motions from earthquakes, as well as the influence of factors such as variable stress drops and site conditions, which have not been included in our calculations, the modelled ground motion shown here should be considered as indicative only, rather than encompassing the fully extent of possible ground motions in this magnitude and distance range. However, we do find good general agreement, between our calculations and many observations of small earthquakes in the British Isles.

Figure 5.4: Modelled peak ground velocity (solid coloured lines) plotted as a function of hyocentral distance .The grey dashed lines show the limits for acceptable vibrations from blasting specified in BS 6472-1 and BS 7385-2. The squares and triangles show observed horizontal and vertical ground motions.

Figure 5.4: Modelled peak ground velocity (solid coloured lines) plotted as a function of hyocentral distance .The grey dashed lines show the limits for acceptable vibrations from blasting specified in BS 6472-1 and BS 7385-2. The squares and triangles show observed horizontal and vertical ground motions.


Contact