See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/268454964

Modeling detection thresholds of microseismic

monitoring networks

Conference Paper September 2012

DOI: 10.1190/segam2012-1069.1

CITATIONS READS

2 29

4 authors, including:

Alex Goertz Nima Riahi

Petroleum Geo-Services University of California, San Diego
47 PUBLICATIONS 190 CITATIONS 21 PUBLICATIONS 35 CITATIONS

SEE PROFILE SEE PROFILE

Toni Kraft
ETH Zurich
37 PUBLICATIONS 403 CITATIONS

SEE PROFILE

All content following this page was uploaded by Toni Kraft on 14 February 2016.

The user has requested enhancement of the downloaded file.

 Modeling detection thresholds of microseismic monitoring networks
Alex Goertz, Nima Riahi, Toni Kraft (ETH Zurich), Marc Lambert (Spectraseis)
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

Summary features of cultural noise. These estimates can be used to

select the proper bandwidth of recording equipment with
We present a method to estimate the detection threshold of respect to instrument response and sampling rate. The
seismic monitoring arrays, based on the estimated spectral method aims at providing simple yet objective criteria for
amplitude of microseismic events, and the expected noise the design of frac monitoring networks, and defining the
level at the recording station. The aim is to develop most relevant medium parameters that should be known a-
objective criteria for network optimization, such as the best priori in the network design process. For illustration
combination of surface and downhole networks, or the purposes, we present a generic example, loosely based on a
optimal depth of shallow borehole monitoring arrays. We shale play.
present some simplified generic examples for illustration of
the general principles, and discuss the most critical Spectral amplitude of microearthquakes
parameters that should be known beforehand for an
informed decision on network design. We compute the source spectrum using the model of Brune
(1970,1971),

Introduction
 =  
.

  

Different array geometries for monitoring microseismicity
of hydrofrac operations have been the subject of a The low-frequency displacement spectral amplitude 
sometimes heated debate over the past few years (e.g., relates to the seismic moment (Boatwright, 1984) as
Warpinski, 2010, Duncan & Eisner 2010). Surface

monitoring arrays oftentimes have a lower detection  =       ,

threshold owing to greater source-receiver distances and
higher levels of cultural noise. While much closer to the where R denotes the distance of the travel path between
action in a typically quieter environment with source S and receiver R. Letters and denote the density
correspondingly lower detection thresholds, borehole arrays and shear wave velocity, respectively. The factor 
are subject to very constrained geometries due to the depends on the source radiation pattern and the free-surface
prohibitive cost of drilling monitoring wells. amplification. For the sake of simplicity, we assume a
In this paper we present a simple methodology to estimate constant factor of 1.1, based on values given by Deichmann
the detection threshold of a microseismic monitoring array. (2006). We translate the seismic moment to moment
We compare estimates of the anticipated spectral amplitude magnitude using the definition of Hanks & Kanamori
of an earthquake of given moment magnitude with (1979),
"
! = $%&   6.07 , -. /01.
expected ambient noise levels either at the surface or in a
borehole, based on representative broadband recordings of #
surface ambient noise. The statistical description of the The corner frequency 2 scales with magnitude and depends
noise level allows calculating a probability of detecting an on the stress drop . Assuming the source spectra to
event of given magnitude. By calculating the surface wave represent S arrivals, we use (Edwards et al., 2010)
eigenfunctions for a given velocity model, we can estimate
5 #
the frequency-dependent decay of the noise amplitude with 4 =    .
depth. From the latter, we can derive an estimate of either a .6 7
threshold magnitude for a given deployment depth, or a To consider attenuation, we multiply the source spectrum
minimum deployment depth for a given maximum :;
9
magnitude threshold. We also consider the fact that with a factor of 8 < , where t denotes the traveltime
stacking a multitude of densely spaced surface receivers between source and receiver and Q is the average quality
can effectively lower the detection threshold of such arrays. factor of the medium. Finally, we convert the displacement
While we concentrate on the seismic noise for the detection spectrum to velocity by multiplying with 2f, and square it
threshold, the sensitivity of the used recording equipment to obtain a power spectral density (PSD) that can be
has to be considered as well if it is to be expected that the compared to ambient noise spectra. We do not consider the
seismic noise could be lower than the instrument noise. directional cosines between the receiver orientation and the
Aside from the detection threshold, we can also estimate emergence angle of the arrival, which is equivalent to
the anticipated dominant frequency of microseismic events assuming that the full vectorial character of the wave field
and frequency ranges that might be masked by typical is captured with multicomponent receivers.

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 1
 Modeling Detection thresholds of monitoring networks

Estimating the ambient seismic noise level only in the upper 500 m, and plays a role in selecting the
optimal depth of shallow-hole monitoring arrays. At deeper
The ambient seismic wave field has a unique spectral
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

levels, other constituents of the ambient wave field such as

character with some common features that can be observed body waves or tube waves become the dominant part and
worldwide. This is illustrated in Figure 1 that shows the the eigenfunction decay is not applicable anymore.
monthly average noise level at a quiet broadband station in
a vault in Switzerland. The spectral average has been
calculated using the method of McNamara & Buland
(2004). The most prominent feature is the ocean
microseism peak at about 5 seconds, caused by ocean
gravity wave energy coupling into the solid earth. Cultural
noise is emerging at periods below 1 s, where we can
observe a splitting of the distribution into two modes,
reflecting the diurnal variation of human activity. The
instrument response and sampling rate of the sensor
prevents the analysis of frequencies above 50 Hz. White
dashed lines denote the 10th and 90th percentile of the
distribution. We can use these percentiles to estimate the
probability of detecting a microseismic event.

Figure 2: Depth decay of the noise power of fundamental-mode

Rayleigh waves for a layered velocity model.

Shale gas example

To illustrate the above considerations, we use numbers for

a typical North American shale play at a target depth of
8000 ft (2440 m). We assume an average bulk quality
factor of 100 (optimistically), and a shear wave velocity of
2454 m/s at the source. These numbers are loosely based on
values encountered in the Barnett. For the purpose of this
calculation, we assume a stress drop of 1 MPa. Densities
are estimated using Gardners (1974) rule. Our noise
estimates are based on 24-hr averages from broadband
measurements near a frac operation. The noise level is
Figure 1: Monthly average noise level calculated using PQLX for a
quiet broadband station in a Swiss vault. typically much stronger near the wellhead of the treatment
well and decays away from it. Our on-pad example was
To account for the possibility of enhancing the detection recorded about 100 m from a treatment well, and the off-
capability of densely spaced surface arrays with N stations pad example is about 2 km away in a quiet open-range
through stacking, we subtract a factor of $%&  / from the area. To illustrate the influence of different parameters, we
power spectra in some of the examples (/ in amplitude assume three network configurations: (i) a densely spaced
corresponds to N in power). This approach is somewhat surface array of point receivers, (ii) a sparse network of
simplified in the sense that we consider the stacking to be borehole stations in 300 m (1000 ft) deep dedicated
ideal, i.e., well-behaved, or well-known velocity models monitoring wells, and, (iii) a deep borehole array at target
and statics. depth within a few hundred feet of the perforation zone.
Noise decay with depth Figure 3 shows a comparison of the measured ambient
noise distribution with source spectra for different
Ambient noise consists to a large part of surface waves, and magnitudes at the indicated source-receiver distance. Near
decays with depth, depending on the frequency. We the well pad (Figure 3 top), the median noise spectrum is
estimate this depth decay by forward-calculating the mostly below the Mw -0.5 curve, translating to a > 50 %
eigenfunctions of the fundamental-mode Rayleigh wave for probability of detecting an Mw -0.5 event. A Mw 0.0 event
a layered velocity model using a propagator matrix method would be always detected, and a Mw -1.0 event would be
(Aki & Richards, 2002). We notice that over the frequency rarely detected ( 10% probability). Around the perimeter
range of interest for microseismic monitoring (mainly of the array, the noise is lower, with less variance, and the
above 5 Hz), the depth decay of surface waves is relevant

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 2
 Modeling Detection thresholds of monitoring networks

listening distance is greater. This greatly increases the not likely, that in this case the seismic noise drops below
probability of detecting a Mw -0.5 event, and a Mw -1.0 the instrument noise level. We assume an instrument noise
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

event would now be detected with a probability of > 50 %. level of -190 dB, corresponding approximately to state-of-
A Mw -1.5 event would never be detected. In order to the-art borehole sensors. In reality, the noise level at depth
consider the possibility of stacking a dense surface array may be higher due to body- or tube waves (Goertz et al.,
we assume an array of 200 stations and subtract a bulk 23 2011).
dB from the noise power spectra. This translates into a
decrease of the overall detection threshold by almost 1
magnitude unit (Figure 4). In reality, it is to be expected
that the stacking is frequency dependent. While it may
work well at the low frequencies, it will be increasingly
difficult to achieve a SNR improvement at higher
frequencies.

Figure 4: Surface recording away from the pad, assuming a 23 dB

improvement in SNR by a 200-fold stack (blue curves).

Figure 5: Detection threshold at 300 m depth. Noise spectra are

multiplied with the fundamental-mode Rayleigh eigenfunctions to
account for the decay of surface waves with depth. Imposing a
lower limit of -190 dB accounts for instrument noise.
We observe that the strong 10-30 Hz noise band that is
Figure 3: Surface detection threshold close (top) and 2 km away
visible in Figure 3 and 4 is now fully attenuated (Figure 5).
(bottom) from a treatment well. Source spectra in 0.5 magnitude The overall detection threshold did not decrease compared
steps (black to orange) are compared to median and 10th/90th to the 200-fold stack due to the instrument noise limit
percentile of seismic noise observed during a frac (green). Peterson which underscores the importance of high instrument
(1993) low noise model (gray) shown for reference. sensitivity. For the deep downhole array, we assume
Cultural noise consists predominantly of surface waves and receiver levels at distances of 100-500 m to the sources in a
decreases dramatically with depth. We simulate the similar formation. In this environment, we expect the
detection threshold for a sparse shallow borehole network instrument noise floor to be the main factor limiting the
in 1000 ft (300 m) deep monitoring wells by multiplying detection threshold. We calculate the instrument noise floor
the noise spectra with the forward-modeled eigenfunction for a 15 Hz omnidirectional high-temperature phone after
for fundamental-model Rayleigh waves. It is possible, if Rodgers (1992). We observe in Figure 6 that the detection
threshold is now greatly reduced to Mw -3.3 or greater at

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 3
 Modeling Detection thresholds of monitoring networks

close distances (100 m), respectively Mw -2.3 at 500 m

distance. Note the increased frequency content of the
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

spectra, warranting a faster sampling rate. The advantage of

more sensitive equipment (if available) is evident,
particularly at the lower frequencies which are important
for accurate magnitude determination. The increase of the
detection threshold with distance depends on geometrical
spreading and attenuation such that, simplified for constant
velocity,
2
>?@ log  E .
F

Comparison to expected seismicity level

We model a magnitude distribution by drawing a fixed
number of events from a Gutenberg-Richter statistic with a
b-value of 1.5. We equally distribute 10,000 events
between 0 2000 ft from the receiver array, and then apply
the distance-dependent detection threshold criterion as
stated in the above equation. This procedure results in
about 1000 recorded events with magnitudes from Mw < -4
to Mw = -1.2 (Figure 7). This synthetic distribution
corresponds roughly to actually recorded event
distributions in the Barnett (Maxwell et al., 2006). To
summarize our results, we plot the obtained best detection
thresholds for the surface and shallow borehole array in
Figure 7. We note that neither the 200-station surface array
nor the shallow borehole array would have been successful
in this particular case. Note however, that this result should
not be generalized as it strongly depends on the underlying
parameters which vary significantly from play to play and
from project to project. We have used generic values that
appear reasonable, for demonstration purposes only. For
example, surface arrays have been used successfully in the Figure 6: Deep downhole recording at 100 m (top) and 500 m
Barnett (e.g. Lakings et al., 2006). Differences can be due (bottom) distance compared to instrument noise floor (blue).
to different target depths, velocities, or Q values. In
addition, surface or near-surface borehole arrays may offer
better location accuracy which may offset pure detection
threshold considerations.

Conclusions

We have developed a fast and simple forward-modeling

method to assess the capability of microseismic network
configurations to detect a certain magnitude event.
Comparison with the distribution of ambient seismic noise
levels allows an estimation of the detection probability. We
compare modeled detection thresholds with the expected
seismicity rate based on Gutenberg-Richter statistics.

Acknowledgments Figure 7: Forward-modeled seismicity cloud (blue) after applying

detection threshold for deep downhole recording (green) compared
to surface and shallow well array thresholds (cyan & magenta).
Part of this work was co-funded by Spectraseis and the
Low Frequency Seismic Partnership (LFSP). Mid-continent
frac monitoring data examples are courtesy of Spectraseis.

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 4
 http://dx.doi.org/10.1190/segam2012-1069.1

EDITED REFERENCES
Note: This reference list is a copy-edited version of the reference list submitted by the author. Reference lists for the 2012
SEG Technical Program Expanded Abstracts have been copy edited so that references provided with the online metadata for
each paper will achieve a high degree of linking to cited sources that appear on the Web.
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

REFERENCES
Aki, K., and P. G. Richards, 2002, Quantitative seismology, 2nd ed.: University Science Books.
Boatwright, J., 1984, Seismic estimates of stress release: Journal of Geophysical Research, 89, 6961
6968.
Brune, J., 1970, Tectonic stress and the spectra of seismic shear waves from earthquakes: Journal of
Geophysical Research, 75, 49975009.
Brune, J., 1971, Correction to tectonic stress and the spectra of seismic shear waves from earthquakes:
Journal of Geophysical Research, 76, 5002.
Deichmann, N., 2006, Local magnitude, a moment revisited: Bulletin of the Seismological Society of
America, 96, 12671277.
Duncan, P., and L. Eisner, 2010, Reservoir characterization using surface microseismic monitoring:
Geophysics, 75, no. 5, 75A13975A146.
Edwards, B., B. Allmann, D. Faeh, and J. Clinton, 2010, Automatic computation of moment magnitudes
for small earthquakes and the scaling of local to moment magnitude: Geophysical Journal
International, 183, 407420, doi: 10.1111/j.1365-246X.2010.04743.x.
Gardner, G., L. Gardner, and A. Gregory, 1974, Formation velocity and density The diagnostic basis
for stratigraphic traps: Geophysics, 39, 770780.
Goertz, A., K. Cieslik, E. Hauser, G. Watts, S. McCrossin, and P. Zbasnik, 2011, A combined
borehole/surface broadband passive seismic survey over a gas storage field: 81st Annual
International Meeting, SEG, Expanded Abstracts, 14881492.
Hanks, T., and H. Kanamori, 1979, A moment magnitude scale : Journal of Geophysical Research, 84, no.
B5, 23482350.
Lakings, J., P. Duncan, C. Neale, and T. Theiner, 2006, Surface-based microseismic monitoring of a
hydraulic fracture well stimulation in the Barnett Shale: 76th Annual International Meeting, SEG,
Expanded Abstracts, 605608.
Maxwell, S., C. K. Waltman, N. Warpinski, M. J. Mayerhofer, and N. Boroumand, 2006, Imaging seismic
deformation induced by hydraulic fracture complexity: SPE paper 102801-MS, doi:
10.2118/102801-MS.
McNamara, D. E., and R. P. Buland, 2004, Ambient noise levels in the continental United States: Bulletin
of the Seismological Society of America, 94, 15171527.
Peterson, J., 1993, Observation and modeling of seismic background noise: U.S.G.S. Open File Report
322/93, http://earthquake.usgs.gov/regional/asl/pubs/files/ofr93-322.pdf, accessed 21 May 2012.
Rodgers, P., 1992, Frequency limits for seismometers as determined from signal-to-noise ratios, Part 1:
The electromagnetic seismometer: Bulletin of the Seismological Society of America, 82, 10711098.

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 5
 Warpinski, N., 2010, The physics of surface microseismic monitoring:, www.halliburton.com, accessed
April 2012.
Downloaded 05/28/14 to 192.33.104.99. Redistribution subject to SEG license or copyright; see Terms of Use at http://library.seg.org/

2012 SEG DOI http://dx.doi.org/10.1190/segam2012-1069.1

SEG Las Vegas 2012 Annual Meeting Page 6
View publication stats