Professional Documents
Culture Documents
Ma 2018
Ma 2018
International Journal of
Rock Mechanics and Mining Sciences
journal homepage: www.elsevier.com/locate/ijrmms
A R T I C L E I N F O
Keywords:
Mining-induced seismicity
Mainshock
Seismic monitoring
B-value
⁎
Corresponding author.
E-mail address: xuma@vt.edu (X. Ma).
https://doi.org/10.1016/j.ijrmms.2018.02.003
Received 1 June 2017; Received in revised form 5 December 2017; Accepted 3 February 2018
1365-1609/ © 2018 Elsevier Ltd. All rights reserved.
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
value cannot achieve the goal of providing underground mines a full microseismic events from July 2007 to March 2012, the temporal
picture of seismic hazards. Restraints and uncertainties involved in magnitude distribution and cumulative number of which is shown in
seismicity forecasting on historical data from multiple sites perhaps Fig. 2a. In addition, Fig. 2b and Fig. 2c show the spatial distribution of
yield imprecise results. Rock failures in the proximity of maximum microsesimic events, mainshocks, and stations at Kidd Mine. Moment
stress concentration are usually found surrounding the edges of mining magnitudes and temporal distributions of microseismic events are ex-
excavations. The length of cracks and stress both increase gradually hibited as well. In addition, Fig. 3 shows the frequency of magnitude
during the first regime and then develop faster to achieve the critical distribution for the events of Creighton Mine and Kidd Mine. The cutoff
stress. Seismic events are triggered when the rupture develops. Fracture magnitude of seismic events for computing the b-value was assessed by
planes are formed parallel to the stope faces.30 The stress drop arises on analyzing the magnitude of completeness, which is defined as the
the fracture plane associated with seismic events, redistributing stress lowest magnitude at which 100% of the events are detected in temporal
to adjacent areas of the rock mass.31 and spatial scales.32 By visualizing the frequency distribution of mag-
The objective of this article is to present the change of b-value as- nitude, it was inferred that the magnitude of completeness for
sociated with mining-induced seismicity sequences, and to describe Creighton Mine should be larger than −1.6 and that for Kidd Mine
relationships of b-value and mainshocks within hard-rock underground should be over −1.9 according to their maximum frequency windows
mines. We found that mining-induced seismicity from two mine sites of magnitude. Further, a rigorous method, originally used in crustal
agreed with the law of Gutenberg-Richter frequency-magnitude very earthquake studies, for obtaining the accurate magnitude of com-
well. By comparing the temporal evolution of b-value for consistent pleteness was firstly used in mining-induced seismicity, validating that
seismicity sequences of these two mines, a uniform pattern of b-value it could be well applied to mining-induced seismicity for better un-
change before the occurrence of mainshocks was recognized. The b- derstanding correlations between the b-value variations and main-
value was computed using the seismicity within the same time scale for shocks. The computing process of determining the optimal magnitude
each mine. The maximum likelihood method was used to describe the of completeness for mining-induced seismicity sequences is explained
regression between frequency and magnitude of mining-induced seis- below.
micity and evaluate the goodness of fit. We validated that the mining- An important empirical observation on seismic events is the
induced seismicity obeyed the Gutenberg-Richter law and the max- Gutenberg-Richter law, which interprets the proportional relationship
imum likelihood estimation method could be applied to determine the between the magnitude M and the cumulative number of seismic events
magnitude of completeness for computing b value. The correlation with magnitude larger than M:
between mainshocks and temporal variations of the b-value can be
potentially used to assess long-term mining-induced seismic hazards. logN = a − bM (1)
Table 1
Times, locations, and magnitude of mainshocks at creighton mine.
Mainshocks Date Time North (m) East (m) Depth (m) Moment magnitude
65
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
Table 2
Times, locations, and magnitude of mainshocks at kidd mine.
Mainshocks Date Time North (m) East (m) Depth (m) Moment magnitude
5 18000
Microseismic events
4 Mainshocks 16000
Cumulative number
2 12000
1 10000
0 8000
-1 6000
-2 4000
-3 2000
-4 0
2007 2008 2009 2010 2011 2012
Year
(a)
(a)
(b)
(b)
(c)
(c)
Fig. 2. (a) Temporal distributions of microseismic events and mainshocks at Kidd Mine.
Fig. 1. (a) Temporal distributions of microseismic events and mainshocks at Creighton (b) Plan view and (c) longitudinal view of microseismic events, mainshocks, and stations
Mine. (b) Plan view and (c) longitudinal view of microseismic events, mainshocks, and at Kidd Mine.
stations at Creighton Mine.
66
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
(a)
(a)
(b)
Fig. 3. Frequency distribution with moment magnitude of seismic events at (a) Creighton
Mine (b) Kidd Mine. (b)
Fig. 4. Cumulative numbers of aftershocks are given as functions of magnitude for (a) the
using maximum likelihood method is interpreted in Fig. 5 and the first mainshock at Creighton Mine and (b) the first mainshock at Kidd Mine. Lines are the
following descriptions. best fit of Gutenberg-Ritcher relation.
We employed the estimation method, developed by Wiember and
Wyss35, that describes the self-similarity of seismicity processes and from Creighton Mine and a data set of 90 days of Kidd Mine, respec-
implies a power-law distribution of seismicity. According to the im- tively. As shown in Fig. 6, the values of goodness of fit at Creighton
plication of power-law distribution of seismicity (Fig. 4), this method Mine and Kidd Mine increase first and achieve the peak around moment
considers the estimate of magnitude of completeness and its influence magnitude of −1.5. They start dropping gradually once the moment
on the b-value. We used the maximum likelihood method to determine magnitude is over about −1.5. For this data set at Creighton Mine, the
an optimal magnitude of completeness for each data set in a selected optimal assumed Mc = −1.48 can explain 87% of the data variability
period for both Creighton Mine and Kidd Mine. By evaluating the (Fig. 6). Similarly, the optimal assumed Mc = −1.47 can explain 92%
goodness of fit of one time period, we picked an optimal magnitude of of the data variability for this picked data set of Kidd Mine (Fig. 6).
completeness. First, we estimated the b- and a- value using filtered Consequently, we determined that the magnitude of completeness for
events with M ≥ Mi based on Gutenberg-Richter law35,36. Mi represents this randomly picked data set of Creighton Mine is −1.48 and −1.47
the selected magnitude for estimating the b- and the a- value. To con- for this data set of Kidd Mine. Note that similar analyses of goodness of
struct a perfect fit of Gutenberg-Richter law, a series of estimated b- fit have been conducted for all data sets and the magnitude of com-
value, a value, and their corresponding Mi were used to construct a pleteness for each data set are varied because each data set has its own
synthetic distribution. Then, the absolute difference between the ob- magnitude of completeness based on the goodness of fit analysis.
served and synthetic distribution in each magnitude bin (bin = 0.1 for After determining the magnitude of completeness, we applied the
Creighton Mine and Kidd Mine) was calculated to estimate the goodness magnitude of completeness and computed the a-value and b-value, as
of fit: well as standard error and standard deviation, of each data set of
Creighton Mine and Kidd Mine using method developed by Shi and
M
max
⎛ ∑Mi Bi − Si ⎞ Bolt36. The standard error of b is
R (a, b, Mi ) = 100 − ⎜ × 100⎟
∑i Bi
⎝ ⎠ (2) σ (b) = 2.30b2σ (M ) (3)
where Bi and Si are the observed and synthetic cumulative number of where b is the estimated b-value. σ (M ) represents the standard devia-
events in each magnitude bin. We randomly picked a data set of 14 days tion:
67
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
Y
Mining-induced
seismicity
Obtain high
N quality goodness
of fit?
Fig. 5. Flowchart illustrating the procedure of constructing data sets, determining the magnitude of completeness, computing b-value and related statistical parameters.
3. Results
where M is the mean value of all magnitude values that are satisfied We analyzed the b-value change with mainshocks for Kidd Mine and
with Mi > Mc, and n is the number of these magnitude values. The found a similar pattern of temporal evolution of b-value as Creighton
standard error and standard deviation of each data set are important to Mine using mining-induced seismicity of the entire mine. The first
evaluate the data variability and confidence level of calculation signature of an elbow point (Fig. 7b) forecasting mainshocks appeared
68
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
1.2 4. Discussion
2nd mainshock period
1.1 1st mainshock period According to the case studies of these two underground mines, we
found that b-value increased to a peak value up to the occurrence of
1 mainshocks caused by mining-induced seismicity within a certain
period. The b-value initially declined to a relative low level in a certain
b-value
0.9
period before mainshocks and grew to higher values with approaching
the mainshocks periods. This provides critical information for seismic
hazards forecasting and mitigation. According to these case studies, we
0.8
found an elbow point of the b-value evolution, which was caused by the
The elbow point of the decrease and return of the b-value, near the occurrence of mainshocks.
0.7 b-value variations at These studies validated inherent similarities of mining-induced seis-
Creighton Mine
micity and crustal earthquakes in terms of source mechanism, although
0.6 the triggered mechanism and regions of influence are on different
04/2011 05/2011 06/2011 07/2011 08/2011 09/2011 10/2011
Date scales.
(a) Given the evidence of the temporal variation of b-value in crustal
earthquakes, there are mainly two patterns of b-value change asso-
1.6
ciated with mainshocks.22 First, the b-value increases to a peak and
2nd mainshock period
1.5 1st mainshock period then drops preceding aftershocks. Aftershocks occur during the de-
1.4 crease. Second, a high b-value appears prior to the mainshock. Overall,
the b-value is influenced by either the previous mainshock or the
1.3
mainshock after the peak of b-value. The investigation of b-value
b-value
1.2 change in Creighton Mine demonstrates that b-value evolving with the
The elbow point of the b-value
1.1 variations before mainshocks occurrence of mainshock agrees with the trend in the crustal earth-
in 2009 at Kidd Mine quakes study discussed above.
1
The elbow point of the b-value A previous study claimed that research on b-value spatial change is
0.9 variations before mainshocks
in 2011 at Kidd Mine more important because b-value change spatially could indicate stress
0.8 drops in the spatial analyses.28 The hypothesis is that b-value variation
3rd & 4th mainshocks period
0.7
is conversely correlated with stress change and mainshocks occur where
07 08 08 09 09 10 10 11 11 12 the largest gradient in b-value exists. In order to ensure accurate b-
20 n 20 c 20 n 20 c 20 n 20 c 20 n 20 c 20 n 20
ec u e u e u e u e u value estimate, analyzing spatial change in the b-value requires con-
ct-D pr-J ct-D pr-J ct-D pr-J ct-D pr-J ct-D pr-J
O A O A O A O A O A
siderable available induced seismicity to fully measure the spatial b-
Date
(b) value change, as induced seismicity needs to be spatially divided into
multiple zones or volumes. To accumulate enough energy release in
Fig. 7. Temporal change of the b-values with standard error at (a) Creighton Mine (b)
Kidd Mine. Continuous line with unfilled circles: b-values through all time periods; circles
different regions, probably a longer duration is required to satisfy the
filled red: the specific time period with occurrence of mainshocks. needed amount of induced seismicity in each region. Thus, temporal
analyses on b-value change would still play an important role because it
allows an investigation of the seismic hazard even if induced seismic
during the period from late 2008 to early 2009, which was prior to the
data acquired over short time period is available. If significant b-value
first mainshock period and the second mainshock period. The second
change can be identified for underground mines, patterns of which can
one (Fig. 7b) exhibited during the period from 2010 to 2011, which was
reflect potential locations such as faults and ongoing excavations, so as
prior to the third and the fourth mainshocks. As shown in Fig. 7b, the
to filter out potential mainshocks and enhance the mining safety. More
first and the second periods with mainshocks experienced the largest
analysis is required with multiple mainshocks over time to confirm the
and the second largest b-value. There was a decline of b-value prior to
pattern obtained with these case studies, so as to ensure that this
soaring to the highest level so that an elbow point of b-value evolution
methodology is robust and can be reliably applied to underground
was formed. The b-value increased by 21% from the previous period
mines.
before mainshock and surged to nearly 1.52 at the period when the first
mainshock occurred. It slightly dropped to 1.49 at the next period when
5. Conclusion
the second mainshock was triggered. After the b-value continuously
experienced a high level with mainshocks, it reduced gradually to about
Through computing and examining the b-value variations by using
1.2 through 6 months and fluctuated at this level during the following
the optimal magnitude of completeness, we identified significant
periods until approaching the period with the mainshocks occurring in
changes of the b-value associated with mainshocks. The b-value tends
2011. Before the occurrence of mainshocks in 2011, the b-value
to drop before the occurrence of mainshocks and then significantly
dropped to 0.9 within 6 months. Then, it grew to over 1.0 in the period
grows to a higher level near or within the mainshocks period. Likewise,
with mainshocks in 2011. Due to the initial drop before the occurrence
an elbow point appears during the b-value change prior to the period
of mainshocks and following increase of b-value within the mainshock
with mainshocks. The appearance of an elbow point in the b-value vs
periods, an elbow point of the b-value change appeared before main-
time plot could be used as an indicator of the increasing probability of
shocks period in 2011. While we used seismicity within 14 days as a
incurring a mainshock. These findings confirm the implication that b-
unit to yield a b-value for Creighton Mine, seismicity within a quarter
value can be used to forecast potential mainshocks and mitigate
was used to produce a b-value for Kidd Mine, as the average seismic
mining-induced seismic hazards. The comparison between observed
rate at Creighton Mine was nearly 6 times higher than Kidd Mine. At
and predicted cumulative number of seismic events could be applied to
least a thousand seismic events need to be used to enable yielding a
examine the goodness of fit, ensuring statistical significance for de-
statistically reliable b-value from previous investigations.
termining the magnitude of completeness of mining-induced seismicity.
A reasonable examination of magnitude of completeness ensures
accurate regression between cumulative number of seismic events and
the distribution of magnitudes. According to our case studies, every
69
X. Ma et al. International Journal of Rock Mechanics and Mining Sciences 104 (2018) 64–70
magnitude of completeness could be estimated for each data set based wave splitting analysis: an example from a mining setting. Geophys J Int.
on the goodness of fit using maximum likelihood method. It is re- 2011;187:848–860.
14. Young R, Collins D, Reyes-Montes J. Quantification and interpretation of seismicity.
commended that standard error and standard deviation should be Int J Rock Mech Min Sci. 2004;41:1317–1327.
evaluated after determining the magnitude of completeness of mining- 15. Lu C, Dou L, Zhang N, Xue J, Wang X. Microseismic frequency-spectrum evolutionary
induced seismicity. rule of rockburst triggered by roof fall. Int J Rock Mech Min Sci. 2013;64:6–16.
16. Ma X, Westman EC, Fahrman BP, Thibodeau D. Imaging of temporal stress redis-
tribution due to triggered seismicity at a deep nickel mine. Geomech Energy Environ.
Acknowledgements 2016;5:55–64.
17. Kranz R. Microcracks in rocks: a review. Tectonophysics. 1983;100:449–480.
18. Gutenberg B, Richter C. Magnitude and energy of earthquakes. Ann di Geofis.
The authors would like to acknowledge the Canadian Mining 1956;9:1–15.
Industry Research Organization for providing data and financial sup- 19. Wesnousky S. Crustal deformation processes and the stability of the Gutenberg-
port for this study. We thank Vale and Glencore for their overall in- Richter relationship. Bull Seismol Soc Am. 1999;89:1131–1137.
20. Shcherbakov R, Turcotte DL, Rundle JB. Aftershock Statistics. Pure Appl Geophys.
volvement and contributions to this study. This work is also supported
2005;162(6–7):1051–1076.
by NIOSH (contract 200-2011-40313) through the Capacity Building 21. Isacks B, Oliver J. Seismic waves with frequencies from 1 to 100 cycles per second
and Ground Control Research for the Mining Industry program. recorded in a deep mine in northern new Jersey. Bull Seismol Soc Am.
1964;54:1941–1979.
22. Smith W. The b-value as an earthquake precursor. Nature. 1981;289:136–139.
References 23. McGarr A, Green R. Measurement of tilt in a deep‐level gold mine and its relationship
to mining and seismicity. Geophys J Int. 1975;43:327–345.
1. Urbancic T, Trifu C, Young R. Microseismicity derived fault‐Planes and their re- 24. Mogi K. Magnitude-frequency relation for elastic shocks accompanying fractures of
lationship to focal mechanism, stress inversion, and geologic data. Geophys Res Lett. various materials and some related problems in earthquakes (2nd paper). Bull Earthq
1993;20:2475–2478. Res Inst. 1962;40:831–853.
2. Urbancic T, Trifu C. Recent advances in seismic monitoring technology at Canadian 25. Lockner D. The role of acoustic emission in the study of rock fracture. Int J Rock Mech
mines. J Appl Geophys. 2000;45:225–237. Min Sci. 1993;30:883–899.
3. Malek F, Espley S, Yao M, Trifu C. Management of high stress and seismicity at Vale 26. Boettcher M, McGarr A. Extension of Gutenberg‐Richter distribution to Mw− 1.3, no
Inco Creighton Mine. 42nd US Rock Mech Symp. Paper ARMA-08-386. lower limit in sight. Geophys Res. 2009;36:L10307.
4. Hudyma M, Potvin Y. An engineering approach to seismic risk management in 27. McGarr A. Violent deformation of rock near deep-level, tabular excavations—seismic
hardrock mines. Rock Mech Rock Eng. 2010;43:891–906. events. Bull Seismol Soc Am. 1971;61:1453–1466.
5. Brady B, Leighton F. Seismicity anomaly prior to a moderate rock burst: a case study. 28. Urbancic T, Trifu C, Long J, Young R. Space-time correlations of b values with stress
Int J Rock Mech. 1977;14:127–132. release. Pure Appl Geophys. 1992;139:449–462.
6. Abdul-Wahed M, Heib M Al, Senfaute G. Mining-induced seismicity: seismic mea- 29. Vallejos J, McKinnon S. Correlations between mining and seismicity for re-entry
surement using multiplet approach and numerical modeling. Int J Coal Geol. protocol development. Int J Rock Mech Min Sci. 2011;48:616–625.
2006;66:137–147. 30. Cook NGW. Seismicity associated with mining. Eng Geol. 1976;10(2):99–122.
7. Convertito V, Maercklin N, Sharma N. From induced seismicity to direct time‐- 31. Shaw B. Generalized Omori law for aftershocks and foreshocks from a simple dy-
dependent seismic hazard. Bull Seism Soc Am. 2012;102(6):2563–2573. namics. Geophys Res Lett. 1993;20:907–910.
8. Bourne S, Oates S, Elk J. A seismological model for earthquakes induced by fluid 32. Woessner J, Wiemer S. Assessing the quality of earthquake catalogues: estimating the
extraction from a subsurface reservoir. J Geophys Res. 2014;119:8991–9015. magnitude of completeness and its uncertainty. Bull Seismol Soc Am.
9. Bommer J, Crowley H, Pinho R. A risk-mitigation approach to the management of 2005;95(2):684–698.
induced seismicity. J Seismol. 2015;19:623–646. 33. W. Goebel TH, Schorlemmer D, Becker TW, Dresen G, Sammis CG. Acoustic emis-
10. Bourne S, Oates S, Bommer J, Dost B. A Monte Carlo method for probabilistic hazard sions document stress changes over many seismic cycles in stick-slip experiments.
assessment of induced seismicity due to conventional natural gas production. Bull Geophys Res Lett. 2013;40(10):2049–2054.
Seism Soc Am. 2015;105:1721–1738. 34. Schorlemmer D, Wiemer S, Wyss M. Variations in earthquake-size distribution across
11. Baker J, Gupta A. Bayesian treatment of induced seismicity in probabilistic seismi- different stress regimes. Nature. 2005;437:539–542.
c‐hazard analysis. Bull Seism Soc Am. 2016;106:860–870. 35. Wiemer S. Minimum magnitude of completeness in earthquake catalogs: examples
12. Young R, Maxwell S. Seismic characterization of a highly stressed rock mass using from Alaska, the Western United States, and Japan. Bull Seismol Soc Am.
tomographic imaging and induced seismicity. J Geophys Res. 2000;90(4):859–869.
1992;97:12,361–12,373. 36. Shi Y, Bolt BA. The standard error of the magnitude-frequency b value. Bull Seismol
13. Wuestefeld A, Kendall J, Verdon J. In situ monitoring of rock fracturing using shear Soc Am. 1982;72:1677–1687.
70