Page 1 of 14
Analysis of time of day seismicity at mine A
19th January 2010, revised 4th June 2014
Abstract
All mining-induced seismic events can be viewed as aftershocks of stope blasts, or of previous events. The
classical Omori law for the decay rate of aftershocks provides a good fit to seismic data over time scales of a
few seconds to 13 days after stope blasts. This result is based on analysis of seismic data from mine A were
centralised face blasting was practiced.
The post-blast seismicity decays to a level close to a background level after several hours and then continues
to decay. Therefore, seismicity during the day shift will be minimised by delaying the re-entry time after the
blast as much as is practical. This could be achieved for example by initiating the blast as soon as the stope
workers are well clear of their working areas.
Few development blasts go off at the same time as the face blasts. As development blasts are not formally
identified in the seismic data base, their presence complicates the seismic analysis. Furthermore, more than
one half of the seismic events with Magnitudes less than 2.0 are lost during the face blast period between
about 18:00 and 21:00, presumably due to difficulties in associating and analysing the seismicity during
periods of very high event rate. Special effort will be needed to solve this problem when the system is
upgraded.
Background
Mine A was on a system of centralised blasting. This system was introduced to concentrate a greater
proportion of seismicity into the blast window. If all blasting takes place within a few minutes in each day,
then the decay of seismicity can be studied to better understand the time-dependent rock mass response to
blasting and advancing the face. This report consists of an analysis of a recent event catalogue that represent
seismicity at Mine A after centralised blasting was well established.
Current analysis
Introduction
The purpose of the current analysis was to identify the times of centralised blasting from the seismic records
and to study the post-blast seismic response to look towards engineering applications such as re-entry times
and understanding the time response to face advance or blasting.
As the exact time of initiation of the daily blast was not available, a large part of my analysis consisted of a
search for the time of the daily blast. In hindsight, if the blasting times were available and a simpler analysis
followed, some of the results presented here might not have been forthcoming.
The analysis involves a few forms of seismic analysis such as time of day, frequency-Magnitude distributions,
and aftershock behaviour.
Aftershock behaviour will be studied using the Omori power law behaviour that has been shown to describe
the rate of post-blast seismicity (e.g. Spottiswoode, 1990, Lynch, 2005 and Kgarume et al 2010). The Appendix
shows that all mining seismicity could be considered to be aftershocks of the face advance.
Page 2 of 14
Data and software development
Seismic data was provided by the mine. It consisted of 81552 events recorded between 1st May 2007 and 4th
December 2009 inclusive. Only the 71327 events within an on-reef square 4608 m on a side in and around the
study area of Spottiswoode et al (2009) were used in this study. The spatial distribution is indicated in Figure
1. The following analysis is based on these events.
Figure 1 Plan view of seismicity at mine A superimposed on the mine outlines. Symbols indicate blast
sequences, as identified in this study, in blue and M>2.5 events in yellow through to red.
The following is a partial list of the processing of the seismic catalogue that I performed during this study. I
needed to write over 700 lines of code to achieve it. The code has sufficient documentation to make it usable
for other data sets and viable for alterations where needed. The reasons for this coding will appear within this
report.
1.
2.
3.
4.
5.
6.
For purposes of plotting seismicity, reduce the number of smaller events;
List time differences and identify temporal clusters (Type 1 sequences);
Measure the spatial spread of the temporal clusters;
Split the temporal clusters into spatio-temporal clusters and list for 3D plotting;
List number of events by Magnitude in categories of time of day and of temporal clusters;
List the maximum number of events in each day and within time windows (5, 10, 15 … 60 minutes),
yielding Type 2 sequences;
7. List the inferred blast times by day of week and by hour of day;
8. Stack seismicity according to the start of Type 2 sequences; and
9. Compare blast times estimated during step 2 with those estimated during step 6.
I used EXCEL to plot values in tables of derived data and MinView3D to plot seismicity.
I will argue that Type 1 sequences are mostly development blasts and Type 2 sequences occur in response to
face blasts.
Page 3 of 14
Analysis
The distribution of the number of events by Magnitude, commonly called the frequency-Magnitude
distribution or Gutenberg-Richter plot, is shown for all the events in Figure 2. The seismic catalogue did not
contain event Magnitudes, so Magnitudes were estimated using
ML = a*log(Energy) + b*log(Moment) + c
(1)
where the coefficients a, b, c take recommended values of a = 0.344, b = 0.516, and c = -6.572 (Durrheim et al,
2007). Magnitude is abbreviated as M in this report.
When tens of thousands of seismic events are plotted together, they appear very crowded, as seen in Figure 1.
The crowding is especially acute for small Magnitude events so the number of smaller events was reduced. A
factor of 10(M-4)/2 was used to reduce the number of events in the range M±0.05 as illustrated in Figure 2 and
plotted spatially again in Figure 3.
Freq-Mag with reduction using b=0.5
10000
b=0.5
N (M +- 0.05)
1000
all
reduced
10(M-4)/2
100
b=1.0
10
1
-3
-2
-1
0
1
Magnitude
2
3
4
Figure 2 Distribution of events by Magnitude (“all”). “Reduced” has been adjusted by dividing all values by a line given
by Log(N) = 2 – M/2 (equivalent to a “b” value of 0.5, or a factor of 10(M-4)/2) as shown.
Figure 3 As for Figure 1 with the number of events reduced as suggested in Figure 2
Search for time of blasting: Part 1
I tried to identify the time of the centralised blasting by searching for times of high event rate. After studying
the listing shown in part in Figure 4 (left), I selected an event rate of one event every 15s over nine or more
events as these sustained event rates were distinctly higher during some periods than during others. As can
be seen in Figure 4 (left and right), there were occasions when more than one such sequence occurred in a
Page 4 of 14
day. This problem was not entirely unexpected as Kevin Riemer had noted that development blasting was
often not tied in to the centralised blasting system and the seismic system recorded sequences of
development blasts.
The spread of each temporal cluster was measured in terms of the median inter-event distance, measured
between the projections of events onto reef. As can be seen in Figure 5, there are clearly two populations,
one with a spread of 100m or less that dominated the sequences that occurred between midnight (0:00) and
6:00 and then between 12:00 and 18:00. All values of median inter-event distance that exceeded 100m
occurred between 18:00 and 21:00, corresponding to the peak in seismicity rate also shown in Figure 5.
These temporal clusters consisted of 38% of all events and occupied only 0.30% of the total time. They
contributed a negligible amount of the total amount of total seismic moment or energy release.
Time to nine following events, s
1500
1000
500
0
0
5
10
15
20
Days after 2007/05/01 00:00:00
260
Cumulated number events
Increasing event number
Cumulated number events
2000
240
220
200
180
160
147500
148500
149500
150500
151500
Seconds after 2007/05/01 00:00:00
Figure 4 Method for selecting times of high event rate shown in text (left) and graphically (right). Right bottom is
enlarged from the red box in right top. The lines at right bottom represent a high event rate that must be exceeded for
9 events or more in succession.
Page 5 of 14
Space-time distributions of >= 10 event sequences
spaced <=15s apart
25000
Number of events per hour
hour
Median inter-event distances
10000
1000
100
Distribution of all events
events by
by hour
hour of
of day
day
20000
15000
10000
5000
0
0
10
00
03
06
09
12
15
18
21
6
12
Hour of day
18
21
24
00
Hour of day
Figure 5 Median inter-event distributions of blast sequences (left) and distribution of all events by hour of day (right).
Frequency-magnitude distributions
The time and spatial variations in the clustered events as shown in Figure 5 suggests that it might be useful to
divide events in terms of time clustering and space clustering. The frequency-Magnitude distributions in
Figure 6 are based on these divisions:

Time boundaries chosen as 6:00, 12:00, 18:00 and 21:00.

Clustering with spread <=100m and spread >100m and events not included in the clusters.
Figure 6 shows many distinct features:
1. The most distinctive feature is the contrast between the time-clustered data and the rest of the data
at right. The time-clustered data at right showed a very sharp peak at M=-1.3 and few events with
M>-0.6. The graph for all events also shows a hump at M=-1.3 that rises above the trend line marked
“b=0.5”. This hump disappears when the time-clustered events are removed.
2. The peak at M=-1.3 is clearly visible in two of the time bands at left, namely {12-18} and {18-21} hours,
but is not visible at time periods {21-6} and {6-12} hours. Note that there was much less time
clustering at these latter times than in the two active time bands within {12-21} hours.
3. The b-value for 2.0≤M≤4.0 is about 1.0 whereas it drops to 0.5 for -0.6≤M≤2.0.
4. The b-value of 0.5 for -0.6≤M≤2.0 is very pronounced for the three time periods outside of {18-21}.
However, the data for the period {18-21} hours has fewer events with M<1.3 than would be expected
if b=0.5 applied. If the data fell on a perfect b=0.5 line, as sketched in the dashed red construction line
then many more events would have been recorded in the hours {18-21}. In fact, 5124 were recorded
whereas 11120 would have been expected for -0.6≤M≤2.0, an apparent loss of events of over 50% in
this time period and Magnitude range.
Page 6 of 14
10000
b=0.5
All
b=0.5
1000
100
b=0.5
<100m
>100m
N (M+-0.05)
N (M+-0.05)
1000
10000
All
{6-12}
{12-18}
{18-21}
{21-6}
Time ranges in hours
Not clustered
Groups of >=9 events
<=15s apart
100
b=1.0
b=1.0
10
10
1
1
-3
-2
-1
0
1
Magnitude
2
3
4
-3
-2
-1
0
1
Magnitude
2
3
Figure 6 Frequency-Magnitude distributions by time of day in hours and by temporal and spatial clustering
The temporal clusters were then split into spatial clusters through successive agglomeration of events within
100m of one another. Despite the large values of the median inter-event distances, with many exceeding
1000m (Figure 5), the temporal clusters broke into only a few spatial sub-clusters, most often only two spatial
sub-clusters, as seen in Table 1.
Table 1 Distribution of spatial sub-clusters with temporal clusters
Number of spatial sub-clusters per temporal cluster
1
2
3
4
Number of temporal clusters
1345
243
24
7
I now refer to these spatial sub-clusters of the temporal clusters as spatio-temporal clusters.
Figure 7 Spatio-temporal clusters of events. Size scales with Magnitude and colour ranges from blue to red for
increasing time.
Search for time of blasting: Part 2
Part 1 of the search for the time of blasting yielded sequences of events that were spread from 12:00 to 21:00
and from 00:00 to 06:00, were spatially localised and had small Magnitudes (almost all with M<-0.6). This is
the familiar pattern for development blasting and not for the seismicity that follows face blasting. The familiar
pattern is normally shown as a histogram of the number of events in each hour of the day (Figure 5).
4
Page 7 of 14
I attempted to look the times of face blasting using the same approach that I used to identify the Type 1
sequences using events with M>-0.6, but without much success. I then found more success by searching for
the maximum number of events with M>-0.6 within fixed time periods, namely 5, 10, 15, …, 60 minutes.
Figure 8 shows the maximum number of events that took place within some of the chosen time periods in
Mondays through to Fridays. Through a process of iteration and inspection, I selected the “best” estimate of
the time of each daily blast to be the largest number of days within a time period while keeping the number
that fall earlier than 16:00 or later than 22:00 from Monday to Friday below one percent of the total. The
selection used for identifying blast times was four events within 10 minutes. 41.3% of days were not identified
and 0.5% of identified times fell outside of the expected time window.
The time of day distribution of types 1 and 2 estimates of blasting time is shown in Figure 9 and compared on a
day to day basis in Figure 10.
Frequency of occurrence of events within
selected time windows
Number of days
200
150
100
5
10
15
20
30
40
50
0
0
4
8
12
16
20
Maximum number within N minutes
Figure 8 Histogram of the maximum number of events with M>-0.6 in each day that occurred within time periods of
N=5, 10, 15, 20, 30 and 40 minutes. The arrows indicated expected tendencies for events that cluster in time.
Distribution of types 1 and 2 event sequences
Number of "blast" events
600
500
400
1
300
2
200
100
0
0
6
12
Hour of day
18
24
Figure 9 Time of day distribution of types 1 and 2 event sequences.
Figure 10 is annotated as if the two ways of identifying blasting are sensitive to development and face blasting
for types 1 and 2 respectively. The text boxes contain speculative suggestions regarding the difference in time
between the development blasts and the inferred face blasting. Further discussion is provided in the
Discussion below.
Page 8 of 14
Cumulated percent
100
Development blasting
slightly delayed after
face blasting
75
Night shift
development
blasting ?
50
Development blasting
12:00 to 18:00 and face
blasting 18:00 to 21:00
Early face
blasting?
25
0
-24
-18
-12
-6
0
6
12
18
24
Blast times (face - development), hours
Figure 10 Cumulated differences between the times of inferred face blasting and development blasting.
Time-dependent response of seismicity to face blasting
The blast times estimated in the previous section were used to stack cumulated seismicity to estimate an
average behaviour of seismicity after the blasts. The stacking process was explained by Kgarume et al (2010)
and by Spottiswoode (2009). Cumulated seismicity is shown below in terms of time as a linear scale on the left
and as a log scale on the right. For example, data for each day of the week is shown in Figure 11. Linear and
logarithmic scales were chosen to illustrate the applicability, or otherwise, of the following model for the
amount of seismicity following a blast:
n(t )  B 
K
(c  t ) p
(2)
Where
n(t) = number of events per second and
B, K and p are constants.
The first part of Equation (2) represents a constant rate of background seismicity and the second part is the
well-known Omori relationship. Kgarume et al (2010) found that the value of p at the study area was so close
to 1.0 that we can assume that p=1.0. Equation (2) can then be integrated to:
t
 n(t )dt  Bt  k (log
10
(c  t )  log 10 (c))
0
(3)
 Bt  k (log 10 (1  t / c)
Where k=ln(10)×K. k is preferred to K here as it can be measured directly off the linear-log plots on the right of
Figure 11 and Figure 12.
The numerical labels within regular pentagons in Figure 11 and Figure 12 are interpreted here as:
1. A high initial rate of seismicity at values of time t over the first few hours This appears as a log-type
curve on the linear scale on the left and a straight line on the logarithmic scale at the right;
2. A constant background rate after a few hours. This appears as a straight line on the linear scale on the
left and an exponential curve on the logarithmic scale at the right;
3. An accelerated rate soon before 24 hours. This is more obvious on the left-hand graphs and is caused
by the next day’s blast taking place less than 24 hours after the current day’s blasts.
Page 9 of 14
As can be seen in Figure 11, a c value of 40s provides a reasonably linear behaviour for the earliest post-blast
events of over 2½ orders of magnitude in time (from 40s to 10 000s). This is strong support for the assumption
of p = 1.0. Also note that the seismicity following blasting during each day of the week was similar, both in
terms of the Omori decay and the background rate.
Events as aftershocks of face advance
Events as aftershocks of face advance
30
25
Cumulated events / day
Cumulated events / day
30
3
20
15
Mon
Wed
Fri
2
10
Tues
Thur
Sat
1
5
20
6
12
18
24
30
36
42
Tues
Thur
Sat
3
15
2
10
5
0
0
Mon
Wed
Fri
25
1
0
48
10
Time after estimated face blast, hr
100
1000
10000
100000
40s+time after estimated face blasts, hr
1000000
Figure 11 Cumulated seismicity with time on a linear (left) and log (right) scale, stacked for each day of week. The
numbers in the pentagon are explained in the text.
I stacked the data further by averaging the seismicity following the blasting from Mondays to Fridays. I then
followed a visual curve-fitting exercise in EXCEL by applying Equation (3) to the stacked data, as shown in
Figure 12. As the first event in a sequence is assumed to be the time of the face blasts, stacking generates one
event at t=0: a small correction was needed in the curve-fitting to incorporate this offset. The event rate as
the slope of the modelled curves in Figure 12 is shown in Figure 13. Hours are used for the time scale for
greater ease of engineering interpretation.
Comparison between model & data
Comparison between model & data
25
3
{Mon-Fri}
Model
20
Cumulated events / day
Cumulated events / day
25
15
2
10
5
1
{Mon-Fri}
Model
20
15
10
5
0
0
0
6
12
18
24
10
Time after estimated face blast, hr
100
1000
10000
100000
40s+time after estimated face blasts, s
Figure 12 Comparison between observed and modelled seismicity following the inferred blasting time
10000
50
Modelled event rate (M>-0.6)
Modelled event rate (M>-0.6)
Events/day
Events/day
40
30
20
1000
100
10
10
0
0
3
6
9
12
15
18
21
24
1
0.01
Hours after face blast
Figure 13 Event rate derived from modelled seismicity as shown in Figure 12
0.1
1
10
Hours after face blast
100
Page 10 of 14
To avoid any dependence of size distributions on time of day, including overcoming the difficulties posed by
missed small events during the blasting period, data for M>1.0 events only is shown in Figure 14 and Figure 15.
The main difference is that the Omori-type behaviour dominates the seismicity rate for longer, leading to even
more benefit for lengthening the re-entry period after the face blasting.
Comparison between model & data
Comparison between model & data
5
{Mon-Fri}
Model
4
Cumulated events M>1. / day
Cumulated events M>1. / day
5
3
2
1
0
{Mon-Fri}
Model
4
3
2
1
0
0
6
12
18
24
10
100
Time after estimated face blasts, hr
1000
10000
100000
30s+time after estimated face blasts, s
Figure 14 Comparison between observed and modelled seismicity following the inferred blasting time, M>1.0 events
only.
20
1000
Modelled event rate (M>1.0)
Modelled event rate (M>1.0)
100
Events/day
Events/day
15
10
5
10
1
0
0
3
6
9
12
15
18
Hours after face blast
21
24
0
0.01
0.1
1
10
Hours after face blast
100
Figure 15 Event rate derived from modelled seismicity as shown in Figure 14.
Mine A stopped all production over the Easter weekend and the Christmas-New Year period each year. This
gave us the opportunity to study the seismicity following blasting over longer periods than 24 hours during the
week or 48 hours from Saturday to Monday. Estimated values of B and k are listed in Table 2 and post-blast
seismicity over the year-end break shown in Figure 16.
Table 2 Table of values of k and c for seismicity following blasting
Period
Days after
blast
1
B, events/day
k, events/decade
Mondays to Fridays
Number of periods stacked, or
start of blast
396
3.5
7.6
Saturdays
64
2
3.5
6.5
Easter 2008
2008/03/20 16:33:02
4
1.8
1.3
Easter 2009
2009/04/09 14:56:27
5
2.6
0.75
Christmas 2007
2007/12/21 17:47:38
13
3.2
2.5
Christmas 2008
2008/12/23 17:47:19
13
2.6
1.3
Page 11 of 14
Seismicity rate around year-end break
600
Y07/09
Y08/09
600
400
Y07/09
bl
bl
800
Cumulated M>-0.6 events
Cumulated M>-0.6 events
Seismicity rate around year-end break
1000
blasts before and
after year-end break
200
Y08/09
Sunday
500
blasts before and after
year-end break
0
-30
-20
-10
0
10
20
30
40
Days after break
400
-5
0
5
10
Days after break
15
20
Figure 16 Influence of the cessation of mining over the Christmas-New Year period.
Discussion and Conclusions
Centralised stope blasting has provided us with the opportunity of considering whether all mining-induced
seismicity could be considered to be aftershocks of production blasts. Indeed, post-blast seismicity on a daily
basis follows the Omori aftershock pattern (n(t) = K/(c+t)p). As previously reported by Kgarume et al (2010),
the value of p is close to 1.0. This allows us to use a standard spread-sheet program (EXCEL) to plot cumulated
seismicity against c+t to find a suitable value for the constant c. The constant c is a function of event
Magnitude and therefore most likely to be caused by the effective Magnitude threshold being elevated at
times of high event rate. This is underscored by the apparent loss of more than half of events with M<2.0
during the blasting time.
When Omori’s law is studied for aftershocks of earthquakes, it is widely recognised that the constant c reflects
incomplete recording of seismic events that occur within the coda waves of earthquakes. In other words,
seismic networks miss many events. The situation when applying Omori’s law to stope blasting is somewhat
different as the blasting can take up to many minutes to complete.
Seismic events with M<-0.6 are currently unusable for understanding induced seismicity as most of them are
development blasts.
All mining-induced seismic events can be viewed as aftershocks of stope blasts, or of previous events. The
classical Omori law for the decay rate of aftershocks provides a good fit to seismic data over time scales of a
few seconds to 13 days after stope blasts.
Mature mining takes place within a cloud of seismicity that is induced not only by the most recent production
blasts, but also within the decay of seismicity from earlier face blasts. As the seismicity that immediately
follows the face blast occurs in volumes of rock with a combination of high stress and high stress change, one
would expect that further mining would move the active areas out of the aftershock zone and therefore that
this zone would only exist for a limited amount of time. It is perhaps surprising that this could be as long as
several months, as suggested by the amount and character of seismicity that takes place during the end-ofyear
break.
Recommendations
The post-blast seismicity decays to a level close to a background level after several hours and then continues
to decay. Therefore, seismicity during the day shift will be minimised by delaying the re-entry time after the
blast as much as is practical. This could be achieved for example by initiating the blast as soon as the stope
workers are well clear of their working areas.
Any investment in improving the seismic network needs to be used to overcome the problems of event loss
and misidentification of events.
Page 12 of 14
References
Durrheim, R.J., Cichowicz, A., Ebrahim-Trollope, R., Essrich, F., Goldbach, O., Linzer, L., Spottiswoode, S.M.,
Stankiewicz, T. and van Aswegen, G. (2007) Minimising the Rockburst Risk (Phase 2): Output 3 - Guidelines,
Standards and Best Practice for Seismic Hazard Assessment and Rockburst Risk Management, Safety in Mines
Research Advisory Committee, Final Project Report, 6 March 2007
Kgarume, T.E., Spottiswoode, S.M. and Durrheim, R.J. (2010) Statistical properties of mine tremor aftershocks,
Pageoph Special Issue: Induced Seismicity.
Lynch, R.A. (2005) Proactive approaches to rock mass stability and control, MHSC final report for project SIM
02 03 02.
SM Spottiswoode. (2000) Aftershocks and foreshocks of mine seismic events. 3rd international workshop on
the application of geophysics to rock and soil engineering, GeoEng2000, Melbourne Australia.
Spottiswoode, S.M., Linzer, L.M. and Majiet, S. (2008) Energy and stiffness of mine models and seismicity,
Accepted by 1st Southern Hemisphere International Rock Mechanics Symposium, Perth, Australian Centre of
Geomechanics, 16-19 September 2008, pp693-707.
Spottiswoode, S.M., Milev, A., Linzer, L.M. and Majiet, S. (2009) Evaluation of the design criteria of Regularly
Spaced Dip Pillars (RSDP) based on their in-situ performance, Final Project Report for Mine Health and Safety
Council, contract SIM 04 03 01.
Page 13 of 14
Appendix: Are all mine events aftershocks of blasting or of one
another?
In the main body of the report, seismicity was accurately modelled as a constant background level
superimposed on Omori power-law behaviour after the daily face blast. What is source of the “constant
background level”? This Appendix will consider whether the background seismicity could be the tails of
Omori-type seismicity rate.
Equation (3) can be rewritten and extended, as shown in Equation (4), to represent seismicity following many
days cumulated seismicity.
t
n
 n(t )dt   k n  (log 10 (1  (t  n  24hrs) / c))  log 10 (1  n  24hrs / c))
(4)
n 0
0
Where n=0 represents today and n>0 is earlier days,
kn is set as 10.0 when there has been a blast on day n and set as zero if no blast has occurred, and
c = 20s.
The second term of Equation (4) removes all the seismicity up until the moment before today’s blast.
The effect of previous days blasting is shown in Figure 17 in which the modelled seismicity following 16 powers
of two ( 1 to 32768). Figure 17 illustrates the short-term and long-term behaviour of Equation (4) that
matches the data:
At short times, cumulated seismicity increases as the logarithm of t+c at small times (t <~one hour) as
illustrated in Figure 17 (right) while cumulated seismicity increases approximately lineally at long times
(t >~three hours) (Figure 17, left). The approximate linearity at long times is underscored when the difference
between the first and last day in is taken (Figure 17, left).
Effect of long-term Omori response
Effect of long-term Omori response
200
200
Cumulated events per day
160
140
1 year
120
1 month
100
1 week
1 day
80
60
40
ear
90 y
20
s-
firs
y
t da
Cumulated events per day
90 years
180
90 years
180
160
140
1 year
120
1 month
100
1 week
80
1 day
60
40
20
0
0
0
6
12
18
24
1
10
Hours after blast
100
1000
10000
100000
seconds after blast
Figure 17 Application of equation (4) for power of two in the number of days (n).
I now extend the comparison between observed and modelled seismicity over the normal 24-hour production
cycle to the end-of year break. The upper part of Figure 18 is copied from Figure 16 to allow visual comparison
with modelled seismicity when a 10-day break is taken after 900 days of mining. Two characteristics not
observed during normal production are seen to be in common between the observed and modelled curves in
Figure 18:
1. The rate of seismicity slows down by a similar amount; and
2. The rate of seismicity after the break is slightly less than before the break.
Page 14 of 14
Seismicity rate around year-end break
600
Y07/09
Y08/09
800
600
400
Y07/09
bl
bl
Cumulated M>-0.6 events
Cumulated M>-0.6 events
Seismicity rate around year-end break
1000
blasts before and
after year-end break
200
0
500
blasts before and after
year-end break
400
-30
-20
-10
0
10
20
30
40
-5
0
5
10
Days after break
Days after break
8000
6000
4000
2000
Events
15
20
4200
Long break and Sundays
Cumulated seismicity
Cumulated seismicity
Y08/09
Sunday
bl
Sunday
Long break and Sundays
3800
3400
3000
Events
bl
Sunday
2600
0
0
10
20
30
40
Days
50
60
70
25
30
35
40
45
Days
Figure 18 Influence of the cessation of mining over the Christmas-New Year period (above) and seismicity modelled
using the Omori law following daily blasts over 100 days (below).