Aftershocks of the largest earthquakes can be found by waveform cross correlation. Several examples of successful recovery of the biggest aftershock sequences are presented

1. Recovering large-scale aftershock sequences using
waveform cross correlation at the primary IMS
seismic stations
Ivan Kitov, IDC/SA/SM, ivan.kitov@ctbto.org

2. Outline
• Principal features of aftershock sequences of catastrophic earthquakes:
Tohoku as the greatest challenge
• Waveform cross correlation (WCC) processing to build preliminary
cross correlation standard event list (XSEL)
• Global set of efficient master events
• Examples of interactive analysis of the XSEL for aftershocks
sequences
• Automatic Spot Check and Interactive Spot Check (Spot Check Tool)
• Tentative results of four aftershock sequences recovery
• Tuning of CC-detection, Local Association, and Conflict Resolution
parameters
• Discussion

3. Aftershocks: experiment setting
Area setting
sequence start end lat,deg lon, deg
Tohoku 2011070 2011076 30N - 46N 134E-148E
Nepal 2015115 2915121 25N - 30N 83E-88E
Chile 2015259 2015265 28S - 34S 67W-75W
PNG 2018056 2018062 1S - 11S 136E-148E
IDC Origin lat,deg long, deg mb D, km nass
Tohoku 05:46:20 38.435 142.520 5.51 0.0 122
Nepal 06:11:24 28.159 84.703 5.98 0.0 74
Chile 22:54:29 -31.517 -31.517 5.86 0.0 82
PNG 17:44:40 -6.044 142.800 5.85 0.0 63
Main shock IDC solution
Total area of the aftershock zones can be
larger than 100,000 sq.km. The aperture is
larger than 600 km. Seismic stations at shorter
distances have varying azimuth to aftershocks.
The master event approach needs to cover the
whole are for efficient processing.
All four events were shallow. Body wave
magnitude scale is saturated near 6.0 and does
not represent the size of these earthquakes
well. The number of stations in the zones of
existing seismic waves (excluding the shadow
zone between 100 and 115 degrees) varies
over the studied events.
Four aftershock sequences with different best seismic station sets

4. Tohoku REB : 2011070-2011076
Major challenge
Time distribution of mb and nass
2.5
3.5
4.5
5.5
6.5
0 1 2 3 4 5 6 7
mb
day
0
20
40
60
80
100
120
140
0 1 2 3 4 5 6 7
nass
day
Body wave magnitudes of the REB events during the first day are much higher, and then the average magnitude
decreases. The same effect, but less prominent, is observed in the number of associated phases. The biggest events
generate extremely large number of secondary phases, which mask primary phases from smaller events, lead to
multiple (combinatorial explosion) mis-associations into invalid event hypotheses in SEL3 and scanner, and also
create noise coherent with valid signals, i.e. negative conditions for waveform cross correlation.

5. REB – SEL3 comparisonas a benchmark for comparison
with the XSEL,which is competingwith SEL3/scanner
1. How many SEL3 events were built?
2. How many SEL3 events match REB by evid and how many were rejected?
3. How many SEL3 events match REB by 2+ phases within 10 s?
4. What in the distribution of SEL3-REB distances?
5. What is the distribution of SEL3-REB depth difference?
6. How many arrivals associated with SEL3 are in the REB by arid?
7. What is the rate of correct interpretation (e.g. N-phases renamed) of SEL3 phases associated
with the REB by arid?
8. What is the distribution of arrival time difference?
Same questions are applicable for the event hypotheses created by scanner, and we do not know
how many event hypotheses were created by scanner (e.g.~120-150 per day of normal
seismicity, and by a factor of 10 to 20 larger for the Tohoku sequence?)
Main question: how can we reduce the analyst workload related to the
largest sequences and how can we estimate the improvement?
This is a multi-dimensional task lacking quantitative estimate

6. SEL3/scanner problems for largest aftershock sequences
1. Extremely high detection rate close to saturation
2. However, too high detection threshold – actual REB detections missing
3. Large distance between consecutive arrivals with similar characteristics –
appropriate detections suppressed/rejected
4. High rate of phase misinterpretation due to similarity in characteristics
5. High rate of phase mis-association into false events
6. As a result, combinatorial explosion in the number of false event hypotheses
(especially for scanner)
7. Extremely high event hypotheses rate close to saturation
Diagnosis: too many poor event hypotheses (rejection workload), not
enough good event hypotheses (creation from scratch workload)

7. Four aftershock sequences:general statistics
Tohoku day REB nass (P) SEL3 by evid
SEL3
nass by arid REB only
1 2011070 743 14375 887 540 14272 6182 203
2 2011071 714 12044 946 559 14082 5952 155
3 2011072 574 8781 712 440 19715 4694 134
4 2011073 444 6544 603 354 7477 3289 90
5 2011074 392 4978 522 303 5954 2585 89
6 2011075 321 4157 481 257 4923 2163 64
7 2011076 322 3829 477 241 5182 2139 81
Total 3510 54708 4628 2694 71605 27004 816
Nepal day REB nass (P) SEL3 by evid
SEL3
nass by arid REB only
1 2015115 211 2620 262 132 2882 1411 79
2 2015116 83 935 239 58 978 494 25
3 2015117 49 417 154 31 351 189 18
4 2015118 26 180 174 17 146 84 9
5 2015119 22 215 175 18 220 116 4
6 2015120 16 109 183 7 78 39 9
7 2015121 13 124 182 11 126 70 2
Total 420 4600 1369 274 4781 2403 146
Chile day REB nass (P) SEL3 by evid
SEL3
nass by arid REB only
1 2015259 36 406 220 13 396 162 23
2 2015260 312 3385 391 184 3556 1531 128
3 2015261 119 1154 256 73 1280 513 46
4 2015262 82 994 233 58 1142 516 24
5 2015263 64 605 195 36 722 314 28
6 2015264 48 571 210 37 880 347 11
7 2015265 37 364 211 30 461 213 7
Total 698 7479 1716 431 8437 3596 267
PNG day REB nass (P) SEL3 by evid
SEL3
nass by arid REB only
1 2018056 122 1326 243 80 1397 589 42
2 2018057 195 1633 284 133 2102 810 62
3 2018058 90 826 250 69 1020 451 21
4 2018059 71 605 225 53 946 357 18
5 2018060 53 317 195 40 426 168 13
6 2018061 56 466 222 44 706 278 12
7 2018062 33 223 171 27 320 113 6
Total 620 5396 1590 446 6917 2766 174
The Tohoku sequence is the most difficult challenge to automatic and interactive processing. From 23% (Tohoku) to
38% (Chile) of REB events were added by analysts using scanner or from scratch. This work also included a huge
number of newly associated phases. Thousands of event hypotheses were rejected. Scanner?

8. REB and SEL3: Tohoku 2011070 + 2011071
Locationdifference
SEL3 events (matching REB by evid) are distributed over the globe and the REB events are within a
relatively small area. The distance between the SEL3 and REB events varies from several km to 153
degrees. There are 201 SEL3 events beyond 4 degrees and 34 in the range between 15 and 153 degrees.
SEL3 events – blue circles
REB events - black circles
SEL3 event in the REB by evid
1.00E+00
1.00E+01
1.00E+02
1.00E+03
0 5 10 15
#
distance, degrees
SEL3/REB distance
(total 1099 evid
matched, 34 between
16 and 180 degrees)

9. REB and SEL3: 2011070 + 2011071
Locationdifference
Days 2011070 and 2011071 are characterized by the highest rate of
aftershocks: 743 REB events on March 11 and 714 on March 12. On
average, there were ~35 events per hour. During the first hour after
the main shock, from 6 am to 7 am, there were found 37 REB
events, and during the next hour – 45 REB events. The noise level
after the main shock was too high to detect small- and many mid-
sized events. The rate of 1 event per 1.3 minutes within relatively
small area is a challenge for DFX and GA. The SEL3 is likely
saturated.

10. REB and SEL3: 2011070-2011076
Depth and magnitude difference
Depth distribution Magnitude distribution
-Depth of SEL3 events (matching the REB by evid) is larger (overestimated?) compared to the REB events
- Many REB events have depth fixed to 0, according to strict IDC location rules. This effect leads to larger
distances between the REB and SEL3 event with the same evid.
- Magnitude distributions for both sets are similar with small deviation at low and high mb values
1.00E-04
1.00E-03
1.00E-02
1.00E-01
1.00E+00
0 200 400 600 800
pdf
depth, km
SEL3
REB
1.00E-04
1.00E-03
1.00E-02
1.00E-01
1.00E+00
2 4 6 8
pdf
mb, km
SEL3
REB

11. REB and SEL3: 2011070-2011076
Origintime difference
Frequency distribution of time differences between origin times of sequential events
Because of very high rate of events and signals, the SEL3 misses many events with short difference
between origin times. This is likely due to the rule prohibiting similar P-wave arrivals within 90 s, and
the signals from the aftershock area are similar. This rule suppresses detection of valid signals from
events close in time and space: subsequent arrivals with shorter time difference at a given station are
highly likely tx instead of P. The REB events are added by analysts likely using scanner results (success
rate?)
0
20
40
60
80
100
120
140
160
180
200
0 100 200 300 400
#
origin time difference, s
SEL3
REB
0
20
40
60
80
100
120
0 10 20 30 40 50
#
origin time difference, s
SEL3
REB

12. REB and SEL3: 2011070+2011071
Arrival time difference
0
200
400
600
800
1000
0 20 40 60 80 100
#
arrival time difference, s
REB
SEL3
Frequency distribution of time difference between arrival times of sequential arrivals at IMS stations
- The REB events have large number of arrivals at the same station within 20 s (likely added by analysts)
- The SEL3 has a significant leap in arrival time difference around 90 s (DFX rule)
- Small time difference between arrivals in a challenge for any association and conflict resolution process
0
10
20
30
40
50
60
70
80
90
0 20 40 60 80 100
#
arrival time difference, s
AKASG MJAR WRA
KSRS ZALV TXAR
USRK

13. REB arrival time differences
Frequency distribution of time difference between arrival times of sequential arrivals at IMS stations
The number of arrivals at the same station within 20-30 s decreases with time in the Tohoku sequence
The complexity of automatic/interactive processing has to decrease accordingly
0
100
200
300
400
500
0 20 40 60 80 100
#
tdiff, s
2011070
2011071
2011072

14. REB and SEL3: SNR (2011070)
SEL3 does not allow poor detections
Frequency distribution of SNR of first P-arrivals (Pg, Pn, P, PKP, PKPab, PKPbc) at IMS stations
- The REB events have large number of arrivals with SNR values below DFX detection thresholds (added by
analysts) and the peak is between 3 and 4
- The SEL3 distribution has a peak between 4 and 5 and DFX is not able to find any low-SNR arrivals
1.E+00
1.E+01
1.E+02
1.E+03
1.E+04
0 5 10 15 20
#
SNR
REB (first P)
SEL3 (first P)
0.0
0.2
0.4
0.6
0.8
1.0
0 10 20 30 40 50
CDF
SNR
REB (first P)
SEL3 (first P)

15. REB and SEL3: arrival time residuals (2011070)
Distribution of arrival time residuals of the first P-arrivals (Pg, Pn, P, PKP, PKPab, PKPbc) at IMS stations
- The REB events have arrival time residuals for the defining phases within 2 s for P, 2.2 s for Pg/Pn, and no
limit for non-defining phases. The REB pdf demonstrates deficiency of arrivals between 2 and 3 s
- The SEL3 events are characterized by exponential decay in the pdf, which corresponds to the stochastic
distribution of arrival times
1.E-05
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
-15 -10 -5 0 5 10 15
pdf
tres, s
0
1,000
2,000
3,000
4,000
5,000
6,000
7,000
8,000
9,000
-15 -10 -5 0 5 10 15
#
tres, s
ToolargetraveltimeresidualsinSEL3mightleadtocombinatorialexplosion

16. REB and SEL3: 2011070 to 2011076
2065
893
2788
177
186
1331
1442
1436
129
285
288
2123
1389
1181
2780
1338
101
52
1821
2480
1219
118
2480
1585
1877
586
1863
2038
875
518
92
433
2509
880
686
1863
707
2167
75
3194
2137
2520
944
513
1369
27
114
758
334
721
59
157
117
1126
798
571
1419
839
51
5
1315
333
568
59
1279
288
834
373
1044
1083
608
116
10
255
1228
645
246
963
491
1584
95
1503
1013
1161
0
500
1000
1500
2000
2500
3000
3500 AKASG
ARCES
ASAR
BDFB
BOSA
BRTR
BVAR
CMAR
CPUP
DBIC
ESDC
FINES
GERES
GEYT
ILAR
KBZ
KEST
KMBO
KSRS
KURK
LPAZ
MAW
MJAR
MKAR
NOA
NRIK
NVAR
PDAR
PETK
PLCA
ROSC
SCHQ
SONM
STKA
TORD
TXAR
ULM
USRK
VNDA
WRA
YKA
ZALV
#
REB
SEL3
Number of associated arrivals at primary IMS stations for the REB and SEL3 (matching REB by evid)
The most sensitive stations for the aftershocks of the Tohoku earthquake are WRA, ASAR, and ILAR. Stations
MJAR and KSRS are likely too close to the aftershock area, and thus, slightly blinded. Stations FINES, KURK,
PDAR, SONM, USRK, YKA, and ZALV demonstrate good performance.

17. REB and SEL3: 2011070+2011071
5
40
49
67
82
95
99
74
52
60
50
44
32
51
36
54
11
27
8
12
6
10
0
20
40
60
80
100
120
5
10
15
20
25
30
35
40
45
50
55
60
65
70
75
80
85
90
95
100
105
110
115
120
125
130
#
Weight (based on historical station participation rates in REB)
REB
SEL3
0
5
10
15
20
25
30
35
40
45
0 10 20 30 40
REB
nassoc
SEL3 nassoc
Frequency distribution of event weights in the REB and SEL3
Event weights in the REB are significantly higher than in the
SEL3 events with the same evid. Both distributions are bi-modal.
There are several REB events with very low weight considering
all historical REB events within the same area.
The number of associated stations in the
REB are significantly larger than in the SEL3
events with the same evid.
Number of associated stations in the REB
and SEL3

18. Examples of interactive analysis: new REB events
(available – IDC lebview account)
October 5, 2011 57.9526ºN 32.5197ºW,
origin time 23:02:10. REB - 36 aftershocks
The main shock
Date Time mb(IDC) ndef Dist, km
05/10/2011 23:06:39 3.32 4 49.52
05/10/2011 23:51:58 3.31 5 28.16
05/10/2011 23:54:20 3.80 6 38.17
05/10/2011 23:55:51 3.65 8 13.8
06/10/2011 00:01:38 4.02 10 5.36
06/10/2011 00:02:31 3.59 10 4.79
06/10/2011 00:05:23 3.56 7 35.97
06/10/2011 00:10:45 3.46 3 76.62
06/10/2011 00:39:18 3.42 4 36.28
06/10/2011 00:45:54 3.52 8 9.17
06/10/2011 01:07:19 3.46 8 39.36
06/10/2011 02:31:30 3.37 5 31.8
06/10/2011 02:34:16 3.35 6 61.11
06/10/2011 07:36:33 3.32 5 26.43
06/10/2011 07:47:04 3.59 6 56.75
06/10/2011 07:48:25 3.54 9 24.72
06/10/2011 10:37:51 3.61 5 27.02
06/10/2011 13:17:19 3.54 4 11.75
06/10/2011 14:12:58 3.38 6 8.48
06/10/2011 14:17:29 3.51 8 15.83
06/10/2011 19:43:05 3.40 5 30.34
06/10/2011 19:47:26 3.61 5 15.18
06/10/2011 19:57:48 3.56 6 19.81
06/10/2011 21:12:34 3.51 8 32.3
06/10/2011 21:15:49 3.46 5 18.9
06/10/2011 23:49:25 3.62 9 25.47
New REB-ready events
26 new REB-ready events were added to 38 REB
events including the main shock. The Reviewed
Event Bulletin has been extended by 67%:
57
57.5
58
58.5
-36 -34 -32 -30
lat,
deg
lon, deg
20 new
events found
by main
shock
34 REB
events found
by main
shock
6 new events
found in the
second
iteration
3 REB events
found in the
second
iteration

19. The main shock and master events Examples of new REB-compatible events
-2
0
2
4
6
8
86 88 90 92 94 96
lat,
deg
lon, deg
Distribution of REB events (black circles) related
to the Sumatera earthquake (green square) between
April 11 and May 24, 2012. Sixteen master events
are shown by red diamonds. They were preselected
using SEL3 available on April 13 and slightly
reviewed by an experienced analyst. The events at
the far periphery of the aftershock zone are likely
related to low location accuracy for 3 to 5 station
events and probable mis-timing /mis-association of
low-amplitude arrivals.
Date Origin
time
Lat,
deg
Lon,
deg
Depth,
km
nass ndef mb
04/11/2012 8:51:33 2.44 92.59 0 7 7 4.95
04/11/2012 9:32:02 2.20 93.09 0 8 7 4.29
04/11/2012 10:05:05 -0.26 92.59 0 7 7 4.06
04/11/2012 13:42:53 2.10 93.62 0 17 17 4.56
04/11/2012 16:25:48 3.31 92.71 0 7 7 3.88
Overall, experienced analysts reviewed 409 XSEL hypotheses built by
sixteen master events and found 119 new REB events. Assuming
homogeneous distribution of aftershocks over space and similar
performance of all masters, one can estimate the number of new REB-
ready events in the sequence as ~650. With all uncertainty in the
aftershock distributions and the imperfectness of master event choice
and coverage, the number of missed events is likely compared with the
number of REB events. It is highly important that these events are
smaller, and thus, are most challenging for CTBT monitoring. It should
be also mentioned that the reviewed XSEL events are most difficult for
interactive analysis since they are characterized by weaker signals and
smaller number of stations. Most of hypotheses associated with the
REB aftershocks have four and more defining stations from the full set
of seven. Therefore, the REB events are much easier to review
interactively.
Examples of interactive analysis: new REB events
(available – IDC lebview account)

20. Aftershock sequence recovery: general approach
1. Use waveform cross correlation (WCC) with master events from the REB/SEL3 and signals at
primary IMS stations as waveform templates
2. Use WCC programs of the XSEL prototype pipeline
3. Use Local Association and Conflict Resolution programs as in routine prototype XSEL
production
4. Use only the REB masters events from the past: more than 2 days before the data day (48
hours rule for the REB production). Expedite REB creation is also considered.
5. It is allowed to use SEL3 master events from the same day without interactive review
6. An REB event is matched by an event in the XSEL when 2 or more phases are within 10 s
from each other at two or more different stations (statistics with 1 match is also helpful)
7. Compare the number of matched events: the rate of matched events is a priority
8. Calculate total number of events in the XSEL: balance the analysts workload with that
expected with the SEL3 and scanner seed events for interactive analysis (to be determined)
9. Estimate the trade off between the rate of matched and additional XSEL events depending on
Local Association and Conflict Resolution parameters
10. Evaluate the quality of the REB events using Spot Check Tool

21. Routine XSEL processing
WCC detection Local Association Conflict Resolution
1. Varying template length
2. Varying # of working channels
3. Time gap between detections
4. Frequency bands: 0.75 Hz to 8
Hz. Choice depends on station:
ARCES vs. CMAR
5. Adaptive detection (SNRcc=
STA/LTA) threshold for CC
traces
6. STA and LTA definition
7. STA and LTA length
8. FK using CC-traces
9. Standard detection and FK for
array stations (e.g., for
screening of CC-detections)
1. Area of master responsibility
2. Spacing between nodes of virtual
event locations
3. Minimum number of associated
stations (e.g., ndef=3)
4. Association window (e.g., 20 s)
5. Origin time residual for defining/
associated arrivals (e.g., ±4 s)
6. Weight of XSEL events > Wmin as
a sum of station probabilities
(from the REB empirical
statistics)
7. Quality of detections SNRcc
8. Quality of detections SNR
9. Residual of relative magnitude at
stations
When two XSEL events from the same of
different masters have detections at
the same station within tolerance
limit (LA tuning parameter), 20 s :
- Compare weights, Wi and Wj, of the
competing XSEL events and save
this detection with the bigger event.
- When Wi and Wj are equal – the
number of associated stations wins
- When both above equal - standard
deviation of the origin time
(scattering) of the competing XSEL
events
- Then remove events with parameters
below EDC

22. Selected pre-processingelements
1. Select an appropriate set of representative master events: # of
associated stations, SNR of signals, average SNR, etc.
2. Estimate/update the threshold for valid event hypotheses weight in a
given region
3. Check availability and quality of waveform templates
4. Estimate mutual cross correlation quality of repeating events in the past
5. Optional: use dimensionality reduction techniques (e.g., PCA) to
improve cross correlation and suppress noise

23. REB and master (historicalREB) events
REB and historical master events 2011070 REB and (fresh REB) master events 2011073
Red circles – 743 REB events, black squares – 138
master events. Masters – historical REB
Red circles – 444 REB events, black squares – 150
master events. Some masters – REB from 2011070

24. Multichannel
waveform
template
–
high
quality
signal
Continuous multichannel cross correlation. For arrays, only vertical or 3-C channels (e.g., ARCES) are used. For
3-C stations, all three channels can be used, but only vertical channel is preferable at teleseismic distances.
Searching for similar (repeating) signals in continuous waveforms. Matched signals have to be synchronized (in
terms of arrival time difference on individual sensors of an array) with those in the master event at many stations
Waveform Cross Correlation

25. Waveform Cross Correlation: Detection
An example of cross correlation detection: DPRK 2006 vs. DPRK 2009. DPRK06 as a template. Station AKASG,
BP filter between 0.8 Hz and 2.0Hz. Good cross correlation suggests very sharp signals at the average CC-trace
and the most efficient STA can be short, i.e. different from STA for regular (physical) seismic signals. LTA has to
be frozen.
STA
LTA
Threshold: SNR=STA/LTA >3.0
Cross Correlation Coefficient, CC
Valid (matched) signal

26. For all valid arrivals at primary stations, which are
found with a given master event, origin times, OTij, are
calculated. The empirical travel times from the master
event to the relevant primary stations, TTij, are
subtracted from the arrival times, ATij :
OTij = ATij – TTij
Luckily,
TTij = TTj !
Empirical travel times from a master event to seismic
stations are characterized by zero modelling errors and
very low measurement errors. These conditions allow
extremely accurate relative location.
Azimuth and slowness of the CC-detections are
borrowed from the master event. The phase type is
determined by slowness or also inherited: Pg, Pn, P,
PKP, PKPab, PKPbc.
Local Association (LA) and REB EDC
• For cross correlation standard event list, XSEL, one arrival for a
given master can participate only in one event hypothesis
• The XSEL includes events with associated detections at 3 and more
primary stations, i.e. only REB-compatible events are created.
These stations are defining.
XSEL events match REB Event Definition Criteria

27. Additional location grid for LA: relative location
when master and XSEL events are far away
MASTER
EVENT
-10
-5
0
5
10
-10 -5 0 5 10
N
E
dtk = S · dk – travel time correction
OTk
ij = ATij - TTj + dtjk – corrected origin time
From origin time residuals to relative location
• k nodes, rectangular or circles;
• grid size from ~1 km to 500 km;
• spacing from meters to 10-20 km
• Average OT and RMS OT residual are calculated
in each node
• No depth resolution! Same as in master event.
s
Relative location - minimum RMS(resOTk
ij )

28. # MASTER HH MM SS
OT
STDEV LAT,deg LON,deg D, km D_diff MB
MB
STVEV NASS
NEW/
REB REB_ORID
#
MATCH EDC
EV
WEIGHT REB
35 5146632 12 38 17.24 0.97 37.644 141.447 46.2 0 4.43 0.14 10 SEL 7314825 9 23.7 46.5 83.2
40 7314825 12 38 5.38 -1 36.964 142.159 0 -1 4.62 -1 18 REB
227 7253528 12 38 41.63 -1 37.84 140.52 219.5 -1 4.34 -1 8 SEL
STA DELTA SEAZ HH MM SS.SS tres, s TT EVENT SNRCC SNR dRM CC
STA
WEIGHT
REB
res, s
MJAR 2.79 80.1 12 38 59.43 -1.133 43.3 6.7 1.7 4.34 0.148 8.5 -1.44
USRK 10.32 131.2 12 40 35.01 -0.320 137.5 6.1 2.1 4.56 0.325 3.3 8.20
KSRS 11.01 88.4 12 40 53.47 0.799 155.1 6.3 3.2 4.58 0.226 4.0 -4.45
SONM 28.00 99.8 12 44 0.01 1.146 341.6 8.9 12.9 4.51 0.592 5.5 0.62
ZALV 42.12 90.0 12 46 1.3 1.088 462.5 5.2 10.4 4.60 0.494 4.1 0.10
CMAR 41.68 54.8 12 46 2.62 1.093 464.2 5 2.6 4.21 0.322 3.0 -99.00
KURK 46.21 81.9 12 46 31.63 -1.409 495.1 4.8 6.3 4.56 0.322 3.2 2.75
ILAR 49.45 272.0 12 46 59.31 -0.164 521.7 6.9 2.4 4.30 0.411 5.5 -3.85
FINES 68.98 50.7 12 49 12.47 0.201 655.0 6.4 6.7 4.29 0.197 5.9 -0.25
AKASG 74.49 50.2 12 49 45.42 -1.300 689.4 5.2 4.7 4.35 0.157 3.5 0.38
# MASTER HH MM SS
OT
STDEV LAT,deg LON,deg D, km D_diff MB
MB
STVEV NASS
NEW/
REB REB_ORID
#
MATCH EDC
EV
WEIGHT REB
36 3630439 12 40 41.6 0.17 37.34 142.078 80 0 3.94 0.07 4 REB 7319395 3 23.7 24.2 42.6
132 7319395 12 40 31.71 -1 37.218 142.126 0 -1 4.23 -1 8 REB
STA DELTA SEAZ HH MM SS.SS tres, s TT EVENT SNRCC SNR dRM CC
STA
WEIGHT
REB
res, s
SONM 27.73 99.7 12 46 25.03 -0.038 343.6 5.3 3.7 3.95 0.347 5.5 1.24
WRA 57.19 6.8 12 50 21.02 -0.254 579.9 5.4 4 3.98 0.254 8.5 0.58
ASAR 60.91 6.9 12 50 47.39 0.202 605.9 4.8 2.9 4.01 0.248 6.8 -99.00
NVAR 74.90 305.5 12 52 12.42 0.090 690.7 6.1 2.8 3.83 0.390 3.4 -0.14
# MASTER HH MM SS
OT
STDEV LAT,deg LON,deg D, km D_diff MB
MB
STVEV NASS
NEW/
REB REB_ORID
#
MATCH EDC
EV
WEIGHT REB
211 5579102 15 28 10.14 1.24 36.811 141.312 62.1 0 3.91 0.25 7 NEW 0 0 23.8 29.4 0
STA DELTA SEAZ HH MM SS.SS tres, s TT EVENT SNRCC SNR dRM CC
STA
WEIGHT
REB
res, s
KSRS 10.75 96.6 15 30 40.43 -1.361 152.0 4.9 2.2 4.32 0.232 3.9 -99.00
PETK 21.21 220.4 15 32 36.45 0.440 266.0 6.2 2.1 3.85 0.326 3.0 -99.00
MKAR 44.68 82.7 15 36 13.32 1.210 482.3 4.5 0 3.70 0.156 7.7 -99.00
KURK 46.72 83.9 15 36 27.16 0.057 497.0 4.4 2.1 -2.22 0.195 3.2 -99.00
ILAR 51.00 271.6 15 36 54.18 -2.206 526.6 6.2 3 3.88 0.600 5.6 -99.00
BVAR 51.39 80.2 15 37 4.07 1.523 532.8 4.5 2.1 -1.98 0.188 3.0 -99.00
STKA 66.99 359.5 15 39 4.04 0.338 654.0 4.8 2.3 4.20 0.499 3.0 -99.00
Examples of XSEL event hypotheses
Three examples of XSEL event hypotheses for
2011070 and their comparison with the REB and
SEL3:
The upper event has arrivals at 10 associated
stations, nine of them are within 10 s from the
corresponding REB arrivals. The XSEL event was
obtained by master orid=5146632. The REB event
has 18 associated stations and was built from a
SEL3 event (same evid) with 8 associated stations.
The REB and SEL3 solutions are presented. The
XSEL event weight 46.5 and the REB events has
weight 83.2
Second XSEL event has 4 associated stations, three
of them are REB matching arrivals. This REB was
built by analysts.
The lower event is a new one without any REB
matching phases. Its weight 29.4, which is larger
than the threshold of 23.8.

29. XSEL and master events coverage
2011070 2011073
The 2011070 XSEL (red dots) covers broader area
together with master events (black squares). Each
master event covers a circle with 180 km in radius.
The 2011073 XSEL is more compact. Several master
events at the periphery do not find any XSEL events.

30. XSEL - REB
2011070 2011073
The 2011070 XSEL (red squares) well covers the area
of REB events (blue circles)
The 2011073 XSEL is more compact and is based on
the fresh REB-2011070 events.

31. XSEL to REB comparison: 2011070+2011071
1.00E-04
1.00E-03
1.00E-02
1.00E-01
1.00E+00
2 3 4 5 6 7
pdf
mb, km
XSEL (days 2011070+2011071)
SEL3
REB
1.00E-04
1.00E-03
1.00E-02
1.00E-01
1.00E+00
0 100 200 300 400 500 600 700
pdf
depth, km
XSEL (days 2011070+2011071)
SEL3
REB
The XSEL for two days with the most
intensive seismicity has magnitude
distribution similar to the REB and
SEL3 for the same days except dozens
of low magnitude (2 to 3) events.
These small events are candidates to be
rejected by magnitude threshold. This
procedure may reduce the number of
XSEL event hypotheses without loss of
REB matching events.
The depth distribution of XSEL events
is somewhere between these for REB
and SEL3. Such a distribution looks
most reasonable from physical point of
view and considering the IDC zero
depth rule. The XSEL events depth is
the same as in corresponding masters
and thus the distribution is not smooth
as the SEL3 one.

32. XSEL to REB comparison: 2011070+2011071
The XSEL and REB for two days have
similar distribution of origin time
difference above 40 s. Below 40 s, the
XSEL has much larger number of
events. This is curse of automatic
processing and the reason of the 90 s
rule for SEL3. To match the REB
events with close origin times and
arrival times at IMS stations, the
Conflict Resolution procedure has
very narrow (10s) window for arrival
comparison. In this situation, many
similar events are created by different
masters, and these events look like
split events in the REB.
The distance between the REB-
matching XSEL events and
corresponding REB events is much
lower than that for the SEL3. The
depth of XSEL is a reason of the
location difference.
1.E+00
1.E+01
1.E+02
1.E+03
0 100 200 300 400
#
origin time difference, s
XSEL (days 2011070+2011071)
SEL3
REB
1.E-04
1.E-03
1.E-02
1.E-01
1.E+00
0 5 10 15
pdf
distance, degrees
SEL3/REB distance (total 1099 evid
matched, 34 between 16 and 180
degrees)
XSEL/REB distance (1435 events 2+
stations matched)

33. XSEL events vs. 170 master events
1 2 1 5 4 1
26
4 3 4 11 4 5 5 1 3 5 2 4
37
7 5 2 1 1 2 2 3 4 2 7 15 22 11 4 10
27 17
1 9
22
4 12 12 6 12 3 1
39
9 14 10 1
37
10
23
157
0
50
100
150
200
2028064
2604362
2727886
2799817
3177488
3249904
3259547
3310612
3364090
3438674
3457374
3475712
3476405
3490685
3532601
3547879
3614694
3684499
3746744
3883968
4048126
4163880
4248354
4282557
4302220
4315590
4500219
4534999
4558103
4599624
4639635
4665958
4695402
4713649
4731959
4733764
4736552
4745421
4775048
4795947
4842130
4845952
4849634
4880765
4918874
4946519
4973353
4995126
5008223
5064540
5065746
5065835
5084439
5086677
5109998
5114453
5114746
4 5
10 13
22
17
7
65
16
44
62
7 6
12 12 16
52
6
16
5 2
8
14
66
2
16
10
17
6 9
2
50
10
4
39
22
6 5
11
33
2
7 8
20
3
57
4
15
47
9
4 7 3
15
2
22 18
0
10
20
30
40
50
60
70
5116774
5117283
5124691
5131661
5144711
5146632
5169658
5177063
5177618
5185670
5234339
5348450
5357792
5357844
5359131
5359691
5371655
5372040
5413985
5421442
5442380
5445440
5445944
5446043
5516791
5566912
5574462
5579102
5580737
5583159
5586870
5606337
5690229
5699331
5759897
5766718
5766722
5767364
5808213
5834393
5906082
5931881
5933344
6033058
6046778
6102469
6107267
6115038
6115702
6285222
6297984
6307989
6360192
6383833
6390833
6423351
6456825
18
12 15 14
49
8
52
1 4 4
27
2
29
12
27
6 4
10 8
25
5
19
5
17
21
53
23
1 1
11 9
17 16 15
1 1
29
18
22
1
5
15 13 12 14
2
39
19
6
14
2 2
25
37
10
5
12
6
0
10
20
30
40
50
60
6456825
6463260
6480571
6480776
6480847
6541207
6551342
6577414
6577725
6578718
6601425
6636645
6637401
6643212
6678527
6715522
6731526
6737613
6763824
6824225
6854873
6883829
6914290
6917227
6925501
6946758
7052906
7090820
7103359
7117441
7139253
7153977
7154647
7154649
7171368
7176481
7178960
7194775
7198939
7201511
7224404
7224406
7224465
7224893
7227056
7227411
7228359
7228660
7229218
7229279
7251911
7252081
7252130
7253253
7253338
7253392
7256467
7258239

34. Statistical results: Tohoku
Tohoku REB
#
masters XSEL
XSEL/
REB
REB
match
%REB
match
Quality
check, SC
QC-SC REB
match
% SC REB
match
LA REB
match
% LA REB
match
2011070 743 138 987 1.33 549 0.74 866 689 0.93 635 0.85
2011071 714 193 1470 2.06 611 0.86 815 656 0.92 675 0.95
2011072 574 203 1242 2.16 505 0.88 657 563 0.98 548 0.95
2011073 444 150 981 2.21 413 0.93 556 437 0.98 431 0.97
2011074 392 214 762 1.94 356 0.91 466 389 0.99 373 0.95
2011075 321 159 551 1.72 284 0.88 368 316 0.98 302 0.94
2011076 322 214 611 1.90 274 0.85 356 308 0.96 287 0.89
The Tohoku aftershock sequence is
the largest by the number and daily
rate of REB events. Overall success
(REB matching) rate is much larger
than in the SEL3 (from 74% to
93%). On 2011070, the automatic
Spot Check Tool confirmed 689
from 743 events (93%). Therefore,
54 events might be revisited with
interactive (spot check) analysis to
re-confirm. For the days 3 through
7, the QC confirmation rate was
around 98%.
During LA processing, before CR,
there were 635 REB events
matched by preliminary events
hypotheses (85%). The success rate
of LA match is high for all days
except the first one, likely because
of lower ambient noise level.
When the fresh REB events from
the same sequence were used on
days 4 to 7, the success rate
increased to 93%, with the total
number of XSEL events two times
larger when in the REB.
Overall, the matching rate and the
excess of XSEL events demonstrate
good performance.

35. Statistical results: Nepal
Nepal REB
#
masters XSEL
XSEL/
REB
REB
match
%REB
match
Quality
check, SC
QC-SC REB
match
% SC REB
match
LA REB
match
% LA REB
match
2015115 211 55 332 1.57 170 0.81 200 185 0.88 179 0.85
2015116 83 62 126 1.52 75 0.90 92 78 0.94 77 0.93
2015117 49 62 68 1.39 44 0.90 51 49 1.00 44 0.90
2015118 26 24 44 1.69 25 0.96 27 26 1.00 26 1.00
2015119 22 35 59 2.68 21 0.95 22 22 1.00 22 1.00
2015120 16 44 42 2.63 15 0.94 17 14 0.88 15 0.94
2015121 13 45 54 4.15 13 1.00 13 13 1.00 13 1.00
The success (REB matching) rate is
high from 81% to 100%. This
aftershock sequence was well
covered by the IMS seismic
network. Only the first day with
the main shock is a challenge for
processing with 211 events in total,
i.e. less than on the 7th day of the
Tohoku sequence. The automatic
Spot Check Tool confirmed 88% of
the events on 2015155, and
practically all events during other 6
days. During LA processing, there
were 179 from 211 REB events
matched by preliminary events
hypotheses. When REB events
from the same sequence were used
on days 4 to 7, the success rate
increased to 95% - 100%. For the
first 4 days, the number of XSEL
events was only by ~ 50% to larger
than the number of REB events.
During the last 3 days, the excess
of XSEL events was 150% to
300%, with the total number of
XSEL events ~50.
Overall, the matching rate and the
excess of XSEL events are close to
the most efficient performance.

36. Statistical results: Chile
The overall success (REB
matching) rate is much larger than
in the SEL3 (from 69% to 88%).
The first 66 min after the main
shock on 2011259 gave 36 REB
events. Considering the fact that the
closest IMS stations are all 3-C,
this was a really difficult problem.
Other days were not so hard to
match and the success rate was
above 80%. (except 2015063).
During LA processing, there were
102 REB events matched by
preliminary event hypotheses, 3 of
which were rejected by CR and
were not promoted to the XSEL.
When the REB events from the
same sequence were used on days 4
to 7, the success rate increased to
95%, with the total number of
XSEL events two times larger when
in the REB. For the first 3 days, the
number of XSEL events was only
by~ 25% to 35% larger.
Overall, the matching rate and the
excess of XSEL events are close to
the most efficient performance.
Chile REB
#
masters XSEL
XSEL/
REB
REB
match
%REB
match
Quality
check, SC
QC-SC REB
match
% SC REB
match
LA REB
match
% LA REB
match
2015259 36 41 54 1.50 25 0.69 36 32 0.89 29 0.81
2015260 312 87 482 1.54 257 0.82 332 290 0.93 276 0.88
2015261 119 102 169 1.42 104 0.87 129 110 0.92 107 0.90
2015262 82 59 102 1.24 72 0.88 90 78 0.95 73 0.89
2015263 64 134 77 1.20 44 0.69 69 58 0.91 45 0.70
2015264 48 95 77 1.60 42 0.88 49 44 0.92 42 0.88
2015265 37 36 50 1.35 32 0.86 35 35 0.95 32 0.86

37. Statistical results: PNG
The overall success (REB
matching) rate is high for this
aftershocks sequence (from 81% to
95%), which is not a challenge by
the number of events per hour and
total number. The highest rate was
on 2018056 and this day is
characterized by the lowermost rate
of matched events of 81%. The
automatic Spot Check Tool
confirmed only 103 from 122
events for this day. Therefore, 19
events might to be revisited with
interactive analysis During LA
processing, there were 102 REB
events matched by preliminary
events hypotheses.
When the REB events from the
same sequence were used on days
4 to 7, the success rate increased to
>94%, with the total number of
XSEL events two times larger
when in the REB. For the first 3
days, the number of XSEL events
was only by~ 25% to 35% larger.
Overall, the matching rate and the
excess of XSEL events are close to
the most efficient performance.
PNG REB
#
masters XSEL
XSEL/
REB
REB
match
%REB
match
Quality
check, SC
QC-SC REB
match
% SC REB
match
LA REB
match
% LA REB
match
2018056 122 60 128 1.05 99 0.81 105 103 0.84 102 0.84
2018057 195 83 243 1.25 170 0.87 176 178 0.91 172 0.88
2018058 90 64 124 1.38 76 0.84 93 89 0.99 77 0.86
2018059 71 38 109 1.54 67 0.94 69 69 0.97 67 0.94
2018060 53 66 104 1.96 50 0.94 53 53 1.00 50 0.94
2018061 56 78 110 1.96 53 0.95 55 55 0.98 53 0.95
2018062 33 27 70 2.12 31 0.94 33 33 1.00 31 0.94

38. Tuning LA and CR: Tohoku mostintensive days
Optimal setting of LA and CR parameters is a challenge for the largest aftershock
sequences. The first 3 days of the Tohoku sequence is the best example of potential
tuning benefits. The effect of increasing or decreasing thresholds can be
counterintuitive when the flux of detections and origin time is saturated and XSEL
event hypotheses compete with many rivals.
We have changed the
following LA and CR
parameters:
1. area of LA from 60 km to
300 km in radius;
2. Travel (origin) time
residual from 3 s to 5 s;
3. Window for CR from 10 s
to 60 s;
4. The max SNRcc in the
event – from 5.5 to 6.5;
5. Minimum weight from 20
to 24;
6. Minimum sum of SNRcc
in a 3-station event from 14
to 17.
Tohoku REB
#
masters XSEL XSEL/REB REB match
%REB
match
LA REB
match
% LA REB
match
Standard LA and CR parameters
2011070 743 138 987 1.33 549 0.74 635 0.85
2011071 714 193 1470 2.06 611 0.86 675 0.95
2011072 574 203 1242 2.16 505 0.88 548 0.95
Lowered thresholds for LA and CR parameters
2011070 743 138 1208 1.63 587 0.79 709 0.95
2011071 714 193 1423 1.99 623 0.87 696 0.97
2011072 574 203 1341 2.34 504 0.88 555 0.97
Increased thresholds for LA and CR parameters
2011070 743 138 725 0.98 530 0.71 654 0.88
2011071 714 193 758 1.06 544 0.76 657 0.92
2011072 574 203 792 1.38 479 0.83 547 0.95

39. Further tuning LA and CR: Tohoku mostintensive days
1. Change STA = 0.8 s for more effective (CC-) noise suppression
2. Vary CC-detection thresholds (from 2.4 to 3.4)
3. Filter CC-detections with standard SNR
4. Vary LA parameters – grid, thresholds in the EDC
5. Vary Conflict Resolution parameters, e.g. time window, event
magnitude
6. Select various sets of masters for various scenarios of larger
aftershock processing
7. Compare with REB (and SEL3) for optimal set of parameters

40. Further tuning LA and CR: Tohoku mostintensive days
Alternative strategies to reduce the IDC analyst workload and
corresponding sets of master events
1. Historical REB as master events (as many as needed, up to 400, tested before)
2. Best historical REB as master events selected by CC-procedure (126)
3. SEL3 biggest events (122)
4. SEL3 events distributed over regular (global) grid (102)
5. REB events built from the set of SEL3 events regularly distributed over the
grid (102)
6. REB events near regular grid (65)
7. REB events built from the biggest SEL3 events (159)
8. Benchmark: Best REB for the given day as master events selected by CC-
procedure (126)

41. Further tuning LA and CR: Tohoku 2011070
REB REGULAR LA and CR setting
STATISTICS OF XSEL-REB MATCH
MASTERS CASE Threshold LA CR Setting #REB (6-24) #XSEL #REB/#XSEL #REB Match
Match XSEL/
REB
LA Match LA Match
SEL3 GRID 3.2 REGULAR 734 647 0.88 553 0.75 717 0.98
SEL3 GRID 3.0 REGULAR 734 658 0.90 568 0.77 725 0.99
SEL3 GRID 2.8 REGULAR 734 684 0.93 582 0.79 726 0.99
SEL3 GRID 2.4 REGULAR 734 839 1.14 602 0.82 732 1.00
SEL3 BIGGEST 3.2 REGULAR 734 671 0.91 555 0.76 723 0.99
SEL3 BIGGEST 3 REGULAR 734 692 0.94 571 0.78 727 0.99
SEL3 GRID 3.2 LOWER 734 738 1.01 561 0.76 722 0.98
SEL3 GRID 3.0 LOWER 734 778 1.06 560 0.76 731 1.00
SEL3 GRID 2.8 LOWER 734 841 1.15 588 0.80 731 1.00
SEL3 GRID 2.4 LOWER 734 990 1.35 609 0.83 734 1.00
SEL3 BIGGEST 3.2 LOWER 734 773 1.05 575 0.78 726 0.99
SEL3 BIGGEST 3.0 LOWER 734 825 1.12 577 0.79 731 1.00
SEL3 BIGGEST 2.4 LOWER 734 951 1.30 603 0.82 733 1.00
SEL3 GRID 2.4 LOWEST 734 1321 1.80 639 0.87
SEL3 BIGGEST 2.4 LOWEST 734 1317 1.79 637 0.87

42. Further tuning LA and CR: Tohoku 2011070
REB REGULAR LA and CR setting
MASTERS CASE Threshold LA CR Setting #REB (6-24) #XSEL #REB/#XSEL #REB Match
Match XSEL/
REB
LA Match LA Match
REB BEFORE 3 REGULAR 734 836 1.14 584 0.80 731 1.00
REB BEFORE 2.8 REGULAR 734 808 1.10 587 0.80 730 0.99
REB BEFORE 2.4 REGULAR 734 1131 1.54 619 0.84 733 1.00
REB FROM_SEL3_GRID 2.4 REGULAR 734 926 1.26 618 0.84 733 1.00
REB FROM_SEL3_GRID 3 REGULAR 734 737 1.00 573 0.78 729 0.99
REB GRID 2.4 REGULAR 734 893 1.22 611 0.83 734 1.00
REB GRID 2.6 REGULAR 734 819 1.12 597 0.81 732 1.00
REB GRID 2.8 REGULAR 734 712 0.97 588 0.80
REB GRID 3 REGULAR 734 705 0.96 583 0.79
REB GRID 3.2 REGULAR 734 660 0.90 569 0.78
REB GRID 3.4 REGULAR 734 639 0.87 559 0.76
REB BIGGEST 2.6 REGULAR 734 901 1.23 608 0.83 733 1.00
REB BIGGEST 3 REGULAR 734 773 1.05 587 0.80 733 1.00
REB BEST 2.4 REGULAR 734 941 1.28 611 0.83 731 1.00
REB BEST 2.8 REGULAR 734 796 1.08 592 0.81 733 1.00
REB BEST 3 REGULAR 734 752 1.02 578 0.79 733 1.00
STATISTICS OF XSEL-REB MATCH

43. Further tuning LA and CR: Tohoku 2011070
REB LOWERLA and CR setting
MASTERS CASE Threshold LA CR Setting #REB (6-24) #XSEL #REB/#XSEL #REB Match
Match XSEL/
REB
LA Match LA Match
REB BEFORE 3.0 LOWER 734 1018 1.39 589 0.80
REB BEFORE 2.8 LOWER 734 1003 1.37 589 0.80
REB BEFORE 2.4 LOWER 734 1423 1.94 629 0.86 734 1.00
REB FROM_SEL3_GRID 2.4 LOWER 734 1110 1.51 623 0.85 734 1.00
REB FROM_SEL3_GRID 3.0 LOWER 734 874 1.19 582 0.79 732 1.00
REB GRID 2.4 LOWER 734 1089 1.48 618 0.84 734 1.00
REB GRID 2.6 LOWER 734 972 1.32 611 0.83 734 1.00
REB GRID 2.8 LOWER 734 929 1.27 579 0.79 731 1.00
REB GRID 3.0 LOWER 734 824 1.12 578 0.79 729 0.99
REB GRID 3.2 LOWER 734 807 1.10 591 0.81 725 0.99
REB GRID 3.4 LOWER 734 693 0.94 559 0.76 711 0.97
REB BIGGEST 2.6 LOWER 734 1062 1.45 618 0.84 725 0.99
REB BIGGEST 3.0 LOWER 734 906 1.23 598 0.81 734 1.00
REB BEST 2.4 LOWER 734 1113 1.52 616 0.84 732 1.00
REB BEST 2.8 LOWER 734 968 1.32 611 0.83 733 1.00
REB BEST 3.0 LOWER 734 877 1.19 586 0.80 734 1.00
REB BEFORE 3.0 LOWEST 734 1559 2.12 624 0.85
REB FROM_SEL3_GRID 2.4 LOWEST 734 1523 2.07 632 0.86
REB GRID 2.4 LOWEST 734 1378 1.88 645 0.88
REB BIGGEST 2.6 LOWEST 734 1342 1.83 626 0.85
REB BEST 2.4 LOWEST 734 1551 2.11 639 0.87

44. Examples of XSEL event hypotheses
# master hh mm ss RMS res, s lat lon depth mb stdev mb Nass Nmatch NREB NEW/REB orid
6 7336003 6 9 13.71 2.19 39.706 142.99 49.6 5.94 0.27 11 0 0 NEW 0
sta delta azim hh mm ss res,s tt SNRcc SNR SNRIDC dRM CC ST_WGHT res_REB filter
MJAR 4.16 51.05 6 10 24.9 -0.777 71.9 2.7 5.9 -1 1.583 0.101 10.0 -99 1
SONM 27.36 95.29 6 14 57.8 2.843 341.15 3.5 10.9 -1 1.708 0.275 7.9 -99 3
ZALV 41.15 87.00 6 16 47.89 -3.393 457.63 2.5 12.8 -1 1.9 0.236 8.9 -99 2
ILAR 47.36 272.86 6 17 39.96 2.487 503.92 2.7 8.7 -1 1.599 0.378 8.0 -99 2
WRA 59.19 7.30 6 19 10.27 -3.101 599.96 3.8 1.8 -1 1.479 0.25 9.8 -99 3
YKA 61.66 300.48 6 19 20.16 -0.356 607.06 2.8 9.2 -1 1.521 0.299 4.7 -99 2
ASAR 62.91 7.42 6 19 39.27 1.114 624.99 2.9 6.9 -1 1.38 0.185 7.9 -99 5
STKA 70.59 0.66 6 20 29.37 2.193 673.92 3.2 3.3 -1 1.416 0.471 1.8 -99 3
NVAR 73.15 306.62 6 20 33.33 -2.834 682.82 2.8 4.8 -1 1.073 0.343 3.8 -99 3
BOSA 127.69 61.08 6 28 16.73 1.483 1141.53 3.7 2.5 -1 0.955 0.571 0.5 -99 2
PLCA 154.74 277.06 6 29 8.14 0.342 1194.29 3 2.9 -1 1.18 0.475 0.5 -99 2
# master hh mm ss RMS res, s lat lon depth mb stdev mb Nass Nmatch NREB NEW/REB orid
7 7336208 6 11 22.25 1.46 37.004 141.251 58.8 5.51 0.15 12 9 16 REB 7321927
sta delta azim hh mm ss res,s tt SNRcc SNR SNRIDC dRM CC ST_WGHT res_REB filter
MJAR 3.10 66.07 6 12 1.14 2.1 37.25 3 10.7 -1 1.31 0.091 10.0 -99 2
KSRS 11.06 84.27 6 13 51.23 -1.713 151.1 2.7 2.1 2.8 0.956 0.132 6.8 2.93 2
SONM 27.67 98.28 6 17 6.34 1.231 342.73 2.9 3.6 4.2 0.92 0.178 7.6 3.7 4
CMAR 41.99 53.75 6 18 59.72 -0.287 458.22 3.5 2.5 2.7 0.994 0.094 3.1 0.36 2
ZALV 41.69 88.98 6 19 3.68 -1.992 463.82 2.9 2.8 -1 0.906 0.239 8.7 -99 4
ILAR 48.73 272.39 6 20 8.28 1.761 524.12 4.1 2.5 3.2 0.885 0.315 7.9 -0.37 4
GEYT 63.78 61.46 6 21 48.47 0.712 625.77 3 2.3 -1 0.681 0.129 2.5 -99 3
YKA 63.04 300.15 6 21 49.13 1.321 626.01 3.9 2.8 3.2 0.926 0.299 4.6 -0.71 2
NVAR 74.31 305.77 6 22 57.03 -1.434 696.67 2.6 3.7 5.7 0.797 0.187 3.8 -0.4 5
BRTR 78.63 50.21 6 23 17.46 -0.864 716.4 3 1.8 4.2 0.863 0.416 3.6 3.58 2
TXAR 89.44 313.83 6 24 16.79 1.037 773.35 2.6 2.6 2.6 0.77 0.176 4.9 -3.13 3
LPAZ 146.06 314.85 6 30 56.91 -1.872 1176.63 4.4 10.2 12.7 0.916 0.658 3.6 -0.2 5
# master hh mm ss RMS res, s lat lon depth mb stdev mb Nass Nmatch NREB NEW/REB orid
8 7336059 6 13 1.79 1.01 37.339 141.688 42.7 5.78 0.46 18 17 32 SEL 7321936
sta delta azim hh mm ss res,s tt SNRcc SNR SNRIDC dRM CC ST_WGHT res_REB filter
MJAR 2.94 74.72 6 13 45.29 -1.307 44.95 3.1 4.5 2 1.228 0.095 10.0 -4.7 2
KSRS 11.08 86.78 6 15 40.65 1.232 157.88 5.2 3.8 1.5 1.291 0.207 6.8 -6.56 3
SONM 27.92 99.16 6 18 46.54 -0.16 345.24 4.1 9.3 8.7 1.365 0.369 7.6 -0.3 3
ZALV 42.00 89.53 6 20 45.85 -0.529 464.46 2.7 7.3 6.7 1.146 0.183 8.7 0.6 5
CMAR 41.85 54.44 6 20 47.17 -0.235 465.44 5.6 10.6 10.6 1.601 0.329 3.1 -0.45 2
NRIK 43.20 108.92 6 20 57.06 0.358 474.92 3.8 2.2 5.2 1.407 0.349 1.5 -1.3 3
ILAR 49.15 272.07 6 21 45.18 0.347 522.81 4.5 23.9 27.8 1.612 0.598 7.9 -0.55 2
WRA 57.32 7.07 6 22 43.84 -1.832 583.71 6.9 12 -1 2.808 0.228 10.0 -99 3
YKA 63.47 299.89 6 23 27.35 0.229 625.18 4.7 19.8 19.2 1.506 0.349 4.6 -0.71 2
GEYT 64.00 61.94 6 23 29.99 -0.229 628.44 3.5 7.5 5.9 1.112 0.299 2.5 -0.15 2
NVAR 74.61 305.38 6 24 37.88 0.196 696.11 3.7 12.8 11.3 1.031 0.342 3.8 -1.05 4
BRTR 78.94 50.58 6 25 0.29 -0.21 718.57 4.4 7.1 8 1.223 0.414 3.6 -0.01 4
ULM 79.15 316.64 6 25 1.61 0.032 719.88 3 5.6 8.7 1.186 0.524 1.3 -1 3
SCHQ 85.04 337.43 6 25 31.27 0.099 749.23 3.2 6.1 10.2 1.145 0.351 1.1 1.05 4
ESDC 97.23 26.91 6 26 29.28 -0.051 807.78 2.7 4 6.8 0.772 0.027 0.8 -0.84 3
VNDA 115.16 342.41 6 31 39.1 3.207 1114.35 3.6 2.1 6.8 0.489 0.608 0.2 -3.1 2
DBIC 126.35 32.99 6 31 59.53 -0.467 1138.1 6.5 7.3 16.6 1.089 0.633 0.7 -0.5 4
LPAZ 146.32 314.11 6 32 38.42 -0.679 1177.46 3.2 9.8 11.7 0.972 0.669 3.6 0.3 2
The “NEW” XSEL event does not
match any REB event by two phases,
but has larger standard SNR (as
calculated by XSEL internal procedure
similar to that in the DFX – column
SNRIDC) than the XSEL event
“REB”, which matches the REB event
with orid=7321927 built by analysts
from scratch. The XSEL event marked
as “SEL” matches the REB event
(orid=7321936) built from a SEL3
seed event (same evid) and has the
largest SNR and SNRcc values.
The “NEW” XSEL events are likely
much better candidates for new REB
events than the seed events created by
scanner.

45. XSEL à la carte
1. Master set: historical REB, SEL3, expedite REB events built from
SEL3
2. Detection: threshold and various attributes control the number of
phases to associate
3. Local Association: control the number of events per master
4. Conflict Resolution: control the distance between XSEL events in
time and space
5. Spot Check: find better solutions for / reject poor XSEL events
Target function – analysts workload

46. • Waveform cross correlation (WCC) is used to detect signals and build aftershock events in four large
aftershock sequences with the area up to 100,000 km2
• This aftershock study uses routine XSEL processing. Historical master events from the REB are allowed as
well as contemporary (not reviewed) SEL3 events. The choice can be made automatically.
• Large number of master events (up to 500) was used for the largest region of the Tohoku earthquake
aftershocks
• Strategies with expedite creation of a reasonable number of REB events are considered
• For the largest aftershock sequences, the preliminary estimates of the XSEL-REB daily match rate varies from
70% to 100%
• The XSEL is closer to the REB than the SEL3 and suggest less work (must be quantitatively estimated) on
event creation/confirmation and event rejection (e.g., scanner)
• The use of the best SEL3 events is almost as efficient as the REB masters
• The distribution of master events over a regular grid is likely the best choice
• Further improvements are possible without large additional computations. The learning curve is still steep.
Discussion