Cross-correlation
The purpose of the cross-correlation is to refine the exact time difference between two picks at a common station of an event pair, which allows for higher resolution event (re)locations. This time difference becomes the observed differential time part of a double-difference (see Relocation Process). When the cross-correlation is disabled or the correlation coefficient is below the configured threshold then the observed differential times are derived from the phase times (difference of their absolute travel times).
The cross-correlation step is optional due to the expensive computation time. The improvement in hypocenter solution quality doesn’t always improve drammatically, so we suggest to enable it only after the other parameters of rtDD have been optimized. The cross-correlation can be disabled by setting the crossCorrelation.maxStationDistance and/or crossCorrelation.maxInterEventDistance to 0. The cross-correlation is also skipped if the RecordStream parameter is not properly configured, because the waveform data will not be available.
Eval-xcorr command
The --eval-xcorr option should be used to properly configure the cross-correlation parameters. The optimization process involves running --eval-xcorr with different configuration and analyzes the results. The goal is to have as many matches as possible (increase correlation coefficient) avoiding bad/false matches (lag values higher than the expected pick time uncertainty are probably an indication of false matches): this is a trade-off.
Example of --eval-xcorr command:
scrtdd --eval-xcorr station.csv,event.csv,phase.csv --profile myProfile --verbosity=3 --console=1
Example output:
[...]
---FINAL STATS--
Total xcorr P phases #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
41877 61126 0.77 (0.10) 0.77 (0.10) -0 ( 36) -0 ( 15)
Total xcorr S phases #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
30730 66292 0.85 (0.06) 0.85 (0.06) -1 ( 42) -0 ( 28)
Xcorr P phases by inter-event distance in 0.10 km step
EvDist [km] #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
0.00-0.10 4764 4508 0.82 (0.10) 0.84 (0.09) 0 ( 28) -0 ( 11)
0.10-0.20 13906 18615 0.78 (0.10) 0.78 (0.10) 0 ( 32) 0 ( 14)
0.20-0.30 13633 21365 0.76 (0.10) 0.76 (0.09) -1 ( 41) 0 ( 17)
0.30-0.40 5445 9591 0.75 (0.09) 0.75 (0.09) -5 ( 44) -1 ( 20)
0.40-0.50 2360 3807 0.75 (0.10) 0.75 (0.09) 5 ( 40) 1 ( 18)
0.50-0.60 1010 1727 0.75 (0.10) 0.75 (0.10) 9 ( 35) 3 ( 14)
0.60-0.70 553 1114 0.75 (0.10) 0.73 (0.10) 3 ( 37) 1 ( 14)
0.70-0.80 142 247 0.74 (0.11) 0.70 (0.09) 20 ( 37) 4 ( 10)
0.80-0.90 60 142 0.76 (0.12) 0.77 (0.13) 14 ( 36) 2 ( 10)
0.90-1.00 4 10 0.79 (0.11) 0.80 (0.10) 25 ( 28) 7 ( 1)
Xcorr S phases by inter-event distance in 0.10 km step
EvDist [km] #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
0.00-0.10 4024 4472 0.89 (0.06) 0.91 (0.05) -0 ( 37) 0 ( 25)
0.10-0.20 10521 20221 0.85 (0.06) 0.86 (0.06) 0 ( 40) -0 ( 28)
0.20-0.30 9961 23181 0.84 (0.06) 0.84 (0.05) -0 ( 45) -0 ( 31)
0.30-0.40 3802 10307 0.83 (0.06) 0.83 (0.05) -7 ( 47) -4 ( 30)
0.40-0.50 1430 4307 0.83 (0.06) 0.83 (0.05) 0 ( 43) -0 ( 28)
0.50-0.60 572 2002 0.83 (0.05) 0.83 (0.05) 16 ( 40) 10 ( 30)
0.60-0.70 327 1357 0.83 (0.05) 0.83 (0.05) -7 ( 42) -7 ( 30)
0.70-0.80 64 287 0.82 (0.05) 0.84 (0.04) 4 ( 33) 2 ( 24)
0.80-0.90 26 150 0.83 (0.05) 0.84 (0.05) 13 ( 45) -3 ( 26)
0.90-1.00 3 8 0.80 (0.05) 0.83 (0.03) 9 ( 8) 14 ( 2)
XCorr P phases by event to station distance in 3.00 km step
StaDist [km] #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
3-6 22 17 0.64 (0.04) 0.64 (0.04) 19 ( 23) 6 ( 8)
6-9 13487 12847 0.77 (0.11) 0.77 (0.11) 0 ( 26) 0 ( 9)
9-12 7942 9451 0.79 (0.11) 0.81 (0.11) -1 ( 21) 0 ( 9)
12-15 7238 11380 0.73 (0.09) 0.72 (0.08) -0 ( 31) 0 ( 18)
15-18 994 679 0.76 (0.07) 0.76 (0.07) -5 ( 64) -2 ( 36)
18-21 3067 4218 0.78 (0.08) 0.79 (0.07) -1 ( 35) -0 ( 21)
21-24 1318 5599 0.73 (0.10) 0.73 (0.09) 2 ( 90) 2 ( 66)
[...]
XCorr S phases by event to station distance in 3.00 km step
StaDist [km] #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
3-6 35 5 0.89 (0.04) 0.89 (0.04) -5 ( 16) -3 ( 14)
6-9 8307 14072 0.87 (0.06) 0.88 (0.05) 0 ( 28) 0 ( 22)
9-12 7740 13058 0.84 (0.06) 0.84 (0.05) -0 ( 44) -0 ( 33)
12-15 3384 10766 0.83 (0.06) 0.83 (0.05) -1 ( 36) 0 ( 21)
15-18 1207 1849 0.83 (0.06) 0.83 (0.05) 1 ( 52) 1 ( 37)
18-21 527 2601 0.82 (0.07) 0.81 (0.06) -0 ( 54) 2 ( 39)
21-24 2411 7921 0.82 (0.06) 0.81 (0.05) -1 ( 48) -2 ( 36)
[...]
XCorr P phases by station
Station #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
8D.RAW1. 4254 5109 0.75 (0.09) 0.74 (0.09) -0 ( 35) 1 ( 26)
8D.RAW2. 6436 4086 0.75 (0.10) 0.74 (0.09) 0 ( 22) 0 ( 10)
8D.RAW4. 2902 3045 0.79 (0.08) 0.79 (0.07) -0 ( 30) 0 ( 20)
C4.CERNS. 0 0 0.00 (0.00) 0.00 (0.00) 0 ( 0) 0 ( 0)
CH.AIGLE. 41 211 0.80 (0.08) 0.84 (0.06) -6 ( 78) 1 ( 31)
CH.DIX. 687 5012 0.75 (0.09) 0.75 (0.08) 1 ( 48) 0 ( 21)
[...]
XCorr S phases by station
Station #CC #Skip meanCC (MAD) medianCC (MAD) meanLag (MAD) medianLag (MAD)
8D.RAW1. 2539 7670 0.83 (0.06) 0.83 (0.05) -0 ( 25) -0 ( 16)
8D.RAW2. 6861 3732 0.88 (0.06) 0.89 (0.05) -0 ( 28) 0 ( 23)
8D.RAW4. 79 838 0.86 (0.06) 0.87 (0.06) -5 ( 32) -5 ( 27)
CH.AIGLE. 113 288 0.82 (0.04) 0.82 (0.04) -1 ( 85) -5 ( 82)
CH.DIX. 2394 2606 0.84 (0.06) 0.85 (0.05) 1 ( 35) 0 ( 22)
[...]
#CC: number of cross-correlations performed#Skip: number of cross-correlations whose results do not account for the computation of the statisticscoeff: correlation coefficient between phase waveformslag: cross-correlation lag between phase waveforms in milliseconds
There could be several reasons why the cross-correlation between 2 phase waveforms is not considered for computing the statistics: the correlation coefficient is below the configured threshold (see crossCorrelation.x-phase.minCCCoef), the SNR of one or both the waveforms is below the configured threshold (see crossCorrelation.snr.minSnr), the waveform data for one or both the phases is not available and in general when the it is not possible to perform the cross-correlation. It is possible to know the exact reason by looking at the logs at debug level (–verbosity=4).
The SNR is particularly important to reject bad automatic picks or picks detected via cross-correlation (see Pick update and Phase detection via cross-correlation), but but it is not so relevant when relocating manually reviewed origins since the picks are checked already and bad ones discarded.
Reusing cross-correlation results
When cross-correlation settings are not changed, it might be useful to reuse the cross-correlation results to save processing time. Both the --eval-xcorr and --reloc-catalog options save a xcorr.csv file after finishing their execution (thay overwrite it if already present!). That file contains the computed cross-correlation results and can be given back to rtDD via the command line option --xcorr-cache. It is safe to change the value of crossCorrelation.x-phase.minCCCoef and reuse the cross-correlation results to see how performance change at varying correlation coefficient threshold.
Waveforms inspection
The --dump-wf option will make rtDD dump to disk the waveforms of the catalog passed as argument. Those files are in miniseed format and can be viewed with an external tool (e.g. scrttv waveform.mseed) or obspy). The waveforms are written to disk after the filtering and resampling have been applied:
scrtdd --help
--dump-wf arg Dump processed waveforms of the catalog
passed as argument in the current
working directory.The catalog can be a
single file (containing seiscomp origin
ids) or a file triplet
(station.csv,event.csv,phase.csv). Use
in combination with --profile.
e.g.:
scrtdd --dump-wf station.csv,event.csv,phase.csv --profile myProfile --verbosity=3 --console=1
17:59:28 [info] Writing ev1.8D.RAW2..HHT.Sg.manual.mseed
17:59:28 [info] Writing ev1.CH.SAYF2..HGT.Sg.manual.mseed
17:59:28 [info] Writing ev1.CH.SENIN..HHT.Sg.manual.mseed
17:59:28 [info] Writing ev1.XY.LEO01..HHT.Sg.manual.mseed
17:59:28 [info] Writing ev1.XY.LEO01..HHZ.Sg.manual.mseed
17:59:28 [info] Writing ev1.FR.OGSI.00.HHZ.Pg.manual.mseed
17:59:28 [info] Writing ev1.GU.REMY..HHZ.Pg.manual.mseed
17:59:28 [info] Writing ev1.CH.FIESA..HHZ.Pg.manual.mseed
17:59:28 [info] Writing ev1.CH.TORNY..HHZ.Pg.manual.mseed
17:59:28 [info] Writing ev1.8D.AMIDI..EHZ.Pg.manual.mseed
17:59:28 [info] Writing ev2.CH.DIX..HHT.Sg.manual.mseed
17:59:28 [info] Writing ev2.8D.RAW2..HHZ.Pg.manual.mseed
17:59:28 [info] Writing ev2.CH.SAYF2..HGZ.Pg.manual.mseed
17:59:28 [info] Writing ev2.CH.STSW2..HGZ.Pg.manual.mseed
[...]