Seismic Noise due to concrete slab (jack hammer)


I was very interested in following up Nelson's results on channels that with the right threshold, have a high used ratio if used as vetoes for inspiral triggers. By following just the first example I found (a BS opt lever channel for H1), I found a jackhammer incident that definitely merits its own DQ flag. I am sure there's a goldmine of information in the rest, but I propose here a flag for this particular incident.


Betsy describes in this elog entry concrete slab work done near the corner station on July 20 2006 between ~16:30 UTC and 17:50 UTC.
Many of Nelson's triggers for H1 BS opt lever in the 1st year fall in this interval: there are 21 (of 48 total) triggers in 837451574-837451604 (17:26 UTC), where the Qscan suggests a jack hammer at work:





H1 and H2 have one science segment each affected:
H1-1854   6486 s    837447900- 837454386   2006 07/20 16:24:46 - 07/20 18:12:52 utc 

H2-1425   5568 s    837446883- 837452451   2006 07/20 16:07:49 - 07/20 17:40:37 utc


I downloaded seismometer data (H0:PEM-LVEA_SEISZ) for 837440000-837460000 (14:13-19:46 UTC), and produced a spectrogram that shows noise in the ~35-55 Hz noise consistent with a jack hammer in the time indicated by Betsy:




Nelson's "BS opt lev" triggers are in the minute 193 of the plot shown above.
I produced a time-frequency spectrogram with a finer 0.125s resolution for times between 837446823 and 837454446 (when either H1 or H2 were in science mode), used a threshold of 50 (in arb fft units for this resolution) in the 45 Hz bin, and produced a list of the triggers that repeated at least once within a second, with the new lines for each time skipped b/c it did not repeat. The large multiplicity blocks of time (more than 10 consecutive triggers) are definitely jackhammer noise from the Qscan appearance, with the earliest one starting at 16:29 UTC; however there some with fewer repetitions that are also suspicious, and seem to have a simultaneous glitch in H1 at least. Rounding all the times (without selecting any further for multiplicity) I created a list of 371 integer GPS times that I recommend using to create a DQ flag: JHflag.txt. There may be some times that are not due to a jackhammer, but they are all selected by high seismic noise at 45 Hz, with more than one trigger within a second; the vetoed times are 5%  of the interval selected. A plot of the coarser spectrogram (that makes the noise visible, the finer one has time bins too thin), the raw time series and the time series of the 45 Hz bin (for both coarse and fine spectrogram, in bue and black) are shown below, with a zoomed version too: