This study sets out to try and determine the nature/source of a series of ‘pops’ that appeared on the Mount St. Helens seismograph stations commencing at 23:50 UTC on the 27th January 2012. This was later shown to extend into the 28th.
The original ‘pops’ were brought to my attention on the Pacific Northwest Seismic Network Facebook page The comment that accompanied the image below was:
One last spectrogram for today. Pops, probably explosions, near Mount St. Helens. In this case, the difference in timing between when the signals arrive on different stations is due to the slow speed (300 m/s) of sound in air.
Clearly the seismograms were displaying a singular set of events as the pattern was apparent on several stations as observed by PNSN. My response to this as you can see on FB was:
Can you define probably explosions please? Gunfire, blasting, what? Hardly blasting at Studebaker Ridge I would not have thought yet that ‘appears’ to be close to the the primary signal source. June Lake is a long way from there. That has to be some explosion to travel across the top of the St Helens summit which is directly in line and still register on a seismo the other side over 7 miles away. A quick visual indicates the source to be immediately west of STD otherwise FL2 and SHW could not get the signal at the same time. This should also be apparent on PB.B201 if that is the case at about the same strength as SHW and FL2. This would mean it was not quite as far to JUN.
This set the scene for the study as I wished to know the nature of these signals.
This morning PNSN had responded:
I hadn’t looked at that level of detail, you’re right, although TDL looks even closer. Soon we will have a location option to use a velocity model based on sound waves, which would give a location. Odd that the arrivals would be so clear so far for an air wave.
In order to better see the timings of the signal arrivals I downloaded SAC data files for the following stations:
I then processed each file as follows:
CC.SUG Removed telemetry signal @16:58.84 (+23h50m) – 17:58.88 (duration 1:00.04) and replaced with silence. The reason for doing this is because the massive telemetry signal prevents amplitude increment to get an audible record. The spectrum after removal was as below with a strong signal @40Hz (Not normally present on these seismos) and the often present 22.5Hz signal. (Apologies for the speeling era ‘artrefact’ on the image 🙂 ) The 18 to 20Hz signal is not so common. I filtered every above 18Hz out using a low-pass filter and then amplified the signal.
Whilst manually taking off the times for the pops a signal was encountered that appeared to be seismic in origin. The little bump on the very bottom of the spectrum is the event.
42:58.30 start of 2Hz signal. Approximate end 47:25.39 (Add 23:50 to these thus 00:32:58.30 etc as the trace started at 23:50)
I played the file to listen to the sound, however nothing could be heard. This is always the case with distant earthquakes. (You can discover the reason here) This event seems to be a teleseismic (distant) event because of this so I looked at the USGS listings and found a quake in Baja California M3.3 which might be a candidate being 03:54m before the signal. Earthquake details shown below in case the USGS page has gone.
This is 234 seconds and the approximate distance between SUG and the Baja quake is 1,655km giving a travel time of ~7 km/sec and therefore within the parameters for P-wave travel.
The sound of the file post processing if you are interested can be downloaded here. This is @40 x speed. Note you do not hear the M3.3. If you have Audacity or some other sound processor set the rate to 8000 samples per second and you will then hear it.
UW.ELK was next and this had no telemetry to be removed. The frequency of the signals seemed to be higher on this station running to 28Hz instead of 18Hz so I did a low-pass on 28Hx to remove the signals above. The pops are there but they are too faint to be useful so unless needed later I have discarded ELK.
UW.FL2 was next and the 28Hz low-pass filter was applied. This had usable signals.
I went through the following seismos, after a total logic failure earlier by which means I pronounced it as P-wave speed. [Extra note added after: Had I read my own initial comment on the PNSN site again – “A quick visual indicates the source to be immediately west of STD” – I would have saved myself some face since that is pretty much where it turned out to be. Never mind the whole thing has been a great learning experience, and only by learning can you grow.]
|ELK||08:30||5s*300m = 1.5km|
|FL2||09:06||39s*300m = 11.7km|
|SUG||09:10||45s*300m = 13.5km|
|JUN||09:31||66s*300m = 19.8km|
The distance is where the intersect should be to find the location. Due to rounding of the latitude and longitude and only using 2 decimal places this is not highly accurate but should be thereabouts.
With UW.SHW nice and close I thought I would double check by reversing the calculation. SHW is about 9.3 km from the approximate intersect and thus at 300 m/s should be showing an arrival time of 31 seconds after TLD. I make if about 08:59 on the spectrum which is only 3 seconds out. Considering the inaccuracies due to rounding etc that is acceptable and proves the source location to be in the approximate area circled. (Approximate) Location in Google Maps 46.29, -122.31
So what is the sound? Explosions? Maybe. Long travel distance @19.3km. That would have to be some huge explosion.
Here is a section of that @500 samples per second. SUG Pops WAV
And this is the spectrum to go with it (remember x 5) Messy because I had to increase the amplitude so you could hear it.
Kudos to the PNSN for the initial assessment which seems to be right but raises the question why, how big and by whom?
If I take nothing else away from this I have found a very rough method of locating a signal source which I need to improve upon and refine. 🙂