Continued Landslide Monitoring

Continued Landslide Monitoring

February 11, 2018

by Steve Malone

Seismic monitoring of a slow landslide near Union Gap in eastern Washington continues with several new analysis techniques being developed and applied.  For details on the PNSN mobilization and monitoring procedures see the extensive Observations and updates for the January period.  This current post is a continuation of reports on our observations and analyses.

Update on February 15, 2018 (A mystery and forensic seismology)

Update: February 11, 2018 (A review of monitoring and a new monitoring parameter)

Here is a list of products available on the web that, along with other techniques, we routinely use to review the seismic activity related to the landslide:

  1. Raw Webicorders (12 hour-long, 10 minute-trace seismograms)
  2. Spectrograms (time versus frequency)
  3. RSAM  (Realtime Seismic Amplitude Monitor)
  4. REDPy (cluster identification over time)
  5. UGAP QuickShake (realtime seismograms)
  6. Tuned event counts (very local, tiny earthquake counts over time. Described below)

We also have access to realtime cameras, daily reports from an automatic motion detection system, and we participate in weekly discussions with other organizations monitoring the landslide. Should we notice unusual seismic activity associated with the slide we will immediately notify the monitoring group and participate in discussions regarding interpretations of the situation.

Number 6 above is a new monitoring tool.  To help us quickly recognize changes in the number of tiny earthquakes we record we are now using an automatic "tuned event counter".  Alex Hutko of our staff, who is a specilist in studying seismic noise and characterizing the quality of seismic data, has developed the system to try to automatically minimize the detections of cultural signals and maximize the detection of landslide related, tiny earthquakes. Using a process of filters and amplitude thresholds this counter runs on all of our seismic stations.  The top part of the display shows counts of only events that record on more than one station.  Here is an annotated snapshot of part of the display for today. Of course, as time goes on the hourly updated display will change.

There are several things to note on this plot.  The numbers along the bottom are day-of-the-year (Julian day).  When there are gaps in the data for one station the combination plot at the top will show no data even if the other station has data.  Because of a problem in our waveserver there is frequently a short gap at the end of many days giving a drop in the counts (seen on the 10 day plot, not on this example above). Note, in the bottom plot the "power" at a station (not the number of events but the overall signal level in the seismogram) has a very diurnal pattern, due to the the diurnal nature of traffic on the interstate highway.

One thing we are not sure how to interpret is the apparent notable increase in the average number of counts on station UGAP3 and perhaps UGAP6 (though the record is shorter) over the past month.  Unfortunately there was a period of time when UGAP3 had electrical problems that looked a little like tiny events to the automatic counting software (day 24 through day 27).  However, if one excludes that time period there apears to be an increase in the raw counts over the whole period from about 25 events per hour to around 50 events per hour.  This level has stayed steady for the past week or two.  Perhaps this averge count level is representative if the speed of the landslide which was slowly increasing up until mid January when it stabilized at about 7.5 cm/day.

Another thing to note is that the count levels on UGAP5 are quite high, UGAP6 a bit less and UGAP3 and UGAP2 less so.  The station UGAP5 is located just a few meters from the largest part of the landscape scarp and, based on approximate event locations covered in an earlier post, one might expect that near this station is where most events are taking place.  The count numbers are consistent with this.

 Update of February 15, 2018 - A Mystery and Forensic Seismology

Whenever seismologists see seismic signals unrelated to earthquakes our interest is piqued.  The seismic stations monitoring the Rattlesnake Ridge landslide near Union Gap, WA are filled with non-earthquake signals. Most we know the source for and consider to be "noise" (ie windy periods, trains, trucks, cars, helicopters and other aircraft). Recently we noticed a strange signal that showed up on all of our stations in the area and lasted all night (Feb 12 -13).  Here is one view of when it occurred.

This is a two day helicorder-type seismogram (one line is 30 minutes) for the time around the night of Feb 13.  Remember the times along the left edge are in Universal Standard Time (GMT or Z) which is 8 hours ahead of Pacific Standard Time (PST).  The approximate period of the strange signal is from 10pm PST on Feb 12 until about 8am PST on Feb. 13.  In this very compacted view one can just see that there is something different (larger signals) than at other times.  We can easily see that there is this elevated signal level on all stations in the vicinity via the RSAM Plot.  What about the character of that signal?  If we zoom in to a part of the signal we can look at it in both the time (seismogram) and the frequency (spectrogram and spectra) domains.  Here is an example of a couple of minutes at about 3:30 am PST on Feb 13.

Note in the spectrogram the many redish lines in the frequency range of 60 - 80 Hz.  These can be seen as the bumps in the spectral plot at the bottom as well.  Also, note the significant peak at about 18 Hz.  This is particularly bothersome to our event detection and counting software that filters the data from 15 - 40 HZ.  This 18 Hz signal, while it varies a little in strength, is almost always there during this period and records on both our UGAP monitoring stations and stations 1.1 km (UGP4) and 4.2 km (YA2) away.  Below are the same sorts of plots for a two minute period almost exactly a day later; early morning of Feb 14.

The scale is the same as before.  It's obvious that the overall signal levels are much lower, and the many peaks, particularly the one at 18 Hz, are gone.  One can also see near the beginning of the seismogram a small event (probably from the landslide) that would have been totally obscured the night before.

So, what is this mystery signal? We have not seen it on other nights.  None of the other monitoring techniques have reported anything unusual going on with the landslide, and from years of looking at seismograms we don't think it could have anything related to the landslide. We were very suspicious that it must have an anthropogenic origin.  It must be either a very strong source, such as a boring machine or rock crusher, or a bunch of distributed sources.  We know of nothing in the area that is big enough or would be operating in the middle of the night for the former.  A possibility we thought of was that the people at the Yakima Firing Range were doing some sort of night time manuvers and driving dozens of tanks all over the place all night.  Calling them turned up no activity.  After sending out queries to some people who live in the area we got a response from Ray Wolverton who lives only a few miles southwest of the landslide.  Here is part of the e-mail he sent to us:

I think I might have the answer to the constant mysterious seismic signal in that area.  The farmers were running their frost fans in that area-----some of them starting around 8:00 or 9:00 p.m. and continuing until around 8:00 a.m. the next morning.  I remember looking out my living room window that morning and seeing at least 30 of those machines operating within a mile or two of the landslide area. Typical size of these machines is 100+ horsepower engine driving an 18-20 foot prop. These machines create the noise and turbulence of a helicopter. When running, these can be heard/felt for considerable distances.

Indeed, we think Ray is correct. Several of these fans going at about the same speed could easily induce vibrations we would pick up several kilometers away.  There was still a bit of a question as to why we had not seen this before.  Looking at the temperature data from the Yakima airport for the past month or so helped to convince us even further that Ray was right.  Here is the plot for February.

Note that the only time the temperature was significantly below freezing for an extended period was the night of the 13-14th. While January had many cold days, the temperature never went much below 30 degrees F and then only for short periods.  In talking with Ray he even predicted that we might see these strange signals again.  The forecast is for very cold nights early next week, and so the frost prevention fans are likely to be turned on again.  Thanks for your help, Ray.