Tag Archives: seismic

We Are ... Seismic Noise

2014-10-28 13.27.05

Over the last few months construction crews have been hard a work tearing into the building adjacent to mine on the Penn State campus. Lots of demolition has been happening as the old building is completely cleaned out and being rebuilt. Some of the noise has been so strong that we could feel it next-door. As a data-nut, my first thought was "I'm going to look at this on our seismometer!"

At the base of Deike building (the geoscience building), we have a seismometer. The station, WRPS "We aRe Penn State", has been in operation on an isolated pier for some time, so we have lots of data to look at! For our purposes, I downloaded the entire month of October for 2013 and 2014. There are some hours/days that are missing, but we'll ignore those and work with what we have. This is a common problem in geoscience!

First let's just make a plot of this year's data. Each square represents one hour (24 squares in a row), and each row represents one day. Missing data is the lightest shade. The squares are colored by the strength of the seismic energy received during that hour; the darker the square, the more energy received.

WRPS_2014_HourlyYou'll immediately notice that there is always more noise starting about 11 UTC, which is the 7-8 AM hour locally. This is about when people are coming into work, vibrating the ground and buildings on campus as they do. The noise again seems to die off about 21 UTC or the 5-6 PM hour locally. This again makes sense with people leaving work and school. This isn't split finely enough to look for class change times on campus, but that could always be another project.

The other thing to point out is the dates of October 4-5,11-12,18-19,25-26. These are the weekends! You notice there is less of the normal daily noise traffic with fewer people on campus and construction halted. There is a repeating noise event at 11 UTC on the 1st, 12th, 20th, and 27th. I'm not sure what that is yet, but looking at more months of data may indicate if that event is associated with equipment starting up, or is really random.

While these daily life trends are interesting, they have been observed before. This whole discussion started with construction and how it was affecting the noise we saw on our local station. To examine this, I made a stacked power spectral density plot. Basically, this shows us how much energy is recorded at different frequencies. The higher frequencies would be human activity.

WRPS_PSD

We can see that the curves from 2013 and 2014 are very similar, with the exception of the 11-16 Hz range. In that range, the energy is higher in 2014 than in 2013 without construction by about a factor of 10. That range makes sense with construction activity as well! The energy remains elevated even after the main bump out to 20 Hz.

You might be thinking that such a bump could be due to anything. That's not necessarily true considering that we have stacked a month's worth of data for each curve. To show how remarkably reproducible these curves are, I made the same plot for the same times with a station in Albuquerque, New Mexico.

ANMO_PSD

 

In the Albuquerque plot, the two years are very similar, nothing like the full order of magnitude difference we saw in University Park. There are obviously some processing effects near 20Hz, but those are not actual signal differences, just artifacts of being near the corner frequency.

That's it for now! If there is interest, we can keep digging and look at signals resulting from touchdowns in football games, class changes, factories, etc. A big thank you to Professor Chuck Ammon as well for lots of discussion about these data and processing techniques.

Quick Test of Geophone Response

I just wanted to post a quick article about geophones.  Geophones are essentially instruments that allow us to measure the velocity or acceleration of the ground.  Yes, seismometers do this, but generally when we refer to geophones we are talking about single sensor (almost always vertical sensing) devices used for seismic imaging in oil/gas exploration.  I've talked about seismic surveys before (here for example). The "element", or the actual sensor is pictured below.  These sensors have a magnetic element on a spring inside a coil of wire.  Motion of the magnet (resulting from ground motion) generates a small electrical potential in the coil.  If I can find a cheap element/case on eBay I'll do a teardown of one in the future.  The signal generation happens through a process called "electromagnetic induction", described by Michael Faraday in 1831!  Want to know more about induction? Head over to the wikipedia page or shout out and we can put together a demonstration.

Dr. Ammon, whose office is next door, brought over an old element that he wanted to compare with our seismometers in the basement of the building.  Not knowing the output voltage range well, we hooked it up to a Rigol DS1102E oscilloscope on my desk.  I set the trigger of the oscilloscope (when it started collecting data) to just above ground potential so that any appreciable motion will trigger data recording.  We recorded the voltage output of the sensor about 6800 times per second!

The sensor element from a geophone.  (Image: Ebay)

Below is the waveform collected from hitting my desk with moderate force.  Surprisingly these elements put out +/-4 Volts! When shaking the element to it's limits we were seeing voltages of around +/- 10 Volts.  To me this indicates there are many turns in the coil and a very strong, probably rare earth, magnet inside.  Measurement of the coil resistance or a teardown will tell if this is correct! I've also included the power spectral density for those of you interested.  These figures tell us about the frequency response of the instrument.  Depending on how the spring system is setup, the oscillator is very sensitive to some frequencies and not so sensitive to others.  These diagrams help us characterize this response.  

Collected waveform from hitting my desk.
Power Spectral Density

Power Spectral Density: Zoomed in

Sorry for the short post, but I just wanted to share a quick desktop experiment!

Texas Fertilizer Plant Explosion Shakes the Ground

Another quick post of some interesting data concerning the explosion at the fertilizer plant in Texas.  Yesterday (4/17/13) there was a large explosion at approximately 7:50 pm local time.  As of early this morning reports showed around 179 people hospitalized, 24 in critical condition, and 5-15 fatalities.  Currently 3-5 firefighters and one law enforcement officer are unaccounted for.  Over 60 homes were damaged by the very large blast.

The best video I've seen so far is attached below, the explosion happens around 30 seconds in.  Also below is the initial emergency services traffic.

 

Finally, we can look at data from the Amarillo seismic station (US.AMTX).  I've pulled down the data and filtered it to show all frequencies above 1Hz.  We expect the explosion to produce mostly high frequency signals and attenuate, or lose strength, quickly (why I didn't see the explosion on any other stations such as US.WMOK in Oklahoma).  It looks like there are 6 main pulses of energy (possibly tanks failing?) very quickly and the large explosion in a period of around 10 seconds.

If you want to look at the data yourself I've made the SAC file available here or you can download the data from IRIS and duplicate the filtering with the following OBSPy code:

EDIT: The USGS posted a transportable array station that was closer to the event (seismogram below) that shows both the fast ground waves and the slower air blast.  They classify this as a magnitude 2.1 event on the event page, but it's really a larger explosion than that hints at as magnitude is only based upon ground motion.

Seismic Evidence From the Russian Meteorite Explosion

Today we're going to follow up on the last blog post about the explosion of a meteorite over Chelyabinsk, Russia.  The process of figuring out precise infrasound arrival times is quite a tricky process, the travel times depend on winds, humidity, and many other atmospheric variables that are hard to constrain over such a long travel path.  I've had several fantastic discussions with Dr. Charles Ammon here at Penn State to try to obtain the infrasound data that was collected near the blast, but so far we have not been able to get it.  When/if we do, expect another posting.

The focus of this post will actually be the seismic data near the blast.  There are many seismometers all over the Earth that record the motion of the ground many times a second.  After some discussion of the infrasound and seismic data available with Dr. Ammon, we found some really nice, simple results that would make a great laboratory assignment for an introductory seismology or geoscience class.  The activity could range from reading times of arrivals on provided graphs for a non-majors class, to filtering and grid searching to estimate the precise detonation location for a more advanced class.  I've provided the data and some thoughts on it below.

We'll consider data from five seismic observatories, the station names are ARU, BRVK, KURK, OBN, and ABKAR.  Below is a map showing the station location, distance to the blast (red star), and a seismogram from that station.  The seismogram shows how the ground is moving through time, in this case I'm showing the "Z" component.   This really just means we're looking at how the ground is moving up and down, though these stations also record North/South and East/West movement.  What we see is ground motion caused by the shock wave hitting the ground and that ground motion propagating away.

Fig. 1 - Map view of the seismic stations used.  Distance from the explosion, time after the explosion to a phase arrival, and arrival order (rank) are shown along with the seismogram.  All seismograms begin at the instant of the explosion.

It's common sense to expect the energy from the explosion to arrive at a later time at stations further away, which it does.  Notice how the sharp peak corresponds to distance? We can actually make a plot of this and learn some more from the data.  To do this, pick a feature that is easily identified in each waveform (we used the first trough) and record how many seconds after the blast it arrives at the instrument.  We then plot that on the x-axis of a graph and the distance of the station from the blast on the y-axis.  The result should be something like that shown in figure 2.  Now we can use some basic math to figure out how fast this energy was traveling.  The red line on the figure is the "best fit line" to the data.  We use some basic statistics (a linear regression) to make this line, but any plotting program will do it for you.  A line has a slope (how steep it is) and a y-intercept (where it touches the y-axis when x is zero).  The slope of a line is how much the y values change per a certain change on the x axis, often taught as "rise over run" in the classroom.  The slope of this line turns out to be about 3km/s.  That's a pretty reasonable speed for surface waves (which these are) through the ground!

Fig. 2 - The distance from the blast against arrival times.  This data indicates the surface waves traveled about 3km/s, a reasonable speed.

If we could pick out a "p-wave" in the data (difficult for reasons we will discuss), the intercept of the line would be the height above the ground that the blast happened.  I haven't seen a really good estimate of the height, probably because the p-wave is hard to find and the speed of the meteorite. The meteorite was traveling about 40,000 mph when it exploded.  It's hard to imagine something moving that fast, so let's change around the units: that's something like 11 miles every second!

The p-wave could be hard to see because 1) it's going to be relatively small, and 2) there are waves from an earthquake in Tonga arriving about the same time as the meteorite explosion.  We know the waves we picked aren't from the tonga event, those would have arrived at all the stations at almost the same time because they were reflecting off the Earth's core.  It would be an interesting project to play with trying to pick p-waves and/or estimate their arrival window by guessing the height of detonation.

We don't have to stop here though.  This morning I saw this youtube video, a compilation of people recording the shockwave.  The meteorite had streaked past, exploded, and they were recording this when the shock wave hit.  Shockwaves behave in a funny way, but luckily it's been studied a lot by the government.  Why? Nuclear weapons! Seismologists are commonly employed to determine if a nuclear test has taken place, and estimate it's size, location, etc.  A lot of very interesting information on air-blast and it's interaction with buildings can be found in the book "The Effects of Nuclear Weapons".  The book has lots of formulas and relations that could make many interesting lab exercises, but we'll just discuss reflection in this post.

A shock wave is really a front of very high air pressure that is propagating through some material.  The high pressure is followed (in a developed shock wave) by a small, longer, suction, then a small overpressure.    I've tried to locate meteorological observations and so far have only found hourly observations.  If we can find short term observations we would expect to see wind rushing away from the blast, then more weakly towards it, then very weakly away from the blast.  By knowing those wind velocities we could estimate the pressure differential that caused the shock.  The local airport (station USCC) does report hourly average winds (data here).  There is a small bump in the average winds between 9-10am local time, when the meteorite entered.  The lack of a gust report though makes this observation a bit too shaky to use for a pressure estimate.  

Shock waves move faster than the speed of sound if they are a high enough "overpressure", or the pressure above atmospheric.  Shock waves will reflect off the ground when they reach it, as shown in figure 3.  The overpressure in the region of "regular reflection" is much higher than the overpressure of the shock wave due to a combined stacking effect.  There can also be complicating patterns such as "Mach Reflections".  

Fig.3 - The initial pressure wave (solid lines) and the reflected shock (dashed lines).  Image from "The Effects of Nuclear Weapons"

What's interesting about all this is the audio of the clips at about 20 and 40 seconds into the YouTube video.  Notice these clips contain two bangs.  The first clip with two shocks could be reflection off the building behind the camera, the second shock follows the first very close and is very loud.  The next clip has a significant delay though.  At any height above the surface the initial reflection occurred on, there will be a delay between the initial and reflected shock.  If we knew the location of this video it would help constrain the shock location.  (After some google searching I can't locate the "Assorty" store in the footage anywhere.)

Overall with the observations of glass breaking over such a large area, we can assume the reflected pressure was probably in the area of 1psi.  This means the initial overpressure was very small at the ground.  Could you work backwards from the estimate of 500 kiltons TNT? Sure! That's a topic for another day or for your students in lab! Be sure to check out the book "The Effects of Nuclear Weapons", many campus libraries have it, Penn State has it online even.

Below is a link to a zip file that contains the .SAC files for the seismic stations (starting at detonation time and low pass filtered as well as raw data) and high quality figures.  If I end up writing up a lab from the event, expect the data and lab to be on my academic website.  A review of literature on the Tunguska event may be helpful as well!

Zip file of data.

Teaching Field Camp Week 1 - Norman, OK

For the next 3.5 weeks I'll be a teaching assistant for the University of Oklahoma geophysics field camp.  The point of the camp is to teach senior geophysics students how field data is collected, processed, interpreted, and applied to the problem.  This is an important capstone class because prior to now students just see geophysical data as equations, numbers, and options in software and on paper.  Now they must hike in the field, observe the geology, collect the data, and finally figure out what it all means.

Week 1 was done in Norman, OK back at the school.  Monday the students listened to lecture on geophysical methods, were introduced to the equipment, and finally were tasked with using differential GPS on the North Oval of campus.  Differential GPS is much more sophisticated than the GPS in your car.  Each unit costs ~$80,000, and one is mounted on a tripod and remains stationary throughout the day.  This station is referred to as the base, and is the most crucial link in any geophysical survey.  The second station is mounted in a backpack and is the rover.  Students walk around with the rover collecting data points, then at the end of the day the base station is used to calibrate the rover data.  We know the base station doesn't move during the day, but it appears to in the data.  This is because GPS locations are highly susceptible to changes in atmospheric humidity, irregularities in the satellite orbits, and a number of other factors.  Without going into more detail, look below at the Excel plot of the oval before and after correction.  Data points are much closer (within centimeters) after correction, and those centimeters make all the difference in some survey environments.  This plot came from one of our students reports that was turned in during the week.

The next objective was to collect a seismic line over a branch of the fault system that slipped during the earthquake sequence of November 2011 in central Oklahoma.  Setting out a seismic line is a long, arduous task, so the students needed a practice day.  We setup a short (~300m) line by the school's duck pond.  Below is a time-lapse video I took of the practice session on Tuesday.

The next two days were collecting the real data in Prague, OK with Friday reserved for processing.  Without going into great detail of how we setup and collected that data I'll say that 72 geophones were deployed every 10m.  Geophones are small seismometers effectively that only measure the motion of the ground in one direction (up and down in this case).  After processing the data we get an 'image' of what's going on underground.  Are the rocks bent (folded), broken (faulted), or otherwise layered/interesting.  We expected to cross the branch of the fault responsible for some of the stronger aftershocks.

Below are some of the processed images from a student.  This is a rough processing and can be improved with more time, but that is beyond the scope of what is expected in the field.  The faults are marked by yellow lines and indicated places were the rock has broken and slipped.  Also notice the folded layers to the left of the section.  More work and interpretation is needed to obtain further geologically useful interpretations.

Expect more posts as we re-group in Cañon City, CO and begin working on gravity, magnetics, and ground penetrating radar.

Seismic Survey with iPhone Application

Recently I was on a sedimentary petrology field trip to Galveston, TX.  While we were standing on the beach the class dug a trench to examine some sedimentary structure and I saw an opportunity to try something very interesting...a seismic survey with an iPhone.

After talking with another geophysics major, Dustin, we got four phones and downloaded the iSeismo application.  We knew the layer we were looking for was about 1ft down and was not dipping much so we quickly set out the phones as shown below.  (Line length ~10x depth we wanted to image.)  For a seismic source we first tried a hammer but then ended up using one geologist who jumped, and we collected three shots.  All were from the same location as we were neglecting the dip so a reverse shoot was not necessary.

After we returned I quickly plotted up the data, and to my amazement saw seismic arrivals at ALL iPhones! Then I saw a problem.  The data is time stamped, but when the iPhone syncs with the network time it is not as accurate as we had hoped.  The data were seconds off when I stacked the arrivals on top of one another.  So, without an accurate way to line it up I could not solve for velocities and depths of layers, but for a proof of concept this is a step in the right direction.  This also shows just how quick and easy it is to collect seismic refraction data! With some software modifications or syncing mechanisms this could be repeated with the possibility of better results.  Overall it proves the versatility of both the method and the iPhone.  Below are plots from the iPhone accelerometer in the x,y,z directions for the first and last phones in the line. Thanks to the sedimentary petrology class, Dustin, Dr. Keranen, and Dr. Elmore.