This summer I am working with ultra-low frequency electromagnetic (ULFEM) data collected from 5 different recording sites in the greater San Francisco Bay Area. I will first catalog all the previously recorded data from each of the sites, indicating periods of time when particular channels at the sites were and weren’t functioning or recording, when batteries have died and when any changes were made to the setup of the site; the ultimate goal of the log will be to determine when periods of “good” data and “bad” data occurred. Then I will look more specifically at the “good” data to pick out particular events that we are pretty certain were being recorded, like lightening strikes and solar storms. By getting good at identifying known events, we will have more success identifying any anomalies that may be precursors to earthquakes.
I have been very very busy lately. I have finally begun looking at EM data, which has been pretty exciting!! The way I narrowed down the data I wanted to look at was by picking 15 earthquakes using my beautiful earthquake graph (see in previous blog post) and then looking in my catalog (see in previous blog post) to see which ones had good data. Then I went through to find co-shaking signals in the EM data. We are assuming that if we do not see the actual earthquake waves arriving at our station, we most likely won’t see any precursors. We aren’t sure about this, but it was a way to narrow down the earthquakes. So I found 4 earthquakes that definitely showed those co-shaking signals and went from there.
It took a lot longer to find co-shaking signals that I thought. At first I was just looking in the EM data around the time I calculated the earthquake to have arrived at each station, but I wasn’t sure what scales to be looking at the data at. I then figured out a way to plot seismic data with the EM data by converting and cutting my EM data so the times matched up. This was great because I saw the earthquake arriving clearly on the seismograms and then could look down to see what the EM data looked like at that time and if the signal changed at all. Below is an example of co-shaking signals. The top plot is one channel of seismic data and the bottom three are EM data from channels T1-T3 from our Parkfield station. You can clearly see the earthquake arrival displayed in the EM data. When I began to find plots like this, it was very exciting, and I knew I was close to being able to look at the actual EM data!
As I discussed in my previous post, the first analysis we wanted to do was to look for the pulsations Bleier described in his paper as what he believes might be earthquake precursors. I had a neat opportunity to meet with Bleier and his team at Quakefinder to hear more about their company and to ask them questions I had about the pulsations and the analysis! The meeting was very interesting and pretty helpful. Since I had not begun looking at the data yet, I only had some basic questions to ask them, but it was helpful getting confirmation that what I thought they meant in their paper was actually mostly what they meant, for example about how they found the pulsations and where I could start. A big thank you to Tom Bleier and Clark Dunson for taking the time to help me out!
I decided to look at our EM data in the week preceding the 10/31/2007 Alum Rock M 5.4 earthquake. This was the earthquake that Bleier found pulsations before and wrote about in his 2009 paper. I wanted to see what we could find on our data! I ran into some trouble getting the EM data using the ULFEM program that I installed onto the computer at Stanford earlier this summer. I ended up spending a lot of time trying to figure that out until we decided it was probably easier for me to download the data using the ULFEM program at USGS that doesn’t have the same problems. Darcy and Jonathan, my advisors at USGS, had me compose an email to the designer of the ULFEM program in hopes that he could help us with some of the bugs on the Stanford computer. It is a very helpful program and would be nice if future students working on this project could use it, as it would require much less programing, since it downloads and plots the data in multiple ways on its own.
Once I got the EM raw data I wrote a script to help me filter and plot it. I needed to create a threshold above and below the mean in order to have a place to start to look for pulsations. After talking with Tom and Clark at Quakefinder, I decided to create the thresholds using 2 or 3X the standard deviation of the two-hour chunk of data I was looking at. I then went through, in increments of about 36 seconds looking for pulsations that exceeded my thresholds. This process is what I am working on now and has been going pretty slowly unfortunately, as there are a lot of pulses and it takes a while to save the pulse, record how long it is, whether it’s unipolar positive or negative or bipolar and find it’s amplitude. I have found some things to report on the pulsations though and was able to do some of that before my abstract was due, so I included that in there! You can see an example below of what two hours of my EM data look like, with thresholds indicated in dotted lines.
I have been recording pulsations that exceed 3X and 4X the standard deviation of the chunk of data I am examining. Essentially, the majority of the pulsations that exceed the thresholds are unipolar and the duration generally ranges from about 3 to 10 seconds long, usually averaging around 6-8 seconds. As of right now I am systematically going through and each day looking at two hours in the morning (9-11 UT), when the electric train BART is mostly dormant (making our data less noisy), and two hours in the afternoon (21-23 UT), when BART is on. I can say that these pulsations seem similar to Bleier’s pulsations in that they exceed a threshold of some kind and have similar durations. However, Bleier’s pulsations stand out much more from their data, which either means they are seeing much larger pulsations or our data is much noisier. And as I am looking through our data, I notice that the pulsations I’m picking out aren’t completely distinguishable from other spikes in the data; other spikes look similar in shape and duration, just that they don’t reach the same amplitude to put them over the thresholds. I am not convinced of anything yet either way for if these pulsations are related to the earthquake because I don’t feel like I’ve had the chance to look through as much data as I would like!
What I would like to do with the data next would be to find a quiet week (in terms of earthquake activity) and pick a day to analyze the same way I’ve been doing for pulsations before Alum Rock. I want to compare a “quiet” day to see if the amount of pulsations has a significant change. In addition, so far I have been only looking on 1 channel of data from 1 station. I want to compare channels of data to see if I see pulsations occur on all 3 channels at once. If I only saw a pulsation on one it might indicate that the spike is merely something related to the machinery.
I have definitely been working on a ton of different things this summer. I am feeling pretty overwhelmed with what I need to accomplish before the end of next week, but I am also very excited thinking about everything I have done! I am probably most excited about my work with MATLAB. When I actually think about the different things I have used MATLAB to do, I realize how much I have actually learned about it! I have a whole folder of MATLAB scripts that I have written mostly myself to do various tasks I needed to do this summer. While they are not the most efficient scripts, I have definitely learned a ton and hope that in the future I can learn more tricks and way to improve on my coding so what I write is more proficient.
Sorry I haven’t posted in a while! I took the week off to head home for the 4th of July and am working to get back into the swing of things. It was a very nice, relaxing week with lots of good food, company, swimming and sun.
Last week back at work, I read more thoroughly through a few papers, which analyze electromagnetic data and conclude they see precursor anomalies, in order to come up with a list of processes they did to replicate on our data. The first analysis I am going to focus on is looking for a certain type of pulsation characterized in Bleier et al. 2009. What I understand so far is that the pulsations are characterized by having amplitudes greater than twice the site background noise, are unique in duration (1-30 seconds) and exhibit strange singular polarities and bi-polar waveforms. I am going to look in our data for pulses like this.
End of last week and this week I have been figuring out how to download my raw data and filter it. Caroline, another intern in my lab, taught me the basics of SAC in order to determine the time a bunch of earthquakes (chosen using my earthquake graph shown above) arrived at our stations so that I would know where to look in our EM data. I had to use some filters on the seismic data to see the arrivals more clearly. I figured out how to use what is called a bandpass butterworth filter. The code I type in SAC is “bandpass butter corner 0.33 10”, where “corner” and the numbers indicate the range of frequencies I want to KEEP with my data. I am very entertained every time I type butter! I also figured out how to open our raw EM data in MATLAB and have written a bandpass butterworth filter in MATLAB to filter our EM data in the same way as the seismic data because the frequency of the seismic and EM data are about the same.
Simon also has me beginning to write up and organize my abstract and things that will go on my poster for AGU, like an introduction, any background and graphs I think I will have. This is making me realize that I am halfway through my internship, which is pretty intense! Since I haven’t produced any concrete graph or results recently I having been feeling slightly like I am not moving forward, but when I take a step back, like I did when writing this blog post, it makes me realize how many different things I have been working on to get me ready to produce actual graphs and results. I’m excited to see what I find!
I had a really great time out in the field for two days with Leah and about 20 other people from USGS this week; Eva was there too for one of the days! We were working on a seismic survey down south along the coast in Ano Nuevo Park. We were working on a 600 meter array along a service road with a beautiful view of the ocean off in the distance! Leah was the PI on the project, so it was pretty neat to see all these people (many of them adults!) working to get data for her to analyze! I worked on putting in and taking out geophones, using a hand auger for the p-wave shot source and avoiding the poison oak plants that were EVERYWHERE (I currently have some itching on my elbow, but I’m pretty sure it’s a couple of mosquito bites…fingers crossed.) As exhausted as I was from two very long and active days, it was nice to get out of the office and see a beautiful coastal area and cool geophysics field techniques! Thanks Leah!
Back to work at my cubicle wasn’t too bad, however. I am still working on figuring out what earthquakes make sense to look at in our data. I have a pretty nice looking graph that gets more and more colorful as I add earthquake events to it! In response to Katie’s question on my previous post, I will most likely start by working off previous work in the sense that I will look at how other people have analyzed their data and begin by doing that to some of our data. It’s looking like we really haven’t had any earthquakes recently enough that would be large enough to show us anomalies over the noise of the area on our stations. I will most likely not see any anomalies associated with earthquakes, but even that will be good to report so we can begin to establish the size and proximity of earthquake needed to see any possible anomalies. Here is what my graph is looking like:
I talked with Simon about a good range of earthquakes to examine based on the graph. I first have to catalog a little more so that I can make sure we have good data from our stations on the dates of the earthquakes I would be looking at. Then my plan is to find lightning data and look at the data closely in order to characterize lightning events. This will help me not only become more familiar with the data but also help us rule out anomalies we might see that are actually just lightning strikes. Next I’ll probably look at solar storms and other exogenic geomagnetic events.
Wow, so, cool story: I just found out I get to look at ULFEM data before, during and after known earthquakes to see if I see anything! I am actually looking for possible earthquake precursors, which I didn't think I got to do this summer!! Simon has me pulling earthquake data from NCEDC and calculating the km distance from the earthquake to our stations, then graphing the magnitude of the earthquake versus the log of the distance to see which earthquakes make sense to look for in the data. Some earthquakes will be too small and far away to most likely show anything on our machines, so I need to figure out which earthquakes it makes sense to investigate further. I am still going to be cataloging and characterizing other events, like solar storms and lightning, but I will now also be exploring data around earthquakes! I'll definitely keep you updated on how this goes; I'm pretty excited.
This week I really chugged along on the excel spreadsheet, cataloging past data. I think I have come up with a pretty clear way to organize it, color-coding what is good data, bad data and things we need to take another look at. As I am getting in more of a rhythm, I find I am able to download and catalog a years worth of data for all the stations in a day. Hopefully I will get even faster at it as I become more used to the data I am looking at.
Below is an example of the type of graphs I am looking at. It shows a week worth of data from the Jasper Ridge station (JRSC). T1, T2 and T3 show the magnetic data, while Q2, Q3, Q5 and Q6 show the electric data. The magnetic data looks pretty good; the scale is on a good order of magnitude, between 104-106, and BART (the train that runs in the area) shows clearly, as the periods with less noise are when it is not running, from midnight to 4am. I write “BART obvious” quite a lot: BART is very noisy. There is something going on with the electric data however. Q2 and Q3 have too small a range to be picking up anything more than machine noise of some kind, probably from one part of the station. Q5 and Q6 are showing a good range, but the signal looks unnatural; most likely one of the two batteries the station switches between is no longer working and thus the step pattern is from it switching from a dead battery to a working battery. This is the kind of cataloging I am doing, but just shorthand, in a chart with color-coding so it can be read easily.
I have been exploring the area a little more as well. This weekend, among other fun things like tennis and a pool day, I went to the Stanford movie theatre (where they only show old classic black and white movies) to see Monkey Business with Cary Grant, Ginger Rogers and Marilyn Monroe! It was really silly and entertaining; I hope to see many more movies there this summer. I’ve also been sending out TONS of postcards so if anyone would like a pretty postcard of Stanford or Palo Alto, send me your address in the comments or to my email (firstname.lastname@example.org) and I’ll send you one!
Also, tomorrow and Wednesday I’ll be out doing field work with Leah and her mentor at USGS! I’m pretty excited to have a couple days outside, out of my cubicle. I’ll let you know how that goes in my next post!
After a busy, but interesting orientation week in Socorro, I headed out to sunny Palo Alto, California! Work began right away this past Monday. It was nice to arrive on campus and have Eva (another IRIS intern) working at the desk right next to me! We have some quality lunch time together sitting outside. We've made it our goal to eat in a difference place every day for lunch; we'll see how long we are able do that this summer! I met with all three of my advisors the first day, Simon Klemperer, a geophysics professor at Stanford University, first and then Darcy Karakelian McPhee and Jonathan Glen, both geophysicists, over at the USGS. I am looking forward to working with all of them; it's interesting talking out plans because each of us bring different ideas and suggestions to the conversation, which I like. As the week has gone on, I am surprised with how much more familiar I've already become with the field site setup and data I will be working with, as well as some of the background on the controversial subject of earthquake precursors. I'm looking forward to understanding it all even more!
Darcy and Jonathan spent a lot of time with me on Monday and Tuesday, showing me the program I would use to pull the ultra-low frequency electromagnetic data (ULFEM) from where it is stored to then be able to plot it. I learned that my first step this summer will be to catalog all the data they have collected at each of the five sites. I will go through week by week and make note of when a channel at a station seemed to be working or not and note any odd-looking periods of time that should be revisited and analyzed further. We also discussed that my work after the log will most likely be looking more closely at certain areas of the data to pick out known events, like lightning strikes and solar storms.
My goals right now for the summer include the following:
I'm staying with cousins here in Palo Alto, which has been nice because I go home to a full house and people to relax with. I'm excited for this weekend to go play some tennis and swim and explore downtown more. Right now I'm feeling nervous and thrilled about spending my summer out here in Palo Alto; we'll see how the summer goes!