Splitlab and San Fran

July 29th, 2010

Well I've been making some serious steps forward and some minor steps back.  After hours of fiddling around with it and a few emails from the creator, we managed to crack the code on Splitlab.  It turned out that to make it work I had to rename all the sac files into a specific format, search all the earthquakes in the CMT catalog within the same distance that I had requested on my sod file, and then associate the sac files I had with those earthquakes before I could look at the seismograms.  After that, there was a minor issue with calculating the phase arrivals but that turned out to be a permissions issue with a java file.  So, once I got them all read in, I started to do some actual splitting.  As I previously said, a lot of the signals from this stations were very noisy and I couldn't actually see arrivals on about half of them.  The ones that I did do were difficult to get good results.  The splitting relies on the fact that SKS waves are totally radially polarized after they go through the core.  This means that all of the energy is initially carried on the radial component of the wave.  However, when it goes through the anisotropic zone, the wave is split into the fast and slow directions.  Once the program calculates the best fit for the dt and phi, it tries to 'undo' the splitting.  If the dt and phi are exactly right, the transverse component would be flat (no energy) and all the energy would be on the radial component, as it originally was.  It was often impossible to achieve this with my data.  Once corrected, the particle motion should change from being elliptical to linear (also, didn't look that great).  I did the splitting anyway as best I could for the permanent station, ELK, in Elko.  The results had some subtle trends: the azimuth of the fast direction seemed to be somewhere between -60 and -90 degrees.  The dt values were significantly scattered and showed no obvious trends.  We decided to increase the distance window of the earthquakes in the hopes of increasing the backazimuthal variation.  It is necessary in two-layer shear waves splitting to have a wide variety of backazimuths because there are 4 (instead of 2) directions that you must avoid because they will lead to 0 splitting in that layer.  We also finally figured out how to apply a filter in Splitlab.  It was amazing how much a .05-.2 Hz filter improved the look of the data. In some cases it went from being completely unusable to a seismogram that produced a good split.  

All in all, it took a long time doing things wrong to figure out how to do them right (or at least better).  I guess that's how these things go, though.  So tomorrow I am starting over on my ELK data to hopefully produce a more convincing result.  

In other news, my friend from Rice came to visit me and we spent Saturday in San Francisco and Sunday in Santa Cruz. It was awesome to see her and despite some car trouble (our rental car broke and wouldn't accelerate..or reverse!) we had a great time.  Here are some pics and I will try to put some pictures from Splitlab up tomorrow.

Example of a good splitting result. Note that the waveforms corrected for delay time line up nicely, the corrected radial and transverse waves show minimal energy on the transverse component, and the particle motion has gone from being elliptical to linear.  Parameters are reasonably well-constrained.  

The house from Full House!!!!

Santa Cruz

Golden Gate Bridge