40m QIL Cryo_Lab CTN SUS_Lab TCS_Lab OMC_Lab CRIME_Lab FEA ENG_Labs OptContFac Mariner WBEEShop
  PSL  Not logged in ELOG logo
Message ID: 434     Entry time: Mon Dec 20 00:08:57 2010
Author: Jan 
Type: Summary 
Category: PEM 
Subject: 2-day seismic analysis 

One of my DAQs (18bit, 1V range, NI 6289) in the seismo lab is recording data from a T240 broadband seismometer on the PSL table. Here, I am going to present plots for the first two days of data taking. I think that there are a couple of interesting features that should eventually be explained using additional seismometers.

The first datum is taken shortly after 2010-12-18 00:00 UTC. The seismometer is oriented along the sides of the table. If I am not mistaken, then the long side of the table should be approximately parallel to the East-West cardinal.



2010-12-18 (Saturday UTC, Friday/Saturday PT)

The plot that I like to look at first is my "hourline" plot. It is inspired by ancient seismology print-out styles. It shows 24h of displacement. Each hour is represented by 200 data points per channel properly filtered. The absolute value of the plot has no information. Different normalizations are chosen at different days to make the largest displacement of the day still fit without intersecting other hour lines. What you can see here is that the vertical channel is very stationary and weaker than the horizontals. The disturbances within the first hours is Tara working at or near the table and causing strong tilts. We will see more detail in this plot at the next day.


The next step is usually to look at the spectra. I like to produce quiet-time and total-time averages. I need to apologize about the labeling. I just noticed that there is still "ASD" on the y-axis. An amplitude spectral density would have different units (m/s/Hz). So these are root power spectral densities (sometimes also called linear spectral densities). In any case, the quiet-time spectrum is just barely below the global NHNM (new high-noise model) of Peterson. This is scary. If your measurement depends on seismic noise, then look for another lab. Characteristic for near coast sites is the double peak at oceanic microseismic frequencies (also seen at the Virgo site). Now, the higher-freq one is not easy to explain. I can only imagine that there is evanescent coupling of ocean wind waves (which do not produce significant microseisms far from the coast) to ocean bottom near the coast. The lower-freq peak is the usual ocean swell, freq doubled peak. You can also see Tara's low-freq tilts comparing the two plots. We will be able to understand more about the >1Hz spectrum in following plots.


The H/V ratio is the ratio of horizontal spectra to the vertical spectrum. It can tell you a lot if you know the seismic speed. If you don't know seismic speeds, then it can still be helpful to identify local anthropogenic sources. A seismic field "randomizes" over large propagation distances, which means that it loses its inital polarization (elliptical for Rayleigh and linear for body waves in many cases). The PSL ratios are not much different for the two horizontal directions. It is difficult to say what it means, but my first guess would be that all structures in the ratios are not due to source characteristics, but due to site characteristics (layers that cause resonances, the table itself,...). But it could also be related to sources.Only a seismic array can help here.


I am only showing two of the three spectral histograms. The level of spectral variation is actually lower than expected (low-freq tilts are the exception). The fact that you don't see much blue in the two plots tells you that the seismic field is quite stationary (well, at least in relation to its high absolute value). I have absolutely no explanation for this. I am very curious to observe the variations over longer periods of time. Especially the vertical displacement is surprisingly stationary. We will see this better in the time frequency plots.


I usually look at normalized time-frequency plots, but the unnormalized versions seem more exciting to people who don't look at them very often. Again, only two out of three channels. You can see many of the features that we also found in earlier plots (tilts, ocean microseisms, high-freq disturbances). Aparently, the near-coastal wind must have lost some strength during these 24h. Its peak was losing power at evening (keep in mind the 8h time difference). I don't understant part of the >10Hz vertical spectrum. What disturbance (small bandwidth) is losing power over night?

Plot11_2010-12-18.jpg Plot13_2010-12-18.jpg

Finally, I plotted a few particle trajectories. There is a low-freq plot with horizontal data <1Hz. It was taking during night when Tara had already left. So the polarization is natural. Although together with the other three plots (>20Hz, all three channels combinations) I get the idea that this polarization could be because of the table. One should eventually compare with data taken from the ground. I intend to program a trajectory histogram since there is more informaition in the high-freq plots than can be represented in these plots. So stay tuned. It requires a serious effort (a seismic array) to understand seismic polarization if you don't have any other site-specific information. Anyway, polarization is already interesting and can be used to give you a factor 2 in (seismic-noise limited) sensitivity if you orient your suspension system along the right direction! On the PSL table, you should avoid alignment parallel to the short side.






ELOG V3.1.3-