Tuesday, November 25, 2008

What's shaking?

This cartoon is posted on the Franklin & Marshall seismograph display:

In the old days of a drum recorder, we often lost data when we forgot to change the paper. Now that the recording is digital, we are doing better. And we have evidence for that. Mitch, maybe you can explain those QUACK statistics?

1 comment:

  1. QUACK is a useful database that DMC provides to examine station data quality. One of the tools there is called a PDF which stands for Probabilistic Density Function. The colorful graph shows the probability of signal over time in units of decibels (acceleration^2/Hz) versus period.

    Go to http://www.iris.washington.edu/servlet/quackquery/ , select Explore Data, and select network LD. There will be a number of stations and channels. To start select station PAL and channel BHZ. On the right side are the various measurements we can examine. Select NEIC PDF Plot, and then a data range (get the whole year of 2008).

    The other features you can see are the minimum, maximum and modal values for the time range, and the high and low noise models as described by Petersen (1993) "Observations and Modeling of Seismic Background Noise" which are composite station spectra of worst and best stations.

    To get a sense of what we are looking at, transient events of low probability will be in magenta, 3-12% is in blue to cyan, and 12-24% is in the green to yellow. The higher the color band, the more consistent the noise level of the station. Also, the narrower the band, the less there is variation.

    Taking a look at the frequency ranges, we look at ranges of 1-10 Hz for most local events, .8-5Hz for regional events (hundreds of kilometers distance), and longer periods for teleseisms (thousands of kilometers distances).

    By eyeballing it, the 1-10Hz range for Palisades is about -135Hz most often. Cultural noise will dominate this band, and if looked at hourly, a pattern will emerge in many stations with higher noise during the 9-5 work week.

    The two curves between 1 second and 10 seconds period represent microseismic background peaks. Inland stations will have lower curves in those bands, but most stations will show some effect. Palisades being close to the coast shows it prominently. Seasonally this band will also change with stronger wind on wave action during the winter months than the summer months.

    The next range from 10 seconds to 100 seconds period will differ for many instruments. Those instruments with a natural period of 30 seconds will have higher instrumental noise at longer periods than a sensor designed for longer periods (100-120 seconds with the sensors deployed in the LCSN). Temperature and pressure changes will have a long-period effect on a sensor and can seen here with stations that have better insulation and vaults than those without.

    Compare the FMPA HHZ channel for 2008 which shows the noise floor rising at 10 seconds and longer periods. While the 40T has a natural period of 30 seconds on the models used in network, the self-noise of the instrument is high enough that it is above the background noise. It will still be sensitive to those longer seismic waves from larger events. There is also high frequency noise at FMPA where from 5-10Hz the modal level rises from -130Hz to -110Hz.

    CUNY has a high frequency noise problem above 0.5 Hz where the level of noise exceeds the high noise model. The only real solution to this site is relocation.

    HCNY shows bimodal noise in the high frequency due to a nearby rail line, but because it's on massive rock in a temperature stable cave (Howe Caverns), the long period background is very low.

    Each site should be looked at for what noise sources are nearby including trees, buildings, physical plants, pumps and traffic.


Wave to us!