Hi Jack,


We used "spectrograms" of the sort that you described in
studying the 1989-1990 eruption of Mt. Redoubt, Alaska.  They were
the best method that we found to monitor and distinguish "LP"
events (with emergent arrivals and lower frequency content) from
"VT" events (volcano-tectonic events with higher-frequency
content and sharp arrivals).  Swarms of LP events tended to precede
eruptions.


It would be great to have a tool available to amateur seismologists and
educators that would allow spectrograms to be easily generated.  Do
you think this could be done without the use of Matlab?


Cheers,

John


At 07:01 AM 2/25/2005, you wrote:

There's a different way to look
at seismic records that is particularly

interesting for microseisms.  For a long time I've been using the


specgram function of Matlab to look at both microseisms and quakes. 


Specgram essentially divides the signal record into blocks of time 

and performs an FFT on each block.  It then displays the FFT
amplitude 

as a gray scale (or other color map).  The Y axis of the display is


increasing frequency, the X-axis is time, and brightness of each pixel


corresponds to the amplitude of the signal at that time and
frequency.

Essentially you get an image showing how each frequency component
changes

with time.  


This type of display is frequently used in speech analysis, passive
sonar,

and probably other fields.  This is not to be confused with the
simple

FFT function implemented by many of the data acquisition programs
that

gives a line of amplitude versus frequency, and which is useless 

by comparison.  


I was amazed at the different information available in this type 

of display compared with looking at a time series.  You can see


amplitude and frequency shifts of the microseisms (presumably as 

storms change location and intensity).  You can see frequency shifts


of the (dispersed) surface waves of a quake as it arrives.  

I have identified quakes by looking at the specgram display that I 

couldn't make out looking at the time series because they were buried


in high-frequency noise.


You can also see interesting higher-frequency signals, including line


spectra that shift and come and go mysteriously (probably cultural noise


of some type).


The representation allows you to easily distinguish body and surface
waves 

by their spectra, but because the FFT is done on blocks of data it is


not useful for calculating very accurate arrival times.


If anyone's interested I can dig out some old data and post a
picture.  It 

would be pretty easy to implement the algorithms in one of the data 


acquisition/display programs....


Jack


__________________________________________________________


Public Seismic Network Mailing List (PSN-L)


http://www.seismicnet.com/maillist.html
for more information. 



##################################/ John C. Lahr

#################################/ Emeritus Seismologist 

################################/ U.S. Geological Survey

===========================/ Geologic Hazards Team, MS966

##############################/ PO Box 25046

#############################//##############################

############################//###############################

     Denver, Colorado 80225-0046 /################################

             Phone: (303) 215-9913 /=============================

               Fax: (303) 273-8540 /##################################

                     lahr@........ /###################################

                                          /####################################

                                 http://jclahr.com/science/