Analog seismic data have been recorded by numerous pen and paper drum-type seismographs throughout the world over the past 100 years or so. Digitizing these data is necessary for studying historical earthquakes and for studying the characteristics of extreme storms and their occurrence in climate change analyses. SeisDig is an interactive Matlab based digitizing tool, designed to digitize entire seismogram "tif" images with a minimum of user interaction. Use of this software is subject to certain copyright restrictions.
Seismograms from the Berkeley Seismological Laboratory (BSL)
archives were electronically "scanned" at 400 dpi (dots/inch) using an
IDEAL FSS8000 full length scanner and stored as "tif" files. Each
scanned seismogram image consists of a series of quasi-parallel
sequential 30-minute or 60-minute traces that generally cover about a
day, with two partial traces at the beginning and end of each image.
One-minute timing marks usually cause small gaps at the end of each
minute. In addition to high amplitude earthquake signals, the arrival of
high amplitude storm waves at nearby shorelines [Bromirski and
Duennebier 2002] causes elevated noise levels called "microseisms",
both types of signals resulting in the occurrence of adjacent trace
intersections, overlaps, and reduced line thickness that complicate
digitization
(Figure 1).
Figure 1 (left): [click images to enlarge] A portion of the seismogram for Dec. 17, 1940 viewed with SeisDig shows the trace variation observed in pre-1962 Galitzin seismograms, with an example of faded trace lines at the seismogram edge that sometimes occurs in older seismograms. The scanned traces (thick white lines) show examples of adjacent trace intersections during relatively high amplitude noise levels. The projected digitized traces (blue lines) successfully track past these irregularities. The magenta trace is the current (active) trace under inspection. The Edge Line (vertical green line) is shown on the left, its position adjusted to conserve the total elapsed time of the seismogram. The full seismogram image is shown in the locator panel at the lower right, with the portion of the image shown in the main image panel outlined by the red box and the active trace identified by the dashed magenta line. The right Edge Line is set at the well defined trace ends. The paper seismogram physical trace length in inches is shown at the top of both the main SeisDig and the locator panels.
Digitized data are exported to disc in either standard or Matlab Seismic Analysis Code (SAC2000) formats. The SeisDig Export window (Figure 2) allows trace inspection, checking time conservation, identification of a particular event, and the input of necessary header information.
Figure 2 (right): The "Export (Preview)" pop-up window. The top panel shows the concatenated output of the digitized data for the full seismogram shown in Figure 1, containing 48 traces of approximately 30min duration each. The portion of the digitized trace outlined by the red locator box (600s) in the upper panel is expanded in the main panel for detailed inspection. The yellow dots indicate the sampled amplitudes to be exported. Note the semi-regular coda-like arrivals, indicative of microseisms generated by the arrival of ocean swell at nearby shorelines.
The temporal variation of raw spectral levels for digitized data for Jan. 1941 is shown in Figure 3. Energy concentrations on Jan. 8-9, 19-20, 22-23, and 24-25 are characteristic of double-frequency microseisms (high amplitude energy in the [0.09,0.25] Hz band at twice the wave frequency).
Figure 3 (left): Power spectral variation of digitized Wilip-Galitzin long-period seismograms during Jan. 1941, not corrected for either instrument response or calibration constants. Vertical stripes indicate either missing data or time periods where the scanned image quality was not adequate for digitizing.
Installation and Matlab/System
Requirements
SeisDig is available as a "tar" file
containing the Matlab m-files and sample image files. It should be
untared in the user's matlab directory, creating a SeisDig
directory with a data subdirectory containing sample image files, and
should be added to the user's MATLABPATH. Minimum system requirements:
Matlab V.6.0 or higher with Image Processing, Signal Processing, and
Statistics Toolboxes, and 256 Mb memory; Matlab v.6.1+ with 512+ Mb
memory recommended.
Image File Formats
Acceptable input image file formats are those that Matlab
function "imread" supports. These include TIF, JPEG, GIF, BMP, PNG, HDF,
PCX, XWD, ICO, CUR. Images of these types can be imported directly
without requiring conversion or modification.
Please email pbromirski@ucsd.edu to request the tar file. Users Manual is available in PDF format below.
Bromirski, P.D. and F.K. Duennebier (2002).
The near-coastal
microseism spectrum: Spatial and temporal wave climate
relationships,
J. Geophys. Res., 107(B8), 2166,
doi:10.1029/2001JB000265.
Bromirski, P.D., R.E. Flick, and N. Graham (1999).
Ocean wave height
determined from inland seismometer data:
Implications for
investigating wave climate changes in the NE Pacific,
J.
Geophys. Res., 104(C9), 20,753-20,766.