Geoscience Reference
In-Depth Information
(e.g. Kanamori et al. 2006). Here, however, we centre our interest in the investiga-
tion of full waveforms preserved on old seismograms, which can, in general, give
us a more complete picture of the earthquakes process. Consequently, the objective
is to convert our analogue record, supported on paper, to a digital time series of
seismic ground motion, ready to use for any of our waveform analysis tools. The
first step of such a procedure, just like for the purpose of conservation and digital
storage of old records, is the scanning of the seismogram as a raster image.
The first key decision is the dpi density the raster image should be acquired to
preserve the resolution of the original seismogram. The answer is tied to the dimen-
sions of the trace to be extracted: It is unlikely to find traces thinner than 0.1 mm
on smoked paper. To have a good definition of the trace on a digitalized image,
we should have at least 3 pixels covering the thickness of the trace (i.e. 762 dpi).
Such a resolution allows to properly defining the centre of the trace. In the case of
photographic paper records, line width is much larger, and just half this estimate is
enough. SISMOS and EUROSEISMOS projects adopted a basic scanning density
of 1,016 dpi, this is, 4 pixels in 0.1 mm, for all records.
As seismograms are always monochrome records (black on white for photo-
graphic records, white on black on smoked papers, a unique color on white in the
case of ink paper records) grayscale scanned images are enough to keep the infor-
mation about the trace without any loss of resolution. For the case of film scanning,
dpi density should be adjusted to the scale of the filmed seismogram to maintain
the resolution of the original image. Such parameters (1,016 dpi, grayscale) impose
the record dimensions: For the example of a WWSSN record (900
×
300 mm) the
scanned image size will be
440 MB. The efficiency of file compression algorithms
depends on the image characteristics, and is usually good only for photographic
or ink recordings. Only recently, image processing with standard PC's, and the
management of databases containing thousands of these files at large facilities has
become functional. Finally, prior to trace extraction, it is useful to optimize the
characteristics of the raster image enhancing the contrast, brightness and other pa-
rameters adjustable within standard image processing software.
Following, the waveform of the seismic trace of interest (usually just a frag-
ment of the section contained in the image) must be extracted and pre-processed
for further seismic analysis. This involves the digitization itself and several steps of
trace correction. On early studies involving the use of digitized old seismograms,
the digitization of the trace was performed from the original records, or enlarged
copies obtained with photographic techniques, with digitizing tables (ex.: Adams
and Allen 1961, Howell 1966, Wickens and Kollar 1967, Batll o et al. 1997). Even,
some studies used digitized points obtained directly on the seismograms measuring
with a rule (Samardjieva et al. 1997). Actual procedures typically avoid the use
of special hardware and involve the use of computer software to extract the traces
from the scanned images. Several commercial or freeware programs, not specially
designed for seismological purposes, are available for this step. Most of them in-
volve the manual picking of points. Whichever will be the program, control on the
original scale of the record, i.e., accurate control of the exact coordinates of the
picked digitization points, must be carefully maintained.
∼
Search WWH ::
Custom Search