Geology Reference
In-Depth Information
5
3
4
6
4
Fig. 4.
Field photograph of the Picco di Carnizza section
illustrating the lines of grey-scale analysis used by Cozzi
et al
. (2005, Geological Society of America data repository
item 2005157). The three lines of section (dark, sub-parallel
lines) in this image were used as the basis for the spectral
analysis of Milankovitch periodicities in the succession.
Sampled at 2000 dpi, the grey-scale data series derived
from this image is the product of over a 10,000-fold
reduction in stratigraphic resolution.
Fig. 5.
Graphical representation of the signifi cance of
differences in the angle between a rectilinear pixel grid
of a scanned image and the apparent up-section direction
recognized on a fi eld photograph. Five traces of equal
length are positioned on the pixel grid where the number
of cells intersected by those traces ranges from a minimum
of three to a maximum of six. Such dependence upon
angular position can have a major control on the interpre-
tation of spatio-temporal characteristics interpreted from a
grey-scale series.
ANGLE OF GREY-SCALE ANALYSIS
The analysis of grey-scale data taken from
outcrop photographs is open to several criticisms.
First, the scale of fi eld photographs (e.g. Fig. 4)
can result in an astonishingly large reduction in
the resolution of data available for analysis.
Second, variation in grey-scale value could be
interpreted to be driven by differential weath-
ering taken to be facies-specifi c in its character.
Yet, establishing a correlation between grey-scale
intensity and directly observed lithological com-
position is diffi cult. That is, without a set of
baseline data that ties composition directly to
grey-scale intensity through fi eld checking, the
validity of any analysis drawn from the data is at
best suspect.
The somewhat obvious drawbacks discussed
above mask other fl aws in the grey-scale approach.
An essential requirement of the spectral analysis
technique is that there be a clear understanding
of the spatio-temporal spacing of the data in the
series. Figure 4, however, illustrates several prob-
lems. First, the composite section used in the
spectral analysis was created from three separate
lines taken at different angles across the image.
Apparently, this was done in an attempt to keep
the line of analysis perpendicular to strike as it
appears on different faces of the outcrop. The result,
however, is that the sampling across the rectilin-
ear grid of pixels results in different amounts of
vertical section per pixel (Fig. 5). As illustrated by
the fi ve arrows of equal length, the angle between
the section and the grid controls how many
pixels are represented in the section, and thus
the average thickness represented by each pixel.
In this simple example, the number of intersected
pixels ranges from a maximum of six to a min-
imum of three. In the actual scan of the Picco di
Carnizza, the differences in angle between the
three traces are not as large as illustrated in Fig. 5,
but the integrated effect over hundreds of pixels
is signifi cant even when dealing with small angu-
lar differences. As such, differences in the pixel-
thickness relationship between outcrop segments
introduce artifi cial trending to the dataset. The
angle of section problem, while real, is perhaps
minor in the Picco di Carnizza example. There
remains, however, a second more signifi cant prob-
lem associated with pixel scaling in the analysis.
ANGLE OF THE OUTCROP
The use of fi eld photography as a data source
for quantitative analysis presents a number of
complex problems, the most obvious and most
vexing is the complex geometry of the outcrop
surface. Natural outcrops such as the Picco di
Carnizza or artifi cially created outcrops such
as road cuts are subjected to the effects of
