' image identified
from darker to
change indicated
This processing
f MA 4D where
A horizontal line,
the North Bliss
There were small
the Camp Hueco
ding the various
cas could indicate
; however, these
iagnitude than the
he playas of the
described in the
rocessing (section
ct classes to be
Che histograms in
g each field point
EY
S
RSS
SEN
NX
NWN
A
VA
%
2
%
z
7
AA
JA
5
4
7 [A
| m E p "
p p, p A
Z ré 4 [ki % Li
10 11 12 13 14 15
ass Value
SNS
listograms
ted a few trends.
ipacts exhibited a
ectral class (i.e.
her hand, peaked
13 and 14). The
pacts may have
oints), but a peak
how higher class
ipacts and lower
ce impacts. The
h Recent impacts
iore exposed soil
9, the lower class
mpacts may be
lunal spaces.
These trends are also in agreement with
the basic observation made in all the sub-projects
that an increase in soil exposure increases the
brightness of the signature recorded by the
satellite. ^ Conversely, the more interdunal
vegetation that is present tends to "darken" the
basic soil signature found in the study area. These
findings indicate that the age classes may be a
very good indicator of vegetation regrowth and
vehicle impacts.
7. DISCUSSION
7.1. Detection
The results of this first sub-project
answered basic questions of detectability of
impact. In some respects, detection was
dependent on level of use. This was particularly
true of roads which fell into three categories
(paved, unpaved, and less frequented trails).
Perhaps most notably, the spatial resolution of the
satellite sensors did not necessarily represent the
limiting factor in detecting impact.
Both Landsat TM and SPOT-XS imagery
were shown to be useful for detecting areas of
heavy impact and decreasing levels of impact away
from unimproved roads in the area. Given
appropriate ground truth, the observed halo effect
may offer an opportunity to quantitatively measure
the level of training impacts.
7.2. Monitoring
Assessment of the use of image
differencing for monitoring change due to training
impacts was complicated by a lack of accurate,
quantifiable maneuver rate data. The main
difficulty stemmed from inability to separate
natural changes from man-made changes.
The method of image differencing requires
accurate image registration and spectral calibration
of the multi-date images. Minimal seasonal spread
in image dates is also desirable. Anniversary dates
should be used where possible.
For further studies of this type on Ft.
Bliss, imagery should be selected from mid to late
June because it is typically the driest time of the
year and offers the highest sun angle. Periodic
aerial photo surveys and ground surveys should be
developed and maintained for monitoring changes
in the landscape. This work would also benefit
from local knowledge of the terrain, of seasonal
changes, and how these changes influences the
imagery.
131
7.3. Classification
This sub-project served as a preliminary
classification of impacts due to vehicle activities.
While statistical correlation was not made, a
positive relationship between spectral classes and
age of impact appeared evident. Image
classification should be accompanied by extensive
field work, followed by rigorous procedures of
accuracy assessment. In this case, however,
constraints of data content and quality prevented
full development of the classification. The spatial
distribution of the survey points in the study area
was also limited. Results would be improved with
more widely distributed points. Most of the
analysis relied on data that were collected for other
purposes (i.e. without satellite image classification
in mind). Thus, the work here can be considered a
first-approximation with many possibilities for
further analysis.
The number of spectral classes chosen for
this study was arbitrary. Determination of the
optimum number of information classes would
make these results more reliable. Finally, and
possibly most importantly, if data on the level of
impact is desired, a ground survey should be
conducted to collect specifically this kind of
information. Recommendations for such a survey
would include sampling points which would
adequately represent the various types of terrain
found in the MAs. In addition, the survey should
be conducted as close to a Landsat overpass as
possible. A rigorous classification of the
multispectral image should be performed and the
survey points and spectral classes then correlated.
Associations of particular levels of impact with
spectral classes might be found. If this is the case,
these classes could be isolated and used to produce
an impact classification map of the MAs. Land
use planners would then have a powerful tool to
aid in maintaining the terrain for continued realistic
training exercises.
8. CONCLUSIONS
This work sought to determine the utility
of remote sensing for detecting, monitoring, and
classifying training impacts at Ft. Bliss. The
following conclusions were drawn:
* The spectral capability of the satellite sensor was
more of an influence than its spatial resolving
power in detecting trails and other impacts.
* A "halo" effect existed in the multispectral
imagery which can be used to identify gradations
in intensity of impact.
* SPOT's spatial resolution helped to detect
impacts; the advantages of Landsat's higher
spectral resolution was inconclusive.