recognition. Forest association variance in heavily
dissected terrain is revealed by preliminary research
results by Yuill, 1992, Mills et al, 1963, and others.
USDA/USFS preliminary investigations, and Yuill, 1992,
indicate that a one-half meter to fifteen meter (0.5 - 15.0)
IFOV/multi-spectral electro-optical systems are useful in
identifying vegetation species and associations.
Previously, TM data were evaluated utilizing an optical
image processing system. This parallel set of processes
was performed by digital Fast Fourier Transforms (FFT)
and filters (Oberly and Brumfield, 1991). Results showed
a three to five data number variance, in a very low
frequency sinusoidal wave, for the TM band 1 data.
Further, utilization of the MIPS digital image processing
systems on the October 2, 1989 TM band 2 data
demonstrated that more apparent patterns were present in
the West Virginia forested mountain data (Bloemer and
Brumfield, 1992). These periodicity and interference
patterns can greatly impact geobiophysical modeling of
data, e.g., forest canopy parameters such as the 'red-edge
effect, in assessing the health of the forests(Oberly and
Brumfield, 1991; Bloemer and Brumfield, 1992; Gross, et
al, 1987). Further, these parameters in pattern
recognition, must be taken into account, for they can affect
the validity of a model's outcome(Peterson, et al, 1988;
Wessman, et al, 1989; Card, et al, 1988). The contention
of these researchers is that the periodicities and higher
spatial frequencies occurring in nature can be effectively
utilized to enhance feature extraction given the spatial
resolution to identify these higher frequency associations.
This research investigates varying spatial frequencies
without modifying the spectral nature of the data other
than transforming from an optical to a digital domain.
Utilizing spatially adjusted variables can lead to more
reliable modeling of geobiophysical data for forestation
cycles, geologic and hydrologic processes for long term
forest ecological assessment and global change impact
evaluation (Wriggley, et al, 1985). These variables may
involve the biosphere, atmosphere and climate, particularly
with emphasis on altitude, latitude, slope, aspect and
micro-climate variations in mountainous terrain (Tucker, et
al, 1985).
This paper is part of a scientific research program in the
early stages of design and strategic implementation
planning and investigation. Because of West Virginia's
dissected plateau mountainous terrain, the identification of
forest associations in current SPOT, Landsat TM and MSS
systems do not provide spatial resolutions of suitable
accuracy in forest association/species identification or
physiognomic analysis in ecosystem evaluation,
management or global change mitigation(Bloemer, et al,
1994). One result, as a component of the research, is the
US Forest Service's evaluation of very high resolution
multi spectral electro-optical systems from aerial platforms
for imaging forest associations and species for visual
interpretation and identification with a high rate of
success(Yuill, 1992). These electro-optical systems have
similar bands as NASA's LANDSAT designs involving
visible and near infrared spectra. Their IFOV range is
typically 0.5m to 11m. We contend that higher resolution
systems are increasingly using spatial frequencies that are
likely to be problematic but can enhance feature extraction
60
in the digital and optical domains (Oberly and Brumfield,
1991; Bloemer et al, 1994 and Yuill, 1992).
Digital and optical feature extraction of sensor data from
aerial platforms are investigated, compared and evaluated
for pattern recognition. This is accomplished via very high
spatial resolution simulated multi-spactral data from
scanned photography for forest associations. Naturally
occurring periodicities of geologic, hydrologic, climatic and
biologic features tend to have higher frequencies in
mountainous terrain. Therefore, these researchers
contend that understanding the expression of naturally
occurring spatial frequencies is germane to understanding
forest ecological and geologic patterns in mountainous
terrain associated with higher spatial frequency multi
spectral system data. This research is focused on the
forest associations data for future generation simulator,
aerial and orbital sensor systems design and engineering
for longer term evaluation of the ecological conditions
associated with existing research programs, particularly,
those using forest and earth resources in mountainous
terrain.
RESEARCH METHODS AND TECHNIQUES:
In nature there are spatial frequencies related to
geologic, hydrologic, climatic and biologic features
particularly in mountainous terrain. A data collection
system with current non military orbital high IFOV sensor
systems cannot resolve the high spatial frequencies and
is, therefore, inadequate in forest species and association
identification in mountainous terrain. These systems and
data were investigated at the participating universities and
NASA with the optical transform instrumentation and
digital image processing software available in a
geobiophysical modeling environment (Oberly and
Brumfield, 1991). It is proposed to look at the sensor data
in question through parallel investigations of optical and
digital transforms and feature extraction techniques,
partially funded through a NASA EPSCOR grant. Cluster
analysis, discriminate analysis, analysis of variance and
field analysis are utilized in evaluating natural and sensor
identified spatial frequencies.
Study Area - Rationale for Appalachian Mountain Site
Selection
The Spruce Knob USGS 7.5 minute quadrangle is chosen
for its representative eastern United States mountainous
forested conditions with high natural spatial frequencies
and periodic disturbances by forest management
practices, agricultural and recreational impacts. These
mountains are northeast/southwest trending of folded and
heavily eroded sedimentary rocks. Stream patterns are
trellis accented with natural and manmade ponds/lakes.
Strata of Pennsylvanian and Ordivician age underlie the
area. Northern deciduous hardwood forest and red spruce
with patches of red pine compose the 1000 m plus
elevation vegetation communities(Adams et al, 1989).
Further, sources of periodicities may well be the
underlying geology. faults, joints, fractures, commonly
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
-7L" (0D Q9 "CO CX «2 —'v € (qv vy
Mr
2e TVS HM) ss Or SO 0 2 0 go ry
©
t0 OQ (C)0 QR TAQ]