field,
| from
uated
/ high
from
urally
c and
es in
chers
urally
nding
inous
multi
in the
ilator,
ering
itions
Alarly,
inous
sd to
atures
action
ensor
s and
iation
s and
s and
and
in a
and
r data
al and
ques,
luster
e and
ensor
n Site
hosen
iinous
incies
»ment
These
d and
IS are
lakes.
ie the
pruce
| plus
1989).
> the
monly
referred to as lineaments in remotely sensed images.
They are a source of natural cyclicity in spectrally
analyzed images. Alternating stratigraphic lithologies in
folded mountainous terrains with colluvial sluff would be
expected to contribute a natural periodicity, and,
furthermore, vegetation establishment and patterns are in
part controlled by slope, aspect, micro-climate and
elevation as well as chemistry and moisture associated
with rock strata and soil type. The orientation of stream
pattern development in low order streams can be
influenced by the direction of systematic jointing in rock.
Without better understanding of the contribution of natural
biotic and geologic frequencies, a full evaluation of images
periodicities and spatial frequencies is not likely to be
attained. The comparative feature extraction and
evaluation techniques developed are expandable to other
mountainous regions of the world.
Data-requirement for Simulated Multistage Spatial and
Spectral Bands in a Near Temporal Time Frame
Simulated multi spectral sensor data derived from digitized
1993 and 1995 CIR Photography ranging from 550 nm, to
850 nm band with 0.1 m and 11.0 m IFOV resolution are
employed. The CIR 1:12000, 1993 and 1:6000, 1995
photography was optically scanned with a Nikon AX-1200
flat bed scanner with Scantouch and ADOBE Photoshop
software at 2400 , 1400, 650, 150, and 50 dpi with digital
output of 1400, 550, 200, 120, 60 and 30 dpi respectively.
Also using differing filters for red, green, and blue with f
stop settings at normal or increased to 0.75 f stop
increase for scanning for separate bands corresponding to
0.5-0.6nm, 0.6-0.7nm, 0.7-0.85nm of the CIR respectively
was achieved with this system. These data are then
compared with ground control data collected and GPS
registered in July and October of 1995. These data, along
with elevation, are incorporated in a geobiophysical
modeling system software for performing the various
digital multi-variate mergers analyses .
Analysis
The data sets are subjected to principal components
analysis, supervised sampling procedures for the same
aerial extent for each site with close geographic position
maintained for signature analysis in feature separability,
histogram analysis of the samples and a comparative
analysis of each sample variance for category separability
and cluster based feature extraction techniques for
spatial evaluation(Yuill, et al, 1991; Bloemer, et al, 1994;
Oberly and Brumfield, 1991). The features and sample
areas represented are: maple beech (6096/4096)
association; maple beech (4096-6096) association; maple
spruce (6096/4096) association; red pine spruce (8096/20)
association; red pine spruce (8096/1096) association;
spruce yellow birch (8096/2096) associations; field meadow
shrub rock (50%/20%/20%) association; field meadow rock
soil (40%/30%/20%) association; road/limestone meadow
(90%/10%) association; open canopy maple beech
(30%/40%/30%) association. These features are then
compared to the GPS registered ground control field data
in a geobiophysical modeling system for spatial, statistical
and mathematical analysis for evaluation. Computer
61
graphic displays are utilized for comparative evaluation of
sample data sets.
RESULTS AND CONCLUSIONS
The differing IFOV's ranging from 0.1-11m with
comparative analysis as stated in technique as given in
the example frequency histograms for infra red, red and
green respectively of figures 1, 1.6m and 2 , 0.2m,
demonstrate that larger IFOV than about 1.0 m results in
a more fragmented spectral frequency sample set than a
set less than one meter. The characteristics of the spatial
distribution of the vegetation assemblages, compared to
field mapping of the vegetation, suggest an intermittent
discontinuity of spectral frequency that has resulted from
integration of the spectral energies apparent at higher
resolutions (IFOV). In fact, histogram display of 2.6m
IFOV resulting from groupings of frequencies comparable
to ground cover distribution provided a classification
similar to higher frequency (IFOV) cluster classification
(Figures 3 and 4). These results are particularly
noticeable for sample features that are spatially and
spectrally similar (Figures 1 and 2). However, it should be
noted that at higher IFOV, the frequency of the data
numbers increases while the apparent variance
decreases; the separability of the spectral types increase
for the mean of the type in each spectral band (Figures
5 & 6). This suggests that increased separability of
spectral features that are spectrally similar, figure 7 & 8,
may be further separated by higher spatial frequencies
that provide more of a spectral continuum which may be
spatially associated. For a fixed number and width of
Multi spectral bands, with decreasing IFOV, the data
become more continuous and the individual data
frequency variance decreases in a constant sample area
of fixed size within a particular vegetation category. This
provides better opportunities for characterization of the
sample within its population and the variance of the
population.
The next generation of air and space borne sensor
platforms need significantly higher spatial and spectral
resolution if foresters, earth scientists and planners are to
monitor, inventory and evaluate the resource conditions
and rejuvenation capacity that mountain forest
ecosystems provide. Factors, which contribute to the
degeneration of the mountainous regions of the world,
must be investigated, to mitigate the degradation of the
intricate cycles that support forest ecosystems. These
results demonstrate the validity that higher spatial
resolutions are necessary to monitor and evaluate the
higher frequency variability of mountainous terrain in a
timely fashion for longer term forest ecological processes
interaction in global change.
REFERENCES
Adams H S., Stephenson J.L., 1989. Old-growth red
spruce communities in the Mid-Appalachians, Vegetation,
85, pp.45-56.
International Archives of Photogrammetry and Remote Sensing. Vol. XXXI, Part B7. Vienna 1996
