847
beetle infestation with an accuracy of 91 percent
using 1:3000 scale photographs. Encouraging
results have also been given in another report
(Murtha, 1985) in which colour infrared large-scale
photography (1:2000) was used to help distinguish
between healthy and unattacked Engelmann spruce
(Picea englemannii Parry).
The objective of this study was to differentiate
between green attacked and unattacked trees using
imagery acquired with an airborne MEIS (Multi
element Electro-optical Imaging Sensor).
A previous study showed that MEIS data at 1.4-m
resolution was sufficient to permit the detection of
more trees with the appearance of red attack than
3.4-m resolution MEIS imagery and conventional
aerial photography (Kneppeck and Ahern 1989),
emphasizing the need for data with high resolution.
2. STUDY AREA
The area chosen for the Investigation is located
approximately 35 km southeast of Cranbrook, British
Columbia near latitude 49° 23’ N, longitude 115° 17’
W in the Elk Forest. The terrain is fairly flat with
elevations ranging from 823 m to 884 m above sea
level. During the site selection process several
areas containing red attacked trees were monitored
during the summer months for signs of beetle
spread to adjacent trees. Image data was then
acquired over those areas exhibiting extensive
green attack (symptoms, as mentioned earlier,
included pitch tubes and boring dust found near
the base of trees). Trees under attack were mainly
lodgepole pine. Other species, not attacked by the
mountain pine beetle, were black cottonwood
(Populus trichocarpa Torr. & Gray), white spruce
(Picea glauca (Moench) Voss), western larch (Larix
occidentalis Nutt.), and douglas fir (Pseudotsuga
menziesii (Mirb.) Franco).
3. DATA ACQUISITION
Four lines of MEIS data were acquired on high-
density tape September 1, 1989 from 579 m, 1539 m,
and 3414 m above ground level. The corresponding
instantaneous field of view was 0.4 m, 1.0 m, and
2.2 m, respectively. The MEIS data consisted of a
blue band (448 nm), a green band (549 nm), a red
band (676 nm) and five infrared bands (711, 722,
735, 751, and 875 nm). The minimum and maximum
limits of each band are provided in Table 1. These
bands closely resembled those determined by Ahern
(1988) to have potential for detecting mountain pine
beetle stress at an early age. Colour infrared aerial
photography was obtained concurrent with the MEIS
data.
The overflight was timed such that it was
subsequent to the period of beetle spread, which is
usually from mid to late July. One month following
the MEIS data acquisition trees were checked on the
ground for mountain pine beetle green attack. A
tree was considered at the green attack stage if all
of the following conditions were met:
needles were still green
reddish coloured boring dust was found at the
base of trees and pitch tubes were formed
around entrance holes on the bark
by cutting into the tree with an axe, blue stain
fungi was found in the sapwood.
A total of 256 green attacked trees and 250 healthy
trees were identified in four principal areas which
we shall refer to as Galloway, Jaffrey, Rosen Lake
and Elko. Every tree was tagged with a number,
and its location was marked onto an acetate sheet
attached to a colour infrared photograph. A record
was kept of each tree which included its location,
tag number, and general health and a 35 mm
photograph of each tree was obtained.
4. DATA ANALYSIS
This section reports on the analysis conducted to
date for the 0.4 m resolution MEIS imagery acquired
over Jaffrey and Galloway.
The MEIS data for all lines were copied from high-
density tape to computer compatible tape.
Subscenes were made of each of the four areas
mentioned earlier and histograms were generated
for each band. A colour infrared image was made
using bands corresponding to the infrared (751
nm), red (676nm), and green (549 nm) portions of
the spectrum, displayed as blue, green and red
respectively. The data were linearly contrast
stretched before display.
Training areas were made for 156 trees whose
locations had been marked on the colour infrared
photographs. The training area for each tree
included only fully sunlit pixels, and excluded
shadowed or partially shadowed pixels. Statistics
extracted for each tree included the mean digital
signal level for each band. Table 2 provides a
summary of the number of attacked and unattacked
trees used in the analysis. For each tree the mean
digital signal level per band was transferred into
the RS/1 statistical package.
It was apparent from the data that the digital signal
level (DSL) of attacked and unattacked trees was
generally highest in the 711 nm band. To reduce
variations caused by differing illumination between
trees, the DSL of each band was divided by the DSL
of the 711 nm band for every tree. Using this
normalized data, the average DSL for the attacked
and unattacked trees was calculated. Figures 1 and
2 show the average DSLs of attacked and
unattacked trees in Jaffrey and Galloway
respectively. The standard deviations of the DSLs
for each band were also calculated and displayed on
Figures 1 and 2 as the average of the standard
deviation for the attacked trees and the unattacked
trees.
Figures 1 and 2 show that there are noticeable
differences between attacked and unattacked
lodgepole pines at both sites. However, the
differences in digital signal level between the two
sites are larger than the differences between
attacked and unattacked trees within a single site.
While the Galloway trees generally exhibit a spectral
shape which is typical for healthy green vegetation
(see Ahern, 1988 for samples of foliar spectra of
attacked and unattacked lodgepole pine), the
Jaffrey trees have higher visible wavelength
Intensities and lower near infrared intensities than
one would expect. This difference between the
Galloway trees and the Jaffrey trees is also visible
on the colour infrared photography acquired during
the mission: the Jaffrey trees have, in general, an
appearance which is more gray and less pink than
the Galloway trees. These differences suggest a
lower level of vegetation vigour for the Jaffrey
sample, which may be related to different growing
conditions.
Additionally, the differences between attacked and
unattacked trees are generally smaller than the
standard deviation of an individual measurement.