Section mapped spruce budworm infestation during
overflights with a helicopter and by means of photo
interpretation. Four damage classes were mapped
based on percentage of crown affected: light
(0-25%), moderate (26-50%), heavy (51-75%) and severe
(76-100%).
A spruce budworm infestation is a difficult remote
sensing detection problem. Damaged spruce needles,
which turn reddish-brown, may occur on only a portion
of the tree. Due to the resolution of the sensor, a
significant difference in signature between healthy
and infected trees may not occur. This detectable
reddish-brown stage may also only have a short bio
window of four to eight weeks depending on the sever
ity of weather conditions. For example, heavy rain
or winds, may knock the needles off the trees. Past
infestations in Saskatchewan have been of limited
extent in isolated areas further increasing the dif
ficulty of detection. Locally, these small areas,
however, can have an important economic impact on
stud mills or plywood mills.
High resolution multispectral data from an electro-
optical pushbroom scanner have been successfully used
by Ahern et al. (1986) to map current year defolia
tion in New Brunswick, but only at one level -
presence or absence of defoliation.
STUDY AREA
The study site is located approximately 70 km south
of the town of Hudson Bay in eastern Saskatchewan,
Canada, covering an area of approximately 25 km 2
along the Etomami River (NTS map sheet 63D/7, 63D/8)
(Figure 1). The area is generally of low relief with
numerous small lakes and wetland areas. The forest
Figure 1. Location of study area - scale 1:250,000.
cover is mixed hardwood-softwood with aspen (Populus
tremuloides, Midix), white spruce, balsam (Abies
balsamea, Mill.), black spruce (Picea mariana,
Mill.), and jack pine (Pinus banksiana Lamb.) domi
nating. White spruce, which is the main species
affected by the budworm, occurs in mixed stands asso
ciated with trembling aspen. Pure stands of white
spruce rarely occurred in the study area.
DATA ACQUISITION
Three transects of MEIS-II data were obtained at the
beginning of July 1985 at 5.5 m resolution (Table 2).
Colour infrared aerial photography was acquired in
the latter half of July and Thematic Mapper data at
the end of August at which time the red-brown stage
of the needles was still evident although some
needles had fallen off the trees. Ground data were
collected at the end of July. Airborne, oblique
colour slides were taken during the ground data
gathering period by helicopter. A number of stops
were also made to check individual trees for levels
of infestation.
DATA ANALYSIS
MEIS-II data
The MEIS-II data were received in the form of com
puter compatible tapes and were transferred from tape
to disc on a DIPIX ARIES-II* image analysis system.
Visual inspection of channel one of a total of eight
channels showed a low signal-to-noise ratio and so
was eliminated from subsequent analysis. The
remaining seven channels were contrast stretched by
using a linear or log transform and displayed on the
colour video monitor. A colour image formed using
channel 7 (red), channel 5 (green) and channel 4
(blue) showed a shift in hue towards the edges of the
swaths. Subsequently the edges were ignored and only
the central portions of each swath analyzed. The
natural colour image with channel 2 (red), channel 4,
(green), and channel 5 (blue) showed an area of bud
worm infestation which corresponded to the areas of
infestation on the colour infrared aerial
photographs.
A number of biomass index enhancements using the
formula:
BI = SQRT ((F(1) - F(2)/(F(1) + F(2)) + 1.0
where F(l) and F(2) represent the various channels,
were also produced. The biomass index produced off
set and gain values which were used to ensure that
the output feature values lay within the intensity
range of 0 to 255. Various biomass indices, using
different channels, were combined through the red,
green, and blue guns of the video display monitor
(Table 3). Principal component and Martin Taylor
enhancements were also produced (Taylor, 1973).
Supervised maximum likelihood signature generation,
unsupervised maximum likelihood signature generation
and a real-time parallelipiped classification were
performed.
Thematic Mapper data
Data from quadrant two, path-row 35-24 taken on
August 26, 1985 on a computer compatible tape was
transferred to disc on the ARIES-II image analysis
system. The data were rotated 13 degrees and
resampled from 30 m to 10 m to compare them to the
* Mention of or failure to mention commercial trade
names does not imply endorsement or criticism by
the Government of Canada or the Government of
Saskatchewan.