Summary of Application to:
Forest Insect Defoliation in Pennsylvania
Darrel L. Williams Mark L. Stauffer
Background and Objective of the Test: Several forest insect infestations have increased expo-
nentially in recent years and are threatening the major forest regions of the United States.
Epidemic forest insect populations often cause substantial economic losses by reducing an-
nual incremental growth and increasing tree mortality as a result of feeding upon favored tree
species. The responsible regulatory agencies must have accurate, timely and efficient methods
of detecting and mapping damage to the forest canopy in order to suppress these infestations
and maintain them at endemic levels. These agencies have been keenly interested in using
the repetitive and synoptic coverage provided by the Landsat satellites for monitoring wide-
spread forest insect damage. In response to this need, a research effort has recently been
initiated at NASA/GSFC to develop and evaluate digital image processing techniques that
will facilitate the assessment of the intensity and spatial distribution of forest insect damage
in Northeastern U.S. hardwood forests using Landsat remotely sensed data.
Study Sties: Recent outbreaks of heavy gypsy moth defoliation in the forests of central and
eastern Pennsylvania make this a particularly attractive study area for this test application of
Landsat imagery. A 30 by 30 kilometer Landsat sub-image of the Harrisburg, Pennsylvania
area is being utilized as a study site. This area is located in the Appalachian Ridge and Valley
physiographic province. In general, the ridges are heavily forested by mixed deciduous tree
species with oaks predominating, while the valleys are dominated by agricultural land uses.
The Landsat frame designated as path 16, row 32 covers this area. Landsat imagery repre-
sentative of healthy forest stand conditions was obtained on 19 July 1976, while 27 June
1977 imagery of the same area depicts several thousand hectares of gypsy moth defoliation
(Figs. 5.1 and 5.2).
Approach and Results: Forest insects generally attack a particular tree species or species as-
sociation having a fairly well-defined and documented geographic distribution. Thus, the
delineation of tree species via Landsat multispectral signature classification is of minor im-
portance for forest canopy damage detection. The recent development of techniques has
therefore been heavily slanted towards isolating and assessing changes in the forest canopy
due to fluctuations in the amount of leaf material per unit area.
Preliminary experimentation with the geometrically registered 1976 **non-defoliation" and
1977 **defoliation" Landsat sub-images of the Harrisburg area has shown that accurate isola-
tion of areas of defoliation is enhanced by subtracting the corresponding MSS bands of these
"before" and "'after" images to yield *'difference" images (Figs. 5.1c and 5.2c). The areas
fluctuating from more to less leaf canopy biomass between 1976 and 1977 appear as signifi-
cantly lighter shades of gray in the MSS band 5 difference image and as substantially darker
shades of gray in the MSS band 7 difference image. Similar trends have also been noted for
the MSS band 4 and 6 difference images. Color composites created by using 3 of the 4 pos-
sible "difference" images can depict the areas of change in unique color tones which are even
easier to interpret than the individual black and white displays. Unfortunately, leaf biomass
changes in agricultural crops are similarly represented in the difference images and are there-
fore potential sources of commission errors. However, almost all Landsat investigators
utilizin
“forest
tions. .
used t«
forest"
mask 0
applied
minate
multip!
by “VS
analysi
lineatir
from tl
Similar
tion di
approa
Theref
demon
Furthe
Ref
