T
increased expo-
nited States.
reducing an-
on favored tree
icient methods
ese infestations
ted in using
nitoring wide-
ecently been
chniques that
insect damage
of central and
application of
Pennsylvania
idge and Valley
eciduous tree
ral land uses.
agery repre-
while 27 June
h defoliation
5 OT species as-
3 Thus, the
of minor im-
echniques has
rest canopy
oliation” and
accurate isola-
bands of these
). The areas
pear as signifi-
ntially darker
sen noted for
of the 4 pos-
which are even
7, leaf biomass
and are there-
vestigators
- 1743 -
Pennsylvania/Williams and Stauffer
utilizing digital classification techniques have reported the ability to accurately delineate
“forest” from ‘‘non-forest’ using Landsat imagery representative of healthy forest condi-
tions. A hybrid classifier combining parallelopiped and maximum-likelihood algorithms was
used to classify the 1976 “non-defoliation” Landsat sub-image into “forest” and “‘non-
forest” categories (Fig. 5.3a). This forest cover classification map was used to create a binary
mask of “1’s” for forested areas and '**0's" for non-forested areas. This binary mask was then
applied cell by cell to the 1977 Landsat sub-image of defoliation conditions in order to eli-
minate all non-forest cells (Fig. 5.3b). In simple mathematical terms, all 1977 pixel values
multiplied by **0's" become zeros and are represented as black, while those cells multiplied
by “1’s” are unchanged in value and their measured radiances are still available for further
analysis. This masking approach eliminates the potential of errors of commission when de-
lineating forest insect defoliation damage, as all non-forest land areas have been removed
from the data set.
Similar techniques could be utilized to isolate and monitor the various types of forest altera-
tion discussed in the other sections in this report. The major advantage of this type of
approach is that only the particular land use(s) of interest is maintained within the data set.
Therefore, future processing costs are reduced as less image data has to be analyzed. Also, as
demonstrated, the potential of commission errors is often reduced or eliminated.
Further Information
References
Williams, D. L. and M. L. Stauffer. 1978. Monitoring gypsy moth defoliation by
applying change detection techniques to LANDSAT imagery. Proc. of the Sym-
posium on Remote Sensing for Vegetation Damage Assessment. American Society
of Photogrammetry, Falls Church, Virginia. 7 p.
Williams, D. L. 1975. Computer analysis and mapping of gypsy moth defoliation
levels in Pennsylvania using LANDSAT-1 digital data. Proc. of the NASA Earth
Resources Survey Symposium. Vol. 1-a: Technical Session Presentations. NASA/
Johnson Space Center, Houston, Texas. pp. 167-181.
Experimenters
Darrel L. Williams, National Aeronautics and Space Administration, Goddard Space
Flight Center, Earth Resources Branch, Greenbelt, Maryland, 20771, U.S.A.
Mark L. Stauffer, Computer Science Corporation, 8728 Colesville Road, Silver
Springs, Maryland, 20907, U.S.A.