194 
were appropriate for conducting gypsy moth defoliation 
assessments. These VI techniques, which had originally 
been developed to estimate agricultural and rangeland 
standing green biomass (Tucker, 1979), were investigated 
because: (a) the problem of estimating defoliation is 
essentially a problem of estimating the remaining green 
biomass or leaf area index in the forest canopy, and (b) 
it had been suggested in the literature that the use of 
these techniques reduced variations in scene reflectance 
properties caused by terrain relief (Vincent, 1973; Goetz 
et al., 1975; Justice, 1978), which is prevalent throughout 
the forested regions in Pennsylvania. All of the Vi's 
tested by Williams et al., consistently discriminated 
heavy defoliation (60 - 100% canopy removed) from healthy 
forest. However, it was noted that areas of moderate 
defoliation (30 - 60% canopy removed) were confused with 
healthy forest. 
The success of the work described above led to the initi 
ation of the joint research project with Pennsylvania in 
late 1979. As part of the JRP effort, Nelson (1981) 
continued to examine these same Vegetation Index techniques 
to determine the most cost-effective and accurate method 
for defoliation assessment. He found that the simplest 
transformation, known as the Ratio Vegetation Index (RVI), 
was as effective as all other transformations for discrimi 
nating among forest cover conditions. Typical accuracies 
for separating heavy defoliation from a combined moderate 
defoliation/healthy forest cover condition fell in the 
range of 75 to 80 percent. The RVI, which is derived 
simply by dividing the MSS band 7 response by the MSS band 
5 response for each pixel, was also the least computation 
ally intensive method. Therefore, Nelson suggested that 
this technique be selected for subsequent assessments. 
FOREST DEFOLIATION ASSESSMENT PROCEDURE 
The work completed by Williams and Stauffer (1978), Williams 
et al. (1979) and Nelson (1981) provided the framework for 
automated defoliation assessments using Landsat multi- 
spectral scanner data. The procedure, as outlined below, 
requires four steps. 
Step 1 - Creation of Healthy Forest Classification Mask 
Cloud-free, summertime Landsat imagery over the selected 
forest site is obtained prior to insect infestation. 
Using computer-aided analysis techniques, this image is 
classified into two major cover types: forest and non 
forest. Pixels classified as forest are assigned a value 
of one. All other pixels are assigned a value of zero. 
This classified image is called the "1/0 forest/non-forest 
mask". 
Step 2 - Application of Forest Classification Mask to a 
Defoliation Image 
A Landsat image that corresponds to the geographic location 
of the cloud-free image (Step 1) and that has been 
collected during or immediately following peak defoliation 
is obtained. This image is digitally registered, or 
overlaid, onto the "1/0 forest/non-forest mask". This