193 
Hardwoods that have had over 60 percent of their foliage 
removed may refoliate as the summer progresses. Hardwoods 
suffering less than 60 percent defoliation do not 
refoliate; conifers cannot regenerate their canopy within 
a growing season. Hence there exists an optimum viewing 
window of one or two weeks in late June-early July when 
defoliation has peaked and the damage is most noticeable. 
Data collected one or two weeks before or after this 
optimum viewing period may also be useful for delineating 
areas under attack. However, Landsat data which will be 
used to monitor gypsy moth defoliation must be obtained in 
this June-July time frame. 
Use of Landsat to Monitor Gypsy Moth Defoliation - A Review 
During the past decade, considerable research has been 
directed towards examining the use of Landsat satellite 
data to monitor gypsy moth defoliation of hardwood forests. 
Initial efforts concentrated largely on photointerpretive 
techniques, but the advantages of digital analysis were 
quickly realized and a rapid shift towards more sophisti 
cated digital image manipulation techniques took place. 
Rohde and Moore (1974) reported that gypsy moth defoliation 
could be delineated by manual interpretation of Landsat 
color composite images. However, the ability to quantify 
degrees of defoliation was hindered by uncalibrated bright 
ness and tonal changes. The authors suggested that digital 
processing of remotely sensed data might improve mapping 
accuracy. 
Another Landsat-based study on defoliation assessment 
included an investigation by Talerico et al. (1978) which 
described a quantitative photographic approach for deline 
ating various levels of insect defoliation by applying 
advanced photometric calibration techniques to aerial 
photography and Landsat imagery. They concluded that 
Landsat data were not only more economical, but also 
better than high altitude film for mapping defoliaton. 
In 1975, Williams reported a study which used a digital 
analysis procedure to map areas of heavy and moderate 
defoliation and healthy forest in eastern Pennsylvania. 
Classification results were subjectively analyzed and 
found to be representative of actual ground cover. 
However, errors of commission in which agricultural cover 
types were classified as heavy defoliation decreased 
classification performance. Later investigations by 
Williams and Stauffer (1978) significantly reduced these 
errors by utilizing registered, multitemporal Landsat 
imagery collected before and during defoliation. The 
image depicting healthy stand conditions (i.e., before 
defoliation) was utilized to separate all forest pixels 
from non-forest cover types, and the image collected 
during defoliation was utilized to assess the levels of 
defoliation only within the pixels previously identified 
as forest. 
A variety of image manipulation and data transformation 
techniques, known as Vegetation Indices (VI), were later 
evaluated by Williams et al. (1979) to determine if they