197 
The Pennsylvania Landsat mosaic was generated using the 
ten scenes listed in Table 1 and illustrated in Figure 3. 
These scenes were selected using the following guidelines: 
(1) summertime imagery acquired between May and September; 
(2) cloud-free data, or a maximum cloud cover less than or 
equal to 10%; (3) no apparent defoliation or other forest 
disturbance; (4) most recently acquired data which met 
guidelines 1 through 3 above; and (5) near anniversary 
coverage (i.e., all scenes from the same month of year if 
possible). 
The creation of the Pennsylvania mosaic and associated 
data base layers was complicated somewhat by the fact that 
the state lies within two UTM zones. This meant that all 
data layers within the data base had to be subdivided, 
registered to the appropriate UTM zone, and stored as two 
separate entities. The 78° meridian, which approximately 
bisects the State, serves to delineate UTM Zone 17 in the 
western half of the State from UTM Zone 18 in the eastern 
half of the State (Figure 2). 
To facilitate the image-to-map mosaicking process and 
associated geodetic accuracy assessment, each UTM zone was 
subdivided into four quadrants that were roughly equivalent 
to the following U.S.G.S. 1:250,000 scale maps: 
UTM Zone 17 
QUAD 1 - Cleveland, OH 
QUAD 2 - Canton, OH 
QUAD 3 - Warren, PA 
QUAD 4 - Pittsburgh, PA 
UTM Zone 18 
QUAD 5 - Williamsport, PA 
QUAD 6 - Harrisburg, PA 
QUAD 7 - Scranton, PA 
QUAD 8 - Newark, NJ 
The image-to-map registrations conducted by JPL personnel 
were done on a quadrant by quadrant basis, and the 
appropriate quadrants within each UTM zone were merged 
together to form the two halves of the Pennsylvania mosaic. 
In order to evaluate the geodetic accuracy of the completed 
mosaic, Goddard personnel selected a number of ground 
control points in each of the eight quadrants using 
1:24,000 scale U.S.G.S. topographic maps. The UTM 
coordinates for each ground control point were identified. 
The exact locations of these points on the mosaic image 
were determined using a series of display functions on the 
ESL/Interactive Digital Image Manipulation System (IDIMS) 
facility at Goddard (ESL, 1980). These locations were 
then used in an IDIMS utility function (ALLCOORD) to 
determine the ground control point residual values for 
each point in both the line direction (i.e., north/south 
offset) and the sample direction (i.e., east/west offset). 
Table 2 summarizes the number of control points identified 
in each quadrant, the average line and sample residuals, 
and the "worst case" line and sample offsets for that 
quadrant. In interpreting the results, note that a 
residual value of 0.0 in both the line or sample direction 
would be indicative of a perfect image-to-map registration. 
A review of Table 2 indicates that in most cases the sample 
(east/west) offset is greater than the line (north/south)