identified and corrected, a mosaic was produced which exactly overlaid
the data base when images printed by a Versatec electrostatic plotter
were visually compared. It was found that the mosaicking process was
very sensitive to "bad" control points, where the automatic correlation
procedure had selected the wrong matching point. These matching points
have to be located by comparing mean residuals and correlations of
pairs of points, using subjectively-determined criteria. Inclusion of
bad points resulted in local errors in the overlay.
8. COSTS
It is estimated that the direct cost of producing a half-state (six-
frame) mosaic of approximately 5250 lines by 6100 elements would be on
the order of $8,000, of which approximately $3,000 are computer costs
at University user rates. This estimate excludes the cost of the data.
While the cost of producing the mosaic is significant, the advantages
of having current Landsat data in geographically-corrected form and
registered to past data are also significant, particularly in reducing
the cost of subsequent processing.
9. ANTICIPATED USES
The immediate application of the layered mosaic is for state-wide
annual assessments of defoliation of forests (Williams et al. , in manu
script.) We anticipate that the data will be of interest and value to
a wide variety of land management and monitoring agents throughout the
state, as well. Possible applications include:
1. Monitoring of forest change. Two-thirds of Pennsylvania is
covered in forest and much of the timber is approaching commercial ma
turity. Together with increased mortality due to insect attacks, in
creased mineral, coal and oil exploitation, and increased competition
from other land uses, large scale changes in Pennsylvania forests are
occurring and can be monitored by Landsat. We have research under way
to determine optimum change-detection procedures using the data base
as the mid-date in three-date analyses.
2. Soil mapping. The value of Landsat data for improving existing
soil maps in Pennsylvania is under investigation. Digitized soil maps
can be easily overlaid on the data base and comparisons made without
further rectification.
3. Updating existing land use data bases. Techniques have been de
veloped at ORSER for interfacing the Landsat data base (or derived data)
with existing geographic information systems (GIS's). This involves
the user defining a grid or polygon pattern (such as the grid-cell pat
tern of an existing GIS). Classified Landsat data can then be extract
ed through this pattern and area statistics summarized by polygons.
Since most existing land-use data bases are at the same map projection
as the Landsat data base, further expensive geometric correction should
be unnecessary.
4. Adding existing digitized topographic data, road networks, state
forest boundaries, etc. It is known that several types of digitized
data, either in raster form (e.g., digital terrain data) or in line or
polygon form (e.g., roads, jurisdictional boundaries), are currently
available. Many are already stored at the PSU Computation Center.
Most of these are in projections conformant with the data base and
could be added as data layers, if desirable.
5. Constructing special purpose land cover maps. The cost of pro
ducing land cover maps for small geographic areas such as watersheds,