83 
AN EVALUATION OF LANDSAT DATA FOR INPUT TO A 
STATE GEOGRAPHIC INFORMATION SYSTEM 
Elizabeth M. Middleton 
Eastern Regional Remote Sensing Applications Center 
NASA/Goddard Space Flight Center 
Greenbelt, MD 20771 
Belden G. Bly III John M. Garber 
Computer Sciences Corporation Department of State Planning 
Silver Spring, MD 20910 301 West Preston Street 
Baltimore, MD 21201 
ABSTRACT 
A cooperative study was undertaken by the Eastern Regional Remote Sen 
sing Applications Center and the Maryland Department of State Planning 
(DSP) to determine the correspondence of land use data in a special 
Maryland Automated Geographic Information (MAGI) System 4.6 acre grid 
cell data base with Landsat categorized land cover data geocorrected to 
1.54 acres/pixel. The DSP 21-category land use scheme was reorganized 
into eight categories: forest, crop/pasture, water, transitional, 
commercial/industrial/institutional, medium density residential, low 
density residential, and wetlands. The MAGI System land use data, 
encoded in separate data layers for primary and secondary use, were 
tabulated several ways to represent a range in interactive and compu 
tational strategies: (a) primary land use interpreted as either 0 or 
100 percent per cell (common strategy by users); (b) primary land use, 
adjusted to recorded proportions between 0 and 100 in increments of 20 
percent; (c) the primary and secondary land use together, adjusted to 
the same recorded proportions as above (P' + S'); and (d) the (P' + S') 
for cells with specific MAGI land use labels only. In general, the 
results indicated that with procedure (c), a 90 percent or better cor 
respondence of the MAGI and Landsat data could be achieved for 4 of 7 
categories. A high correspondence was achieved for 6 of 7 categories 
and a near acceptable correspondence for the seventh category with pro 
cedure (d). An additional finding was that the Landsat results 
degraded to the MAGI grid cell size were considerably less satisfactory 
than the full resolution data, with decreases in correspondence of up 
to one order of magnitude for some cover types. These results have 
been translated into strategies for users of the MAGI System. 
INTRODUCTION 
Land cover information interpreted from Landsat data have become an 
important source for updating state land use surveys, and for integra 
tion with other types of data in georeferenced data bases (Boyd, Jones, 
and Dasgupta, 1981). Recognizing this potential, the Governor in 1977 
designated the Maryland Department of State Planning (DSP) the agency 
responsible for coordinating remote sensing technology within the state 
institutional framework. In order to evaluate Landsat digital data, 
categorized by land cover, for input to the Maryland Automated Geo 
graphic Information (MAGI) System, the state georeferenced data base, 
a study was undertaken jointly by DSP and the Eastern Regional Remote 
Sensing Applications Center (ERRSAC) as part of the NASA technology 
transfer program to state and local governments. The primary objec 
tive of this study was to compare and evaluate the overall accuracy of 
a land cover inventory derived from Landsat data with DSP's detailed