Cheryl A. Brantley, Manager Production Services 
Rob Robinson, Senior Applications Manager 
Michael Bradley, Image Processing Specialist 
Andrew S. Bury, Applications Analyst 
ERDAS, Inc. 
2801 Buford Highway, Suite 300 
Atlanta, Georgia USA 30329 
(404)248-9000 
(404)248-9400 FAX 
ISPRS Commission IV 
USING ANCILLARY DIGITAL DATA TO IMPROVE LAND COVER CLASSIFICATION 
ABSTRACT 
The initial field work tasks associated with land cover classification efforts are usually expensive and time- 
consuming. By using available digital data in conjunction with the imagery to be classified, labor intensive ground 
truthing can be reduced or even eliminated. Techniques have been developed to integrate data bases such as USGS DLG 
(Digital Line Graph), Census TIGER, and USGS LUDA (Land Use/Land Cover) with imagery to target the selection of 
training sample sites. In this paper, several case studies will be used to illustrate these techniques and introduce advanced 
image processing and geographic information systems (GIS) software functions which increase classification accuracy 
rates while decreasing the overall time needed to conduct major land cover analyses. 
INTRODUCTION 
Today, GIS users are feeling the pressure of demands for 
more accurate and up-to-date data bases to support a growing 
variety of applications. For years the primary input source 
for GIS has been paper topographic and thematic maps (soils, 
land use, etc.). However, as the industry grows and software 
applications become more advanced, there is an increasing 
need for more specialized information. As GIS plays a 
greater role in planning and decision making, there is an 
emphasis on data bases to display actual current conditions. 
GIS coverages can no longer afford to be 10, five, or even 
one year out of date. Yet the cost of manually updating a 
large data base on an annual basis, makes it difficult, if not 
impossible, for many federal, state and private sectors to meet 
these demands. Cutbacks in personnel and spending, 
coupled with legislation mandating that correct geographic 
information be maintained, are placing many agencies in a 
dilemma. Factors such as soils makeup, wetlands, 
endangered species, crops and forests must be monitored. 
Therefore, many agencies, both public and private, are 
turning to the use of satellite data as a reliable source to 
calculate current conditions of vegetative and anthropogenic 
information. 
The use of satellite mapping for land use/land cover is most 
effective when compiled for a large study area. Over the last 
several years, ERDAS, Inc., has participated in several large 
scale projects which have ranged in size from four to 12 
scenes of satellite data. In our experience with projects of 
this size, we have found that the use of ancillary data is 
essential for breaking the data into logical analytical 
components. The use of this ancillary information increases 
the speed of the process time and helps refine the accuracy of 
the data. Inexpensive and widely available data sources make 
it possible to greatly increase the speed, ease and accuracy 
with which a satellite land use/land cover can be completed. 
This paper will focus on ancillary data as a tool to be used in 
the successful creation of large scale land use/land cover 
classifications. It will also examine the characteristics of data 
used in ERDAS Production Services projects, and discuss the 
positive uses and difficulties associated with each data type. 
104 
ANCILLARY DATA AS A TOOL IN REMOTE 
SENSING ANALYSIS 
Ancillary data can be described as any supplemental data that 
might be used to enhance or become an additional part or 
section of the primary data base. Ancillary data can range 
from aerial photography to out-dated GIS coverages or maps. 
Usually several different types of ancillary data may be used 
to examine satellite data. Below are examples of readily 
available ancillary data: 
» USGS Digital Line Graph Data (DLG) 
* United States Bureau of Census TIGER Data 
* USGS Topographic maps 
* USGS Digital Elevation Model (DEM) 
* 1:250,000 Land Use and Land Cover Data (LUDA) 
* United States Department of Agriculture - ASCS 
35mm color compliance photography 
* National High Altitude Photography (NHAP-2) 
* National Aerial Photography Program (NAPP) 
The data listed above is available through federal agencies 
and, with the exception of aerial photography, can be 
purchased at a relatively low cost. Also, as state and local 
agencies begin to understand the need to coordinate activities, 
more spatial data bases will become available to the private 
and public sectors. The cost to compile and create computer 
based GIS is forcing states to begin to encourage or mandate 
the cooperation and sharing of spatial data base information. 
This, in effect, will set up a framework where more private 
and public sector businesses can move into the GIS user 
community. Therefore, it is important that the GIS user be 
cognizant of digital data bases available in their particular area 
of interest. In many cases, such data information can be 
used to reduce the amount of work and associated cost 
required to complete a project. 
The u: 
and ac 
the de 
helpfu 
data is 
how t 
systen 
one fi 
impler 
in one 
envirc 
marke 
update 
Softw: 
raster 
innove 
such a 
linked 
vector 
softw: 
vector 
transp 
displa 
satelli 
intercl 
data f 
update 
condit 
PR( 
The v 
direct] 
the sc 
Regar 
to ana 
data w 
image 
inform 
In wo 
metho 
base d 
the pre 
be cre: 
Once 
establ 
potent 
shoulc 
much 
the sy 
much 
the da 
compl 
Will it 
issues 
neede« 
The f 
comp] 
studie 
circun 
ancille 
study 
comm 
ability 
speed: 
proce: 
classi 
neede 
ground