B Fo DD DO N uw DD TS
ty
ue
| 0
n © ® +
LANDSAT-TM DATA FOR MUNICIPAL ENVIRONMENTAL PLANNING ?
STUDIES OF VEGETATION INDICES IN THE URBAN AREA
Matthias Achen
Assistent Professor, Department of Geography
University of Heidelberg, Germany
ISPRS-Commission VII
ABSTRACT
The aim of this study is to determine the accuracy with
which the percentage of vegetation cover (PVC) in urban
areas, can be recorded by digitally processing Landsat-TM
data. The area analyzed consists of the settled area of Hei-
delberg, Germany; the reference data were yielded from the
analysis of aerial photography. Various vegetation indices
(NDVI, PVI, SAVI, PC2) were determined using geocoded
TM data. The correlation with the PVC referent (n— 12,585)
was calculated for every index on the basis of 50-m grid
squares on a scala calibrated from 0 to 100; they are found
on the level of r=0.92. Comparative advantages are demon-
strated by PVI in the lower PVC range and by NDVI in the
upper PVC range. SAVI and NDVI display relatively
similar results for the area analyzed; masking the settled area
prior to the principal component transformation is favorable
for PC2. Residues are largely caused by the dependency of
all indices on the composition of surface materials and on
shadow formation and by radiometric and geometric errors
in the TM data. Moreover, the vegetation cover partially
changed between the time of the satellite and the aerial
pictures; this is why the results for all indices were in reality
(even) more favorable. There is even a correlation in the
range of r=0.97/0.98 between the indices and PVC for
selected test areas made on the basis of 99 (building)
quadrangles. The results of the study show that even on the
municipal level, data from satellite photographs could
already be used today for environmental planning.
Keywords: Landsat-TM, Image Processing, Vegetation Indi-
ces, Percentage of Vegetation Cover, Urban Area, Munici-
pal Environmental Planning, Heidelberg (Germany)
1. INTRODUCTION
This article reviews the most important results of an empiri-
cal investigation of the accuracy with which the proportion
of the total surface area covered by vegetation, i.e., the per-
centage of vegetation cover (PVC), can measured by digi-
tally processing Landsat-TM data with the help of vegetation
indices (ACHEN 1992). The settled area of Heidelberg is
analyzed, an area displaying the basic types of urban con-
struction and land use. Visual analysis of color infrared ae-
rial photographs provided the reference data.
The state of technical data and information on the environ-
mental situation in urban areas requires decisive impro-
vement; this forms the general backdrop to the study. At
least in the Federal Republic of Germany, official statistics
in this sector do not yet include relevant, spatially related
data. However, since such data is indispensable to the incre-
asingly important political concept of "ecologically oriented
urban development”, a suitable procedure for measuring so-
called environmental indicators must be found. Procedures
are evaluated in terms of expense, time required, relevance,
accuracy, and degree of differentiation of data, as well as
the resolution and precision of geometric information
(RADERMACHER 1989, 19).
Basically, there is no procedure that optimally satisfies these
requirements. In data collected on the ground, differentiated
201
indicators (such as the amount of vegetation) can be very ac-
curately measured; however, the cost in time and money
preclude the extensive use of this procedure. Visual evalua-
tion of (color infrared) aerial photographs (depending upon
their scale) must be judged as less accurate and differentia-
ted, and its costs in time and money are only somewhat lo-
wer; this is presently the procedure most frequently used for
collecting environmental data at the municipal level. Though
the digital processing of satellite photograph data must be
evaluated as at best satisfactory in resolution, precision, and
the differentiatedness of data, its expenditures of time and
money are considerably lower. Detailed data have yet to be
published on the accuracy of such studies. Generally, one
must seek an "ends-means-oriented procedure" (HUBLER
1986, 465) for data collection. The aforementioned criteria
have to be weighed against each other in a procedure where,
given limited government financial resources, the
importance of "expenditure of time and money" as a cri-
terion is not to be underestimated.
The urban area cannot yet be counted among the objects of
research for which digital satellite remote sensing has be-
come a significant source of information. This may be ex-
plained in part by the incongruence between land use and
land cover and the high percentage of mixed pixels provided
by currently available sensory systems. Thus, the author is
only aware of two publications on vegetation indices in ur-
ban areas: the study by KERL (1989) using Landsat-TM
data on Munich, in which, however, the statistical evidence
is incomplete; and the study by FORSTER (1983/1985a)
using Landsat-MSS data on Sydney, in which the results are
only suitable in part for operationalization in environmental
planning.
This study thus aims to study the quality of various vegeta-
tion indices in measuring the percentage of vegetation cover
(PVC) as an parameter relevant to urban ecology
(KENNEWEG 1975) and to carry out this study on two le-
vels: for 50-m grid squares covering the entire settled area
of Heidelberg (31.5 km^) and for (building) jjuadrangles
from selected test areas in Heidelberg (1.16 km^). The em-
pirical studies have been supported by the German Institute
for Aeronautical and Space Research (Deutsche Forschungs-
anstalt für Luft- und Raumfahrt; Oberpfaffenhofen) and the
city of Heidelberg.
2. AREA INVESTIGATED
The area analyzed consists of the settled area of Heidelberg,
defined here on the basis of the land use plan. It is primarily
made up of residential, commercial, mixed, and public buil-
ding areas. Furthermore, areas categorized as special buil-
ding areas, areas of public utilities and waste management
plants, public green spaces, allotments, cemeteries, traffic
areas, and water areas were considered part of the settled
area if they were largely enclosed by areas from the first set
of categories. The same holds for small areas of agricultural
use. On the other hand, smaller-sized settled areas located a
clear distance from other settled areas were excluded. In this
way, the settled area of Heidelberg corresponds roughly to a
"restricted area" (geschlossene Ortschaft) and represents an
area of about 31.5 km^^