Study of Sealed Surfaces and Diurnal Temperature Variations
in Urban Environments with Airborne Scanner Data
Ramon Franck, Boris Prinz, Rafael Wiemker, Hartwig Spitzer
Il. Institut fir Experimentalphysik, CENSIS
Universitat Hamburg
Mail address: KOGS, Vogt-Kolln-Str. 30, 22527 Hamburg, Germany
WWNW: http://kogs-www.informatik.uni-hamburg.de/projects/Censis:html
(franck, prinz,wiemker) Gkogs.informatik.uni-hamburg.de
Commission VII, Working Group 3
KEY WORDS: Diurnal Temperature Variations, Unsupervised Classification, Thermal Infrared Imaging
ABSTRACT
The composition of surfaces in urban areas has a strong influence on urban climate and therefore on the quality of life. In
general sealed surfaces will cause a stronger daytime heating in comparison to vegetated areas. Using remotely sensed data it
is possible to monitor the extent of surface sealing and its effect on surface temperature.
Furthermore, the determination of thermal behaviour with regard to solar heating can be used to discriminate between different
surface types. In this study we investigate the usefulness of remotely sensed temperatures of a part of the diurnal brightness
temperature curve for unsupervised classification purposes in an urban environment, in particular for sealed and unsealed
surfaces. Thermal image data were taken with the thermal IR channel of the Daedalus AADS 1268 scanner at four different
times (4:30, 8:30, 11:30, 12:30 MET). We propose a straightforward method to classify a scene into vegetated and non
vegetated regions based solely on thermal data. No spectral bands in the frequency range of reflected sunlight were used,
accepting worse classification accuracies for the benefit of a more accurate interpretation of the results.
1 INTRODUCTION
Thermal infrared measurements of vegetation surface temper-
atures are widely used to calculate heatfluxes (Gash 1987)
and to determine surface conditions and heating properties
(e.g. urban heat island, (Kim 1992)). On the other hand
thermal data has been used for classification purposes in
only very few investigations (e.g. (Leak & Venugopal 1990),
(Lambin & Strahler 1994)), requiring a deeper understanding
(Price 1981). This is due to the fact that the surface tempera-
ture is not a material property like albedo but results from the
balance between incoming and outgoing energy fluxes. Since
many passive remote sensing systems acting in the frequency
range of reflected sunlight have also the ability to measure
thermal infrared radiation, a useful source of information is
wasted by not using this data.
In order to calculate material properties from thermal
measurements we have to examine the thermal response of
a substance with regard to time-varying incoming power.
This response which is mainly determined by the thermal
inertia of the substance. So temperature measurements
at different times of a day provide an appropriate basis
for classification purposes. To illustrate this idea we have
calculated part of the diurnal temperature curves of two
substances having different values of thermal inertia (Figure
1). The underlying model is described in (Price 1977). Note
the distinguishability of the two curves, particularly based
upon different morning slopes and noon temperatures.
Temperoture [^C]
Time (MET)
Figure 1: Modeled surface temperature curves of two surface
types urder solar irradiation (daily mean temperatures can-
not be calculated with this model). All parameters but ther-
mal inertia P are arbitrary chosen and equal for both sur-
face types (dashed line: P= 0.1 cal/cm? s! /? ?C, solid line:
P= 0.05 cal/cm? s!/?C).
228 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
If s
so
vis
of 1
(H
this
veg
dat
the
we
tio!
Th
12€
anc
pla
par
flig
res
dat
ten
Atr
SA
atn
me
brig
itie
cla:
not
dat
wa:
age
iste
(W
cau
anc
per
the
In |
con
3.1
To
we
rep
era
Gra
des
res
ore
gra
