MONITORING HIGH MOUNTAIN SNOWCOVER USING DATA FUSION
TECHNIQUES
Harold HAEFNER and Jens PIESBERGEN
Remote Sensing Laboratories, Department of Geography
University of Zürich, 8057 Zürich, Switzerland
e-mail address: {haefner, piesi}@geo.unizh.ch
Fax No. +41 1 635 68 42
ISPRS Commission VII Working Group 4
Snow Cover Monitoring, Data Fusion Techniques, ERS SAR, Landsat TM, Radarsat
ScanSar
KEY WORDS:
ABSTRACT
A continuous snowcover monitoring in time and space represents the basis for all snowhydrologi-'
cal applications, and is an important factor for studying climatic change as well. For applica-
tions in high mountain regions, the unstable weather conditions ask for the use of SAR satellite
data. But SAR imaging properties require specific methods to reduce or even eliminate the relief
induced distortions to ensure a continuous areal snowcover mapping. In addition, characteristic
problems have to be solved regarding classification methods and accuracy when working on differ-
ent scales. Three test sites in the Swiss Alps, reaching from very large to very small scale,
have been investigated. Studies based solely on the use of single frequency, single polarization
synthetic aperture radar (SAR) data demonstrate that wetness maps can be compiled, and a moni-
toring of the snow melting process is possible over a long-term period. Analyzing advantages and
disadvantages of the use of EO and SAR sensors, it can be concluded that EO data are more suit-
able for snowcover mapping purposes. On the other hand, SAR systems are absolutely necessary to
retrieve certain snow parameters and to continuously monitor the melting process. The goal of
our project is to take advantage of the synergism of EO and SAR systems using data fusion tech-
niques, and to monitor the melting process over large mountainous areas. The completion of the
first objective consists of a rigorous calibration of the remote sensing data in respect to the
geometry and the radiometry, using digital elevation models and sensor orbital data. While in
optical imagery, illumination and atmospheric corrections are additionally considered, a fully
thematic information of SAR data can only be achieved applying the optimal resolution approach
(ORA). In a further step, EO and SAR data are fused pixel by pixel, using a colour transforma-
tion, in order to reduce misclassifications. First, a simple fusion model is applied, and then
the basic model is extended, to take into account the temporal attributes between data sources
acquired at different dates based on the multitemporal optimal resolution approach (MORA). The
method decreases the misclassification significantly. Results of snowcover mapping and monitor-
ing from the medium size test area GRISONS are presented to illustrate the methodological prin-
ciples.
a combination with SAR data is the only possi-
bility to achieve a continuous monitoring dur-
1. INTRODUCTION
The seasonal snowcover has a significant im-
pact on geoecological processes and human ac-
tivities. These effects get of special impor-
tance in high mountain terrain. But snow is
not only a valuable resource, it can also turn
into a menacing natural hazard. Furthermore,
its changes are an important parameter for
studying climatic change.
Hence, a continuous snowcover monitoring in
time and space forms the basis for all snowhy-
drological applications. But in high mountain
regions, the unstable weather conditions ham-
per the use of optical EO data. Consequently,
ing the accumulation and melting process of
the snowcover in small individual basins or
over large mountain areas. Another possibility
would be to relay on radar sensors only. But
snow mapping based solely on a single fre-
quency, single polarization, C-band SAR system
(such as ERS) is not suitable for operational
purposes. SAR imaging properties require spe-
cific methods to eliminate or at least reduce
the relief induced distortions. Multifrequency
polarimetric radar data (SIR-C/X-SAR) allow a
much better classification of the areal extent
of the snowcover, even of dry snow, as well as
an assessment of the water equivalent. It is
350 International Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
adva
and
syst
diti
solv
over
To .:
clas
accu
ate
data
site
1).
aski
larg
regi
els
tion
tics
avai
Tabl
Snow
or £
cuss
Haef
Stei
Afte
rect
1997
sis
red,
blue
type
Snow
Refl
(cha
Snow
rela
soil
