The detection of water bodies is relatively a simple task
compared to other classification processes. However, problems
arise also in this field, like the identification of borderlines of
land and water, the detection of water under vegetation, and
difficulties of analysis caused by waves on the water surface
when active microwave images are used. To overcome these
problems different approaches had been developed with varying
results. The problem of detecting stagnate water under
vegetation is seemingly the one causing the highest uncertainty,
and remains unsolved.
Beside the problem of the high vegetation, below which remote
sensing techniques can hardly identify water bodies, clouds are
another natural phenomena limiting the use of optical sensors. If
the cloud cover is significant above the area to be analyzed,
infrared and microwave sensors can be used. Passive
microwave sensors usually have poor spatial resolution,
therefore their use in flood monitoring is almost non-existing.
The high-resolution synthetic aperture radar (SAR) sensors on
the platforms of ERS-2, JERS-1 and RADARSAT can penetrate
clouds, darkness, even light rainfall from, therefore are a
promising tool for flood monitoring.
The term multisensor remote sensing is applied here for
applications where images acquired by different sensors are
used for the same analysis. Multisensor remote sensing is often
combined with multitemporal technique, where a time series of
images (taken by the same sensor) is analyzed.
SATELLITES AND SENSORS FOR FLOOD
MONITORING
A number of different kinds of satellites have been used for
monitoring inundated areas recently. Four main characteristics
of satellite imagery that highly influence the application of
satellite-based remote sensing for flood monitoring are: (1)
spatial resolution (or pixel size) of the image, (2) return period
of the satellite, (3) availability of images and (4) costs of the
data. The sensors used in flood monitoring can be divided into
three large groups: (a) high-resolution optical systems, like
Landsat or SPOT, (b) meteorological satellites, like NOAA
AVHRR, and (c) active microwave sensors, like the SAR
sensors onboard ERS-1/ERS-2, JERS-1 or RADARSAT.
Further characteristics of importance are wavelength, view
angle and polarization (microwave systems) of sensors. High-
resolution optical systems have several channels in the visible
bands of the electromagnetic spectrum, meteorological satellites
have different visible and infrared bands and SAR sensors
usually have single wavelength. Almost all satellites have fixed
view angle, with the operational exception of RADARSAT.
Ideal solution for flood monitoring would be a series of images
with high spatial resolution, frequent (at least daily) overpasses
covering the whole flooded area with multiband observations
(including visible, near-infrared, infrared and active microwave
channels) at daytime and nadir with continuous, (near) real-time
reception for low price. At present, and very likely either in the
near future, none of the operational satellites provide such data
sets, giving scientist plenty of think about what images to use
for different applications.
Spatial resolution. High-resolution optical systems like
onboard the Landsat and SPOT satellites have high spatial
resolution (Table 1) and therefore are widely used for various
mapping and land-use classification processes. In case of flood
monitoring, a resolution of a few meters to a few tens of meters
(high-resolution) would be needed for damage assessment, for a
precise modeling of the dynamics of the event, and for other
small scale purposes. On the other hand, only low resolution
images with a pixel size of a few hundred meters to several
kilometers have the ability to map and monitor areas frequently
enough to get an overview of the dynamics of the flooding. In
the assessment of inundated areas, the restriction is usually the
thick cloud cover over the flooded areas, through which the
optical sensors are not able to capture useful data of the flood
extent. Meteorological and geostationary satellites are less
frequently used for monitoring flood events, mainly because of
their sparse spatial resolution. The resolution of the images
taken by the AVHRR sensors is 1.1 km at nadir, which reduces
the possible applications of these types of images to regional
and global scale monitoring only. Geostationary meteorological
satellites have even lower ground resolution, and therefore their
application to flood monitoring is limited. The ground
resolution of SAR images is as high as of the first category, and
often even better: images with 10-m resolution can be acquired
by RADARSAT.
Landsat NOAA ERS-2/
Parameter SPOT AVHRR JERS-1 RADARSAT
Ground
resolution 30-80 1,100 30/18 10-100
(m)
Return
period 8-16 0.5 35/44 24
(day)
Frequency VI, IR VI, IR M M
Look angle Fix Fix Fix Var.
Quick data Not Avail. Avail. Not Avail. Avail.
ordering
x aAA
EA of one 3,500 Jo |
a scene / 100 os 2,700-4,500
(ECU, 3.100 y
approx.) jg 900
VI: visible, IR: infrared, M: microwave, Avail.: available
Table 1: Availability of images from different satcllites/sensor
systems
Return period of satellites. Landsat, SPOT and other Earth-
observing satellites have the unfavorable property of a relatively
long return period of more than one week. Considering the
frequent thick cloud cover over the flooded areas, even if the
satellite has an overpass in the required time frame, useful
information may not be obtained due to the clouds. A big
portion of flooding occur in a short time period, often in a few
days, therefore monitoring a single flood event exclusively by
these high-resolution images is limited, and is often impossible.
At present, the third generation of the TIROS experimental
satellites are in operation called NOAA-12 and NOAA-14,
designed for meteorological observations. Orbiting the Earth on
a sun-synchronous orbit, images can be acquired four times a
day over the same location using both satellites. Thus, the
temporal frequency is seemingly adequate for hydrological
applications. The NOAA satellites are able to take images twice
a day, usually one at daytime and one at nighttime. Nighttime
images have less information content, but even daytime images
can lack of information of the land surface if the cloud cover is
significant. Therefore, another critical point of flood monitoring
if optical sensors are used is to get cloud-free or only partly
cloudy images over the area to be analyzed. The return period of
satellites carrying SAR sensors is 35 days (mission-dependent)
for ERS-2, 44 days for JERS-1, and 24 days for RADARSAT.
In the case of the latter, it is possible to acquire images from the
same location even in 3-6 days intervals, using the unique
property of changing the view angle of the sensor on orbit, but
780 Intemational Archives of Photogrammetry and Remote Sensing. Vol. XXXII, Part 7, Budapest, 1998
A 3 ("To oc QU e
--— "^ ^35 A —… 6 f^ Ay " As "+
4 "= (rj
mnc m. 3 m. A
"-. bum
a
