228
G. Kite and A. Pietroruro
sensed data. At a simple level, river flows at specific return periods may be related to
geomorphological characteristics. For example, Inglis (1947) suggested that the 100year flood is a linear function of the river's meander length; if the meander length can
be measured from a satellite image then the flood flow can be estimated. Engman and
Gurney (1991) recommend the use of the visible red band (MSS band 5, TM band 3)
for discerning stream channel networks. Solomon (1993) recommends the use of
ERS-l synthetic aperture radar (SAR) for delineating river networks in densely
forested areas, taking advantage of the SAR's all-weather day/night capability and its
ability to detect water beneath forest canopies.
Similarly, the flow in a wide shallow river may be estimated from the channel width
measured by satellite. Leopold and Maddock (1953) suggested that width, W, is
related to discharge, Q, as:
W=aQb
where a and b are constants. Smith et al. (1995) used the C-band (5.3 Ghz) ERS-l
SAR to measure the width of a heavily-braided section of the Iskut River in British
Columbia twenty-eight times in the period April 1992 to December 1993. Each image
was radiometrically calibrated and fitted to the Goddard Earth model ellipsoid and an
"effective width", We, of the river was computed as a simple count of those pixels
classified as water within a 10km x 3km control area divided by the length of the river
section. A regression equation of the form We=aQb was then developed between the
"effective width" and the corresponding river discharges, Q, measured by Water
Survey of Canada at each overpass. The processed SAR pixel resolution ofERS-l is
12.5m and the "effective width" of the Iskut River varied between 100 and 1100 m
during the measured period. The corresponding discharges varied between 240 and
6350 m 3 /s. Smith et al. (1996) extended this study to a total of 41 SAR images of the
Iskut, Taku and Tanana Rivers with similar types of relationship and suggest that
discharges of ungauged rivers (with strongly braided channels) can be estimated
within a factor of 2.
If a river is wide enough and has a stable stage-discharge relationship, then the
discharge may also be estimated from measurements of river level using a radar
altimeter. For example, Cudlip et al. (1992) used Seasat data to produce a river
surface elevation profile for the Amazon River from 32 altimeter crossings of the river
over a 17 day period in July 1978. The accuracy of the river elevations is estimated
at +/- 10 em to +/- 20cm. If a rating curve exists for a location along this profile, then
the river discharge could be computed. Koblinsky et al. (1993) concluded, however,
that Geosat radar does not provide sufficient accuracy or coverage to estimate river
levels in the Amazon Basin and recommended that TOPEXIPoseidon could be used
with better results.
For runoff over shorter time periods it is necessary to develop more complex models. For example, to simulate monthly runoff in a basin in southwestern France,
Strubing and Schultz (1985) used NOAA A VHRR infrared images to estimate
different densities of cloud cover and to develop a "mean daily temperature-weighted
cloud cover index". The mean monthly runoff for a particular month was then a linear
function of the sum of the daily product of the cloud index and a system response
G. Kite and A. Pietroruro
sensed data. At a simple level, river flows at specific return periods may be related to
geomorphological characteristics. For example, Inglis (1947) suggested that the 100year flood is a linear function of the river's meander length; if the meander length can
be measured from a satellite image then the flood flow can be estimated. Engman and
Gurney (1991) recommend the use of the visible red band (MSS band 5, TM band 3)
for discerning stream channel networks. Solomon (1993) recommends the use of
ERS-l synthetic aperture radar (SAR) for delineating river networks in densely
forested areas, taking advantage of the SAR's all-weather day/night capability and its
ability to detect water beneath forest canopies.
Similarly, the flow in a wide shallow river may be estimated from the channel width
measured by satellite. Leopold and Maddock (1953) suggested that width, W, is
related to discharge, Q, as:
W=aQb
where a and b are constants. Smith et al. (1995) used the C-band (5.3 Ghz) ERS-l
SAR to measure the width of a heavily-braided section of the Iskut River in British
Columbia twenty-eight times in the period April 1992 to December 1993. Each image
was radiometrically calibrated and fitted to the Goddard Earth model ellipsoid and an
"effective width", We, of the river was computed as a simple count of those pixels
classified as water within a 10km x 3km control area divided by the length of the river
section. A regression equation of the form We=aQb was then developed between the
"effective width" and the corresponding river discharges, Q, measured by Water
Survey of Canada at each overpass. The processed SAR pixel resolution ofERS-l is
12.5m and the "effective width" of the Iskut River varied between 100 and 1100 m
during the measured period. The corresponding discharges varied between 240 and
6350 m 3 /s. Smith et al. (1996) extended this study to a total of 41 SAR images of the
Iskut, Taku and Tanana Rivers with similar types of relationship and suggest that
discharges of ungauged rivers (with strongly braided channels) can be estimated
within a factor of 2.
If a river is wide enough and has a stable stage-discharge relationship, then the
discharge may also be estimated from measurements of river level using a radar
altimeter. For example, Cudlip et al. (1992) used Seasat data to produce a river
surface elevation profile for the Amazon River from 32 altimeter crossings of the river
over a 17 day period in July 1978. The accuracy of the river elevations is estimated
at +/- 10 em to +/- 20cm. If a rating curve exists for a location along this profile, then
the river discharge could be computed. Koblinsky et al. (1993) concluded, however,
that Geosat radar does not provide sufficient accuracy or coverage to estimate river
levels in the Amazon Basin and recommended that TOPEXIPoseidon could be used
with better results.
For runoff over shorter time periods it is necessary to develop more complex models. For example, to simulate monthly runoff in a basin in southwestern France,
Strubing and Schultz (1985) used NOAA A VHRR infrared images to estimate
different densities of cloud cover and to develop a "mean daily temperature-weighted
cloud cover index". The mean monthly runoff for a particular month was then a linear
function of the sum of the daily product of the cloud index and a system response