226
G. Kite and A. Pietroniro
also be used to identify ground features since the height and shape of the waveform
are highly dependent on the surface roughness.
Birkett and Kite (1997) used data from the NRA and SSALT radar altimeters onboard the joint NASAICNES TOPEXIPoseidon satellite to measure levels of Lake
Athabasca, Great Slave lake and Great Bear Lake in the Mackenzie Basin for September 1992 to December 1995. Results showed that the altimeters agreed well with
the recorded lake levels except for periods when the lakes were ice covered. At such
times, the altimeters record a dip in altimetric "lake surface" during January to April
and then a recovery to normal in May, suggesting that the altimeter is recording the
effect of an increasing ice and snow thickness during the winter decreasing to zero
during melt. The raw altimeter waveform data were extracted for pass 095 across
Lake Athabasca for 1994/95. At the end of April the echoes are broad-peaked but
during May, when the ice begins to melt, the waveforms change shape, becoming
multi-peaked, and reverting to normal broad-peaked during July, when the lake is
completely ice-free.
Lake Victoria in East Africa has an interesting history of major changes in lake
level (Kite, 1982) and it is important that accurate measurements of the lake level be
obtained. Figure 10.5 plots water level vs. time for a position in Lake Victoria about
100km west of Kisumu for a series of TOPEXIPoseidon satellite passes from June
1992 to December 1993. The data show a plausible pattern of lake level changes but
no surface-based measurements are currently available for comparison.
Stuttard et al. (1994) conclude that at the moment, the application of radar altimetry
for routine lake level monitoring of a specific lake is not appropriate on the grounds
of cost, complexity and accuracy. The role of radar altimetry is seen as monitoring a
large series of closed lakes worldwide for long-term climatic change studies.
So far, this section has described methods of estimating lake levels directly from
remotely sensed data. However, there is also a less direct method; many hydrological
models use remotely sensed data as inputs, along with ground-based data, to simulate
river flows and lake levels. Such models are described in more detail in Chap. 5 of
this publication as well as in other publications such as Schultz (1994), Kite and
Pietroniro (1996) and Schultz (1996). As an example, the SLURP model (Kite, 1995)
has been applied to simulate the inter-annual and seasonal variation in levels of a
prairie wetland near Saskatoon, Saskatchewan, Canada (Su et aI., 1997). This model
uses NOAA A VHRR or Landsat data to divide a basin into sub-areas of different land
cover. For each vegetation type the model then carries out a vertical water balance
and accumulates water from each land class and sub-area. The NDVI vegetation
activity index is used to compute evapotranspiration for each land class. Figure 10.6
shows an example of the simulated and observed levels of the prairie slough.
10.6 River Levels and Flows
While streamflow may be measured in the field or may be computed in a hydrological
model, many watersheds are ungauged and flow estimates from satellite data would
be extremely useful. While streamflow cannot be measured directly from satellite it
may, in some circumstances, be computed from parameters estimated from remotely
G. Kite and A. Pietroniro
also be used to identify ground features since the height and shape of the waveform
are highly dependent on the surface roughness.
Birkett and Kite (1997) used data from the NRA and SSALT radar altimeters onboard the joint NASAICNES TOPEXIPoseidon satellite to measure levels of Lake
Athabasca, Great Slave lake and Great Bear Lake in the Mackenzie Basin for September 1992 to December 1995. Results showed that the altimeters agreed well with
the recorded lake levels except for periods when the lakes were ice covered. At such
times, the altimeters record a dip in altimetric "lake surface" during January to April
and then a recovery to normal in May, suggesting that the altimeter is recording the
effect of an increasing ice and snow thickness during the winter decreasing to zero
during melt. The raw altimeter waveform data were extracted for pass 095 across
Lake Athabasca for 1994/95. At the end of April the echoes are broad-peaked but
during May, when the ice begins to melt, the waveforms change shape, becoming
multi-peaked, and reverting to normal broad-peaked during July, when the lake is
completely ice-free.
Lake Victoria in East Africa has an interesting history of major changes in lake
level (Kite, 1982) and it is important that accurate measurements of the lake level be
obtained. Figure 10.5 plots water level vs. time for a position in Lake Victoria about
100km west of Kisumu for a series of TOPEXIPoseidon satellite passes from June
1992 to December 1993. The data show a plausible pattern of lake level changes but
no surface-based measurements are currently available for comparison.
Stuttard et al. (1994) conclude that at the moment, the application of radar altimetry
for routine lake level monitoring of a specific lake is not appropriate on the grounds
of cost, complexity and accuracy. The role of radar altimetry is seen as monitoring a
large series of closed lakes worldwide for long-term climatic change studies.
So far, this section has described methods of estimating lake levels directly from
remotely sensed data. However, there is also a less direct method; many hydrological
models use remotely sensed data as inputs, along with ground-based data, to simulate
river flows and lake levels. Such models are described in more detail in Chap. 5 of
this publication as well as in other publications such as Schultz (1994), Kite and
Pietroniro (1996) and Schultz (1996). As an example, the SLURP model (Kite, 1995)
has been applied to simulate the inter-annual and seasonal variation in levels of a
prairie wetland near Saskatoon, Saskatchewan, Canada (Su et aI., 1997). This model
uses NOAA A VHRR or Landsat data to divide a basin into sub-areas of different land
cover. For each vegetation type the model then carries out a vertical water balance
and accumulates water from each land class and sub-area. The NDVI vegetation
activity index is used to compute evapotranspiration for each land class. Figure 10.6
shows an example of the simulated and observed levels of the prairie slough.
10.6 River Levels and Flows
While streamflow may be measured in the field or may be computed in a hydrological
model, many watersheds are ungauged and flow estimates from satellite data would
be extremely useful. While streamflow cannot be measured directly from satellite it
may, in some circumstances, be computed from parameters estimated from remotely