222
G. Kite and A. Pietroniro
satellite and lake altitudes, lake latitudes and locations of the centre-lake pixels.
Errors in measured areas were determined to be in the order of 3-4% when compared
to ground measurements.
Time series of lake areas can provide a record of climatic change; in particular, the
areas of closed lakes (without river outflows) display considerable variation. The
areas of lakes greater than 100 km 2 may easily be measured from Landsat MSS or TM
to an accuracy of 1 % (Harris, 1994); however, there is a need for frequent and regular
coverage to obtain sufficient cloud-free passes. The cost of such numbers of Landsat
images is high and the use of NOAA A VHRR is preferable provided that sufficient
accuracy can be maintained. Harris (1994) describes the development and use of a
technique, which uses sub-pixel edge detection, based on local iso-luminance contours (LIC). Edge detection is done within a moving circle of pixels. For each neighborhood, the brightest and darkest pixel intensities are determined and, if the difference exceeds a specified threshold, an edge with iso-luminance equal to one half the
sum of the brightest and darkest pixel intensities is plotted. The resulting disjointed
edge plot is transformed to a continuous boundary using a line-joining algorithm.
Using the LIC technique on a degraded Landsat image showed comparable accuracy
to using simpler techniques on the full image. In order to use the LIC technique for
more areas, Birkett and Mason (1995) constructed a global database of large
(> 1 00km2) lakes (1,403) and evaluated the availability of satellite passes for each
lake.
Remotely sensed data may also be used to discover lakes. In 1974-75 an airborne
radio-echo sounding survey indicated the presence of a large previously-unknown
lake beneath about 4 km of ice in central East Antarctica. Analysis of changes in
surface slope derived from waveform products of ERS-l radar altimeter data confinned the presence of this lake and allowed estimations of the lake area (10,000 km 2 )
and mean depth (125 m); comparable in size to Lake Ontario (Kapitsa et aI., 1996).
Previously-known relationships between water depth and some other measurable
parameter can be used to convert remotely sensed lake areas into lake volumes (Higer
and Anderson, 1985). In a study of the Florida Everglades, Higer and Anderson
(1985) used channels 5 and 7 to classify a Landsat MSS scene into water and vegetation. A two-band false-colour technique was used with ground measurements of the
variation of vegetation density with depth to classify the MSS image into 10 waterdepth classes. The volume of water in the area was determined by summing the
products of class area and depth.
The properties of natural microwave emissions from the earth can be used to distinguish between open water and lake ice. As a development of earlier work using the
DMSP SSMII passive microwave sensor to measure snow water equivalent, Walker
has used SSM/I 85 Ghz brightness temperatures to discriminate between areas of ice
cover and open water on large lakes in Canada (Great Slave, Great Bear, Great Lakes)
and to monitor spatial and temporal patterns of ice freeze-up and decay. Low brightness temperatures over known lake areas indicate the presence of open water (Walker,
1997).
Quite apart from the measurement of existing lakes and reservoirs, remotely sensed
data may be used to select future reservoir locations (Schumann and Geyer, 1997).
G. Kite and A. Pietroniro
satellite and lake altitudes, lake latitudes and locations of the centre-lake pixels.
Errors in measured areas were determined to be in the order of 3-4% when compared
to ground measurements.
Time series of lake areas can provide a record of climatic change; in particular, the
areas of closed lakes (without river outflows) display considerable variation. The
areas of lakes greater than 100 km 2 may easily be measured from Landsat MSS or TM
to an accuracy of 1 % (Harris, 1994); however, there is a need for frequent and regular
coverage to obtain sufficient cloud-free passes. The cost of such numbers of Landsat
images is high and the use of NOAA A VHRR is preferable provided that sufficient
accuracy can be maintained. Harris (1994) describes the development and use of a
technique, which uses sub-pixel edge detection, based on local iso-luminance contours (LIC). Edge detection is done within a moving circle of pixels. For each neighborhood, the brightest and darkest pixel intensities are determined and, if the difference exceeds a specified threshold, an edge with iso-luminance equal to one half the
sum of the brightest and darkest pixel intensities is plotted. The resulting disjointed
edge plot is transformed to a continuous boundary using a line-joining algorithm.
Using the LIC technique on a degraded Landsat image showed comparable accuracy
to using simpler techniques on the full image. In order to use the LIC technique for
more areas, Birkett and Mason (1995) constructed a global database of large
(> 1 00km2) lakes (1,403) and evaluated the availability of satellite passes for each
lake.
Remotely sensed data may also be used to discover lakes. In 1974-75 an airborne
radio-echo sounding survey indicated the presence of a large previously-unknown
lake beneath about 4 km of ice in central East Antarctica. Analysis of changes in
surface slope derived from waveform products of ERS-l radar altimeter data confinned the presence of this lake and allowed estimations of the lake area (10,000 km 2 )
and mean depth (125 m); comparable in size to Lake Ontario (Kapitsa et aI., 1996).
Previously-known relationships between water depth and some other measurable
parameter can be used to convert remotely sensed lake areas into lake volumes (Higer
and Anderson, 1985). In a study of the Florida Everglades, Higer and Anderson
(1985) used channels 5 and 7 to classify a Landsat MSS scene into water and vegetation. A two-band false-colour technique was used with ground measurements of the
variation of vegetation density with depth to classify the MSS image into 10 waterdepth classes. The volume of water in the area was determined by summing the
products of class area and depth.
The properties of natural microwave emissions from the earth can be used to distinguish between open water and lake ice. As a development of earlier work using the
DMSP SSMII passive microwave sensor to measure snow water equivalent, Walker
has used SSM/I 85 Ghz brightness temperatures to discriminate between areas of ice
cover and open water on large lakes in Canada (Great Slave, Great Bear, Great Lakes)
and to monitor spatial and temporal patterns of ice freeze-up and decay. Low brightness temperatures over known lake areas indicate the presence of open water (Walker,
1997).
Quite apart from the measurement of existing lakes and reservoirs, remotely sensed
data may be used to select future reservoir locations (Schumann and Geyer, 1997).