10 Remote Sensing of Surface Water
223
Land use data classified from a Landsat TM scene of the Priim basin, Germany were
used with a digital elevation model in a GIS to identify more than 9000 potential
reservoir locations and to determine the characteristics of selected sites.
10.4 Wetlands
While the contrast between water and vegetation makes the delineation of lake areas
relatively easy, the similarity in reflectance between vegetation and the surrounding
ground (Fig. 10.1) complicates the measurement of vegetation-laden swamps and
wetlands. Remote sensing techniques have increasingly been used for wetland assessment and since the 1980's, remotely sensed data from satellites has been considered the most important tool for the identification and monitoring of wetlands (Wang
et aI., 1998). Klemas et aI.( 1993) note that the advantages of using satellite imagery
for mapping wetlands over conventional aerial photography include timeliness and
reduced costs. Jensen et aI., (1986) showed that Landsat MSS data was useful in
mapping the location and area of wetlands, and that Landsat MSS was well suited for
regional assessment of inland wetlands.
For larger areas, Mason et aI. (1992) have shown that in near-infrared (channel 2)
data from an A VHRR image of East Africa the clear-water Lakes Turkana and Albert
are distinct but the 10-50,000 km 2 Sudd swamp is barely distinguishable. However,
using channel 4 (thermal infrared) data, the Sudd shows very clearly because of the
higher thermal inertia of the water/vegetation compared to the land. Using daily
sequences of 30-minute Meteosat images, Mason et aI. (1992) found that the greatest
contrast between water, vegetation and land occurred during early afternoon. A simple
thresholding technique was used on the brightness histograms of early afternoon
images for the year 1988 to show inter-annual variations in Sudd area. A fundamental
limitation to the technique was found during the rainy season (May to October) when
the increased atmospheric humidity and the wet ground reduced contrast between the
swamp and its surroundings. Improvements to the thermal imaging were made by
using AVHRR channe13 (3.7/l1Il) with its higher resolution and higher contrast.
The location and measurement of water hyacinth in Lake Kyoga, Uganda has been
addressed by classifying SPOT images into 6 categories of concentration (RCSSMRS,
1995). The contrast between water, vegetation and land in the infrared has also been
used to monitor the development of lake deltas. Haack (1996) describes the use of the
Landsat MSS near-infrared (0.8-1.1llm) channel with images from 1973 to 1989 to
study the increasing size of the Omo River delta in Lake Turkana, Kenya, using the
fact that the newly-formed delta was covered with active vegetation.
Radar sensors can also be utilized in wetland identification and have been used
successfully for mapping. However, space-borne SAR platforms are limited to single
frequency and single polarization, limiting their usefulness for vegetation discrimination. None the less, JERS-l L-band radar imagery has been applied to mangrove
vegetation mapping (Aschbacher et aI., 1995). Adam et aI., 1998 successfully used
Radarsat C-HH data to map open and flooded vegetation in a large wetland region.
In their study, they identified floodwater distribution by segmenting the images into
223
Land use data classified from a Landsat TM scene of the Priim basin, Germany were
used with a digital elevation model in a GIS to identify more than 9000 potential
reservoir locations and to determine the characteristics of selected sites.
10.4 Wetlands
While the contrast between water and vegetation makes the delineation of lake areas
relatively easy, the similarity in reflectance between vegetation and the surrounding
ground (Fig. 10.1) complicates the measurement of vegetation-laden swamps and
wetlands. Remote sensing techniques have increasingly been used for wetland assessment and since the 1980's, remotely sensed data from satellites has been considered the most important tool for the identification and monitoring of wetlands (Wang
et aI., 1998). Klemas et aI.( 1993) note that the advantages of using satellite imagery
for mapping wetlands over conventional aerial photography include timeliness and
reduced costs. Jensen et aI., (1986) showed that Landsat MSS data was useful in
mapping the location and area of wetlands, and that Landsat MSS was well suited for
regional assessment of inland wetlands.
For larger areas, Mason et aI. (1992) have shown that in near-infrared (channel 2)
data from an A VHRR image of East Africa the clear-water Lakes Turkana and Albert
are distinct but the 10-50,000 km 2 Sudd swamp is barely distinguishable. However,
using channel 4 (thermal infrared) data, the Sudd shows very clearly because of the
higher thermal inertia of the water/vegetation compared to the land. Using daily
sequences of 30-minute Meteosat images, Mason et aI. (1992) found that the greatest
contrast between water, vegetation and land occurred during early afternoon. A simple
thresholding technique was used on the brightness histograms of early afternoon
images for the year 1988 to show inter-annual variations in Sudd area. A fundamental
limitation to the technique was found during the rainy season (May to October) when
the increased atmospheric humidity and the wet ground reduced contrast between the
swamp and its surroundings. Improvements to the thermal imaging were made by
using AVHRR channe13 (3.7/l1Il) with its higher resolution and higher contrast.
The location and measurement of water hyacinth in Lake Kyoga, Uganda has been
addressed by classifying SPOT images into 6 categories of concentration (RCSSMRS,
1995). The contrast between water, vegetation and land in the infrared has also been
used to monitor the development of lake deltas. Haack (1996) describes the use of the
Landsat MSS near-infrared (0.8-1.1llm) channel with images from 1973 to 1989 to
study the increasing size of the Omo River delta in Lake Turkana, Kenya, using the
fact that the newly-formed delta was covered with active vegetation.
Radar sensors can also be utilized in wetland identification and have been used
successfully for mapping. However, space-borne SAR platforms are limited to single
frequency and single polarization, limiting their usefulness for vegetation discrimination. None the less, JERS-l L-band radar imagery has been applied to mangrove
vegetation mapping (Aschbacher et aI., 1995). Adam et aI., 1998 successfully used
Radarsat C-HH data to map open and flooded vegetation in a large wetland region.
In their study, they identified floodwater distribution by segmenting the images into