20 Remote Sensing of African Lakes: A Review
407
spatial resolution; visible and NIR) as well as from IKONOS (in the 4 m spatial resolution version; visible and NIR) and showed that despite the relatively low spatial
resolution, MERIS was able to detect changes in lake extent (kappa coefficients, as
per Cohen 1960, ranging from 0.92 to 0.95).
Other lakes have also drawn attention. Booth and Oldfield (1988) used Landsat
MSS NIR Band 7 to estimate declining Lake Kariba area in five images from 1981
to 1985 coincident with a drought in the region. Classification was done by manually
judging an appropriate water cut off DN value in each image.
Ouma and Tateishi (2006) used visible and NIR bands from Landsat TM and
ETM+ images to develop a new Water Index (WI) based on Tasseled Cap Wetness
(TCW) and Normalized Difference Water Index (NDWI) to rapidly map shorelines
in five Rift Valley lakes in Kenya (Baringo, Bogoria, Elementaita, Naivasha and
Nakuru) for images in 1986 and 2001. Ground-based observations showed that the
WI had an accuracy of 98.4 % outperforming TCW by ∼ 22 % and NWDI by ∼ 43 %.
Harris (1994) used the NIR band of AVHRR (Channel 2) to detect shoreline
changes at Lake Abiyata (Ethiopia) with a local isoluminance contour (LIC) method.
The technique also made use of a re-sampled NIR band from Landsat MSS (band 7).
Also for Lake Abiyata, Legesse and Ayenew (2006) used Landsat MSS, TM, ETM+
and SPOT data, as well as data from the Moderate Resolution Imaging Spectrometer
(MODIS)—which exists in two versions, carried onboard the NASA orbital platforms
Terra (MODIS Terra) and Aqua (MODIS Aqua)—to visually display changes in the
lake area from 1978 to 2005. The band used is not specified but the very dark
appearance of the lake relative to bright land in the figures indicates it was an NIR
band.
Finally, Böhme et al. (2006) examined intra- and inter-annual areal changes
in Lake Urema (Mozambique) from 1979 to 2000 by a Maximum Likelihood
Classification augmented with NDVI, Principal Component Analysis (PCA) and
Tasseled-Cap-Transformation (TCT) based on visible and NIR bands from MSS,
TM, ETM+, and ASTER.
20.2.2 Thermal Sensing Applications
The spectral differences between lake water and surrounding land in the visible and
especially NIR bands are the basis for shoreline delineation discussed in the previous
section. It is also possible to use the thermal part of the spectrum to do similar work.
The main benefit of using thermal imagery is the ability of the technique to work at
night, due to the fact that the electromagnetic radiation being received by the sensor
is emitted from the target and not reflected from a source (sun).
Schneider et al. (1985), in addition to the work on visible and NIR bands discussed
above, used two of the AVHRR thermal channels (4 and 5; wavelengths 10.3–11.3
and 11.5–12.5 μm, respectively) to differentiate Lake Chad’s open water from flood
plains, deltas and surrounding land.
407
spatial resolution; visible and NIR) as well as from IKONOS (in the 4 m spatial resolution version; visible and NIR) and showed that despite the relatively low spatial
resolution, MERIS was able to detect changes in lake extent (kappa coefficients, as
per Cohen 1960, ranging from 0.92 to 0.95).
Other lakes have also drawn attention. Booth and Oldfield (1988) used Landsat
MSS NIR Band 7 to estimate declining Lake Kariba area in five images from 1981
to 1985 coincident with a drought in the region. Classification was done by manually
judging an appropriate water cut off DN value in each image.
Ouma and Tateishi (2006) used visible and NIR bands from Landsat TM and
ETM+ images to develop a new Water Index (WI) based on Tasseled Cap Wetness
(TCW) and Normalized Difference Water Index (NDWI) to rapidly map shorelines
in five Rift Valley lakes in Kenya (Baringo, Bogoria, Elementaita, Naivasha and
Nakuru) for images in 1986 and 2001. Ground-based observations showed that the
WI had an accuracy of 98.4 % outperforming TCW by ∼ 22 % and NWDI by ∼ 43 %.
Harris (1994) used the NIR band of AVHRR (Channel 2) to detect shoreline
changes at Lake Abiyata (Ethiopia) with a local isoluminance contour (LIC) method.
The technique also made use of a re-sampled NIR band from Landsat MSS (band 7).
Also for Lake Abiyata, Legesse and Ayenew (2006) used Landsat MSS, TM, ETM+
and SPOT data, as well as data from the Moderate Resolution Imaging Spectrometer
(MODIS)—which exists in two versions, carried onboard the NASA orbital platforms
Terra (MODIS Terra) and Aqua (MODIS Aqua)—to visually display changes in the
lake area from 1978 to 2005. The band used is not specified but the very dark
appearance of the lake relative to bright land in the figures indicates it was an NIR
band.
Finally, Böhme et al. (2006) examined intra- and inter-annual areal changes
in Lake Urema (Mozambique) from 1979 to 2000 by a Maximum Likelihood
Classification augmented with NDVI, Principal Component Analysis (PCA) and
Tasseled-Cap-Transformation (TCT) based on visible and NIR bands from MSS,
TM, ETM+, and ASTER.
20.2.2 Thermal Sensing Applications
The spectral differences between lake water and surrounding land in the visible and
especially NIR bands are the basis for shoreline delineation discussed in the previous
section. It is also possible to use the thermal part of the spectrum to do similar work.
The main benefit of using thermal imagery is the ability of the technique to work at
night, due to the fact that the electromagnetic radiation being received by the sensor
is emitted from the target and not reflected from a source (sun).
Schneider et al. (1985), in addition to the work on visible and NIR bands discussed
above, used two of the AVHRR thermal channels (4 and 5; wavelengths 10.3–11.3
and 11.5–12.5 μm, respectively) to differentiate Lake Chad’s open water from flood
plains, deltas and surrounding land.