reefs (e.g. the Bahamas, Belize, Lesser Antilles, etc.). In
more general terms, the Caribbean has a higher propor-
tion of deep reef to reef ¯at than the Indo-Paci®c (Done
1983), and the transferability of results to such areas is
unclear. However, because there is less light attenuation
at shallower depths, all forms of optical remote sensing
would logically be favoured in the Indo-Paci®c.
Turbid waters are possibly the greatest constraint to
any coastal habitat-mapping programme utilising opti-
cal remote-sensing methods. Where coral reefs occur in
waters of high suspended sediment concentration (e.g.
Hong Kong), light transmittance through water is in-
adequate for describing the coverage of reef habitats.
Even where sucient light penetration exists, compen-
sation for the eects of variable depth becomes more
complex as turbidity increases. Lyzenga's depth-invari-
ant model requires clear water, and high water-turbidity
exerts a major (though currently not quanti®ed) eect
on the applicability of the model (but see Spitzer and
Dirks 1987). Horizontal Secchi distance at a depth of
0.5 m in the Turks and Caicos Islands was of the order
of 30 to 50 m and PAR (photosynthetically active ra-
diation) attenuation coecients were 0.108 and 0.065 for
the upper 5.5 m and lower 4.5 m of the top 10 m of the
water column, respectively (AJE and CDC unpublished
results). At present it is not possible to give a threshold
turbidity at which satellite imagery will be ineective. In
the absence of such information, we point out that Tassan
(1996) has described a depth-invariant model for water of
greater turbidity than that required by Lyzenga (1981).
Conclusions
Whilst some of the results presented here were expected
(e.g. the superiority of CASI over satellite imagery), at
least three of the conclusions were surprising. First,
given the disparity in spatial resolution, we did not ex-
pect satellite imagery to compare so favourably against
aerial photography. Second, although CASI imagery
was expected to map benthic habitats with greater ac-
curacy than satellites (and to some extent, aerial pho-
tography), it was surprising that CASI was capable of
mapping a comprehensive range of benthic habitats with
such high accuracy (72 to 93%). Third, merging the
spectral qualities of Landsat TM with the spatial reso-
lution of SPOT Pan was expected to provide an opti-
mum satellite-based approach to habitat mapping. This
was not the case, and may re¯ect the diculty of
merging satellite data sources and the overall limitation
of satellite imagery for benthic habitat-mapping. It is
hoped that the conclusions presented above will help
practitioners match their coastal habitat-mapping ob-
jectives to the most appropriate sensor(s).
Acknowledgements This research was funded by the U.K. Overseas
Development Administration's Environment Research Pro-
gramme. Seagrass studies were also funded by NERC Grant No.
GR9/02233 to C. Clark and T. Malthus. We are very grateful to the
Ministry of Natural Resources of the Turks and Caicos Islands for
reefs (e.g. the Bahamas, Belize, Lesser Antilles, etc.). In
more general terms, the Caribbean has a higher propor-
tion of deep reef to reef ¯at than the Indo-Paci®c (Done
1983), and the transferability of results to such areas is
unclear. However, because there is less light attenuation
at shallower depths, all forms of optical remote sensing
would logically be favoured in the Indo-Paci®c.
Turbid waters are possibly the greatest constraint to
any coastal habitat-mapping programme utilising opti-
cal remote-sensing methods. Where coral reefs occur in
waters of high suspended sediment concentration (e.g.
Hong Kong), light transmittance through water is in-
adequate for describing the coverage of reef habitats.
Even where sucient light penetration exists, compen-
sation for the eects of variable depth becomes more
complex as turbidity increases. Lyzenga's depth-invari-
ant model requires clear water, and high water-turbidity
exerts a major (though currently not quanti®ed) eect
on the applicability of the model (but see Spitzer and
Dirks 1987). Horizontal Secchi distance at a depth of
0.5 m in the Turks and Caicos Islands was of the order
of 30 to 50 m and PAR (photosynthetically active ra-
diation) attenuation coecients were 0.108 and 0.065 for
the upper 5.5 m and lower 4.5 m of the top 10 m of the
water column, respectively (AJE and CDC unpublished
results). At present it is not possible to give a threshold
turbidity at which satellite imagery will be ineective. In
the absence of such information, we point out that Tassan
(1996) has described a depth-invariant model for water of
greater turbidity than that required by Lyzenga (1981).
Conclusions
Whilst some of the results presented here were expected
(e.g. the superiority of CASI over satellite imagery), at
least three of the conclusions were surprising. First,
given the disparity in spatial resolution, we did not ex-
pect satellite imagery to compare so favourably against
aerial photography. Second, although CASI imagery
was expected to map benthic habitats with greater ac-
curacy than satellites (and to some extent, aerial pho-
tography), it was surprising that CASI was capable of
mapping a comprehensive range of benthic habitats with
such high accuracy (72 to 93%). Third, merging the
spectral qualities of Landsat TM with the spatial reso-
lution of SPOT Pan was expected to provide an opti-
mum satellite-based approach to habitat mapping. This
was not the case, and may re¯ect the diculty of
merging satellite data sources and the overall limitation
of satellite imagery for benthic habitat-mapping. It is
hoped that the conclusions presented above will help
practitioners match their coastal habitat-mapping ob-
jectives to the most appropriate sensor(s).
Acknowledgements This research was funded by the U.K. Overseas
Development Administration's Environment Research Pro-
gramme. Seagrass studies were also funded by NERC Grant No.
GR9/02233 to C. Clark and T. Malthus. We are very grateful to the
Ministry of Natural Resources of the Turks and Caicos Islands for
การแปล กรุณารอสักครู่..
