Abstract:
This research
specifically
aims
to
investigate
the
most accurate
spectral
indices
in
extracting
wetlands
geospatial
information
taking
South
Kalimantan, Indonesia,
as
an
example
of wetlands
in
tropical
areas.
Ten
spectral
indices
were
selected for
testing
their
ability
to extract
wetlands,
those
are
NDVI, NDWI,
MNDWI,
MNDWIs2,
NDMI, WRI, NDPI,
TCWT,
AWEInsh,
andAWEIsh.
Tests
were
performed
on
Landsat
8
OLI
path/row
117/062
and
117/063.
The threshold method which
was
used
to
separate
the wetland
features from the
spectral
indices
imagery
is
Otsu
method. The results of this research showed
that
generally
MNDWIs2
was
the
most
optimal spectral
indices
in
wetlands
extraction.
Especially
tropical
wetlands
that
rich with
green
vegetation
cover.
However,
MNDWIs2
is
very
sensitive
to
dense
vegetation,
this feature has
the
potential
to
be detected
as
wetlands.
Furthermore,
to
improve
the
accuracy
and
prevent
detection of the
dryland
vegetation
as
wetlands,
the threshold value should be determined
carefully.
