wnloaded from
d calibrated to
ered relatively
idation points.
ations running
'S. ACTOSS each
ion cover and
on cover was
ssistance of a
which a secchi
3 points across
then compared
Land Monitor
nial vegetation
Ihe inundation
nce to create an
isets were then
e 263 wetland
culated. When
face water in a
/etlands which
classified as
than 12 times
es, 2008). The
l.
for separating
) values were
surface water
vith increasing
appear in the
sites recorded
e surface water
Yt surface water
onds to point 2
s covered with
surface water, much of this water was covered by a stand of
emergent halophytic vegetation. The halophytic vegetation
covered 4096 of the site and much of the foliage was desiccated.
This caused an increase in reflectance of Band 5.
Figure 3: Inundated site with dense coverage of emergent
halophytic vegetation.
From visual interpretation of the regression in Figure 2 a
threshold value of 28 was derived. To further examine the
relationship between surface water cover and Band 5 values,
Band 5 values were extracted along transect lines which
intersected field sites 1, 2 and 3 (Figure 4). Photographs from
the 3 field sites, located in Lake Logue, are shown in Figure 5.
| 1
100 -
| |
| 80 -
| |
jo. 604
|
$7]
8 | 2
| 40 - Not inundated 3
| 20 inundated Mt ter tees tto
| o eo r- © © ~~ e © D ©
<r oo e nr ER «e e < co
| -— = a Q e e e
| Distance (meters) |
; Figure 4: Band 5 values extracted along a transect line which
includes 3 sample points with the derived threshold value in red.
Figure 5: Photographs from sites 1, 2 and 3 (left to right) at
Lake Logue.
The Lake Logue sites 1, 2 and 3 recorded average water depths
of 5, 10 and 25 mm, percent water coverage of 70, 60 and 100
and percentage vegetation cover of 50, 20 and 10 respectively.
International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, Volume XXXIX-B8, 2012
XXII ISPRS Congress, 25 August — 01 September 2012, Melbourne, Australia
A dramatic drop in Band 5 values was observed between sites 1
and 2. While depth and percent water cover are marginally
different between the 2 sites vegetation cover drops from 50 to
20 %.
The inundation threshold was applied to the validation image
and checked against the 16 field sites. The overall accuracy of
the classification was 87.5%. All errors were sites that were
predicted to be “not inundated” but were found to be
“inundated”. Field sites with greater than 50 % surface water
coverage were classes as inundated.
3.2. Hydroperiod classification
The criteria for a wetland to be classified as “Permanently
inundated" is that it must be more than 10 % inundated for 70 %
of the time (or 12 out of the 17 years assessed). Of the 263
wetlands assessed in this study only 2 met this criterion. The
remaining wetlands were classified as “Seasonally inundated”.
The total area inundated across all wetlands was also assessed
(Figure 6). Area inundated appears fairly consistent across the
time period. The year 2000 is anomalous. This can be explained
by the heavy winter rainfall in 1999 which helped push the total
rainfall, at the nearby Jurien Bay weather station, for this year to
970.8 mm. The mean annual rainfall is 536.3 mm (Bureau of
Meteorology, 2011).
Area (ha) Inundated
2000
1500 4
Area (ha)
S
e
e
500 4
0 :
0 ONU «v DO dw O Q O' x i «o r o o © +
OHO DIO Q OO O O oO OOo ooo SE =
9oQ B OO 0 0 0 00 OoO0O 0 oO o oco oO o
- = = + + + ON ON A A As Nas as a AU N
Year
Figure 6: Area of inundation per year.
4. DISCUSSION
Landsat band 5 was successfully used to separate not inundated
from inundated plots. Vegetation cover was found to be the
major confounding factor. At 2 plots, high vegetation cover and
high surface water cover were recorded. In both cases band 5
digital counts were significantly higher than sights with low
vegetation cover. Furthermore, these sites could not be
separated from vegetated sites that were not inundated.
The reflectance of vegetation in band 5 is influenced by
moisture content and thickness of the leaf (Lillesand & Kiefer,
1994). Vegetation will absorb more light in Band 5 than bare
soil and will reflect more than water. The percentage of
vegetation cover at which band 5 reflectance increases is
difficult to discern from this study, but it is likely to be heavily
dependent on the type of vegetation. Both 40 96 coverage of
woody halophytic vegetation and 70 96 coverage of lush grassy
vegetation were sufficient to dramatically increase band 5