2.1. Study Area
COOLIMBA (
\ x 4 ZF
je
LEEMAN, og nc
lí ^
N
Western
Australia
{
S. a
‘
f
GREEN HEAD \
JURIEN say} I Wetlands within
~ as ^] ESSEN the study area
a
>
T .
da
CERVANTES /
7e
0 3 6.9 12 E
m -— ——ÀWaumm Kilometers \ - 4
Figure 1: Cervantes to Coolimba coastal wetlands.
Cervantes and Coolimba in Western Australia. Wetlands
included in the study area are shown in Figure 1. The study area
covers wetlands within approximately 12 km of the coast
between
2.2. Wetland Boundaries
The hydroperiod analysis was restricted to areas within mapped
wetland boundaries. Boundaries were manually digitized from
aerial photography. Areas of water accumulation (derived from
a 10 m Digital Elevation Model), soil maps and four winter
Landsat images were used as reference data sets. À total of 263
individual wetlands were mapped. The accuracy of the wetland
boundaries was then checked in the field and is the subject of an
ongoing study.
2.3. Landsat imagery
Calibrated and co-registered Landsat imagery captured in
summer and supplied by the Land Monitor Project (Furby,
2010) were used in the analysis. The study area was covered by
the Landsat path/row, 113/81. Image dates used in the analysis
were 11/2/88, 13/1/90, 22/2/92, 26/1/94, 1/2/96, 26/3/98,
19/1/00, 9/2/02, 17/1/03, 23/2/04, 9/2/05, 27/1/06, 14/1/07,
2/2/08, 18/12/08, 22/1/10, 10/2/11.
2.4. Threshold development
A single Band 5 threshold value to separate inundated from not
inundated needed to be defined. Field work to determine this
threshold was carried out between 15 and 19 August 2011.
Comparing field results from the summer Land Monitor
imagery was not possible as conditions had changed
considerably since the date of image capture. For this reason a
Landsat image captured 5/8/11 was used to compare field
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
measurements to index values. The scene was downloaded from
the United States Geological Survey (USGS) and calibrated to
Land Monitor values.
A total of 16, 90 m by 90 m plots which covered relatively
homogenous index values were chosen as validation points.
Plots were chosen in easily accessible locations running
perpendicular to estimated water body boundaries. Across each
site the average water depth, percent vegetation cover and
percent inundation were estimated. Vegetation cover was
estimated by visual interpretation with the assistance of a
template (Hnatiuk et al., 2009). The depth to which a secchi
disk was visible and salinity were recorded at 3 points across
the site and averaged. These field measures were then compared
to index values.
2.5. Hydroperiod classification
Perennial vegetation maps produced by the Land Monitor
Project for each year were used to remove perennial vegetation
within wetland boundaries from the analysis. The inundation
threshold was applied to the Land Monitor sequence to create an
inundation map for each year. These raster datasets were then
converted to polygons. The area of each of the 263 wetland
polygons inundated in each year was then calculated. When
wetlands had greater than 10% coverage of surface water in a
given year they were considered inundated. Wetlands which
were inundated 12 or more times were classified as
“permanently inundated”. Those inundated less than 12 times
were classified as “seasonally inundated” (Jones, 2008). The
total area inundated each year was also calculated.
3. RESULTS
3.1. Threshold development
To determine an appropriate Band 5 threshold for separating
inundated from not inundated areas Band 5 values were
regressed against field measures of percentage surface water
coverage. In general Band 5 values decrease with increasing
surface water coverage, however 2 outliers appear in the
regression (points 1 and 2 in Figure 2). These sites recorded
high vegetation cover along with high percentage surface water
coverage.
200
160
t: 120
"0
$ 80 1 25
e O C
Ó O
40 Ó
0 20 40 60 80 100
surface water cover 96
Figure 2: Regression of Band 5 values and percent surface water
cover.
Figure 3 is a photo taken at a site which corresponds to point 2
in Figure 2. While 100 percent of the site was covered with
Int
surface
emergen
covered
This cau
Figu
From v
threshol
relation
Band $
intersec
the 3 fie
100
80
60
Band 5
40
20
Figur
include
Figu
The L:
of 5, 1
and pe