with LULC
pplied
enario, codes for
ds to year 2050
yields generated
port. Cambridge
Climate Hotspot
peace Southeast
., 2006. Creating
Capes: cloud and
Bingner, R.'L.,
n Guidelines for
ed Simulations.
J. R., 2005. Soil
on Version 2005.
ural Experiment
ation - Cagayan
inicating Climate
2G Achievement
. (WEPA) *Plans
lative to Water
nt.” www.wepa-
))
sitivity Analysis.
of Megacities to
al Conference on
la, PHL.
nal Engineering
ngton, DC, USA.
e (EROS) Center
1 (SRTM)-
nd Data Availa
«shed Hydrology
09-1000.
mments of Dr.
ie study and to
ology (DOST)
iis research.
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
HYDROPERIOD CLASSIFICATION OF CERVANTES COOLIMBA COASTAL
WETLANDS USING LANDSAT TIME SERIES IMAGERY
R. J. van Dongen *, G. A. Behn *, M. Coote *, A. Shanahan* and H. Setiawan
* Dept, of Environment and Conservation, Perth, Australia — (ricky.vandongen, graeme.behn, michael.coote, anne.shanahan,
hery.setiawan) @dec.wa.gov.au
Commission VIII, WG VIIU/4
KEY WORDS: Environment, Hydrology, Monitoring, Landsat, Temporal
ABSTRACT:
Geomorphic classification of wetlands relies on information regarding landform and hydroperiod. Additional attributes of wetland
size, vegetation cover and salinity can be collected for lower order classification. Hydroperiod is important in determining features
that characterize the ecological function of the wetland. This study examines how Landsat time series imagery was used to describe
the hydroperiod of wetlands along the coastal zone between Cervantes to Coolimba in Western Australia. Inundation extent maps
were derived using 17, Landsat Band 5 images captured between 1988 and 2011. The number of times a wetland basin was more
than 10 % inundated was then calculated. This inundation frequency dataset was then the basis for hydroperiod classification.
Wetlands were classified as being permanently inundated if they had more than 10% inundation in 12 or more of the 17 images
available (approximately 70%). A Landsat image captured 2 weeks prior to field work was used to determine Band 5 thresholds to
identify areas of inundation and assess classification accuracy. Field measurements of surface water cover, depth, secchi depth,
vegetation cover and salinity were taken across 16 plots located close to inundation boundaries. The accuracy of the inundation maps
was highly dependent on the degree of vegetation cover. Mapping accuracy was heavily influenced by vegetation cover and achieved
an overall accuracy of 87.5%. The resulting hydroperiod datasets provides an accurate record of inundation frequency which can be
used to aid classification of wetlands and also allows changes to inundation frequency over time to be assessed.
1. INTRODUCTION
Landsat series of satellites has an extensive archive, with data
available from 1972, and has been used for mapping and
monitoring long term environmental processes (Kuhnell et al.,
1998; Caccetta et al., 2007). Indices can be derived from a
combination of Landsat spectral bands to provide quantitative
information on land cover (Furby et al. 2008). For direct
comparison of index values over time, to facilitate land cover
monitoring, consistent image processing techniques are
required. In the of the south west of WA, consistently processed
Landsat data, captured at one and two year intervals, is available
from the Land Monitor Project (Furby, 2009).
In this study the Landsat time series was used to assess the
hydroperiod of wetlands along a coastal zone in Western
Australia. The methodology was taken from Jones (2008). In
that study the hydroperiod of wetlands in adjacent catchments
was assessed as part of a wetland evaluation program.
1.1. Hydroperiod
Wetlands are defined by Boulton & Brock (1999) as "any area
of temporarily or permanently waterlogged or inundated land,
natural or artificial, with water that is standing or running,
ranging from fresh to saline, and where inundation by water
influences the biota and ecological processes occurring at any
time". Hydroperiod (the period of water availability) is an
important attribute of wetlands and is used to differentiate
between a lake (permanently inundated basin) and a sumpland
(seasonally inundated basin) (Semeniuk, 1995). It is also is
Important in determining features that characterize the
ecological function of the wetland.
1.2. Classification Band Selection
In remote sensing studies of wetlands the most common indices
used to map and monitor inundation are the tassel cap wetness
and Normalised Difference Water Index (NDWI) (Lacaux et al.,
2007; Weiss and Crabtree, 2011), a Band 4 (near infrared) and 5
(mid infrared) ratio. Near infrared and mid infrared bands are
well suited to locating and delineating water bodies due to
strong absorption of water in this region of the spectrum
(Lillesand, 1994). Water absorbs less in the visible parts of the
spectrum and can be used to map bethic material and estimate
turbidity or chlorophyll concentration (Jensen, 2007). Landsat
Band 5 was used in this study due to its documented strong
absorption by water and for consistency with Jones (2008).
The physical attributes to which moisture indices respond is
often poorly defined. The quantitative influences to indices that
map “wetness”, which is a relatively common term in the
remote sensing field, are rarely described in detail. Brom et al.
(2011) state that the NDWI "expresses surface moisture
conditions". While this definition may be useful when analyzing
relative change over a time sequence the lack of a quantitative
reference makes the assessments difficult to use in ecological
studies. A focus on percent of surface water appears to be a
more robust and well defined approach than mapping
“wetness”. It has been used successfully by Rover (2010) and
Weiss and Crabtree (2011). Surface water maps or inundation
are also more easily understood and applicable in the ecological
domain. Lacaux et al. (2007) used maps of surface water in
Senegal to study the potential transmission of Rift Valley Fever.
2. METHODOLOGY