1
MAPPING THE SPECTRAL AND SPATIAL CHARACTERISTICSOF MOUND SPRING WETLAND VEGETATION: A NOVELINTEGRATED HYPERSPECTRAL APPROACH
Davina White
and
Megan Lewis
The University of Adelaide,School of Earth and Environmental Sciences,Waite Campus, South Australia 5064Phone: +61 0(8) 8303 8112; Fax: +61 (0)8 8303 6717davina.white@adelaide.edu.au megan.lewis@adelaide.edu.au
Abstract
The Great Artesian Basin mound springs are unique wetland ecosystemssupporting a diverse range of rare and relic endemic flora and fauna. They areof great ecological, scientific and economic importance and are culturallysignificant for indigenous Australians, providing a vital source of water in
Australia’s arid interior. In recent decades the ecological sustainability of the
springs has become uncertain as demands (mining and pastoral waterallocations) for this precious water resource increase. Our research within the
Allocating Water and Maintaining Springs in the Great Artesian Basin
programis developing methods using hyperspectral remote sensing for mapping andmonitoring the sensitivity of spring vegetation to water allocations and land use.Hyperspectral remote sensing provides the spectral detail necessary todiscriminate wetland vegetation, showing great potential in wetlands previousstudies. This paper presents an overview of our approaches and results usinghyperspectral analysis of mound spring wetland vegetation.HyMap airborne hyperspectral imagery was acquired in March 2009, coincidingwith a comprehensive field campaign of spectroradiometry measurements andbotanical survey. We are developing and evaluating a number of approaches tomap the extent, distribution and diversity of mound spring wetland vegetation,including Spectrally Segmented Principal Component Analysis (SSPCA),narrow band indices and Spectral Angle Mapper (SAM). SSPCA was applied toNDVI masked vegetation portions of the HyMap imagery with wavelengthregions spectrally segmented for VIS-NIR, 450-1,350
nm
, SWIR1, 1,400-1,800
nm
, and SWIR2, 1,950-2,480
nm
. VIS-NIR PCs 2, 3 and 9 identified keyvegetation discrimination wavelength features, i.e., green peak, chlorophyllabsorption, VIS-NIR and red-edge contrasts. SWIR1 PC 3 mapped
Phragmites
successfully, coinciding with
Phragmites
in survey field plots verified with 30 cmdigital aerial photography. SAM successfully detected
Phoenix dactylifera
(datepalms), an invasive species, within the wetland vegetation. Narrow band NDVI,green:red ratio and the moisture index are currently being investigated todetermine their capability of delineating mound spring wetland extent.
2
These results demonstrate that hyperspectral data and techniques are capableof revealing considerable spectral variation within spring wetland vegetation,providing a good basis for discriminating species and communities.
Introduction
The Great Artesian Basin (GAB) mound springs (Figure 1) support a unique anddiverse range of rare and relic endemic flora and fauna. They also provide avital source of water in the arid inland heart of Australia (Mudd, 2000; Badman,1996; Boyd, 1990). In recent decades the ecological sustainability of the springshas become uncertain as demands on the GAB for this precious artesian waterresource increase. The impacts of existing water extractions for mining andpastoral activities along with their land use impacts are unknown. This situationis further compounded by the likelihood of increasing demand for extractions inthe future.
Figure 1: Location of Australian GAB spring groups (graphic courtesy of Travis Gotch,South Australia Arid Lands Natural Resource Management Board) and a false colourcomposite (blue: 467
nm
; green: 559
nm
; red: 890
nm
) mosaicked HyMap image of theDalhousie spring group study site.
Previous mapping and monitoring of wetland vegetation associated withselected mound springs in the south-western portion of the GAB has relied onvisual interpretation of aerial photography and considerable field work. Thisapproach is particularly limiting for discriminating wetland plant communities,perennial from ephemeral wetland plants and from dryland vegetation.
3
This paper presents an overview of the research we are conducting within the
Allocating Water and Maintaining Springs in the Great Artesian Basin
program,developing and evaluating hyperspectral approaches for mapping andmonitoring the sensitivity of mound spring vegetation to mining and pastoralwater allocations and land use. Hyperspectral remote sensing provides thespectral detail necessary to discriminate wetland vegetation and has showngreat potential in previous studies for mapping wetlands vegetation (Torbick andBecker, 2009; Hestir et al., 2008; Zomer et al., 2009). Our research isdeveloping several hyperspectral approaches for mapping the extent,distribution and diversity of mound spring wetland vegetation includingSpectrally Segmented Principal Component Analysis (SSPCA), narrow bandindices and Spectral Angle Mapper (SAM).
Methods
Data collection and pre-processing
HyMap airborne hyperspectral imagery was acquired in March 2009, comprising126 wavebands with a bandwidth of ~15
nm
and wavelength range of 450 -2,500
nm
(Kruse et al., 2009; Cocks et al., 1998), a 3 m spatial resolution, andswath width of 1.5 km. The raw HyMap imagery was radiometrically corrected(converted to apparent surface reflectance) along with geometric correction andcolour balancing of swaths to form a seamless mosaic. Digital aerialphotography at 30 cm ground resolution was acquired concurrently with theHyMap imagery, primarily for image analysis validation purposes.The analyses reported in this paper were performed on mosaicked HyMapimagery covering approximately 171 km
2
over the Dalhousie Springs grouplocated in northern South Australia on the Northern Territory border (Figure 1).The HyMap image acquisition coincided with a comprehensive field campaignof field spectral measurements and a botanical survey. Sample plots of 9 × 9 mwere recorded, which included vegetation cover and composition based on amodified version of the Braun-Blanquet relevé method as recommended byTiner (1999). Five target spectra representative of the vegetation within each ofthe 9 x 9 m sample areas were acquired using an analytical spectral devices(ASD) FieldSpec® Pro full range (FR) spectroradiometer, together withreference panel measurements. Vertical digital photographs were also acquiredfor each field spectroradiometer target, capturing the top of canopyinstantaneous field of view (TOCIFOV) of the spectroradiometer to quantifyvegetation cover.
Search History:
Result 00 of 00
00 results for result for
p.
More From This User
Notes
Load more
