Brief description
Changes in seagrass coverage in Cockburn Sound from 1967 to 1999 were assessed from aerial photographs using modern mapping methods with the aim of accurately determining the magnitude of change in hectares of seagrasses between 1967 and 1999.Lineage
Maintenance and Update Frequency: notPlanned
Statement: - Mapping of seagrasses -
To determine seagrass distribution, submerged vegetation was mapped from recent and historical aerial photography. Distribution of seagrass assemblages and reef were then determined in 1999 from towed underwater video. We define assemblages of seagrasses as multi-species assemblages dominated, or characterised, by single or multiple species as determined from their relative abundance in video footage.
In Cockburn Sound, the seagrass species Amphibolis antarctica, Amphibolis griffithii, Posidonia australis, Posidonia coriacea, Posidonia sinuosa, Halophila ovalis, Heterozostera tasmanica and Syringodium isoetifolium were components of the seagrass assemblages. We did not map to single species, although a single species assemblage is composed of more than 70% of that species. Results from mapping of aerial photographs and underwater video footage were then combined in a GIS to create coverage maps of seagrass assemblages, reef and unvegetated sand. We describe our maps as coverage maps, rather than maps of seagrass cover to reduce the confusion between the area covered by a single species of seagrass and the area occupied by an assemblage of seagrasses than is dominated by one or a few species. This is not presence-absence mapping, as vegetated assemblages do not exclude all unvegetated habitat.
To determine seagrass distribution, submerged vegetation was mapped from recent and historical aerial photography. Distribution of seagrass assemblages and reef were then determined in 1999 from towed underwater video. We define assemblages of seagrasses as multi-species assemblages dominated, or characterised, by single or multiple species as determined from their relative abundance in video footage.
In Cockburn Sound, the seagrass species Amphibolis antarctica, Amphibolis griffithii, Posidonia australis, Posidonia coriacea, Posidonia sinuosa, Halophila ovalis, Heterozostera tasmanica and Syringodium isoetifolium were components of the seagrass assemblages. We did not map to single species, although a single species assemblage is composed of more than 70% of that species. Results from mapping of aerial photographs and underwater video footage were then combined in a GIS to create coverage maps of seagrass assemblages, reef and unvegetated sand. We describe our maps as coverage maps, rather than maps of seagrass cover to reduce the confusion between the area covered by a single species of seagrass and the area occupied by an assemblage of seagrasses than is dominated by one or a few species. This is not presence-absence mapping, as vegetated assemblages do not exclude all unvegetated habitat.
Statement: - Study area and mapped regions -
Cockburn Sound is a sheltered marine embayment 16 km long × 9 km wide, and consists of a deep central basin (17-22 m deep) surrounded by shallow platforms. The shallow platforms vary in width from 50 m to 3 km and are where seagrass meadows are found (see thumbnail). The seagrass mapping area was delineated as shallow platforms to a depth of 10 m and covered an area of approximately 3667 ha.
Coverage of all submerged vegetation was separately calculated for three regions called Cockburn Sound East, South and West. Cockburn Sound East encompassed most of the eastern bank from Woodmans Point to south of James Point. Cockburn Sound South contained the shallow bank near Rockingham, the southern sand flats and Careening Bay, Garden Island. Cockburn Sound West encompassed the western bank north of Careening Bay and shallow waters north of Garden Island. Benthic features are difficult to resolve from aerial photographs at depths greater than 10 m as seagrasses become sparsely distributed at these depths. Hence, the 10 m isobath and the coastline are used to delineate the mapping boundaries within each of these regions
Cockburn Sound is a sheltered marine embayment 16 km long × 9 km wide, and consists of a deep central basin (17-22 m deep) surrounded by shallow platforms. The shallow platforms vary in width from 50 m to 3 km and are where seagrass meadows are found (see thumbnail). The seagrass mapping area was delineated as shallow platforms to a depth of 10 m and covered an area of approximately 3667 ha.
Coverage of all submerged vegetation was separately calculated for three regions called Cockburn Sound East, South and West. Cockburn Sound East encompassed most of the eastern bank from Woodmans Point to south of James Point. Cockburn Sound South contained the shallow bank near Rockingham, the southern sand flats and Careening Bay, Garden Island. Cockburn Sound West encompassed the western bank north of Careening Bay and shallow waters north of Garden Island. Benthic features are difficult to resolve from aerial photographs at depths greater than 10 m as seagrasses become sparsely distributed at these depths. Hence, the 10 m isobath and the coastline are used to delineate the mapping boundaries within each of these regions
Statement: - Data sources -
In 1999, flights to acquire aerial photographs were purpose-flown in late-February to early-March when conditions were optimal. These conditions included maximum seagrass leaf cover; maximum water clarity; minimum haze from the adjacent industrial zone; minimum turbidity from dredging, river outflow or industrial activities; minimum cloud cover; weak prevailing winds; and incident sun angle at 20-30°. Colour photography with a yellow lens filter was obtained at two altitudes. High altitude imagery was collected at 10,000 m (scale, 1:55,000) for accurate ortho-rectification. The resulting imagery was used as a rectification base for the low-altitude imagery that was obtained at an altitude of 3,800 m.
Mapping of the historical seagrass coverage in Cockburn Sound was conducted using rectified and mosaicked imagery obtained in: 1994 (majority of imagery obtained on 4 and 8 January 1994 and some on 6 January 1995); 1981 (imagery from 13 June 1981); 1972 (imagery from 2 May 1972); and 1967 (imagery from 20 March 1967).
- Image geo-referencing and rectification -
Images were initially rectified (Datum: WGS-84) using ERMapper-6.1 to a resolution of 2.0 m (Earth Resource Mapping, 2000). The images were initially scanned into three colour bands (red, green and blue). The red band of the colour images contained almost no information for marine benthic habitat, as red light was severely attenuated through the water column. The green band was better than the red, though images had low contrast due to atmospheric attenuation. The blue band provided the best contrast between vegetated and unvegetated habitats and therefore was the only band used for mapping. It was converted to 256 greyscales.
Rectified imagery was processed using Geographic Resources Analysis Support System (GRASS) GIS software (Neteler, 1998). GRASS is a raster-based GIS and the 2.0 m cell size and its spatial location (spatial integrity) was strictly maintained in rectified imagery. Also the same spatial integrity was kept consistent among the rectified imagery for the different years mapped (spatial consistency).
- Automated mapping -
A computerised, semi-automated, greyscale segmentation mapping method called Spann-Wilson segmentation (Spann and Wilson, 1985) was employed to map submerged vegetation. Spann-Wilson combines locally adaptive segmentation (local centroid) with pre-processing using multi-level quad-tree smoothing. The segmentation method was implemented using the Xite image-processing software (The University of Oslo, 1999). The size of the moving histogram window and the number of quad-tree smoothing levels were controlled by the operator, although for this exercise were set to 324 m2 and 3-4. Spann-Wilson segmentation was very effective in defining the boundaries of seagrass meadows reefs and unvegetated sand from greyscale images. It reduced the original 256 greyscales from the aerial photographs to four greyscales. The operator then chose a greyscale which most closely coincided with the visually interpreted boundary between vegetated and unvegetated habitat.
- Control rules for mapping -
The vegetated areas were distinguished from the unvegetated areas as they had a distinct photo-tone of medium to dark grey. To enable consistent coverage mapping across the study area, a series of control rules were used (Kendrick et al., 2000). These control rules were as follows:
- Isolated vegetated patches less than 30 m2 in area were not mapped.
- Vegetated patches, that were greater than 30 m2 and less than 100 m2 were mapped as separate patches when the distance between one patch and another was greater than the diameter of the patch.
- Vegetated patches, that were greater than 30 m2 and less than 100 m2 were mapped as a single meadow when the distance between one patch and another was less than the diameter of the patch. Unvegetated areas within the meadow with an area greater than 100 m2 were mapped.
- Vegetated patches greater than 100 m2 were mapped and the edges of these areas were traced accurately. Unvegetated regions within these patches with areas greater than 100 m2 were mapped.
The control rules were applied automatically during mapping with control rules 2?4 processed during the Spann-Wilson segmentation step to a 324 m2 moving window, and control rule 1 applied after the segmentation step to 1 km2 areas using a PERL script.
In 1999, flights to acquire aerial photographs were purpose-flown in late-February to early-March when conditions were optimal. These conditions included maximum seagrass leaf cover; maximum water clarity; minimum haze from the adjacent industrial zone; minimum turbidity from dredging, river outflow or industrial activities; minimum cloud cover; weak prevailing winds; and incident sun angle at 20-30°. Colour photography with a yellow lens filter was obtained at two altitudes. High altitude imagery was collected at 10,000 m (scale, 1:55,000) for accurate ortho-rectification. The resulting imagery was used as a rectification base for the low-altitude imagery that was obtained at an altitude of 3,800 m.
Mapping of the historical seagrass coverage in Cockburn Sound was conducted using rectified and mosaicked imagery obtained in: 1994 (majority of imagery obtained on 4 and 8 January 1994 and some on 6 January 1995); 1981 (imagery from 13 June 1981); 1972 (imagery from 2 May 1972); and 1967 (imagery from 20 March 1967).
- Image geo-referencing and rectification -
Images were initially rectified (Datum: WGS-84) using ERMapper-6.1 to a resolution of 2.0 m (Earth Resource Mapping, 2000). The images were initially scanned into three colour bands (red, green and blue). The red band of the colour images contained almost no information for marine benthic habitat, as red light was severely attenuated through the water column. The green band was better than the red, though images had low contrast due to atmospheric attenuation. The blue band provided the best contrast between vegetated and unvegetated habitats and therefore was the only band used for mapping. It was converted to 256 greyscales.
Rectified imagery was processed using Geographic Resources Analysis Support System (GRASS) GIS software (Neteler, 1998). GRASS is a raster-based GIS and the 2.0 m cell size and its spatial location (spatial integrity) was strictly maintained in rectified imagery. Also the same spatial integrity was kept consistent among the rectified imagery for the different years mapped (spatial consistency).
- Automated mapping -
A computerised, semi-automated, greyscale segmentation mapping method called Spann-Wilson segmentation (Spann and Wilson, 1985) was employed to map submerged vegetation. Spann-Wilson combines locally adaptive segmentation (local centroid) with pre-processing using multi-level quad-tree smoothing. The segmentation method was implemented using the Xite image-processing software (The University of Oslo, 1999). The size of the moving histogram window and the number of quad-tree smoothing levels were controlled by the operator, although for this exercise were set to 324 m2 and 3-4. Spann-Wilson segmentation was very effective in defining the boundaries of seagrass meadows reefs and unvegetated sand from greyscale images. It reduced the original 256 greyscales from the aerial photographs to four greyscales. The operator then chose a greyscale which most closely coincided with the visually interpreted boundary between vegetated and unvegetated habitat.
- Control rules for mapping -
The vegetated areas were distinguished from the unvegetated areas as they had a distinct photo-tone of medium to dark grey. To enable consistent coverage mapping across the study area, a series of control rules were used (Kendrick et al., 2000). These control rules were as follows:
- Isolated vegetated patches less than 30 m2 in area were not mapped.
- Vegetated patches, that were greater than 30 m2 and less than 100 m2 were mapped as separate patches when the distance between one patch and another was greater than the diameter of the patch.
- Vegetated patches, that were greater than 30 m2 and less than 100 m2 were mapped as a single meadow when the distance between one patch and another was less than the diameter of the patch. Unvegetated areas within the meadow with an area greater than 100 m2 were mapped.
- Vegetated patches greater than 100 m2 were mapped and the edges of these areas were traced accurately. Unvegetated regions within these patches with areas greater than 100 m2 were mapped.
The control rules were applied automatically during mapping with control rules 2?4 processed during the Spann-Wilson segmentation step to a 324 m2 moving window, and control rule 1 applied after the segmentation step to 1 km2 areas using a PERL script.
Statement: - Groundtruth surveys -
Detailed ground truth surveys were conducted between late-February and May 1999. These groundtruth surveys were undertaken using a differential GPS combined with diver-operated towed video (manta tow) and a downward-looking surface deployed (drop-down) video. Following the completion of the formal groundtruth surveys, a series of opportunistic dives were also undertaken to establish the assemblage type in specific areas of interest. A total of four manta tow transects (8 km in total) and 114 drop-down video locations were surveyed during the groundtruthing exercise.
The manta tow videos were analysed by pausing the video at 20 s intervals (corresponding with the differential GPS waypoints). At each of these video pauses the percent of the image each seagrass species and habitat type was recorded. For the drop-down videos, percent representation of species of seagrass and habitats was averaged across the total video footage obtained from each drop unless the species composition varied noticeably. If it was variable, the drop was sub-divided to more accurately represent the seagrass coverage.
Seagrass species recognised in the video footage were A. antarctica, A. griffithii, P. australis, P. coriacea, P. sinuosa, H. ovalis, H. tasmanica and S. isoetifolium. Habitats other than seagrasses that were recognised were limestone reef and unvegetated sand.
It was not possible to separate the species of P. sinuosa and Posidonia angustifolia from the video data, as this requires an examination of the rhizome fibres (Cambridge and Kuo, 1979). Groundtruth dives indicated that P. sinuosa was generally more common than P. angustifolia (greater than 90% except in the Cockburn Sound East region where P. sinuosa cover was approximately 70%). In the present mapping exercise, these two species were both mapped as the P. sinuosa assemblage.
Detailed ground truth surveys were conducted between late-February and May 1999. These groundtruth surveys were undertaken using a differential GPS combined with diver-operated towed video (manta tow) and a downward-looking surface deployed (drop-down) video. Following the completion of the formal groundtruth surveys, a series of opportunistic dives were also undertaken to establish the assemblage type in specific areas of interest. A total of four manta tow transects (8 km in total) and 114 drop-down video locations were surveyed during the groundtruthing exercise.
The manta tow videos were analysed by pausing the video at 20 s intervals (corresponding with the differential GPS waypoints). At each of these video pauses the percent of the image each seagrass species and habitat type was recorded. For the drop-down videos, percent representation of species of seagrass and habitats was averaged across the total video footage obtained from each drop unless the species composition varied noticeably. If it was variable, the drop was sub-divided to more accurately represent the seagrass coverage.
Seagrass species recognised in the video footage were A. antarctica, A. griffithii, P. australis, P. coriacea, P. sinuosa, H. ovalis, H. tasmanica and S. isoetifolium. Habitats other than seagrasses that were recognised were limestone reef and unvegetated sand.
It was not possible to separate the species of P. sinuosa and Posidonia angustifolia from the video data, as this requires an examination of the rhizome fibres (Cambridge and Kuo, 1979). Groundtruth dives indicated that P. sinuosa was generally more common than P. angustifolia (greater than 90% except in the Cockburn Sound East region where P. sinuosa cover was approximately 70%). In the present mapping exercise, these two species were both mapped as the P. sinuosa assemblage.
Notes
CreditThis project was funded by a consortium of industries and Western Australian Government Departments: Cockburn Cement Pty Ltd., Departments of Environmental Protection, Commerce and Trade and Resources Development, Fremantle Port Authority, James Point Pty Ltd., Kwinana Industries Council, Royal Australian Navy, Water Corporation of Western Australia and Waters and Rivers Commission of Western Australia.
Purpose
To set up a baseline for future monitoring of seagrass loss in Cockburn Sound.
To set up a baseline for future monitoring of seagrass loss in Cockburn Sound.
Issued: 11 09 2017
Data time period: 1967 to 1999
text: westlimit=115.6; southlimit=-32.30; eastlimit=115.8; northlimit=-32.05
text: uplimit=10; downlimit=0
Subjects
63 605002 |
63 617002 |
63 617003 |
63 617008 |
63 618001 |
63 618004 |
63 618005 |
63 619004 |
Amphibolis antarctica |
Amphibolis griffithii |
BIOLOGICAL CLASSIFICATION | PLANTS | ANGIOSPERMS (FLOWERING PLANTS) | MONOCOTS |
Biological Sciences |
BIOSPHERE | ECOSYSTEMS | MARINE ECOSYSTEMS | BENTHIC |
BIOSPHERE | ECOSYSTEMS | MARINE ECOSYSTEMS | COASTAL |
Benthic habitat |
Biotic taxonomic identification |
EARTH SCIENCE | BIOLOGICAL CLASSIFICATION | PLANTS | MACROALGAE (SEAWEEDS) |
Ecology |
Environmental Science and Management |
Environmental Sciences |
Environmental Management |
Environmental Sciences Not Elsewhere Classified |
Halophila ovalis |
Heterozostera tasmanica |
Marine and Estuarine Ecology (Incl. Marine Ichthyology) |
Other Environmental Sciences |
Posidonia australis |
Posidonia coriacea |
Posidonia sinuosa |
Syringodium isoetifolium |
biota |
oceans |
research aeroplane |
User Contributed Tags
Login to tag this record with meaningful keywords to make it easier to discover
Other Information
(DATA ACCESS - download all survey data (1967-1999) as zipped Shapefiles)
global : 4739e4b0-4dba-4ec5-b658-02c09f27ab9a
Identifiers
- global : 81edfca0-9d59-11dc-a0ca-00188b4c0af8
