Title

Port Phillip Bay seagrass mapping at nine aerial assessment regions in April 2010 

Alternate title

SEAGRASS_PPB_9REGIONS_APRIL10 

Purpose

The objective of the Baywide Seagrass Monitoring Program is to detect changes in seagrass health outside expected variability. Seagrass is being monitored at several spatial scales including aerial mapping undertaken annually in April/May. This data presents seagrass mapping from aerial photography at one of nine detailed aerial assessment regions established in the main seagrass areas in Port Phillip Bay. The seagrass mapping will be compared to the previous year's mapping and historic mapping to detect changes in percent cover outside expected variability. This dataset replaces SEAGRASS_ALTONA_2010, SEAGRASS_BLAIRGOWRIE_2010, SEAGRASS_CURLEWIS_BANK_2010, SEAGRASS_KIRK_POINT_2010, SEAGRASS_MUD_ISLANDS_2010, SEAGRASS_PT_HENRY_WEST_2010, SEAGRASS_PT_RICHARDS_2010, SEAGRASS_ST_LEONARDS_2010 and SEAGRASS_SWAN_BAY_2010. 

Credit

David Ball 

Supplemental Information

History: See seagrass mapping at this site from 2008 and 2009. Relationship to other Datasets: This data forms part of a range of seagrass health measurements being recorded across Port Phillip Bay as part of the Baywide Seagrass Monitoring Program. Additional data collection at six of the regions (all excluding Altona-Point Cook, Curlewis Bank and Point Henry West) includes field-surveys using fixed quadrats at seagrass plots within the mapping area. Current Design Issues: The mapping design is described in the Baywide Seagrass Monitoring Program Detailed Design (CDP_ENV_MD_022 Rev5 - Port of Melbourne Corporation, 2010). Related Documents: None Underwater video ground-truthing 2010, April 2010 aerial photography 

Status

Completed  

Horizontal Accuracy

200m  

Code

4283 

Maintenance and update frequency

Annually  

Topic category

• Biota
• Farming
• Oceans

Use limitation

. 

Classification

Unclassified  

Statement

Dataset Source: Aerial photography flown by Aerometrex Pty Ltd on 17 April 2010 and video ground-truthing by DPI Queenscliff from 15 April-27 May 2010. Dataset Originality: Primary & Derived 

Description

Collection Method: Aerial photo interpretation, underwater video ground-truthing 

Description

Source description not available 

Description

Source description not available 

Description

Aerial photography of the Port Phillip Bay seagrass areas was flown by a consultant (Aerometrex Pty Ltd) 17 April 2010. As far as possible, the photography was flown under optimum conditions for seabed habitat. The orthorectified aerial photography was delivered to Fisheries Victoria Fisheries Research Branch (FRB) as unprocessed geotiff files for each frame and a colour balanced mosaic of all the photography in a compressed ECW format. The unprocessed geotiff files were used for the aerial mapping. Seagrass habitat mapping was undertaken with the remote sensing software ENVI. A subset of the area to be mapped was extracted from the photography and exported to a separate ENVI Image file. For all regions except Point Henry West, the mapping approach applied a principal components transformation to the red, green, blue and infra red colour bands, except Blairgowrie where the transformation was not applied to the infra red band, to create a principal component image. An unsupervised classification was then applied to the whole image at the Blairgowrie and Point Henry West regions, the first PC band of this image at the Altona-Point Cook, Kirk Point, Mud Islands and Point Richards-Bellarine Bank regions and the first two PC bands of this image at the Curlewis Bank, St Leonards and Swan Bay regions to transform it into a thematic map. The output from this process was a single band image with 15 values representing the number of categories defined by the unsupervised classification. The classified image was overlayed on the aerial photography to allow a visual comparison. Groups of pixel values within the classified image were then assigned to a thematic classification based on the dominant seabed type (i.e. sparse seagrass, medium-dense seagrass or sediment) using the 'Combine Classes' function n ENVI. The classified image was then re-calculated so that a single pixel value corresponded to a single habitat type. The small pixel size of the aerial photography (0.3 m) meant that even the classified image retained isolated pixels or noise in the data. Following the unsupervised classification, a median filter was applied to smooth out any image noise in the data. The final phase of the seagrass mapping was to convert the classified raster image to a vector layer (shapefile) in ArcGIS. The classified raster data was made up of a grid of pixels with each pixel assigned a value representing the habitat feature at that location. Groups of adjacent pixels with the same value represented a single habitat feature such as a seagrass bed. A "smoothing" algorithm was applied during the raster to vector conversion process. The seagrass habitat shapefile from the raster to vector conversion was imported to a Geodatabase and polygon topology was created. To remove extraneous polygons generated by the raster to vector conversion, the Eliminate command was used to remove all polygons less than 2 m2 at all regions except Blairgowrie, where polygons less than 1m2 were removed. The layer was then checked on screen and some manual edits were made to seagrass polygons which did not accurately match the aerial photography. A text attribute "Habitat" was then added to the layer and the habitat description was added for the relevant codes. The Geodatabase feature class was then exported back into a shapefile format to allow distribution. Seagrass mapping from the aerial photography was ground-truthed with underwater video. The ground-truthing sites were positioned to maximise coverage of field observations across the range of seagrass habitats present and to target any areas of uncertain classification. Video sites were also positioned amongst bare sediment to assist in assessing the mapping accuracy. The video was interpreted by a benthic ecologist at FRB to identify the dominant habitat. For each seagrass site the seagrass density was classified as Bare <1% cover, Very Sparse 1-10% cover, Sparse 11-20% cover, Medium 21-60% cover, Dense 61-100% cover. Seagrass densities <10% are typically below the level able to be interpreted from aerial photography. Seagrass epiphyte cover (%) was also estimated from the video. The video point data was overlayed on the seagrass habitat mapping to determine the mapping accuracy. Initially, nine separate layers were created, one for each region of Port Phillip Bay. Later, ArcToolbox's Merge tool was used to merge these nine shapefiles into a single shapefile, and a consistency check was applied to the final shapefile's attributes. 

Classification

Unclassified  

Attribute Quality

Comments

Description: The habitat layer attributes describe the dominant habitat (substratum and vegetation) and vegetation density. The accuracy of the mapping was determined by comparison to point observations from underwater video. Determination: Overall mapping classification accuracy was determined by generating an error matrix from the video ground-truthing overlayed on the mapping. The overall mapping accuracies for the regions are: Altona-Point Cook: 80% based on 15 video ground-truthing points. Blairgowrie: 88% based on 26 video ground-truthing points Curlewis Bank: 100% based on 15 video ground-truthing points Kirk Point: 67% based on 21 video ground-truthing points Mud Islands: 100% based on 15 video ground-truthing points Point Henry West: 100% based on 12 video ground-truthing points Point Richards-Bellarine Bank: 86% based on 35 video ground-truthing points St Leonards: 100% based on 13 video ground-truthing points Swan Bay:95% based on 20 video ground-truthing points. 

Positional Accuracy

Comments

Description: Positional accuracy of the mapping is dependent on the accuracy of the source aerial photography, and the accuracy of the habitat boundaries is related to the mapping accuracy. Determination: The estimate of positional accuracy for the aerial photography was determined by comparison to known ground-control points. Overall mapping accuracy was determined by comparison with video-ground-truthing, and no quantitative measure of positional accuracy was undertaken. 

Conceptual Consistency

Comments

Polygon topology was created in ArcGIS and all polygons were checked to ensure they had labels and an attribute value. 

Missing Data

Comments

All of the regions mapped are covered by this layer Completeness Verification: The dataset has been verified for all attributes. 

Excess Data

Comments

Attributes describe the dominant habitat type but are not exhaustive. The habitat classification groups all seagrass and macroalgae into single categories as it is not possible to derive species type from aerial photography interpretation.