Wetland Vegetation of the Macquarie Marshes 2022

Brief description

This wetland vegetation map is produced from air photo interpretation techniques and imagery acquired in June 2022.

Map development began with the collection of high-resolution aerial colour (Red-Green-Blue) imagery. The imagery was provided as an orthographic mosaic (ie a straight down view) with a 40 cm ground sampling distance covering the whole study area. This formed the primary input of information for vegetation extent mapping.

Several interpreters were then trained in Aerial Photographic Interpretation (API) to visually analyse the imagery to identify and delineate different vegetation types. The Aerial Photographic Interpretation separated vegetation types using spectral characteristics, colour, texture, shape, spatial patterns and associations with predictive environmental layers (such as flood frequency categories, elevation and geomorphology type). Existing survey data was also used to help identify vegetation types from imagery. This included BioNet species data, floristic data and other grey literature. Oblique aerial handheld photos captured from a helicopter were also sourced from another project to inform the aerial imagery interpretation. A subset of the available oblique handheld photos was selected to correspond to the timing (within two years) of the 40cm aerial imagery acquired for vegetation map development. The subset of oblique handheld photos adopted to inform the air photo interpretation included photos collected between January 2022 to April 2023 at the Macquarie Marshes.

A polygon layer divided into small regions was sourced to overlay on the 40cm aerial imagery. This spatial layer was produced using the Definiens eCognition software package. The polygon layer was generated with a computer-based image analysis tool known as segmentation. Inputs to the segmentation tool included a set of raster datasets with a 5m grid cell size. The segmentation tool produced a spatial layer of ‘segments’ or very small polygons based on the combined spectral and textural features of the input rasters (Roff et al., 2022). The segmented layer was overlayed on the 40cm aerial imagery. Interpreters then manually selected groups of segments and assigned classes (‘attributes’) to the polygons to delineate vegetation patterns. The use of the segmented spatial layer enabled more efficient mapping, as interpreters did not have to manually draw polygon linework with a mouse.

Vegetation patterns were interpreted from the high-resolution 40cm aerial imagery at a scale of 1:25 000 for non-flood dependent vegetation and at a scale of 1:10 000 for wetland communities. The minimum map unit (smallest polygon) was 2 ha.

Selected polygons from the segmentation process were initially assigned to an artificial class referred to as a Vegetation Photo Pattern (VPP), analogous to NSW Vegetation Classes (for more information on NSW Vegetation Classes see https://www.environment.nsw.gov.au/topics/animals-and-plants/biodiversity/nsw-bionet/the-nsw-vegetation-classification-framework ).
The VVPs were aligned with plant community types (PCTs) as described in the NSW BioNet Vegetation Classification Database (see https://vegetation.bionet.nsw.gov.au/).

The accuracy of the map wetland vegetation functional groups was assessed using 505 independently collected field validation points.

The overall accuracy was 0.74 and the Kappa statistic was 0.67.

Each wetland PCT was also aligned to a vegetation functional group corresponding to the vegetation objectives in the Macquarie Marshes Long Term Watering Plan.

Accuracies and 95% confidence intervals for map individual map classes were:
Non woody wetland: 0.89 (0.84 to 0.94)
Flood dependent woodland: 0.71 (0.61 to 0.81)
River red gum forest: 0.24 (0.00 to 0.41)
River red gum woodland: 0.73 (0.64 to 0.81)
Terrestrial vegetation: 0.73 (0.65 to 0.81)
Non-native or other (includes pasture, cropping, infrastructure, dams): Not assessed. No field survey data.

This mapping project was funded by the NSW Water for the Environment Program.

Full description

Data Quality Statement
Download Package