Survey sites
The mapping team acquires survey site data from established BioNet records such as Systematic Flora Survey data. The mapping team may also conduct surveys to collect floristic site data for upload into BioNet and other landscape-based information that is essential to remote sensing interpretation.
Segmentation
We use sophisticated feature recognition software, originally developed for medical imaging technology to analyse satellite imagery and aerial photography to delineate small patches of vegetation that have the same visual characteristics. This technique is generally known as segmentation because the process divides the image into segments and then automatically draws polygons or lines around each of these areas of similarity according to a defined set of rules. It effectively replicates what humans do when interpreting imagery, but saves considerable time and cost compared with manually drawing lines.
The whole image is segmented; including non-vegetation features such as water bodies, infrastructure and buildings so that we can construct a pre-clearing native vegetation map as well.
Structural groups
Next, remote sensing experts use their skills and experience to assign each of the segments to a vegetation group called Vegetation Photo Patterns. Patterns include multiple categories that are observable by aerial photographic interpretation, for example: Dry Sclerophyll Forest, Coastal Floodplain Forest, Heathland, Rainforest or Wetland. We use a range of clues to help identify the Patterns; including survey sites, previous mapping and topographic layers like geology and altitude.
Patterns are created for the whole landscape so we can create a pre-clearing PCT map.
Incorporating detailed mapping
The SVTM has been designed to readily incorporate new information without the need for whole-scale remapping. In fact, the map already includes some existing mapping. These sources have used the same or equivalent vegetation types but may have mapped to higher levels of detail or had more intensive field verification.
Fine-scale vegetation mapping is often undertaken independently and for very specific purposes, including management of Threatened Ecological Communities. We can help other agencies, local government or private consultants to maximise their compatibility for incorporation into the SVTM.
Computer modelling
We use sophisticated computer models to assign each PCT into an appropriate landscape Pattern. The models identify the mostly likely PCTs within each Pattern.
We also ensure that our models undergo rigorous internal validation and cross-checking and we regularly set aside a pool of samples to independently test the model performance.
The extant and pre-clearing maps
We use the NSW native vegetation extent layer to mask out areas that are not native vegetation. This provides us with the extant map of PCTs.
Natural native grasslands are added back into the extant map using another layer called the Seasonal Disturbance Image. This helps us locate natural grasslands which cannot be reliably mapped using aerial photographic interpretation - unlike most other Pattern types.
Derived communities and areas that are not native vegetation; such as pine forests, plantations, weeds, open water, agricultural paddocks, built infrastructure, urban areas, roads, dams and exposed earth are excluded from the extant map.
Expert review
Experienced ecologists review the draft PCT map to check for any issues like ecological inconsistency, gross errors, spatial inconsistencies and discontinuities. Errors and omission are manually corrected by expert landscape ecologists.
NSW State Vegetation Type Map: Technical Notes provides more technical detail about how the map is created.
