Contacts

Identifiers and Keywords

ANZLIC Id

ANZVI0803009332 

Jurisdiction

Victoria 

Topic category

• Environment
• Biota
• Imagery base maps earth cover
• Farming

Resource Constraints

Use limitation

Data is available under a Creative Commons licence. For access please email coordinated.imagery@delwp.vic.gov.au 

Classification

Unclassified  

Creative Commons Attribution 4.0 (CC-BY)

• 
• License Text

Other Dataset Details

Date ( Revision )

2022-01-25

Status

Completed  

Purpose

The Tree Extent is intended to show both the presence and density of tree cover at a large scale. The potential uses are many, however, the initial purpose is to provide a detailed, comprehensive and consistent statewide dataset. Tree_Extent provides the most detailed statewide picture of tree cover available to date and is an excellent source of data for any applications requiring the identification of tree cover across the state. The dataset also lends itself well to map presentation and makes an excellent backdrop for other thematic information shown on maps at a large scale. 

Credit

John White, Cat Gilbert, Ross Honeyman, Emil Skender. 

Format

Digital 3 

Supplemental Information

History: The aerial photography was sourced via Department of Environment Land Water and Planning Coordinated Imagery Program. Relationship to other Datasets: Vicmap Vegetation Tree Density is sourced from this tree_extent layer. Tree_Density is a polygon layers which has generalised the tree_extent layers and contains additional attribution describing the separation of tree cover (Dense, Medium and Sparse). This layer also includes extra information such as the source date of the aerial imagery used to map the tree cover. Current Design Issues: TREE_EXTENT is derived from high resolution aerial photography and machine learning utlising a semantic segmentation model. Due to a degree of variability in the input imagery in some areas the machine learning model did not perform as well in some areas and resulted in poor predictions or lower accuracy of the tree extent. This particularly occurred in the vicinity of Dargo High Plains (Eastern Victoria) and Wyperfeld and Big Desert National Parks (North West Victoria) where there is low reliability of the mapping of tree cover. These areas should be improved with the introduction of new imagery. The processing of the imagery was broken into small units of 1km x 1km in size. This has resulted in some edge effects visible in the layer where the prediction/mapping of tree cover did not consistently cross the tile boundary. Two different semantic segmentation models were used across the state to increase the accuracy as one model could not be used consistently due to the variation in the tree cover. This has led in some areas a join issue where tee cover mapping across the model boundaries is not constant. Related Documents: None 

Maintenance and update frequency

Irregular  

Spatial

Spatial representation type

Grid  

Description

General - Victoria 

N

S

E

W

Temporal

Begin date

2013-12-08 

End date

2020-11-03 

Quality

Conceptual Consistency

All polygons are automatically generated in a raster to vector conversion. All polygons are closed and labelled consistently. All relationships between attributes are logical. 

Missing Data

Statewide Completeness Verification: After processing the state-wide tree extent data waterbody and selected crop masks were applied to remove any incorrect tree classification. No additional human intervention post processing was performed on the data. Intersection over Union score (IoU) was 80% across the whole of the state. 

Excess Data

Primary input was derived from aerial photography which has a pixel resolution of 20cm (or was downsampled to 20cm from a higher resolution). 

Attribute Quality

This raster dataset classifies tree presence as a cell value of 1 and tree absense as a cell value as 0. Intersection over Union score (IoU) over the whole of the state was 80%. The IoU is the amount of area shared by the human delineated data (test set) and machine learning prediction (intersection) to the total area covered by either (union). 

Positional Accuracy

The positional accuracy is determined by the geometric rectification of the high resolution aerial photography. The aerial photophaphy metadata can be found using the associated ANZLIC ID in the metadata for Vicmap Vegetation Tree Density. On average it was +/- 80cm RMSE 

Lineage

Description

A presence/absence tree cover dataset is derived from the Vicmap Vegetation Tree Extent statewide dataset (20cm pixels) . This classification was based upon a semantic segmentation machine learning approach where each pixel was assigned if it was tree cover or not with the resulting raster output having a cell value of one for existence of tree cover and zero for none. A mask was then applied to remove areas incorrectly classified as tree cover in water bodies and some crop types. 

Statement

Dataset Source: Vicmap Vegetation Tree Extent has been created using machine learning and high resolution aerial photography. Dataset Originality: Derived 

About the Metadata Record

Metadata identifier

f6800447-ef34-5f66-acaa-77a5f2936546

Resource Type

Dataset  

Date info ( Revision )

2022-01-25

Standard Name

ISO 19115-3:2018 

Profile Name

DELWP Profile 

Profile Version

Version 1 

Profile Date

2019-05-24  

Metadata Constraints

Classification

Unclassified