Title

Native Vegetation - Modelled Quality (Site condition and landscape context) 2005 

Alternate title

NV2005_QUAL 

Purpose

To provide a strategic view of native vegetation quality across the state. Site Condition ranges are best considered as relative rather than actual values and site based habitat hectare assessments are required to ascertain actual vegetation condition at site To provide a consistent assessment of landscape context across the state; and where applicable to replace the field assessment of landscape context that is part of a site-based habitat hectare assessment 

Supplemental Information

Relationship to other Datasets: NV2005_EXTENT is part of a series of native vegetation datsets, as follows: NV2005_EXTENT - Modelled extent of native vegetation (very broad categories) NV1750_EVC - Mapped pre 1750 ecological vegetation classes (EVCs) distribution NV1750_EVCBCS - Mapped pre1750 EVC distribution with Bioregional Conservation Status derived from NV1750_EVC and VBIOREGION100 and depletion statistics, also contains pre1750 geographic occurrence of EVCs NV2005_EVCBCS - Extant EVC distribution with Bioregional Conservation Status derived from NV1750_EVCBCS and NV2005_EXTENT , also contains pre1750 geographic occurrence NV2005_QUAL - Modelled Native Vegetation Quality - includes modelled site condition and patch-based landscape context NV2005_LSIMP - Landscape Scale native vegetation importance and connectivity derived from NV2005_EVCBCS, NV2005_QUAL and landscape connectivity Current Design Issues: In general the modelled dataset tends to underestimate the condition of vegetation at high quality sites and overestimate the condition of vegetation at low quality sites. As such, it is inappropriate to use the data for informing decisions of a statutory nature. Landscape context is wholly dependent on the accuracy of the Native Vegetation dataset upon which it is based , in this case NV2005_EXTENT. This version scores 80% of the landscape context score, a further 20% ie. 5 points associated with low disturbance history (such as areas of old growth) has not yet been included. Future Design Issues: Inclusion of scores for low disturbance Related Documents: None 

Status

Completed  

Horizontal Accuracy

25m  

Code

4283 

Maintenance and update frequency

Unknown  

Title

DIGITAL ESRI grid DIGITAL TIFF image DIGITAL Oracle/SDE NV2005_QUAL.AVL, NV2005_QUAL_SC.AVL, NV2005_QUAL_LC.AVL - ArcView3.3 Legend NV2005_QUAL.LYR, NV2005_QUAL_SC.LYR, NV2005_QUAL_LC.LYR - ArcGIS 9 Layer File NV2005_QUAL_LEGEND.tif 2 

Topic category

• Biota

Use limitation

Public Access This is a landscape scale dataset, site verification is required for site based projects Planning or investment decisions at the site-scale should use some form of ground-truthing. Note: in general the modelled dataset tends to underestimate the condition of vegetation at high quality sites and overestimate the condition of vegetation at low quality sites. As such, it is inappropriate to use the data for informing decisions of a statutory nature. 

Classification

Unclassified  

Statement

Dataset Source: Site Score: Each of the 7 site component models have been constructed using "Site Score" data collected from approximately 18,000 sites across Victoria. Only terrestrial sites were surveyed - wetland condition/quality was not assessed in the field. At these same locations the values of many environmental (climate, soils, terrain etc) and spectral data (landsat bands) were extracted within a GIS). Relationships between site score components assessed at point locations during field survey and the environmental variables, were modelled using artificial neural networks. Neural networks were built for the individual condition variables using the Neural Networks module within the Statistica software package (StatSoft Inc., Tulsa, Oklahoma). For each output the modelling procedure involved the following steps. The input data (environmental variables) and output data (site component score indices) were standardized to means of zero and units of standard deviation. The dataset was then divided into training (half of the field data), selection (one quarter of the field data) and test data (one quarter of the field data). The training data were used to build the neural network, the selection data were used to check that the procedure was not over-fitting (i.e. to prevent models fitting just the idiosyncrasies of the training data) and the test data were used to validate the final fit of the model. Models were built from 200 random starts and the 50 models with the best statistical fit (highest R2 values) were saved. The best performed site component models based on both statistical fit, input variable parsimony and visual assessment were selected. These models were then used to predict and map site component scores for every 25 m grid across Victoria. These models were summed to produce an initial "Site Score" model. Finally the site score for naturally treeless vegetation (such as grasslands and shrublands) was rescaled to compensate for the absence of tree related site score components. Landscape Score: Dataset was generated from a spatial analysis of NVE_2005. Dataset Originality: Derived 

Description

Site Condition Score Model of the weighted sum of site component scores as per Parkes et al (2003). Component surfaces derived from neural-net modelling (using Landsat spectral data from 1991 to 2005) and a further 40 ancillary environmental variables including climate and terrain Landscape Score: 1. NV2005_EXTENT was reclassified into habitat (highly likely native vegetation & wetland habitat) and non-habitat 2. Habitat was separated into small and/or narrow linear patches based on distance from edge; and larger rounded patches 3. Area and maximum thickness are calculated for each patch and a patch size and shape rating is applied 4. Percentage habitat is calculated for five neighbourhoods around each cell and summarised into a single cell rating of surrounding vegetation 5. Weighted distance is calculated to 4 different patch size/shape classes and summarised into a single cell rating of connectivity. Distance is weighted such that distance cost is incurred between native habitat and no distance cost is incurred when habitat is present 6. 3, 4 and 5 are combined into a single rating of landscape context for existing habitat 

Classification

Unclassified  

Attribute Quality

Comments

In general the modelled dataset tends to underestimate the condition of vegetation at high quality sites and overestimate the condition of vegetation at low quality sites. As such, it is inappropriate to use the data for informing decisions of a statutory nature. The Landscape Context class accuracy is dependent on the accuracy of the source extent dataset.(Landscape Context processing steps for methodology) 

Positional Accuracy

Comments

Accurate 

Conceptual Consistency

Comments

The same approach to assessing site condition is applied across the state, however the site score for naturally treeless vegetation (such as grasslands and shrublands) was rescaled to compensate for the absence of tree related site score components. The same approach for calculating Landscape Context has been applied to the complete NV2005_EXTENT dataset 

Missing Data

Comments

Complete Statewide Coverage Completeness Verification: For Site Condition, the neural network modelling process, involving many thousands of groundtruthing points, along with expert validation Landscape Context is a spatial analysis of NV2005_EXTENT, as a result is dependent on the accuracy of the outer extent boundary of this dataset (see NV2005_EXTENT metadata for further information) 

Excess Data

Comments

Site Condition ( 0-75). has been classified into the following broad classes: 0,1-20,20-30,30-40,40-50,60-75 Landscape Context (0-25) currently does not include any score for lack of disturbance (eg. old growth) and as such the maximum rating is 20. Landscape Context has been classified into the following broad classes: 0-5, 6-10,11-25,16-20 Vegetation Quality has been classified into the following classes: 1-20,21-30, 31-40,41,50,51,60,61-70 and 71-100