Model overviewThe indicative Australian Urban Development Risk Model is based on an assumption that recent-past trends in urban expansion (i.e the transition from non-urban land use to urban land use) will continue linearly, and that parameters associated with past expansion are valid predictors of future expansion. The model is underpinned by a conceptual logic, derived within ERIN, based on known datasets and their reasonable association with patterns of urbanisation. Specifically, we predict a higher urban development risk for non-urban locations with:proximity to existing high urban development areashigh increasing trend in street address densityland uses evidently prone to urbanisation andattractive geomorphology.The model is stratified by Australia’s 109 Significant Urban Areas and eight Greater Capital City Statistical Areas (ABS, 2016) and the model output is limited to these zones.Users should note there are likely to areas of high urban development risk beyond these zones, as discussed in the limitations section below.The model draws on multiple datasets to derive values for the parameters above and then combines them into a single index, with a value for every cell in a 9-second grid (about 1.2 million 250 x 250m cells). Derivation of parameter values is described below, followed by the approaches used to combine them into the index, classifying values for mapping, and combining with non-index masks to make the model spatially complete. Model parameters:The model is based on four parameters or predictor variables. For each parameter the field name (in the GIS data spatial attribute table) is provided in square brackets.1. Proximity to existing high urban development areas [NEAR_DIST]This parameter assumes continuation of 2006-2016 trends in urban development within a given Significant Urban Area or Greater Capital City Statistical Area. Locations close to an urban fringe which had expanded significantly during this period are at higher risk of urbanisation. Identifying past change from non-urban to urban involved comparing 2006 and 2016 ABS mesh blocks data. These datasets use a land use classification comprising 10 categories which were reclassified into urban(commercial, education, medical, industrial, residential, transport) and non-urban(parkland, water, primary production, other). Centre points of all 2016 urbanmesh blocks were compared with 2006 non-urbanmesh blocks to identify new urbanmesh blocks, or those which had changed from non-urbanto urban.These new urbanmesh blocks were used to attribute individual cells in the 9sec grid.Distances were then calculated between each new urbancell and its nearest 2006 urbanmesh block.Means (2006 dist. mean) and standard deviations for these distances were derived within each of Australia’s 109 Significant Urban Areas and eight Greater Capital City Statistical Areas (Areas). Larger mean values indicate greater urban expansion over the period for the Areain question. The standard deviations indicate how variable the expansion was within the Areaand, as shown in the table below, were used to account for uncertainty.To extrapolate a risk rating, all 2016 non-urbancells were converted to points and analysed for their distance to the closest 2016 urbancells. This distance was then compared to the relevant Areamean (2006 dist. mean) as described above. Where a 2016 non-urbancell is closer to an urbancell than the mean distance for conversion in the period 2006-2016, it is rated at higher risk, and particularly so for an Areawhere standard deviations are lower.The following table shows thresholds and parameter values. Conditions for cellParameter Value2016 dist. ≥[2006 dist. mean + standard deviation] 0.12016 dist. ≥[2006 dist. mean] but 0.42016 dist. ≥[2006 dist. mean - standard deviation] 0.72016 dist. –standard deviation 12. Increasing trend in street address density [setGnaf]A density of 60 addresses per 250m cell roughly equates to a ‘quarter acre block’urban landscape. This parameter assumes the continuation of trends in street address densification apparent during 2009-2016. Lower density locations (less than 30/cell) are considered non-urban and not at risk. Moderate density locations demonstrating significant increase during 2009-2016 are considered at high risk of urbanisation.The Geocoded National Address File (GNAF) dataset was used to derive both the number of addresses/cell for February 2016, and the increase from May 2009 to February 2016. The following thresholds and parameter values were applied.Conditions for cellParameter ValueLow density, low densification areas (ie all other than the following)02016 GNAF density ≥30 addresses/cell and density change (2009-2016) ≥2012016 GNAF density ≥40 addresses/cell and density change (2009-2016) ≥1012016 GNAF density ≥60 addresses/cell13.Land uses evidently prone to urbanisation [setLandUse]This parameter builds on the analysis used for Parameter 1. At assumes that past high likelihoods for urbanisation associated with certain land use types in different Areaswill persist. The 2016 new urbancells from Parameter 1, above, were compared to a 9 sec grid of the 2006 land use categories (derived from 2006 mesh blocks). For each land use, in each Area(ie a Significant Urban Areas or Greater Capital City Statistical Areas) the proportion urbanised was calculated as a number between zero and one. This number was directly applied as a parameter value for all 2016 non-urbancells. For example, a non-urbancell on a land use which had demonstrated a 60% chance of conversion to urbanin the period 2006-2016, would be scored at 0.6 for this parameter.4. Attractive geomorphology [setSlope]This parameter assumes that past preferences for urbanisation of lower slope areas will continue, given lower costs associated with developing such sites. Slope was calculated,from Geoscience Australia’s 1sec digital elevation model, as the mean slope across each 9sec cell. The following thresholds and parameter values were applied, and are based on a limited research effort into accessible building codes for new dwellings.Conditions for cellParameter ValueSlope ≥20 0.1Slope ≥12 and 0.4Slope ≥6 and 0.7Slope 1Derivation of the indexParameters were combined with equal weight on the assumption that each makes an equal contribution to our capacity to predict future urban expansion. However, individual parameter values are included in the GIS dataset to allow weights to be adjusted to suit particular analyses.Index Value = 0.25 x (proximity to high urban development) + 0.25 x (street address density) + 0.25 x (urbanising land use) + 0.25 x (attractive geomorphology)Classification of index values The index derives values for all cells between zero and one. These were classified into five equal-sized categories from “very low” to “very high”risk.Derivation of mapping unitsMapping units comprise the five risk categories, masked by non-index values for protected areas and existing-urban areas, as follows:Cells identified as protected, either through their inclusion in the Collaborative Australian Protected Areas Database or as ‘offset’areas in an EPBC Strategic Assessment area are ascribed a value of zero.Cells assessed as 2016 likely-urban for the NEAR_DIST parameter, attributed as ‘residential’in the Mesh Block layer,and with greater than 60 GNAF addresses are predicted to be existing-urban areas and ascribed with a value of 1.Non-index maskIndex valueRisk categorySuggested RGB for map colours: Protected from development38, 115, 00 0.2, Very low56, 148, 00.2 0.4, Low152, 230, 00.4 0.6, Moderate255, 255, 00.6 0.8, High255, 163, 430.8, Very high255, 0, 01, Existing urban239, 228, 190
