How It's Worked Out

Overview

This page sets out the method behind the homes, people, jobs and land in the results: its formulas, its sources and how far to trust each. Numbers are estimates for testing options, not a planning opinion.

What's Here Now

Homes and people. The smallest published counts are ABS mesh blocks, and a site is rarely a whole one. So each mesh block is shared by its address points:

homes on the site = Σ mesh block's homes × (its addresses on the site ÷ its addresses)

A mesh block with no addresses is shared by area, and the result says so. People work the same way. The counts are from the Census of August 2021, and addresses are today's, so homes built since aren't counted. The counts include occupied and unoccupied homes. A mesh block isn't a site, and this is an estimate you can change.

Workplace jobs. No open count is fine enough for a site, so they're not known here until you enter them.

Land. Constrained land is road and unparcelled land, Floodway, Rural Floodway and Public Acquisition Overlays, and public park and transport zones. Developed land is a lot with at least 40 m² of building footprint. The rest is vacant. A lot whose building is missing from the footprint data reads as vacant.

Existing − Lost + New = Total

Total = Existing − Lost + New

• On a lot, what's built replaces what's there, so all of Existing is lost, unless you tick Keep what's here.
• On a precinct, 5 ha or more: Lost = Existing × (redevelopment + abandonment).

What's unknown stays unknown, and one unknown site leaves the scenario's total unknown. "Lost" means no longer on the site. A job there may move elsewhere, so it isn't people out of work.

New Homes and Jobs

new homes on a precinct = vacant land × lots share × homes per hectare of lots + developed land × redevelopment × homes per hectare of lots

Vacant land gives up streets, parks and civic land. Developed land already has them, so the redeveloped share counts in full. A lot uses all its land.

Homes per hectare of lots come from a type's building mix, whose shares are of land, not homes.

People are homes × people per home by kind of home (ABS Census 2021, G36, for the area, blended with Greater Melbourne's).

Workplace jobs are non-residential floor space by use ÷ floor area per job. Homes count as none, and working from home isn't counted.

Streets, Parks and Net Developable Area

street reserve = carriageway + verges (Clause 56.06, Table C1) block to the centre line = (block width + reserve) × (block length + reserve) street share of a block = (that − block width × block length) ÷ that parks = target × (1 − civic) ÷ (1 + target) Net Developable Area (NDA) = 1 − parks − civic streets = street share of a block × NDA lots = NDA − streets new street km per hectare of NDA = (block width + block length) ÷ 1,000 × 10,000 ÷ block to the centre line lane-km = new street km × lanes

How this was changed from Envision Tomorrow. The method comes from its Scenario Builder (v3.7.6), in metres and with Victorian values. Its shares are added together. Here, parks and civic land come out first and the street share applies to what's left, because a Precinct Structure Plan's NDA includes local streets and its open space target is a share of NDA. No Envision Tomorrow numbers are used.

Block sizes, each type's street, civic shares and redevelopment rates are our assumptions, marked as such, and you can change them in Assumptions. Clause 56 is for residential subdivision, and stands in for commercial and industrial streets.

Transport cost is Infrastructure Victoria's $70,000 for each home on vacant land, covering roads, intersections and public transport together. It isn't the cost of these streets alone.

How Households Travel

These are in Transport Impact. They describe today's households in places like this, not who'll live here. People who prefer walking choose walkable places, which isn't controlled for.

Why not Envision Tomorrow's model. Its household model needs what's within about 400 to 1,600 m of each home. The open VISTA travel survey only says which council and ring a home is in, so it can't be repeated. The documents in the copy we read report no test on places left out of the fit, though later work on it may.

Cars per household, for 359 SA2s in Melbourne and Geelong:

ln(cars per household) = −0.539 − 0.106 ln(homes a hectare of urban land) − 0.043 (has a train station) − 1.409 (share of intersections with four or more arms) + 0.141 ln(median household income) + 0.431 ln(household size)

Tested on places left out of the fit, by SA3: R² 0.80, against 0.53 from income and household size alone. With whole SA4 regions left out: 0.75 against 0.43. Beyond 8 km from the centre: 0.60 against 0.40. These measures were picked using the same tests, so that's slightly optimistic: all seven measures gave 0.79. Across places, those twice as dense have about 7% fewer cars per household, other things equal. That's a pattern at one time, and people's choice of where to live is part of it, so it isn't what adding homes would cause. The 80% range comes from the held-out error for whole areas, and leaves out uncertainty in the coefficients and differences between one building's households and its area's.

Driving, walking and public transport, from 23,866 weekday VISTA households in 34 councils, 2012 to 2025 with COVID years marked, with council density as the only measure of place:

"Used" means density's gain beat 95% of runs with councils shuffled and reached 0.01. With seven measures of place on only 34 councils, some signs came out wrong, so one measure was used. Fit is low, as it is for any one household's day. Density here stands in for everything about a place, so its effect can't be read as what building more homes would cause, and the app shows a range: from no change, to the whole difference between places of those densities. Density is held within the councils fitted, 0.7 to 74 homes a hectare.

Baselines are each council's weighted weekday figures, 2022 to 2025 where it has 150 or more households, otherwise 2012 to 2025. A council with under 100 households isn't shown. A household's driving is capped at 300 km a weekday in these means, as a long trip away isn't everyday driving. Geography: a council or SA2, not a site, and a scenario's sites all count in one area. The Census of August 2021 fell in Melbourne's lockdown, so its journey to work isn't used.

Parking demand is cars per household × new homes, beside Clause 52.06. For planning, not a parking study.

Activity is minutes walking and cycling for transport on a weekday, including walking to public transport, measured by council, with MET values of 3.5 for walking and 6.8 for cycling. It also gives the share of adults, 20 and over, reaching 30 minutes on the surveyed day. The Australian guidelines ask for 30 minutes or more on most days, and one day can't show who meets that. Other activity isn't counted. No disease, body weight or mortality figures are given.

Trips kept inside a mixed-use site: not estimated. Transport for NSW's guide says to apply a capture rate, and no Australian rate was found.

Road Crashes Around a Site

Fatal and injury crashes reported to police within 400 m of the sites, on Victorian roads, 2022 to 2025, beside the council's and Victoria's rate per resident. Earlier years include lockdowns, property damage alone isn't recorded, and about 8% of 2025's crashes had no location yet.

No change is projected. Envision Tomorrow moves a local rate with driving and density. Here, surroundings did predict crash rates on places left out (R² 0.39 against 0.16), but crashes are counted where they happen and people where they live: the highest rate is in an industrial area with few residents. So adding residents would make a place look safer on paper. Speed, which drives severity, isn't in the model either.

How Far to Trust It

In measured checks, road and other constrained land ran 24% to 41% on six of eight places tried, against street shares of 19% to 32% here. That's a rough check, not a calibration.

Downloads and Units

Screens use hectares and metres. Results by site download as GeoJSON or CSV in WGS84 (EPSG:4326), with areas in hectares. In the API, painted_ha, vacant_ha and developed_ha sit beside the older acres fields, which stay for saved data. net_developable_ha means the lots alone, and nda_ha is the NDA.

Where to See It

• Site card: What's here now, with its figures to change, and a tick to show the buildings there on the map.
• Results: the sum, per household, land budget and downloads.
• Compare: Existing and scenarios.
• Assumptions: each number behind a type, with its sums.

Sources