top of page

Resources

Finding Housing Realism Within Zoning Optimism: The AI approach to Missing Middle planning

  • Team Forecaz
  • May 30, 2025
  • 4 min read

KEY TAKEAWAYS:

  • Up-zoning alone does not deliver housing. Most parcels in established suburbs will not redevelop due to high site coverage and unfeasible demolition requirements. AI-driven urban growth modelling identifies the specific parcels where Gentle Density/Missing Middle development is feasible, giving planners a realistic view of actual housing supply.


  • When applied to Toowoomba Regional Council's inner suburbs, the Forecaz Gentle Density (Missing Middle) model scenario exceeded the State’s SEQ Regional Plan dwelling diversity targets, projecting a 7% increase in low-rise attached dwellings relative to detached by 2046, compared to the regional plan's 5% target.


Bradley Rasmussen, Forecaz CEO, presented at the Planning Institute of Australia's national Congress 2025 in Darwin. Making the case that zoning optimism and development realism are two different things, and showing how AI-driven urban growth modelling bridges the gap between them through a practical case study applied to Toowoomba Regional Council's inner suburbs.



Congress Overview

PIA Congress 2025 in Darwin
PIA Congress 2025 in Darwin

The PIA Congress 2025 brought planners from across Australia together in Darwin to address some of the most pressing challenges facing the profession. Bradley Rasmussen took the stage to present "Finding Housing Realism within Zoning Optimism," a session focused on one of the central tensions in housing policy: zoning for more dwellings does not automatically mean more dwellings get built.


Queensland’s Regional Housing Context

The context is significant. The updated ShapingSEQ Regional Plan, which took effect in December 2023, sets out Queensland Government's strategy to manage South East Queensland's growth to 2046. SEQ's population is forecast to grow from 3.7 million to 5.9 million, a 57% increase. The dwelling stock needs to grow from 1.5 million to 2.4 million, requiring 665 new dwellings to be delivered every week for the next 25 years. The plan targets 60% of new dwelling growth in attached dwellings and proposes Gentle Density, also called Missing Middle, as the approach to achieving low-rise attached housing supply. Dwelling diversity sub-targets have been set for each Local Government, with 29% of new growth expected to come from low-rise attached dwellings.


What Is Gentle Density / Missing Middle Housing?

Gentle Density/Missing Middle housing covers a range of built forms: secondary dwellings, side-by-side duplexes, stacked duplexes, terrace houses, townhouses, manor house apartments, and low-rise apartments. The planning intent is to increase density incrementally within established suburbs without abruptly changing the building scale.


The limits of up-zoning in Established Suburbs

The problem the presentation addressed is what happens when planners apply higher zoning densities to existing residential areas. Up-zoning a suburb on paper produces a significant theoretical increase in ultimate development capacity. In practice, many of those lots will never redevelop. A typical established suburb contains properties where existing houses occupy the full width of the parcel with high site coverage. For a developer to build a duplex on such a lot, they would first need to demolish a structurally sound home on a parcel that may offer limited development yield. In most cases, that development simply will not occur.


This is the core problem that urban growth modelling addresses: identifying which lots will actually develop, and when.


Case Study: Toowoomba Regional Council Gentle Density Scenario

Forecaz applied this approach as a live case study for Toowoomba Regional Council's inner urban consolidation areas. Forecaz used its’ platform and Bayesian Network (BN) modelling methodology. The TRC Gentle Density scenario used historical aerial imagery to map development eras across the Toowoomba Urban Extent, identifying pre-1946, pre-1971, and pre-1994 housing areas. Scenario planning zones were created across these areas: Low-Density Residential for inner pre-1994 suburbs and Medium-Density Residential zones for inner pre-1994 and pre-1946 core areas.


Density Parameters and Development Criteria

Density parameters were then assigned to each zone. Low Density Residential lots qualified for two detached dwellings on parcels with 20% or less site coverage and at least 800 square metres of developable area. Low-Medium Density lots qualified for three or more units on parcels with 30% or less site coverage and at least 800 square metres. Large Low Density lots with 30% or less site coverage and at least 1,200 square metres qualified for four or more units.


Bayesian Network Model Assessment

The BN model then assessed each parcel against a set of weighted development factors. Parcels needed low site coverage between 20% and 30%, good street frontage of at least 18 metres, and a developable area above 800 square metres. Corner lots and multi-frontage parcels were also favoured. Parcels already subject to a Development Application were excluded. These factors were classified and combined into a seven-level development potential score that fed each property's development propensity score within the model.


Scenario Results and Housing Supply Outcomes

The scenario results showed a well-distributed pattern of Gentle Density/Missing Middle development across the projection cohorts from 2026 to 2041, with very limited development options remaining at 2046. The model confirmed these density settings would support 15 to 20 years of infill housing supply. The dwelling types delivered showed an even distribution between two detached dwellings (splitter or auxiliary configurations) and multiple attached dwellings, producing a natural mix across the suburb without concentrations or abrupt density changes.


Comparison with ShapingSEQ Targets

When the TRC model results were compared to the ShapingSEQ diversity targets, the model exceeded the regional plan's expectations. The SEQ Regional Plan projected a 5% increase in low-rise attached dwellings relative to detached by 2046. The TRC Gentle Density model achieved a 7% increase in low-rise attached to detached over the same period, moving from 27% to 34% of total dwelling stock in low-rise attached form.


Key Conclusion from the presentation

The core conclusion from the presentation was direct: zoning sets the ceiling for what development is permitted. Urban growth modelling establishes what development is actually feasible. The gap between those two numbers is where planning decisions need to be made with care. Applied at the parcel level, AI-driven propensity modelling gives planners a spatially grounded, evidence-based view of how much Gentle Density or Missing Middle housing their zoning decisions will realistically deliver and over what timeframe.


Bradley noted he attended the Congress to deepen Forecaz's understanding of different planning approaches and to bring those insights back into the development of the Forecaz platform's modelling capabilities.



See how Forecaz applies AI to test your council's Gentle Density/Missing Middle zoning policy before it is formalised.

Book a discovery session with the Forecaz Team today.




 
 
bottom of page