Content uploaded by Benjamin Thorp
Author content
All content in this area was uploaded by Benjamin Thorp on Aug 13, 2023
Content may be subject to copyright.
Finding good densities:
An urban morphological search for Goldilocks density in Melbourne.
MSD Minor Thesis
Benjamin Thorp
Supervisor: Dr. Elek Pafka
Melbourne School of Design, University of Melbourne
May 2023
Abstract
Density is a deceptively simple quantitative concept - how much of a thing is there in an
area - with major impacts on transportation, economics and environmental performance.
Yet it also has a highly complex qualitative dimension critically influencing streetlife,
place character and diversity.
This study went looking in Melbourne for the missing middle or “Goldilocks” density
(Alter, 2014) . It has then examined four case study neighbourhoods and assessed
them against criteria for “livable density” developed by Sim, (2019). Sim’s framework
was critically evaluated and adapted to remove overlapping criteria and to make it more
suited to the urban morphology existing in the Australian context.
The research found very little medium density urban fabric. There are distinct clusters
of medium density but the predominant pattern of density is high-rise at activity centres
and urban renewal sites rather than neighbourhoods with an even spread of medium
density. The case studies selected however did rate well overall on the framework for
liveable density. All cases are relatively low density by world standards featuring
population densities of around 90 people per hectare and floor area ratios less than 1.0.
Goldilocks density does exist in Melbourne but it is in a low-rise, lower-medium density
form. The implication here for planners and the community is that this type of lower
density will need to be spread over a larger area necessitating change in more suburbs.
The case studies in this research demonstrate that densities of 50 dwellings a hectare
can produce highly livable environments in the Australian context but to ensure a quality
built form, planning needs to adopt more nuanced density measures. These measures
should include floor area ratios, site coverage, open space ratios and measures of
interface quality.
Word Count: 10,324
Contents
1. Research aims and research questions ................................................................... 2
1.1 Research aims ...................................................................................................... 2
1.2 Research questions ............................................................................................... 3
2. Literature review ......................................................................................................... 3
2.1 Density debates ..................................................................................................... 3
2.2 Melbourne context ................................................................................................. 6
2.3 Density in planning theory - historical overview .................................................... 7
2.4 Urban morphology ................................................................................................. 9
2.5 Quantitative aspects - density measures ............................................................ 10
2.6 Qualitative aspects of density .............................................................................. 11
3. Methodology: ............................................................................................................ 14
3.1 Approach ............................................................................................................. 14
3.2 Process ............................................................................................................... 15
4. Findings ..................................................................................................................... 23
4. 1 Case selection .................................................................................................... 23
4. 2 Case study observations .................................................................................... 28
4.3 Quality assessment ............................................................................................. 32
4.4 Measures ............................................................................................................. 36
5. Discussion ................................................................................................................. 40
5.1 RQ1 - Case selection .......................................................................................... 40
5.2 RQ2 - Operationalising method ........................................................................... 41
5.3 RQ3 - Relationships ............................................................................................ 43
6. Conclusion ................................................................................................................ 45
6.1 Limitations ........................................................................................................... 45
6.2 Implications for policy .......................................................................................... 46
6.3 Further research .................................................................................................. 47
Figures and tables ..................................................................................................................... 48
Reference list ................................................................................................................ 49
Appendix I : Process for producing population grid
Appendix II : Process for producing FAR grid
Appendix III : Short list neighbourhoods
Appendix IV : Synthesis of Sim’s criteria
Appendix V : Case Studies - Visual data sheets
1
1. Research aims and research questions
1.1 Research aims
The aim of this research is to explore the outcomes of different densities in a particular
urban context. The research addresses the problem of how to densify Australian cities
while maintaining a high level of liveability, a complex social policy challenge with no
absolute solution, a ‘wicked problem’ (Rittel and Webber, 1973). Suggested solutions
often involve an awareness of a need to find a “missing middle” (Mares, 2018) but this is
difficult to determine and overseas precedents are often used that do not reflect the
local context (policy, cultural, legal, historical).
The concept of liveability is associated with a range of outcomes including - walkability,
public transport, public open space, housing affordability and employment, (Arundel et
al, 2017). Some of these can clearly be related to urban form but others are
independent social constructs that may be beyond the scope of contemporary planning
or design practice to influence. This research focuses on urban form - built form, public
realm and urban structure and is positioned within the field of urban morphology.
What this research seeks is greater densities - where density is enough to generate
liveable density where urbanity and the qualities of a dense urban environment are
valuable in themselves. These higher densities are needed for two key reasons:
• To house projected population without further expansion of the urban growth
boundary.
• To provide diversity and choice of living environment with an option that is between
suburban and hyper-dense living.
2
1.2 Research questions
This research seeks to explore the overall question of: which densifying residential
areas of Melbourne exhibit an urban morphology that produces a higher quality, more
liveable, ‘goldilocks’ medium density?
The following research questions will be pursued:
● RQ1: Where in Melbourne can we expect to find medium density housing?
● RQ2: Can a method be operationalised to assess the qualities of medium
density at the neighbourhood scale?
● RQ3: What relationships are observable between qualitative outcomes in
different neighbourhoods?
2. Literature review
2.1 Density debates
Density has come to occupy a prominent place in urban planning discourses in
Australian (Taylor et al, 2017). This prominence in the discourse is often due to a
focus on perceived negative outcomes of higher density yet in a global density context,
Australia has some of the lowest urban densities. The most dense (large cities) are
the Indian, subcontinent countries (Bangladesh, Pakistan and India) followed by other
members of the ‘global south’ in Africa and South America. China is close to the world
average as are Japan and Europe before the Anglosphere (Australia, Canada, United
States, United Kingdom) which are considerably below the average (Demographia,
2015). Even within this group, Australia is notable for its lower densities, Spencer et al
(2015) found less medium density and more lower density areas when comparing
Australian to British and Canadian cities.
3
For the Anglosphere countries, the lower densities reflect 20th century development
based on a low-density suburban model of detached, low-rise housing, car
transportation and separation of land uses (Taylor et al, 2017). This cultural
preference for suburban housing led to attempts to stem the spread of apartments in the
inter-war period such as the introduction of ‘single family’ zoning in the US and similar
attempts to regulate out flats in some Australian jurisdictions (Taylor et al, 2017). In the
post-WWII period suburban expansion rapidly accelerated in Australia. From the 1960s
and the introduction of the Strata Title acts there was some increase in apartments, with
the appearance of the suburban “six-pack” blocks (Taylor et al, 2017).
Starting in the late 1960s, there emerged a viewpoint in urban planning that saw less
value and more cost in low-density expansion (Taylor et al, 2017). These views came to
be expressed in policy through what can be broadly termed ‘compact city’ or ‘smart
growth’ policy. A leading voice in this view were Newman and Kenworthy (1989) who
drew links between sprawling cities, automobile use and energy consumption and
pollution. Planning policy changed to reflect this now dominant view with a broad shift
in Australian planning policy to encourage urban consolidation (Taylor et al, 2017).
Compact city policy has been criticised strongly by scholars with alternate views. Mees
(2009) has questioned the accuracy of density measures used to draw correlations with
energy use and has advocated for improved public transportation for suburbia rather
than increased densification. Likewise Neuman (2005) has criticised the empirical
evidence for the sustainability claims of compact city advocates and suggested that
urban form alone cannot make cities sustainable. Scholars such as McFarlane (2021)
have concentrated on the ‘knowledge politics’ of urban densification and how claims of
environmental sustainability can hide socio-economic changes such as gentrification
and displacement of low income groups. McFarlane advocates for a questioning of the
dominant density narrative and the emergence of an alternative knowledge base around
urban density. Gren et al (2018) reviewed scientific literature on the environmental
effects of compact city or ‘smart growth’ policy and found inconclusive evidence for the
4
effectiveness of smart growth with 34% of studies finding negative effects and an
inconsistency in the way density is measured.
Berghauser Pont and Haupt, 2021 conducted a literature review of 229 papers
examining the outcomes of urban density across economics, health, ecology and
transport. While most papers reviewed use simplistic density measures, some trends
come through the literature. The following were found to be similar views:
- Energy use - higher density is associated with lower energy use but socio-economic
status can be a factor.
- Transport - higher density contributes to increased public transport use and increased
active transport.
- Economics - Density aids agglomeration economics lifting productivity.
- Environment - Within an urban area density negatively impacts the environment but
generally reduces per capita greenhouse gas emissions thus having a positive global
impact.
For each of these views there were found to be papers that offered an alternative
viewpoint as well. Taylor et al (2017) found similar rationales for densification around
reducing environmental impact, sustainable transport, providing well located affordable
housing, sociability as well as quality and liveability. They also found strong objecting
views for each such as that expressed by Mees in regard to density and sustainable
transport.
In this research the questions around the impact of density on the environment,
sustainable transport, housing affordability and economics will be put to the side with
the scope of the research being on the quality of urban form and livability.
5
2.2 Melbourne context
Melbourne/Naarm is a city of close to 5 million people (Australian Bureau of Statistics,
2021a) occupying the traditional lands of the Wurundjeri people of the Kulin nation. The
colonial city was established in the 19th century, growing outward from the central
‘Hoddle Grid’ on the framework of an extensive tram and rail system that drove the
pattern of urban development up until World War II. From then car-based development
has dominated and the city has relatively low densities (Dovey and Woodcock, 2014).
Melbourne is a sprawling city with some higher density inner suburbs, a very large lower
density suburban area and some semi-urban edges. This is a common pattern of
density in Australian cities according to Spencer et al (2015), their comparison of
densities found Melbourne in 2011 had 55% of its population living at densities in the
range 4-30 people per hectare (p/ha) and 39% at 30-60. They found only 1% of the
population live at densities above 100 p/ha.
The desire for greater compaction since the late 1970s has been manifested in
significant policy changes as well as considerable academic research. Rowley (2017)
identifies the emergence of activity centre policy and a desire to constrain outward
growth in this period leading to the first version of Plan Melbourne in 2002, Melbourne
2030. This plan introduced an urban growth boundary (UGB) and focused growth at
‘activity centres’ nodes of commercial and residential development. Subsequent
iterations of this metropolitan plan have re-established a desired growth pattern to
constrain outward growth while allowing higher density infill along transport corridors
and at activity centres in commercial zones (DELWP, 2017).
Recent versions of Plan Melbourne have introduced an added layer with the “20 minute
neighbourhoods” concept of neighbourhoods across residential zones that can support
local living (DELWP, 2017). This policy is about having a minimum accessibility to jobs,
recreation and services within a walkable catchment or walkable to public transport that
provides that access (DELWP, 2022). The objective of the policy is that all of Melbourne
should have this minimum of “20-minute neighbourhood” similar to a minimum wage.
6
As yet “20 minute neighbourhoods” lacks detailed policy implementation including
thresholds of density or mixed uses that are required for its operation or enforcement of
those through the planning system (Rowley, 2017, p.194). Morley and Pafka (2023)
found little consistency amongst planning professionals in what densities are desirable
to meet 20-minute city policy goals.
Recent projections forecast that Melbourne’s population will grow to reach 8.5 million in
2056 (Australian Centre for Population, 2020). The challenges of accommodating this
population in line with the policy described above are considerable. Several key
studies have explored how densification could occur in this context to meet the
requirements of this population increase. City of Melbourne (2010) examined the
potential of growth along tram corridors with medium density typology creating what the
key author Rob Adams has elsewhere referred to as a ‘linear Barcelona’. Dovey and
Woodcock (2014) looked at the potential of activity centres modelling various built form
configurations that explored both the development capacity and urban design
outcomes. Examining the potential of the suburban fabric, Newton et al (2022) has
explored the potential of densification within residential zones in what he calls the
‘greyfield’ suburbs of the middle ring. Newton’s concept of “greening the greyfields”
involves using lot amalgamation and building within the general 3 storey constraint of
the residential zones to create a denser and more environmentally sustainable urban
fabric (Newton et al, 2022).
2.3 Density in planning theory - historical overview
The concept of density has concerned urban scholars since the birth of urbanism in the
late 19th century. Cerdá, the originator of the term 'urbanisation' in the 1850s and the
designer of the Barcelona expansion grid sought to address a range of urban issues
including health using regulation of maximum densities as a key lever (Soria y Puig,
1999). Similarly, in the early 20th century the Garden City movement proposed even
lower densities as a way of managing overcrowding and improving health outcomes in
the industrial city (Berghauser Pont and Haupt, 2021).
7
A critical change was the work of Jacobs (1961) who identified the “need for
concentration” as one of four preconditions for streetlife vitality. Jacobs' work marked a
change in approach where minimum densities were desired rather than a maximum
(Berghauser Pont and Haupt, 2021). These approaches to urban density proscription
are summarised in the chart below from Pafka (2022). It demonstrates a very broad
range of density targets at multiple scales, some setting maximum caps on density,
others setting only a minimum. Ideas of preferred density have evolved through the
development of urbanism and continue to do so.
Figure 1. Density
targets in urban
theory: Source: Pafka
(2022).
8
2.4 Urban morphology
Urban morphology can be defined as the study of the urban form and processes that
transform it including the structure of blocks, lots, streets and buildings that constitute
urban form. Urban morphology explores the interrelationships between these elements
and their dynamic interactions in a process of “urban morphogenesis” (Mouden, 1997).
With its focus on physical/material properties of cities, urban morphology is a key
approach through which this thesis engages with the research questions. Central to the
study of urban morphology is the work of Jacobs (1961) who proposed an
understanding based on generators of diversity. Her four conditions for streetlife -
mixed uses, short blocks, concentration (density) and aged buildings - describe the
fundamental aspects of urban morphology.
This thesis will take urban morphology as an approach to analysing urban density with
the intention that findings from the research can feed back into the practice of urban
design. Marshall and Caliskan (2011) describe the relationship between urban
morphology and urban design. They construct a joint framework for urban morphology
and design that explores their relationship with each other and the physical world.
According to Marshall, it is the “incremental, emergent nature of the urban fabric”
(p.420) that makes urban design different from architectural design and thus requires an
analytical approach through urban morphology that is more rigorous than a simple
examination of precedent.
There are a range of approaches to the study of urban morphology that have been
influential on questions of urban density. While there are several traditions of urban
morphological study - French, Italian, British - (Kristjansdottir, 2019) this thesis will
largely utilise sources from the British and Dutch schools. A key reference point here is
Spacematrix (2021) by Berghauser Pont and Haupt. These authors use density as a
key lens to understand urban morphology using calculation and diagrammatic
representation to bridge the gap between measurements used by planners and
designers and physical form.
9
2.5 Quantitative aspects - density measures
A range of approaches have been taken over time towards the measurement of urban
density. The most common approach used is to calculate the number of either
dwellings or people per hectare. The commonly used approach of dwellings per
hectare has been critiqued on several fronts including that it does not take into account
factors such as building size, internal or external populations (Pafka, 2013). Other
densities that can be calculated include floor area or plot ratio, open space, ground
coverage and street network density (Berghauser Pont and Haupt, 2021).
A critical issue occurs with the scale at which density is calculated. Net densities
include just private parcels whereas gross densities can include large areas of public
spaces and unbuilt space such as roads or waterways depending on the size of the
area chosen; however, the definition can vary (Churchman, 1999). Berghauser Pont
and Haupt (2021) attempt to capture the properties of different urban morphologies at a
range of scales. The authors take a multi-scale approach using three main scales -
cadastral lot, “island” (which is a block) and “fabric” which is a consistent section of
urban fabric composed of multiple blocks. The last is effectively a “gross” measurement
whereas the first two are a net measure. Their method is dependent however on
consistent sections of urban fabric rather than heterogenous fabrics. Others have used
regularised grids to average density across statistical areas, an early example in
Melbourne being Fooks (1946). The difficulty of comparing differing reference areas at
different scales remains vexed.
Dovey and Pafka (2014) create a model of urban density that connects multiple density
variables. A key distinction is drawn between density that is measurable (as listed
above) and the experience of ‘intensity’, a perceived character resulting from the degree
of density and other factors. Using a range of measures such as coverage, building
height, floor area ratio (FAR), dwelling density, external density, internal density and
open space ratio it is possible to see relationships between these variables across a
10
range of urban morphologies. Combined, these measures are indicators of a particular
‘density assemblage” (p.10) where individual densities are mediated by factors such as
use, height or interface conditions. They warn against a reductionist approach of using
any single measure in isolation and advocate for an understanding of the complex
nature of urban density and greater focus on its relationship to intensity.
Berghauser Pont and Haupt (2021) also take a multiple measure approach but focus on
two key measures - Floor Space Index (the same as FAR) and ground space index
(GSI) constructing a chart they call the ‘Spacemate’ that captures these key properties
and by inference, height and proportion of open space to floor area. Open space to
floor area they call Open Space Ratio (OSR).
2.6 Qualitative aspects of density
The discussion above highlights the complex nature of density measures. Also of great
complexity is the range of urban qualities that result from certain densities. A number of
authors have sought to engage with this qualitative question, sometimes also relating
the qualitative dimension back to a quantitative measure.
Uytenhaak (2008), focuses on the quality of internal and external spaces created
through differing density prescriptions and the resulting built form typologies.
Berghauser Pont and Haupt (2021) examine five key “performances” - parking, daylight
access, noise pollution, air quality and urbanity - to settings within their Spacemate.
They find a strong relationship between daylight access and OSR.
11
Others have focused more broadly on qualitative outcomes in terms of ‘liveable
density’. A recent work on this topic is Soft City by David Sim (2019), a long-time
collaborator of urbanist Jan Gehl. In this work Sim advocates strongly for a form of
medium density urbanism in the form of the perimeter block typology, that is common in
continental Europe and South America. Sim identifies 9 criteria for liveable density:
1. Diversity of built form
2. Diversity of outdoor spaces
3. Flexibility
4. Human scale
5. Walkability
6. Sense of control and identity
7. A pleasant microclimate
8. Smaller carbon footprint
9. Greater biodiversity.
On these criteria the perimeter block typology is argued to offer a highly liveable form of
urban density - its courtyards, many entrances and accessways, diversity of forms and
uses and medium scale contribute to make a diverse, human scaled, comfortable and
‘soft’ urban environment.
Sim’s criteria are a useful starting point for an assessment of liveable density. However,
his highly illustrated, evocative text presents a somewhat idealistic version of urban life
and does not take into account some of the inherent conflicts within complex urban
environments (Pafka, 2019). There is also an issue of scale in application of the
criteria. While the first seven can be seen as reflective of best practice urban qualities
at the neighbourhood scale, the last two - carbon footprint and contribution to
biodiversity - have a global dimension inherent in them.
Eberle and Troger (2014) look specifically at density and its relationship to the
perceptual quality of “atmosphere”. They define atmosphere as “sensory mood or
12
ambience of a location or space” (p.36). This sense is, they say, related to the material
density but is not dependent on it. Through their study they explore European
neighbourhoods from Berlin, Munich, Vienna and Zurich across a range of 9 density
categories based on gross FAR. A range of other measures are also used including site
occupancy, ratio of public to private space and building height. For each case study
images are shown of street level views and orthogonal views and the quantitative
measures are mapped. For each of the 9 categories, the relationship between the
quantitatively measured factors and the observable atmosphere is discussed. Their
conclusion identifies that the relationships are determined by a complex range of social
factors and how they interact with the spatial layout and built form of the case study,
related to the key density factor of FAR.
Jacobs (1961) focussed on the qualities of density in terms of outcomes for streetlife
and walkability. Jacobs posited density as providing a sufficient concentration of people
for a vibrant city life that avoided the problems of low density ‘gray’ areas and
“semi-suburbs’ (pp.208-210) but also avoided too high density that required built form to
be constructed too efficiently, reducing diversity. The latter phenomenon is observable
within many redevelopment areas in Melbourne. Dovey and Symons (2013) have found
that where high-rise building typologies with high densities occur in Melbourne’s
Southbank hinterland they do not result in vibrant, intense, streetlife but in “density
without intensity”. A lack of spatial diversity and permeability in building interfaces
creating a dull and lifeless street environment.
13
3. Methodology:
3.1 Approach
The proposed method takes an urban morphological research approach utilising the
material properties of the built form of the city as the ‘dataset’.
This research is situated within the field of urban design, as such the epistemological
approach is one that matches research approaches in this field. The methodological
approach will be using metrics in combination with morphological mapping to explore
the qualities of urban fabrics. Using an observational approach, research findings are
obtained inductively from the dataset and these form the basis for further discussion.
Analysis will be through comparison and examination of relations between
measurements and observations to infer implications, through inductive rather than
deductive reasoning (Bryman, 2012).
The research will use two critical frameworks to undertake the investigation. Firstly, the
multiple measure and multi-scalar approaches to quantitative measurement of density
developed by Pafka (2022) will be used to search for ‘goldilocks’ density within the
Melbourne metropolitan area. Secondly, the ‘liveable density’ framework developed by
Sim in Soft City (2019) will be operationalised to test the qualities of the discovered
density case studies.
The following section describes the process through which a series of case studies will
be examined to acquire two key datasets - quantitative measures of urban density and
qualitative measures of resulting urban design outcomes. These will then be compared
and contrasted through a comparison matrix.
14
3.2 Process
The research has proceeded in three steps: (1) the selection of case study areas, the
searching from the metropolitan scale down to the neighbourhood scale to discover
examples of liveable density. (2) the analysis of case studies including density
measures as well as qualitative assessment. (3) comparative assessment of the case
studies utilising the ‘soft density’ framework of Sim (2019). The following is a detailed
description of each of these.
Step 1: Case selection
An initial examination of Census data from 2021 (Australian Bureau of Statistics,
2021b), at the mesh block level was used to position a 20kmX20km square in GIS over
the key areas of population density. As discussed in the literature review, differing
reference areas can introduce bias in comparisons, to remove the spatial bias of census
geographies a 1 ha area grid (100 m by 100m) was created and mesh block data
averaged to this grid. The detailed technical explanation of this process is attached as
Appendix I. This grid data was then used to calculate population densities per ha at the
neighbourhood scale of a 500m by 500m grid (25 ha). The 1 ha grid process was also
used to create a map of density by FAR for comparison purposes. The detailed
technical process for this can be found at Appendix II
15
Figure 2: Population density data grids:
20km X 20 km square 1 Ha grid (top) and 25 ha grid (bottom).
16
These neighbourhoods were then filtered in the following sequence:
1. Remove neighbourhoods with less than 50 people per hectare (p/ha) to create a
‘long list’. The list was 227 neighbourhoods.
2. Remove neighbourhoods that are outside walkable access to PT network (800m
buffer train, 400m buffer tram).
Figure 3: Filtering sequence 1 and 2.
Filter 1 (left) and Filter 2 (right, red squares removed).
3. Comparison of a long list to FAR map suggested many of these were not of
significant density so the threshold was increased to 75 p/ha.
4. Remove neighbourhoods with 50% of dwellings in high-rise (above 8 storeys).
Census data at the SA1 level (Australian Bureau of Statistics, 2021b) GIS.
Figure 4: Filtering sequence 1 and 2.
Filter 3 (left) and Filter 4 (right, red squares removed).
17
This resulted in a short list of 37 neighbourhoods. The dataset for these
neighbourhoods can be viewed at Appendix III. A further filtering process was applied:
5. Remove neighbourhoods with significant influence of taller buildings.
Threshold reduced to 25% of dwellings in high-rise (above 8 storeys). Census
data for dwelling type at the SA1 level used (ABS, 2021b).
6. Remove neighbourhoods that are isolated from other dense neighbourhoods,
not being adjacent to another dense neighbourhood.
Figure 5: Filtering sequence 5 and 6.
Filter 5 (left) and Filter 6 (right), red squares removed in both.
This left 5 medium density clusters: East Brunswick, North Melbourne, Fitzroy,
Richmond and St Kilda.
Figure 6: Filtering final cluster
18
The fieldwork pilot was conducted in East Brunswick. East Brunswick was not
advanced as a case as the pilot showed the neighbourhood was mostly suburban with
just a small section of higher density.
Step 2: measuring
Within each cluster the most morphologically typical neighbourhood was selected. In
some cases the square was adjusted to clip out non-residential or high-rise areas. The
following measures of built form and urban structure were made within each case study
neighbourhood
Measure
Abbreviation
Formula
People per hectare
p/ha
total dwellings in subject area/area in hectares
Dwellings per hectare
dw/ha
total dwellings in subject area/area in hectares
Floor Area Ratio
FAR
gross floor area/subject area m
2
Site Coverage
SC
total area/building footprint area
Open Space Ratio
OSR
unbuilt area/gross floor area
Gross Floor Area per person
m
2
/p
gross floor area/number of people
Average Height
AveH
mean of heights in subject area
Average lot size
AveL
mean of cadastral lot area
Percentage Public Realm
PR%
total non-lot area/total area
Table 1: Density measures and formula.
Population and dwelling data (p/h and dw/h) were computed from Census data
(Australian Bureau of Statistics, 2021b). Remaining measures were calculated from
GIS data available through Data Vic (2023). A field work visit was made to each case
study during March/April 2023. The area was walked and observations made. A
selection of images from Google Streetview and NearMap was used to supplement field
observations.
19
This included mapping of spaces and interfaces. Thematic analysis of Sim’s (2019)
criteria, identified a number of overlapping built form elements - interfaces, spaces,
permeability, division between public and private. These were mapped at typical areas
within the neighbourhoods using a unique scale developed for the project. An example
can be seen below for North Melbourne at figure 7. Maps were made at two 200m by
200m typical areas for ‘interfaces’ and ‘spaces’. The ‘spaces’ map recorded public and
private open space, hybrid public/private spaces (Carmona, 2010) and levels of
enclosure. The interfaces map recorded levels of facade permeability and transparency
(Dovey and Wood, 2015) as well as potential occupation of edge zones.
Figure 7: Case study mapping examples
20
Step 3: assessing
To conduct an urban quality assessment an appropriate theoretical framework was
required. With the focus of the research being on urban morphology and built form, the
framework developed by Sim (2019) for assessing ‘liveable density’ of urban fabrics
was considered the best fit for the research aims. Sim has 9 criteria but only 7 are used
here. Criteria 8 “smaller carbon footprint” and criteria 9 “greater biodiversity” have been
excluded as they are “global environmental factors” (Pafka, 2020) not easily accessible
at the scale of a neighbourhood. For each criteria Sim has 7-8 sub criteria of “what to
look for”. To allow the framework to be operationalised as an assessment method,
Sim’s sub criteria were thematically analysed to look for overlaps and synergies. These
are represented in the diagram below:
Figure 8: Thematic analysis of Sim criteria.
21
These overlapping categories were used to further refine Sim’s 7-8 sub-criteria into a
2-4 synthesised sub-criteria per main criteria represented below in figure 8. The
analysis can be seen at Appendix IV.
Main Criteria
Synthesised Sub Criteria
How assessed
1. Diversity of built
form
Type
Perspective view/census data
Small grain
Map/data
Façade variation
Streetview image/in person observation
2. Diversity of
outdoor space
Public
Spaces/interfaces map
Private and Communal
Spaces Map
Flexible and hybrid
Spaces/interfaces map
3. Flexibility
Multi-purpose Spaces
Spaces Map
Flexible interface b/w public and private
In-person ob
Backs with room for growth
Spaces Map
4. Human Scale
Small components
Streetview image/in person observation
Eye level detail and multi-sensory stimulus
Streetview image/in person observation
5. Walkability
Permeable building interface
Interfaces map
Connection to ground to ground plane
Built Form Map
6. Sense of control
and identity
Enclosed and defined spaces
Spaces Map
Identifiable interface between public and
private
Interfaces map
Small units and subdivisions
Streetview image/in person observation
Architectural expression especially on
corners
Streetview image/in person observation
7. A pleasant
microclimate
Consistency of greenery
Streetview image/in person observation
Solar gain and shade
Perspective view
indoor/outdoor spaces
Spaces Map
Table 2: Sim criteria synthesised and assessment method.
22
This set of sub-criteria were rated based on the visual and mapped data acquired during
fieldwork (see “how assessed” column above). The rating was made with a simple
three point scale - good, average, poor by the researcher who has worked
professionally as an urban designer for 5 years.
4. Findings
4. 1 Case selection
The case study selection process has highlighted the difficulty in locating a medium
density fabric within the Melbourne metropolitan area at the neighbourhood scale. The
pattern of development observed has resulted in higher buildings above 8 stories in
clusters amongst suburban housing or more-dense high-rise fabrics. Neighbourhoods
influenced by these higher buildings are excluded from this study, thus restricting the
dense areas that have met the criteria. This includes large areas such as the CBD,
Southbank, Docklands and the St Kilda Road strip (shown as red in figure 8 below).
However, it also includes parts of South Yarra, Fitzroy, Carlton, Richmond, Kensington
and Abbotsford where there are social housing towers and/or large recent
developments (orange in figure 8 below).
Where medium density (50-100 p/ha) does occur, it is observable in this research that it
likely occurs along tram routes or at activity centres, forming either linearly along tram
corridors or in small clusters at transit nodes or redevelopment sites. This pattern is
observable in figure 9 below. Long list neighbourhoods with greater than 50 p/ha are
shown overlaid as squares. Only rarely is there a consistent pattern of density across a
large area where many dense neighbourhoods have clustered together. The largest
cluster of medium density observable in figure 9 is between St Kilda south to Elwood
and east to Caulfield. In this area there is a spread of neighbourhoods with greater than
50 people p/ha and clusters of greater density again above 75 p/ha such as the area
that has been chosen at St Kilda as a case study. There are surprising clusters of
emerging density in the west and north that
23
indicate higher density is spreading from the traditional inner city areas. These are
shown in the figure below as: 364 Essendon,132 Maribyrnong and 607/608 West
Brunswick (numbering references GIS database).
Figure 9: Showing ‘long ‘list’ overlaid on 1 ha population density grid map .
For comparison a FAR map was also produced to analyse the building density across
the study area. This is likely the first grid based FAR map made of Melbourne. This
can be seen below as figure 10. This map follows a very similar pattern to the
population per hectare map however there are nuanced differences. The density is
more in areas
24
that include a large mixed use component that contributes to the FAR but not the
population density. Density by FAR does follow very similar patterns overall with a
spread along linear corridors from the centre. Surprising clusters occur further from the
centre where there are large industrial areas or educational institutions that include
large floor area buildings. For predominantly residential areas the FAR pattern is very
similar to the pattern of residential density.
Figure 10: FAR data grids 1 ha, 20km X 20km square.
25
Figure 11: Dwellings p/ha plotted against population p/ha for ‘long list’ .
The chart at figure 11 highlights the extreme nature of the density pattern in Melbourne.
It shows population density for the 227 ‘long list’ neighbourhoods above 50 p/ha, most
are between 50 and 100 p/ha but there are a number of neighbourhoods with much
higher densities. The neighbourhoods at right end of the scale are the central city and
surrounds where there are a significant number of high-rise residential towers. The four
selected case studies are shown in red.
Multi-scalar mapping is inherent in the method of searching for case study
neighbourhoods used in this study through mapping at the 1 ha (100m by 100m) and
25 ha (500 m by 500m) scale and metropolitan scale. Through the data generated it is
also possible to explore how the density alters across multiple scales from the block, to
the neighbourhood, to the district (defined here as a 2km square), to the metropolitan
26
area (defined here as 20 km by 20 km). These can be seen plotted on the chart below
at figure 12. Richmond and St Kilda show the most consistency (straightest line)
relative to the other cases, North Melbourne and Fitzroy. This likely reflects that the
former cases are located deep within areas of relatively consistent density whereas the
latter are close to either much higher density (CBD) or much lower density (parks).
Figure 12: Multi-scalar density of the cases, p/ha.
27
4. 2 Case study observations
The following paragraphs are a summary of these observations for each
neighbourhood. Full visual data sheets from fieldwork can be seen at Appendix V.
North Melbourne
This area exhibits a very high built form diversity and there are two distinct fabrics. The
first fabric is the modernist walk-up apartment blocks around O’Shannessey Street with
courtyards. The second fabric is the tighter and denser urban blocks where
contemporary lift apartments and townhouses have infilled around historic terraces and
cottages and down laneways in some cases. There are some advantages and
disadvantages to both. The streets are a mix of very wide tree lined streets with
medians in combination with very narrow laneways. These uniquely have building
frontages to them making habitable an otherwise access space. A varied access to
open space and varied microclimate are the key weaknesses of this case. In some
cases there are open spaces around buildings but in others the buildings cover 100% of
the lot area. Microclimate varies as the extent of tree planting in the streets and within
lots varies across the case study with extensive tree canopy in the north-west and less
in the south.
Figure 13: North Melbourne, oblique view Near Map (2023).
28
Fitzroy
This area of Fitzroy is an historic area with rows of 1 and 2 storey terraces. Larger grain
lots where this pattern has been broken have been filled with much larger medium
density lift apartments, many with high architectural quality. There is some difference in
the pattern of terraces, 1 storey in the south-west corner and 2 storey in the north-east.
The area is visually diverse and spatially varied (such as street to lane widths). Building
types have some inbuilt flexibility. The boundary between public and private is well
defined but often that can mean blank walls and buildings turning their back on the
street. A high level of security/blank interface at street level is an especially poor
outcome of many of the larger infill buildings. There are many small porches and
balconies but many are too small to support activity and show less signs of use. Open
space is lacking and greenery is inconsistent meaning a poorer microclimate and less
open space diversity.
Figure 14: Fitzroy, oblique view Near Map (2023).
29
Richmond
This is the most fine grain case study with the exception of one block which has much
larger lift apartments. This block lifts the case study into a higher density threshold
although the area still has a reasonable level of density even without this block. There
is very high built form diversity and variety in edge zones and entrance arrangements
across this case study. The streets are high quality and walkable with consistent visual
interest and greenery. Although there is one significant open space being Dame Nellie
Melba Memorial Park the case study is generally lacking public open space diversity or
shared, communal spaces.
Figure 15: Fitzroy, oblique view Near Map (2023) .
St Kilda
This is an area with predominant two and three storey walk-up apartments from the
early to mid 20th century. The forms and designs are varied creating a rich visual
experience. Walkability is encouraged with a well connected neighbourhood and
walk-up buildings. There is a strong definition between public and private space with
30
each apartment building a separated entity with secure fencing and locked gates.
There is less diversity of outdoor spaces and flexibility here though with spaces highly
controlled with little room for hybrid public/private spaces. There is opportunity for
communal space around apartments but this does not appear to often be utilised as
such. A lack of balconies and sheltered outdoor spaces reduces outdoor microclimate
comfort. Greenery is rich where possible including in the streets. Grey Street is more
car dominated but side streets are quiet. A predominance of walk-up apartment type
with very few terraces/townhouses or single detached dwellings makes this case study
distinct from the other cases .
Figure 16: St Kilda, oblique view Near Map (2023) .
31
4.3 Quality assessment
An assessment was undertaken using the adapted version of Sim’s (2019) criteria for
liveable density as described above in the method section. The results of this
assessment are in the table below:
Main Criteria
Synthesised Criteria
North Melb
Fitzroy
Richmond
St Kilda
1. Diversity of built
form
Type
Small grain
Façade variation
2. Diversity of
outdoor space
Public Realm
Private and Communal
Flexible/hybrid
3. Flexibility
Multi-purpose Spaces
Flexible interface b/w public and private
Backs with room for growth
4. Human Scale
Small components
Eye level detail and multi-sensory stimulus
5. Walkability
Permeable building interface
Connection to ground plane
6. Sense of Control
and identity
Enclosed and defined spaces
Identifiable interface between public and private
Small units and subdivisions
Architectural expression especially on corners
7. A pleasant
microclimate
Consistency of greenery
Solar gain and shade
Indoor/outdoor spaces
Good
Average
Poor
Table 3: Assessment against adapted David Sim criteria.
32
There are some clear patterns observable in the results presented above. For diversity
of built form all cases performed well and exhibited very diverse typologies the
exception being St Kilda where there is a dominance of a similar apartment type. There
is considerable historical layering within these areas. Where there is dominance of one
era this can present as poorer visual variation at street level. North Melbourne with the
greater spread through historical eras rates best for facade variation. All cases are
lacking though in the 4-6 storey height range advocated by Sim. They tend to have
great diversity up to 3 storeys then jump in height to 7 or 8.
Figure 17: Built form diversity North Melbourne left and Richmond right.
Author's photo.
All cases rate equally poorly on Diversity of Open Space. Most have only one local
park within the 500m by 500m square, however streets are often high quality public
spaces. There is a deficiency in private open spaces as well which are often just very
small yards or balconies. Many in-between spaces around buildings are used for
parking or otherwise not developed as open space (see mapping at Appendix V). The
predominant low-rise, high site coverage fabric means less open space. There is
usually a very clear and distinct difference between public and private space and few
semi-private spaces. Hybrid spaces do exist but they are often small, on the edge of
buildings and often used for parking.
33
Figure 18: A rare example of a courtyard that is semi-public, North Melbourne.
Author's photo.
Most of the opportunity for Flexibility observed occurs in the historical terraces and
townhouses (front and back) and many have been infilled already with small extensions
or have signs of use in front yards or porches. Newer developments of lift apartments
generally lack opportunities for flexibility other than balconies (which are often too small
or oriented poorly for use). Walk-up apartments can have more potential but often this
is currently an undercroft or in-between space used for carparking.
Figure 19: Example of flexible undercroft space used for parking.
Author's photo.
The cases generally match well to the Human Scale. Generally the cases have a lot of
visual interest however, this can be intermittent at eye level as there are large sections
of blank walls and security-focussed design of interfaces. Greenery often adds
34
considerable visual interest where this is not provided by built form variety. As the
cases are predominantly low-rise 1-3 stories they are well scaled to eye-level
experience. This can change in Fitzroy and Richmond though with much larger-grain
infill developments being less human scaled.
Figure 20: Good eye level detail and human scaled elements: North Melbourne.
Author's photo.
All cases perform well on Walkability . This measure captures the walkability of built
form elements rather than the neighbourhood walkability which is not assessed (all
cases being in walkable locations close to public transport and amenities). Having a
large proportion of walk in/up/through buildings, the built form often exhibits permeable
interfaces with many stairways and doors and a close connection to the ground plane.
Fitzroy and Richmond lack walk-up, being mainly walk in/through (terraces) or lift
apartments. The array of door locations in the case study maps shows these
connections (see mapping at Appendix V).
Sense of Control is well maintained in all cases through a well defined public/private
division often defined by fencing and security elements. Expressive architecture on
corners and in most built form gives a clear sense of identity. All lack the layering of
levels of control that Sim (2019) favours, there is little shared communal space. Missing
is the shared garden or courtyard space generally.
35
The cases perform poorly on Pleasant Microclimate . The low rise form with a variety
of roof forms means control of wind is good. However, all cases rate poorly as street
microclimate varies considerably. Some streets have good sun and shade, others are
very exposed or shaded. Greenery is rich in places but non-existent in others - such as
laneways of minor streets. There is little rain protection with awnings not usually
present. Indoor/outdoor possibilities are limited to some extent by low-rise high site
coverage form with less facade facing to the outdoors.
Figure 21: Good microclimate North Melbourne, shade of tree, indoor/outdoor space.
Author's photo.
4.4 Measures
The case studies were measured using a set of density measures as described in the
methodology.
RATIOS PEOPLE
RATIOS BUILT FORM
OPEN SPACE
GRAIN
p/ha
dw/h
p/dwe
m
2
/p
FAR
SC
AveH
OSR
PR %
AveL
Richmond
94
50.4
1.9
66
0.63
0.47
1.3
0.84
21.37%
210
Fitzroy
79
48.7
1.6
109
0.87
0.46
1.9
0.62
33.70%
238
St Kilda
95
64.6
1.5
71
0.69
0.36
1.9
0.93
23.96%
436
Nth Melb
92
53.6
1.7
65
0.60
0.34
1.8
1.10
36.30%
213
p/ha: People per hectare. dw/ha : Dwellings per hectare. p/dw: People per dwelling. m
2
/p : Gross Floor
Area per person. FAR : Floor Area Ratio. SC: Site Coverage. AveH : Average Height. OSR: Open Space
Ratio. PR%: Percentage Public Realm. AveL: Average lot size
Table 4: Summary of density measures - cases.
36
Population density can be seen to be closely related to building type and dwelling
count. There is likely very little cultural or use variance in these cases . Dwelling
density is closely related to residential population with household size ranging from 1.5
to 1.9. People per dwelling is fairly consistent as is GFA per person at about 65-70
m2 with Fitzroy higher for the latter due to floor area of commercial areas influencing the
data.
FAR is consistent at about 0.6-0.7 with Fitzroy higher due to the floor area of mixed
uses. The connection to residential density may be hard to make sometimes due to
mixed uses. Site Cover shows some differences between cases with Fitzroy and
Richmond high and North Melbourne and St Kilda lower. Average Height is very low
averaging below 2 in each case.
There is considerable variation in open space measures with OSR large for North
Melbourne where wide streets and open spaces around buildings exist. North
Melbourne is also highest for Public Realm Percentage .
Grain Size is quite consistent, the exception being St Kilda where larger lots exist with
the dominant apartment typology.
A further interesting finding is that there is a greater or lesser range of performance for
some neighbourhoods across criteria compared to others. The results in table 3 above
show that the most noticeably varied is Fitzroy which within criteria can rate well or
poorly. Richmond is more consistent, North Melbourne and St Kilda are consistent on
some quality measures and not on others. These differences in performances are in
many ways explained by the diversity across the neighbourhoods. This can be
appreciated in the maps below at figure 22.
37
Figure 22: Case studies comparison maps, 500m by 500m neighbourhoods.
While overall density measures may be broadly similar, the pattern these take across
the 500m by 500m area is varied. Fitzroy shows considerable variance - the centre
having a distinct lack of population density due to commercial uses. North Melbourne is
similarly diverse with the top left corner showing much lower site coverage due to
medium-rise apartment blocks with courtyards here. St Kilda by contrast shows the
most uniformity in population density and FAR.
38
The maps at figure 22 demonstrate the critical relationships that different densities have
with each other. Residential population generally follows FAR but can vary when the
density of commercial uses or street width varies (not measured in this study). Site
coverage can be seen to be critical in mediating the outcome of FAR. The site
coverage measure shows great variance in pattern across the neighbourhoods as well
as clear differences between the cases overall - North Melbourne and St Kilda are low,
Fitzroy and Richmond higher. The relationship between site cover and FAR largely
determines the building typology, reflecting whether the built form adopted is of a certain
configuration (Berghauser Pont and Haupt, 2021). These configurations will vary over
time and across cultures but are influenced in similar ways by the relationship between
site coverage and FAR. Below the shortlisted cases are shown plotted on the
“spacemate” chart used by Berghauser Pont and Haupt. It can be seen that the cases
cluster together despite a number of outliers, different combinations of site cover and
FAR as described above determine the position of cases within the broad cluster .
Figure 23: Spacemate, short-list cases.
Case study neighbourhoods shown in red.
39
5. Discussion
5.1 RQ1 - Case selection
● Where in Melbourne can we expect to find medium density housing?
This research has found that Melbourne has a gap in a density range, the often
discussed and much sought after ‘missing middle’ or ‘goldilocks’ density. The gap
manifests itself the most in the range 75 - 100 p/ha. While it is possible to find
‘goldilocks’ density as the case studies show, it is within a limited range in terms of
density level and geographically limited.
Suburban fabrics with population densities of less than 50 people p/ha are most
common across Melbourne. Areas similar to the case studies are found along tram
corridors, adjacent to activity centres and in some inner city historic suburbs. These are
represented in this study by the ‘short list’ cases and exhibit population densities of
75-100 p/ha and gross FAR of 0.5-1.0. There is then a steep jump to higher densities in
neighbourhoods dominated by high-rise towers, with up to 400 people p/ha and FAR of
5 or 6.
Comparing these densities to European cities reinforces the particular range of density
in Melbourne. Eberle and Troger (2014) identify 9 density categories, the first three
have FARs that can also be observed in Melbourne neighbourhoods - “single-family
house Idyll 1” (< 0.4), “single family house Idyll 2” (0.4-0.6) and “Urban Apartments in
Green Areas 1” (0.6-0.9). The last substantially matches the St Kilda case in urban
form as well as FAR. There is then a series of categories not represented in Melbourne
before reaching the final category “Commercial Centres” (> 2.7). Similarly, Berghauser
Pont and Haupt (2021) examine case studies from the Netherlands, Germany, Spain,
United Kingdom and Sweden finding FARs broadly in the range 0.5-2.5.
These authors analysed case studies that are consistent fabrics of one built form
typology and this is another key difference to the Melbourne context. This research has
40
shown that it is rare to find a consistency in built form across a neighbourhood. All the
case studies exhibit built form diversity (terraces, townhouses, apartments) rather than
homogeneity. The heterogeneity is most pronounced in Fitzroy and Richmond where
FAR and population density vary across the neighbourhood from the low point in the
scale through to the highest band. St Kilda is the only case study that exhibits the kind
of homogeneity that is similar to the European examples.
This heterogeneity is also manifest when examining density at multiple scales. The
change in density across the scales from the metropolitan down to the “urban
experiential density” (Pafka, 2022) has a critical effect on walkability across
neighbourhoods as well as the experience or perception of density at street level. North
Melbourne and Fitzroy lack consistency at the multiple scales in contrast to Richmond
and St Kilda. At least in terms of a consistency beyond the immediate neighbourhood, it
is these cases that exhibit multi-scalar ‘goldilocks’ density particularly the large cluster
from St Kilda south to Elwood and East to Caulfield.
5.2 RQ2 - Operationalising method
● Can a method be operationalised to assess the qualities of medium density at the
neighbourhood scale?
This research has shown that it is possible to make an assessment of the quality of
density at the neighbourhood scale that brings out some of the qualitative nuance within
these neighbourhoods. David Sim’s nine criteria of liveable density were
operationalised into a method to assess qualities at the neighbourhood scale. This
involved a synthesis of Sim’s framework, development of mapping and observational
techniques and the operationalising of these in field. The results of this process
demonstrated in figure 9 above show critical and interesting differences between the
tested criteria and across the case study neighbourhoods. This data allows more
careful consideration of the benefits and costs of urban density and provides an
evidence base for qualitative outcomes of certain built forms.
41
The framework used was most successful in identifying poorer performance on the
criteria of diversity of open space and pleasant microclimate . These performances
clearly rated poorly in the assessment generally being a uniform poor performance
across case studies. The criteria of diversity of built form , human scale and
walkability conversely show a generally high performance across case studies. It is
the remaining criteria of flexibility and sense of control and identity where there are
ambiguities that are in part due to the assessment methodology. Across these criteria
performance varies considerably between the cases and within the sub-criteria. To an
extent this is due to the difficulty of assessing the way the built form mediates these
qualities. Both are critically influenced by the built form interface condition and the
programming of non-built space on the edges of buildings. In assessing the quality of
outcome here it is necessary to observe very subtle transitions between public and
private spaces, subtle degrees of permeability across thresholds and temporal
variations in the use of spaces. In depth consideration of these, such as that conducted
by Gehl (1977) in examining some of the same case study neighbourhoods, was
beyond the scope of a project such as this. The best this more high level assessment
can achieve is to apply theory developed by others to a more macro scale.
This research has also shown the importance of building typology as an influence on
urban morphological outcomes in a context where other elements such as lot sizes or
street widths might be relatively uniform. Building typology plays an influential role
across all criteria, not just diversity of built form . For instance the existence of the
terrace house type influences the performance on flexibility since this typology
provides backs with room for growth as well as semi-private front yards that can be a
flexible interface between public and private. Since building typology is so critical, it is
somewhat problematic that Sim’s framework was developed originally for consideration
of the European walk-up perimeter block type not generally found in Australia. As Pafka
(2020) highlights, Sim’s approach somewhat overemphasises the successful qualities of
walk-up perimeter blocks discounting for instance the contributions of lift apartments to
creating a diverse building mix. This research has found no existing walk-up perimeter
blocks within the case studies and it was only through adapting Sim’s sub-criteria
42
considerably that the framework could be operationalised in the Australian context -
for instance by looking for adaptable communal space on the outer edges of buildings
rather than interior courtyards.
Another key limitation of Sim’s framework is its concentration on the building scale in
what presents as an urban assessment framework. Many of the criteria involve
elements that are interior to buildings such as interior stair access. There is also
tendency to miss connections between elements at the neighbourhood scale, for
instance diversity of open spaces is considered important but not how they connect
across the neighbourhood to form a continuous public realm. Sense of control and
identity is carefully considered at the building scale but the overall character of the
neighbourhood such as streetscapes with unique identity is not considered.
5.3 RQ3 - Relationships
● What relationships are observable between qualitative outcomes in different
neighbourhoods?
This research has located what might constitute ‘goldilocks’ density in the Melbourne
context. The case studies chosen as representative of the clusters of medium density
within inner Melbourne meet many of the criteria set by David Sim and present as the
best case examples of medium density living in this context. A key finding of this
research discussed above is that these dense neighbourhoods are found in a relatively
narrow band of density, at the lower end of what would constitute ‘medium’ in an
international context. All four cases could be described as low-rise, high site coverage
contrasting to the mid-rise, low site coverage types described in the European cities
studied by Eberle and Troger (2014) or Berghauser Pont and Haupt (2021).
Typologically the case studies are similar but within this there is enough variance to see
some insights into the underlying urban morphology that has created the observable
qualitative outcome. This does reveal a series of observable relationships between
performances on the criteria.
43
Low rise, high site coverage fabrics result in less outdoor space in the private realm, the
strong performance of all cases on diversity of built form, flexibility, human scale and
walkability is largely due to a low rise, fine grain typology - the tightly packed terraces
and small apartment buildings that characterise the cases create the conditions for
visual interest, a high level of access from the ground plane and low human scaled
forms. But in order to be dense at all this typology needs to cover much of the ground
plane reducing the outdoor space in the private realm. This could be compensated for
in the public realm but in these cases the public realm is dominated by mobility being
roads, streets and lanes used primarily for transport.
Microclimate performance is closely related to outdoor space with the cases performing
generally poorly for this criteria as well. Outdoor space is needed for tree plantings and
for shaded outdoor areas to exist close to dwellings. Some cases do better on some
aspects of microclimate however, St Kilda performs better with enough space allowing
significant tree plantings. A strong sense of control is created in most cases through a
clear definition between the public and private realm however, this is often through high
levels of facade impermeability and security which negatively affects performance on
flexibility and human scale.
A critical density measure that can reveal more of the qualitative relationship between
FAR and site coverage is the Open Space Ratio or OSR (Berghauser Pont and Haupt,
2021). This is simply the proportion of built floor area to open ground space. The
lowest OSR measured is by Fitzroy and this matches its worse performance on outdoor
space and microclimate. North Melbourne has the highest OSR, its rates for these are
not significantly higher than Fitzroy but this is in part influenced by the change in urban
morphology within this case. The relationship between OSR and qualitative outcome
could likely be observed more clearly in other neighbourhoods. Shortlisted cases such
as 993 Tennyson Street and 1031 Inkerman/Balaclava have OSR around 1.7,
significantly higher than the case studies. While these areas have not been subject to
detailed assessment they were visited as part of the selection process and observed to
44
have significantly increased outdoor spaces and strong microclimate qualities. The full
shortlist dataset is available at Appendix III.
6. Conclusion
6.1 Limitations
This research has gained some insights into urban density in Melbourne but there are
limitations. A key difficulty was constructing a framework to assess the qualitative
outcomes of built form and urban structure. Sim’s Nine Criteria for Liveable Density
were used as a starting point but this method has a bias towards a particular building
typology (European perimeter block) and in its concentration on the building scale,
neglects the scale of the neighbourhood that is of most interest to urban design.
Through this research the framework was adapted and operationalised but this did not
remove these biases.
Another challenge was obtaining quantitative data. GIS layers were sourced from
publicly available databases but the data currently available for building footprints was
not current. The most recent dataset obtainable was 2020 and this misses some newly
constructed buildings reducing FAR in some cases. This dataset was used to generate
key built form measurements such as FAR, site cover, OSR and average stories. A
second quantitative data limitation is the lack of land use data available, only the City of
Melbourne currently conducts a comprehensive survey of land uses, with the exception
of North Melbourne the case studies were outside that municipality. With such data it
would have been possible to consider functional mix as a variable and this would have
been useful as its relationship with residential density and urban structure is critical.
45
6.2 Implications for policy
A key implication for policy from this research comes from the finding that the
‘goldilocks’ medium density occurs in a narrow band of locations. While there are some
interesting emerging medium density areas (ie. Brunswick or Maribyrnong), the case
studies occur in well established, inner city suburbs. Spencer et al (2015) suggest
planning strategies should encourage medium densities over a larger area rather than
excessive concentration. This research shows examples of how this can be achieved.
Since significant urban renewal areas such as Fisherman’s Bend and Arden Macauley
are earmarked for much higher densities, the goldilocks density will need to occur in
established suburbs. This is where there is a strong link between this research and
that of Peter Newton’s Greening the Greyfields (2022). Through showing examples of
density that has achieved up to 50 dw/ha in the Melbourne context with relatively low
heights and high quality, this research demonstrates what the ‘greyfields’ could
accommodate and in a preferred form. The area identified in this study between St
Kilda, Elwood and Caulfield demonstrates how this has been achieved at a district
scale.
A key finding of this research is that diversity is certainly one of the key qualitative
properties of Melbourne's inner-urban form creating the kind of variety and flexibility
sought in Sim’s framework, yet in a different way. Rather than relying on one building
typology, the case studies in this research show a high level of built form heterogeneity.
During the research this made the task of measuring and assessing density levels
difficult. Likewise, it makes management through density prescriptions difficult as a
certain setting will not necessarily result in a particular built form outcome. The
over-reliance on dwelling density measures in planning documents has been
extensively criticised but finding an appropriate alternate measure is difficult. The
research suggests that some combination of FAR, site cover, OSR as well as measures
of interface quality would be desirable.
46
Another issue raised by this heterogeneity is that it likely contributes to increased
perceptions of higher density without genuine density increases. For example, many of
the suburbs filtered out in the searching process contain higher buildings that are
enough to provoke community unease despite being in a context where the overall
neighbourhood density remains relatively low. The case studies in this research can
help to demonstrate that density can be high quality and liveable within the Australian
context. The research recognises the strong desire for this type of density that is often
expressed by urban communities in contrast to the dominant mode of high-rise urban
forms. There is a clear need to rebuild the image of density emphasising its positive
qualities and this research can be part of that story.
6.3 Further research
There are several avenues for further research that are opened up through this
research. One aspect is the link between the densities observed, their qualities and
streetlife. Density is one of four preconditions for streetlife vitality (Jacobs, 1961) and
examining pedestrian activity across the case studies would help advance
understandings of how density has activated streetlife.
Likewise, further examination of the interrelationships between densities and other
aspects of the urban assemblage such as mixed uses would reveal in more detail how
these interact with density. While this thesis focussed on predominantly residential
areas the influence of commercial uses within these areas altered the pattern and
performance of density to some extent.
Finally, shortlisted cases numbering 37 were measured and superficially assessed as
part of the searching phase of this study and these represent a useful dataset for further
study. For instance, these exhibit a greater range of OSR than the selected case
studies and deeper qualitative assessment of these would likely reveal further insight
into this measure and its implications.
47
Figures and tables
Figures
Figure 1. Density targets i n urban theory..................................... ...................................... 8
Figure 2: Population density data grids: .......................................................................... 16
Figure 3: Filtering sequence 1 and 2. .............................................................................. 17
Figure 4: Filtering sequence 1 and 2. .............................................................................. 17
Figure 5: Filtering sequence 5 and 6. .............................................................................. 18
Figure 6: Filtering final cluster .......................................................................................... 18
Figure 7: Case study mapping examples ........................................................................ 20
Figure 8: Thematic analysis of Sim criteria. ..................................................................... 21
Figure 9: Showing ‘long ‘list’ overlaid on 1 ha population density grid map. .................... 24
Figure 10: FAR data grids 1 ha, 20km X 20km square. ................................................... 25
Figure 11: Dwellings p/ha plotted against population p/ha for ‘long list’. .......................... 26
Figure 12: Multi-scalar density of the cases, p/ha. .......................................................... 27
Figure 13: North Melbourne, oblique view Near Map (2023). .......................................... 28
Figure 14: Fitzroy, oblique view Near Map (2023). .......................................................... 29
Figure 15: Fitzroy, oblique view Near Map (2023). .......................................................... 30
Figure 16: St Kilda, oblique view Near Map (2023). ........................................................ 31
Figure 17: Built form diversity North Melbourne left and Richmond right. ....................... 33
Figure 18: A rare example of a courtyard that is semi-public, North Melbourne. ............. 34
Figure 19: Example of flexible undercroft space used for parking. .................................. 34
Figure 20: Good eye level detail and human scaled elements: North Melbourne. .......... 35
Figure 21: Good microclimate North Melbourne, shade of tree, indoor/outdoor space. .. 36
Figure 22: Case studies comparison maps, 500m by 500m neighbourhoods. ................ 38
Figure 23: Spacemate, short-list cases. ........................................................................... 39
Tables
Table 1: Density measures and formula. ......................................................................... 19
Table 2: Sim criteria synthesised and assessment method. ............................................ 22
Table 3: Assessment against adapted David Sim criteria. ............................................... 32
Table 4: Summary of density measures - cases. ............................................................. 36
48
Reference list
Australian Bureau of Statistics. (2021a). Census Quickstats:
https://www.abs.gov.au/census/find-census-data/quickstats/2021/2GMEL
Australian Bureau of Statistics. (2021b). Census Data:
https://www.abs.gov.au/census/find-census-data
Alexander, C., Ishikawa, S., Fiksdahl-King, I., & Angel, S. (1977). A Pattern Language .
Oxford University Press.
Alter, L. (2014), Cities need Goldilocks housing density - not too high or low, but just
right , The Conversation:
https://www.theguardian.com/lifeandstyle/2014/apr/16/cities-need-goldilocks-housing-de
nsity-not-too-high-low-just-right
Batty, M. (2013). The New Science of Cities . Cambridge.
Bryman, A. (2012). Social Research Methods , (4th ed.), Oxford.
Carmona, M. (2010). Contemporary Public Space, Part Two: Classification. Journal of
Urban Design,15 (2), 157-173.
Churchman, A. (1999). Disentangling the concept of density. Journal of Planning
Literature, 13 (4), 389-411
49
City of Melbourne, (2010). Transforming Australian Cities, For a More Financially
Sustainable Future, Transportation and Urban Design , City of Melbourne.
Data Vic, (2023), Discover and Access Victorian Government Data:
https://www.data.vic.gov.au/
Demographia, (2015), World Urban Areas 11th Edition,
http://www.demographia.com/db-worldua.pdf
Dovey, K., & Wood, S. (2015). Public/private Urban Interfaces: Type, Adaptation,
Assemblage. Journal of Urbanism: International Research on Placemaking and Urban
Sustainability 8(1), 11-29.
Dovey, K., & Woodcock, I. (2014). Intensifying Melbourne: Transit-Oriented Urban
Design or Resilient Urban Futures , Melbourne School of Design.
Dovey, K., & Pafka, E. (2014). The Urban Density Assemblage, Urban Design
International , 19(1), 66-76.
Dovey, K., & Pafka, E. (2016). The science of urban design, Urban Design International ,
21 , 1-10.
Dovey, K., & Symons, F. (2013). Density without intensity and what to do about it:
reassembling public/private interfaces in Melbourne’s Southbank hinterland, Australian
Planner, 51 (1), 34-46.
50
DELWP, (2017), Plan Melbourne 2017-2050 , Department of Environment, Water and
Planning (DELWP).
DELWP, (2022), 20-minute neighbourhoods, Plan Melbourne:
https://www.planning.vic.gov.au/policy-and-strategy/planning-for-melbourne/plan-melbo
urne/20-minute-neighbourhoods
Eberle, D., & Troger, E. (2014). Density and Atmosphere, On Factors relating to Building
density in the European City . Birkhauser Basel.
Ewing, R., & Handy, S. (2009), Measuring the Unmeasurable: Urban Design Qualities
Related to Walkability, Journal of Urban Design,14 (1), 65-84.
Fooks, E. (1946). X-Ray the city! The density diagram basis for urban planning , The
Ruskin Press.
Gehl, J. (1977). The Interface between Public and Private Territories in Residential
Areas , [Masters Thesis, University of Melbourne].
Gren, A., Colding, J., Berghauser Pont, M., & Marcus, L. (2018), How smart is smart
growth? Examining the environmental validation behind city compaction, Ambio, 48 ,
580-589.
Jacobs, J. (1961). The Death and Life of Great American Cities . Vintage Books.
51
Kristjansdottir, S. (2019), Roots of Urban Morphology, International Journal of
Architecture and Planning , 7, 15-36.
Marshall, S., & Caliskan, O. (2011), A Joint Framework for Urban Morphology and
Design , Built Environment , 37 (4), 409-426.
Mares, P. (2018), No place like home: repairing Australia’s housing crisis. Melbourne:
Text Publishing.
McFarlane, C. (2021). Future Densities: Knowledge, Politics and Remaking the City. In
A. Amin, & M. Lancione (Eds.), Grammars of the Urban Ground. Duke University
Press.
Mees, P. (2009). Transport for Suburbia: Beyond the Automobile Age . Taylor & Francis
Group.
Morley, M., & Pafka, E. (2023), ‘Density Done Well’ in pursuit of 20-Minute
Neighbourhoods: Navigating Fluid Discourses in Melbourne. Urban Policy and
Research, 41 (2), 148-163.
Moudon, A. (1997), Urban Morphology as an emerging interdisciplinary field, Urban
Morphology , 1, 3-10.
Nearmap, (2023). Updated Aerial imagery: https://www.nearmap.com/au/en
52
Neuman, M. (2005), The Compact City Fallacy, Journal of Planning Education and
Research 25 , 11-26.
Newman, P., & Kenworthy, J. (1989). Cities and Automobile Dependence: An
International Sourcebook . Gower Technical.
Newton, P., Newman, P., Glackin, S., & Thomson, G. (2022), Greening the Greyfields:
New Models for Regenerating the Middle Suburbs of Low-Density Cities , Palgrave
Macmillan.
Oliveira, V. (2016). Urban Morphology: An Introduction to the Study of the Physical
Form of Cities . (2nd ed.), Springer.
Pafka, E. (2013), “Nothing Gained by only Counting Dwellings per Hectare: A hundred
years of confusing urban densities”, State of Australian Cities Conference, Sydney.
Pafka, E. (2014), Urban density rhythms: How does urban morphology mediate
streetlife intensity over time, PhD thesis , [Doctoral dissertation, University of Melbourne]
Pafka, E. (2020). Book Review, Soft City by David Sim, 2019, Island Press, Built
Environment, 46 (4), 651-652.
Pafka, E. (2022), Multi-scalar densities: from the metropolitan to the street level, Urban
Design International, 27 (1), 53-63.
53
Berghauser Pont, M., & Haupt, P. (2005), The Spacemate: Density and the
Typomorphology of the Urban Fabric, Nordisk Arkitekurforskning, 4.
Berghauser Pont, M., & Haupt, P. (2021). Spacematrix: Space, Density and Urban Form
(2nd ed.). nai010.
Rowley, S. (2017). The Victorian Planning System, Practice, Problems and Prospects .
The Federation Press.
Rittel, H., & Webber, M. (1973). Dilemmas in a general theory of planning. Policy
Sciences, 4 (2), 155-169.
Sim, D. (2019). Soft City: Building density for everyday life . Washington.
Soria y Puig, A. (1999). Cerda: The Five Bases of the General Theory of Urbanization.
Electa.
Spencer, A., Gill, J., & Schmahmann, L. (2015, December), Urban or Suburban?
Examining the density of Australian Cities in a global context. [Paper presentation].
State of Australian Cities Conference 2015. Gold Coast.
Stanley, J., & Hansen, R. (2020), People love the idea of 20-minute neighbourhoods. So
why isn’t it top of the agenda? The Conversation:
https://theconversation.com/people-love-the-idea-of-20-minute-neighbourhoods-so-why-
isnt-it-top-of-the-agenda-131193
54
Taylor, E., Hurley, J., & Dodson, J. (2017). Getting Dense, why has urban consolidation
been so difficult. In N.Sipe, & K.Vella, (Eds.), The Routledge handbook of Australian
urban and regional planning . Routledge. London.
Uytenhaak, R. (2008). Cities full of space: Qualities of density , 010 Publishers.
55
