|
Designing for Context:
the use of 'space syntax' as an interactive design tool in urban developments
Mark David Major and Tim
Stonor
Published in the Proceddings of the14th Inter-Schools Conference on
Development, Edinburgh, Scotland, 24-25 March 1997
Keywords: Configuration, Design, Modelling, Movement, Urban
Abstract
Computer modelling is an
increasingly important way for architects and developers to visualise
the mechanical and visual form of their design proposals before they
are actually built. However, many computer modelling programmes are
static in that they can only respond to predetermined design inputs
- too often the building is designed first, then 'verified' by computer.
As a result, computer modelling becomes an end-product of the design
process rather than an integral decision-making part of it. Recently,
a set of techniques developed for the configurational analysis of space,
known as space syntax, have formed the basis for a new generation of
software programmes through which computer modelling has become both
interactive and dynamic. Space syntax takes as its subject the layout
of streets and public spaces in settlements which, when seen together,
form the network of space through which people move in towns and cities.
Modelling these networks has proven to be effective in predicting patterns
of pedestrian and vehicular movement and levels of space use in urban
areas. These findings form the basis of space syntax as a powerful,
new urban design tool - a tool founded on the simple idea that the way
we design space is fundamental to the way we use it. Using this technique,
the probable outcome of design decisions can be forecasted during the
design process. Designs can then be modified so they will achieve levels
of movement and space use appropriate to the functions desired on the
site, i.e. high levels of movement for retail streets or lower levels
for residential ones. This is demonstrated in a recent project for Zaanstad
on the outskirts of Amsterdam where configurational analysis was central
to development of a long-term strategy for urban growth in the area.
A Vision for Zaanstad
In January 1996, the Space
Syntax was represented at the second Zaanstad design workshop
in Holland. The theme of this workshop was 'movement' and a computer
model of the network of routes in Zaanstad was used to sketch the likely
effects of changes in this network on patterns of movement. Following
this workshop, the Space Syntax was commissioned by Battle
McCarthy Consulting Engineers to pursue a detailed space syntax analysis
of the town with a view towards enhancing pedestrian movement as an
integral part of the strategic plan for urban development in the town
over the long-term, see Figure 1.
Click here to see Fig. 1
The Space Syntax is a research-based consultancy which specialises in
the analysis and design of buildings and cities. At the heart of its
work is the finding that patterns of movement and interaction within
the built environment are directly related to design and, most importantly,
to patterns of space. Over the past twenty years, the company has developed
computer-based modelling techniques to describe spatial patterns and
through intensive observational studies has developed a detailed understanding
of the relationship between design and activity. Its work covers building
interiors as well as public, outdoor space and the relationship between
spatial design and movement has been found for vehicular traffic as
well as pedestrian activity. Many recent studies by the Space Syntax
have revealed the importance of movement patterns both to
the success of public spaces in urban areas and to levels of communication
and interaction within building interiors.
This paper presents the study of Zaanstad carried out by the Space Syntax
. It shows how proposals for new public spaces and building
layouts in Zaastad were tested using spatial models to assess the effects
of designs on existing patterns of movement. It is demonstrated how
this is possible by first; showing how research using space syntax techniques
over the last twenty years had led to the development of computer models
which are interactive and readily amended and second; how these can
be used in the course of the design process to assist in design development
and refinement in bring shape to strategic masterplans for urban areas.
A series of computer simulations demonstrate how changes can be made
in the urban fabric of the town during its normal course of development
to improve the accessibility of primary local routes and links between
local areas. The point was not to lay down 'hard and fast' routes about
the form of development but to provide a general framework for the layout
of stratgeic routes in which development could take place. Following
a brief description of the space syntax method, the existing spatial
configuration of Zaanstad is described and how, by using the computer
modelling capabilities of space syntax, a number of steps can be taken
to outline a strategy for incremental urban growth in the area to develop
a long-term 'Vision for Zaanstad'.
Configurational Analysis
of Space
Configurational modelling,
or space syntax, is a set of techniques for representing space in a
building or city. For example, movement in a city tends to approximate
lines so one representation of space is a series of all longest and
fewest lines of sight and access, see Figure 2a. The stationary use
of space in a building or city by people will tend towards convexity,
i.e. the mathematical definition of all points being visible to each
other as in a group of people gathered in a circle, thus another representation
of space could be the collection of all 'fattest' two-dimensional 'lumps
' of space in a system, see Figure 2b. Finally, the potential for seeing
and moving can be represented as isovists, or fields of vision, which
are all the visible and accessible spaces which we might move to from
a point or particular set of points in space, see Figure 2c. We can
make more complex representations of space by using any combination
of these three depending on the problem to be researched, see Figures
2d and 2e.
Fig. 2 
Fig. 3
This is how space can be represented however, some clarification is
required on what we mean by configuration. For example in Figure 3a,
two objects are in a mathematical relationship to each other so it can
said that 'a' is to 'b' as 'b' is to 'a'. However, once this relationship
is established with regards to a third, in this case the surface of
the earth, there is a configurational relationship. For example in Figure
3b, we can say that 'a' is to 'c' as 'b' is to 'c' but in order to reach
'b' from 'a' we have to pass through 'c'. This can be seen more clearly
in the corresponding graphs. Next the idea of permeability or connection
can be introduced. If the objects are in a symmetrical relationship
where all spaces are maximally shallow from each other, we can say that
'a' is to 'b' as 'b' is to 'c', see Figure 3c. Finally, if ' b' is placed
on top of 'a', an asymmetrical relationship results, see Figure 3d.
This is because to reach 'b' from 'c' you have to pass through 'a' but
this is not necessary to pass from 'c' to 'a'. This is what is meant
by configuration; a relational system where any local change in that
system has global effects across the system.
In previous studies, the Space Syntax has found that axial
lines are most revealing representations of urban systems. Figure 4
shows an axial model of Greater London, approximately between the North
and South Circular roads.
Click here to see Fig. 4
We can then ask the computer to mathematically measure the relationship
between every space in London to every other space relativised for the
size of the system. A measure called 'global integration' because it
measures relationships globally across the system. The computer colours
the map from dark through to light with dark lines being the most integrated,
or shallow, and the light colours being the most segregated, or deep,
see Figure 5.
Click here to see Fig. 5
This computer model of Greater London shows that the most integrated
space is Oxford Street. Let us be clear at this point that the model
is a purely mathematical representation of pattern - land use, density,
and other functional uses have not been taken into account though we
can later build these into the model. Despite this, the model provides
us with a very realistic picture of London. We can also ask the computer
to provide a more localised picture of the city by examining only those
spaces up to three steps away from every space, or 'local integration',
see Figure 6. By doing this, the model highlights a series of locally
important shopping streets throughout the city.
Click here to see Fig. 6
Natural Movement
One of the most common accusations
made against modern urban developments is that they lack 'vitality'
and at worst turn into 'urban deserts' devoid of people even in the
middle of the day. It is a longstanding question whether the design
of these modern schemes is in any way at fault, or whether the blame
rests with the way that the schemes are managed, or the socioeconomic
status of the area and so on. The multivariate nature of this sort of
socio-spatial question has made it seem almost impossible to resolve
this question one way or the other. However, these new techniques for
describing and quantifying the geometric and topological form of urban
space allow us to approach some of the simpler questions at the heart
of this issue.
Since we can now describe and quantify radically different spatial designs
on the same basis we can begin to 'control' the design variable in studies
of other aspects of urban function. It is possible to detect effects
of spatial design on patterns of pedestrian movement by observing pedestrian
flow rates at a number of locations and then using simple bivariate
statistics to look for the relationship between configurational and
flows. A large number of studies have now established that integration
is consistently the strongest predictor of pedestrian flow rates (see
Hillier et al, 1993, for a comprehensive review of the evidence). Integrated
spaces carry greater pedestrian flows than more segregated ones. The
effects are strong and consistent. For example we can see this in Figure
7 which shows an area of London called Barnsbury where detailed observations
of pedestrian movement patterns have been made several times over the
last decade.
Click here to see Fig. 7
In this study, each street segment was observed for a total of 50 minutes
at different times of day and on different days of the week. The all
day mean hourly pedestrian flow is noted on each segment. The scattergram
between the measure of 'local integration' and the square root (a statistical
transformation to correct to a normal distribution)of pedestrian flow
rates in Figure 8 shows a strong correlation. In this case the model
is much larger than that shown, extending approximately two kilometres
away from the observation area in all directions. The correlation is
remarkably strong at r=.734, p<.0001.
Fig. 8
The key discovery is that the correlation is between pedestrian flows
and a purely spatial measure of the pattern of the street grid. No account
has been taken of the location of attractors or generators of movement
in constructing the measure of integration. It seems that movement patterns
result naturally from the way the spatial configuration of the street
grid organises simplest routes (involving fewest changes of direction)
to and from all locations in an area. Of course, this runs counter to
the premises of traffic modelling which holds that the key facts in
urban systems are the distributions of activities and land uses that
'generate' or 'attract' flows between different geographic locations.
Our results show that the primary fact is the pattern of space, and
that if there is a relationship between land uses and pedestrian flows
(which there certainly is - you find more people on streets with shops
than on streets without). This is most likely to be due to retailers
choosing their shop sites with care in order to take advantage of the
opportunities for passing trade provided by the natural movement pattern
resulting from the grid. These findings would suggest a marked shift
in emphasis from existing urban theory is required away from attraction
and rigid zoning of land uses to a dynamic mixture of land uses building
on the potentials for all types of movement but especially through movement,
what we call 'natural' movement.
Zaanstad Today
The space syntax method was
used to develop a strategy for enhancing pedestrian movement in Zaanstad
as part of a long-term development plan in the town. A spatial model
of Zaanstad was constructed and processed by computer and then used
as the basis for testing design schemes incorporating new strategic
connections across the system. The spatial model in Figure 9 represents
the existing system of public space in Zaanstad made up of 2503 lines
of sight and movement. (While this model contains all possible routes
through space, other models have been made to correspond with pedestrian,
vehicular and cycle routes).
Click here to see Fig. 9
The colours of the lines relate to this important spatial property of
integration. Looking closely at the lines of movement, an 'integration
core' of blacks and dark grey lines can be identified. This picks up
the main traffic routes in the area, including the Coentunnel Weg and
the Provinciale Weg. The core extends to include the Gedempte Gracht
main shopping street, Vincent van Gogh Weg and the Dr. H. G. Scholten/
Heijermansstraat route. In effect it covers most of the spatial system,
from centre to eastern edge and from north to south. However, strong
segregation can be seen in the light grey and white lines of the largely
residential areas to the west of the railway tracks and in the southern
streets of Wormer and in the Westerspoor, Zuiderhout and Achtersluispolder
industrial areas adjacent to the Noordzeekanaal.
Of these segregated areas, by far the largest is that of the Wester
Koog, Wester Watering and Westerspoor housing areas. Their segregation
is caused in the main by two effects. First, they have very poor east-west
connections across the railway tracks to the eastern part of Zaanstad.
Unlike the bridges across the Zaan - which occur on long, strategic
lines of movement - those across the railway tracks are either just
off or significantly distant from important east-west routes. For example,
the Bloemwijk tunnel just misses an alignment with Lelie Straat while
the pedestrian route across the tracks inside the railway station avoids
the very strategic Gedempte Gracht alignment. Second, the housing areas
to the west of the tracks are very spatially fragmented. In general,
the individual lines of movement here are shorter and less connected
than lines in the older parts of the system to the east.
The 'Strip' Effect, or Linear
Intelligibility
Detailed analysis of the
integration structure of Zaanstad reveals a pattern of development which
prioritises a series of linear routes - or ribbons of growth - through
the area. These routes include J J Allanstraat, Dorps Straat, Ros Molen
Straat and, most importantly, the 'High Street' ribbon of Gedempte Gracht/Peperstraat,
see Figure 10a and 10b.
Click here to see Fig. 10
Recent space syntax studies have detected similar patterns in other
cities of the world suggesting that this property of 'linear intelligibility'
is a more fundamental aspect of city growth than was previously realised.
When an area grows linearly around a single route or series of routes,
there appears to be a tendency for these lines to become privileged
within the overall pattern of the grid. The result is known as a 'strip'
effect whereby the principal route and all spaces which connect to it
become locally strategic within the larger system of spaces and serve
to distribute large-scale movement into the fine-scale structure of
the grid.
One celebrated example of the strip effect is Las Vegas, see Figure
11a and 11b. On the ground, the strip is read as an intelligible system
which is easy to understand and move along. In the analysis, it can
be detected in a scattergram relating integration with another spatial
property called 'connectivity' which is a simple measure of the number
of connections any single space makes with its neighbours.
Click here to see Fig. 11
In the scattergram, Las Vegas Boulevard and all the routes which connect
directly to it are picked out as black dots. These form a linear set
of points at a steeper angle than the overall scatter, representing
the strip effect as a local intensification of the grid. A similar correlation
between global integration and local integration can be found for Zaanstad
when the Gedempte Gracht/Peperstraat routes and all the routes which
connect directly to it are highlighted (Note: The use of connectivity
for Las Vegas instead of local integration is due to the greater shallowness
of the American grid pattern which results in global and local integration
tending to be very similar to one another whereas connectivity reveals
greater differentiation between the local properties of spaces in the
grid). This finding suggests that one strategy for future development
in Zaanstad might be to relate future growth directly to the detailed
morphology of the existing spatial system.
A strategy such as this is centred upon the High Street and how this
can be further developed is demonstrated below by first, extending its
alignment west across the railway tracks to continue the route and second,
adding urabn development along its length to enhance the strip effect.
Linking East-West
Figure 12 shows two possible
options for the extension of the High Street west across the railway
tracks. In Option 1A, the High Street is extended as far as the Economics
College and is linked into the existing street layout there. In Option
1B, these links are developed towards the north so that connections
are made into the southernmost residential area. The aim here is to
develop the site to the west of the railway without affecting the relatively
segregated nature of the housing areas.
Click here to see Fig. 12
Figure 13 shows Option 2A involving a strategy whereby a regular grid
replaces the existing layout and a strategic new link is introduced
through the centre of the residential area along the De Watering canal.
This integrates the heart of the residential area and relates it directly
to the extended High Street. Option 2B develops this to embrace both
the residential areas to the north and the industrial sites to the south.
In both options, the continuance of the strip effect can be seen in
the linear arrangements of black dots within the scattergrams.
Click here to see Fig. 13
Linking North-South
In the set of computer simulations
shown in Figure 14, the nature of the north-south links into the residential
areas has been developed in stages so their impact can be measured incrementally.
Stage 1 simulates two links: a simple connection along the western boundary
of the area as far as Ina Dammanstraat, and a stronger, more central
two-step link from the new grid to Dalerveen Houtveldweg which then
continues north along a cycle route though to Wester Koog. Stage 2 straightens
out the central route to a single step between the grid and Dalerveen
Houtveldweg. Along the western boundary, Annie V. Ees Straat is extended
south to link with the route created in Stage 1. Stage 3 continues to
strengthen the central route by straightening and extending it further
north as far as Houtveldweg. Along the western boundary, the Annie V.
Ees Straat route is connected north into S. Kooymanstraat. Stage 4 simulates
a final development to the area with the central route straightened
further.
Click here to see Fig. 14
It can be seen that the effect of incrementally extending the two routes
is first, to strengthen the area-to-area connections between the proposed
grid and the
residential areas and second, structure a pattern of integration in
the residential areas which integrates from the centre out, not from
edge-to-centre as at present.
Ease of Accessibility
Analysis shows that extending
the High Street west of the railway station and connecting a new route
north from it into the residential areas significantly improves accessibility
for the rest of Zaanstad.
The two maps shown in Figure 15a and b demonstrate the degree of this
improvement by measuring the 'local catchment area' of the High Street
before and after the changes are made. This is done by first, selecting
all streets which are within three changes of direction from the High
Street (i.e. all the routes which go into calculating the local integration
of the High Street) and second, asking the computer to measure depth
from this set of spaces. The original set of lines are then coloured
black, those one step away are coloured dark grey and so on until the
deepest spaces are coloured white. Besides the striking before-and-after
visual representation, this technique provides numeric data to demonstrate
how much more accessible the High Street becomes and how the local catchment
area increases significantly in size. In the existing situation, ease
of accessibility from the High Street is primarily focused to the east
of Zaanstad with some north-south accessibility being generated through
the Provinciale Weg and H. Gerhard Straat routes. The average 'depth'
of the system from the High Street is 8.4 steps and the mean depth of
the local catchment area is 5.4.
Click here to see Fig. 15
In the proposal, ease of accessibility to the High Street is significantly
increased to the west, northwest and southwest. This increase happens
as a direct result of extending the High Street and making a direct
route through the heart of the residential areas in the new grid layout.
It is also helped by Westzanerdijk which runs along the southern boundary
of the proposal. The High Street becomes shallower with a mean depth
of 7.8 and the shallowness of the local catchment area drops to 4.8.
In fact, the size of the local catchment area of the High Street grows
from 212 spaces in the existing situation to 280 spaces in the proposal.
This represents a 32% increase in the local catchment area of the High
Street. Overall, mean global integration in Zaanstad as a whole has
been significantly increased from 0.77 to 0.82.
Long-term Development in
Zaanstad
The computer models shown
in Figure 16 demonstrate the potential for long-term growth, area development
and incremental improvements in the spatial structure of Zaanstad. Stage
1 models the impact of new links from Provinciale Weg into the residential
areas of Wester Watering and Westerspoor. These links have the limited
effect of moderately integrating the western part of the system. Stage
2 shows the potential for linear development in the western residential
areas adjacent to and running alongside the railway. This creates a
secondary north-south route from Dalerveen Houtveldweg to the extended
Ebbehout route in the south.
   
Click here to see Fig. 16
Stage 3 models the potential for urban 'in-fill' at the Valk Straat
Industrial Estate. A layout is formed which links directly to Provinciale
Weg to reinforce its role as a distributor of east-west movement and
then Valk Straat is extended as a central north-south route through
the heart of the site. Stage 4 shows a later stage of growth in Zaanstad
where new area-to-area links are introduced both north-south and east-west
to build on the High Street extension west of the railway station.
A Vision of Zaanstad
The computer model in Figure
17 presents a summary of the new developments, incremental improvements
and long-term options for growth modelled in the study as a complete
vision for Zaanstad which aims to enhance pedestrian movement in the
area.
Click here to see Fig. 17
This vision includes an extension of the High Street west of the railway
station through the heart of a semi-regular grid development, the urban
in-fill of the Valk Straat site, linear development in the west along
and adjacent to the the railway tracks and the introduction of new area-to-area
links on east-west and north-south alignments. Together, these developments
have the effect of:
- greatly increasing the
accessibility of the High Street routes for pedestrians and enhancing
retail opportunities in the area, see FIGURE 17b;
- creating new local areas
to the west of the railway, in the Valk Straat site and in the proposed
new grid which for pedestrians will be easy to read and move around
in, see FIGURES 17c AND 17d;
- integrating the isolated
western residential areas of Wester Watering, Westerspoor and Wester
Koog; and,
- generating a new heart
for Zaanstad focused on both the railway station and the extended
High Street.
Conclusions
This paper has demonstrated
just one example of how space syntax computer modelling can be used
as a dynamic and interactive design tool during the design process by
architects, planners and local authorities rather than as an 'after
the fact' demonstration of how urban developments will appear once built.
The models are dynamic. They provide architects with information about
the likely effects of alternative design strategies because the computer
models successfully account for actual patterns of movement and urban
function within the town. Also, because they are easily amended they
can be used to test several different design strategies to determine
which are appropiate within the urban context they are to be built.
Previous studies by the Space Syntax have also demonstrated
that many of the things considered wrong about our cities also can be
linked to a lack of movement, i.e. failed retail, distribution and patterns
of crime as well as higher social outcomes in anti-social behaviour
and perceived social malaise. The basis for generating sustainable levels
of movement in the city would appear to be fundamentally configurational
in nature, that is it involves the spatial layout of streets in relation
to their larger urban context. Indeed, much of our research suggests
that to begin to understand, and correct for, many of the ill-understood
problems of our cities we must first understand configuration and its
effects in patterning a basic level of movement through the route structure
of cities.
Space syntax computer modelling is continually advancing our knowledge
on this relationship between the configuration of urban form and the
pattern of urban function within cities through a series of high-profile
urban projects around the world including the World Squares for All
masterplan for the Trafalgar Square, Whitehall and Parliament Square
areas in central London with Foster and Partners and the Richard Rogers
Partnership proposals for the South Bank in London, see Figure 18. It
seems apparent to us from these projects that what architects and urban
designers really require is computer modelling programmes which can
be applied during the design process in making decisions about the layout
of streets and the locating of uses to better ensure that future urban
developments bring positive benefits to our cities.
Click here to see Fig. 18
Bibliography
Hanson, Julienne and B. Hillier.
(1987) 'The Architecture of Community: Some New Proposals on the Social
Consequences of Architectural and Planning Decisions' from Architecture
and Behaviour. Volume 3, Number 3, pp. 251-273.
Hillier, Bill, (1996). Chapter 4, 'Cities As Movement Economies' from
Space Is The Machine. Cambridge University Press, Cambridge, England.
Hillier B., Penn A., Hanson J., Grajewski T., Xu J. (1993) Natural Movement:
or configuration and attraction in urban pedestrian movement, Planning
and Design: Environment and Planning B, Pion, London.
Hillier B. & Hanson J. (1984) The Social Logic of Space, Cambridge
University Press, Cambridge, England.
Stonor T. and Major M.D. (1996) A Vision for Zannstad, Space Syntax
Report. Copies available from the Space Syntax
|
