You are on page 1of 4

The Bartlett School of Graduate Studies

MSc Advanced Architectural Studies 2007-8

Using DepthMap for urban analysis: a simple guide on what to do once you have an
analysable map in the system.

Bill Hillier

Before we begin
A few things you need to know before you start
- Depthmap can do axial analysis, but you don’t need it. Segment analysis is similar, but more
powerful and can do more things
- The default analysis in Depthmap is segment angular analysis. That means the units of analysis are
street segments and the ‘distance’ relation between them is the amount of angular change from
one segment to the other – so a straight connection between two segments is a 0-degree
connection, and a series of 0-degree connections will be a straight line. So in a sense, the analysis
finds the line structure. We call this least angle or geometrical analysis, in contrast to the fewest turns or
topological analysis we have in axial analysis (and in Axman), or shortest path or metric analysis, which
can also be done in DepthMap.
- we use least angle analysis because this has been shown to correspond most closely to how
people navigate urban space (see Hillier & Iida 2005 Network and psychological effects in urban
movement – an essential read)
- You shouldn’t use metric distance within the calculations of integration and choice (it won’t give
you much useful information). You should use metric distance (along the roads) as the radius for
the measure. In this context by radius we mean how far the calculations are reaching. So the
measures you will get from the analysis set out below will be least angle segment analysis within a metric
radius.

Preparing the map for analysis


Start with an axial map in dxf format. Open depthmap. Then go:
o File – New – gives a blank page. Best use a white background at
this stage
o Layer – Import – your dxf file
o Tools – Line – Make axial map from drawing layers
o Tools – segment – Make segment map form axial map, clicking on
Remove axial stubs less than 25% of line
o You now have a segment map
o Good to change to black background at this stage.

Analysis of the map


Go: Tools ----- Segment ------ Run segment analysis

A box appears
o Make sure Tulip is clicked
o click Include choice (betweenness)
o click Metric for Radius Type
o then under List of radii insert
n,250,500,750,1000,1250,1500,1750,2000,2500, 3000,4000,5000
with a comma but no gap between each (no comma at the end)
o (some people prefer (n,200,400,800,1200,1600,2000,2400,3200 –
you decide)
o click Include weighted measures and scroll to segment length
o click OK and the analysis will run. A box will tell you how
long it will take.

NOTE – if you don’t have a proper metric (scale) on your map, then DepthMap will allow you
to set integer radii i.e. 1,2,3…..n where 1 is a single one of the units in which the map was drawn

Using DepthMap for urban analysis_1401081 Page 1 of 4


The Bartlett School of Graduate Studies
MSc Advanced Architectural Studies 2007-8

Creating the integration measures


You now have the analysis.
o The measures you need are the simple choice and mean depth measures
o Ignore for the time being those with Norm, and Segment Length Weighted (and the
combination of the two).
o NOTE: the mean depth measures are meaningless as they are and need to be normalised
for the number of segments within the radius by using the formula box to go Node
Count/Mean Depth for each radius, calling it, say for radius 1000 metres,
NC/MDr1000metric (node count over mean depth for a metric radius of 1000 metres).
This in effect emulates the RA equation in The Social Logic of Space and is in effect the
integration measure for segment angular analysis.

So what does it all mean ?


You now have a whole set of integration and choice measures at variable radii. What can we do
with these? The first step to recall what the measures mean.

o Integration measures how close each segment is to all others within the radius, using the
least angle measure of distance, so it is a measure of how accessible each segment is from
all the others, and so how much potential it has as a destination for movement i.e. you
would put a shop in an easily accessible location from all other segment, because people
make many short trips and few long trips, so the closer a segment is to all others the
more promise it offers as a destination. You could say that integration measures the
destination potential for a segment at that radius.

o Choice, in contrast, measure the degree to which each segment lies on least angle routes
between all other pairs of segments within the radius, so it measure the through-movement
potential of each segment within that radius – the passing trade potential perhaps ? – in
contrast to the to-movement potential measured by integration.

So the two measures deal with the two components of human movement i.e we select a
destination, more often a close one and less often a distant once – and we select a route i.e. the
series of segments will pass through. As we have shown that people navigate the grid using
something like a least angle model, then it is not surprising that the destination and route potential
measured by integration and choice with the least angle definition of distance so often give a
good account of real movement.

We would expect local movement to be best accounted for by a local radius choice measure –
800 metres is the current favourite with space syntax limited – but in, say, a market areas with a
local intensified grid, we may find it works best at a lower radius, perhaps 400. Larger scale
movement such as vehicular movement should best be reflected in a higher radius measure.

These potentials will be reflected in maps. The more you reduce the radius of the choice
measure, the more it will identify a smaller scale and denser network of paths – BUT you may
need to adjust the colour range (you find it under View) to bring out the pattern of local
differences. There is nothing wrong with this. The numbers all stay the same. All you are doing is
clarifying the local differences by making the colour spectrum narrower.

Combining integration and choice


You can also combine the integration and choice measures by multiplying one by the other. For
technical reasons you need to take the log of Choice, and add 2 to it (because the original
calculation will give some -1’s), and multiply by NodeCount/MeanDepth, so
(NC/MD)*(log(CH+2)). In the past this has often been done by (1/MD)*(log(CH+2)) and this
is the same measure. With the combined measure, which can be done at any radius, you are
seeing the combined potential of a segment as both a destination and a route.

Using DepthMap for urban analysis_1401081 Page 2 of 4


The Bartlett School of Graduate Studies
MSc Advanced Architectural Studies 2007-8

Segment length weighting


BUT if you are looking at a large system, say London, then you will find that using choice with
segment length weighting makes the picture clearer. The reason for this is that block size in cities
tends to grow from the centre outwards, and the segment length weighting version of the
measure partially neutralises this.

What then does ‘segment length weighting’ mean ? It means weighting the choice measure by the
product of the lengths of the origin and destination nodes. The idea is that a long node will have
more buildings on it, and so generate and attract more trips, and a short node fewer. So with
ChoiceSLW you are weighting the movement generated by each pair of nodes by how much
activity we expect both the origin and destination to generate. The lower the radius, the more
differences it make.

A interesting development of this is also weighting the choice measure by the length of the route
between thee origin and destination. This can be done by simply going
ChoiceSLW/TotalSegmentLength (TSL) for that radius. Work on the Choice measure is still
going on.

A theoretical difference between the measures


Another difference between the measures is I think interesting enough to bear in mind when
seeking and interpretation of the results of analysis. Because integration deals with the closeness
of a segment to its neighbourhood we can intuit it in the sense that we can imagine something like
its justified line graph from where we are. Choice cannot be intuited, because it deal with how the
segment we are on feature on routes between locations which might be quite remote. We cannot
imagine some equivalent to the justified graph. For this reason, I think we can think of
integration patterns as constructive, since they have to do with how the city was put together in the
first place, and choice patterns as emergent since they arise as a consequence of how the city was put
together.

Why not use metric distance as the distance measure ?


If we use metric distance as the distance measure, we are assuming that people know how to
calculate metric distance in complex spatial systems. They don’t. They think they are, but the way
they do it is use a geometrical representation of the grid structure, and try to ‘follow their noses’
to the destination by minimising angular deviation. If we do use metric distance within the
measure, as opposed to for the radius of the measure, then we find nonsense. The picture of the
urban system you get from a metric analysis has little or no relation to the observed functional
patterns. This is all explained more fully in Hillier, Turner, Yang and Park (2007) Metric and topo-
geometric properties of urban street networks: some convergences, divergences and new results given at the
Istanbul Symposium, along with some dramatic new things that can be done with DepthMap.

One of these is a metric technique for identifying urban areas by looking at their block structure
and scaling. The measure can be done by:

Go

Tools
Extra Segment Tools
Metric and Topological analysis

Then in the box

Click metric
Inert a single radius (it only does one at a time)
Click ok

Using DepthMap for urban analysis_1401081 Page 3 of 4


The Bartlett School of Graduate Studies
MSc Advanced Architectural Studies 2007-8

The result will look a mess, but in the colour range move the blue colour 3 clicks to the right and
the red Colour 1 click to the left, and it will pick out a patchwork, in which the blue patches are
small block areas and the red large blocks. This is a whole new area of research and you need to
read the Istanbul paper.

Some recent studies


We will now develop these ideas further by looking at three ongoing studies we are carrying out
as part of the Urban Buzz iValul programme. The studies are all about trying to link spatial
design to economic value, all are London focused, and cover:
- spatial aspects of local centres
- spatial dimensions to property values
- spatial factors in urban crime patterns.

All of these are available on the iValul website and can be accessed freely:

Address
ftp://urbanbuzz:8802london@mailgate.spacesyntax.com

username= Urbanbuzz
password= 8802london

Open the UCL Students folder

bh
14.01.08

Using DepthMap for urban analysis_1401081 Page 4 of 4

You might also like