Thanks to Dr Kayvan Karimi, Alasdair Turner and Dr Jake Desyllas for
initiating and continuing an interesting discussion on Visibility Graph
Analysis  (or Isovist Integration analysis, as it has been informally known
among those who have been applying it at UCL).

During the past one and a half  years we have come a long way in developing
not only the method and software that performs this type of analysis, but
also an understanding of its application and some valuable experience of
using it on a range of spatial environments. Throughout the development of
VGA, many ideas about new measures, new tools and new ways of using the
method have emerged from the requirements of various real projects, and
applications for intelligent descriptions. Below are some comments on the
method and its application.

The strength of VGA is in that, based on the relationships between point
isovists, it enables detailed description of local and configurational
properties differentiating among locations in any spatial system.
Whereas axial analysis works best in linear spaces, like urban systems, and
convex syntactic analysis perhaps in cellular layouts, VGA provides the
ability to analyse in detail any continuous space. For example 'open plan'
buildings, where there are more than one possible ways of applying a convex
or axial break-up. 

Although the representation step (e.g. manually drawing axial lines or
dividing into convex units before analysing for configuration) is eliminated
and space is instead covered by a regular grid of points, questions about
representation are inherent in this method too, in issues such as:

- the choice of grid resolution (a matter of capturing detail balanced with
processing time and power) and exact location of grid points, 
- the selection of the height of analysis (for example at 'eye level' for
visibility or 'knee level' for access), 
- the coverage of the grid (for example, for analysing visibility, one can
omit points within - but still maintain visibility across - a void/atrium in
a building or the middle of a vehicular road or length of a river). 
- the choice of boundaries (especially within the urban fabric).

For example, studies of part of Frankfurt and the City of London (Space
Syntax Laboratory, 1999) have shown that a vital factor in VGA is the way
the grid analysed is embedded within the surrounding urban fabric. Another
important factor is the way of averaging values around the observed
locations for evaluating the model against movement rates.
The application of VGA in the previous examples complemented axial analysis
by adding an understanding of the finer scale characteristics of well-used
public spaces, directly relating to patterns of isovist integration, isovist
size (connectivity) and degree of multi-directionality. 

Studies of public building interiors,  such as the Royal Festival Hall
(M.D., 1998), London Department store and Wales Millennium Centre (Space
Syntax Laboratory 1998-9), have shown that a combination of Isovist
Integration patterns analysed for visibility and for accessibility inform
spatial characteristics relating to intelligibility and levels of occupancy
and movement. 

Moving away from linear or planar representations (both topological),
towards a point representation introduces metric properties within the
configurational description. As Alasdair has pointed out, we need to account
for the fact that larger spaces which have more (and more inter-connected)
points achieve higher isovist integration values.   

This is one of the reasons why in urban systems Isovist Integration gives
good results but usually not better than Axial Integration analysis (except
of course in cases where strategic long routes are also wider and
connectivity works in accordance). The Axial Map captures properties of
urban space relating to movement within the logic of its representation. 
It may be possible to develop the methodology further to go from Visibility
Graphs to a topological description.

What seems to be important at this stage is to understand what the method
adds to spatial descriptions and how the results are meaningful. 
VGA can be applied to any kind of spatial system and can differentiate
configurational and local properties between locations, based on visibility
or accessibility relations. However, it is not expected to explain all
issues relating to patterns of space use and the way we experience our
surroundings. 
It will, for example, analyse open plan buildings, where convex and axial
descriptions are not uniquely defined. 
However, it may not give as good a correlation with movement or occupancy as
axial configurational analysis invariably in large urban systems.  
It will not show us differences between adjacent locations on a closed
rectangular space (to use Kayvan's example), even though it is not purely
topological, since in terms of direct visibility or accessibility all
locations are equal. 
Likewise, it will not explain why people often chose to sit with their backs
at a corner in a space, because the isovists in VGA are taken at 360
degrees, not 180 to 110 of a person's visual field at a time (which would
give a corner a higher control).
 
It may soon give us the opportunity to analyse and understand how complex
buildings work in terms of intelligibility in section, as well as in plan,
and to understand visibility relations in three dimensions  (so far VGA
deals only with 2-dimentional plans, although we have analysed configuration
based on accessibility in multilevel systems by linking vertically across
e.g. the Wales Millennium Centre, SSL, 1999).

VGA may also provide a ground for understanding better linear and convex
spatial descriptions and configurational relations. 
It seems to me that the development of VGA does not replace any other method
of analysis, but is perhaps an opportunity for rethinking the role of
methodology in spatial analysis, learning through testing and evaluating,
understanding the limitations and the value of experience gained by its
application. Hopefully with the availability of VGA on a PC platform there
will soon be many  interesting applications and innovations in the field. 

 

To add to Jake's interesting links, 
more examples of the use of Visibility Graph Analysis in research
consultancy and design can be found at
http://www.bartlett.ucl.ac.uk/spacesyntax/news/MovBroBuy_forhtml_070400/MovB
roBuy_forhtml_070400_files/frame.htm and
http://www.spacesyntax.com
 
Also, an application of VGA for the comparative analysis of interior public
spaces in cultural building from my MSc report at
http://users.hol.gr/~glory/rfh_nt_spatial3.htm 
And http://users.hol.gr/~glory/consultancy3.htm shows more examples of the
applications of VGA in research and consultancy in Space Syntax Laboratory.



Maria Doxa

[log in to unmask]



%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%