Dear Martin,
You are correct that there are some inconsistencies in terminology between
many network analysis papers and space syntax. A consequence I suspect of
convergent evolution amongst scientific communities working in relative
isolation from one another in their early years, but also a result of any
area of science inventing terminology that relates to its own field of work.
The terms 'choice', 'control', 'integration' and 'segregation' are used
partly because their meaning relates directly to the aspects of spatial
morphology we study. We use the word 'connectivity' where others would use
'valence' (USA) or 'valency' (Europe) or 'degree' - I don't think that these
things will change, and in fact I think it would be sad if they did as they
reflect something of the history of the ideas.
There are also practical issues as Drew has noted in "ground truthing" plans
of what exactly can be seen or moved across, especially in more open soft
landscapes. These are not actually problems with the axial definition
itself, just the practical problems facing any researcher using any
methodology who wants to analyse 'what one can see, or where one can go'.
The beauty of the axial mapping method is that once you have decided on a
set of boundaries of what areas of space can be moved across and what
cannot, perhaps for different groups of people (wheel chairs and push chairs
may avoid certain surfaces or gradients for example), and perhaps in
different seasons, you can then follow a single and consistent methodology
in producing the network representation itself. In this way the results of
analysis of two different spatial boundaries - with different morphology and
topology - can be directly compared. It is just this that allows these
methods to treat 'what you can see' and 'where you can go' as separate
variables in an analysis, as well as treat the 'where you can go'
differently for different groups of the population, different times of year
etc.
The axial method itself is pretty well defined in comparison to other
mapping methodologies, see for example: Turner, A. and Penn, A. and Hillier,
B. (2005) An algorithmic definition of the axial map. Environment and
Planning B: Planning and Design, 32 (3). pp. 425-444. ISSN 02658135
available at
http://eprints.ucl.ac.uk/archive/00000624/
More importantly it has shown consistent results in correlating with
observed movement.
Now, you ask a question that makes me think you might be getting something
wrong: "how do you vectorise your axial map :)) triple intersections are a
mess!" - in an axial map all intersections are pairwise between two lines.
The intersection translates to a link in the graph. The process is simply
one of digitising the axial lines, saving as .dxf and importing into
Depthmap, or just drawing the lines directly into Webmap, Axman or pretty
much any of the other axial analysis packages. I don't quite understand your
'triple intersection' problem... happy to help if I can though.
All the best,
Alan Penn
Martin Tomko wrote:
> Dear Drew,
> thank you very much for your email and the pointer to your thesis. I
> will have a look at it.
> I wonder if you found about my research from the recent talk of Alex
> Klippel at UCSB or just through the web.
> I am currently exploring the usability of space syntax measures (which I
> found weakly defined and terminologically inconsistent with many of the
> traditional network analysis papers, unfortunately...). I will be
> looking at the connectivity in the network to reconstruct a more natural
> hierarchy of the network to be used in granular route directions.
> As you say, axial maps have a mjor problem with the definition of the
> vista spaces. These can further be inconsistent in time (seasons vs.
> foliage, temporarily obstructions,...), further problems arise from the
> handling of the intersections of the vistas (I think this is not only my
> problem, but how do you vectorize your axial map :)) tripple
> intersections are a mess!, and these are still simple ones...). length
> of the vista space, + you can look at the use of 3D models, terrain
> models, angle of sight,... but hey, one cannot solve all the world
> problems... :)
>
> Please, keep me up to date with your progress, I will be glad to discuss
> mine as well.
> Kind regards
> Martin
>
> ----------------------------------------------
> Ing. Martin Tomko
> PhD. candidate
>
> CRC for Spatial Information
> Department of Geomatics
> University of Melbourne
> Victoria 3010
> Australia
>
> phone +61 3 8344 9179
> fax +61 3 9349 5185
> email [log in to unmask]
> url http://www.geom.unimelb.edu.au/tomko
> url http://www.spatialinformationscience.org
> url http://www.crcsi.com.au
> ----------------------------------------------
>
> Drew Dara-Abrams wrote:
> > Hello,
> >
> > Last year I completed a senior thesis that uses the case study of an
> > American college campus to compare space syntax measures of urban form
> > with behavioral measures of spatial judgment and memory. I found that
> > the global integration of a location on the campus predicts students'
> > accuracy at recalling that location when asked under controlled
> > conditions. The document is available on-line as a PDF, along with a
> > brief summary handout, a demo of the spatial judgment and memory tasks,
> > and the axial map that I used:
> >
> > http://drew.dara-abrams.com/research/
> >
> > I would appreciate any questions, comments, or suggestions you may have.
> >
> > I am currently revising this work at University of California, Santa
> > Barbara, and readying it for publication. While preparing to run the
> > study on this campus, I have been wondering about how to best use axial
> > maps and visibility graphs to characterize outdoor settings that are not
> > simply open streets and solid buildings. For instance, the university
> > campuses that I am considering are filled with landscaping as well as
> > roadways, some of which can be walked across and some of which cannot be
> > traversed but can be seen across. I am planning on doing two sets of
> > models, one that describes spaces that are accessible/can be walked
> > across and a second that describes spaces that are visible/can be seen.
> > As I am doing this, I am trying to assemble some rough guidelines for
> > which physical features to ignore and which to include in space syntax
> > models when "ground truthing" CAD plans, but I don't want to duplicate
> > work that may have been done previously. I would greatly appreciate any
> > literature references or other suggestions on methodology.
> >
> > Thank you for your time,
> > Drew Dara-Abrams
> >
|