Hi,
a general response. The process of representation of two dimensional open space for the purposes of the kind of analysis space syntax does has several different drivers.
1. practicality - one could imagine a very perfect form of representation that was so time consuming and costly to carry out that people simply would not do it and so research would not take place. Maps might be drawn, but too small for what was really needed, or comparative databases of different cities just would not cover enough cases.
2. repeatability - ideally if two different scientists were given the same town to analyse their representations should be the same and give the same results.
3. logical plausibility - there may be a relationship between the form of representation and the kind of measure of the resulting network being used. In the case of angular measures of cost in the network this implies a representation from the 2D open space morphology which is optimised in terms of angular cost reduction from all point to all other points. A global optimisation.
As a researcher the main trade off is between 1. and the other two. Now that road centre line data are fairly comprehensive in their coverage, they certainly help achieve practicality. However, they are not defined for the purposes of these kind of analysis and the morphology of the RCL network can be rather arbitrary especially in the treatment of junctions. This routinely requires a lot of tidying up and the way that tidying happens is very hard to guarantee 2. repeatability. So far as 3. logical plausibility, is concerned only a full axial analysis can guarantee minimum angle change routes from all points to all points, however it is time consuming and if you are using automated line generation, also memory and processing power intensive (it is a global optimisation problem after all).
Now, the differences in actual resulting analyses between RCL analysis and full axial is in general only a very small one, and especially so as maps get larger, hence the drift toward practicality and away from logical plausibility and repeatability as drivers.
Alan Penn,
> On 26 Oct 2014, at 12:32, SUBSCRIBE SPACESYNTAX Anonymous <[log in to unmask]> wrote:
>
> Thanks Jorge and Bin for your responses. Just to share what I finally get from this discussion and my reading (may not be correct though).
> To do angular analysis using road-centre lines map (or in general):
> - Transform road-centre line map into ‘natural streets’ in ArcMap using Axwomen extension, and produce the ‘axial lines’ based on it (of note, this is totally a new ‘definition’ of axial lines which has been introduced by Xintao Liu and Bin Jiang in a paper entitled “Defining and generating axial lines from street center lines for better understanding of urban morphologies).
> - Export the ‘axial lines’ to MIF.
> - Import MIF into Depthmap, and transform to ‘Axial map’, then transform ‘Axial map’ into ‘segment map’. And, run the angular analysis.
> Of note, in this process, I found the first step to be conceptually really ‘important’; as it is a new definition of Axial lines. Any thought about this step? I mean, ‘how’ we decide about ‘axial lines’. Though, I understand, this is a critical issue in space syntax (and many people criticize us on this!). Professor Bin Jiang already showed in that paper that the new definition of axial lines are more suitable (at a city scale) than the old ones.
> I would be really appreciate it to know how you guys do these steps in doing an angular analysis, especially how you define axial lines (you draw by hands based on visibility, or you use any auto process, or you use the new definition I mentioned above) (maybe this could help to reach some consistency among space syntax researchers :)
> Thanks,
|