Dear Kimon,
Thanks for your email. The problem you raise in the rural areas is referred to as “flipping depth” and is discussed in the appendix
of the paper Hillier, B., Yang, T., Turner, A., 2012, “Normalising least angle
choice in Depthmap and how it opens up new perspectives on the global and local analysis of city space”, in The Journal of Space Syntax, vol. 3, issue 2, p. 155-193.
But I will have a closer look at your data next week and discuss it in the group and then hopefully get back to you with an answer.
Kind regards,
Meta
Meta Berghauser Pont
Associate Professor Urban Design
Spatial Morphology Group (SMoG)
Chalmers University of Technology
Department of Architecture and Civil Engineering Sven Hultins gata 6 (room SB-K 395)
SE-412 96 Gothenburg, Sweden
+46 (0) 73 4233637
Dear Meta,
Thanks a lot for sharing the newest version of PST for QGIS 3.- with everyone.
It really is a fantastic tool, particularly when it comes to computation time!
I would like to follow up on Yuru’s question about the difference between depthmapX and PST.
As you’ve mentioned; PST allows the same space syntax analysis based on the axial map (or angular analysis based on the road-center-line map).
I’ve run a couple of Angular Integration analyses in QGIS using PST and the model of Gothenburg from your tutorial data and came a cross some rather unexpected results.
My expectation was that PST Angular Integration would give the same results as depthmapX Angular Segment Integration, however, they appear to not be comparable (r2 < 0.04 and 0.5 respectively).
Please see figure 1 below:
As far as I understand from the PST manual, the user is forced to use either of two ’normalisation’ methods (1. Turner 2007 and 2. NAIN).
I briefly read Turner 2007 paper cited in the manual (
https://journals.sagepub.com/doi/10.1068/b32067)
and he mentioned two measures for integration (or closeness) 1. standard angular closeness and 2. weighted angular closeness (page 544):
Compared to the PST Angular closeness formula (page 11):
The above appear consistent with the formula generally referred to as Closeness Centrality (
https://en.wikipedia.org/wiki/Closeness_centrality)
a normalised form of Bavelas (1950) Closeness, multiplied by N - 1 (Turner seemed to have dropped the -1?), where N is the number of nodes in the graph. PST also includes a 1+ in the divisor to avoid dividing by 0.
However, if it is the same ’standard angular closeness’ Turner referred to in his paper, that is implemented in PST, then why are the results different (r2 = 0.036)? I am sure I am missing something here.
A visualisation of the three measures shows that the PST results have high values in more ‘rural’ areas–at the edge of the system–whereas depthmapX does not. This is also an observation that I encountered in other models that I have tested.
The folder includes a high resolution versions of the figures (GOT_comp_1.png & linear_regression.png), a .CSV file with all data, including the geometry as x1y1x2y2 (GOT_comp.csv), as well as a MapInfo file to be opened in QGIS or depthmapX (GOT_comp.mif &
.mid).
Please note: I have renamed the depthmapX variable in the MapInfo file to Integration_R2K.
Faculty of the Built Environment
University College London UCL
22 Gordon Street
London WC1H 0QB
To unsubscribe from the SPACESYNTAX list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=SPACESYNTAX&A=1
