Evolving Configurational Properties
article published at 12th Space Syntax Symposium at Jiaotong University, Beijing, China
SIMULATING MULTIPLIER EFFECTS BETWEEN LAND USE AND MOVEMENT PATTERNS
Martin Bielik 1, Reinhard Koenig2,3 , Ekaterina Fuchkina1, Sven Schneider1, Abdulmalik Abdulmawla1
1 Bauhaus University, Weimar, Computer Science in Architecture
2 Bauhaus University, Weimar, Computational Architecture
3 Digital Resilient Cities, Austrian Institute of Technology
Abstract
In this paper, we introduce simulation framework for examining the effect of street network configuration on the evolution of the relationship between movement and land use allocation over time. The causal chain introduced in Space Syntax literature suggests that the potential generated by spatial configuration of a street network influences how people move and that these movement flows attract specific types of land uses. These land uses generate in turn additional movement creating an endless cycle of mutual interactions. In Space Syntax, this interaction between movement flow and land use is assumed to work in a positive feedback loop, multiplying the initial potentials given by a street network configuration. The practical consequence of this hypothetical assumption for the Space Syntax method is that the outcome of the feedback loop can be predicted as multiplication of the initial state and therefore doesn’t have to be simulated.
In this paper, we introduce a computational method for testing the multiplier effect hypothesis and identify the cases in which it holds true and those ones in which more detailed investigation considering feedback loops might be necessary. We demonstrate how such investigation based on the simulation of the interactions between movement and land use in time can be operationalized and conduct series of studies exploring the spatio-temporal effects of street network configuration.
We conclude that these exemplary studies show how the presented simulation model can be used to test the core assumption behind the Space Syntax method. We also offer preliminary insides about when and under which conditions it can be reliably applied and when system dynamic simulation might be necessary to predict not only the immediate, but also the long-term effects of street network configurations on centrality, movement and land use distribution.
Keywords
Street network configuration, System dynamics, Centrality, Movement, Land use, Multiplier effects, Natural movement
One sentence summary:
We explore the evolution of the configurational effects of urban form on the movement and land use allocation over time.
Paper overview
In this section, we compiled a short review of the main ideas discussed in the paper. The purpose is to get a general idea that can be deepened by reading the full paper attached below.
1. Background | What do we already know?
A)
“Street network configuration affects how people move”
(Golledge 1995)
B)
“Street network configuration affects where human activities allocate”
(Hansen 1959)
C)
“Movement influences allocation of land uses and allocation of land uses influence movement”
(Hillier 1996a)
D)
“Space Syntax is about bringing the configuration, movement and allocation of activities together”
(Hillier 1996b)
2. Research gap | What do we don’t know?
“Does the effect of street network configuration on movement and land use allocation evolve over time?”
OPTION A)
“The effect of configuration on movement and land use is stable over time, it DOES NOT EVOLVE“
OPTION B)
“The effect of configuration on movement and land use does DOES EVOLVE over time”
Consequences
A) = SIMPLE TO PREDICT
B) = HARD TO PREDICT
3. Research question | What do we want to know?
Question A
Under which conditions is the result of the feedback loop between movement and land use distribution predictable as a function of the multiplier effect?
Question B
Under which conditions does the natural movement potential of the network configuration correct the effect of unequal distribution of land uses (disruptions) throughout the feedback cycle?
“Space Syntax method is based on the assumption that option A holds true. In other words, it is assumed that the outcome of the interaction between land use and movement can be predicted without simulation.”
This assumption is based on:
1. Multiplier effect hypothesis
The result of the feedback loop between movement and land use distribution is predictable as a function of the multiplier effect
(Hillier 1996b)
2. Natural movement hypothesis
The natural movement potential of the network configuration corrects the effect of unequal distribution of land uses (disruptions) throughout the feedback cycle
(Hillier 1999)
4. Research method | Spatio-temporal simulation engine
Left ) UML diagram of the Space Syntax underlying urban dynamics model. a) current model of the feedback loop between movement and land use. b) displays the “multiplier effect” hypothesis.
Right) Exemplary illustration of two iterations of our dynamic simulation. The initial state is an equally loaded network. The centrality values show the angular betweenness centrality with radius Rn (global radius).Value range: high movement potential = red, low movement potential = blue
5. Results | Simulation experiments
We exemplary explore the impact of different conditions on the multiplier and natural movement effect. The conditions explored in this study relates to how we model the effect of land use on movement and what is the initial distribution of land use in the simulation.
Explored conditions & model parameters:
- Definitions of movement potential centrality MPC.
Closeness vs. Betweenness centrality - Analysis radius
Local radius of 600m vs. global radius - Initial land use weightings w
Equally loaded vs. Disrupted network*
For the following experiments, we used the street network of the inner city of Weimar as an exemplary case study area. Weimar is a medieval, mid-size German city with approximately 60.000 inhabitants. The size, historical development and the overall variety of street network patterns makes it a good candidate for testing the proposed simulation model. On the one hand, the size of the city makes it possible to quickly calculate many iterations of the simulation, but on the other hand it is large and diverse enough to let non-trivial configurational patterns emerge.
*Disrupted network in context of this paper means that in the initial stage of the simulation, not all street are loaded with equal amount of land use. In other words some streets produce more movement than others, which is represented by higher weighting of selected graph vertices.
Experiment 1 | Measuring the multiplier effects on equally loaded network
Equally loaded network in context of this paper means that in the initial stage of the simulation, all street are loaded with equal amount of land use. In other words all streets produce same amount of movement, which is represented by equal weighting of all graph vertices.
E.1a
Conditions:
- Equally loaded network
- Betweenness centrality
- Global radius
E.1c
Conditions:
- Equally loaded network
- Betweenness centrality
- Local radius
E.1b
Conditions:
- Equally loaded network
- Closeness centrality
- Global radius
E.1d
Conditions:
- Equally loaded network
- Closeness centrality
- Local radius
Experiment 1 | Summary
Quantifying the multiplier effect by measuring linear relationship between the movement pattern at the first and the last iteration of the simulation
- Multiplier effect is reliable predictor for:
i) global closeness centrality
ii) global betweenness centrality - Strong effect of scale of the centrality model
Multiplier effect Global > Multiplier effect Local
Experiment 2 | Measuring the multiplier effects on disrupted network
Disrupted network in context of this paper means that in the initial stage of the simulation, not all street are loaded with equal amount of land use. In other words some streets produce more movement than others, which is represented by higher weighting of selected graph vertices.
E.2a
Conditions:
- Disrupted network
- Betweenness centrality
- Global radius
E.2c
Conditions:
- Disrupted network
- Betweenness centrality
- Local radius
E.2b
Conditions:
- Disrupted network
- Closeness centrality
- Global radius
E.2d
Conditions:
- Disrupted network
- Closeness centrality
- Local radius
Experiment 2 | Summary
Quantifying the multiplier effect by measuring linear relationship between the movement pattern at the first and the last iteration of the simulation
- Multiplier effect is NOT reliable predictor
- Doesn’t pick up most central streets*.* Non normality & heterocsedascity
Experiment 1&2 | Measuring the natural movement effect
Quantifying the natural movement effect by measuring linear relationship between the movement pattern at the first iteration of equally loaded network and last iteration of the disrupted network.
- Natural movement is reliable predictor for:
i) global closeness centrality
6. Conclusions | Joined multiplier and natural movement effect
Evaluating the impact of scale, initial network loading, and type of centrality measure on natural movement and multiplier effect. In other words, under which conditions is the influence of configuration on movement predictable without simulations and when the simulation is needed.
NO SIMULATION IS NEEDED:
- For global closeness and betweenness centrality in equally loaded networks
- For global closeness and in disrupted networks
SIMULATION IS REQUIRED:
- When running local models
- When running models on disrupted – non equally loaded networks
7. Outlook | Next steps
A) Testing the simulation results against empirical longitudinal data
B) Testing the sensitivity of multiplier and natural movement effect on conditions related to street network configuration and type of land use (i.e. does multiplier effect depends on the type of the street network layout and how we model the impact of movement on land use).
Download the article
Bielik et al. – 2019 – Evolving configurational properties.pdf
Cite the article
Supplementary materials
Grasshopper scripts
Following grasshopper script simulates the interaction between movement flows and land use distribution through street network configuration. The street network used in the example is simplified example of city of Weimar, Germany which was used as an example in the paper. The data for the street network is directly embedded in this script.
Versions of the script:
1. Simplified script, easy to run. This script requires only DeCodingSpaces toolbox and Anemone plugins being installed (see instructions bellow)
evolving configurational properties_simple.gh
2. Full script. This script requires also R language additionally to DeCodingSpaces toolbox and Anemone plugins being installed (see instructions bellow). This version of the script offers some additional visualization and possibility to save the simulation results
evolving configurational properties_full.gh
NOTE: Get sure the GPU acceleration is enabled.
(otherwise is the simulation too slow to execute in real time)
NOTE: Check the requirements section bellow and get sure you get all the plugins installed using the links on this website.(otherwise the version might not fit the version of the script)
Requirements
- Windows 10
- Rhino 6 (64bit)
- Required Grasshopper plugins
Install instructions
Unblocking plugins
After downloading the RequiredGHPlugins_SimAUD19.zip file, check if its unblocked before extracting the zip archive. Right click on the file > Properties > select unblock > select ok
Install components
After unlocking and extracting the RequiredGHPlugins_SimAUD19.zip archive, copy the “SimAUD19 components” folder into the grasshopper component folder. The grasshopper component folder can be found at:
C:\Users\YourUserName\AppData\Roaming\Grasshopper\Libraries
or via grasshopper file menu:
Install user objects
After unlocking and extracting the RequiredGHPlugins_SimAUD19.zip archive, copy the “SimAUD19 user objects” folder into the grasshopper UserObjects folder. The grasshopper component folder can be found at:
C:\Users\YourUserName\AppData\Roaming\Grasshopper\UserObjects
or via grasshopper file menu:
Enable GPU acceleration
The requirement to run the GPU accelerated street network analysis is the CUDA platform enabled NVIDIA GPU.
If this requirement is fulfilled, you have to copy the folders “Alea.CUDA.CT.LibDevice” and “Alea.CUDA.CT.Native.X86.B64.Windows” from the GPU acceleration folder to your Rhino install folder (i.e. “Program Files\Rhinoceros 6 (64-bit)\System\”).
