Correlation between Space Syntax and Vehicle Traffic Volumes

Page 1

Correlation between Space Syntax and Vehicle Traffic Volumes


Made in Milan by Transform Transport Š 2018, Systematica Srl All studies presented in this book are developed by Transform Transport. All rights reserved. Unauthorised use is prohibited. No part of this publication may be reproduced in any form or by any means without the written permission of Systematica Srl.

Team: Filippo Bazzoni, Rawad Choubassi, Efrem Errera, Francesco Scarnera A special thanks to all collaborators of Systematica who contributed to this book.

Systematica Srl Transport Planning and Mobility Engineering

Milan Beirut Mumbai

Via Lovanio, 8 20121 - Milan Italy

T + 39 02 62 31 19 1 E milano@systematica.net

www.systematica.net


Table of contents 1. INTRODUCTION Research goal Four step traffic model (CUBE) Space Syntax and spatial analysis

p. 03 p. 04 p. 05 p. 06

2. TRAFFIC MODEL OF MILAN Traffic volume estimate Available data in the model

p. 07 p. 09 p. 11

3. SPATIAL ANALYSIS OF MILAN Different outputs of spatial analysis Network choice analysis Choice analysis at different distances Normalized measure of choice (NACH)

p. 13 p. 15 p. 17 p. 19 p. 21

4. CORRELATION BETWEEN TRAFFIC MODEL AND SPACE SYNTAX Methodology Outputs of NACH*road capacity Correlation between NACH*Capacity and Traffic Volumes (CUBE)

p. 25 p. 27 p. 29 p. 33

5. CONCLUSIONS AND NEXT STEPS

p. 37

6. REFERENCES

p. 39


Traffic models are extensively used to forecast future traffic conditions with different mobility patterns, land uses or network configurations. Their robustness and reliability is well acknowledged and the majority of national, regional and urban mobility plans are done with the support of this tool. However, their implementation requires a rigorous approach that is time consuming and fairly expensive, as it grounds on large amount of data collection and specialist works, so that not every Municipality or stakeholder can make use of this instrument.


Spatial configurations of networks are also being increasingly used to preliminarily assess potential movements of vehicles and pedestrians. This methodology, born approximately thirty years ago in the UK, is arguing that route choice is strongly affected by the configuration of the space and by the way the different parts of complex system are related to each other, according to the same rules that are applied to mathematical graph theories. This research is aimed to test to extent to which this is true for the Metropolitan area of Milan.


CUBE_Traffic Volume

Space Syntax _Choice

Space Syntax_Normalized Choice

Space Syntax_Normalized Choice*Road Capacity


To what extent road network configurational characteristics can anticipate vehicle traffic movements? The aim of this research is to measure the degree of correlation between predicted traffic volumes (with CUBE software by Citilabs) and the configurational values of the same network depicted through spatial network analysis (using DepthMap software by UCL). The research focuses on the City of Milan since the researchers have access to a fully reliable and well calibrated traffic model, as well as an excellent knowledge of the city. There are two main findings: first, traffic volumes have higher correlation with choice values (calculated with DepthMap) once these are normalized. Second, a significant correlation is found between traffic volumes and normalized choice multiplied per road capacity; this value could find useful applications for traffic prediction analysis.

Examples of elaborations using CUBE and Space Syntax data

02


Traffic Volumes_USA

Pedestrian Movement Analysis

Š Citilabs

Traffic Volumes_Torrance


Traffic Model

A transportation model is the instrument to support decisions when planning and designing transportation systems. Macroscopic forecast modelling platforms are the most powerful tool providing advanced methods for transportation modelling: urban, regional and long distance demand forecasting and assignment. These software are commonly used for developing standard four-step multimodal traffic models (trip generation, trip distribution, mode choice, and assignment). Trip generation models are typically based on demographic data; Distribution models predict trips from homes to destinations such as work places, retail facilities, and other homes; Mode choice models predict the split of travel choices (private car, public transport, etc.); Assignment models predict the actual route taken, given an origin, a destination, and a mode of transport.

Example of four-step model representing traffic volume and pedestrian movements 04


Agent analysis

Segment analysis

Š UCL DepthMap 10

VGA analysis

Axial analysis


Convex analysis_Connectivity

Space Syntax Spatial Analysis

Space syntax investigates relationships between spatial layout and a range of social, economic and environmental phenomena. These phenomena include patterns of movement, awareness and interaction; density, land use and land value; urban growth and societal differentiation; safety and crime distribution (SSN, 2018). From a transport perspective, the most innovative factor of this discipline consists in describing convenient movements not just in terms of time or distance, but also the angular deviation from a straight route from origin and destination. The shortest path is therefore determined not only by metric distance, but also by relative cost of turning, whereby greater turns are costlier: this is confirmed by more than three decades of observations and studies, still ongoing, that generally people tend to move preferring intuitiveness to distance, especially on foot.

Example of Space Syntax analysis (Agent, VGA, Convex, Segment, Axial)

06


CHAPTER 2

Traffic Model Of Milan

Space Syntax and Traffic Volumes

November 2018



Traffic Volume Prediction

The image on the right shows the traffic model of Milan metropolitan area used within this research. This model is a reliable tool that is calibrated to replicate existing traffic conditions with great accuracy. The roads with higher traffic flows are shown in red, primarily the radial and ring road highway system, while in yellow the local roads with lower traffic loads.

09

Traffic volumes Highest Lowest



Data available in the Model

Traffic Volume

Within the model, a variety of data are associated to each road such as width, speed, travel direction, type of road, etc. In particular, within this study the following is considered: Traffic Volumes which represent the amount of vehicles passing through different roads; Road Capacity which represents the maximum vehicle capacity for each road; Road Speed which represents the maximum posted speed limit of different roads.

Traffic volumes 8496 (Highest) 0 (Lowest)

11


Road Capacity

Road Capacity

Speed

Speed

10000 (Highest)

130 (Highest)

0 (Lowest)

0 (Lowest)


CHAPTER 3

Spatial Analysis Of Milan



Different Outputs of Spatial Analysis

Street Network Choice

Within Space Syntax discipline, there is a wide range of configurational parameters that can describe each link of a network. The images on the right show the three main parameters that are currently being used in the majority of the studies. In particular: Choice measures movement flows through spaces. Spaces that record high global choice are located on the shortest paths from all origins to all destinations. Choice is a powerful measure for forecasting forecasting potential pedestrian and vehicle movements. Integration is a measure of mean depth that is specifically adapted for architectural layouts. The global measure shows how deep or shallow a space is in relation to all other spaces. Using integration, spaces are ranked from the most integrated to the most segregated. Integration is usually indicative to how many people are likely to be in a space, and is thought to correspond to rates of social encounter and retail activities (Hillier, 1996a). Connectivity (degree) measures the number of immediate neighbors directly connected to a space or road.

Street Network Choice Most chosen path Least chosen path

15


Road Capacity

Integration

Connectivity

Connectivity

Highest

Highest

Lowest

Lowest


Network Choice Analysis

Network Choice depicts the degree to which the streets form a part of the shortest trip from all origins to all destinations. movement: these streets are inherently important parts of many different journeys. Choice is computed as follows:

Angular Choice reveals the streets that are inherently on many shortest paths, when considering all origins and all destinations within a certain network distance radius. The shortest path is determined not by metric distance, but by relative cost of turning, whereby greater turns are more costly. The spatial network analysis at a distance=n (taking all segments of the network as both possible origins and destinations) highlights the most chosen paths inside the network in red and the least chosen paths in the network in blue. Higher choice value

17

Lower choice value

Street Network Choice Most chosen path Least chosen path



Choice Analysis at Different Distances

Choice 500meters

By changing the scale for choice analysis, results are different depending on the applied distance, describing different configurations that are ideal for different travel modes. Choice distance = 500meters Analysis at very small scales highlight those streets, and particularly junctions, that are ideal for community services, small local businesses that require high pedestrian footfall.

Choice 2500meters

Choice distance = 1000meters This is the ideal distance everyone is willing to walk on foot, and where walking is much more convenient than all other modes; the potential areas for being center of neighborhood activities are identified. Choice distance = 2500 meters Increasing the distance of the analysis is possible to identify the most convenient routes for bicycles and other small electric vehicles, understanding strategic corridors to connect adjacent districts in a sustainable manner. Choice distance = n meters (infinite distance) possible to preliminarily identify arteries convenient for long distance movements, therefore most likely by car or by bus at urban scale. Street network Choice Most chosen path

Least chosen path

19


Choice 1000meters

Choice n meters


Normalized Measure of Choice (NACH)

Correlation (r2) between Choice and Traffic Volume = 0.0990 (at distance=n) Choice TRAFFIC VOLUME (CUBE) 140% 120% 100% 80% 60%

R² = 0,099

40% 20% 0% 0

Normalized choice aims to solve the paradox that segregated designs add more total (and average) choice to the system than integrated ones. It divides total choice by total depth for each segment in the system. This adjusts choice values according to the depth of each segment in the system, since the more segregated it is, the more its choice value will be reduced by being divided by a higher total depth number. This would seem to have the effect of measuring choice in a cost-benefit way (Hillier et al., 2012).

2000

4000

6000

8000

10000

Correlation (r2) between NACH and Traffic Volume = 0.1727 (at distance=n) NACH - TRAFFIC VOLUME (CUBE) 140%

R² = 0,1727

120%

Normalized Angular Choice NACH is defined as follows;

100% 80% 60% 40% 20% 0% 0

2000

4000

6000

8000

where (i, x, j)=1 if the shortest path from i to j passes through x and 0 otherwise.

Finding #1: NACH as higher correlation with Traffic Volumes

21

10000


Space Syntax model representing NACH values at distance=n

NACH Highest Lowest


Graphical representation of standard values of choice is significantly different to values generated by normalized measure of choice (NACH). The latter representation is, in fact, more coherent with real traffic distribution.

Choice_n

Street Network Choice Most chosen path Least chosen path

Space Syntax and Traffic Volumes

November 2018


NACH_n

Street Network NACH Highest Lowest


CHAPTER 3

Correlation between Traffic Model and Space Syntax

Space Syntax and Traffic Volumes

November 2018



Methodology Extracting CUBE Predicted Traffic Volumes

Calculating Space Syntax Choice Values at different Distances The methodology applied to test correlation between forecasted traffic volumes and Space Syntax values consists in five different steps: First step. Extracting traffic volumes from the macroscopic traffic model and calculating choice values in Space Syntax at different distances; Second step. Adding variables to choice values (Road Capacity, Speed and Segment Length) obtained from the macroscopic traffic model; Third step. Calculating normalized measures of choice (NACH) and adding variables; Fourth step. Calculating correlation between values; Fifth step. Analyzing results through chart/data and graphical comparison.

(500, 1000, 2500, 5000, 10000, 15000, 20000, n meters)

Calculating Normalized measure of Choice (NACH)

Analysis of Results (Chart/Data and Graphical Comparison)

27


Choice*Road Capacity

Adding Variables to Choice Values

Choice*Speed Choice*Segment_Lenght

Calculating Correlation (r2)

NACH*Road Capacity

Adding Variables to NACH Values

NACH*Speed NACH*Segment_Lenght

METROPOLITAN AREA (Correlation 0<r2<1) Distance r:

Choice Choice x x road speed capacity

Choice x segm. lenght

NACH x Choice road capacity

NACH x speed

NACH x segm. lenght

NACH

500

0.0126

0.0223

0.0149

0.0291

0.1948

0.0402

0.0016

0.0181

1000

0.0298

0.0510

0.0354

0.0671

0.4388

0.2183

0.0092

0.001

2500

0.0583

0.0930

0.0758

0.1168

0.6315

0.4752

0.0402

0.0391

5000

0.0768

0.1157

0.0933

0.1376

0.6898

0.5643

0.0582

0.102

10000

0.0992

0.1494

0.0908

0.1723

0.7086

0.6007

0.0668

0.1479

15000

0.0962

0.1436

0.0835

0.1641

0.7123

0.61

0.0685

0.162

20000

0.0909

0.1347

0.0799

0.1530

0.7129

0.6132

0.0696

0.1668

n

0.0865

0.1276

0.0791

0.1444

0.7105

0.6126

0.0688

0.1727

28


NACH*Road Capacity Outputs NACH*Road Capacity presents the highest correlation values with forecasted traffic volumes

Several tests were made at different calculation distances. Analysis of results is interesting if we look at charts comparison and graphical restitution. The charts on the right represent only correlation values for the NACH*Road Capacity column (which show the highest values). Graphical representation of results is given in the following page.

METROPOLITAN AREA (Correlation - 0<r2<1)

Choice x road capacity

Choice x speed

Choice x segm. lenght

Choice

NACH x road capacity

500

0.0126

0.0223

0.0149

0.0291

0.1948

1000

0.0298

0.0510

0.0354

0.0671

2500

0.0583

0.0930

0.0758

5000

0.0768

0.1157

10000

0.0992

15000

NACH x segm. lenght

NACH

0.0402

0.0016

0.0181

0.4388

0.2183

0.0092

0.001

0.1168

0.6315

0.4752

0.0402

0.0391

0.0933

0.1376

0.6898

0.5643

0.0582

0.102

0.1494

0.0908

0.1723

0.7086

0.6007

0.0668

0.1479

0.0962

0.1436

0.0835

0.1641

0.7123

0.61

0.0685

0.162

20000

0.0909

0.1347

0.0799

0.1530

0.7129

0.6132

0.0696

0.1668

n

0.0865

0.1276

0.0791

0.1444

0.7105

0.6126

0.0688

0.1727

Distance (m):

29

NACH x speed


Distance=500m

Distance=1000m

120%

120%

100%

100%

80%

80%

60%

60%

40%

R² = 0,1948

R² = 0,4388

40% 20%

20%

0%

0% 0

2000

4000

6000

8000

0

10000

2000

4000

6000

8000

10000

Predicted Traffic Volume

Predicted Traffic Volume

Distance=2500

Distance=5000m

120%

120%

100%

100%

80%

80%

60%

R² = 0,6315

40%

40%

20%

20%

0%

R² = 0,6898

60%

0% 0

2000

4000

6000

8000

10000

0

2000

Predicted Traffic Volume

4000

6000

8000

10000

Predicted Traffic Volume

Distance=10000m

Distance=15000m

120%

120%

100%

100%

80%

80% R² = 0,7086

60% 40%

40%

20%

20%

0%

R² = 0,7123

60%

0% 0

2000

4000

6000

8000

10000

0

2000

Predicted Traffic Volume

4000

6000

8000

10000

Predicted Traffic Volume

Distance=20000m

Distance=n

120%

120%

100%

100%

80%

80% R² = 0,7129

60% 40%

40%

20%

20%

0%

R² = 0,7105

60%

0% 0

2000

4000

6000

Predicted Traffic Volume

8000

10000

0

2000

4000

6000

Predicted Traffic Volume

8000

10000


Distance=500m

Distance=1000m

Distance=10000m

Distance=15000m

NACH*Road Capacity Highest Lowest

Space Syntax and Traffic Volumes

November 2018


Distance=2500m

Distance=5000m

Distance=20000m

Distance=n

GIS elaborations representing NACH*Road Capacity values at different distances


Correlation between NACH*Capacity and Traffic Volumes The correlation between NACH*Capacity values and Traffic Volumes reach results that start from a minimum correlation value (r2 ) of 0.1948 (in case of distance=500m) to a maximum of 0.7129 (in case of distance=20000m).

NACH*CAPACITY (distance=20000m) 120% 100%

Generally, correlations higher than 0.7 are considered statistically significant. As shown in the bottom right chart, this value is reached for values of Capacity*NACH at all distances above 2500 meters, which is in line with travel done by private vehicles.

80% R² = 0,7129 60% 40%

The distance of 20 Km is the one that shows most successful results and it is graphically represented in the image on the right.

20% 0% 0

1000

2000

3000

4000

5000

6000

7000

8000

9000

Predicted Traffic Volume

METROPOLITAN AREA (Correlation - 0<r2<1)

33

NACH x speed

Distance r:

NACH x road capacity

500

0.1948

1000

Metropolitan Area (0 < r2 < 1)

NACH x segm. lenght

NACH

0.0402

0.0016

0.0181

0,6

0.4388

0.2183

0.0092

0.001

0,5

2500

0.6315

0.4752

0.0402

0.0391

0,4

5000

0.6898

0.5643

0.0582

0.102

0,3

10000

0.7086

0.6007

0.0668

0.1479

0,2

15000

0.7123

0.61

0.0685

0.162

0,1

20000

0.7129

0.6132

0.0696

0.1668

n

0.7105

0.6126

0.0688

0.1727

0,8 0,7

0 500

1000

NACH*CAPACITY

2500

5000

NACH*SPEED

10000

15000

20000

NACH*SEG_LENGHT

n NACH


NACH*Capacity Highest Lowest

GIS elaboration representing NACH*Road Capacity values at distance=20Km


Graphical representation of Four-Step Traffic model (left) is very similar compared to the map representing NACH*Road Capacity values at distance=20000m (right)

Four-Step Traffic model

Traffic volume Highest Lowest

Space Syntax and Traffic Volumes

November 2018


NACH*Road Capacity (distance=20Km)

NACH*Capacity Highest Lowest


Conclusion and next steps The main goal of this research is to measure the correlations between traffic volumes and the space syntax analysis outputs. Main findings are: High correlation levels are recorded between traffic volumes and choice values if normalized (NACH); Higher correlation levels are recorded between traffic volumes and Normalized Angular Choice (NACH) values multiplied by Road Capacities (or other variables like speed). Further research might: Carry additional tests and check correlations for different road categories (highways, collector roads, local roads, etc); Test other case studies and compare results with the case of Milan; Introduce other measures in space syntax in order to directly weigh choice values for different parameters, such as land use, population, etc.

37


References CITILABS, (2018). Retrieved on 08/11/2018 through http://www.citilabs.com/software/cube/cube-voyager/. Hillier, B., (1996a). Urban Design International, 1(1): 41-60. Hillier, B., Yang, T., Turner, A., (2012). Advancing depthmap to advance our understanding of cities. In: Greene, M and Reyes, J and Castro, A, (eds.) 8th International Space Syntax Symposium. Pontificia Universidad Catolica de Chile: Santiago, Chile. SSN, (2018). Space Syntax Network. Retrieved on 08/11/2018 through http://www.spacesyntax.net/. Systematica, (2018). Reading Dubai Structure. Retrieved on 08/11/2018 through http://research.systematica.net/research/reading-dubaistructure/.

38


research.systematica.net


Turn static files into dynamic content formats.

Create a flipbook
Issuu converts static files into: digital portfolios, online yearbooks, online catalogs, digital photo albums and more. Sign up and create your flipbook.