0 The Co-Evolution of Land Use and Transport: Theory and Application to London David Levinson...

1

The Co-Evolution of Land Use and Transport: Theory and

Application to London

David Levinson

University of MinnesotaImperial College, London

February 21, 2007

With Bhanu Yerra, Feng Xie, Shanjiang Zhu, Ahmed El-Geneidy

2

General Points• Transportation and Land use are

interdependent shapers of urban form– Concentration of commercial activities

(suburbanization of residences) and prosperity of public transport at the first half of 20th century.

– Decentralization of both commercial and residential activities accompanied by expanding automobile/highway system in second half of 20th century.

3

Organization of Talk

• Movie: http://rational.ce.umn.edu (go to SIMS)

• Orderliness in London• SIGNAL - A Simulation Model of Co-

evolution• Twin Cities & Next steps

http://rational.ce.umn.edu/

4

The Sequence of Development?

This extension of the railway system by means of feeder lines means that in many ways the early development of the system can be viewed, not in terms of booms and slumps, but in rational steps. By the end of 1833, three of the five English provincial towns with a population of more than 100,000 had railway links with London under construction; by the end of 1836 only Portsmouth remained among English towns of over 50,000 population without a line authorized; and by the end of 1837 most towns of more than 20,000 inhabitants were on or close to the route of an authorized railway. - M.C. Reed

5

Orderliness Hypothesis

• H: Places will be connected to the network roughly in order of their population density (density is used to control for area). The places that have the highest population per unit area (or population density) will get the network first.

• Tested in London

6

Background

• 1836 London and Greenwich Railway• 1846 Royal Commission on Railway Termini• 1854 Metropolitan Railway chartered• 1863 Metropolitan Railway opens• 1884 “Circle” closed• 1890 City and South London Railway (first

tube)• Railways not permitted to be developers

except Metropolitan Railway --> Metro-Land• Greenbelt

7

Surface Rail & PopulationCorrelation between Population Density Rank and Train Station Density Rank

of London Boroughs (Local Administrations)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

1820 1840 1860 1880 1900 1920 1940 1960 1980 2000 2020

Correlation

Correlation between Station DensityRank and Population Density Rank(including City of London)

Correlation between Station DensityRank and Population Density Rank(excluding City of London)

8

Underground & PopulationCorrelation between Population Density Rank and Tube Station Density Rank

of London Boroughs (Local Administrations)

0.00

0.10

0.20

0.30

0.40

0.50

0.60

0.70

0.80

0.90

1.00

1860 1880 1900 1920 1940 1960 1980 2000 2020

Correlation

0

5

10

15

20

25

30

Number of Boroughs with Stations

Correlation between Station DensityRank and Population Density Rank(including City of London) (left axis)

Correlation between Station DensityRank and Population Density Rank(excluding City of London) (left axis)

Number of Boroughs with Stations (rightaxis)

9

Boroughs without Underground Service

Year

Boroughs with

Stations (N)

Notable Boroughs (in top N of density) without service at time which eventually get service.

1871 8 Tower Hamlets (2), Southwark(7), Hackney (8)

1881 14 Tower Hamlets (1) , Southwark (6), Hackney (7), Lambeth (9), Lewisham (13), Wandsworth (11)

1891 23 Hackney (6), Greenwich (15), Waltham Forest (16)

1901 23 Hackney (6), Greenwich (16), Waltham Forest (15) missing

1911 23 Hackney (5), Waltham Forest (15), Greenwich (16), Redbridge (22)



1941 25 Greenwich (16), Redbridge (20), Waltham Forest (14)

1951 27 Greenwich (19)





2001 28

10

City of London

0

20000

40000

60000

80000

100000

120000

140000

160000

180000

200000

1750 1800 1850 1900 1950 2000 2050

Year

Density (Employment, Population)

0

5

10

15

20

25

30

35

Rank of Population Density

Population Density of City ofLondon

Employment Density of City ofLondon

Rank of Population Density of Cityof London

11

Leads and Lags

Transport Leads Land Use

Transport Follows Land Use

Developed Area (B) Development densifies in urban area after construction of new transport infrastructure

(A) Constructing new (higher speed) mode in existing urbanized area (e.g. London Transport in early years)

Undeveloped Area (C) Constructing new (higher speed) mode in greenfields, to promote development

Still waiting …

12

Qualitative Model

• Rail first inter-city: connects outside -> in• Underground first connects termini, then other

points in developed area, and finally connects to new suburbs: inside -> out

• Has a decentralizing effect for residences, lowering population density in center, increasing it in suburbs.

• Other factors; entrepreneurs, construction costs, South vs. North (income, rail embeddedness, geology, competition, south already more local than north (London in south of England)

13

SIGNAL• Simulator of Integrated Growth of Network and

Land use– Simulate the co-evolution of land use and road

networks.– Implement a bottom-up process that incorporates

independent route choices of travelers, location decisions of businesses (jobs) and residents (workers), as well as investment decisions of autonomous roads.

– Kept as simple as possible to capture salient components, while enabling us to display and analyze the emergent patterns of land use and network on a large scale.

14

Model Framework

•

User-defined Land uses

User-defined Network

Travel Demand Models

Road Investment

Models

Access and Land use Models

Trip Generation

Trip Distribution

Traffic Assignment

Revenue Model

Cost Model

Investment Model

Accessibility

Employment Location

Population Location

SIGNAL Interface

15

Travel Demand Models1.Trip Generation

2.Trip Distribution

4. Traffic Assignment

Calibrated Stochastic User Equilibrium (SUE)

3. Mode Choice

Single mode is assumed

16

Road Investment Models1. Revenue Model

2. Cost Model

3. Investment Model

ab

17

SONG (fixed Land Use)Base Case: Uniform Initial Speeds and Land use

(U/U) (left) Spatial distribution of uniform speed for the initial

network; (right) Spatial distribution of speed for the network at equilibrium reached after 8 iterations.

QuickTime™ and aTIFF (Uncompressed) decompressor

are needed to see this picture.



18

Uniform initial speeds and random initial land use (U/R)

Spatial distribution of speed for experiment U/R after reaching equilibrium;





19

Random initial speeds and random initial land use (R/R)

(left) Spatial distribution of initial speed for experiment R/R (random initial speeds and random initial land use);

(right) Spatial distribution of speeds for the network after reaching equilibrium; The color and thickness of the link

shows its relative speed or flow.





20

Uniform initial speeds and bell-shaped initial land use (U/B)

Spatial distribution of final speeds for experiment U/B (uniform initial speeds and bell shaped land use)



Trips Produced

Trips Attracted

21

Access and Land use Models•Assumptions

–Residence and employment are the only types of activities, and their totals are kept equal and constant.

–Accessibility to population and to employment are the only factors that affect location decisions.

–People want to live near jobs, but far from other people to maximize available space and to avoid potential competitors for jobs, while employment wants to be accessible both to other employment and to people (who are their suppliers of labor and customers).

22

Access & Land use Models

1. Accessibility

2. Potential

3. Redistribution

23

Measures• Gini

€

rE=Ejdj

2j=1

n∑Ej

j=1

n∑

Equivalent radius (r)

Where

Ej= employment of zone j

dj= distance of zone j to the center of a region

24

Numerical Example • A hypothetical metropolitan area where:

– both the population and employment are distributed over a two-dimensional grid, stretching 10 km in both dimensions, divided into a 20X20 grid lattice of land use cells (400 zones).

– A total of 400,000 people are living in the city, which is equivalent to an average of 1,000 residents in each zone. Total employment equals 400,000 as well (and each resident holds a job).

– Two-way roads connect the centroids of each pair of adjacent zones, thus forming a 20X20 grid of road network as well, comprising 400 nodes and 1520 links.

25

Experiments and Hypotheses Experiments

HypothesesInitially flat land uses become more concentrated, and initially concentrated become less so.

The degree to which land uses are concentrated is reinforced when road networks are allowed to vary rather than remain constant.

€

0 < ΔGini(Ex.1) < ΔGini(Ex.2)

Δr(Ex.2) < Δr(Ex.1) < 0

ΔGini(Ex.3) < ΔGini(Ex.4) < 0

0 < Δr(Ex.4) < Δr(Ex.2)

26

Result: activity density

Evolution of employment without network dynamics

Evolution of employment with network dynamics

27

Results: Gini

0 more equitable

1 less equitable

28

Results: Equivalent Radius

29

Sensitivity

30

Sensitivity

31

Summary• A simple co-evolution model of transportation and

land use which incorporates independent route choices of travelers, location decisions of businesses (jobs) and residents (workers), as well as investment decisions of autonomous roads.

• Experimental results show that the degree of both employment and population concentration is reinforced when road networks are allowed to vary rather than remain constant.

• Contemporary integrated transportation and land use models that neglect road dynamics could underestimate the concentration of land uses.

32

Next Steps

• Application to Twin Cities• Estimation & Calibration of Models

(Markov Chain, Cellular Automata, MCCA, SIGNAL) using detailed empirical dataset.

41

Future Models …

• Modeling– Explicitly consider co-evolution– Dynamic transportation network instead of

static network– Dynamic land use rather than fixed land use– Variable rather than static demand– Bottom-up process (Cellular

Automata/Agent-Based) instead of top-down (Allocation, Equilibrium, Optimization)

42

Thank You

• More available at: Rational.ce.umn.edu

• Email: [email protected]

43

Orderliness

44

Random initial speeds and bell-shaped initial land use (R/B)(top) Spatial distribution of initial speed for experiments R/B (random

initial speeds and bell shaped land use); (bottom) Spatial distribution of speeds for the network after reaching equilibrium





45

Gravity model parameter variations with uniform network and land use (U/U) Spatial

distribution of relative speeds at equilibrium for (top) w = 0.02 (less sensitive to travel cost);

(bottom) w = 0.8 (more sensitive to travel cost).





46

Variation of Average Traffic Flow with Travel Behavior

50

250

450

650

850

1050

1250

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1

w

Average Traffic Flow

10X10

11X11

15X15

Variation of Average Link Speed with Travel Behavior

5

5.5

6

6.5

7

7.5

8

8.5

9

9.5

10

0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1w

Average Link Speed

10X1011X1115X15

Variation of average traffic flow (left) speed (right) with w for 10X10, 11X11

and 15X15 networks.

47

Slice of a bell-shaped downtown land use

pattern

Trips Produced

Trips Attracted

0 The Co-Evolution of Land Use and Transport: Theory and Application to London David Levinson...

Documents

Transcript of 0 The Co-Evolution of Land Use and Transport: Theory and Application to London David Levinson...