Top Banner
 Cellular Automata and Agent-Based Models for Earth Systems Research
50

Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

Apr 26, 2019

Download

Documents

dinhcong
Welcome message from author
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Page 1: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Cellular Automata and Agent-Based Models for Earth Systems

Research

Page 2: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Outline

• Introduction to Modelling and Simulation• CA

­ theory and application­ examples

• ABM

­ theory and application­ examples

Page 3: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

General Principles

• Natural systems can often be represented as continua• These can be represented by continuous discrete fields or by equations 

describing rates of change• Mathematically rates of change are expressed by differential 

equations.• Sometimes precise analytical solutions exist but often they must be 

solved numerically.• Advection and diffusion processes describe the rate of change of 

quantities in time and space.• They are best represented by partial differential equations and 

frequently solved numerically using finite differences/ finite elements.

Page 4: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Modelling vs. Simulation

• Modelling: the act of abstracting from the real world and specifying it in some formalism

• Simulation: running the model

Page 5: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Discrete vs. Continuous

• Time, Space, & Attributes• Discrete as approximation of continuous• Not either/or

Page 6: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Modelling Framework

Following Zeigler et al. 2000

     DEVS

DESS

DTSS

ABM

CA

     Discrete Event/Time/Equation Simulation System

Page 7: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

 Complexity Theory

• Not a theory• Chaos theory – Edward Lorenz• Related to emergence

Page 8: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

 Emergence

• Complex behaviour emerges from simple interactions

• Inter­scale emergence vs. intra­scale emergence

emergent global structure

Page 9: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   Stolen from : http://necsi.org/projects/mclemens/cs_char.gif

Page 10: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Outline

• Introduction • CA

­ theory and application­ examples

• ABM

­ theory and application­ examples

Page 11: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Cellular Automata (CA)

• Discrete dynamical systems

• Discrete = space, time, and properties have finite,  countable states

• Complexity is bottom up

Page 12: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA background

• Early research in the 40’s• Popularised by The Game of Life• Now used in modelling physical and human systems, e.g.  

­ soil erosion­ vegetation dynamics­ urbanization/ land use change­ sand piles

(And studied by a bunch of people obsessed with discovering all of the possible patterns that can be created by CA)

Page 13: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA components

• Cellular Space or Lattice• Cell States• Neighbourhood• Transition Rules• Discrete Time

if (some condition holds)     do x

finite set of cell states

Page 14: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Spaces• Traditionally Raster• Vector• Graph• Higher dimensional spaces?

Page 15: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA Neighbourhoods

Moore:

Page 16: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA Neighbourhoods

von Neumann:

Page 17: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA Neighbourhoods

Arbitrary:

Page 18: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   game of life glider

Page 19: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Game of Life Rules

1. A dead cell with exactly 3 live neighbours becomes alive

2. A live cell with 2 or 3 live neighbours stays alive; otherwise it dies. 

t = 4?

Page 20: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Page 21: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   Rinaldi E (1999)."The Multi­Cellular Automaton: a tool to build more sophisticated models. A theoretical foundation and a practical implementation" in Rizzi P. e Savino M. (eds) On the edge of the Millennium. Proceedings of Computer in Urban Planning and Urban Management 6th International Conference F. Angeli 1999 (in pubblicazione) e in Proceedings ESIT Creta (in pubblicazione)

Page 22: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA Applications: urban growth

http://www.geog.umd.edu/resac/urban­modeling.htm

Model of Future Growth in the Washington, DC­Baltimore Region 1986­2030 using the SLEUTH model

Page 23: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

SLEUTH growth coefficients

• dispersion coefficient­ spontaneous or road influenced growth

• breed coefficient­ new spreading centre or road influenced growth

• spread coefficient­ edge growth from spreading centre

• slope coefficient­ lower slopes are easier to build on

• road gravity coefficient­ distance from road influences growth

Page 24: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA results

http://www.geog.umd.edu/resac/urban­modeling­animation1.htm

sloperoads

excluded areas

Page 25: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CA Applications: Soil Erosion

• RillGrow 2 by Favis­Mortlock

http://www.soilerosion.net/rillgrow/

Page 26: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

CELLULAR AUTOMATA IN INTEGRATED MODELLING

Change in cropland area (for food production) by 2080 compared to baseline (%) for the 4 SRES storylines and HADCM3

After: Schröter et al. (2005). Ecosystem service supply and vulnerability to global change in Europe. Science, 310 (5752), 1333­1337

Page 27: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Analysis of CA Output

• Plot cell attributes

• Plot number of cells in certain state

• Use metrics for describing spatial pattern

time

no.

Height

Mass

Land Use….etc

e.g. patch size metrics

Page 28: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Outline

• Introduction • CA

­ theory and application­ examples

• ABM

­ theory and application­ examples

Page 29: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Agent Based Model (ABM)

A representation of a system in which agents interact with each other and their environment using a set of rules 

• Also called multi­agent systems (MAS)

Page 30: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

if (some condition holds)     do x

ABM Components

• Space (environment)• Agent(s) – rules defining interaction and neighbourhoods

• Discrete Time

Page 31: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

what is an agent?

Represents:

• some discrete thing in the world (usually a living thing)• something with behaviour

Representation:

• Physically ­ Geometric object• Programmed – an object with attributes and behaviour

Page 32: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

agent

• behaviour:­ Rational – deterministic / Stochastic­ e.g. BDI algorithm

• communication:­ Stigmergic­ Message passing

Page 33: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

deterministic agents

+same initial conditions

=same final state

+stochasticity

Page 34: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Environmental Examples

?

Page 35: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Types of ABM

• Fixed behaviour model vs. evolutionary model­ e.g. genetic algorithms

• Top down vs. or plus bottom up

Page 36: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Example – urban land use in East Anglia

• Endogenising the planning process

felled forest

inland bare ground

continuous urban

suburban/rural development

ruderal weed

tilled land

coniferous woodland

deciduous woodland

scrub/orchard

dense shrub moor

bracken

rough/marsh grass

meadow/verge/semi­natural

mown/grazed turf

grass heath

saltmarch

beach and costal bare

inland water

sea/estuary

unclassified

0 6030 km

Source: Lilibeth Acosta­Michlik and Corentin Fontaine; funded by the Tyndall Centre

Page 37: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

interactions

patches

agents

actions feedbackfeedback

Page 38: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Agent­environment interaction and   are β γ

parameters affecting preferences for landscape and service amenities, respectively

After: Caruso, G., Peeters, D. and Cavailhès, J. and Rounsevell, M.D.A. (2007). ‘Spatial configurations and cellular dynamics in a periurban city’. Regional Science and Urban Economics, 00, 000­000 (in press)

Page 39: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

agentsprivate sectorpublic sector

localplanners

national policyregional

development

non-residential residential

propertydevelopers

individuallandlords

tourismactvities

commercialcorporations

industryinvestors

tourists

non-residential

publicservices industrial

buildings

commercialcentres

individualretailers

hotels

sites ofvisit

residential

appartments

individualhouses

patches

feedback feedback

Page 40: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

1 2 3 4 5 6 7 8 9 10 11 12isolated student HA1 ++ + +++

single person HA2 + +++ ++ ++ +couple HA3 ++ + + +++ ++ +++ ++

couple with dep. children HA4 + ++ +++ + +single­parent family HA5 +++ ++ ++ +

couple with non­dep. children HA6 +++ ++ + +all retired HA7 + + +++ ++ +

CLUSTERS

Residential agents

• Socio­economic data analysis• Agent profiles (household types) & location trends

Page 41: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Legend

LSOA_EA_clust12_4fact<all other values>

clust_ward_12.CLUSTER<Null>

1 = HA6 ­ ... ­ HA7

2 = ... ­ HA3/6 ­ HA7

3 = ... ­ ... ­ HA3/4/6

4 = HA7 ­ ... ­ HA3/6

5 = HA5 ­ HA4 ­ ...

6 = ... ­ HA5/7 ­ ...

7 = HA3/4 HA2 ­ ...

8 = HA2 ­ HA3 ­ ...

9 = ... ­ HA2/5 ­ HA4

10 = HA3 ­ HA1/2 ­ HA4

11 = ... ­ HA3 ­ HA1/2

12 = HA1 ­ ... ­ HA5/7

Household agent location preferences

Demographics and coastal zone pressures

Page 42: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Residential model runs

R.I.P.

1

14%

cities

2

20%

suburbs

3

39%

periurban& rural

4

25%

coast

stage

% of pop

concentra

te

mainly in

Structu

re

Model run animation

Page 43: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Planning agent interactions

Page 44: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Infrastructureprovision

Built environment(Type 2: „implementors“

Top­down (Type 1: „policy developers“)

Bottom­up: (Type 3 „lobbyists“)

Environmentalorganisations

Propertydevelopers

Cultural/naturalheritage

Communityforums

Governmentalorganisations

Conceptual planning model

Page 45: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

ABM as Computational Laboratory

• Testing hypotheses• Testing methodologies• Is your ABM deterministic or has it got a 

stochastic component?• How many simulations is enough?• How do we interpret model results?• Statistical analysis of results

Page 46: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Analysis of ABM Output

• Plot agent attributes• Plot number of agents of certain type• Spatial pattern metrics

­ temporal considerations (at a time or over time)

Page 47: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

Difference between CA and ABM

?

Page 48: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

What is the goal of modelling?

• to predict the represented system?

• to understand and explain the represented system?

Page 49: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

ReferencesGeneral Modelling:

• Zeigler, B. P., H. Praehofer, and T. G. Kim, 2000. Theory of Modeling and Simulation: integrating discrete event and continuous complex dynamic systems. Academic Press, San Diego.

CA:

• Torrens, P. M. 2006. Simulating Sprawl. Annals of the Association of American Geographers 96 (2):248­275.

• Coulthard, T. J., M. J. Kirkby, and M. G. Macklin, 2000. Modelling Geomorphic Response to Environmental Change in an Upland Catchment. Hydrological Processes, 14: 2031­2045.

• Favis­Mortlock, D. T., J. Boardman, A. J. Parsons, et al., 2000. Emergence and Erosion: a model for rill initiation and development. Hydrological Processes, 14: 2173­2205.

• Fonstad, M. A. (2006). Cellular automata as analysis and synthesis engines at the geomorphology­ecology interface. Geomorphology, 77, 217­234.

• Langton, C., 1986. Studying Artificial Life with Cellular Automata. Physica D, 22.

• Shiyuan, H. and L. Deren, 2004: Vector Cellular Automata Based Geographical Entity. Proceedings of the 12th International Conference on Geoinformatics ­ Geospatial Information Research: Bridging the Pacific and Atlantic.  University of Gavle, Sweden, 7­9 June.

Page 50: Cellular Automata and Agent-Based Models for Earth Systems ... · Cellular Automata and Agent-Based Models for Earth Systems Research Outline • Introduction to Modelling and Simulation

   

References

ABM:

• Benenson, I. and P. M. Torrens, 2004. Geosimulation: Agent­based Modeling of Urban Phenomena. John Wiley and Sons, Ltd, London.

• Brown, D. G., S. E. Page, R. L. Riolo, et al., forthcoming. Agent­based and Analytical Modeling to Evaluate the Effectiveness of Greenbelts. Environmental Modelling & Software.

• Brown, Daniel G., Riolo, Rick, Robinson, Derek T., North, M., and William Rand (2005) "Spatial Process and Data Models: Toward Integration of Agent­Based Models and GIS" Journal of Geographical Systems, Special Issue on Space­Time Information Systems 7(1): 25­47

• Gimblett, H. R., Ed., 2002: Integrating Geographic Information Systems and Agent­Based Modeling Techniques for Simulating Social and Ecological Processes. Sante Fe Institute Studies in the Sciences of Complexity, Oxford University Press.

• Parker, D.C., Manson, S.M, Janssen, M.A., Hoffmann, M.J., and Deadman, P. 2003 Multi­agent systems for the simulation of land­use and land­cover change: a review Annals of the Association of American Geographers, 93(2). 314­337.

• Papers on the RePast site: repast.sourceforge.net/papers