Modelling urban dynamics - moving cities
Erasmus Intensive Programme (IP)
Agents-based modeling for processes and dynamics in landscape geography
17 February until 2 March 2013
At AGROCAMPUS OUEST ANGERS - France
2. Human actions are altering the terrestrial environment
at unprecedented rates, magnitudes, and spatial scales.
Land-cover change stemming from human land uses
represents a major source and a major element of
global environmental change.
The world population is increasing day by day, this
increases the demand for food and housing.
The cities nowadays are growing to points where before
were agricultural areas, agricultural areas are moving to
forest areas, the space will end at some point. Avoiding
the collapse starts with good planning.
To help planning one must be able to predict. Our crystal
balls are complex computer models.
3.
4. Land cover: attributes of the Earth‟s land
surface and immediate subsurface, including
biota, soil, topography, surface and
groundwater, and human (mainly built-up)
Structures
Land cover conversions: replacement of one
cover type by another and are measured by a
shift from one land-cover category to another,
as is the case of agricultural expansion,
deforestation, or change in urban extent.
[Cabral 2012]
5. Land cover and changes are visible in
remotely-sensed data or by generating
evidence from secondary statistics, such as
(agricultural) census data
Land-use as well as land-management
information, in contrast, is mainly gained
through detailed ground-based analysis,
although land use can be inferred in remotely-
sensed data under certain circumstances
[Cabral 2012]
6. Main tropical deforestation fronts in the
1980s and 1990s.
The map is based on three data sets:
(a) the deforestation hotspots in the humid tropics of
the Tropical Ecosystem Environment Observation
by Satellite (TREES) project,
(b) a time series analysis of tree cover based on 8-km
resolution data from the National Oceanic and
Atmospheric Administration‟s advanced very high
resolution radiometer (AVHRR, and
(c) the Amazon Basin deforestation maps derived
from time series of Landsat Thematic Mapper
(LandsatTM) data.
These maps were overlaid and combined to identify
areas where high rates of deforestation were measured
by several of the datasets. Green areas are intact
forests. The map indicates the number of times each
0.18grid was identified as being affected by rapid
deforestation by the different datasets (orange, pixels
detected as hotspot by one dataset, red, pixels
detected as hotspot by two datasets, black, pixels
detected as hotspot by three datasets)
[Lambin 2003]
9. Legend of previous slide‟s picture:
Population density in 1995 and most populated and changing cities
from 1990 to 2000.
The map is based on the 2001 revision of the “World Urbanization
Prospects”, which provides population estimates in cities of more
than 750,000 inhabitants for the years 1990 and 2000, and the
“gridded population of the world”, which provides population
estimates in 1995.
The first dataset focuses on megacities whereas the second
includes less populated areas.
Green circles represent the most changing cities between 1990 and
2000, blue circles the most populated cities in 2000, and red
circles the most changing and populated cities. The background
color scale represents the population densities in 1995 (from less
than five inhabitants in gray to more than 1750 inhabitants/km 2in
dark orange)
[Lambin 2003]
10. “In 2000, towns and cities housed more than 2.9 bilion people,
nearly half of the world population.”
“Urban population has been growing more rapidly than rural
population worldwide”
“the number of megacities, defined here as cities with more than 10
million inhabitants, has changed from one in 1950 (New York) to 17 in
2000, the majority of which are in developing countries”
“ It is estimated that 1 to 2 million ha of cropland are being taken
out of production every year in developing countries to meet the land
demand for housing, industry, infrastructure, and recreation ”
[Lambin & Geist 2006]
12. Guia de Portugal in 1924:
« Finalmente, no extremo N. dessa longa artéria, como remate da cidade moderna, abre-se o diadema da Rotunda
(Praça Marquês de Pombal), de pavimento em empedrado lisboeta, com letreiros alusivos à acção reformadora do
estadista, plantada de acácias do Japão (Sophora japonica), e de onde a vista se enfia através da Avenida e as ruas da
Baixa até morrer nos montes azuis da outra banda. De aí irradiam em diferentes direcções uma série de avenidas: a
de Brancaamp, que vai dar à praça do Brasil [Largo do Rato]; a de Fontes Pereira de Melo que comunica com a praça
do Duque de Saldanha, servindo de ligação aos modernos bairros conhecidos por Avenidas Novas; e a do Duque de
Loulé, que liga directamente a Rotunda com o largo do Matadouro, servindo de limite N. ao bairro de Camões.»
Praça Marquês de Pombal, Lisbon, 192...
Source: unkown
13. Praça Marquês de Pombal, Lisbon, 195...
Photo by: António Passaporte, in Postais de Lisboa, [Lisbon], C.M.L., [1998].
15. Praça Marquês de Pombal, Lisbon, nowadays...
in Bing maps bird‟s eye view
16. Rome was not built in a day.
Europeans still travel on Roman roads.
17. Planning for the best use of land and its resources should take fully into consideration the long-
term consequences of each type of use in order to stretch out most beneficially the well-being
of society in the future, and to protect the integrity of the land and its biota.
Reversible land-use: leaves the land, after use, essentially as it was before; little or no man-
induced modification remains.
Terminal land-use: commits the land to a chosen particular use, and any attempt at reversal
requires either time-scales that are long compared with the expected lifespan of the social and
political institution, or a commitment of resources that is too high for society to consider worth
bearing. Examples of terminal land-use are location of metropolises and sites of toxic and/or
radioactive waste disposals; by its nature the list grows monotonically.
In between these two extremes of reversible and terminal land-use, the bulk of land-use is
sequential, in which each use of land changes its potentials and configurations, and these
changes are mainly irreversible.
18. Major transport investments tend to be the
most durable and also involve the longest
time lags between planning and completion.
The same is true for the appropriation of
open space for human settlements.
The common feature of these changes is their
virtual irreversibility.
19. Provision and lack of food, feed, fiber and
timber
Disease risk and human health
Atmospheric Chemistry, Climate Regulation
and life support functions
Agro diversity and biodiversity loss
Soil quality
Fresh water hydrology, agricultural water use
and coastal zones
[Lambin & Geist 2006]
20.
21. Several factors can be used to explain a land use change in
general:
◦ Biophysical factors
Climate, relief, hydrology, and vegetation ...
◦ Economic and technological factors
input and output prices, taxes, subsidies, production and transportation
costs, capital flows and investments, credit access, trade …
◦ Demographic factors
population composition and distribution, namely changes in urbanization
and in household size acknowledge the importance of indirect or
consumptive demands on the land by an increasingly urbanized
population
◦ Institutional factors
Institutions (political, legal, economic, and traditional) and their
interactions with individual decision-making. In particular, government
policy plays a ubiquitous role in land change, either directly causative or
in mediating fashion
[Cabral 2012]
22. “ A model is a an abstraction of an object,
system or process that permits knowledge to
be gained about reality by conducting
experiments on the model ”
[Clarke 2003]
23. “Modeling is essential for the analysis, and
especially for the prediction, of the dynamics of
urban growth.
Yet the successful application of a model in one
particular geographical area does not necessarily
imply its successful use in another setting where
local characteristics, territorial constraints and the
classic site and situation properties of economic
geography ensure that different development
paths have been followed.”
[Silva 2001]
24. In the last decades different modeling techniques
have been developed for better understanding
and predicting urban expansion, such as Cellular
Automata (CA).
CA-based models have been commonly used in
exploring various urban phenomena, such as
urbanization, urban form change, urban growth
effect, etc.
CA is individual-based models designed to
simulate systems in which states, time, and space
are discrete.
[MadhanMohan 2012]
25. Simply stated, an automaton (plural:
automata) is a self-operating machine.
CA is a discrete dynamical system that is
composed of an array of cells, each of which
behaves like a finite-state automaton.
26. The data requirements for parameterization,
calibration and validation of urban models are
intense due to the complexity of the models
and their objectives.
A simple and well-known example of a
cellular automata is John Conway's Game of
Life.
27. These models are not concern on the
dynamics of multiple categories of urban land
that leads to simulation errors.
To overcome this drawback we use Agent
Based Models (ABM). It is an effective and
process based model which is used for
modeling the real world applications like
urban growth.
[MadhanMohan 2012]
28. SLEUTH
◦ Stands for the input needs for driving the model:
Slope
Land cover
Exclusion
Urbanization
Transportation
Hillshade
Uses historical geospatial data for calibration of its
parameters, when running SLEUTH the rules of the CAs are
calibrated to historical urban spatial data.
Actually SLEUTH incorporates two different models such as
Clarke Urban Growth Model (UGM) and Deltatron Land
Use/Land Cover Model (DLM).
29. SLEUTH – Advantages
◦ SLEUTH is a self-organizing CA model and, therefore, the
coefficients that control growth may vary according to numerous
factors.
◦ Due to its independent scalability, transportability, and
transparency which has become a popular tool in modeling, the
increase of urban extent over time and recreating the past or
forecasting growth into the future.
SLEUTH – Drawbacks
◦ The simulations performed with SLEUTH underestimate the
emergence of urban.
◦ SLEUTH produces some of the errors, because this model does not
consider the dynamics of multiple categories of urban land.
◦ The number of patches generated by SLEUTH is much lower and
patches are larger and more clustered. This model produces an
excess underestimation of new growth and an overestimation of
infill growth.
30. FCAUGM
◦ Stands for Fuzzy Cellular Automata Urban Growth
Model
◦ Fuzzy logic combined to CA provides a proper
framework for expressing and mapping the urban
growth dynamics.
◦ FCAUGM is generally capable of simulating and
predicting the complexities of urban growth.
31. FCAUGM - Advantages
◦ The fuzzy logic gives the evaluation closer to the
complex reality of regional planning.
◦ Assignment of the weights to the different indicators can
be taken into consideration in an environmental impact
using fuzzy logic, in order to obtain a significant
homogeneity and objectivity.
FCAUGM – Drawbacks
◦ Direct employment of fuzzy logic lies in the way
knowledge is captured, i.e. by employing man-made
rules.
◦ The construction of a manual, expertly guided rule-base
is a complex task due to the presence of a high number
of inter-dependent variables.
32. MOLAND
◦ The MOLAND model represents processes at three spatial levels: Global, Regional and
Local. Greater Dublin Area example.
[Lavalle 2004]
33. The macro-model consists of 4 strongly linked sub-
models representing the Economic, Demographic, Land
use and Transportation sub-systems.
◦ The economic sectors are aggregated into four main categories:
Industry,
Services,
Commerce, and
Port activities
The population is assigned to four residential categories:
◦ Residential continuous dense,
◦ Residential continuous medium,
◦ Residential discontinuous urban,
◦ Residential discontinuous sparse
34. The MOLAND macro-model consists of 4 sub-systems:
Economy, Demography, Land use and Transportation
[Lavalle 2004]
35. MOLAND – Advantages
◦ MOLAND is the model that comes nearest to real
percentages of each type of growth and it has
produced growth patterns closer to the real ones
than the other models.
◦ Indeed, it produced more realistic urban patterns
near the road network than the model of White et
al. since it considered various types of roads (
consequently, various coefficients) in the
calculation of the effects of road type on the
transition potential, instead of a single type of road
and, consequently, a single coefficient.
[MadhanMohan 2012]
36. MOLAND – Drawbacks
◦ The cell-to-cell correspondence between the real
and the map simulated using MOLAND is low. Yet,
establishing the correct location of each simulated
land use cell is very difficult because of path
dependence and stochastic uncertainty.
[MadhanMohan 2012]
37. There are some common limitations for all CA-
based models:
◦ They focused on the reproduction of spatial patterns and
this is the obscurity to model socioeconomic dynamics
and decision making processes regarding land use.
◦ None of these models consider the dynamics of multiple
categories of urban land and they may produce
simulation errors such as those found in the visual
inspection of results, caused by the lack of
differentiation between a single-family detached house
with a large garden and a group of single-family
detached houses.
◦ These shortcomings are overcome by Agent Based
Models (ABM), which are process based.
38. An agent-based model is a generalization of
cellular automata in which agents are able to
move around the space, rather than being
confined to the cells of a raster.
Agent based modeling refers to a modeling
concept which is closely linked to the
modeling techniques of object orientation
(OO).
40. Agents are a representation and a simplification
of complex (including human) behavior, this
representation is established by defining rules
used by the agents to pursue a goal (or goals).
Agents must be able to communicate among
each other, dependent on the behavior-rules one
applies in the model; also agents communicate
with the simulated environment.
42. An agent is „alive‟ in its environment. An agent
has direct interactions with its world.
The agent may act on the environment, which in
turn provides perceptions to the agent.
The complexity of these interactions is primarily
constructed in the agent, and not in the
environment. An agent observes and interprets
the world. The environment may change, but the
agent will have to observe these changes.
43. Agent-based models are useful in conceptualizing land use
changes and urban growth.
Each agent, in such models, acquires its momentum from factors
like the configuration of the land use of its neighbors, the cost
of living, cost of transportation, accessibility and other factors
determining the quality of life.
The spatial environment in the model includes land use
attributes (slope, land use, excluded, urban, transportation, hill
shade), land price distribution, surrounding environment.
A spatial environment includes a virtual real estate market,
social network, government policies and casualty problems
45. ABMS Software Packages
◦ AgentSheets
◦ AndroMeta
◦ AnyLogic
◦ Ascape
◦ Breve
◦ Cormas
◦ DEVS: Discrete Event System Specification
◦ EcoLab
◦ FLAME: FLexible Agent Modelling Environment
◦ GAMA: Gis& Agent-based Modelling Architecture
◦ JAS: Java Agent Based Simulation Library
◦ LSD: Laboratory for Simulation Development
◦ MAML: Multi-Agent Modelling Language
◦ MATSim
◦ MASON: Multi-Agent Simulation of Neighbourhoods
◦ MASS: Multi-Agent Simulation Suite
◦ MetaABM
◦ MIMOSE
◦ MobiDyc: Modélisation Basée sur les Individus pour la Dynamique des Communautés
◦ Modelling4all
◦ NetLogo
◦ Open StarLogo
◦ RePast: Recursive Porous Agent Simulation Toolkit
◦ Repast Simphony
◦ SimPack
◦ SimPy
◦ SOARS: Spot Oriented Agent Role Simulator
◦ StarLogo
◦ SugarScape
◦ Swarm
◦ VisualBots
◦ Xholon
◦ ...
46. Cellular automata (CA)modelling is one of the recent
advances in spatial–temporal modelling techniques in the
field of urban growth dynamics. It is commonly used in
exploring various urban phenomena, such as urbanization,
urban form change, urban growth effect, etc.
The limitations of the existing CA models are overcome by
using Agent based Modelling (ABM).
ABMs would certainly provide a more realistic
representation of complex urban organization, as well as
provide us the flexibility to vary urban quantities and
population characteristics.
ABM can be used as an effective model for modeling the
urban growth dynamics.
47. [Cabral 2012]Cabral P. (2012) - Land use and
cover changes: general issues and modelling
approaches - Erasmus Intensive Programm 2012
.
[Lambin 2003] Lambin E. et al. – Dynamics of
land-use and land-cover, change in tropical
regions.
[Lambin & Geist 2006] Lambin E. and Geist H.
(Eds) (2006). Land-Use and Land-Cover Change:
local Processes and Global Impacts. Springer
48. [Silva 2001] Silva, E.A & Clarke, K.C. 2001-
Calibration of the SLEUTH urban growth
model for Lisbon and Porto, Portugal, in
Computers, Environment and Urban Systems
#26 (2002), pg. 525–552
[MadhanMohan 2012] MadhanMohan, S. et al.
2012 – Analysis of various urban growth
models based on cellular automata, in
[IJESAT] International Journal of Engineering
Science & Advanced Techonology volume 2,
issue 3, p.453-460
49. [Lavalle 2004] Lavalle, C. et al. 2004-The
MOLAND model for urban and regional
growth forecast - A tool for the definition of
sustainable development paths, European
Commission - Joint Research Centre
[Huigen 2003] Huigen, M. G.A. 2003 - Agent
Based Modelling in Land Use and Land Cover
Change Studies – Interim Report IR-03-044,
International Institute for Applied Systems
Analysis Schlossplatz 1 A-2361 Laxenburg,
Austria
50. [Mathur 2007] Mathur, P. et al. 2007- Agent-
based Modeling of Urban Phenomena in GIS,
University of Pennsylvania
51. Erasmus Intensive Programme (IP)
Agents-based modeling for processes and dynamics in
landscape geography
17 February until 2 March 2013
At AGROCAMPUS OUEST ANGERS - France
Notas del editor
Tip: Add your own speaker notes here.
Have you ever collected pictures of your town? How it was in the past and how is it now?
The world is changing every single day in term of land use.As you can see in the fifties Lisbon was a city full of trees. This was the modern aspect at the time.
Lisbon! At night! Taken from the International Space Station by Canadian Astronaut Commander Chris Hadfield.
Fuzzy logic is a form of many-valued logic or probabilistic logic; it deals with reasoning that is approximate rather than fixed and exact
The macro-model consists of 4 strongly linked sub-models representing the Economic, Demographic, Land use and Transportation sub-systems. The economic sectors are aggregated into four main categories:Industry,Services,Commerce, andPort activitiesThe population is assigned to four residential categories:Residential continuous dense,Residential continuous medium,Residential discontinuous urban,Residential discontinuous sparse
Let’s see what happens when we don’t see everything surrounding: http://www.youtube.com/watch?v=8HJQ7XXYJSA
The agents or component parts "live" in some topological space (e.g. farmers, political institutions, predators and prey may live in a two or three dimensional world). Agents perform their activities in this environment. Agents communicate among each other and can cooperate in fulfilling their activities. The communication and actions of the autonomous agents is rule-based. Every agent possesses rules that enable it to deal with specific situations. For example, if an agent is approached by another agent and is asked whether it wants to participate in an exchange, the agent will search for a rule within its “register” that applies to the proposed exchange. If the agent has selected a rule, this rule allows the agent to negotiate in the exchange. A rule is a method of the agent.
This representation is established by defining rules, which the agent uses to pursue a goal (or multiple goals). The rules together represent the ‘rational’ behaviour of the agent. In order to simulatean agent model we let the agents communicate with each other. Communication in agent simulation is how an agent modeller intuitively sees the interactions between real-life (e.g. other agents, but also the environment) entities. Agents must be able to communicate among each other, dependent on the behaviour-rules one applies in the model; also agents communicate with the simulated environment.