A cellular automata model for the study of small-size urban areas

Similar documents
Towards a cellular automata based land-use transportation model

Cell-based Model For GIS Generalization

Lecture 7: Cellular Automata Modelling: Principles of Cell Space Simulation

Development of modal split modeling for Chennai

Methodological issues in the development of accessibility measures to services: challenges and possible solutions in the Canadian context

CHANGES IN THE STRUCTURE OF POPULATION AND HOUSING FUND BETWEEN TWO CENSUSES 1 - South Muntenia Development Region

Changes in the Spatial Distribution of Mobile Source Emissions due to the Interactions between Land-use and Regional Transportation Systems

Geographic Information Systems (GIS) in Environmental Studies ENVS Winter 2003 Session III

Joint-accessibility Design (JAD) Thomas Straatemeier

Lecture 7: Cellular Automata Modelling: Principles of Cell Space Simulation

Brazil Experience in SDG data production, dissemination and capacity building

Improving rural statistics. Defining rural territories and key indicators of rural development

Tomás Eiró Luis Miguel Martínez José Manuel Viegas

Population Change. Alessandro Alasia Agriculture Division Statistics Canada. (ICRPS) Summer School 2009

Merging statistics and geospatial information

Generic Success Criteria

Labour Market Areas in Italy. Sandro Cruciani Istat, Italian National Statistical Institute Directorate for territorial and environmental statistics

Urbanisation Dynamics in West Africa AFRICAPOLIS I, 2015 UPDATE

DEFINING AND MEASURING WORLD-METRO REGIONS FOR INTERNATIONAL COMPARISONS

CONSTRUCTING THE POVERTY AND OPPORTUNITIES/PUBLIC SERVICES MAPS INFORMATION MANAGEMENT. Background: Brazil Without Extreme Poverty Plan

The Spatial Structure of Cities: International Examples of the Interaction of Government, Topography and Markets

Size matters: issues and challenges of local development with a special focus on small and medium sized towns in Czechia

Interactive Visualization Tool (InViTo)

TRAVEL DEMAND MODEL. Chapter 6

Urban settlements delimitation using a gridded spatial support

M. Saraiva* 1 and J. Barros 1. * Keywords: Agent-Based Models, Urban Flows, Accessibility, Centrality.

COMPREHENSIVE LAND-USE MANAGEMENT UNDERSTANDING THE INTERRELATIONSHIP BETWEEN SPATIAL PLANNING, LAND MANAGEMENT AND LAND ADMINISTRATION

Nature of Spatial Data. Outline. Spatial Is Special

Urban Process and Future Development of Colombo Metropolitan Area, Sri Lanka: An Application of Geospatial Techniques

CIV3703 Transport Engineering. Module 2 Transport Modelling

Representation of Geographic Data

Freight Facilities Impact on Land-Use A Temporal Analysis. Ben Zietlow Bradford Winkelman Mohamad Mzanar

A Constrained CA Model to Simulate Urban Growth of the Tokyo Metropolitan Area

The Building Blocks of the City: Points, Lines and Polygons

Rural Pennsylvania: Where Is It Anyway? A Compendium of the Definitions of Rural and Rationale for Their Use

Implementing a Spatial Decision Support System for Puerto Rico: The Political Process

METROPOLITANIZATION PROCESSES IN SPAIN: FROM THE TRADITIONAL CITY TO THE "CITY OF CITIES."

Unit No-1&2. Settlement Geography

A Micro-Analysis of Accessibility and Travel Behavior of a Small Sized Indian City: A Case Study of Agartala

- World-wide cities are growing at a rate of 2% annually (UN 1999). - (60,3%) will reside in urban areas in 2030.

Operational Definitions of Urban, Rural and Urban Agglomeration for Monitoring Human Settlements

New Prospects for Peripheral Rural Regions Helmut Hiess Glasgow, 19th of May 2010

The Use of Spatial Weights Matrices and the Effect of Geometry and Geographical Scale

São Paulo Metropolis and Macrometropolis - territories and dynamics of a recent urban transition

Understanding China Census Data with GIS By Shuming Bao and Susan Haynie China Data Center, University of Michigan

Simulating urban growth in South Asia: A SLEUTH application

USING DOWNSCALED POPULATION IN LOCAL DATA GENERATION

MOBILITIES AND LONG TERM LOCATION CHOICES IN BELGIUM MOBLOC

Subject: Note on spatial issues in Urban South Africa From: Alain Bertaud Date: Oct 7, A. Spatial issues

ENVIRONMENTAL VULNERABILITY INDICATORS OF THE COASTAL SLOPES OF SÃO PAULO, BRAZIL *

Analysis of a high sub-centrality of peripheral areas at the global urban context

Poland, European Territory, ESPON Programme Warsaw, 2 July 2007 STRATEGY OF THE ESPON 2013 PROGRAMME

A Logistic Regression Method for Urban growth modeling Case Study: Sanandaj City in IRAN

Data Collection. Lecture Notes in Transportation Systems Engineering. Prof. Tom V. Mathew. 1 Overview 1

Typical information required from the data collection can be grouped into four categories, enumerated as below.

Vanessa Nadalin Lucas Mation IPEA - Institute for Applied Economic Research Regional Studies Association Global Conference Fortaleza 2014

Labour MarketAreas: ThePortuguese case

EXPLORING THE FUTURE WATER INFRASTRUCTURE OF CITIES

Chapter 10 Human Settlement Geography Book 1 Class 12

Vincent Goodstadt. Head of European Affairs METREX European Network

How the science of cities can help European policy makers: new analysis and perspectives

Land Use Planning and Agriculture: Austrian Experiences and Challenges I

PRIMA. Planning for Retailing in Metropolitan Areas

Changes in population and industries in the rural areas of Finland: from analysis of administrative regions to a GIS based approach

Spatial profile of three South African cities

Volume Author/Editor: Gregory K. Ingram, John F. Kain, and J. Royce Ginn. Volume URL:

Urban-Rural spatial classification of Finland

Land accounting perspective on water resources management

Stepping away from trend analyses for regional integrated planning and modelling

Declaration Population and culture

Assessment of Directed Land Use Development in Chennai Metropolitan Area

Census Geography, Geographic Standards, and Geographic Information

University of Sheffield Department of Town & Regional Planning. Urban Design & Place- making Credit Value: 20 Level: 2

Achieving the Vision Geo-statistical integration addressing South Africa s Developmental Agenda. geospatial + statistics. The Data Revolution

PLANNING (PLAN) Planning (PLAN) 1

THE BRAZILIAN HOUSING PROGRAM - MINHA CASA MINHA VIDA - EFFECT ON URBAN SPRAWL

An Introduction to Scientific Research Methods in Geography Chapter 3 Data Collection in Geography

Global City Definition

Simulating household location choice at different geographical levels with UrbanSim

Land Use/Cover Changes & Modeling Urban Expansion of Nairobi City

Measuring Agglomeration Economies The Agglomeration Index:

Implementation of the ESPON 2020 cooperation program. 16 January 2017 Anneloes van Noordt

From second home areas to a permanent city: a planning strategy to renovate tourist destinations on the Mediterranean coast

Types of spatial data. The Nature of Geographic Data. Types of spatial data. Spatial Autocorrelation. Continuous spatial data: geostatistics

Methodology used in preparing the estimates and projections of the urban and rural population and the economically active population

Brazil Paper for the. Second Preparatory Meeting of the Proposed United Nations Committee of Experts on Global Geographic Information Management

Identification of Urban Areas Using Raster Spatial Analysis

Integrated Strategies in regional development

P. O. Box 5043, 2600 CR Delft, the Netherlands, Building, Pokfulam Road, Hong Kong,

Introducing GIS analysis

Intimate Infrastructures

Classification and Evaluation on Urban Sprawl Quality. in Future Shenzhen Based on SOFM Network Model

Vol. 51 No. 4 JOURNAL OF ZHENGZHOU UNIVERSITY July 2018 I Y2014-5

Mapping Welsh Neighbourhood Types. Dr Scott Orford Wales Institute for Social and Economic Research, Data and Methods WISERD

Developing Spatial Data to Support Statistical Analysis of Education

Note on Transportation and Urban Spatial Structure

CORRELATION OF PADDY FIELD FOR LAND BOUNDARY RECORD

A BASE SYSTEM FOR MICRO TRAFFIC SIMULATION USING THE GEOGRAPHICAL INFORMATION DATABASE

The importance of visioning in urban strategic planning By Peter Robinson. MILE Master Class Urban Strategic Planning March 2016

Spatial configuration analysis to improve the accessibility to light-rail stations

Transcription:

Context and Motivation A cellular automata model for the study of small-size urban areas Centre de Política de Sòl i Valoracions Barcelona 23 January 2007 Nuno Norte Pinto School of Technology and Management Polytechnic Institute of Leiria Portugal Context [1] A rapid growth of the Portuguese urban areas was registered in the past two decades Important changes in the national planning legal framework took place in the early 1990s By the late 1980s a great majority of the Portuguese municipalities did not had any type of master plan. Only in 1995 all the municipalities had their municipal master plans legally approved Significant evolution occurred in small and peripheral urban areas, from an unsatisfactory situation in terms of urban quality public spaces and facilities, infrastructures to a satisfactory level of urban quality 2 36 Context and Motivation Context and Motivation Context [2] Motivation [1] Small urban areas were poorly served by public facilities and infrastructures in the early 1990s. They lacked urban functions and, more generally, urban quality The use of modelling as an effective planning tool was not (and still is not) a common practice among the Portuguese planning practitioners Their potential for attracting new populations was very low and generally they have lost population in a constant decay since the 1950s As a result of the use of the European Union s structural funds, these small urban areas made an important effort during the past 15 years to build new public facilities, network infrastructures, and public areas The administration still has the need for more planning technicians that could integrate traditional planning approaches with modern modelling techniques Modelling is believed to be a valid and yet unexplored tool for understanding the Portuguese urban phenomena Some of these small municipalities presented the highest growth rates for population and constructed urban areas during the last decade It is also challenging to develop a modelling approach that could be able to simulate land use change in a context of rapid growth, based only in recent data 3 36 4 36 Context and Motivation Motivation [2] The use of Cellular Automata [1] The research on CA and urban studies has increased both in Two main issues were considered at early stages of model design: theoretical and operational perspectives all over the world. Many issues regarding CA and its applications to geography are still under intensive research The application of CA models, usually to large metropolitan areas, is based on the use of remote sensing maps and regular cell spaces. This approach is thought to be unfitted for the common Portuguese urban area The use of irregular cells based on census tracts is assessed, as well as the application of CA models to problems focused on small urban areas How could the growth of small urban areas be studied following an approach based on urban modelling? How could the model be tailored to reproduce problems that are intrinsically simple, no matter what modelling approach would be chosen? Small urban areas tend to have less land uses, thus less urban functions than large ones The modelling technique chosen should be able to replicate this simple set of land use interactions 5 36 6 36 1

The use of Cellular Automata [2] Cellular Automata structure Why the choice over CA instead of other microsimulation techniques (such as agent-base simulation, ABS) that are also capable of reproducing spatiality for different types of problems? Two main characteristics CA are a very intuitive and straightforward technique and they are easily applied to spatial problems of any kind The technique is based on a set of rules that operate over a spatial (cellular) structure thus reproducing spatial phenomena that evolves over space. It also simulates the evolution over time Spatiality Data Availability CA have five major components Both CA and ABS are able to simulate well spatial relations based on sets of simple relations CA can use location-based data, as it simulates interactions between locations (and its land uses) ABS simulates individual behaviours (households, land parcels, trips), demanding disaggregate information that is harder to collect and to validate Cells Cell space Cell neighbourhood Transition rules Time 7 36 8 36 Cells and Cell Space [1] Cells and Cell Space [2] The structural base of CA is the cell unit and the spatial structure created by a finite set of cells CA models usually are based on regular square cells that are obtained from remote sense maps (the cell is the pixel) The usual range of resolutions used on CA models varies from 500 500 square meters up to 25 25 square meters Regular cells may not reproduce well the Portuguese urban structure. The consideration of regular square cells could became a problem, not only because of the resolution of the cell, but mainly because of data availability for this spatial partition Information is referenced to spatial census tracts that are drawn considering the urban structure If a high resolution was considered, the model would became closer to agent-based simulation than to CA with the cell s dimension becoming similar to the dimension of the land parcel Data availability If the choice was for a low resolution of the cell, (100 100 m 2 or larger) the cell would become less representative of the urban form, particularly in the definition of urban areas Spatial representativeness 9 36 10 36 Cells and Cell Space [3] Cell States [1] The establishment of the set of land use states followed the main goal of keeping the model simple, as close to the CA formulation as possible Small urban areas are characterized by more simple urban structures and less diversity of land uses and functions Only six aggregate cell land uses were considered : Urban Low Density (ULD) Urban High Density (UHD) Industrial (IND) Non-urbanized urban areas (N-UUrb) Non-urbanized industrial areas (N-UInd) Restricted areas (Rest) Cell Space based on Census Tracts 11 36 12 36 2

1 0.5 0-0.5-1 1 0.5 0-0.5-1 Cell States [2] It was also considered the existence of a period of inactivity for a cell after it has suffered a change of state Urban areas were classified by their construction densities This classification is close to the planning rules commonly applied by municipal master plans, based on construction factors that are area dependent The consideration of all urban functions aggregated under a general urban land use incorporates those areas occupied by infrastructures, public facilities, and public space Cell suitability is assumed to be a binary variable that takes the value 1 if the cell is suitable for a given land use (both physically and legally) and 0 otherwise Cell States [3] Accessibility is assessed through the calibrated distance from the cell to the main functional centres of the territory (depending on the distance in time): the civil parish, the municipality main town and the main industrial area AccMeas i = α acc Di, MunicipalMainTown + βacc Di, CivilParish + γ acc Di, IndustrialArea Road network was classified in three classes: main roads, secondary roads and local accesses. Each cell is served by at least a local access road 13 36 14 36 Cell Neighbourhood [1] Cell Neighbourhood [2] One of the most distinctive characteristics of CA is its dependence on the interactions that are generated between each cell and its neighbours Neighbourhood effect was defined as a measure of the level of interaction between two locations within a certain distance that decays as the distance between land use location increases The first law of geography: everything is related to everything else, but near things are more related than distant things (Waldo Tobler) This value is set to -1 if two land uses manifest repulsion, 0 if they do not interact, and 1 if two land uses manifest attraction This relation is valid for land uses and for urban functions: two locations with similar or different land uses generates attraction or repulsion that varies with the distance between them These relations are very difficult to assess, as they depend on several interdependent factors such as land value, housing demand, public facilities location among several others The classical formulation considers that the neighbour effect is related only with the set of cells that are at a small distance to a given cell. The neighbourhood distance was considered as a model parameter Neighbourhood Effect 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Neighbourhood Effect 0 0.5 1 1.5 2 ULD-Rest Distance (km) ULD-IND Distance (km) 15 36 16 36 Transition Rules [1] Transition Rules [2] This process takes into account all the characteristics of each cell by the form of a measure of the transition potential, a calibrated value of accessibility, suitabilities and neighbourhood effect within the considered neighbourhood with a stochastic perturbation t t t Poti, s = f ( Acci, Suiti, s, NEffi, s, ξ ), i C; s CS Pot t i,s is the transition potential for cell state s of cell i at time step t Acc i is the accessibility measure for cell i (constant during the entire simulation period) Suit i,st is the suitability value for cell state s of cell i at time step t NEff i,st is the neighbourhood effect for cell state s of cell i at time step t ξ is the stochastic perturbation parameter The model is based on a population density demand approach. The model calculates the increase of population during the historical period from reference data and distributes that population over the territory under consideration Four transition rules were considered. Cells can transit from state 1 (ULD) to state 2 (UHD) state 4 (N-UUrb) to states 1 and 2 state 5 (N-UInd) can transit to state 3 (Ind), with the criterion changing from population to employment Basically, one cell on state n can transit to state m after verifying a set of conditions: The cell is not considered in the construction transitory condition There still is population to distribute over the territory Cell is at state n and its highest transition potential is associated to state m at time step t, chosen within the sub-set of cells that present the highest transition potential and are available for transition 17 36 18 36 3

Measuring CA performance The calibration process [1] Contingency matrixes and associated kappa measures (kvalue) were used to assess the quality of the simulation Calibration is based on an optimization procedure that uses a new algorithm called Particle Swarm The contingency matrix compares the map resulting from the simulation with a reference map for the same year The choice for an optimization approach is related with the high number of parameters of the model The consideration of the entire set of cell states for the calculation of the kvalue value would produce a distortion on its significance as an indicator of a good simulation To avoid this distortion, a modification of the kvalue measure was considered, called ModkValue This simplification consists in considering only the cell states that take part in the urban change dynamics, that is, not considering any cell state that is forced to remain in the same state throughout the simulation (and during real world evolution) or that does not participate in the land use dynamics Given such complexity there are 38 calibration parameters that are highly interdependent the use of sensitive analysis for each parameter would not reproduce the behaviour of one parameter when considering the others The main goal of an optimization approach is to ensure an extensive search for solutions in the multi-dimensional space through the minimization or maximization of an objective function related with the fitness function that assesses the model results 19 36 20 36 The calibration process [2] Definition The fitness function chosen to assess the quality of the model results was the modified kappa value, ModkValue The objective function is the maximization of ModkValue, which can assume the optimum value of 1 absolute coincidence between simulation and reality Particle Swarm is based on a group of 60 particles that fly through the search space during a maximum of 100 iterations Each particle has 38 dimensions, one dimension per calibration parameter A set of 20 theoretical problems was produced to test the performance of the CA model. The main goal is to evaluate the performance of the model for a group of theoretical problems, generated in a randomly manner under a set of conditions that could replicate likely real-world spatial structures The final goal of this procedure is to generate spatial structures that can be considered similar to an average small Portuguese municipality, both in its maximum dimension and in its number of cells Each particle has a memory of its past search history, usually called the cognitive component, stored as the individual best position Each particle also knows the search results of the swarm, called the social component, stored as the best position of the entire swarm These territories are square areas with their side dimension varying from 10 km to 20 km and were structured based on cells that intend to represent census spatial tracts. The number of cells varies from 800 to 1200 21 36 22 36 Problem 08 [1] Problem 08 [2] Final Land Use Map Simulation Land Use Map kvalue=0.841 23 36 24 36 4

Problem 11 [1] Problem 11 [2] Final Land Use Map Simulation Land Use Map kvalue=0.811 25 36 26 36 Definition Reference 1991 Condeixa-a-Nova registered a 48% growth rate for built area and a 17.8% growth rate for population for the period 1991/2001 The cell space has 1433 cells resulting from the intersection of census tracts with urban areas for 1991 and 2001 The set of cell states was identical to the one used for the test problems with 6 aggregate cell states Population densities were set to 6.6 and 13.2 inhabitants per hectare for state ULD and for state UHD in 2001, respectively The calibration procedure was based on a set of 60 particles with a maximum number of 100 iterations, for 10 years-long time steps 27 36 28 36 Reference 2001 Calibration 2001 kvalue=0.621 29 36 30 36 5

Prospective analysis Reference 2001 A prospective analysis was made after the calibration process considering the reference data from the national population censuses of 1991 and 2001 This prospective analysis was produced with the set of calibration parameters obtained from the calibration phase Two prospective periods of 10 years were considered to obtain simulations for the years 2011 and 2021 The population growth rate was considered constant of 10% in the two prospective periods 31 36 32 36 Simulation 2011 Simulation 2021 33 36 34 36 Conclusions Conclusions Model Results [1] Model Results [2] The CA model presented promising results for the set of theoretical test problems, although a series of improvements must be implemented to overcome some shortcomings The assessed performance for this model is in line with the results of some other works published for equivalent measures Results must also be considered good because they are close to commonly accepted thresholds Future developments must focus the assessment of more sophisticate and accurate measures of accessibility, suitability and neighbourhood relations The use of irregular cells based on census tracts is considered an important development of the approach. It shifts the focus of the model from land use usually obtained from remote sensing image to spatial information on both land use and demographics census tracts contain disaggregate demographic and socio-economic information and can be easily classified in terms of land use Another innovative element of the study is the consideration of land use demand as a function of observed and expected population densities. It can be said that urban growth depends on population growth and on the variation of construction density Neighbourhood must also be matter of further research 35 36 36 36 6