Download as pdf or txt
Download as pdf or txt
You are on page 1of 26

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/233884682

Dynamic GIS and Strategic Physical Planning: A Practical Application

Chapter · January 1999


DOI: 10.1007/978-3-662-03954-0_5

CITATIONS READS

57 352

4 authors, including:

Guy Engelen Stan Geertman


Flemish Institute for Technological Research Utrecht University
123 PUBLICATIONS 6,665 CITATIONS 294 PUBLICATIONS 6,165 CITATIONS

SEE PROFILE SEE PROFILE

C.G.A.M. Wessels
Nexpri BV
16 PUBLICATIONS 103 CITATIONS

SEE PROFILE

All content following this page was uploaded by Guy Engelen on 17 March 2016.

The user has requested enhancement of the downloaded file.


5 Dynamie GIS and Strategie Physieal
Planning Support: A Praetieal Applieation

Guy EngelenI, Stan Geertman2 , Petra Smits 3 and Coen Wessels 4


I Research Institute for Knowledge Systems (RIKS), Tongersestraat 6, PO Box 463,
6200 AL Maastricht, Netherlands
2 Faculty of Geographical ScienceslNexpRI, University of Utrecht, Heidelberglaan 2,
3508 TC Utrecht, Netherlands
3 Research Institute for Knowledge Systems (RIKS), Tongersestraat 6, PO Box 463,
6200 AL Maastricht, Netherlands and NexpRI, University of Utrecht, Heidelberglaan 2,
3508 TC Utrecht, Netherlands
4 Netherlands Expertise Centre for Geographical Information Processing
(NexpRI), University of Utrecht, Heidelberglaan 2, 3508 TC Utrecht, Netherlands

5.1 Introduction

Information and communication technology is a booming sector, not just in financial


and employment terms, but also in the products and facilities which have been created
in recent times. Product life cyc1es continue to get shorter, as exemplified by the
continuous updates of software, and potential fields of application continue to widen.
The technology is no longer envisaged as just an essential component of the gross
national product or as a competitive commodity within our national knowledge
infrastructure. Increasingly, it is considered as a tool which lightens workloads and
helps us to manage more complex activities in our working lives. One example of
the latter is the increasing role technology is playing in the field of physical planning.
It is known both from theory and practice that with policy-making for physical
planning there is an increasing need for more instrumental and methodologie al support
(see, for example, Nijkamp and Scholten 1993, Webster 1993, 1994, Wright et al.
1993, Holmberg 1994, van Lammeren 1994, Masri and Moore 11 1994, Worral11994,
Eweg 1994, Geertman 1996). In fact, according to Alexander (1984, 1986) and also
Jankowski and Richard (1994), little attention has been given up until now to the
development of methods and techniques which can support the planning process,
especially the generation and evaluation of alternative spatial intervention scenarios.
Most of the general purpose policy support instruments have their origins in the 1960s
and 1970s, inc1uding methods such as cost-benefit and multi-criteria analyses. The
same applies to systems which can handle the locational features of data and spatial
phenomena explicitly. In this chapter, the concept of geographical information
technology is used as an umbrella term to indicate information systems which can
handle the locational features of data and spatial phenomena explicitly, of which GIS
is one particular type.

J. Stillwell et al. (eds.), Geographical Information and Planning


© Springer-Verlag Berlin Heidelberg 1999
88 G. Engelen et al.

It has been the case that tools within geographieal information technology, CAD
and GIS for example, have experienced rather isolated developments and applications.
Thus, present demands for more integrated use of these instruments pose many
problems (Cowen 1988, Kevany 1989, Newell and Theriault 1990, Dave and Schmitt
1994, Schuur 1994). This applies to the integrated use of spatial simulation models
and GIS (Batty 1992, Batty and Xie 1994, Chou and Ding 1992, van Deursen 1995,
Fedra 1995) and to the integration of GIS and WWW technology (e.g. the use of
analytical GIS funetionality in multimedia WWW Java applications). Nevertheless,
the integrated use of separate instruments of information and communieation
technologies in new ways also holds real promise for policy support in spatial planning
and deeision-making (Liggett and Jepson 1995). This chapter will focus on the
integrated use of spatial simulation modelling and geographical information
teehnology to provide policy support within the field of strategie physical planning.
Spatial intervention in the buHt environment has been a common practice since
the beginning of mankind. In 4,000 BC, the first eities were buHt in Mesopotamia.
The Egyptians buHt their pyramids, the Greeks raised their temples and the Romans
constructed their aqueducts and road networks. In the twentieth eentury, small and
large scale spatial intervention in the already heavHy built-up environment continues:
European-wide networks of railroads and highways; eity-wide networks of tramways
and undergrounds; newly planned neighbourhoods, eities and airports; the large scale
renovation of outdated industrial areas; and so on. Recent examples in the Netherlands
of major spatial interventions inc1ude the waterworks (Zuiderzeewerken) for the new
town of Almere, and plans for some extensive new international railroad networks
(e.g. Betuwelijn; HogeSnelheidsLijn). One striking feature ofthese Dutch examples
is the extensive time involved in the process of deeision-making; another issue is the
contentious outcomes of the preparation phase, given the eomplexity of the problem
and the impact the outcomes will have. For instance, the route of a highway will
influence its near surroundings in a negative (noise pollution) and a positive
(connectivity) way for decades thereafter.
In the Netherlands, several steps have recently been undertaken to reduee the time
involved in the proeess of deeision-making (e.g. streamlining the legal and politieal
procedures; curtailing the participation process). Nevertheless, there still is a
considerable discrepancy between the time involved in the decision-making process
and the quality of the outcomes of this poliey preparation phase. Moreover, the politieal
dimension of the deeision-making process lacks a c1ear and full overview of all the
pros and cons of the potential alternative deeisions. According to Jeffers (1988), this
shortcoming has to do with both the insuffieient generation of a wide diversity of
potential (partial) solutions to the problem at hand and the failure to examine the
consequences of these (partial) solutions thoroughly. From the literature it appears
this is not just a Dutch problem but much more a general one. In fact, it is quite a
striking feature when one is aware of the time and money involved in tackling the
complexity of this kind of spatial policy problem and when one considers the high
expectations assoeiated with the introduction of computers into the policy process.
Whilst new geographical information technology instruments promised to be panaceas
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 89

for lots of spatial design and research problems, these promises have not been fulfilled
either in terms of efficiency (time) or the quality of the end results.
It is the purpose of this chapter to have a closer look at this quality deficiency by
studying the role of policy support of geographical information technology in the
physical planning and decision-making process. More specifically, attention will be
focused on the present and potential role of the integrated application of GIS and
spatial simulation modelling. There are three key questions we wish to answer. Firstly,
what is common practice in terms of policy support where spatial intervention in the
buHt environment is concemed? An initial review of the literature suggests that support
has been largely qualitative. Secondly, what is the potential role of geographical
information technology in this process? Current practice indicates that there is a
substantive need for more support in this highly complex policy preparation phase.
Thirdly, how can the policy support function of geographical information technology
be extended through the integrated application of GIS and spatial simulation modelling?
The literature shows a growing number of examples where significant value has been
added through integrated use. These three questions provide the chapter with its structure.
In Section 5.2 we consider the support role of geographical information technology in
policy preparation and its potential. The added value of integrating GIS and a dynamic
spatial model based on the theory of cellular automata is the topic of Section 5.3. A
case study of the suggested integrated approach is explained in Section 5.4. Conclusions
and recommendations for further study are given in the final section.

5.2 Spatial Policy Preparation Methodology

Aceording to an old but still valid definition, physical planning is concemed with
finding the best possible reeiprocal adjustment of space and society in favour of
society (Rouge 1947). In this ehapter, attention is foeused primarilyon the very
early phase of the physical planning and decision-making process. The prime activities
at this early stage of poliey preparation are aimed at the more precise specifieation of
often ill-structured problems (Rittel and Webber 1984) and towards exploring the
first tentative solutions to the spatial problem at hand ('sketch planning'). The
complexity of spatial problems is a eonsequenee of several faetors (Radford 1988,
cited in Massam 1992):
• the absence of deterministic and eomplete information about the set of options,
impacts and interest groups;
• the diffieulty of assigning numerieal ratio values to indieate the relative
attractiveness of options for all the criteria; and
• the fact that different interest groups are involved in the search for a preferred site
and that each interest group may have different perceptions about the problem and
may have more than one (possibly mutually conflieting) objective to whieh potential
outcomes are compared.
90 G. Engelen et al.

Looking at the current prevailing practice to handle this complexity from a


methodological viewpoint, it can be argued that this process of policy preparation
for spatial intervention in the built environment is rather one-sided. The outcome of
this policy formulation phase normally consists of the generation of primarily
qualitative spatial intervention scenarios. Spatial researchers are usually given the
task of building an inventory of the relevant figures for the area under study, now
and as expected in the future. Descriptive scenarios are often the prime outcome of
this research activity, which include demographic, econornic and social statistics
and written descriptions of some alternative ways in which the spatial intervention
in the built environment will occur in the future together with their consequences.
These figures and descriptive scenarios form inputs for the work of spatial designers
who will draw/create the different functional settings of the intended intervention,
resulting in alternative spatial intervention scenarios. In practice, this procedure has
several shortcomings. Firstly, communication between spatial designers and
researchers is often unsatisfactory, resulting in negative consequences in terms of
the outcomes of the preparation phase. This communication can be characterised as
one-off, one-way (from research to design) and is defective because often there is no
clear relationship between the output of the research activities (intended as inputs
for the design activities) and the output of the design activities. This inconsistency
can be explained partly by the differential working habits and methods ofboth working
groups (e.g. creative, intuitive versus analytical, systematic), and partly by the
hierarchical relationship described to both methods (design takes precedence over
research) (Geertman 1992, Schuur 1994). However, it should be recognised that
each of these activities makes a unique contribution to spatial policy support and it
would help if there was greater complementarity between the designers and the
researchers. Moreover, replacing the one-off, one-way communication by a more
frequent and mutually beneficial type of communication can also contribute to
improving the quality of the combined outcome of the procedure.
A second shortcorning of the planning procedure is the lack of accountability of its
end results (Faludi and van der Valk 1994). Due to the way in which the spatial
design part of this procedure is fulfilled (mostly filled with implicit choices and
assumptions), the outcomes of this procedure, the alternative spatial intervention
scenarios, lack the possibility of being evaluated for their differential consequences
and, in this way, compared systematically. Because of the societal consequences of
most of these spatial interventions, this is unacceptable in an open and democratic
society. Thirdly, the procedure remains quite traditional. Despite the fact that, in the
case of the Nethedands, geographical information technology was introduced in the
rnid-1980s, most of these planning procedures are still performed (almost) without
the help of these kinds of instruments. This gives an indication of the conservatism
that characterises this process. Moreover, in the few working procedures where use
has been made of the technology (especially CAD and GIS), it appears to be that just
a lirnited part of the potential functionality and application of this technology is
used, primarily restricted to database management and cartographic map-making
(Geertman 1998). However, geographical information technology can perform, in
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 91

principle at least, a much more profound policy support role. In general terms,
geographical information technology can provide analysis, design and visualisation
tools so that planners can see and analyse what will be the results and consequences
of particular design options. Notwithstanding that, it is widely recognised that the
present generation of this technology is much better equipped to perform the 80-
called 'representation' function (management and cartographic presentation of
locational data and related phenomena) instead of the so-called 'what-if' function
(analysing alternative spatial intervention scenarios and their consequences). Quite
recently, several attempts have been undertaken to improve this imbalance. One of
these is the coupling of spatial simulation models to a GIS. In this way, a so-called
dynamic GIS is created, which can simulate the spatial dynamics of reality much
better than solelyaGIS (e.g. see White and Engelen 1997). Moreover, this integration
opens the way to a much more quantitative approach to planning which can
complement the present qualitative procedures. In the next section, we will take a
doser look at the coupling of a spatial simulation model to a GIS and the benefits
which result from this integration.

5.3 The Integration of GIS and Dynamic Spatial Models

Although geographical information technology enables the representation and the


what-if functions to be used in the early phase of the physical planning process, its
spatial analytical functionality is still poorly developed in comparison with
cartographic functionality. "GIS have still notfully realized their potential as systems
to support and facilitate spatial modelling processes. They continue to handle the
temporal dimension very poorly." (Wagner 1997, p.219). "Missing almost entirely
are non-localised spatial notions such as spatial organisation, configuration, pattern,
spatial process, spatial dynamies, restructuring, transformation, change. Yet these
are all notions that are central in urban and regional studies, and they underlie
urban and regional planning especially at the strategie level" (Coudelis 1991, p.1S).
It is questionnable how effective physical planning can be without some capability
to anticipate the changes in activities, flows and land uses that are the result of the
intrinsic growth potential of the system itself and the planning interventions imposed
upon the system. The only way to get some insights into these is through a dynamic
model that represents the mechanisms of change. In the late 60s and early 70s, large
scale dynamic models for urban and regional planning applications boomed. Many
of them had a chequered record, partly because of the naive expectations about their
ability to make detailed and precise predictions, and partly because of inept modelling
paradigms. Since then, the elaboration of new scientific paradigms based on
phenomena such as complexity, self-organisation, chaos and fractals has generally
emphasised the fact that exact prediction in complex socio-economic or socio-
environmental systems is not possible. As a result, a new breed of models has been
92 G. Engelen et al.

developed which treat socio-economic systems as integrated, dynamic and complex


systems. The main purpose of the models is to serve as exploration and thinking
tools, that enable the user to leam about the complex nature and dynamic behaviour
of the real system and to find out how it is critically bounded, rather than to make
definite statements about the future state of the system modelled. These models go
back to the roots of modelling; they represent a simplified version of reality, they
represent the complex web of essential processes responsible for the dynamics of
the system, but their structure is kept simple and transparent. The further assumption
is that this simplicity and transparency will help analysts and planners alike to deepen
their understanding of the fundamental processes that structure the system for which
they are planning.
Cellular Automata (CA) models are a particular example of a class of models
which are extremely simple in their definition, yet are capable of generating very
complex behaviours. They can be thought of as simple dynamic systems in which
the state of each cell in an n-dimensional array depends on its previous state and the
state of the cells within its neighbourhood, according to a set of stated transition
mIes. They are, by their definition, very appealing objects to spatial scientists, since
they explicitly refer to the concept of space. A CA model has the following features:
• an (in principle) infinite n-dimensional Euclidean space, divided up into an array
of identical cells. For most geographical problems a two-dimensional space
consisting of square cells is chosen;
• each cell is in one of k possible discrete states at a given time;
• (a single set of) transition mIes describe the new state of each cell as a function of
its own state and the state of the cells in its neighbourhood;
• the neighbourhood is a fixed template which is applied to each cell of the ceHular
space at each iteration; and
• the system is closed, time progresses uniformly, and at each discrete time step, all
cells change state simultaneously as described in the transition mIes.
Because the system is discrete and iterative, and involves interactions only within
local regions rather than between all possible pairs of cells (as is the case for most
interaction based spatial models), a CA model is very efficient computationally. It is
thus possible to work with arrays containing hundreds of thousands of cells. The
best known and most studied CA model is Life (Gardner 1970). It is a 2 (states), 8
(neighbours) CA model which mimics a form of life on a two-dimensional array.
Cells are either dead or alive (the states). They are born if they are surrounded by
sufficient (=3) life neighbours to act as parents, and die ifthere is overcrowding (>3)
or a lack of company «2). It was Tobler (1979), who introduced Life into geography
and called it the Geographical Model. Yet, despite the fact that Life-like models can
display very complex behaviours - evolving towards stable states, or to chaotic states,
with all sorts of transients - it is too simple a model to describe geographical problems
realistically. Some spatial scientists have trled to add detail and constraints to Life in
order to turn it into an operational model for geographical applications (Roy and
Snickars 1993, Batty and Xie 1994). Other weH known CA models are Forest Fire
which models diffusion processes (see for example: Bradbury et al. 1990, Green
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 93

1993) and the logistic-like models developed by Couc1elis (1988) mimicking the
behaviour of rodent populations (see also Itami 1994).
While the early applications of CA models in the spatial sciences remained rather
conceptual and theoretical (Couc1elis 1985, 1988, Phipps 1989, Cecchini and Viola
1990, 1992), most recent applications are developed with an aim to realistically
represent geographical systems, both in terms of the processes modelIed and the
geographical detail represented. But, this trend has come with an increase in the
complication of the models developed, and, in their effort to build more realistic and
operational models, "there is practically no dejining characteristic of standard CA
that researchers..... have not been able to discard" (Couc1elis 1997, p.167). One of
the very essential relaxations to the standard CA definition is the introduction of the
finite non-homogeneous cell space: a bounded cell space consisting of cells having
different attribute values representing physical, environmental, social, economic
infrastructural or institutional characteristics of the cello This has allowed CA models
to be linked, conceptually and practically, with GIS (White and Engelen 1993, Batty
and Xie 1994, Clarke et al. 1997). In addition to the very similar representation of
geographical space, state of the art raster GIS systems provide the user with a sufficient
set of operators to define CA models: user-defined filters, overlay, (re-)c1assification,
and a scripting language to put all the operations in a sequential order. As a result,
most recently, a number of authors have suggested ways to build CA functionality
into GIS and/or GIS functionality into CA (Itami 1994, Takeyama 1996, Wagner
1997, Wu 1998, Lohman de Savomin 1998). At the present time, the examples of
CA models developed in GIS remain simple, because GIS do not, as yet, provide the
flexible operators required to define complex CA rules and lack the simulation engines
to ron the models at practical speeds. Yet, the disappearance of these technical
obstac1es and the further integration of CA into GIS, could bring dynamic capabilities
to the GIS user community in the near future.
Another avenue towards dynamic GIS is to be expected in the further advances
made in the domain of software component technologies, which will permit the
straightforward integration of specialised CA software modules in GIS packages or,
vice versa, GIS modules in CA packages. Along these lines, the Research Institute for
Knowledge Systems (RIKS) has developed GEONAMICA®, a simulation package
geared to build and ron CA models whose overall dynamics, particularly the total
growth and decline of different land use categories, are constrained by processes
represented in another type of dynamic model. GEONAMICA® supports very much
the idea that CA models are a type of modelling capable of dealing very efficiently
with short range dynamic processes. But, they have their lirnitations; in particular,
their performance can be greatly enhanced if they are used in combination With other
models and techniques. Coupling CA models to GIS has the great advantage of adding
geographical detail and reality to the models; and coupling CA models to dynamic
models of another type, in particular dynamic spatial interaction based models, has the
great advantage of enabling, on the one hand, the representation of spatial interaction
processes that are beyond the reach of the CA neighbourhoods, and on the other hand,
the introduction of the results and by-products of the local CA dynamics into the long
range interaction mechanisms.
94 G. Engelen et al.

The processes that are active at each level are described by means of appropriate
dynamic models: a systems dynamics model at the national level, a dynamic spatial
interaction based model at the regional level and a CA at the locallevel. To that end,
GEONAMICA® features a simple custom-designed GIS which has all the functionality
required to enter, import, edit and export the maps required to run the CA models, it
has the functionality required to define and run CA models, and it features a built-in
object oriented simulation language to develop integrated dynamic models that are
coupled to the CA models and to the GIS. With GEONAMICA® rather sophisticated
models can be developed. An example of such a model is the LeejOmgevings Verkenner.
It concems an integrated socio-economic and environmental model representing the
Netherlands at three coupled levels: the national level, the regional level consisting of
forty interacting economic COROP regions and the cellular level consisting of 351,000
grid cells, each covering 25 hectares (Fig. 5.1). COROP regions are more or less
economically homogeneous administrative regions centred around major cities in
the Netherlands.

National level:
1 region

Regional level :
40 COROP-regions

Local level :
351.000 cells (25hal

Fig. 5.1. Structure of the LeejOmgevingsVerkenner model


Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 95

In the remainder of this chapter we will concentrate on the application of a relatively


simple CA model to the IJmond region in the Netherlands. The IJmond model has
been implemented by means of GEONAMICA®. It has the general characteristics
of the models developed by White, Engelen and Uljee (White and Engelen 1993,
1994,1997, White et al. 1997, Engelen et al. 1993, 1995, 1996, 1997a, 1997b, 1998)
and consists of a cellular model only. The total land claims for each function modelled
and at each simulation time step, are calculated exogenously to the model and entered
as time series. The CA model is specified as follows:
• the area modelled is represented by means of a matrix of 120 (north-south) by 142
(east-west) cells each covering an area of 100 by 100 metres;
• the model describes the changing land use in the region for the period 1989-2010.
Cells are in one of 14 possible states representing the dominant land use in the
cello Agriculture (1), urban residential (2), rural residential (3), industry (4), offices
(5), distribution and transport (6), mixed activities (7), and natural and cultural (8)
are the so-called land use 'junctions' in the model. The CA model will fully describe
their temporal and spatial dynamics. The other categories: protected nature (9),
recreational (10), institution al (11), water (12), roads (13), and railways (14) are
the so-called 'features' in the model. They are static elements in the model, but
their presence and position will influence the dynamics of the functions;
• the neighbourhood is larger than for traditional CA models.1t is a circular template,
with a radius of eight cells and covering a total of 196 cells; and
• the transition rules are typically written as distance functions and represent, in a
general way, locational preferences and spatial interaction mechanisms, including
push and pull forces, inertia, agglomeration and competition effects (Fig. 5.2).

100
.
-
_ -.- Urban residential
Rural residential
. . . . ..., .. ·x··· Industry
" .,
"., - -e-- Distribution and transport
...+ ... Mixed activities

o
\
.-._._. -_. '
-~ ~~t~~~
.*~.,
:liik ~ *.~
c::
o ....-;.
i~ ~lif:."'<"~"'l
'e;;
"S
i ..
... ' " ~··'
c-
a.> .....:i-.:).:.; t-'
c:::
," , IJ '
I "

-100 l 2 3 4 5 6 7 8

Distance fram the centre cell


Fig. 5.2. Rule set for the urban residential function used in the IJmond model
96 G. Engelen et al.

Positive values represent an attraction force, while negative values express arepulsion
force. From the graph it is clear how these forces change with the distance. The urban
residential function is attracted by itself and by the rural residential function, while it is
repelled by industry, distribution and transport and by mixed activities functions. Cells
within the neighbourhood are weighted differently depending on their state, y, and
their distance, d, from the centre cell for which the neighbourhood is defined. For each
cell, its potential for transition to each active cell state, z, is calculated as follows:

(5.1)

where

tNz =LLW
d j
tl.
z-J,d d,1 (5.2)
and

v=l+[-ln (rand)]a (5.3)

where:
t p.. is the potential for transition to state z at time t;

Sz is the suitability of the cell for state z; with (0 ::;; Sz::;; 1);

t Zz is the zoning regulations for the cell for state z at time t;

t N z is the neighbourhood effect at time t; ifNz < 0 then Sz is replaced by (1-Sz);

W z,y,d is the weighting parameter applied to cells in state y in distance zone d;

t ld,j = 1 if cell i in distance zone d is in state y at time t, otherwise t ld,j =0;


s, p and n are parameters expressing the importance of Sz, Zz, and Nzin the calculation
of the transition potential. In the model they are used as binary switches and get the
value 0 or 1;
v is a stochastic disturbance term, where (0 ::;; rand::;; 1) is a uniform random variant; and
a is a parameter that allows the size of the perturbation to be adjusted.
The total amount of cells, Tz, in state z at each simulation time step is defmed extemally
to the model. To decide which Tz cells will be in state z at each time step, an allocation
rule is used. First, all the transition potentials to all possible states and for all cells are
calculated and ranked from highest to lowest. This results in an ordered list with a
number of elements equal to the number of cells times the number of possible states (in
our application 120*142 cells times 8 possible states =136,320 elements). Next, starting
with the highest value (the first) from this list, the Tz,cells with the highest potential for
transition to each particular state z' are identified. Then, the rule states that a cell will
change to the state for which its potential is highest, unless it is not among the Tz'
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 97

highest values. In the latter case, the cell will change to state ZN for which its potential
ranks second, and so forth. Cells for which potentials are not among the Tz highest for
any of the states will remain or return to the state 'agriculture'.
From expression (5.1), it will be dear that, next to the generic neighbourhood
effect, N z, ceIl-specific information may be introduced into the calculation of the
transition potential. In particular, the ceIl' s suitability, Sz, to support the specific land
use (state), and the zoning regulations, Zz, that apply for the land use, may be
introduced into the calculation to be controlled. The parameters s,p and n in expression
(5.1) allow the importance of the suitability and zoning in the calculation to be
controlled. TypicaIly, the information on the initial land use, the suitabilities and the
zoning regulations for each land use, and the network structure, are obtained from a
GIS. In order to develop the suitability maps, a technique based on multi-criteria
analysis is applied in our particular example of IJmond. Section 5.4 of this chapter
deals with the case study and the results of the model in more detail.

5.4 The Case Study ofUmond/Zuid-Kennemerland

5.4.1 The Umond/Zuid-Kennemerland Region

The study area IJmond/Zuid-Kennemerland (Fig. 5.3) covers an area of 12 by 14.2


km on the northern border of the Randstad, where the Randstad shades off into the
less urbanised part of the North-Holland province. It has a large diversity of land
uses and landscapes with a strong north-south orientation. The North Sea borders
the region in the west. The dunes flatten out from the west to the east, where vegetation
cover and building areas increase in importance. Fen pastures are located in the east
of the area. As part of the Randstad, the region is highly urbanised. Haarlem in the
south and IJmuiden in the north are the main urban centres. Haarlem is the largest
city of the two, consisting of differentiated residential neighbourhoods and offering
numerous facilities for work, transport, health care and recreation. IJmuiden has an
east-west orientation following the North Sea Canal and is very much an industrial
city. Next to these two main cities, the study area has several medium-sized and
smaller towns.
Economically speaking, industry is the main employer. The steel mill of Hoogovens,
north of the North Sea Canal, generates a lot of industrial activities. The port, which
is planned to expand near IJmuiden, attracts industrial activities due to its rapid
transport facilities for raw materials. At present, the offshore and transport sector get
a growing share of employment. As weIl as the industrial activities, the business
areas in Beverwijk and Haarlem also generate important amounts of employment.
The economic contribution of the agricultural sector has decreased as urbanization
and the expansion of natural and recreational areas has taken in considerable amounts
of agricultural land. In future, it will keep on decreasing as parts of the remaining
98 G. Engelen et al.

agriculturalland will be claimed by the city of IJmuiden and the port of Arnsterdam
(which is to the north east of the study area).

Source: ANWB Randstad noord en 't Gooi


Fig. 5.3. The Dmond/Zuid-Kennemerland study area

The beach, the dunes and the recreational area of Spaamwoude are of supra-regional
recreational interest. The dunes are a protected naturalarea because of their high
level of biodiversity. The fen pastures of Spaamwoude, the transition areas between
the dunes and the inner dunes, as weH as the transition areas between the inner dunes
and the fen pastures represent high natural values too. SmaHer areas of natural interest
are the river banks. However, the ecological quality of the natural areas decreases
because of groundwater extraction for drinking water. Another cause of the decline
of the biodiversity in the study area is the fragmentation of habitats due to
transportation infrastructure, consisting of two highways (A9, A22) and a railway,
and the increasing distances between remaining habitats.
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 99

From a planning perspective, it is a c1ear intention to bring a halt to the fragmentation


of habitats through the establishment of an ecological network consisting of nature
reserves linked by means of ecological corridors. Similarly, in future, the public
transport system will be extended, as from a sustainability point of view, connections
to the railway (stations) are preferred to highway connections. New urban and other
spatial developments will get good access to the public transport infrastructure in
order to limit traffic congestion within the region.

5.4.2 The Model and the Data

In order to model the DmondlZuid-Kennemerland region, several map layers have


been created in a GIS. These maps consist of 120 by 142 cells, each covering 1
hectare. Thus, each map is composed of 17,040 cells.

Land Use Maps. For the sake of the simulation, two land use maps have been
composed, defining the spatial structure of the region in 1989 and 1997. Data was
made available by Foundation Recreation, a leisure innovation centre, from BORIS
(Basis Openlucht Recreatie Informatie Systeem). The 1989 map is directly derived
from the detailed land use map available from the Central Bureau of Statistics (CBS
Bodemstatistiek), while the 1997 map is a manually updated version of the 1989
map. The 1997 analogue 1/25,000 topographic sheet was used for this update. The
original 26 land use categories have been rec1assified into eight function and six
feature land use categories. In the model, a function is an active and dynamic land
use, which can increase, decrease, or change location during the simulation. A feature
is a static land use, which will remain unchanged during the simulation, but which
will influence the behaviour and location of functions. The defmition of the land
uses and their c1assification are based on a survey carried out by TNO (TNO 1997a,
1997b, Langerak et al. 1997).

Suitability Maps. In the GIS, a suitability map is composed for each function using
a multi-criteria analysis (see Eastman 1997, for example). The relevant criteria are
derived from the survey mentioned earlier (TNO 1997a, 1997b). Of interest in the
determination of a cell's suitability is its position relative to the transport infrastructure,
its position relative to water for transport and cooling purposes (water and transport
network) and its position relative to the natural areas (green network). The cells in
the suitability maps express, by means of a value in the range 0-1 (0 = unsuitable; 1=
highly suitable), the capacity ofthe cell to support the function concemed. Fig. 5.4
shows an example of the suitability map for the distribution and transport function.

Zoning Maps. The actual spatial policies relating to the region are available from
development plans. In the CA model, zoning maps for each land use function were
derived from these development plans. The cells of a zoning map express by means
of values in the range 0-3 whether location is permitted (value =0), or permanently
100 G. Engelen et al.

forbidden (value = 3) for the function concerned. The intermediate values 1 and 2
refer to a 'temporarily not allowed' status, meaning that the cells are available for
development in future development stages to be defined by the analyst, for instance:
from 2000 (value =1) and 2005 (value = 2) onwards. This option has been used to
implement the expansion of the port and the nearby changing development plans for
industry, mixed activities, and distribution and transport from 2005 onwards. Fig.
5.5 shows a zoning map for the these functions.

00.0
00.1
0.2
0.3
. 0 .4
. 0.5
.0.6
.0.7
. 0.8
. 0.9
. 1 .0

Note: See http://www.geog.leeds.ac.uklpublications/GIPEP/ for a colour version


Fig. 5.4. Example of a suitability map for the distribution and transport function

As explained in Section 5.3, the overall spatial claims of the different land use
functions are entered into the model as time series: they constrain the cellular
dynamics. Table 5.1 shows the data derived from the TNO Inro study (TNO 1997a)
and used in the IJmond/Zuid-Kennemerland model. From the table it is clear that
natural and culturalland as well as rural residential is not growing in the region in
the period 1989-2010. Rather all residential development is of the type urban
residential. In absolute numbers, the latter activity claims most land. In relative terms,
both offices and distribution and transport expand considerably. For offices, all the
growth takes place after 1997, whilst the distribution and transport sector expands
most before 1997.
Dynamic GIS and Strategic Physical Planning Support: A Practical Application 101

Note: See http://www.geog.leeds.ac.uk/publications/GIPEP/ for a colour version

Fig. 5.5. Zoning map for the industry, distribution and transport and mixed activities functions

Table 5.1. Spatial claims for land use functions

1989 1997 2005 2010 % change


1989-2010

Urban residential 1663 1823 1966(*) 2029 22.0


Rural residential 211 211 211 211 0.0
Industry 523 523 535 543 3.8
Offices 21 21 35 43 104.7
Distribution and transport 156 235 264 282 80.8
Mixed activities 241 309 351 377 56.4
Natural and cultural 1715 1715 1715 1715 0.0
Note: (*) Actually planned VINEX locations

5.4.3 Calibration of the Model

Since land use maps are available for both 1989 and 1997, it is possible to carry out
a limited calibration of the model for this eight year period. The exercise consisted
in starting off the model with the 1989 data and the 'best' set of CA transition roles
and have it generate as accurately as possible the 1997 land uses. The discrepancies
between the generated 1997 land uses and the actual 1997 land uses are measured on
a cell by cell basis and are further compared by means of the Kappa statistic (calculated
102 G. Engelen et al.

per function). Since features do not change as the result of the dynamics of the model,
only the function cells are relevant in the evaluation of the results of the calibration
exercise. Hence, only the cells coloured in white in Fig.5.6 are the object of the
research. Cells in grey are features; cells in white are functions.

Fig. 5.6. The modelIed area

The first series of calibration runs was performed without zoning maps (p=0 in
equation 5.1). This resulted in a best fit shown in Fig. 5.7 and in values for the Kappa
statistic shown in Table 5.2. Based on the Kappa statistic values, we may conc1ude
that the transition rules obtained from the calibration represent the actuallocational
behaviour of the activities modelled rather weIl. The functions that score a Kappa
statistic value of 1.00 are functions that do not change in size in the period studied.
In other words, there are no cells that need to be added to or removed from the study
area for those functions. However, this does not prevent the model from relocating
all these cells. On the contrary, at each time step, the model will evaluate whether
activities are stilliocated in accordance with the logic of the model. A 'wrong' set of
transition rules could weIl result in the fuH relocation of these activities. But, from
the results of the simulations, we must conc1ude, that no cells were relocated, hence,
that the transition rules represent weH their locational preferences. This result is
partly due to the highly stabilizing effect of the model caused by the inertia parameter
wz.z. o in equation (5.2). This parameter expresses the fact that once an activity has
taken in a particular cell, it will stick to that location as long as possible.
A c10ser look at the maps reveals that, in the absence of zoning maps (Fig. 5.7b),
the development of the Velserbroek urban cluster cannot be accomplished in any of
the simulations performed. The expansion of Santpoort-North towards the East is
too little to compensate for this. Even if the simulation is continued beyond 1997,
this initial expansion will only generate a cluster of sufficient size at a much later
stage. But, developments such as Velserbroek are clearly the result of very strict
zoning regulations, such as they apply in the study area. In a second series of
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 103

calibration runs, we have tested the effect of introducing the zoning regulations (p= 1
in equation 5.1) that applied in the region and for the period studied (1989-1997).
The best set of transition rules obtained from the previous exercise was kept
unchanged. One zoning map was introduced for the urban residential function and
another one for the industry, offices, distribution and transport and mixed activities
functions. For the remaining functions (agriculture, rural residential and natural and
cultural), no zoning maps were introduced in the simulation. As a result, the
discrepancies between the generated 1997 land use map (Fig. 5.7c) and the actual
map reduced considerably. In particular, the model is now able to generate the major
part of the Velserbroek cluster. The functions are underlined in the legend in Fig. 5.7.

a. Actualland use in 1989 b. Simulation in 1997 without zoning maps


.,...--:--...,

c. Simulation in 1997 with zoning maps d. Actualland use in 1997

• Natural and cultural D Protected nature OWater


• Recreational • Roads
• Offices
Rural residential • Distribution and transport • Institutional • Railways
D Industry • Mixed activities

Note: See http://www.geog.leeds.ac.uk/publications/GIPEP/ for a colour version


Fig. 5.7. Results of the calibration run
104 G. Engelen et al.

Table 5.2. Results of the calibration run: Kappa statistic values

Function Values per function of the Kappa statistic


Without zoning maps With zoning maps

Agriculture 0.91 0.97


Urban residentia1 0.91 0.96
Rural residential 1.00 1.00
Industry 1.00 1.00
Offices 1.00 1.00
Distribution and transport 0.85 0.88
Mixed activities 0.81 0.91
Natural and cultural 1.00 1.00

5.4.4. Two Scenarios for 1997-2010

Two scenarios were set up to simulate potential spatial developments in the region
for the period 1997-2010. For both scenarios, the set of transition rules obtained
from the calibration run is used without further changes. The 1997 land use map is
now used as the initial condition and growth coefficients are as shown in Table 5.1.
The first scenario shows a possible evolution of the region in the absence of zoning
regulations. In contrast, in the second scenario, hypothetical zoning plans for the
region are introduced into the model. The maps in Fig. 5.8 are the zoning maps
applied. It is clear from the maps that some cells are exclusively reserved for a
particular function, while others can be taken in by a mixture of activities.
In scenario 1 the urban belt, running from Haarlem in the south to IJmuiden in the
north, fills up completely because of the strong mutual pull forces between urban
clusters (Figs. 5.5 and 5.9a). The urban expansion between Santpoort-North and
Driehuis is striking; the ecological corridor ends up being occupied by urban
residential activities and this occurs despite the fact that north of the North Sea canal
there is very suitable land for urban expansion. However, during the simulation, the
industry, distribution and transport, and mixed activities functions appear on these
locations first, which stops and even repels urban expansion (push force).
The zoning regulations introduced in scenario 2, prevent urban activities from
expanding into the ecological corridors in the region. As a result, expansion of urban
residential activity is now mostly towards the east and to the south (Fig. 5.9b). Due
to the continued intensification of the residential development of the centre of Haarlem
early on in the simulation, the office activities are appearing in a cluster much more
to the south than is the case in scenario 1. This alternative development of the office
cluster starts off at an early stage in the simulation and causes an irreversible change
in the spatial organisation of the region.
Dynamic GIS and Strategie Physical Planning Support: A Practical Application 105

a. Map for the urban residential function b. Map for the offices function
Note: See http://www.geog.leeds.ac.uk/publications/GIPEP/ for a colour version
Fig. 5.8. Zoning maps applied in scenario 2 for the simulation period 1997-2010

a. The result of scenario I, in the b. The result of scenario 2, with the


absence of zoning regulations zoning maps of Figs. 5.5 and 5.8

o Agriculture • Natural and cultural o Protected nature o Water


. ~ential . ~ • Recreational • Roads
Rural residential • Distribution and transport • Institutional • Railways
o Indust'Y • Mixed activities

Note: See http://www.geog.leeds.ac.uk/publications/GIPEP/ for a colour version


Fig. 5.9. Simulated land use in 2010
106 G. Engelen et al.

This shows the importance of the exact timing at which specific areas undergo an
initial small change, usually the appearance of an activity in a single cell, and its
further consequences in the complete reorganisation of much larger areas. This type
of self-organising behaviour of complex systems is what renders the task of physical
planners very difficult but also very necessary. They have to anticipate well in advance
what path of development the system is likely to take and they have to weigh the
pros and cons of such evolution against specific planning objectives. Thus, the model
should be used as a tool to explore the different possible futures of the system. Such
an explorative approach contributes to the design of actions to pilot the system so as
to avoid the worst and hopefully take the most desirable course possible. In this way,
catastrophes can perhaps be avoided and the impact of policy actions can be tested
in an interactive way between the planner and the modelled system. Traditional GIS
systems do not provide the tools that allow this type of exercise.
Typically, in the IJmond case, the simulations run in the absence of zoning
regulations failed to respect the ecological buffer areas which are desired in the
region. Clearly, the introduction of zoning maps that prohibit the development of
socio-economic activities in designated ecological areas did remedy this problem as
was shown in the simulations with zoning regulations. However, if the CA model
presented in this chapter would be further developed for detailed planning purposes,
it would be desirable to extend it with additional transition rules, which would take
into consideration, more directly, a number of important physical planning principles.
This would have the great advantage that the planning principles would produce an
immediate effect in the morphogenesis created by the model; hence the model would
generate activity patterns reflecting the design principles and specifications entered
by the physical planner. We will not further develop this idea here, but will briefly
illustrate it with three examples:
• BuJfer-zones: it is possible to design transition rules that recognise activity clusters
of a specific type and size, check their size against a predefined minimum and
maximum, check their positions relative to one another and check the types and
sizes ofbuffer zones in between clusters. Further, the rules should decide on whether
the clusters can or can not grow and where such growth should take place. In the
dynamics of the system, such rules will clearly influence the size, shape and the
spacing of clusters of activities.
• Fill-in ruIes: in planning practice it is not unusual to fill in gaps in the urban tissue
with specific types of activities with a view to generating entities of a given size or
desired urban mix. Here again, CA rules can be designed that would first fill in
gaps that meet preset conditions before other areas are taken in by expanding
activities.
• Spacing ruIes: for certain types of activities (often social or cultural services) a
minimum or 'optimal' spacing between their geographicallocations is sought.
Transition ruIes could be defined that take into consideration these minimum or
maximum distances and thus would favor the development of specific activities at
specified distances from some land functions and features modelled.
Dynamie GIS and Strategie Physical Planning Support: A Praetieal Applieation 107

5.5 Concluding Remarks

In the introduction of this chapter, three questions were formulated. Tbe first question
relates to the common practice of policy support in the physical planning and decision-
making process concerning spatial intervention in the built environment. It appears
that this common practice must be characterized as rather one-sided in that the more
qualitative (design) working methods precede over the more quantitative (research)
methodologies. Several shortcomings of this one-sided practice have been detected.
First, the communication between the two main parties involved in this process,
spatial designers and researchers, does not work properly (one-off, one-way,
defective). Second, the dominance of the qualitative working methods in practice
conflicts with the need for accountability of its end results (many implicit choices
and assumptions). Third, the described practice is quite old-fashioned. Despite the
fact that the geographical information technology was introduced more than a decade
ago, most of the planning procedures are still performed without the help of this kind
of instrument. Moreover, in the few actual working procedures in which use has
been made of this technology, it appears to be that just a limited part of the potential
functionality and application ofthe technology was used, i.e. database management
and cartographic map-making.
The second question relates to the potential role of geographical information
technology in the physical planning and decision-making process. It was conc1uded
that there c1early appears to be a substantive need for more support in this highly
complex policy preparation phase. In general terms, the contribution it can make in
this procedure is twofold: the cartographic representation function and the analytical
what-if function. Both functions can facilitate the task of making an inventory of all
possible spatial intervention scenarios and identifying their pros and cons. That will
provide a basis on which the implications of an intended policy decision can be
c1arified explicitly beforehand in a conceptual and representational manner, both
spatially (in terms of location) and temporally (in terms of strategy). However, it
was also conc1uded that the what-if function (the spatial analytical functionality,
inc1uding c10se coupling of spatial modelling) is still poorly developed in comparison
to its cartographic representation functionality. It is only quite recently that several
attempts have been undertaken to improve its analytical possibilities, one of which
is the coupling of spatial simulation models (e.g. CA models) to a GIS. In this way,
a dynamic GIS is created which can contribute substantially to the early planning
phase of generating and evaluating alternative spatial intervention scenarios. This
opens the way to a much more quantitative planning procedure which can complement
the present more qualitative physical plan preparation practice.
The third question relates to the way in which the policy support function of
geographical information technology in physical planning and decision-making can
be enhanced through the integrated use of GIS and spatial dynamic modelling. From
the literature it appears that there is a growing number of examples of integrated use of
GIS and spatial simulation modelling. Although these examples hardly make use of
108 G. Engelen et al.

the regular analytical functionality of a GIS Ce.g. network analysis), it shows that this
coupling offers a way to transform the more or less static character of the GIS into a
more dynarnic instrument which can help to detect and visualize the dynarnism and its
direction in changing geographical systems. From the UmondlZuid-Kennemerland
case study, we can conc1ude that CA models coupled to GIS systems allow, in theory at
least, for a type of dynarnic approach which would otherwise be impossible. The
application has shown that these models can be calibrated to represent real world systems
realistically, and that they can be used to test the merits of different policy alternatives.
However, we are far from having the ultimate CA model available that fits the
requirements of the urban or regional planner fully. That is why the authors see it as
their prime task to continue their research and apply the integrated GIS-CA instrument
in a diversity of real world physical planning contexts. These applications should result
in a methodology and supporting tools that make explicit the otherwise irnplicit choices
and assumptions in physical planning practice. That added value of this methodology
and these tools will ultimately determine if a much more complementary qualitative
and quantitative methodology will form the basis of the physical planning and decision-
making process in practice.

Acknowledgement
The authors would like to thank Mrs. drs. luge Uljee for the work she has performed
on the case study of UmondlZuid-Kennemerland.

References
Alexander E.R. 1984. After rationality, what? A revie~ or responses to paradigm breakdown,
Journal ofthe American Planning Association, 50(4): 62-69.
Alexander E.R. 1986. Approaches to Planning; Introducing Current Planning Theories,
Concepts and Issues, Gordon and Breaeh Seienee Publishers, Montreux.
Batty M. 1992. Urban modeling in computer-graphie and geographie information system
environments, Environment and Planning B, 19: 663-688.
Batty M. and Xie Y. 1994. From eells to eities, Environment and Planning B, 21: 31-48.
Bradbury R.H., Laan J.D. van der and MaeDonald B. 1990. Modelling the effeets of predation
and dispers al of the generation of waves of starfish outbreaks, Mathematical Computing
and Modelling, 13: 61-68.
Ceeehini A. and Viola F. 1990. Eine Stadtbausimulation, Wissenschaftliche Zeitschrift der
Hochschule fur Architektur und Bauwesen, 36, 4.
Ceeehini A. and Viola F. 1992. FicTzes fictitious cities: a simulation for the creation of cities,
Paper presented at the International Seminar on Cellular Automata for Regional Analysis,
Dipartimento di Analisi Eeonomica e Soeiale deI Territorio dei Arehitettura di Venezia,
Venice.
Chou H. and Ding Y. 1992. Methodology of integrating spatial analysis/modeling and GIS, in
Proceedings Fifth International Symposium on Spatial Data Handling, IGU, Charleston:
514-523.
Clarke K.C., Hoppen S. and Gaydos I. 1997. A self-modifying eellular automaton model of
historical urbanization in the San Franeisco Bay area, Environment and Planning B, 24:
247-261.
Dynamic GIS and Strategic Physical Planning Support: A Practical Application 109

Couclelis H. 1985. Cellular worlds: a framework for modelling micro-macro dynamics,


Environment and Planning A, 17: 585-596.
Couclelis H. 1988. Of mice and men: what rodent populations can teach us about complex
spatial dynarnics, Environment and Planning A, 20: 99-109.
Couclelis H. 1991. Requirements for planning-relevant GIS: a spatial perspective, Papers in
Regional Science, 70: 9-19.
Couclelis H. 1997. From cellular automata to urban models: new principles for model
development and implementation, Environment and Planning B, 24: 165-174.
Cowen D.J. 1988. GIS versus CAD versus DBMS: what are the differences?, Photogrammetric
Engineering and Remote Sensing, 54: 1551-1555.
Dave B. and Schrnitt G. 1994. Information systems for urban analysis and design development,
Environment and Planning B, 21: 83-96.
Deursen W.P.A. van 1995. Geographical information systems and dynamic models;
development and application of a prototype spatial modelling language, NGS 190, KNAG/
Faculteit Ruimtelijke Wetenschappen, Universiteit Utrecht, Utrecht.
Eastman J. R. 1997. ldrisifor Windows, Users Guide, Clarke University, Worcester, MA,
USA.
Engelen G., Uljee 1. and White R. 1997. Vulnerability Assessment ofLow-lying Coastal Areas
and Smallislands to Climate Change and Sea Level Rise, UNEP United Nations Environment
Program, CARlRCU, Jamaica, pp. 101.
Engelen G., White R. and Uljee 1. 1993. Exploratory modelling of socio-economic impacts of
climatic change, in Maul G.A. (ed.) Climate Change in the 1ntra-Americas Sea, Edward
Amold, London: 306-324.
Engelen G., White R. and Uljee 1. 1997. Integrating constrained cellular automata models,
GIS and decision support tools for urban planning and policy-making, in Timmermans H.
(ed.) Decision Support Systems in Urban Planning, E & F Spon, London: 125-155.
Engelen G., White R. and Uljee I. 1998. De LeefOmgevingsVerkenner. 'Proof of Concept'-
Versie van een 1ntgraal Model voor het Berekenen van het LeefOmgevingsKapitaal in
Nederland, RIVM, Bilthoven.
Engelen G., White R., Uljee 1. and Drazan P. 1995. Using cellular automata for integrated
modelling of socio-environmental systems, Environmental Monitoring andAssessment, 34:
203-214.
Enge1en G., White R., Uljee 1. and Wargnies S. 1996. Numerical modeling of small island
socio-econornics to achieve sustainab1e development, in Maul G.A. (ed) Smallislands.
Marine Science and Sustainable Development, American Geophysical Union, Washington
DC, Coastal and Estuarine Studies, 51: 437-463.
Eweg R. 1994. Computer supported Reconnaissance Planning: 1mplementing a Planning
Methodology with Geographical Information System in Noord-Brabant, The Netherlands,
Landbouwuniversiteit Wageningen, Wageningen.
Faludi A.and Valk A. van der 1994. Rule and Order: Dutch Planning Doctrine in the Twentieth
Century, Kluwer, Dordrecht.
Fedra K. 1995. From spatial data to spatial information: GIS, environmental models, and
expert systems, in JECG1'95 Proceedings 1, AKM, Basel: 264-278.
Gardner M. 1970. The fantastic combinations of John Conway's new solitaire game Life,
Scientific American, 223: 120-123
Geertman S.C.M. 1992. Physical planning and information technology; the integration dispute,
in Proceedings ofthe TwelthAnnual ESR1 User Conference, ESRI, Palm Springs: 133-144.
110 G. Engelen et al.

Geertman S.C.M. 1996. Physical Planning and Geographical Information: A Reconnaissance


ofGeo-IT Methodology, Van Gorcum, Assen.
Geertman S.C.M. 1998. Strategie Physical Planning and Geographical Information Technology:
Some Methodological Reflections (forthcoming).
Green D.G. 1993. Emergent behaviour in biological systems, in Green D.G. and Bossomaier
TJ. (eds.) Complex Systems - From Biology to Computation, lOS Press, Amsterdam: 24-35.
Holmberg X.C. 1994. Geoinformatics for urban and regional planning, Environment and
Planning B, 21: 5-19.
Itami 1994. Simulating spatial dynamies: cellular automata theory, Landscape and Urban
Planning, 30: 27-47.
Jankowski P. and Richard L. 1994. Integration of GIS-based suitability analysis and multicriteria
evaluation in a spatial decision support system for route selection, Environment and Planning
B, 21: 323-340.
Jeffers J. 1988. Decision-thinking about land use, Land Use Policy, 5: 75-78.
Kevany MJ. 1989. Relationship between GIS and CADD technologies, UDMS' 89 Proceedings,
Lisboa: 217-227.
Lammeren R. van 1994. Computergebruik in de Ruimtelijke Planning; Methodologische
Aspecten van Ruimtelijke Planvorming met Behulp van Informatieverwerkende Systemen,
Landbouwuniversiteit Wageningen, Wageningen.
Langerak L., Geertman S. and Wesseis C. 1997. Strategie van de Drie Netwerken als Ruimtelijk
Planconcept, NexpRi Info, Utrecht.
Liggett R.S. and Jepson W.H. 1995. An integrated environment for urban simulation,
Environment and Planning B, 22: 291-302
Lohman de Savornin A. 1998. Landgebruik in beweging. Het gebruik van cellular automata
voor ruimtelijke planning, Doctoraalverslag, Landbouwuniversiteit Wageningen,
Wageningen.
Masri A. and Moore 11 lE. 1994. Integrated planning information systems: context, design
requirements, and prospects, Computers, Environment and Urban Systems, 17(6): 491-511.
Massam B.H. 1992. The Right Place: Shared Responsibility and the Location of Public
Facilities, Longman, Harlow.
Newell R.G. and Th6riault D.G. 1990. CAD and GIS - chalk and cheese?, The 1990 GIS
Sourcebook: 241-245.
Nijkamp P. and Scholten H. 1993. Spatial information systems: design, modelling, and use in
planning, International Journal ofGeographical Information Systems, 7(1): 85-96.
Phipps M. 1989. Dynamical behaviour of cellular automata under constraints of neighbourhood
coherence, GeographicalAnalysis, 21: 197-215.
Prigogine I. 1981. From Being to Becoming, Freeman, San Francisco.
Rittel H. and Webber M. 1984. Dilemmas in a general theory of planning, in Cross N. (ed.)
Developments in Design Methodology, Wiley, Chichester: 135-144.
Rouge M.F. 1947. La Geonomie ou I 'Organisation de l'Espace, Paris.
Roy G. G. and Snickars F. 1993. City Life: a study of cellular automata in urban dynamies,
Paper presented at the 33rd European Congress of the Regional Science Association,
Moscow, August 24-27.
Schuur J. 1994. Analysis and design in computer-aided physical planning, Environment and
Planning B, 21: 97-108.
Takeyama M. 1996. Geo-Algebra: A MathematicalApproach to Integrating Spatial Modeling
and GIS, Unpublished PhD thesis, Department of Geography, University of California,
Santa Barbara.
Dynamie GIS and Strategie Physieal Planning Support: A Praetieal Applieation 111

Tobler W. 1979. Cellular geography, in Gale S. and Olsson G. (eds.) Philosophy in Geography,
Reidel, Dordreeht: 379-386.
TNO Inro, Afdeling Planning. 1997a. IjmondlZuid-Kennemerland - Drie Netwerken Atlas,
Delft.
TNO Inro, Afdeling Planning. 1997b. De strategie van de drie netwerken als ruimtelijk
planconcept, TNO Inro 97/NP-076, Delft.
Wagner D.F. 1997. Cellular automata and geographie information systems, Environment and
Planning B, 24: 219-234.
Webster C.J. 1993. GIS and the seientific inputs to planning. Part 1: description, Environment
and Planning B, 20: 615-748.
Webster CJ. 1994. GIS and the scientific inputs to planning. Part 2: prediction and prescription,
Environment and Planning B, 21: 145-157.
White R and Engelen G. 1993. Cellular automata and fractal urban form: a cellular modelling
approach to the evolution of urban land use patterns, Environment and Planning A, 25:
1175-1199.
White R and Engelen G. 1994 Cellular dynarnics and GIS: modelling spatial complexity,
Geographical Systems, 1: 237-253.
White R. and Engelen G. 1997. Cellular automata as the basis of integrated dynamic regional
modelling, Environment and Planning B, 24: 235-246.
White R, Engelen G. and Uljee I. 1997. The use of constrained cellular automata for high-
resolution modelling of urban land use dynarnics, Environment and Planning B, 24: 323-343.
White R, Uljee 1. and Engelen G. 1998. A nested cellular automata model for the St. John's
Metropolitan Area. (in preparation).
Worrall L. 1994. The role of GIS-based spatial analysis in strategie management in local
government, Computers, Environment and Urban Systems, 18(5): 323-332.
Wright lR, Wiggins L.L., Jain RK. and Kim TJ. (eds.). 1993. Expert Systems in Environmental
Planning, Springer-Verlag, Berlin.
Wu F. 1998. SimLand: a prototype to simulate land conversion through the integrated GIS
and CA with AHP-derived transition rules, International Journal of Geographical
Information Science, 12, 1: 63-82.

View publication stats

You might also like