VU Research Portal
Residential construction, land use and the environment. Simulations for The
Netherlands using a GIS-based land use model
Schotten, C.G.J.; Goetgeluk, R.; Hilferink, M.T.A.; Rietveld, P.; Scholten, H.J.
published in
Environmental Modeling and Assessment
2001
DOI (link to publisher)
10.1023/A:1011531120436
document version
Publisher's PDF, also known as Version of record
Link to publication in VU Research Portal
citation for published version (APA)
Schotten, C. G. J., Goetgeluk, R., Hilferink, M. T. A., Rietveld, P., & Scholten, H. J. (2001). Residential
construction, land use and the environment. Simulations for The Netherlands using a GIS-based land use
model. Environmental Modeling and Assessment, 6, 133-143. https://doi.org/10.1023/A:1011531120436
General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.
• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal ?
Take down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
E-mail address:
vuresearchportal.ub@vu.nl
Download date: 09. Aug. 2022
Environmental Modeling and Assessment 6: 133–143, 2001.
2001 Kluwer Academic Publishers. Printed in the Netherlands.
Residential construction, land use and the environment.
Simulations for the Netherlands using a GIS-based land use model
Kees Schotten
a
, Roland Goetgeluk
b
, Maarten Hilferink
c
, Piet Rietveld
d
and Henk Scholten
d
a
RIVM, Bilthoven, The Netherlands
b
Universidade Pedagogica, Beira, Mozambique
c
Object Vision, Haarlem, The Netherlands
d
Vrije Universiteit, Amsterdam, The Netherlands
E-mail: prietveld@econ.vu.nl
The present generation of geographical information systems supports strategic planning processes in several ways. They are able to
store, manage and analyse the enormous amount of data needed. Another more output-oriented use is the visualisation of the diversity
of locational preferences and perspectives of different interest groups and stakeholders. For the simulation of (more indirect) effects of
autonomous or planned developments land use modelling can be applied. A step further is the definition and implementation of a set of
indicators that show the impact of land use change on different aspects of space and the environment in order to facilitate the (political)
discussions, that are an essential part of strategic planning.
This paper focuses on the application of a GIS-based simulation model in the framework of the Fifth National Physical Planning Report
in the Netherlands. The simulation model generates future land use in the Netherlands given several growth scenarios and a spatial strategy
that comprises both foreseen strategic and autonomous developments. Special attention is paid to residential construction because this is
expected to be one of the major driving forces in land use changes. An analysis of residential construction for the period 1980–1995 reveals
that residential construction has been relatively concentrated in areas close to existing urban areas. New town policies also played a rather
strong role during this period. The presence of natural areas (woods and wetlands) plays a significant though limited role in the choice
where to build new dwellings. The simulation results for the year 2020 are used to assess the effects of land use changes for a range of
environmental indicators.
1. Introduction
Spatial planning has returned to the forefront of public
and politic attention. In Europe this is illustrated by the
recent adoption of the European Spatial Development Per-
spective (ESDP) at the Ministerial conference in Tampere.
At the regional and national level numerous examples can
be given such as the release of the (first) Regional Plan for
Lisbon Metropolitan Area (PROTALM) in Portugal and the
Fifth National Physical Planning Report in the Netherlands.
Spatial or physical planning can be seen as a strategic
spatial policy aimed at finding the optimum adjustment of
space and society. Often the solutions generated by the par-
ties involved, who all have different interests, objectives and
preferences, do not lead to satisfactory results. Therefore,
governments are often involved in the planning process that
is described by Healey [1] as a set of governance practices
for the development and implementation of spatial strate-
gies, plans, policies and projects by regulating the location,
timing and form of development. These practices are shaped
by the dynamics of economic and social change, which give
rise to demands of space.
The present generation of geographical information sys-
tems (GIS) is used to support these strategic planning
processes in different ways. Their ability to store, manage,
retrieve and visualise the enormous amount of data regard-
ing spatial objects needed in the planning process is often
applied (e.g., [2]). Also spatial analytical functions to assess
relevant spatial patterns are widely used. Another develop-
ment is the linkage of GIS with location or allocation mod-
els and more recently the integration of accessible computer
based applications or Decision Support Systems.
Various model approaches can be distinguished in the
field of land use (cf. [3]):
1. Planning models. These models compute an optimal al-
location of land in order to arrive at a maximum value
for some optimisation criterion (for example, maximum
profit of a farm), or to maximise some social welfare ob-
jective strived after by a public sector planner (see, e.g.,
[4]).
2. Individual choice models. These models describe the lo-
cational preferences of individual actors on the basis of
individual decision process [5].
3. Al-models derived from the field of Artificial Intelligence.
Examples of these allocation algorithms are Neural Net-
works (NN), Genetic Algorithms and Cellular Automata
(CA). Applications of Al-models in the field of allocation
planning are, e.g., described in [6–9].
4. Equilibrium models explaining land use in terms of sup-
ply, demand and equilibrium prices. These models are
part of the tradition of the classical approaches of Von
Thunen, Losch and Alonso. In most of these models
transport and accessibility play an important role as de-
terminants of the equilibrium outcomes of land use. Ex-
134 K. Schotten et al. / Residential construction, land use and the environment
amples of operational models used in planning contexts
are those of Anas [10] and Landis [11].
In this paper we will use make use of a model of the latter
category.
An interesting development is that these land use models
are linked to environmental models in order to be able to
assess the environmental consequences of land use change
(for a review see [12]. For example, changes in land use will
have consequences for spatial interaction patterns and these,
in turn, will have an impact on emission and noise caused by
transport (see also [13]).
This paper focuses on the application of a GIS-based land
use model to the Fifth National Physical Planning Report
in the Netherlands. The model generates future land use in
the Netherlands given several growth scenarios and a spa-
tial strategy that comprises both foreseen strategic and au-
tonomous developments. The simulation results, a map with
residential areas and an integrated land use map, are used to
assess the effects of land use changes on a wide range of en-
vironmental indicators. We focus on land use for residential
purposes because this is a very dynamic land use category in
the Netherlands. In terms of modelling, most attention will
be paid to the Land Use Scanner, a GIS-based model for the
Netherlands. Models assessing land use implications for the
environment will only be discussed briefly.
The structure of the paper is as follows. We start with
a short review of the backgrounds of physical planning in
the Netherlands and of possible changes in external condi-
tions (section 2). The land use scanner model is presented in
more detail in section 3. Changes in land use for residential
purposes during the period 1980–1995 are analysed in sec-
tion 4. This is followed by a simulation of future patterns in
residential construction in section 5. Environmental impacts
are analysed in section 6.
2. Background of the study
The Netherlands is among the most densely populated
countries in the world. Given the high externalities that oc-
cur in land use, it is no surprise that the government is quite
active in interfering with the land market. Especially in de-
cisions on the location of residential areas the government
plays a strong role. The long run national policies on land
use are formulated in strategic policy memoranda, the last
of which appeared in 1988. The next document called the
Fifth National Physical Planning Report will be published in
2000. In the preparatory phase the National Institute of Pub-
lic Health and Environment (RIVM) analysed basic uncer-
tainties for the future by means of scenarios. Possible conse-
quences for land use and the environment during the period
1995–2020 (the Fifth Report’s time span) were investigated.
Aim of the study is not to deliver a blueprint for land
use in 2020, nor to simply quantify the impacts of land use
change on the environment, but to establish a objectified
methodology making use of models to enable discussions
on future land use developments. Of course, the discussions
can embrace the necessary operational assumptions made in
order to obtain results and can also be used to gain under-
standing of land use dynamics and (indirect) effects on the
environment. The main goal, however, is to enable the eval-
uation of spatial policies and to assess the environmental ef-
fects of alternative spatial policies.
2.1. Scenarios
An important element that shapes spatial planning prac-
tices is the expected rise or decline in demand for space that
results from the dynamics in economic and social change.
In order to cope with the uncertainties of future demands
for space the concept of scenarios is often used [14]. In
this study three internally consistent and coherent scenarios,
developed by The Dutch Central Planning Bureau [15], are
used to describe demographic, economic, social and policy
developments until 2020. These scenarios can be charac-
terised as follows.
– Divided Europe (DE): protectionistic tendencies in the
different European countries are strong. Economic
growth is low with an annual rise in BNP of approx.
1.7%. The population grows to 16.3 million while the
number of dwellings rises to 7.5 million (currently 15
and 6.2 million, respectively). The Common Agricultural
Policy (CAP) is continued.
– European Co-ordination (EC): The world economy is
dominated by trade organisations resulting in an accel-
erated integration of the European countries. The annual
economic growth is moderate (3.0%) but the growth in
population to a total of 17.7 million is the highest of all
scenarios. The number of dwellings grows to 7.8 mil-
lion. A unified European policy is implemented implying
a certain level of protection of agriculture in the common
market.
– Global Competition (GC): Market forces dominate the
world economy and the economic growth is high (3.3%).
The number of inhabitants rises to 16.9 million. How-
ever, due to a more individualistic life style the number
of dwellings grows to 8.1 million, faster than in the EC
scenario. The CAP is abandoned forcing the agricultural
sector to compete on the world market.
Taking the developments described in the CPB scenarios as
a starting point different sector-specific models are used to
calculate the amount of land needed for agriculture, housing,
industrial and commercial areas in each of the scenarios. The
results are shown in figure 1.
The demands for nature and (rail and motorway) in-
frastructure in the Netherlands are determined by policies
that have already been agreed, so that the claims of these
land use types are identical in each of the scenarios. Na-
ture (including forest) is also the land use type that claims
the largest amount of land until 2020 in each of the scenar-
ios. While the claims for housing, industry and commerce
are less than that for nature, they also vary due to the dif-
ference in foreseen dwellings and economic growth within
K. Schotten et al. / Residential construction, land use and the environment 135
Figure 1. Sectoral land use demands in the three scenarios.
the scenarios. In all scenarios the area used for agriculture
declines, only the rate differs in each scenario; high in the
Divided Europe and Global Competition scenarios and low
in European Coordination. For the Netherlands this results
in a surplus of land in the GC scenario and a more or less
balanced situation in the DE-scenario. In the EC scenario
a situation is reached where the demand for land exceeds
the supply by 46000 ha. This calls for a reconsideration of
the claims computed by the sectoral models. An integrated
analysis of demand for land across various land use types is
necessary to arrive at such a conclusion. Such an integrated
analysis will be the subject of the next section. The model to
be used for this purpose is the LAND USE SCANNER and
it will be discussed in the next section.
3. The LAND USE SCANNER model
3.1. General features
We start with a short characterisation of the properties of
the LAND USE SCANNER. Some characteristic features of
the model are:
• Grid based. The model describes for all grids in a system
the relative proportions of land to be used for a number
of land use types. Model specification and software al-
low large numbers of grids. The present version covers
193,399 grid cells of 500 by 500 meters each, covering
all of the Netherlands.
• Integrated. The model provides an integration framework
for sectoral data bases and sectoral policy proposals by
confronting these inputs in a spatial-analytical context.
• Exhaustive. The model is exhaustive in the sense that all
grids in a spatial unit (in our case a country) are consid-
ered. All types of land use are explicitly considered; thus
there are no “rest” categories left untreated. The model
can be formulated in such a way that transfers of wet
grids (sea, lakes) into land are allowed.
• Dynamic. The model deals with changes in land use tak-
ing into account present land use patterns. The suitabil-
ities of the grids for certain types of land use are not as-
sumed constant, but may change as the result of changes
in land use in the course of time.
• Satellite structure. The model is driven by forecasts at
a national or regional level in terms of variables such as
population, agricultural production, infrastructure, etc.
• Stochastic. The outcomes of the model are to be inter-
preted as expected proportions of land to be used for var-
ious types of land use categories. The use of the model is
not that it predicts land use in particular small grids in the
future. The main use of the model is that it gives the im-
plications for the spatial patterns of land use of processes
such as population growth, production (manufacturing,
agriculture, etc.), and natural conservation.
• Policy oriented. Several types of sectoral policies have
strong spatial implications. LAND USE SCANNER
makes these implications explicit. The model helps solv-
ing questions referring to the types of grids in which ma-
jor policy conflicts can be expected to emerge. It can
also be used to investigate the implications of sectoral
and macro policies for human settlement and land use
patterns.
The property of integration means that the LAND USE
SCANNER can function as a tool to improve communica-
tion between analysts working in various fields of land use
(for example, urban functions versus agriculture versus nat-
ural land use). The model also helps to improve consistency
between projections made in these fields. Thus a potential
use of the LAND USE SCANNER is that is does not only
function as a modelling tool, but also as a communication
tool between analysts in various policy fields.
The following types of land use are distinguished in the
present version of the model:
1. Urban: residential, industrial, roads, railways, and air-
ports.
2. Agriculture: pasture, corn, arable land (potatoes, beets,
cereals), flower bulbs, orchards, cultivation under glass,
and other agriculture.
3. Natural areas: wood, nature.
4. Water.
136 K. Schotten et al. / Residential construction, land use and the environment
We arrive at 15 different land use types. This number of
land use types can be extended. Data are available for more
finely meshed distinctions, thus leading to up to 40 land use
categories.
3.2. Regional constraints, suitability maps and government
interventions
The LAND USE SCANNER model is driven by out-
comes of other sector specific models which generate re-
sults at a much lower spatial detail. The outcomes relate
to the year 2020. The projections for agricultural land use
and urban functions have been made for the various scenar-
ios mentioned in the preceding section. Thus, a check has to
be conducted to ensure consistency of inputs (see below).
Projections of demand for land at the regional level are
available among others for various types of agriculture are
available for agricultural regions, of which there are 14 in
the Netherlands. For residential and industrial areas pro-
jections are given at the level of so-called COROP regions.
There are 40 COROP regions in the Netherlands. For natural
areas regional constraints are included for 66 regions. The
constraints imply that the total amount of land used by na-
ture must be at least the amount which is set as a minimum
amount for each region in government plans.
The demand side of the land market is represented by
suitability maps. A suitability map is generated for each land
use type to indicate the suitability of each grid cell for that
type. These suitability maps can be interpreted as bid-prices.
The suitabilities depend on factors such as:
• soil quality,
• transition costs, given previous land use,
• accessibility of facilities and infrastructure,
• amount of similar land use in the neighbourhood.
In Hilferink and Rietveld [16] more details are given about
the computation of suitability maps.
In addition to suitabilities expressing bid prices of market
actors, governmental planning regulations have an impact on
land use developments. These include:
• policies towards building permits for residential and in-
dustrial land use,
• policies regarding the preservation and development of
natural areas.
These policy interventions can be interpreted as subsidies
and taxes for certain types of land use so that the final market
outcome does not merely reflect the willingness to pay of
actors at the market. Also the intentions of the public sector
to stimulate or discourage land use of particular types are to
some extent reflected by the resulting land use patterns.
3.3. Mathematical formulation
A core variable of the model is the suitability s
cj
for land
use of type j in grid cell c. This suitability represents the net
benefits (benefits minus costs) of land use type j in cell c.
The higher the suitability for land use type j , the higher the
probability x
cj
that the cell will be used for this type. In the
simplest version of our model we use a logit type approach
to determine this probability:
x
cj
=
exp(β · s
cj
)
j
exp(β · s
cj
)
for all c and j. (1)
Thus, when β is zero, all types of land use have the same
probability; i.e., the suitability factors s
cj
do not play any
role in determining these shares. On the other hand, when
β goes to infinite, the limit of probability that the category
with the highest suitability gets the cell is equal to 1.
In terms of expected values, the expected volume of land
use L
cj
for category j in cell c equals:
L
cj
= x
cj
· L
c
for all c and j, (2)
where L
c
denotes the total volume of land in cell c . With
equally sized cells L
c
would of course be equal for all c.Un-
equally sized cells may occur in the case of cells located near
the national border, or cells being partly water (if a transfer
from water to non-water land use is not allowed), or contain
preset land use based on exogenous data, such as infrastruc-
ture developments.
The model as formulated here does not guarantee that the
allocation of space across possible land uses is in accordance
with overall demand conditions. Therefore, side constraints
have to be imposed in order to ensure that at the relevant
levels of aggregation total demand is met.
This leads to a reformulation of the model. Let D
j
be
a restriction on total demand for land use category j .In
addition, let M
cj
denote the expected amount of land in cell
c that will be used for category j taking into account the side
constraints. We then arrive at a doubly constrained model:
M
cj
= a
j
· b
c
· exp β · s
cj
for the constrained j and all c,
(3)
where a
j
and b
c
are balancing factors such that the following
constraints are satisfied:
c
M
cj
= D
j
for the constrained j, (4)
j
M
cj
= L
c
for all c. (5)
Equation (4) guarantees that the expected amount of land al-
located for land use type j equals the imposed amount D
j
.
In addition, equation (5) implies that the sum of the expected
volumes of the various land use types per cell is equal to
the total area of each cell. We use the expression “for the
constrained j ”whenanaggregate constraint has been for-
mulated for the particular land use type j . It is clear that
the constraints may imply that no feasible solution exists.
This can be checked by seeking for a starting solution of the
system in a linear programming context. When no feasible
solution is found, the aggregate constraints have to be re-
considered before the model can be used. For those land use
