Modelling protocol for landuse and climate change
effects
Deliverable D5.3
Part II
2
Deliverable summary
Project title
Acronym
Date due
Final version submitted to EC
Complete references
Groundwater and Dependent Ecosystems: New Scientific and
Technological Basis for Assessing Climate Change and Land-use
Impacts on Groundwater
GENESIS
Contract number 226536
Month 30 in GENESIS
Month 38 in GENESIS
Contact person
Contact information
Authors and their affiliation
Karim C. Abbaspour, EAWAG
[email protected]
Karim C. Abbaspour, EAWAG
Pertti Ala-aho, UOULU
Kiriakos Stefanopoulos, DUTh
Ali Ertürk, IGEM
Project homepage
Confidentiality
Key words
www.thegenesisproject.eu
Public
Modelling, landuse change, climate change
Deliverable 5.3 to present an overview of spatial data, inputs, tools
and method for constructing spatially explicit landuse scenarios. The
landuse models have a common point: to simulate the landscape
dynamic for the future at multiple scales based on different coherent
assumptions. These models can improve the understanding and
sensitivity of key processes of landuse patterns.
There are different landuse modeling approaches based on relatively
complex theories and methodologies. Most models tend to apply
black-boxes procedures. The landuse models are mechanistically
robust, but owing to the complexity of the system and data
interpretation, their results is rather restricted for academic research
purposes.
In this document we review four commonly used models: Land Change
Modeller (LCM), CA_Markov, METRONAMICA and CLUE. However final
model selection remains an open issue. Although all four models can
generally fulfill the same application, they differ in terms of specific
features such as level of complexity, information on model
development, optimization of model input processing, cost, and
functionalities for specific end-user needs.
Based on a comparative assessment of the latest European land cover
maps MODIS, GlobCover, and Corine were assessed as suitable for
spatially explicit accounting of main land cover categories. Main
categories of thematic data needed for landuse modeling as
biophysical and climate datasets, protected areas and transport
infrastructure were reviewed and first selection of sources are
presented.
This report presents an overview of main themes related to landuse
modeling and required inputs. It starts with an overview of existing
applications and review of past experiences. Then the process of land
Summary (publishable) for
policy uptake
3
modeling with practical examples from a test site is presented,
followed by an overview of latest available modeling tools. Afterwards
the document presents a review of the available datasets to be
applied as modeling inputs, including land-cover, thematic data as
well as datasets useful for validation of the modeling output.
4
List of GENESIS partners
Norwegian Institute for Agricultural and Environmental Research (CO)
Bioforsk
Norway
UOULU
Finland
Joanneum Research Forschungsgesellschaft mbH
JR
Austria
Swiss Federal Institute of Technology Zurich
ETH
Switzerland
Luleå University of Technology
LUT
Sweden
University of Bucharest
UB
Romania
GIS-Geoindustry, s.r.o.
GIS
Czech Republic
French National institute for Agricultural research
INRA
France
Alterra - Wageningen University and Research Centre
Alterra
The Netherlands
Helmholtz München Gesundheit Umwelt
HMGU
Germany
Swiss Federal Institute of Aquatic Science and Technology
EAWAG
Switzerland
University of Science and Technology
AGH
Poland
Università Cattolica del Sacro Cuore
UCSC
Italy
Integrated Global Ecosystem Management Research and Consulting Co.
IGEM
Turkey
Technical University of Valencia
UPVLC
Spain
Democritus University of Thrace
DUTh
Greece
Cracow University of Technology
CUT
Poland
University of Neuchâtel
UNINE
Switzerland
University of Ferrara
UNIFE
Italy
Athens University of Economics and Business- Research Centre
AUEB-RC
Greece
University of Dundee
UNIVDUN
United Kingdom
University of Zagreb - Faculty of Mining, Geology and Petroleum
Engineering
UNIZG-RGNF
Croatia
Helmholtz Centre for Environmental Research
UFZ
Germany
Swedish Meteorological and Hydrological Institute
SMHI
Sweden
University of Manchester
UNIMAN
United Kingdom
University of Oulu
5
Table of contents
1. Summary ................................................................................................................................................................7
2. Landuse classification systems ............................................................................................................................11
2.1. The CORINE land cover project of 1990............................................................................................................11
2.2. The GlobeCorine landcover ..............................................................................................................................13
3 Models of landuse change ....................................................................................................................................20
APPENDIX I - Step by step Modeling with Dyna-CLUE .............................................................................................47
Appendix II - Implementation of land use and climate change scenarios at Rokua aquifer ...................................52
1. Conceptual Model For Scenarios .................................................................................................... 52
2. Climate Change Scenarios ............................................................................................................... 52
3. Land Use Scenarios ......................................................................................................................... 54
4. Model parameterization and uncertainties .................................................................................... 57
Appendix III - Implementation of land use and climate change scenarios ..............................................................61
1.1. Climate change implementation.................................................................................................. 62
1.2. Land use change........................................................................................................................... 66
1.3. Assessment of uncertainty of groundwater flow model parameters and components ............. 68
Appendix IV..............................................................................................................................................................70
6
1. Summary
1.1. Overview of Issues Related to Landuse Change
Human use of land resources gives rise to "landuse" which varies with the purposes it serves,
whether food production, provision of shelter, recreation, extraction and processing of materials,
and so on, as well as the bio-physical characteristics of land itself. Hence, landuse is being
shaped under the influence of two broad sets of forces – human needs and environmental
features and processes. These forces are in a constant state of flux and change. Changes in the
uses of land occur at various spatial levels and within various time periods. These changes have
at times beneficial and at times detrimental impacts and effects. The magnitude of landuse
change varies with the time period being examined as well as with the geographical area.
Moreover, assessments of these changes depend on the source, the definitions of landuse types,
the spatial groupings, and the data sets used.
1.2. The Purpose of the Analysis of Landuse Change
The approaches taken for the analysis of landuse change are determined critically by the
analyst’s objectives. The definitions and landuse classification systems used, the theoretical
schemata adopted and the models employed all depend on the main questions and the needs of
the analysis. We discuss here six main categories, which are characteristic purposes of landuse
change analyses. These are: description, explanation, prediction, impact assessment, prescription
and evaluation.
Descriptive studies of landuse change are almost indispensable in any analytical endeavor as a
first step towards more refined analyses. Description of landuse change documents changes from
one type of landuse to another over a given time period and within a given spatial entity.
Changes in both the qualitative as well as the quantitative characteristics of landuse are
described, the level of detail conditioned by the spatial level of analysis and the availability of
necessary data. Descriptive studies of landuse change have provided the impetus for more
thorough investigations of the "why" of these changes as well as for taking actions (policies) to
counteract the negative impacts of the changes.
Description alone, however detailed and thorough it may be, is not enough to provide the basis
for understanding the observed landuse changes or to guide policy and decision. Explanatory
analyses attempt to fill this gap. Explanation attempts to address the question of "why" these
changes have occurred (or, are occurring) and to uncover the factors or forces that bring about
these changes directly or indirectly, in the short or the longer run. The level of explanation
7
offered by any study is a matter of the chosen spatial and temporal level of analysis. Macroanalyses necessarily refer to global changes and take into account global explanatory factors or
determinants of landuse change. As the analysis moves towards lower spatial levels, explanation
moves deeper into the social and psychological dynamics that underlie observed human behavior
and, consequently, landuse change. Similarly, explanatory analyses over long time periods
attempt to reveal the macro-forces that induce landuse changes such as social, cultural and
technological change. On the contrary, short-term explanatory analyses necessarily seek for more
immediate factors affecting human behavior that leads to landuse change although the influence
of the larger macro-forces can be taken into account as conditioning the shorter-term phenomena.
Explanatory studies employ more or less specific theoretical schemata that account for the main
determinants of landuse change and their intricate interrelationships.
In addition to describing and explaining landuse change, an important purpose for conducting
such analyses is to predict future changes in landuse. Predictions may be unconditional or
conditional. Unconditional predictions, also called trend extrapolations, provide future images of
the landuse patterns in an area if past trends continue into the future. Unconditional predictions
may be mechanistic extrapolations of past landuse change or, if they are informed by theory,
they may be more thorough projections of past trends in the determinants and the resulting
landuse change into the future. Conditional predictions of landuse change produce alternative
landuse futures of an area under hypothetical conditions or scenarios. Some analyses are
conducted with the purpose of predicting landuse changes caused by climatic change or by
changes in future population, food and other habits and so on. Conditional predictions, based
usually on scenario analysis, are frequently used in the context of policy making on issues of
global change. In both unconditional and conditional predictions, the critical issues are again the
spatial and temporal level of analysis.
Impact Assessment is another important purpose of the analysis of landuse change. The
contemporary interest is not so much on landuse change itself as is on its various environmental
and socio-economic impacts at all spatial levels. In addition, as policies are designed to address
several of the environmental and socio-economic problems in which landuse change contributes
in one way or another, policy impact assessment has emerged as a significant scientific activity.
The recent policy interest, specifically, is on the broader issue of sustainability of development as
it is impacted by landuse change triggered by proposed or implemented policies. Landuse
changes with adverse impacts – such as land degradation, desertification, depopulation, etc.
contribute negatively to the achievement of long term sustainability as they reduce the natural,
economic, human, and social capital available to future generations.
In a normative perspective, the analysis of landuse change may seek to address the question of
"what should be"; in other words, the purpose is to prescribe landuse configurations that ensure
the achievement of particular goals. Presently, these goals come under the broad search for
8
"sustainable landuse solutions". The purpose of this type of analysis is to indicate those patterns
of landuse that are associated with environmental preservation, economic prosperity and welfare
and social equity.
Finally, evaluation may be undertaken for assessing either past, present or future policy-driven
changes in patterns of landuse in terms of certain criteria such as environmental deterioration (or
improvement), economic decline (or growth), or social impoverishment; or, more generally,
against the criterion of sustainability. The results of these evaluations may be used to suggest
landuse alternatives that would contribute to the attainment of these goals.
1.3. Landuse Change: Bio-Physical and Socio-Economic Drivers
The analysis of landuse change revolves around two central and interrelated questions: "what
drives/causes landuse change" and "what are the (environmental and socio-economic) impacts of
landuse change".
The precise meaning of the "drivers" or "determinants" or "driving forces" of landuse change is
not always clear, commonly accepted and understood by all those who engage in studies of
landuse change. It is almost unanimously accepted that there are two main categories: biophysical and socio-economic drivers. The bio-physical drivers include characteristics and
processes of the natural environment such as: weather and climate variations, landform,
topography, and geomorphic processes, volcanic eruptions, plant succession, soil types and
processes, drainage patterns, availability of natural resources. The socio-economic drivers
comprise demographic, social, economic, political and institutional factors and processes such as
population and population change, industrial structure and change, technology and technological
change, the family, the market, various public sector bodies and the related policies and rules,
values, community organization and norms, property regime. It should be noted that the biophysical drivers usually do not cause landuse change directly. Mostly, they do cause land-cover
change which, in turn, may influence the landuse decisions of land owners/managers (e.g. no
farming on marginal lands). In addition, landuse changes may result in land cover changes
which, then, feedback on landuse decisions causing perhaps new rounds of landuse change (or
changes).
1.4. Landuse and Land Cover Classification Systems
The analysis of landuse change depends critically on the chosen system of landuse and land
cover classification. The magnitude and quality of landuse change is expressed in terms of
9
specific landuse or landuse/cover types. The assessment of the environmental and socioeconomic impacts of landuse change is possible only when the particular environmental and
socio-economic features of the chosen landuse/cover types are specified. If this requirement is
not met, then, the analysis will be of limited value in guiding policy and decision making
especially at lower scales. Hence, the need to discuss available landuse and land cover
classification systems and consider their suitability for the analysis of landuse change at various
spatial and temporal levels. In the following we discuss some existing landuse classification
systems in Europe.
10
2. Landuse classification systems
2.1. The CORINE land cover project of 1990
Land cover and landuse
The distinction between land cover and landuse is fundamental, and the two are often confused.
They are defined as follows:

Land cover is the observed physical cover, as seen from the ground or through remote
sensing, including natural or planted vegetation and human constructions (buildings,
roads, etc.) which cover the earth's surface. Water, ice, bare rock or sand surfaces count
as land cover

Landuse is based upon function, the purpose for which the land is being used
A landuse is defined as a series of activities undertaken to produce one or more goods or
services. A given landuse may take place on one or several pieces of land, and several landuses
may occur on the same piece of land. Defining landuse in this way provides a basis for precise
and quantitative economic and environmental impact analysis, and permits precise distinctions
between landuses if required.
Corine stands for Coordination of Information on the Environment. The EU established Corine
in 1985 to create pan-European databases on land cover, biotopes (habitats), soil maps and acid
rain.
Corine Land Cover (CLC) is a map of the European environmental landscape based on
interpretation of satellite images. It provides comparable digital maps of land cover for each
country for much of Europe. This is useful for environmental analysis and for policy makers.
The CLC1990 project was undertaken as a cross-border initiative by the Ordnance Survey of
Ireland and the Ordnance Survey of Northern Ireland. The aim was to produce a land cover map
for the entire island of Ireland.
The land cover database was based on the interpretation of satellite images for 1989 and 1990,
with land cover types in 44 standard classes. The map was created in GIS ARC/INFO format, at
an original scale of 1:100,000, which was consistent and comparable with similar land cover
databases in other European countries.
The CLC2000 database was created by first assessing and correcting the existing CLC1990 land
cover database and images for geometric and thematic content, then land cover changes were
mapped using 2000 satellite imagery and ancillary data.
A plot of the CORINE landuse map covering the Black Sea Basin and the associated database
indicating the % area of each landuse is provided in Figure 1 and Table 1, respectively. The
landuse classes were reclassified according to the SWAT (Soil and Water Assessment Tool)
landuse database.
11
Figure 1. The CORINE
landuse map
Table 1. % Area of different landuses/covers in the Corine database. The landuse/cover classes
have been matched with that of the SWAT landuse database
AGRR
Agricultural Land-Row Crops
12.77
BSVG
BAREN OR SPARSLY VEGETATED
0.36
CRDY
DRYLAND CROPLAND AND PASTURE
1.19
CRGR
CROPLAND/GRASSLAND MOSAIC
11.99
CRIR
IRRIGATED CROPLAND AND PASTURE
30.46
CRWO
CROPLAND/WOODLAND MOSAIC
13.57
FODB
DECIDUOUS BROADLEAF FOREST
9.68
FODN
DECIDUOUS NEEDLELEAF FOREST
0.02
FOEN
EVERGREEN BROADLEAF FOREST
3.44
FOMI
MIXED FOREST
3.64
GRAS
GRASSLAND
4.37
MIGS
MIXED GRASSLAND/SHRUBLAND
0.00
MIXC
MIXED DRYLAND/IRRIGATED CROPL
0.25
PAST
Pasture
2.64
RICE
Rice
0.01
SAVA
SAVANNA
0.81
12
SHRB
SHRUBLAND
1.36
TUBG
BARE GROUND TUNDRA
0.00
TUHB
HERBACEOUS TUNDRA
0.20
TUWO
WOODED TUNDRA
0.00
UIDU
Industrial
0.19
URLD
Residential-Low Density
1.41
URML
Residential-Med/Low Density
0.01
UTRN
Transportation
0.02
WATB
WATER BODIES
0.96
WEHB
HERBACEOUS WETLAND
0.65
WEWO
WOODED WETLAND
0.01
2.2. The GlobeCorine landcover
The GlobCorine 2005 land cover map has been generated over the period between December
2004 and June 2006, covering a pan-European area. The GlobCorine project, which was initiated
by the European Space Agency (ESA), focused on the production of a land cover map dedicated
to the pan-European continent and driven by the European Environmental Agency (EEA)
recommendations and needs. The GlobCorine project aims to address this issue by making the
full use of the potential of the ENVISAT’s Medium Resolution Imaging Spectrometer
Instrument (MERIS) Full Resolution Full Swath (FRS) time series and by further developing the
GlobCover classification approach.
The GlobCover classification module has to be adjusted to produce from the 300-m MERIS
dataset a land cover product as compatible as possible with the Corine Land Cover (CLC)
aggregated typology which is more land use oriented than the GlobCover legend.
The GlobCorine 2009 land cover map has been generated over the period spanning the entire.
year 2009 (1st January to 31st December). The product covers the European continent. extended
to the Mediterranean basin, as shown in Figure 2.
It is derived from an automatic and regionally-tuned classification of a time series of MERIS
seasonal and annual mosaics. The product nomenclature is as compatible as possible with the
CLC aggregated typology, while presenting an LCCS-based structure. The product is available in
the GeoTIFF format and stored in a zip archive named “GlobCorine_LC_200901_200912.zip”
enriched with additional files.
13
Figure 2. Pan-European extent of the GlobCorine 2009 land cover map
The GlobCorine 2009 land cover map has used on ENVISAT’s Medium Resolution Imaging
Spectrometer (MERIS) Level 1B data acquired in the Full Resolution mode with a spatial
resolution of 300 meters. For the generation of the Level 1B data, the raw data acquisitions have
been resampled on a path-oriented grid, with pixel values having been calibrated to match the
Top Of Atmosphere (TOA) radiance. The GlobCorine 2009 project is based on 12 months of
MERIS Fine Resolution Full Swath (FRS) data, from the 1st January 2009 until the 31st
December 2009. Further information about the ENVISAT MERIS Mission is available at the
MERIS
home
page
ENVISAT
MERIS
Mission
(http://envisat.esa.int/object/index.cfm?fobjectid=1665). Figure 3 shows the GlobeCorine
landuse/cover maps of 2005 and 2009. The area distributions of different classes are given in
Table 2. The classes are based on the classes of SWAT landuse database.
a
b
14
Figure 3. a) GlobeCorine 2005 and b) Corine 2009 landuse/landcover map
Table 2. % Area of different landuses/covers in the GlobeCorine database. The landuse/cover
classes have been matched with that of the SWAT landuse database
Landuse
Code
Landuse Description
% Area GCorine2005
%Area GCorine2009
URBN
Residential
0.84
1.07
AGRF
Rainfed cropland
20.78
25.21
AGIR
Irrigated cropland
0.03
0.05
FRST
Forest
23.82
21.12
RNG2
Range-Brush
0.35
0.19
BSVG
Baren or sparsly vegetated
2.48
1.93
WETN
Wetlands-Non-Forested
0.48
0.54
BARE
Bare areas
0.21
0.28
AGMX
Complex cropland
44.01
38.83
RYER
Russian Wildrye
5.65
9.50
WATR
Water
1.32
1.26
ICES
SNOW OR ICE
0.04
0.00
2.3. The MODIS landcover
The MODIS land cover product is designed to support scientific investigations that require
information related to the current state and seasonal-to-decadal scale dynamics in global land
cover properties. The product consists of two suites of science datasets. MODIS Land Cover
Type (MCD12Q1; Friedl et al., 2002) includes five main layers in which land cover is mapped
using different classification systems (the MLCT product). The MODIS Land Cover Dynamics
product (MCD12Q2; Zhang et al., 2006) includes seven layers, and has been developed to
support studies of seasonal phenology and interannual variation in land surface and ecosystem
properties. Various products in the Collection 5 land cover dynamics product is summarized in
Table 3. Here we discuss the land cover type product only.
15
The MLCT product consists of five different land cover classifications (Table 4). These layers
include the 17-class International Geosphere–Biosphere Programme classification (IGBP;
Loveland & Belward, 1997); the 14-class University of Maryland classification (UMD; Hansen
et al., 2000); a 10-class system used by the MODIS LAI/FPAR algorithm (Lotsch et al., 2003;
Myneni et al., 2002); an 8-Biome classification proposed by Running et al. (1995); and a 12Class plant functional type classification described by Bonan et al. (2002a). In addition to these
classification layers, the MLCT product provides the most likely alternative IGBP class and a
continuous measure of “classification confidence” at each pixel (McIver & Friedl, 2001). A
lower spatial resolution climate-modeling grid (MCD12C1) is produced at 0.05° spatial
resolution for users who do not require the spatial detail afforded by the 500-m land cover
product. The MCD12C1 product provides the dominant land cover type as well as the sub-grid
frequency distribution of land cover classes within each 0.05° cell.
In Figure 4 three MODIS soil maps are presented for the years 2001, 2006, and 2009. The classifications
based on the SWAT landuse database is shown in Table 5.
a
b
c
Figure 4. Modis landuse maps.
a) Modis 2001, b) Modis 2006,
c) Modis 2008
16
Table 3. Summary of different MODIS products.
Satellite
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Terra
Aqua
and
Collection
Date
5
January 04
2010
5
August
2009
12
5
August
2009
04
5
January 06
2006
5
March
2009
Products
04
Sites
Events
MCD 12Q2
Global
Added MCD 12Q2 subsets to global tool
orders. Anew time series plot that shows
the vegetation phenology in the subset area
was also added.
All products
1147
Fixed
subsets
Added stacked time series plot to time
series visualization. Example plot NDVI time
series stack for walker branch site. Stacked
Time Series
All
products
(Except
MCD
12Q2 and MCD
43A)
Global Web
Service
All products
Global
All products
Global
Moving MODIS Web Service to production.
The subsetting Web Service allows users to
create subsets up to 201 × 201 sq.km for
any
location
on
Earth.
http://daac.ornl.gov/modiswebservice
Major update to the global tool. Update
includes addition of Geo TIFF subsets as
default to all global tool orders. The Geo
TIFF subsets can also be reprojected to
Geographic coordinate system.
Added MODIS Land Cover subsets to the
display.
Added Land Cover (MCD 12Q1) histogram
visualization to the fixed site tool. Example
visualization for Walker branch site. Grid
and Histogram plot
5
February
19 2009
MCD 12Q1
1147
Fixed
subsets
5
February
03 2009
MCD 12Q1
1147
Fixed
subsets
Added the Land Cover (MCD 12Q1) subsets
to the fixed sites tool.
5
January 23
2009
All
products
(Except
MCD
12Q1, MCD 12Q2
and MCD 43A)
Global Web
Service
ORNL DAAC released a beta version of
MODIS fixed sites subsetting Web Service.
The subsetting Web Service allows users to
create subsets up to 201 × 201 sq.km for
any
location
on
Earth.
http://daac.ornl.gov/modiswebservice
5
December
14 2008
Global
Capability to create stackable time series
was added to the MODIS global tool.
5
June
2008
13
5
March
2008
10
All
products
(Except
MCD
12Q1 and MCD
12Q2)
All products
All
products
(Except
MCD
12Q1 and MCD
12Q2)
1147
Fixed
subsets
Global
ORNL DAAC added 95 new sites to its
MODIS fixed sites subset visualization tool.
These subsets include sites from NSIDC’s
subset request for GC-Net and IASAO
stations. Many new Flux tower locations are
also in this site list
ORNL DAAC released the MODIS Collection
5 Global subsetting tool. This tool allows
users to create subsets up to 201 × 201
sq.km for any location on Earth.
http://daac.ornl.gov/modisglobal
17
Terra
Aqua
and
March
2008
5
10
All
products
(Except
MCD
12Q1 and MCD
12Q2)
1147
Fixed
subsets
ORNL DAAC released the MODIS Collection
5 subsets for fixed sites. This tool provides
users subsets of more than 1000 sites in
ASCII and Geo TIFF data format. m
http://daac.ornl.gov/modisfixedsites
Table 4. Classifications of MOD12Q1 product.
Forests
Woodlands
Grasses/cereals
IGP
UMD
LAI/FPAR
BGC
PFT
Evergreen
needleleaf forest
(1)
Evergreen
needleleaf forest
Evergreen
needleleaf
forests
Evergreen
needleleaf
vegetation
Evergreen
needleleaf tree
Deciduous
needleleaf forest
(2)
Deciduous
needleleaf forest
Deciduous
needleleaf
forests
Deciduous
needleleaf
vegetation
Deciduous
needleleaf tree
Evergreen
broadleleaf forest
(3)
Evergreen
broadleleaf forest
Evergreen
broadleleaf
forests
Evergreen
broadleleaf
vegetation
Evergreen
broadleleaf tree
Deciduous
broadleleaf forest
(4)
Deciduous
broadleleaf forest
Deciduous
broadleleaf
forests
Deciduous
broadleleaf
vegetation
Deciduous
broadleleaf tree
Mixed forests (5)
Mixed forests
Woody savannas
(8)
Woody savannas
Savannas (9)
Savannas
Grasslands (10)
Grasslands
Grasses/cere
al crops
Annual grass
vegetation
Grass
Closed shrublands
(6)
Closed shrublands
shrublands
Shrub
Open shrublands
(7)
Open shrublands
shrublands
Shrub
Croplands (12)
Croplands
Broadleaf
crops
Cereal crop
Savannas
Shrublands
Croplands and
mosaics
18
Cropland/ natural
vegetation
mosaics (14)
Seasonally or
permanently
inundated
Unvegetated
Brodleaf crop
Permanent
wetlands (11)
Urban and buildup land (13)
Urban and buildup land
Urban
Urban
Urban
Barren or sparsely
vegetated (16)
Barren or sparsely
vegetated
Unvegetated
Unvegetated
Barren
sparsely
vegetated
Water (17)
Water
Water
Water
Water
Permanent snow
or ice (15)
or
Snow and ice
Table 5. % Area of different landuses/covers in the Modis database. The landuse/cover classes
have been matched with that of the SWAT landuse database
Landuse
Code
Landuse Description
%Area
Modis2001
%Area
Modis2006
%Area Modis
2009
AGRL
Agricultural Land-Generic
68.00
68.17
66.88
BSVG
BAREN OR SPARSLY
VEGETATED
0.07
0.04
0.04
FRSD
Forest-Deciduous
6.07
6.19
6.44
FRSE
Forest-Evergreen
0.00
0.00
0.00
FRST
Forest-Mixed
9.93
10.56
10.91
ICES
Ice and snow
0.11
0.11
0.09
OAK
Oak
2.00
1.53
1.68
PINE
Pine
2.31
2.72
2.95
RNG1
Range-Brush, modified
for cold temperature and
leaf area Index
0.68
0.24
0.30
19
RNG2
Range-Brush, modified
for cold temperature
0.99
0.84
0.65
RYER
Russian Wildrye
7.40
7.08
7.54
URBN
Residential
1.66
1.66
1.66
WATB
Water bodies
0.75
0.75
0.71
WETL
Wetlands-Mixed
0.04
0.10
0.13
3 Models of landuse change
3.1. Introduction
Models can be considered as the formal representation of some theory of a system of interest or
more broadly, models can be considered as abstractions, approximations of reality which is
achieved though simplification of complex real world relations to the point that they are
understandable and analytically manageable. Models which treat land and landuse change
explicitly are basically those in which the direct object of model building is landuse change. In
these models, land (and landuse) is conceptualized, at a minimum, as "a delineable area of the
earth’s terrestrial surface". Landuse is characterized by: (a) its areal (stock) and not point
character, (b) its relative immobility, (c) the relative stability of its occupancy, (d) the relatively
high cost of change from one type to another. Hence, models in which land is reduced to a point
in space are not considered landuse models.
As it is the case with all "spatial" models, landuse models employ some type of zonal system for
spatial representation. Each zone is characterized by its particular distribution of landuse types.
The number of zones, however, should be greater than a minimum value to consider the spatial
representation offered by the models as satisfactory. The recent trend is, however, towards
individual land unit-level models which make the use of a zoning system redundant.
3.2. Models of landuse change – Classification
The literature contains a considerable number and variety of models of landuse change where
landuse and its change are treated explicitly and are the direct object of the modeling exercise.
Eight interrelated sources of variation, in a roughly decreasing order of importance, can be
discerned: the purpose of the model, the theory underlying the model, the spatial scale and level
of spatial aggregation adopted as well as the degree of spatial explicitness of the model, the types
20
of landuse considered as principal objects of analysis, the types of landuse change processes
considered, the treatment of the temporal dimension, and the solution techniques used.
There exist the following types of models:
a- Descriptive models report changes in landuse and attempt to predict the factors that are
responsible for the changes. These models are usually applied to large areas where it is difficult
to obtain the data needed to calibrate other models (Mulley & Unruh, 2005; and Jianchu et al.
2005).
b- Stochastic models of changes in landuse consist of probabilistic transition models between
predefined states of the system (Thornton and Jones,1998).
c- Statistical models attempt to identify the factors causing changes in landuse through
multivariate analyses that highlight the exogenous factors of the observed changes (Joshi et al.
2006; Heistermann et al., 2006).
d- Simulation models highlight the interactions between all of elements that comprise the
environmental system. These approaches condense and aggregate complex ecosystems into a
small set equations (Dietzel & Clarke, 2006; Soares-Filho, 2002).
e- Economic landuse models assume that land demand, as influenced by the system of
preferences, motivations, markets, accessibility, and population, is the main determinant of
landuse. These approaches include both micro and macro models. Micro models attempt to
explain landuse changes at the farm level, using linear and non-linear mathematical
programming models (Porteiro et al., 2004).
f- Macro models use partial (Adams et al., 2005) or general equilibrium mathematical models
(Burniaux, 2002). Nevertheless, they have some difficulty in adapting to the spatially
disaggregated schemes that are used to estimate landuse evolution.
g- Integrated spatial models combine the advantages of simulation spatial models with the
qualities of spatial economic models (Verburg et al. 2006; Abildtrup et al. 2006).
h- Finally, spatial interaction models integrate the geographical approach implicit in simulation
models with the consistency of the methodology present in the gravity models of spatial
interaction. In particular, this allows for integration of the consistent interpretations usually
present in economic spatial models.
The above models are further classified in the following four main categories of models:
a. statistical and econometric models
21
b. spatial interaction models
c. optimization models, and
d. integrated models.
3.3 Statistical Models
In a statistical model of landuse change, the study area is usually subdivided into a number of
zones (or, grid cells if a grid system is adopted) the size and shape of each cell depending on the
level of aggregation chosen as well as the availability of data. In the continuous case, for each
zone, the distribution of landuse types (the dependent variables ) as well as the values of other
environmental and socio-economic predictor variables (e.g. population, employment, soil
conditions, slope, climate (temperature, rainfall, etc.) are given. A multiple regression equation
for each landuse type is fit to these data (usually referring to a given year). The general form of
the equation is:
LUTi  a  1 X1   2 X 2  ..........   n X n  
(1)
where LUT is the area of land occupied by landuse type i (in each cell) and X1, X2, … Xn the
predictor variables. The term  is the error term of the statistical model.
This model form can be used to assess the changes in the area covered by a given landuse type
for specified changes in one or more of the predictor variables by substituting their values in Eq.
1 above.
A similar statistical model is used in the CHANGE module of the CLUE model which is
discussed below. The CHANGE module uses linear regression models to estimate the changes in
the area of given landuse types that are caused by changes in the values of environmental and
socio-economic driving factors projected from other modules of the CLUE model.
Discrete statistical models (or, discrete choice models) are used to represent choice situations in
general. In the case of landuse modeling, each landuse type is described as a function of a
number of characteristics, which usually differ from one cell to another. For each cell, the utility
of every landuse type is assessed as a function of these characteristics. The probability of
choosing a particular landuse type in a given cell is calculated as a function of the utilities
associated with the landuse types considered. The most common mathematical forms used in
discrete choice models are the logit and probit models.
In the context of a larger modeling exercise for the analysis of landuse change in Japan,
Kitamura et al. (1997) and Morita et al. (1997) use a multinomial logit model to assess changes
in landuse by type. The model assesses the probability of choice of a particular landuse type in
each of the cells in which the study area is subdivided as a function of the values of a set of
22
predictor/ explanatory variables . These probabilities are interpreted as landuse proportions for
each of a specified number of landuse types. The mathematical form of the model is as follows:
Pij 
exp( Vij )
 exp( Vij )
(2)
i
where
Vij   ik X jk  Ci
(3)
k
where Pij is the landuse proportion of landuse type i in cell j, Vij is the utility of the ith landuse
type in cell j, Xik the kth explanatory variable in cell j, and  ik is the multiple regression
coefficients of the explanatory variables Xjk.
The above formulation calculates first the utility of each landuse type in each cell of the study
area as a linear function of the values of a set of predictor variables and then uses this utility to
estimate the probability of a particular landuse type occurring in each cell. As it was the case
with the previous multiple regression model, changes in the predictor variables calculated from
other modules of the larger model are fed into equation (3) to estimate changes in the utility of
each landuse type. These changes are then used in equation (2) to estimate changes in the
proportion of each landuse type in each cell of the study region.
3.4. Spatial interaction models or gravity models
The spatial interaction modeling tradition draws from the original efforts to model interaction of
human activities in space based on the analogy of the Law of Gravity in Physics. Hence, the
models included in this group are the well known gravity-type models and their newer versions
known more generally as spatial interaction models.
Gravity models of spatial interaction are built to describe and predict the flow of people, goods,
and information across space (Smith, 1995). Applications of gravity models to analyze spatial
interactions have long existed in the literature (e.g., Carey 1858; Reilly 1931; Schneider 1959).
These studies have provided analytical tools that are commonly used in land planning,
geographical study, and regional science (Wilson 1967); demography (Plane 1984); and
commerce and marketing (Deardorff 1998). A comprehensive review of operational gravity
models of spatial interactions applied to urban regions is made by Wegener (1994) and a good
presentation of the evolution of the theoretical bases of these models is undertaken both by Roy
and Thill (2004) and with a larger scope on various economic fields by Roy (2004).
23
The main theoretical question regarding the use of gravity models for spatial interaction arose
from the process of model creation. Theoretical questions regarding these models attempt to
identify minimal behavioral hypotheses that justify a pre-defined intuitive and powerful model.
Gravity models used for spatial interaction perform very well in explaining the spatial interaction
behaviors of large populations. Nevertheless, they perform very poorly at explaining the
behavior of individuals, an attribute due to the lack of information about individual spatial
behavior, rather than a fault of the features of the model.
The gravity model assumes a study region subdivided into a number of zones which are called
origin and destination zones. Origin zones are characterized by activities from which flows
originate (e.g. residential areas where employees live) to reach destination zones (e.g.
employment areas where the employees work). Each zone of the system can be both an origin
and a destination zone. The simplest form of the gravity model which parallels the form of the
corresponding model in Physics is the following:
S ij  k
Pi Pj
d ijb
(4)
Sij denotes interaction (flow) from origin zone i to destination zone j, Pi is the "size" or "mass" of
origin zone I, Pj is the "size" or "mass" of destination zone j, dij is a measure of distance between
zones i and j, b an exponent indicating the effect of distance on the interaction between origin
and destination zones, k is a constant which is empirically determined and adjusts the
relationship to actual conditions
The above formula states that the magnitude of the interaction between zone i and zone j, Sij, is
proportional to the product of the "sizes" or "masses" of the origin and the destination zones and
inversely proportional to a measure of the distance between them. Measures of the "interaction"
term include number of trips between zones, volume of goods transported between zones,
migration flows, etc. The "sizes" or "masses" of the origin and the destination zones are
operationalized variously depending on the application. In the more common applications of the
model – in retail and residential location problems – the "size" of the origin zones is expressed
by the population of these areas or the income of the population (a proxy of their purchasing
power). The "size" of the destination zones is expressed as retail floorspace or revenues of retail
stores or number of employees.
The denominator of the formula contains the critical expression of the effect of distance on the
interaction between origin and destination zones. This is variously known as "friction of space",
"impedance effect of distance", "friction against movement", and so on. The literature contains
an extensive discussion of the distance function as regards: (a) alternative ways to operationalize
the concept of distance in other than metric units – such as in terms of cost, time spent on
commuting between origin and destination zones, multidimensional measures combining time,
money, and effort spent in commuting between zones, (b) the values of the exponent of the
distance function, known also as the "distance decay parameter" – which varies with the purpose
of the interaction (e.g. trip purpose) as well as with distance itself and (c) the use of other
24
functional forms of the distance function instead of the one shown above. As to the latter issue,
Wilson (1967) suggested a negative exponential function e t which reflects the fact that the
exponent (i.e. the magnitude of the effect of distance) varies with distance (Fotheringham 1984).
An alternative, simple form of the model shown in equation (4) is the following:
Sij  kOi D j f ( d ij )
(5)
where, Oi corresponds to Pi above (O standing for Origins), Dj corresponds to Pj above (D
standing for Destinations), f(dij) a general symbol for the distance function. Sometimes the origin
and destination terms are raised to some power to reflect the difference in importance of the
"masses" of origins and destinations. Oi can be considered as the total "production" of interaction
flows out of zone i and Dj the "attraction" of flows by zone j. The above, classical form of the
gravity model does not ensure that the aggregate flows modeled will sum to the total flows
observed in the study region. This is called the additivity condition and it can be expressed
mathematically as:
M
 ( S ij  Oi )
j = 1....M
(6)
i = 1....N
(7)
j 1
N
 ( S ij  Di )
j 1
Drawing on the above, a form of the gravity model which satisfies the additivity condition for
the flows of both the origin and the destination zones is the following:
Sij  Ai B j Oi D j f ( d ij )
(8)
where


M

Ai    B j D j f ( d ij ) 


 j 1





N

B j    Ai Oi f ( d ij ) 
i 1

1
(9)
1
(10)
Based on equation (8), four alternative forms of the gravity formulation can be distinguished
depending on whether information on the interaction sums Oi and/or Dj is available. When either
one or both are not known, Oi and Dj are replaced by "attractiveness" terms Wi and Wj
respectively. The attractiveness terms Wi and Wj can be operationalized in various ways.
Common measures for Wi is the amount of housing available in an origin zone (perhaps of a
25
given quality) and for Wj the number of jobs in destination zones. The four forms of the gravity
model are:
(a) unconstrained – neither Oi nor Dj are given. In this case the model takes the form of equation
(5) where Wi replaces Oi and Wj replaces Dj as follows:
Sij  kWiW j f ( d ij )
(11)
(b) production-constrained – Oi is given but not Dj. In this case the model takes the form:
Sij  Ai OiW j f ( d ij )
(12)
where,
Ai 
1
M
(13)
 W j f ( d ij )
j 1
(c) attraction-constrained – Dj is given but not Oi. In this case the model takes the form:
Sij  B jWi D j f ( dij )
(14)
where,
Bj 
1
N
(15)
Wi f ( dij )
i 1
(d) production-attraction-constrained (or, doubly-constrained) when both Oi and Dj are known. In
this case the model takes the form of equation (8) and Ai and Bj are given by expressions (9) and
(10).
Gravity models are spatially explicit, the degree of spatial representation they offer depending on
the number of zones into which the study region is subdivided. There has been considerable
debate about the proper number and shape of zones and the effects of the zoning system used on
the results of the model. The models are static or quasi-static at best, which means that they do
not account for the dynamics which underlies the observed interactions. In terms of level of
detail of the landuses considered as well as of the spatial behavior modeled, the most common
forms of gravity models concern two main types of landuse – e.g. residential and commercial,
residential and employment, residential and recreation. However, to make the gravity model
more sensitive to the real world variability of human behavior, an important stream of research
effort has been devoted to producing disaggregate versions of the models (depending on the
availability of data). For example, residential (origin) areas are disaggregated by income group
or types/prices of housing; employment (destination) areas are disaggregated by different wage
26
levels and types of products; and, interaction has been disaggregated by various modes of
transport, trip purposes, and stages (Batten and Boyce 1986).
3.5. Integrated models
Integrated models are those models which consider in some way the interactions, relationships,
and linkages between two or more components of a spatial system – be they sectors of economic
activity, regions, society and economy, environment and economy, and so on – and relate them
to landuse and its changes either directly or indirectly. A common characteristic of integrated
models, in addition to their emphasis on integration, is that they are mostly large-scale models.
The range of spatial levels covered starts from the urban/metropolitan and reaches the global.
The spatial coverage of integrated models is closely related to their purpose, focus, and other
structural and design characteristics. The meaning of integration varies with the model purpose
and is reflected in the structure of the integrated model. Five dimensions of integration can be
distinguished broadly:
a. spatial integration – where the horizontal and/or vertical interactions among spatial
levels are emphasized with respect to a phenomenon being modeled
b. sectoral integration – where the model represents the linkages and relationships between
two or more economic sectors of the spatial system of interest such as retail, housing,
transportation, industry, agriculture, etc.
c. landuse integration – in which the model accounts for the interactions between more than
two types of landuse such as residential, commercial, manufacturing, transportation, etc.;
this dimension of integration may be equivalent at times with the sectoral integration
d. economy-society-environment integration – where the model represents the linkages
between at least two of the several components of the spatial system such as economyenvironment, economy-society (e.g. population), economy-energy, etc.
e. sub-markets integration – where models show how different sub-markets of the whole
economy relate to one another; a related type of integration may be considered that
between supply and demand. In this latter case, the related economic models are
distinguished into partial equilibrium (referring to either demand or supply) and general
equilibrium models.
Four integrated models that can be classified as regional level simulation models are presented
below which address the analysis of landuse change directly:

The CLUE modeling framework

The cellular automata modeling framework

IIASA’s LUC (Landuse Change) model, and
27

The IMPEL model
Of these, we will briefly introduce the CLUE and the Cellular automata modeling framework.
The CLUE modeling framework
The CLUE (Conversion of Landuse and its Effects) modeling framework was developed at the
Wageningen Agricultural University in the Netherlands to model landuse changes as a function
of their driving factors (de Koning et al. 1998, Verburg et al. 1999). It has been applied to
analyze landuse/cover changes in several countries such as Ecuador, Costa Rica, Java, and
China. A basic outline of this framework follows based on the several publications of the CLUE
research group. The following section is adopted from Verbrug (2010), which can be
downloaded from:
http://www.ivm.vu.nl/en/Organisation/departments/spatial-analysis-decisionsupport/Clue/index.asp
The CLUE modeling framework is a spatially explicit modeling framework for the analysis of
landuse/cover dynamics at various spatial scales. Its most recent versions incorporate also
dynamic analysis of feedbacks of landuse changes on the local environment, the population, etc.
as it is the case, for example, of agricultural over-use or unsuitable use in sensitive areas. In other
words, the CLUE framework can be described as an integrated, spatially explicit, multi-scale,
dynamic, economy-environment-society-landuse model. The modeling approach of CLUE has
been modified and is now called CLUE-S (the Conversion of Landuse and its Effects at Small
regional extent). CLUE-S is specifically developed for the spatially explicit simulation of
landuse change based on an empirical analysis of location suitability combined with the dynamic
simulation of competition and interactions between the spatial and temporal dynamics of landuse
systems. More information on the development of the CLUE-S model can be found in Verburg et
al. (2002) and Verburg and Veldkamp (2004). The more recent versions of the CLUE model:
Dyna-CLUE (Verburg and Overmars, 2009) and CLUE-Scanner include new methodological
advances.
The model is sub-divided into two distinct modules, namely a non-spatial demand module and a
spatially explicit allocation procedure (Figure 5).
28
Figure 5. Overview of the modeling procedure
The Demand Module calculates the demand for various types of landuses based on the national
level demand for various commodities. National level demand consists of domestic consumption
and exports. Exports are assessed exogenously and they are related to international prices and
national subsidies. Domestic consumption is assessed as a function of population size,
composition (urban and rural) and consumption patterns. The Population Module provides the
necessary demographic input to the Demand Module. Consumption patterns may be related to
macro-economic indicators like GNP, purchasing power and price levels. Demand functions for
separate commodities are estimated based on historical data. The user-interface of the CLUE-S
model only supports the spatial allocation of landuse change. To account for difficult-to-predict
changes in demand, alternative scenarios are formulated which take into account various
population projections and changes in diet patterns. The production volumes demanded for the
separate commodities are translated into areas of the corresponding landuse/cover types using
crop specific yield coefficients. The areas calculated for separate crops are aggregated to broader
landuse/cover types to obtain the demand for land at the level of these aggregate types.
The Allocation Module is based upon a combination of empirical, spatial analysis and dynamic
modeling. Figure 2 gives an overview of the information needed to run the CLUE-S model. This
information is subdivided into four categories that together create a set of conditions and
possibilities for which the model calculates the best solution in an iterative procedure. The next
sections discuss each of the boxes: spatial policies and restrictions, landuse type specific
conversion settings, landuse requirements (demand) and location characteristics.
29
Figure 6. Overview of the information flow in te CLUE-S model
Landuse type specific conversion settings
Landuse type specific conversion settings determine the temporal dynamics of the simulations.
Two sets of parameters are needed to characterize the individual landuse types: conversion
elasticities and landuse transition sequences. The first parameter set, the conversion elasticities,
is related to the reversibility of landuse change. Landuse types with high capital investment will
not easily be converted in other uses as long as there is sufficient demand. Examples are
residential locations but also plantations with permanent crops (e.g., fruit trees). Other landuse
types easily shift location when the location becomes more suitable for other landuse types.
Arable land often makes place for urban development while expansion of agricultural land
occurs at the forest frontier. An extreme example is shifting cultivation: for this landuse system
the same location is mostly not used for periods exceeding two seasons as a consequence of
nutrient depletion of the soil. These differences in behavior towards conversion can be
approximated by conversion costs. However, costs cannot represent all factors that influence the
decisions towards conversion such as nutrient depletion, esthetical values etc. Therefore, for each
landuse type a value needs to be specified that represents the relative elasticity to change,
ranging from 0 (easy conversion) to 1 (irreversible change). The user should decide on this factor
based on expert knowledge or observed behavior in the recent past.
30
The second set of landuse type characteristics that needs to be specified are the landuse type
specific conversion settings and their temporal characteristics. These settings are specified in a
conversion matrix. This matrix defines: To what other landuse types the present landuse type can
be converted or not (Figure 7). In which regions a specific conversion is allowed to occur and in
which regions it is not allowed.
How many years (or time steps) the landuse type at a location should remain the same before it
can change into another landuse type. This can be relevant in case of the re-growth of forest.
Open forest cannot change directly into closed forest. However, after a number of years it is
possible that an undisturbed open forest will change into closed forest because of re-growth.
The maximum number of years that a landuse type can remain the same. This setting is
particularly suitable for arable cropping within a shifting cultivation system. In these systems the
number of years a piece of land can be used is commonly limited due to soil nutrient depletion
and weed infestation.
It is important to note that only the minimum and maximum number of years before a conversion
can or should happen is indicated in the conversion table. The exact number of years depends on
the landuse pressure and location specific conditions. The simulation of these interactions
combined with the constraints set in the conversion matrix will determine the length of the
period before a conversion occurs. Figure 8 provides an example of the use of a conversion
matrix for a simplified situation with only three landuse types.
Figure 7. Illustration of the translation of a hypothetical landuse change sequence into a
landuse conversion matrix.
31
Figure 8. Example of a landuse concversion matrix with the different options implemented in the
model.
Landuse requirements (demand)
Landuse requirements (demand) are calculated at the aggregate level (the level of the case-study
as a whole) as part of a specific scenario. The landuse requirements constrain the simulation by
defining the totally required change in landuse. All changes in individual pixels should add up to
these requirements. In the approach, landuse requirements are calculated independently from the
CLUE-S model itself. The calculation of these landuse requirements is based on a range of
methods, depending on the case study and the scenario. The extrapolation of trends in landuse
change of the recent past into the near future is a common technique to calculate landuse
requirements. When necessary, these trends can be corrected for changes in population growth
and/or diminishing land resources. For policy analysis it is also possible to base landuse
requirements on advanced models of macro-economic changes, which can serve to provide
scenario conditions that relate policy targets to landuse change requirements.
Location characteristics
Landuse conversions are expected to take place at locations with the highest 'preference' for the
specific type of landuse at that moment in time. Preference represents the outcome of the
interaction between the different actors and decision making processes that have resulted in a
spatial landuse configuration. The preference of a location is empirically estimated from a set of
factors that are based on the different, disciplinary, understandings of the determinants of
landuse change. The preference is calculated following:
32
Rki  ak X 1i  bk X 2i  ...
(16)
where R is the preference to assign location i to landuse type k, X1, X2, ... are biophysical or
socio-economical characteristics of location i and ak and bk the relative impact of these
characteristics on the preference for landuse type k. The exact specification of the model should
be based on a thorough review of the processes important to the spatial allocation of landuse in
the studied region.
A statistical model can be developed as a binomial logit model of two choices: convert location i
into landuse type k or not. The preference Rki is assumed to be the underlying response of this
choice. However, the preference Rki cannot be observed or measured directly and has therefore to
be calculated has a probability. The function that relates these probabilities with the biophysical
and socio-economic location characteristics is defined in a logit model following:
 P
log  i
 1  Pi

   0  1 X 1,i   2 X 2 ,i ......   n X n ,i

(17)
where Pi is the probability of a grid cell for the occurrence of the considered landuse type on
location i and the X's are the location factors. The coefficients (β) are estimated through logistic
regression using the actual landuse pattern as dependent variable. This method is similar to
econometric analysis of landuse change, which is very common in deforestation studies. In
econometric studies the assumed behavior is profit maximization, which limits the location
characteristics to (agricultural) economic factors. In the study areas is assumed that locations are
devoted to the landuse type with the highest 'suitability'. 'Suitability' includes the monetary
profit, but can also include cultural and other factors that lead to deviations from (economic)
rational behaviour in land allocation. This assumption makes it possible to include a wide variety
of location characteristics or their proxies to estimate the logit function that defines the relative
probabilities for the different landuse types.
Most of these location characteristics relate to the location directly, such as soil characteristics
and altitude. However, land management decisions for a certain location are not always based on
location specific characteristics alone. Conditions at other levels, e.g., the household, community
or administrative level can influence the decisions as well. These factors are represented by
accessibility measures, indicating the position of the location relative to important regional
facilities, such as the market and by the use of spatially lagged variables. A spatially lagged
measure of the population density approximates the regionally population pressure for the
location instead of only representing the population living at the location itself.
Allocation procedure
When all input is provided the CLUE-S model calculates, with discrete time steps, the most
likely changes in landuse given the before described restrictions and suitabilities. The allocation
33
procedure is summarized in Figure 5. The following steps are taken to allocate the changes in
landuse:
The first step includes the determination of all grid cells that are allowed to change. Grid cells
that are either part of a protected area or presently under a landuse type that is not allowed to
change are excluded from further calculation. Also the locations where certain conversions are
not allowed due to the specification of the conversion matrix are identified.
For each grid cell i the total probability (TPROPi,u) is calculated for each of the landuse types u
according to:
TPROPi ,u  Pi ,u  ELAS u  ITERu
(18)
where Pi,u is the suitability of location i for landuse type u (based on the logit model), ELASu is
the conversion elasticity for landuse u and ITERu is an iteration variable that is specific to the
landuse type and indicative for the relative competitive strength of the landuse type. ELASu, the
landuse type specific elasticity to change value, is only added if grid-cell i is already under
landuse type u in the year considered.
A preliminary allocation is made with an equal value of the iteration variable (ITERu) for all
landuse types by allocating the landuse type with the highest total probability for the considered
grid cell. Conversions that are not allowed according to the conversion matrix are not allocated.
This allocation process will cause a certain number of grid cells to change landuse.
The total allocated area of each landuse is now compared to the landuse requirements (demand).
For landuse types where the allocated area is smaller than the demanded area the value of the
iteration variable is increased. For landuse types for which too much is allocated the value is
decreased. Through this procedure it is possible that the local suitability based on the location
factors is overruled by the iteration variable due to the differences in regional demand. The
procedure followed balances the bottom-up allocation based on location suitability and the topdown allocation based on regional demand.
Steps 2 to 4 are repeated as long as the demands are not correctly allocated. When allocation
equals demand the final map is saved and the calculations can continue for the next time step.
Some of the allocated changes are irreversible while others are dependent on the changes in
earlier time steps. Therefore, the simulations tend to result in complex, non-linear changes in
landuse pattern, characteristic for complex systems.
34
Figure 9. Flow chart of the allocation module of the CLUE-S model
Description of the allocation procedure of the new Dyna-CLUE 2 version
The model is an adapted version of the CLUE-s model (Castella and Verburg, 2007) which is
based on the spatial allocation of demands for different landuse types to individual grid cells.
The version implemented (Dynamic Conversion of Landuse and its Effects model: Dyna-CLUE
version 2.0) combines the top-down allocation of landuse change to grid cells with a bottom-up
determination of conversions for specific landuse transitions. The analysis starts by grouping the
landuse types into two groups: those that are driven by demand at the regional level and those for
which no aggregate demand at the regional level can be determined. In many applications, the
demands can be specified for urban and agricultural landuses (including production forest) while
no specific demand can be determined for the (semi-) natural land cover. The land cover types
for which no demand can be specified are grouped into one, new, category for which the
aggregate change in area results from the dynamics of the other landuse types, i.e., the overall
change in area of this new category corresponds to the net change in the demand-driven landuse
types (Figure 10).
The spatial allocation module allocates the regional level demands to individual grid cells until
the demand has been satisfied by iteratively comparing the allocated area of the landuse types
with the area demanded. Land cover types that are grouped in a new category are allocated
individually but only the sum of the allocated area of the grouped land cover types is compared
with the demand. The allocation procedure allocates at time (t) for each location (i) the
landuse/cover type (lu) with the highest total probability (Ptoti,t,lu). The total probability is
defined as the sum of the location suitability (Ploci,t,lu), neighborhood suitability (Pnbhi,t,lu),
conversion elasticity (elaslu) and competitive advantage (compt,lu) following:
Ptot i ,t ,lu  Ploc i ,t ,lu  Pnbhi ,t ,lu  elas u  compt ,lu
(19)
35
Figure 10. Overview of the Dyna-CLUE model
The conversion elasticity is a measure of the cost of conversion of one landuse type to another
landuse type and applied only to those locations where the landuse type is found at time t. High
values indicate high conversion cost (either monetary or institutional) and thus a higher total
probability for the location to remain under the current landuse type. Low values for Elaslu may
apply to annual crops, grassland and similar landuse types while high values apply to forest,
urban areas and permanent crops for which high costs of establishment have been made.
The competitive advantage is iteratively determined for all landuse types during an iterative
procedure. Values are increased during the iteration when allocated area is smaller than area
demanded while values are decreased when allocated area exceeds the demand. In the case of
increasing demand, the value of the competitive advantage is likely to increase while lower
values are obtained when the demand for a certain landuse type decreases. For the grouped
landuse types, only a value for the competitive advantage for the group as a whole is determined,
as demands are not specified for the individual landuse types within this group.
36
Location suitability and neighborhood suitability can be determined by either empirical methods
(Verburg et al., 2004a), process and expert knowledge (Overmars et al., 2007) and the (dynamic)
analysis of neighborhood interactions similar to constrained cellular automata models (Verburg
et al., 2004b). In case of (semi-)natural landuse types suitabilities are only defined when specific
location requirements are known and relevant. Otherwise a uniform suitability is assigned to all
locations.
The maximization of the total probability is checked against a set of conversion rules as specified
in a conversion matrix (Figure 7). This conversion matrix indicates which conversions are
possible for each landuse type, e.g., the conversion from agriculture to forest is not possible
during one (yearly) time step as a consequence of the time it takes to grow a forest. Conversions
that are excluded by the conversion matrix overrule the maximization of total probability.
Instead, the landuse type with the highest total probability for which the conversion is allowed
will be selected. In addition it is possible to specify that certain conversions are only possible
within delineated areas, such as outside nature reserves. In this case a reference to a map
indicating these zones is made in the conversion matrix. The dynamics of the landuse types
governed by local processes (‘bottom-up processes’ in Figure 1) are also specified in the
conversion matrix. Instead of restricting a specific conversion it is also possible to enforce a
conversion between landuse types. When a specific conversion is expected within a specific
number of years the conversion will be enforced as soon as the number of years is exceeded.
Figure 2 illustrates this for the conversion of shrubland to forest which takes place after a number
of years depending on the growth conditions at the location. Such locally determined conversions
are the result of specific management practices or vegetation dynamics. Due to the spatial
variation in local conditions, these time periods are represented in a map (Figure 11).
Locally determined conversions will, to some extent, interfere with the allocation of the other
landuse types that are driven by the regional demands due to changes in conversion elasticity
upon locally determined conversions, i.e., the conversion to agriculture is less difficult for
recently abandoned agricultural land than for shrubland. The resulting conversion trajectories
will cause intricate interactions between the spatial and temporal dynamics of the simulation.
The specification of the model for different landuse types, location suitability, conversion
elasticity, and conversion matrix is dependent on the specific case study area, spatial and
temporal scale and the purpose of the model. The following section illustrates the functioning of
the model by a specification of the model for the simulation of landuse for the 27 countries of the
European Union at a spatial resolution of 1 km2 for the time period 2000-2030.
37
Figure 11. Simplified land cover conversion matrix indicating the possible conversions during
one time step of the simulation.
Implementation of the Dyna-CLUE model for Europe
The application of the model for Europe includes 16 different landuse types. Although the
landuse types area derived from a land cover map, they also represent, to some extent, the use of
the land cover. Therefore, we refer to ‘landuse types’ in the following. The landuse types are
subdivided into 3 categories. The first category includes landuse types for which a demand is
calculated at the level of individual member states by a macro-economic, multi-sector model
accounting for global trade and agricultural policy (van Meijl et al., 2006) in combination with a
simple projection model for urbanization. The second category contains landuse types for which
the area is expected to be more or less constant in time due to the inability to use these lands for
agricultural or urban purposes, or strict protection to avoid conversion. The third category
contains landuse types the conversions of which are determined by local conditions, especially
the regeneration of natural vegetation. Landuse types in this group are recently abandoned arable
land, recently abandoned grassland, (semi-)natural vegetation and forest. The landuse types in
38
this category are grouped into one single group the area of which is a result of the dynamics of
the agricultural and urban landuse types. Agricultural decline will increase the area of this group
while agricultural expansion and urbanization will occur at the cost of this group. The protected
areas for nature conservation determine the minimum area allocated to these (semi-) natural
landuses. The conceptual transitions between the landuse types in this group are shown in Figure
12. Upon abandonment of agricultural land regeneration/succession of (semi-)natural vegetation
takes place depending on the local conditions that favor or retard the establishment and growth
of natural vegetation.
The subdivision of the regrowth of natural vegetation into three stages of succession is arbitrary
since succession is a continuous process. However, the three stages were chosen because of their
clear morphological and functional differences and frequent use in studies of succession on
abandoned farmland (Pueyo and Beguería, 2007). Occasional grazing on abandoned farmlands,
which is common practice in many parts of Europe, may retard the transition to shrubs and trees
(González-Martínez and Bravo, 2001; Tasser et al., 2007; Tzanopoulos et al., 2007). Also, in
densely populated areas alternative uses may occupy former farmland areas, e.g., hobby farming
and horse-boarding (Gellrich et al., 2008). In this case the landuse remains similar to agricultural
land but does not contribute to agricultural production. Therefore these areas are disregarded in
the demand calculations for agricultural land. Under these circumstances the classification of the
land will remain ‘recently abandoned agricultural area’. Besides the effects of grazing and
population pressure, the re-growth of shrub vegetation on recently abandoned land depends on
local growth conditions for vegetation including soil constraints (Tasser et al., 2007). Recently
abandoned agricultural land is subdivided into recently abandoned grassland and recently
abandoned arable land depending on the previous use. This subdivision is necessary because
succession on grassland takes, under similar conditions, longer due to the closer vegetation
structure that makes the establishment of new species including shrubs and trees more difficult
(Flinn and Vellend, 2005). Also the subsequent conversion of shrubland to forest depends on
local biophysical conditions (Kräuchi et al., 2000; Pueyo and Beguería, 2007). In dry or cold
climates or on very shallow soils the succession of shrubland to forest is extremely slow and may
not occur at all (del Barrio et al., 1997). In these locations shrubland is the climax vegetation
including typical vegetations such as Maquis, Garrigue and Macchia as found in southern
Europe, the Tundra of northern Europe and mountain areas above the treeline. Besides climatic
and soil conditions the time needed for succession into forest is also determined by the dispersal
of seeds (Pugnaire et al., 2006; Tasser et al., 2007) which is approximated by the presence of
forest in the neighborhood.
39
Figure 12. Schematization of the landuse/cover transitions upon abandonment of agricultural
land
All possible conversions indicated in Figure 12 are represented in the landuse conversion matrix
(Figure 11). The matrix indicates that certain conversions are not possible, e.g. the conversions
from agricultural land to shrubland and forest because upon agricultural abandonment the
landuse is first classified as recently abandoned land. Conversion of recently abandoned land into
shrubland is scheduled after a number of years indicated in a map depending on the local
conditions and the processes mentioned above (Figure 13). The parameterization of the time
between the different succession stages is based on a combination of expert rules and biophysical
data. The influence of climate and soil conditions is quantified by calculating an index that
combines potential evapotranspiration during the growing season and constraints based on the
water holding capacity of the soil available to plants, water deficit, temperature restrictions and
water logging occurrence. Spatial information for these variables is derived from the WorldClim
database (Hijmans et al., 2005), the Climate Research Units database (Mitchell et al., 2005) and
the European Soil Database (ESDB). This index is translated into succession periods by
calibration on an expert table of observed and reported succession speed in different
environmental and altitude zones across Europe. The expert table is based on observations of
forest re-growth on abandoned land and review of literature for various case studies (GonzálezMartínez and Bravo, 2001; MacDonald et al., 2000; Poyatos et al., 2003; Pugnaire et al., 2006;
Tasser et al., 2007). In the calibration, it was accounted for that the observed succession times
often correspond with plots that are marginal for agriculture, showing lower succession speed for
natural vegetation as compared to locations on prime agricultural land. This calibration resulted
40
in three maps indicating succession time for recently abandoned grassland to (semi-)natural
vegetation, recently abandoned arable land to (semi-)natural vegetation and for (semi-)natural
vegetation to forest (Figure 13). Based on current grazing intensities and population densities
Other model settings include the definition of the suitability of locations for agricultural and
urban landuse types, conversion elasticities and region-specific constraints representing spatial
policies and planning. Suitabilities where estimated by logit models using the spatial association
of current landuse with a wide range of biophysical and socio-economic variables to represent
location factors (Verburg et al., 2004; Verburg et al., 2006). Conversion elasticities were
estimated based on expert knowledge of the conversion costs for different landuses and spatial
restrictions included NATURA2000 nature reserves, erosion sensitive locations and ‘less
favoured areas’ following the spatial policies included in the scenario description (Westhoek et
al., 2006). More specific details on the configuration of the model are provided in Verburg et al.
(2008) and (WUR/MNP, 2008).
The application of the model to Europe has illustrated the application of the model in the context
of declining agricultural area and regeneration of natural vegetation. The combination of topdown and bottom-up processes in a consistent modeling framework may also be relevant in other
areas and for other processes. Examples of possible applications include the dynamics of tropical
forest landscapes, where large scale logging as result of global demand for timber and
agricultural commodities, interacts with local processes of soil degradation and regeneration of
secondary vegetation. Local processes causing soil degradation may prevent future use of these
soils and therefore need to be taken into account. In addition, the simulation of low-input
agricultural systems with fallow periods as part of the crop rotation may be captured by
combining an assessment of the overall demand for agricultural production with local processes
of soil fertility dynamics.
A step by step procedure for building a Dyna-CLUE model is given in the Appendix I.
41
Figure 13. Number of years needed for the transition of recently abandoned arable land into
(semi-) natural vegetation (A) and for the transition of (semi-) natural vegetation into forest (B).
42
The Cellular Automata Modeling Framework
Another integrated simulation modeling approach draws from the theoretical framework of
Social Physics more specifically from the theory of fractals to model the structure and evolution
of landuse patterns. It applies cellular automata (CA) concepts to model a variety of complex,
dynamic, socio-economic and environmental phenomena (see, for example, Engelen 1988, White
and Engelen 1993). The approach – henceforth called cellular automata approach – to be
presented below is considered to be "quite general in terms of the situations to which it can be
usefully applied". It has been shown to apply to both the urban and larger geographical scales for
the comprehensive, integrated analysis of landuse change. The following section is adopted from
a report by Guy Engelen:
http://www.proland.iung.pulawy.pl/materials/wp1/engelen.pdf
Cellular automata get their name from the fact that they consist of cells, like the cells on a
checkerboard, and that cell states may evolve according to a simple transition rule, the
automaton. A conventional cellular automaton consists of:
- a Euclidean space divided into an array of identical cells. For geographical applications a 2dimensional array is most practical;
• a cell neighborhood. For flow and diffusion processes the 4 or 8 immediate neighbors are
sufficient, but for most socio-economic processes larger neighborhoods are required;
• a set of discrete cell states;
• a set of transition rules, which determine the state of a cell as a function of the states of cells in
the neighborhood;
• discrete time steps, with all cell states updated simultaneously.
A generic Cellular Automata model
Over the past several years, a generic constrained cellular automata model was developed and
applied to urban (White and Engelen, 1993, 1994; White et al., 1997) and regional (Engelen et
al., 1993, 1996, 1998, 2000) cases. This model has the following characteristics:
The cell space
43
The cell space consists of a 2-dimensional rectangular grid of square cells each representing an
area ranging from 50 to 500 m square. The grid size and shape varies according to the
requirements of the application, but is typically less than 500 by 500 cells. The grid may be
larger, but at the cost of long run times. The same applies to the resolution of the model: it is
technically possible to increase the resolution of the CA model, but this requires working on
larger neighborhoods as well, which increases the essential to analyze whether this would lead
to any better results. Very often the basic map material will not be available or it will become
unreliable at high resolution, and the processes modeled are laden with lots of uncertainty. Thus,
a higher spatial resolution might give a false impression of detail and information.
The cell neighborhood
The cell neighborhood is defined as the circular region around the cell out to a radius of eight
cells. The neighborhood thus contains 196 cells (see Figure 14) that are arranged in 30 discrete
distance zones (1, 2 , 2, 5 ,...). Depending on the resolution of the grid, the neighborhood
radius represents distances ranging from 0.4 to 4 km (for grid resolutions ranging from 50 to 500
m). This distance delimits an area that is similar to what residents and entrepreneurs commonly
perceive to be their neighbourhood. It thus should be sufficient to allow local-scale spatial
processes to be captured in the CA transition rules.
Figure 14: For the calculation of the neighborhood effect, a circular neighborhood consisting of
196 cells is applied (left). For each land use function, the transition rule is a weighted sum of
distance functions calculated relative to all other land use functions and features (Right).
44
The cell states represent typically the dominant land use in each cell. A distinction is made
between dynamic, called Land-use Functions, and static elements, called Landuse Features.
Land-use Features will not change as the result of micro-scale dynamics: they do not change
location, but influence the dynamics of the Land use Functions, and thus affect the general
allocation process. For example a Land use function ‘Beach tourism’ will be strongly influenced
by the presence (or absence) of the land use feature ‘Beach’. Our models will operate on a
maximum of 32 states, 16 of which are land use functions and 16 are land use features. Clearly,
raising the number of states in the CA will increase, in theory at least, the number of possible
state transitions of each cell, and defining the transition rules of the model will become more
cumbersome. Again, it requires special attention on behalf of the model developer to keep this
complexity within limits. It is only useful to distinguish between land uses if and only if these
land uses behave differently in space. If however their spatial dynamic is very similar then land
uses can just as well be combined into a single land use function.
The neighborhood effect
The fundamental idea of a CA is that the state of a cell at any time depends on the states of the
cells within its neighbourhood. Thus a neighbourhood effect must be calculated for each of the
land use function states to which the cell could be converted. In our models, the neighbourhood
effect represents the attraction (positive) and repulsion (negative) effects of the various land uses
and land covers within the neighbourhood (see Figure 14). In general, cells that are more distant
in the neighbourhood will have a smaller effect. Thus each cell in a neighbourhood will receive a
weight according to its state and its distance from the central cell. Specifically, the
neighbourhood effect is calculated as:
N j   wkxd I xd
(20)
x d
Where: wkxd is the weighting parameter applied to land use k at position x in distance zone d of
the neighborhood, and Ixd is the Dirac delta function where Ixd = 1 if the cell is occupied by land
use k; otherwise, Ixd = 0
The transition rules
For cellular automata developed on a homogeneous cell space, a vector of transition potentials
(one potential for each function) is calculated for each cell from the neighborhood effect. The
45
deterministic value is given a stochastic perturbation (using a modified extreme value
distribution), such that most values are changed very little but a few are changed significantly:
Pj  vN j
(21)
where Pj is the potential of the cell for land use j, v is a scalable random perturbation term Nj is
the neighborhood effect on the cell for land use j.
For cellular automata developed on a non-homogeneous cell space, the transition potential will
include next to the neighborhood effect also the attributes representing the details of the cell
space.
Once the transition potentials for all cells and all functions have been calculated, the transition
rule is to change each cell to the state for which it has the highest potential - subject, however to
the constraint that the number of cells in each state must be equal to the number demanded at that
iteration. Thus all cells are ranked by their highest potential, and cell transitions begin with the
highest ranked cell and proceed downward. The number of cells required is determined external
to the cellular model in a ‘macromodel’ as will be explained in section 5. It is imposed as a
constraint on the cellular automaton. When a sufficient number of cells of a particular land use
have been achieved, the potentials for that land use are subsequently ignored in determining cell
transitions; the result is that some cells are not in the state for which they have the highest
potential. Each cell is subject to this transition algorithm at each iteration, although most of the
resulting “transitions” are from a state to itself, that is, the cell remains in its current state.
46
APPENDIX I - Step by step Modeling with Dyna-CLUE
Step 1: Is Dyna-CLUE the adequate tool for my research questions?
The exercises, paper and descriptions of the CLUE models should have given you a good idea of what
you can use the model for. Basically, the CLUE modelling framework is developed to spatially allocate
land use changes for visualising the impacts of different scenarios on land use patterns. In case your
research has different objectives, e.g., determining the aggregate quantity of land use change as result
of economic policies, it is better to choose another model.
Step 2: Do you have sufficient information with respect to changes in demand for land use areas at the
aggregate level?
The Dyna-CLUE model requires projections of the change in area for the different land cover types at the
level of the study region as a whole. These may be derived from trend extrapolation (so trends are
needed), from rough scenario assumptions (e.g., a 10% increase of agricultural area over the next 20
years) or from advanced models such as global global economic models or integrated assessment
models (as in www.eururalis.nl). For a specific application it may also be possible to combine methods
for the different land cover types, as long as is made sure that the results are leading to a consistent
change in land areas, i.e., equalling the total area available within the study region. These data should
be prepared before the Dyna-CLUE modelling is started.
Step 3: Build a conceptual model for your study area
The conceptual model should address a number of questions relevant to the design of the model:
Q1: What is the extent of the study area that you want to address?
Q2: What are the land use types that you are interested in (only include land use types for which you
think information is available)?
Q3: List for each of the land use types a number of location factors which you think may affect allocation
decisions?
Q4: Determine for each of the land use types how you will determine the change in area at the level of
the study region (see step 2)
47
Q5: Are there any specific, fixed conversion trajectories that need to be taken into account?
Q6: Are there specific spatial polices to be considered?
The answers to these questions can be filled in the diagram (figure 1) for a schematized model setup
Step 4: Prepare data
-Choose the resolution of your spatial data based on the resolution of your land use data and location
factor data. It makes no sense to choose a resolution for which the location factors do not show any
variation between cells. Furthermore, a high spatial resolution will result in high calculation times.
Calculation times are reasonable at resolutions below 1200x1200 cells.
-Convert all spatial data to a similar projection. Equal area projections are preferred
-Reclassify thematic data to a classification to be used in the modelling, e.g., the land use types or the
classes to be used as location factors. Please note that the first class should always have class number 0.
-Convert all data to a grid with the same extent and resolution (pls note that the upper left corner of
each grid should be located at exactly the same location)
-Prepare a ‘mask’ that contains value 1 inside the study area and ‘nodata’ outside.
-Fill gabs within all data layers, either by adding auxiliary data or by interpolation methods (e.g., ‘assign
proximity’ / ‘eucallocate’).
-Multiply all layers with the mask
-Export all data to ASCII grid data files. It is easiest to directly use the naming conventions of CLUE:
cov_all.0 for the initial land cover and sc1gr[number coding].fil for the location factors
Step 5: Statistical analysis or setting up suitability maps based on decision rules
The procedure for quantifying the role of the different location factors in the suitability for a specific
land use type by statistical analysis is described in Exercise 4.
Step 6: Create a directory for your model application
48
CLUE only needs and produces files within one directory. It is most convenient to create a new directory
for your application and store all the files that you prepare for the model in that directory. You may start
by copying clues.exe and clues.hlp into the directory. Note that if you use ArcView3.x it is not possible to
have spaces within the path name (e.g. c:/documents and settings/clue).
Step 7: Prepare demand / land area claim file
The procedure is described in exercise 2.2. Please note that the total area of all land use types together
may not change or exceed the surface area of the active cells within the study area. Indicate in the top
line the total number of lines in this file, this should equal the number of years to be simulated plus the
initial year. The second line of the file gives the surface area of the land use types in the initial year. This
should equal the area as indicated in the map of the initial year (cov_all.0). Please note that the units
should be equal to the units as specified in the main parameter file, if the reference units of all maps are
in meters the preferred land area unit is hectares. The resulting demand file should be saved in the
simulation directory with the name demand.in* with * being a number or character.
Figure 15. Fill in the grey boxes in order to make a draft setup of your model configuration
49
Step 8: Prepare a region file
In the standard region file all cells that have a land use type in the initial situation are allowed to change.
This standard region file can easily be made by reclassifying the mask made in step 4 to value 0 in all
cells that need to be calculated and ‘no data’/’-9999’ in all other cells. This file needs to be exported to
an ASCII file and saved in the simulation directory as region**.fil where ** may be any name with
multiple characters.
For scenarios in which certain areas are not allowed to change it is possible to create an alternative
region file in which the ‘static’ regions are assigned value -9998.
Step 9: Copy all files with location factors in the simulation directory
If you did not yet do so in step 4 it is now needed to copy all location factors used in the statistical
models into the simulation directory as ASCII grid files named sc1gr*.fil where * stands for the location
factor number. Note that numbering should be consecutive and start with 0.
Step 10: Copy the initial land use map to the simulation directory
This file should contain the initial land use map with land uses numbered consecutive from 0 onwards.
The format is ASCII grid named cov_all.0
Step 11: Set-up the main parameter file
Create within the simulation directory a text file called main.1 (e.g. by opening Notepad and saving an
empty text file). You can now edit this file using the CLUE interface (click clues.exe / file | edit main
parameters) or by editing the main.1 file with a text editor (Notepad/Wordpad etc.). In the CLUE-help
file you can exactly read what parameters need to be defined. Define all parameter settings as adequate
for your case study area.
Step 12: Create the regression parameter file
Create within the simulation directory a text file called alloc1.reg (e.g. by opening Notepad and saving an
empty text file). You can now edit this file using the CLUE interface (click clues.exe / file | edit regression
results) or by editing the alloc1.reg file with a text editor (Notepad/Wordpad etc.). In the CLUE-help file
you can exactly read how the file should be formatted. The file should reflect the results of the statistical
analysis. In case for one land use type no changes in land use are simulated (e.g., a static land use type)
50
it is still needed to define parameters for the regression equation. In that case a regression with equal
values should be indicates, e.g., a constant with value 0.7 and 1 location factor with beta value 0.
Step 13: Create the conversion matrix
Create within the simulation directory a text file called allow.txt (e.g. by opening Notepad and saving an
empty text file). You can now edit this file using the CLUE interface (click clues.exe / file | edit
conversion matrix) or by editing the allow.txt file with a text editor (Notepad/Wordpad etc.). In the
CLUE-help file you can exactly read how the file should be formatted. It is easiest to first conceptually
think which conversions are possible and which are not possible. This can be implemented by the values
1 and 0 in the conversion matrix. After a test run is successfully made it is possible to further specify the
matrix with time lags and other more advanced options.
Step 14: Optional: specify neighbourhood interactions
See the help file for more information. In case neighbourhood interactions are not considered (option 0
for neighbourhood interactions in the main parameter file) these files do not need to be specified.
Step 15: Test the model
Start the model by selecting a demand file, region file and click RUN!
51
Appendix II - Implementation of land use and climate
change scenarios at Rokua aquifer
1. Conceptual Model For Scenarios
Figure 1 presents a conceptual model, in which climate change and land use scenarios, cause and effect
relationships of scenarios and important ecosystems are highlighted (conceptual model development is
described in D5.2 UOULU contribution). Three different codes will be applied in groundwater flow
modeling: HydroGeosphere (Therrien et al., 2010), MODFLOW (McDonald and Harbaugh, 1983) and
Coup-model (Jansson and Karlberg, 2004). Coup model was chosen to simulate the plant-soil system
because of its ability to include snow and soil frost processes in simulations. Coup-model and MODFLOW
will be used in a sequentially coupled manner. HydgoGeoSphere treats subsurface and surface water
flow in a fully integrated manner so it is used to model the aquifer system as a whole.
2. Climate Change Scenarios
Climate change is expected to influence water fluxes in the study in almost throughout the conceptual
model. Changes are expected for groundwater as a result of changes in snow accumulation and melt,
precipitation and evapotranspiration (Okkonen and Kløve, 2010). Downscaled climate data from
GENESIS project partner SMHI complied with the assumptions, showing an increase both in annual
precipitation and annual average temperature.
In principle the effects of climate change are seen in all water flow components, but only the most
important are highlighted in blue (Fig. 1). Firstly amount of annual precipitation is projected to increase
over time, leading to gradual changes to water input to the system as a boundary condition. Annual
season of snow cover period is expected to reduce due to milder winter climate. Changes in
temperature and in water available for evapotranspiration change both lake evaporation and
evapotranspiration. Combined effects of changes in precipitation and evapotranspiration will lead to
changes in timing and amount of groundwater recharge. Changes in recharge affects annual
groundwater table fluctuations, leading to changes in GW-lake interaction. Possible decrease in soil frost
depth and durations is expected to affect soil permeability.
Climate change scenarios are adopted in Rokua aquifer as changes in model boundary conditions.
Parameter values of the models are treated as independent from climate change scenarios. Changes in
52
vegetation caused by climate change scenarios are considered minor compared to changes caused by
land use scenarios.
Figure 1. Conceptual model including relationships of important ecosystems and scenarios of landuse
change and climate change
On top of the esker (recharge area) can be found rare lake types, old forests in natural state and lichen
coverings supporting endangered vegetation and insect species. Ares include also several other nature
types and are protected in NATURA2000 and Finnish national park network. Spring ecosystems in
groundwater discharge area are mostly not included in natural conservation programs, and they are
53
heavily altered (Fig. 1, headwater streams/peatlands). Changes in hydrology are expected to affect lake
and spring ecosystems mentioned above, but modeling the response of biotic components is out of
modeling scope for the study site.
3. Land Use Scenarios
3.1 Current land use and pressures
The main effect of land use to the aquifer is speculated to come from excessive draining of peatlands
located at the discharge area of the aquifer. Peatlands are drained to induce forest growth for the use of
forestry (channels in figure 3 vertical profile). Land use outside the recharge area consists mostly of
forest, low productivity forest and wasteland (Fig. 2). Some agricultural and peat production areas are
present outside the recharge area.
FIGURE 2. DOMINATING LAND USE TYPES AND NATURE CONSERVATION AREAS
Land use in the recharge area consists mostly of forests used for forestry activities (Fig. 2). North and
northeast parts of recharge area are protected with NATURA 2000 network and Finnish national park
network. Small scale anthropogenic development such as second homes, recreation facilities and paved
54
roads has been constructed adjacent to kettle-hole lakes. Top of the esker is used for forestry but not
drained like the surrounding peatlands. Forestry operations locally modify the pine tree canopy density
(e.g. during loggings).
Changes in land use in the area are expected to be related to management of forestry industry both in
the recharge and discharge area. Considerable pressures from urbanization, agriculture or peat
production are not expected. The Rokua area has been accepted to European Geoparks Network (EGN)
in year 2010. EGN aims to protect geodiversity, to promote geological heritage to the general public as
well as to support sustainable economic development of geopark territories primarily through the
development of geological tourism. Partly because of EGN partnership tourism can be expected to
increase in the area. This may involve some development projects in the area, but major changes in land
use are not likely because tourism relies on the nature and geologically unique landscapes.
3.2 Future scenarios
Land use scenarios for the research area are formulated based on research results and discussions with
stakeholders and environmental authorities. Pressures for potential land use changes are thought to
largely depend on the decisions made for managing forestry industry in the area. Land use change
modeling was not seen as a relevant choice to formulate land use scenarios. Land use scenarios for the
area are separated between:
I) land use in the recharge area on top of the esker and
II) land use in groundwater protection area in the discharge area (Fig.3).
55
Figure 3. Landuse scenarios are separated between recharge and discharge areas. Left: map
presentation, right: vertical cross section.
Land use scenarios in the recharge area are related to management of forestry/loggings changing areas
vegetation, resulting in changes in transpiration (Fig. 1). Different stages of forest growth are not
considered, but changes in vegetation are treated as “lumped” overall changes upheld by land use
practices, leading to permanent changes in vegetation structure for the time span of the scenario (100
years). Vegetation is considered not to be affected by climate change. Land use scenarios in the
recharge area are as follows:
1) steady state
- Natural conservation and logging practices are continued as now
- Applied in models: current vegetation and parameters describing it
2) more logging/forestry
- Logging of the area is more intense, natural conservation areas are reduced
- Applied in models: Reduction of pine tree, e.g. LAI or tree coverage -50%
3) more protection
- Natural conservation areas are expanded, logging activities decrease
- Applied in models: increase of pine tree, e.g. LAI or tree coverage similar to
natural conservation areas for the whole recharge area
4) hazard scenario?
- Forest fire (typical to area in natural state) burns parts/all of the forest
- Applied in models: removal of pine tree? Time span of 100 years is not realistic
due to forest growth, dynamic parameters required
Land use scenarios in the discharge area affect primarily groundwater discharge to peatlands (red arrow
in figure 1). Land use changes in peatlands are related to changes in the resistance of the confining peat
layer to groundwater discharge. Changes in vegetation are also contributing to restoration of
endangered spring ecosystems. Changes in hydrology (transpiration and runoff) of the discharge area
would have feedback to areas vegetation and transpiration, but such effects are out of research scope.
Land use scenarios in the discharge area are as follows:
1) Steady state
- New peatland drainages are forbidden, but reopening clogged drainage routes are
allowed with permission granted and evaluated by Evironmental officials.
- Applied in models : resistance for GW discharge trough peat layer remains the
same
2) Protection policy
- All forestry drainage is forbidden in groundwater area
56
-
Area for groundwater protection zone is expanded where seen necessary
Applied in models : resistance of peat layer increases slowly with time, as ditches
get clogged
3) Restoration policy
- Risk areas of excessive groundwater discharge to peatlands are mapped and
ditches are restored to close natural state by damming
- In low risk areas drains are allowed and clogged drainage can possibly be
opened if no risk is detected
- Applied in models: resistance of peat layer is actively increased in critical
discharge areas
Land use scenarios in recharge and discharge areas are combined to follow a certain policy lines (table
1). Policy lines aim to create reasonable combinations of scenarios that would take place simultaneously
in the recharge and discharge areas. Selection of policy lines also reduce the amount of scenarios to be
simulated. Policy lines can be outlined as 1) Steady state 2) Modest protection 3) Extensive protection 4)
Increased logging. Work from WP 6 can be also used add information to scenarios presented. For
example if the Multicriteria Decision Analysis (MCDA) pinpoints some scenario combination as most
acceptable by stakeholders, this can be added to modeling. Attributes and be combined to form
alternative sets different from table 1, so that certain policy lines have two or more scenarios.
TABLE 1. Landuse scenario matrix
Policy line
Recharge area
Discharge area
Steady state
1) steady state
1) steady state
Modest protection
1) steady state
2) protection policy
Extensive protection
3) more protection
3) restoration policy
Increased logging
2) more logging
1) steady state
4. Model parameterization and uncertainties
4.1 model domains affected by climate change and land use scenarios
57
Climate change affects only the meteorological boundary conditions of the model domain. Possible
changes and feedbacks in model parameter values due to climate change are not considered at this
point. Land use scenarios will be reflected in parameters controlling transpiration (LAI, vegetation
height, root depth etc.) and hydraulic properties of the peat layer. Major uncertainty and topic of
research is lack of experience how to model preferential flow through the peat layer at the aquifer
boundary (link from groundwater to headwater streams, figure 1). This compartment of the model is
important, because some of the land use scenarios should be able to affect the parameters/boundary
conditions at this part of the model. Also exact spatial location of such preferential flow exfiltration sites
is hard to pinpoint with the resources available. Novel approaches are possibly needed in modeling
scheme of preferential flow through peat layer. All of the parameters and ranges for individual
parameter values to be changed in land use scenarios are not yet explicitly reported. These will be
specified as modeling process evolves.
4.2 Geological uncertainties
Modeling approaches (coupled Coup-MODFLOW and HGS) share most of the sources of uncertainty.
Each step of building the geological map consists of various sources of error inherent in the methods,
leading to uncertainty in the geological model. Most important varying factors in the geological model
are the hydrological properties of different soils: possible differences in sand in eastern and western
part of the esker, peat layer and possibly continuous gravel core. Initial assumption of the soil is that soil
is homogeneous. As an exception to soil homogeneity, coarser sand and gravel deposits were found
from one deep borehole. Extent of the coarse sand and gravel deposit is unknown (Fig. 2), but deposit
needs to be considered in the model.
Because of the uncertainties in the geological structure three different geological model spaces are
needed:
1) Model with no massive continuation of the esker sand or gravel to the next groundwater
area (as in Fig. 4.)
2) model with continuous sand core to next groundwater area (bedrock deeper than in Fig.
4.)
3) model with continuous gravel core to next groundwater area.
4.3 Different uncertainties in modeling approaches
Some different uncertainty sources arise between the modeling approaches:
1) Uncertainties in Coup-MODFLOW coupling:
58
a. Horizontal flow component from the unsaturated zone to lake systems is not
possible to model with 1D Coup model. Component is possible to exist mainly
during spring melting period, when soil is partly frozen.
b. Need for feedback from Modflow to CoupModel. Is sequential coupling
sufficient, especially adjacent to lakes where depth to GW-table is low?
c. Problem with using MODFLOW as the model code might come with the peat
thickness (Fig.4). As the peat thicknesses are only one tenth of the usual sand
thicknesses in the area, grid structure of the MODFLOW can cause problems
building a working model that actually takes the peat into account.
2) Uncertainties HydroGeoSphere:
a. Capability to simulate Nordic winter conditions with snow and soil frost? Such
processes currently not included in the code.
Figure 4. Geological structure and topography of Rokua Esker (z-axix has 50:1 exaggeration). Yellow is
sand and green areas peatland surrounding the esker. Three problems concerning modeling area
represented are: 1. Variation of sand hydraulic conductivity, 2. Continuation of gravel and, 3. Peat
thickness compared to sand thickness.
4.4 Sensitivity analysis and uncertainty estimation
The sensitivity analysis for the Rokua model should include at least following parameters:
-
Hydraulic conductivities of soil types (peat, sand, gravel)
Drain conduction parameters
Effective conductivity in the unsaturated zone and
59
-
Recharge.
In the coupled (Coup-MODFLOW) procedure recharge is an input from COUP-model in the coupled
model procedure, it is important to define how sensitive the groundwater flow model is for this
parameter (and to the data coming from another model). As there is three different geological model
spaces this sensitivity analysis should be done in each model space. MODFLOW uses parameter
estimation PEST interface to estimate parameters and to give sensitivity analysis. There is also a
possibility to run Monte Carlo simulation in MODFLOW, either with random sampling or using Latin
Hypercube. This way data for more informational sensitivity analysis can be studied. For example HSY
generalized sensitivity analysis (Hornberger and Spear, 1981) could be conducted for the results in
MATLAB to examine if the parameters are behavioural or to form dotty plots. And also with the same
Monte Carlo data also GLUE analysis (Beven and Binley, 1992) is possible.
For both HSY and GLUE the amount of parameters (and model spaces) is quite high and needs a lot of
computational time. Therefore actually conducting the Monte Carlo run might be too problematic. To
simplify Monte Carlo run, parameter amount should be reduced. PEST sensitivity analysis could show
some of the parameters less important and this might be the way to ease up the Monte Carlo run.
In summary there are numerous sources of uncertainty in the groundwater model of Rokua esker. If the
model is used to make predictions how the land use changes and climate conditions effect the
groundwater and lake levels one has to emphasise that predictions are subject to these known (and
possibly unknown) uncertainty factors. Although it takes time and resources to understand how the
model actually works using sensitivity analyses and possibly even GLUE it would be worth the effort.
That is because you get better understanding how all different parts of the model (and input data) might
affect your predictions. And although the predictions might be more inaccurate, or uncertain, model is
not misused and it tries to be as realistic as possible with the data and the model code used and
available.
60
Appendix III - Implementation of land use and climate
change scenarios at Neo Sichirochorio aquifer
1. Implementation of concept in GENESIS case studies: the Neo Sichirochorio aquifer case.
Before the implementation of concept description, a brief statement of Neo Sidirochorio aquifer
groundwater flow model is provided. The results from isotope analysis in groundwater from Neo
Sidirochorio aquifer and surface water from the adjacent ecosystems (Vosvozis river and Ismarida Lake)
indicated that groundwater in the aquifer of the study area can be characterized as “young water”. The
comparison of the isotopic composition of river, lake and aquifer water indicates the hydraulic
connection between those three water bodies. Based on this evidence, conceptual model quantitative
inflows (sources) and outflows (sinks), as well as MODFLOW (McDonald and Harbaugh, 1988) packages
used to simulate them are summarized in Figure 1.
Recharge from Precipitation (RCH)
Inflow from Ismarida Lake (GHB)
INFLOWS
Inflow from Vosvozis River (RIV)
Lateral inflows from north and
southeastern boundary (GHB)
Irrigation (WEL)
OUTFLOWS
Lateral outflows from north and
southeastern boundary (GHB)
61
Figure 1. Major inflows and outflows for study area aquifer.
1.1. Climate change implementation
According to downscaled climate change data for the study area, the major general climate change
trends for the study area are illustrated in Figure 2 and are summarized as follows:


Temperature increment by about 4 oC until 2100.
Precipitation decrement by about 100 mm until 2100.
Figure 2. Precipitation and temperature fluctuations for the study area based on downscaled data from
Alexandroupolis meteorological station.
62
Based on these climate change characteristics, significant effects are expected to be observed in study
area aquifer which are summarized in Figure 3. The general trend illustrated in Figure 3, indicates
decrement in aquifer system inflows and increment in aquifer system outflows.
63
CHANGE IN BASIC
CLIMATE
PARAMETERS
GENERALCLIMATE
CHANGE (CH)
EFFECTS
CH EFFECTS IN GDE
AND PROCESSES
River flow
CH EFFECTS IN
AQUIFER
River inflows
to aquifer
River
Discharge to
Lake
Lake Level
TEMPERATURE
Lake inflows
to aquifer
Evapotranspiration
Crop ETr
Groundwater
outflows for
irrigation
Precipitation
percolation
Groundwater
recharge
River flow
River inflows
to aquifer
River
Discharge to
Lake
Lake Level
PRECIPITATION
Lake inflows
to aquifer
Inflows for all water
bodies
Effective
Rainfall
Precipitation
percolation
Groundwater
outflows for
irrigation
Groundwater
recharge
Figure 3. Effects of climate change in quantitative components of Neo Sidirochorio aquifer.
64
These trends were considered in modeling process by assessing their effects at each MODFLOW package
parameters. A description of model parameters and components affected by climate change is
following.
1.1.1. Inflows from Vosvozis River
STR (stream) package was used in order to simulate the hydraulic connection between Vosvozis river
and study area aquifer. STR package, in contrast with RIV package, accounts for the amount of flow instream and for this, it is more suitable to simulate the interaction between surface streams and
groundwater. The following parameters of STR package are affected by climate change:



Stream Inflow: Inflow rate to the reach of a stream segment, in units of length cubed per time. The inflow
rates for climate change scenarios have been estimated by SWAT (Arnold et al., 1998) model and inputted
in to STR package.
Stream Stage: This parameter corresponds to the free water surface elevation of the surface water body. A
seasonal average stream stage is considered, based on collected monitoring data.
Width: This parameter corresponds to the width of the stream channel in units of length. Similarly to
stream stage, a seasonal average stream stage is considered, based on collected monitoring data.
1.1.2. Inflows from lake
GHB (General Head Boundary) package is used to simulate water inflows from Ismarida Lake. The
concept of using GHB in this case is that flow into boundary cells from the lake is provided in proportion
to the difference between the head in the cell and the reference head assigned to the lake. So, lake level
is the key parameter that has to be considered for climate change impacts assessment in lake-aquifer
interaction simulation with GHB. Ismarida Lake is connected to the sea with an artificial channel
equipped with a wooden sluice which is controlled by local fishermen. The operation of the channel
sluice is made empirically by the local fishermen with a view to maintain sufficient water volume in the
lake for fishery purposes. The common practice of local fishermen is to open sluice at periods of high
inflows in lake from Vosvozis river and close it at periods with low inflows (usually from June-July to
September-October). Seasonal average lake water level was considered for modeling purposes.
1.1.3. Groundwater recharge
SWAT model was used in order to estimate groundwater recharge for the study area aquifer.
Groundwater recharge concerning the “deep aquifer” was found to be most representative for the study
area, because groundwater level is far below soil profile and root zone and no interaction between
aquifer and the simulated soil profiles is taking place. Climate change characteristics such as decreased
65
precipitation and increased temperature are incorporated into SWAT model which simulates all the
related processes (evapotranspiration, water uptake by plants, soil water storage etc.) and results in
groundwater recharge quantities calculation.
1.1.4. Outflows for irrigation
Blaney-Criddle method (Blaney and Criddle, 1950) was used to determine the reference crop
evapotranspiration ETo values in the study area which is expressed by the following formula:
ETo=0.254* p*(32+1.8Ta)
Where, ETo is the reference crop evapotranspiration (mm/day) as an average for a period of one month,
Ta is the mean daily temperature (°C) of the month and, p is the mean daily percentage of annual
daytime hours. According to this equation the increasing temperature trend of climate change
assessment for the study area, is directly incorporated for the calculation of ETo of the several crops
cultivated in the study area and therefore for the calculation of the irrigation needs.
Effective rainfall was calculated with the following equations (FAO, 1978):
Pe = 0.8 P-25 if P > 75 mm/month
Pe = 0.6 P-10 if P < 75 mm/month
Where, Pe is the effective rainfall in mm/month, and P is the cumulative rainfall of the month. According
to these equations the climate change trend of decreasing precipitation will result in lower effective
rainfall values and for this in higher quantities of water for the satisfaction of irrigation needs.
1.2. Land use change
Neo Sidirochorio aquifer is mainly covered by agricultural fields in which cotton appears to be the most
frequently cultivated crop. In order to identify the general land use change trends for Vosvozis River
basin and Neo Sidirochorio aquifer, CORINE land cover data for years 1990 and 2000 was compared.
Unfortunately, CORINE land cover data for year 2006 is not available for Greece. As illustrated in Figure
4, there are no significant land use changes in Vosvozis River basin, as the only major change observed,
corresponds to “Egnatia Odos” national road, the construction of which has been finished.
66
Figure 4. CORINE land cover for year 1990 (left) and 2000 (left). Vosvozis river basin as resulted after
SWAT model delineation and Neo Sidirochorio aquifer are illustrated with a red and a black dashed line,
respectively.
Moreover, future climate change trends were assessed using projected land-use changes from the panEuropean ALARM (Assessing Large scale Risks for biodiversity with tested Methods) GRAS (Growth
Applied Strategy) scenario (Spangenberg 2007). The results are presented in Figure 5 and according to
these, cropland which is the major land cover for the wider study area (82.7%) and urban land covers
are remaining constant until year 2080. Also, since year 2060 and until 2080, grass land and liquid
biofuels are expected to be decreased by 0.7% and 1.8%. This decrement is counterbalanced by surplus
land cover which includes land use/cover that cannot be categorized to the other eight major land
use/cover categories. The major outcome adapted from ALARM GRAS land use change scenario for Neo
Sidirochorio aquifer is that cropland area is remaining constant and no urban expansion is expected.
67
Figure 5. ALARM GRAS land use change scenario results for the wider study area.
One major trend observed in agricultural areas across Greece is the construction of photovoltaic parks.
As the Greek government has given economic motivation to Greek farmers for the construction of
photovoltaic parks within their fields, this land use change trend cannot be ignored for the study area.
Land use change from cropland to solar power parks area was simulated in SWAT model in order to
determine changes: a) in the hydrological regime of Vosvozis river basin and subsequent changes in
inflows to the aquifer from river water loss and b) in groundwater recharge for the study area aquifer.
The major assumptions made for the above land use change simulation are the following three:



Soil albedo increment in order to simulate the increased solar radiation adsorption by
solar panels, as well as soil shading effects.
Decrement in upper soil layer hydraulic conductivity because of soil compaction carried
out from heavy machinery (e.g. excavators) during the construction of solar parks.
Increment in potential runoff volume caused by soil compaction and by the usage of
materials such as concrete. In our case this fact corresponds to increment in Curve
Number coefficient used for runoff volume with SCS method.
1.3. Assessment of uncertainty of groundwater flow model parameters and
components
In general, the complexity of groundwater systems is significant, mainly because of the heterogeneity,
the interaction with other water systems and processes and non–linearity observed in their hydraulic
68
behavior (Meyer and Cohen, 2010). Sensitivity Analysis (SA) is used to the present case study in order to
assess the uncertainty of model parameters and components. Generally, SA corresponds to the
technique used to identify the way that the variation (uncertainty) in the output of a model can be
attributed to the variation in the input parameters of the model (Saltelli et al., 2008). The model
parameters and components related in climate and land use change which are included in SA process,
are presented in Table 1.
Table 1. Model parameter and components affected by climate and land use change, which area
included in SA.
Model
parameter
component
or
Inflows to aquifer
Vosvozis river
from
Degree of uncertainty
Specific parameter change
Medium
Vertical hydraulic conductivity
of the streambed material
Lateral inflows and outflows
from north and southeastern High
boundary
Boundary
conductance
head
and
Inflows from Ismarida Lake
Medium to High
Boundary
conductance
head
and
Outflows for irrigation
Medium
Groundwater pumping rate
Aquifer recharge
Medium to high
Aquifer recharge rate
Because of medium to high degree of uncertainty in model parameters and components presented in
Table 1, SA technique constitutes a valuable method for the determination of model parameter and
components which are more important to climate and land use change assessment, as well as for
understanding the overall behavior of the aquifer. The approach of SA in this case study relies on the
perturbation of the parameters presented in Table 1 and the subsequent assessment of perturbation
effects in groundwater level and water budget.
69
Appendix IV - Implementation of land use and climate
change scenarios at Köyceğiz-Dalyan
6.3 Modelling the Response of the Biotic Components of the GDE to Climate Change
Climate change affects the organisms through two mechanisms, direct effects of the changes in
meteorological and climatic parameters and changes in the physical environment that is partly formed
by habitat forming organisms such as reeds, grass, trees and corals. Traditional mass balance based
ecological models usually have some consideration on the direct effects of meteorological and climatic
parameters such as increase of respiration rates for warmer periods, however in most of the traditional
models there is either an oversimplified consideration of the physical environment on the biotic state
variables or no consideration of it is included.
6.3.1 Direct Effects of the Meteorological, Hydrological and Climate Variables
Temperature is one of the most important meteorological variables for organisms. The instantaneous
effects on the temperature on geochemical reaction rates are represented either as a monoton relation
between the temperature and the reaction rate (such as an Arrhenius type equation) or as an
asymmetric bell-shaped curve where first the reaction rate is increasing with temperature until an
optimum temperature range than does not change until the upper limit of temperature and decreases
with increasing temperature. The first type of temperature relation is mostly used for

Geochemical reactions such as dissolution, weathering, oxidation abiotic mineralization

Growth rate of microorganisms if they survive in the temperatures warmer than the physical
environment and therefore are not subjected to temperature limitation, respiration rates except
photorespiration of microscopic and higher plants
The second type of temperature relation is more complex (Figure x.1) to represent mathematically and
requires more parameters that are needed to be calibrated, however, it allows the modeller to
distinguish between stenotherm and eurtherm organisms (Figure x.2).
70
Process rate
Optimum range
Tolerance limits
Temperature
Figure 1. Realistic representation of the effect of temperature on biological process rate
Figure 2. Response of steotherm and eurytherm organisms to temperature change
71
As seen from Figure 1, the same equations could be used to model the instantaneous effects of other
meteorological and hydrological variables (such as the ones listed below) on biological processes such as
growth, respiration or death as long as a meaningful set of parameters is provided.

Salinity

pH

Light intensity

Water content in the unsaturated zone

Depth of the unsaturated zone

Some micro nutrients like boron that are necessary for growth but have direct inhibiting effect
at high concentrations.
If long term direct effects of the physical environment on the organisms illustrated in Figure 3 are to be
simulated then the modellers need to find a way to include the effects of additional variables as well.
One example is the initiation of mating and spawning cycle of higher level organisms. In order to trigger
spawning, a threshold of degree-days should be exceeded. Other examples are:

Blooming of plants or root development after a number of frost free days since the last days of
winter.

Number of dry/wet days in a wetland or ephemeral river

Number of days with flood events
These variables are climatic variables rather than meteorological variables. Their long-term but direct
effects on the organisms are not considered in most of the traditional mass balance or energy based
ecological models.
72
Value of the environmental
variable
Natural conditions
Climate change
Early time of activity
(blooming, spawning,
etc.) because of the
climate change
Natural time of
activity
(blooming,
spawning, etc.)
Time of the year
Figure 3. Long term effects of the physical environment
6.3.2 Effects of the Changes in Physical Environment on Organisms
These effects are indirect and are not easy to model. Some organisms are considerably affecting the
physical environment. A very common example for such an effect is increase of evapotranstpiration by
trees and hence decreasing of groundwater tables. Other organisms are affecting the physical
environment by forming habitats. For example shrubs and reeds are providing shelter for smaller
organisms from their predators an increase the succession. Trees in a forest form specialized habitats for
different organisms as well. Climate change may affect the habitat forming organisms directly and other
organism that depend on them indirectly. Such relations together with long term reactions of habitat
forming organisms on climate change should be considered for predicting the response of biotic
ecosystem components on climate change.
In addition its habitat forming biotic components, the physical environment has abiotic components as
well. Examples are the hills, riverbeds, deposited sediments, boulders or the soil. The abitotic
environment is affected by climate change as well. Rates of geological, hydrological and geochemical
weathering processes may be altered because of the climate change. Some of these processes are very
slow occurring in geological time scales and their reactions to climate change will be only considerable
after centuries even millennia. On the other hand, processes such as erosion will react much faster to
climate change since it is related to single storm events as well. Climate change may alter the intensity
and frequency of single storm events resulting in export of sediments, nutrients into surface waters as
well as enhance the loss of soil in the non vegetative seasons. Whether slow or fast, the abiotic
73
components will be affected by climate change and their reactions will affect the succession of habitat
forming organisms.
6.3.3 The Temporal and Spatial Scale
The hydrological events usually respond rapidly to meteorological changes and most of them are
completed in days. To characterize the dynamics of those events, time steps on the order of minutes to
hours are required. An exceptional case is when the modellers have to deal with surface water
hydrodynamic related processes, where running the models on smaller time steps such as seconds may
be necessary depending on the conditions. Biogeochemical components of the ecological models that
usually deal with the chemical components (such as sediments, nutrients, gasses, pollutants if any) as
well as organism in the lower levels of the food web (such as bacteria, plankton in case an aquatic
ecosystem is dealt with, microscopic fungi) are strongly dependent on the hydrological transport
processes and therefore should be used with similar time steps or with time steps that are one order of
magnitude longer. Most of the hydrological and biogeochemical processes respond to climate change
relatively fast.
The differences between the organisms in the upper and lower levels of the food web focused on spatial
and temporal model scales are listed below:

The organisms in the upper levels of the food web depend more on the physical environment
than the biogeochemical components. Upper level organisms are generally more vulnerable to changes
or loss of their habitats.

The reactions of the organisms to environmental changes in the upper level of the food web are
usually much slower than those that are in the lower levels. The lower level organism and process rates
of biogeochemical components react to environmental changes in a temporal scale of time steps (hours)
whereas upper level organisms need much longer adaptation time to environmental changes and
usually react to the cumulative effects of many events over longer time. For example, climatic variables
are more important for them as meteorological events or change of the long term average of food/prey
over longer period is more important than instantaneous changes of nutrients. For the same reasons a
longer simulation time is needed to model the dynamics and succession of upper level organisms. Since
their reaction to environmental changes such as the climate change will be relatively slowly a longer
time step such as weeks, months or seasons is sufficient to characterize their dynamics.

Upper level organisms are affected less by transport than the lower level microorganisms or
geochemical components such as nutrients. Microorganisms and geochemical components can only be
transported by hydrological processes with the exception that some of plankton in the aquatic
ecosystem can migrate vertically. Therefore their reaction rate is much more dependent on the spatial
discretization scale on the hydrological transport sub-models. The spatial discretization scale of the
transport sub-model is designed to characterize the spatial heterogeneity of physical transport
processes. Traditionally to save computational time, several control volumes of the hydrological
transport model are lumped together when generating the model domain of the biogeochemical cycling
model, but number of control volumes in the biogeochemical cycling model is generally on the same
order of magnitude or one order of magnitude less than the control volumes in the hydrological
74
transport model. On the other hand, most of the upper level organisms can screen an area in unit time
that is larger than the passive travel support provided by the hydrological transport processes per unit
time. Therefore, for upper level organisms spatially averaged conditions over a larger area are more
important than conditions characterized with a high resolution spatial grid. This situation allows the
modeller to construct the biogeochemical model with smaller number of control volumes if the spatial
resolution of the geochemical components and microorganisms are not needed to be high or to lump
several control volumes in the biogeochemical model together for the ecological sub model dealing with
the upper level organisms.
6.3.4 The Behaviours and Life Cycles of Biotic Components
The response of geochemical components to climate change through abiotic processes is usually
predictable using traditional mass balance model if the key processes are representing the physical
environment correctly. For biotic components; even for relatively simple ones, prediction is more
difficult since they can adapt to environmental conditions. Complex models dealing with organisms were
successfully validated and reproduced the field measurements, however since most of them lack the
adaptation capability of the organism their validity for long terms such as decades where the
environmental conditions and forcing factors are changing due to the climate change. The bottlenecks in
traditional modelling and possible solutions how to deal with them is given below:

Organisms and communities can adapt to environmental changes, especially if those occur
gradually (several orders of magnitudes larger than the model time step) such as the case of climate
change. This adaptation may occur on species level, when individuals on one species develop resistance
to the changing environment over several generations (in case of organisms with short lifetime such as
microorganisms) or the communities may respond to such changes with the shifts in species
distribution. These changes can be simulated by developing algorithms that activate or deactivate
special options during the simulation or that change the numerical values of essential model constants
such as the maximum growth rate or the food assimilation efficiency to emulate the shifts in species
distribution. This solution could be further enhanced if a knowledge based system related to relevant
organisms is coupled with model.

Many organisms have several developmental stages such as eggs, larval stage, juveniles and
adults. Organisms are vulnerable to different environmental conditions in different developmental
stages of their life cycle. Climate change may affect one or more of theses stages. If one of the stages is
disturbed by environmental changes, the succession of the organism may be jeopardized. The affects
may take time. For example environmental conditions may reduce the probability of hatching
considerably or even to zero, but since such organisms have a longer time the effects may be observed
after several years. Life cycles can be simulated using multi-stanza models that consider several ages of
the same organism.
75

If one group of organisms (represented by a state variable in the ecological model) is affected
because of environmental changes such as the climate change other organisms in the ecosystem
(represented by different state variables in the ecological model) will be affected as well and the entire
ecosystem will try to adapt to the new conditions. The interactions among the organisms will be
rearranged to reach a new optimum within the ecosystem under new environmental conditions. This
type of ecosystem behaviour can be simulated by applying optimization algorithms (such as generic
algorithms, price algorithm, simulated annealing), where the set of model constants that represent the
ecosystem behaviours are automatically changed during simulation to optimize the value of one or
several derived model outputs such as primary production, consumption, energy.
REFERENCES
Abildtrup J., Audsley, E., Fekete-Farkas, M., Giupponi, C., Gylling, M., Rosato, P. and Rounsevell, M.
(2006) Socio-economic scenario development for the assessment of climate change impacts on
agricultural land use: a pairwise comparison approach, Environmental Science & Policy, 9: 101 –
115.
Adams, D., Alig, R., Mc Carl, B. A. and Murray, B. C. (2005) FASOMGHG consptual structure and
specification.
electronic
version,
at:
http://agecon2.tamu.edu/people/faculty/mccarlbruce/papers/
Arnold JG, Srinivasin R, Muttiah RS, Williams JR. 1998. Large Area Hydrologic Modeling and Assessment:
Part I. Model Development. JAWRA 34(1):73-89.
Batten, D. F. and Boyce, D. E. (1986) Spatial Interaction, Transportation and Interregional Commodity
Flow Models. In: Handbook of Regional and Urban Economics. Vol. 1. ed. P. Nijkamp, 357-406.
Amsterdam: North-Holland.
Beven, K.J., Binley, A.M., 1992. The future of distributed morels: model calibration and uncertainty
prediction. Hydrol. Process. 6, 279-298.
Blaney HF, Criddle WD. 1950. Determining Water Requirements in Irrigated Areas from Climatological
and Irrigation Data. USDA (SCS) TP-96.
Bonan, G. B., Levis, S., Kergoat, L. and Oleson, K. W. (2002) Landscapes as patches of plant functional
types: An integrating concept for climate and ecosystem models. Global Biogeochemical Cycles,
16.
Burniaux, J. M. (2002) Incorporating Carbon Sequestration into CGE Models: a Prototype GTAP Model
with
Land
Uses,
electronic
version
at:
http//www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=1144
Carey, H. C. (1858) Principles of Social Science. Philadelphia: J. B. Lippincott.
76
Castella, J. C., and Verburg P. H. (2007) Combination of process-oriented and pattern-oriented models of
land-use change in a mountain area of Vietnam. Ecological Modelling 202:410–420.
de Koning, G. H. J., Veldkamp A., Verburg P. H., Kok K., and Bergsma A. R. (1998) CLUE: A Tool for
Spatially Explicit and Scale Sensitive Exploration of Land Use Changes. In Information Technology
as a Tool to Assess Land Use Options in Space and Time, eds. J.J. Stoorvogel and W.T. Borma.
Proceedings of an International Workshop, Lima, Peru, 1997. Quantitative Approaches in Systems
Analysis, No. 16.
Deardorff, A. (1998) Determinants of bilateral trade: does gravity work in a frictionless world, in F.
Jeffrey (ed) The Regionalization of the World Economy, University of Chicago Press, Chicago, pp.
7-28.
Del Barrio, G., Alvera, B., Puigdefabregas, J. and Dfez, C. (1997) Response of high mountain landscape to
topographic variables: Central Pyrenees. Landscape Ecology, 12: 95-115.
Dietzel, C. and Clarke, K. (2006) The effect of disaggregating land use categories in cellular automata
during model calibration and forecasting, Computers, Environment and Urban Systems, 30: 78101.
Engelen G., White R., Uljee I. and Wargnies S. (1996) Numerical modeling of small island socioeconomics to achieve sustainable development, in Small Islands: Marine Science and Sustainable
Development, Coastal Estuarine Stud., vol. 51, edited by G. A. Maul, pp. 437–463, AGU,
Washington, D. C., doi:10.1029/CE051p0437.
Engelen G., Uljee I. and White R. (1998) Vulnerability Assessment of Lowlying Coastal Areas and Small
Islands to Climate Change and Sea Level Rise. Phase 2: Case Study St. Lucia. United Nations
Environment Programme, UNEP/CAR/RCU, Kingston Jamaica.
Engelen G., White R. and Uljee I. (1993) Exploratory Modelling of Socio-Economic Impacts of Climatic
Change. In: Climate Change in the Intra-America's Sea, G. Maul, (ed.) Edward Arnold, London,
pp.306-324.
Engelen, G. (1988) The Theory of Self-Organization and Modeling Complex Urban Systems. European
Journal of Operations Research, 37: 42-57.
Engelen G., van der Meulen M., Hahn B., Mulligan M., Reaney S., Oxley T., Mata-Porras M., Blatsou C.,
Kahrimanis S., Giannoulopoulos P., Mazzoleni S., Coppola A., McIntosh B. and Winder N. (2000)
MODULUS: A Spatial Modelling Tool for Integrated Environmental Decision-Making’, Final Report,
EU-DG12 (Contract ENV4-CT97-0685), Brussels.
FAO
(1978) Effective Rainfall in Agriculture. FAO
http://www.fao.org/docrep/X5560E/X5560E00.htm.
Irrigation
and
drainage
paper.
77
Flinn, K. M. and Vellend, M. (2005) Recovery of forest plant communities in post-agricultural landscapes.
Frontiers in Ecology and the Environment, 3: 243–250.
Fotheringham, A. S. and O’Kelly, M. E. (1989) Spatial Interaction Models: Formulations and Applications,
Kluwer Academic Publishers, Dordrecht
Gellrich, M., Baur P., Robinson B. H., and Bebi P. (2008) Combining classification tree analyses with
interviews to study why sub-alpine grasslands sometimes revert to forest: a case study from the
Swiss Alps. Agricultural Systems, 96:124–138.
González-Martínez SC, Bravo F. (2001) Density and population structure of the natural regeneration of
Scots pine (Pinus sylvestris L.) in the High Ebro Basin (northern Spain). Ann. For. Sci., 58: 277-288
Hansen, M. C., Defries, R. S., Townshend, J. R. G. and Sohlberg, R. (2000) Global and cover classification
at 1 km spatial resolution using a classification tree approach. International Journal of Remote
Sensing, 21: 1331-1364.
Heistermann, M., Muller, C. and Ronneberger, K. (2006) Land in sight? Achievements, deficits and
potentials of continental to global scale land-use modelling. Agriculture, Ecosystems and
Environment, 114: 141–158.
Hijmans R. J., Cameron S. E., Parra J. L., JONES P. G. and Jarvis A. (2005) Very high resolution interpolated
climate surfaces for global land areas. Int. J. Climatol., 25: 1965–1978.
Hornberger, G.M., Spear, R.C., 1981. An approach to the preliminary analysis of environmental systems.
Environmental Management 12, 7-18.
Jansson, P., Karlberg, L., 2004. Coupled heat and mass transfer model for soil-plant-atmosphere systems.
Royal Institute of Technolgy, Dept of Civl and Environmental Engineering, Stockholm, pp. 435.
Available Online: ftp://www.lwr.kth.se/CoupModel/CoupModel.pdf.
Jianchu, X., Xihui, A. and Xiqing, D. (2005) Exploring the spatial and temporal dynamics of land use in
Xizhuang watershed of Yunnan, southwest China, International Journal of Applied Earth
Observation and Geoinformation, 7: 299-309.
Joshi, C., De Leeuw, J., Skidmore, A., Van Duren, C. and Van Oosten, H. (2006) Remotely sensed
estimation of forest canopy density: A comparison of the performance of four methods,
International Journal of Applied Earth Observation and Geoinformation, 8: 84-95.
Kitamura, T., M. Kagatsume, S. Hoshino, H. Morita. 1997. A Theoretical Consideration on the Land Use
Change Model for the Japan Case Study Area. Interim Report IR-97-064. Laxenburg: International
Institute for Applied Systems Analysis.
78
Kräuchi, N., Brang, P. and Schönenberger, W. (2000) Forests of mountainous regions: gaps in knowledge
and research needs. For. Ecol. Manage., 132 (1): 73–82.
Lotsch, A., Tian,Y., Friedl M.A. and Myneni R.B. (2003). Land cover mapping in support of LAI and FPAR
retrievals from EOS-MODIS and MISR: Classification methods and sensitivities to errors.
International Journal of Remote Sensing, 24: 1997−2016.
Loveland,T.R. and Belward,A.S. (1997) The IGBP-DIS global 1k m land cover dataset, DISCover: First
results. International Journal of Remote Sensing, 18: 3291−3295.
Mac Donald D., Crabtree J. R., Wiesinger G., Dax T., Stamou N., Fleury P., Lazpita J. G., Gibon A. (2000)
Agricultural abandonment in mountain areas of Europe: Environmental consequences and policy
response. Journal of Environmental Management 59: 47–69.
McDonald, M.G., Harbaugh, A.W., 1983. A modular three-dimensional finite-difference ground-water
flow model. Open-File Report 83-875. U.S. Geological Survey.
McDonald MG, Harbaugh AW (1988) A modular three-dimensional finite-difference ground-water flow
model: Techniques of Water-Resources Investigations, U.S. Geological Survey, Book 6, Chapter
672 A1.
McIver, D. K. and Friedl, M. A. (2001) Estimating pixel-scale land cover classification confidence using on
parametric machine learning methods. IEEE Transactions on Geoscience and Remote Sensing, 39:
1959−1968.
Meyer PD, and SJ Cohen (2010) "Treatment of Uncertainty in Groundwater Modeling: A (Limited)
Research Perspective." In Proceedings of the World Environmental and Water Resources Congress
2010: Challenges of Change, May 16-20, 2010, Providence, Rhode Island, pp. 729-736.
Mitchell, T.D. and Jones, P.D. (2005. An improved method of constructing a database of monthly climate
observations and associated high-resolution grids. Int. J. Climatol. 25 (6): 693-712.
Morita, H., Hoshino S., Kagatsume M. and Mizuno K. (1997) An Application of the Land Use Change
Model for the Japan Case Study Area. Interim Report IR-97-065, November. Laxenburg, Austria:
International Institute for Applied Systems Analysis.
Mulley, B. G. and Unruh, Jon D. (2004) The role of off-farm employment in tropical forest conservation:
labor, migration, and smallholder attitudes toward land in western Uganda, Journal of
Environmental Management, 71: 193–205.
Myneni, R. B. Hoffman, S., Knyazikhin, Y., Privette, J. L., Glassy, J., Tian, Y. et al. (2002) Global products of
vegetation leaf area and fraction absorbed PAR from year one of MODIS data. Remote Sensing of
Environment, 83: 214-231.
79
Okkonen, J., Kløve, B., 2010. A conceptual and statistical approach for the analysis of climate impact on
ground water table fluctuation patterns in cold conditions. Journal of Hydrology 388, 1-12.
Overmars, K. P., Verburg P. H. and Veldkamp T. (2007) Comparison of a deductive and an inductive
approach to specify land suitability in a spatially explicit land use model. Land Use Policy, 24:584–
599.
Plane, D. (1984) Migration space: doubly constrained gravity model mapping of relative interstate
separation, Annuals of the Association of American Geographers, 74: 244-256.
Porteiro, J., Calado, H., Pereira, M. and Dentinho, T. (2005) Integrated Management on the Sete Cidades
WaterShed, 14th European Colloquium on Theoretical and Quantitative Geography, Tomar
(Portugal).
Poyatos, R., Latron, J., and Llorens, P. (2003): Land-Use and Land-Cover change after agricultural
abandonment, The case of a Mediterranean Mountain Area (Catalan Pre-Pyrenees), Mt. Res. Dev.,
23(4): 52–58.
Pueyo, Y. and Beguería, S. (2007) Modelling the rate of secondary succession after farmland
abandonment in a Mediterranean mountain area. Landscape and Urban Planning, 83(4): 245-254.
Pugnaire, F. I., Luque, M. T., Armas, C. and Gutié L. (2006) Colonization processes in semi-arid
Mediterranean old-fields. Journal of Arid Environments, 65: 591–603.
Reilly, W. J. (1931) The Law of Retail Gravitation. New York: W. J. Reilly.
Roy, J. R. (2004) Spatial Interaction Modelling, A Regional Science Context, Advances in Spatial Science.
Springer.
Roy, J. R. and Thill, J. C. (2004) Spatial interaction modeling. Papers in Regional Science, 83: 339-361.
Running, S. W., Loveland, T. R., Pierce, L. L., Nemani, R. and Hunt, E. R. (1995) A remote sensing based
vegetation classification logic for global and cover analysis. Remote Sensing of Environment, 51:
39−48.
Saltelli A, Ratto M, Andres T, Campolongo F, Cariboni J, Gatelli D, Saisana M, Tarantola S (2008) Global
Sensitivity Analysis: The Primer. John Wiley & Sons.
Schneider, M. (1959) Gravity Models and Trip Distribution Theory. Papers and Proceedings of the
Regional Science Association, 5: 51-56.
Smith, T. E. (1975) A Choice Theory of Spatial Interaction. Regional Science and Urban Economics, 5:
137-176.
80
Spangenberg JH (2007) Integrated scenarios for assessing biodiversity risks. Sustainable Development
15:343–356.
Soares-Filho, B. S., Cerqueira, G. C. and Pennachin, C. L. (2002) DINAMICA - a stochastic cellular
automata model designed to simulate the landscape dynamics in an Amazonian colonization
frontier, Ecological Modelling, 154: 217–235.
Tasser, E., Teutsch, A., Noggler, W., Tappeiner, U. (2007) Land-use changes and natural reforestation in
the Eastern Central Alps. Agriculture, Ecosystems and Environment, 118: 115–129.
Therrien, R., McLaren, R.G., Sudicky, E., Panday, S.M., 2010. HydroGeoSphere A Three-dimensional
Numerical Model Describing Fully-integrated Subsurface and Surface Flow and Solute Transport
(draft).
University
of
Waterloo,
pp.
457.
Available
Online:
http://hydrogeosphere.org/hydrosphere.pdf (20.10.2011).
Thornton, P. K. and Jones, P. G. (1998) A conceptual approach to dynamic agricultural landuse modeling.
Agricultural Systems, 57(4): 505–521.
Tzanopoulos J., Mitchley J. and Pantis J. D. (2007) Vegetation dynamics in abandoned crop fields on a
Mediterranean island: development of succession model and estimation of disturbance
thresholds. Agric. Ecosyst. Environ., 120: 370–376.
Van Meijl, H., van Rheenen, T., Tabeau, A. and Eickhout, B. (2006). The impact of different policy
environments on agricultural land use in Europe. Agriculture, Ecosystems and Environment, 114:
21-38.
Verburg, P. H., de Konig, G. H. J., Kok, K., Veldkamp, A., Fresco, L. O. and Bouma, J. (1997) Quantifying
the Spatial Structure of Land Use Change: An Integrated Approach. Conference on GeoInformation for Sustainable Land Management, Wageningen University, The Netherlands.
Verburg, P. H., Soepboer, W., Limpiada, R., Espaldon, M. V. O., Sharifa, M. A. and Veldkamp, A. (2002)
Modelling the spatial dynamics of regional land use: The CLUE-S model. Environmental
Management, 30: 391-405.
Verburg, P. H. and Veldkamp, A. (2004) Projecting land use transitions at forest fringes in the Philippines
at two spatial scales. Landscape Ecology, 19 (1): 77-98.
Verburg, P. H., Schot, P., Dijst, M. and Veldkamp, A. (2004) Land use change modelling: current practice
and reseach priorities. GeoJournal, 61: 309–324.
Verburg, P. H., Schulp, C. J. E., Witte, N. and Veldkamp, A. (2006) Downscaling of land use change
scenarios to assess the dynamics of European landscapes, Agriculture, Ecosystems and
Environment, 114: 39-56.
81
Verburg, P. H., Eickhout, B. and van Meijl, H. (2008) A multi-scale, multi-model approach for analyzing
the future dynamics of European land use. Annals of Regional Science, 42: 57-77.
Verburg, P. H. and Overmars, K. P. (2009) Combining top-down and bottom-up dynamics in land use
modeling: exploring the future of abandoned farmlands in Europe with the Dyna-CLUE model.
Landscape Ecology, 24(9): 1167-1181. http://dx.doi.org/10.1007/s10980-009-9355-7
Verburg, P. H., Berkel, D. v., Doorn, A. v., Eupen, M. v. and Heiligenberg, H. A. R. M. v. (2010) Trajectories
of land use change in Europe: a model-based exploration of rural futures. Landscape Ecology,
25(2): 217-232.
Wageningen University and Research Centre and the Netherlands Environmental Assessment Agency
(WUR/MNP). (2008) EURURALIS2.0 CdROM. Wageningen University and Research Centre and the
Netherlands Environmental Assessment Agency, Wageningen, The Netherlands.
Wegener, M. (1994) Operational Urban Models, State of the Art. Journal of American Planning
Association. Winter pp. 17-29.
Westhoek, H. J., Van den Berg, M. and Bakkes, J. A. (2006) Scenario development to explore the future
of Europe’s rural areas. Agriculture, Ecosystems and Environment, 114: 7-20.
White R. and Engelen G. (1994) Cellular Dynamics and GIS: Modelling Spatial Complexity, Geographical
Systems, 1: 237-253.
White, R. and Engelen, G. (1993) Cellular Automata and Fractal Urban Form: A Cellular Modeling
Approach to the Evolution of Urban Land Use Patterns. Environment and Planning A, 25(8): 11751199.
Wilson, A. G. (1967) A statistical theory of spatial distribution models, Transportation Research, 1: 253269.
82