Download - A qualitative method for the spatial and thematic downscaling of land-use change scenarios
A qualitative method for the spatial and thematicdownscaling of land-use change scenarios
Sophie Rickebusch a,*, Marc J. Metzger a, Guangcai Xu a,b, Ioannis N. Vogiatzakis c,Simon G. Potts c, Maria Teresa Stirpe c, Mark D.A. Rounsevell a
aCentre for the Study of Environmental Change and Sustainability (CECS), School of GeoSciences, The University of Edinburgh,
Drummond Street, Edinburgh EH8 9XP, UKbCollege of Resources Science and Technology, Beijing Normal University, Beijing 100875, ChinacCentre for Agri-Environmental Research, School of Agriculture, Policy and Development, University of Reading, RG6 6AR, UK
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8
a r t i c l e i n f o
Published on line 8 December 2010
Keywords:
Rule-based approach
Bio-energy crops
Land abandonment
Local knowledge
United Kingdom
a b s t r a c t
Assessing the potential impact of future land-cover changes on habitat quality requires
projections with a fine spatial and thematic resolution. The former is usually addressed by
downscaling methods, often at the expense of the latter. We present a new, rule-based
method to downscale land-use change scenarios to the landscape level while keeping a large
number of land-cover classes (CORINE level 3). The method relies on the interpretation of
European scenario storylines, the observation of past land-use changes, high-resolution
regional data and expert knowledge.
The results give a landscape-level transposition of the scenario storylines which reflects
the local conditions. The method has a number of advantages, such as its potential
application in dialogues with policy-makers and stakeholders. Possible further develop-
ments include automating the rule-based selection to overcome the current limitations of
this method in terms of spatial extent.
# 2010 Elsevier Ltd. All rights reserved.
avai lable at www.sc iencedi rec t .com
journal homepage: www.elsevier.com/locate/envsci
1. Introduction
Land-use/-cover change has increased dramatically in recent
decades due to rapid demographic and societal changes; this is
expected to continue in many European regions (Busch, 2006;
Lambin et al., 2001). Agricultural and urban lands are
particularly susceptible to these changes and affect habitat
quality for both flora and fauna. Increased urbanisation and
peri-urban settlement results in fragmentation, which affects
various ecosystem processes (Lambin et al., 2001). Assessing
the impact of land-cover changes on ecosystem processes or
habitat quality requires data with a fine spatial and thematic
resolution (Conway, 2009). Spatial resolution is necessary to
calculate the degree of fragmentation (patch size, connectivi-
ty) precisely (Fahrig, 2003; Reidsma et al., 2006) and to preserve
* Corresponding author. Tel.: +44 131 651 4449; fax: +44 131 650 2524.E-mail address: [email protected] (S. Rickebusch).
1462-9011/$ – see front matter # 2010 Elsevier Ltd. All rights reservedoi:10.1016/j.envsci.2010.11.003
the representation of land-use types with typically small
patches, such as grasslands, which may be lost at coarser
resolutions (Schmit et al., 2006). Using a finer resolution also
reduces the errors inherent to mapping methods, such as
representing continuous data in a raster (grid) format
(Dendoncker et al., 2008). Without sufficient thematic resolu-
tion it is impossible to assess habitat quality correctly:
‘‘agricultural land’’ may include very different land-cover
types, such as highly intensive monocultures or low intensity,
species-rich pastures (de Chazal and Rounsevell, 2009). This
level of precision is often available for recent land-cover
changes (post-1990), for instance in the CORINE land-cover
(CLC) datasets (Buttner et al., 2002). Even the finest CLC
classification (level 3, 44 classes) is arguably not sufficient for
some ecological applications, especially as the more ‘‘natural’’
d.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8 269
land-cover types, such as forest or shrub, are not very well
differentiated in CLC (Neumann et al., 2007; Waser and
Schwarz, 2006). However, high thematic resolution is less
common for older changes, which makes the calibration of
land-use/-cover change models very difficult. One exception is
the BIOPRESS transect data, which details CLC level 3 land-
cover in 1950, 1990 and 2000 (Gerard et al., 2010; Kohler et al.,
2006; Thomson et al., 2007).
Scenarios providing alternative images of how the future
may unfold are a popular tool for exploring alternative
plausible land-use futures (Ewert et al., 2005; Rounsevell et al.,
2005, 2006; Rounsevell and Metzger, 2010; Verburg et al., 2006).
Rounsevell et al. (2006) describe how the Intergovernmental
Panel on Climate Change (IPCC) Special Report on Emissions
Scenarios (SRES) (Nakicenovic and Swart, 2000) can be
interpreted to derive European land-use change scenarios.
The same methodology was used to develop the ALARM
scenarios (Settele et al., 2005), which we use in this study. At
the European level, the ALARM scenarios show a continuation
of current trends in land-use change, but with differing
magnitudes between the scenarios (Reginster et al., 2010;
Spangenberg et al., in press). Unfortunately, the 100 spatial
resolution (approximately 16 km � 16 km) of these projec-
tions is too coarse for studies relating to biodiversity or
ecosystem services, for instance. A number of quantitative
downscaling methods have been used to refine the spatial
resolution of land-use change projections (Dendoncker et al.,
2006, 2008; Verburg et al., 2008, 2006). However, automated
quantitative downscaling has limitations, such as the
inability to adjust to the local context and the lack of control
over the results. With fewer than 10 land-use categories, the
thematic resolution of many current land-use scenarios also
remains too coarse. Some studies use a finer thematic
resolution in their particular field of interest, e.g. agriculture,
but group all other land-uses into a single ‘‘other’’ category
(Chakir, 2009). As observed by Conway (2009), increasing the
number of classes in land-use change models may in some
cases be detrimental to model accuracy, especially where less
frequent types of conversions are concerned. This may
explain why so little attention has been paid to improving
thematic resolution, despite its importance for ecological and
other applications.
For planners and policy-makers, it is essential to under-
stand how land-use change might affect biodiversity and
other ecosystem services, so that these services can be
preserved and potential synergies or conflicts identified
(Verburg et al., 2009). In turn, planners and policy-makers
can also inform research by bringing local knowledge into the
interpretation of scenario storylines for their region. This
interaction is often overlooked when developing models and
decision-making tools for management purposes (Jakeman
et al., 2006).
We describe here a rule-based downscaling method which
provides plausible projections of land-cover change at the
landscape level, with fine spatial and, more importantly,
thematic resolution. By using a qualitative method, we
circumvent the main problem with calibrating models for
rare events, namely insufficient data. As it relies on visualising
the data, this method enables us to check during the
downscaling process that the results are representative of
local conditions and has the potential to be used for policy
dialogue. Finally, working at the local level allows the use of
detailed data and ancillary knowledge, which are often not
available for large-scale data.
2. Materials and methods
Projected land-use changes from the pan-European ALARM
scenarios (Reginster et al., 2010; Spangenberg, 2007; Spangen-
berg et al., in press) were thematically disaggregated and
spatially allocated in a three-step process. Firstly, we derived
the relative changes in four aggregated land-use categories
from the ALARM scenarios and converted them to absolute
increases or decreases in number of 100 m � 100 m (1 ha) grid
cells within seven transects from the BIOPRESS project (Gerard
et al., 2010; Kohler et al., 2006; Thomson et al., 2007). Secondly,
we disaggregated these changes to the more detailed CLC
categories used in the transects, to obtain the number of 1 ha
grid cells by which each land-cover category will increase or
decrease under each scenario. Finally, we allocated these
changes spatially within the transect using a combination of
scenario-specific rules, regional information and expert
judgement (by the author team at this stage). These three
steps are detailed in Sections 2.2.–2.4.
2.1. Land-cover data and scenarios
We used land-cover data from seven transects in the United
Kingdom (Fig. 1), taken from the BIOPRESS project (Gerard
et al., 2010; Kohler et al., 2006; Thomson et al., 2007). The
transects in that project were chosen to intersect with Natura
2000 sites and are 15 km � 2 km rectangles, though some have
missing sections due to lack of coverage by the aerial
photographs from which the data is derived. The data follows
CLC level 3 (44 classes). Each transect consists of a mosaic of
polygons representing contiguous areas with a common land-
cover history.
For our purpose, the polygon data for each transect was
converted to 100 m (1 ha) grids, one per time step (1950, 1990,
2000). These were then converted to point data, with the points
in the centres of the grid cells. The associated attribute table
contained the land-cover classes for the three past time steps,
to which we added three fields for 2030, one per scenario.
These were initially filled with the land-cover class for 2000.
The ALARM land-use change scenarios (Reginster et al.,
2010; Spangenberg, 2007; Spangenberg et al., in press) give
projections of the percentage cover per 100 grid cell in 2030 for
seven land-use categories, following the three storylines in
Table 1. The land-use categories are: urban, cropland,
grassland, permanent crops, biofuels, forests and land in
succession (i.e. abandoned agricultural land). The construc-
tion of the ALARM scenarios involved three steps: (1)
construction of three alternative scenario storylines (Table
1), (2) estimation of the aggregate totals of land-use change
using a supply/demand model and (3) spatial allocation of
these aggregate quantities using spatially explicit rules
(Rounsevell et al., 2006). The methodology is described in
detail by Reginster et al. (2010) and Spangenberg et al. (in
press).
Fig. 1 – Location of the seven United Kingdom transects used and example of past land-cover changes (1950–1990–2000) in
BIOPRESS transect UK1 (Gerard et al., 2010; Kohler et al., 2006).
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8270
2.2. Change in aggregated land-use categories
To quantify the land-cover changes in the transects, we need
to know the relative changes compared to the current
situation. This is done for five aggregated categories (cropland,
forest, grassland, built-up areas and surplus land). As some
Table 1 – Summary of the three ALARM scenario storylines: GR2007; Spangenberg et al., in press).
In the GRAS (GRowth Applied Strategy) storyline, deregulation and free m
regions. The guidelines of the European planning policies (The Europe
patterns for new urbanisation are urban sprawl and diffuse peri-urban
the free market policy and regional subsidies and compensation withi
areas are only maintained in optimal locations with a comparative ad
areas are preserved, but the NATURA 2000 site network is not enforce
In the BAMBU (Business-As-Might-Be-Usual) storyline, the European plan
Perspective, ESDP) are applied. Compact city development and limits o
agricultural policy (CAP) is maintained, but adapted to avoid overprodu
comparative advantage and a minimum level of agriculture is preserv
current European afforestation policy is maintained. Current protected
In the SEDG (Sustainable European Development Goal) storyline, integra
extensification of agriculture and the encouragement of organic farmi
biodiversity by, for example, limiting nitrogen and pesticide use. Small
rural areas. Planning policies are strict and favour compact settlement
public transport and other energy savings. Current protected areas are
ALARM categories such as ‘‘bio-energy crops’’ and ‘‘surplus
land’’ are not currently present, we cannot calculate relative
change figures for these. For bio-energy crops, we solved this
by calculating the relative change for aggregated categories
(‘‘total crops’’ and ‘‘total forests’’, Table 2) which incorporate
the various bio-energy crop types. These relative changes were
AS, BAMBU and SEDG (Reginster et al., 2010; Spangenberg,
arket objectives lead to the reduction or abolition of zoning in
an Spatial Development Perspective, ESDP) are not applied and the
isation. Liberalisation of food trade is one of the consequences of
n the common agricultural policy (CAP) are abolished. Agricultural
vantage, where the agricultural rent is positive. Current protected
d.
ning policy guidelines (The European Spatial Development
n peri-urbanisation are implemented and enforced. The common
ction. Agricultural areas are maintained in optimal locations with a
ed in traditional landscape for rural development objectives. The
areas are preserved and the NATURA 2000 site network is enforced.
ted social, environmental and economic policies lead to the
ng. This reduces the effect of agricultural intensification on
decreases in agricultural areas help to reduce unemployment in
to reduce travel needs. This provides opportunities for efficient
preserved and the NATURA 2000 site network is enforced.
Table 2 – Correspondence of the aggregated land-covercategories with ALARM and CORINE. There is notnecessarily a direct correspondence between ALARM andCORINE, hence the need for aggregation.
Aggregated ALARM CORINE
Total crops Cropland permanent
crops liquid bio-energy
crops non-woody
bio-energy crops
Arable land (2.1.x)
heterogeneous
agricultural areas
(2.4.x) permanent
crop (2.2.x)
Total forest Forest woody
bio-energy crops
Forest (3.1.x)
Grasslands Grasslands Pastures (2.3.x)
natural grasslands
(3.2.1)
Built-up area Built-up area All artificial
surfaces (1.x.x)
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8 271
then used to calculate the absolute change in 1 ha grid cells per
transect, which was then disaggregated to CLC level 3 (plus the
new bio-energy categories).
Two other land-cover category groups can be discerned.
First, categories which are assumed to be uninfluenced by the
main land-use change drivers (Rounsevell et al., 2005). These
include bare rock, water bodies, glaciers and natural areas,
which were left out of this analysis as they do not change
under the ALARM scenario assumptions. Second, the ‘‘sur-
plus land’’ category, which consists mainly of abandoned
agricultural land. It is defined as land remaining after the
demand for urban uses, agriculture and forestry has been
met. It can therefore be calculated as the difference between
total land-use change and the sum of changes in all other
categories.
The ALARM scenarios were constructed using the relatively
coarse 1 km resolution PELCOM land-cover map, which can
deviate considerably from detailed regional land-use patterns
(Dendoncker et al., 2008; Schmit et al., 2006), especially for
land-uses with small spatial extents. Consequently, using
individual 100 ALARM grid cells to derive future change may
cause problems. For example, the ALARM grid cell might not
contain any urban land-cover, resulting in no change for that
category, while in reality there are small settlements in the
region. This problem can be overcome by calculating average
trends over a wider region. In order to develop a methodology
which could be rolled out across Europe, we chose to calculate
average regional trends for the Atlantic Central and Atlantic
North European environmental zones (Metzger et al., 2005).
For the United Kingdom, these distinguish between the
English lowlands (transects 1, 2, 5 and 6) and the uplands
plus Scotland (transects 3, 8 and 9; Fig. 1).
2.3. Disaggregation to CORINE level 3 categories
The relative proportions of bio-energy to conventional
cropland or forest in the ALARM scenarios were used to
disaggregate the ‘‘total crops’’ and ‘‘total forest’’ categories,
respectively. For example, if under a given scenario the ‘‘total
forest’’ category increased by 50 grid cells from 1500 to 1550
and the fraction of woody bio-energy crops increased from 0 to
10%, then conventional forestry would decrease by 105 cells to
1395 [=1550–(1550 � 0.1)] and woody bio-energy crops would
increase by 155 cells. In this way, it is possible to obtain
absolute changes for the three types of bio-energy crops as
well as arable land, permanent crops and forests.
Since some future land-use categories do not yet exist in
the CLC classification, we created a new level 2 class (2.5) for
bio-energy crops, which is split into three level 3 classes: liquid
bio-energy crops (2.5.1; e.g. oilseed rape), non-woody bio-
energy crops (2.5.2; e.g. Miscanthus grass) and woody bio-
energy crops (2.5.3; e.g. willow). Additionally, we created two
categories for surplus land (abandoned agricultural land)
within the relevant level 2 groups: abandoned arable land
(2.1.4) and abandoned pastures (2.3.2).
2.4. Rule-based spatial allocation
Within each transect, we re-allocated the necessary number of
1 ha grid cells of each type to match the downscaled ALARM
values for each scenario, following a set of rules. Technically,
this was done using GIS (Geographical Information System)
software and working on the point files created from the
transect grids. The points to be changed were selected
manually and a new land-cover category assigned to the
corresponding field in the attribute table. This procedure is
illustrated in Fig. 2, with more details available in Appendix A
(online supplementary material). Once all the changes had
been made, following the allocation rules below, we created
grids of the projected land-cover for 2030 according to each
scenario.
The allocation rules come primarily from interpreting the
ALARM scenario storylines. Additionally, we also took into
account past trends in the BIOPRESS transects, e.g. most
frequent conversions between land-cover classes, polygons
previously subjected to change. Finally, we added expert and
local knowledge (from within the author team), e.g. require-
ments for a given crop type, presence of small landscape
features within other land-use types (heterogeneity). The
changes followed this order of precedence: (1) protected areas,
(2) urban areas, (3) agricultural land, (4) forests and (5)
abandoned land. An example of the detailed allocation rules
for transect UK1 is given in Table 3 and the rules for all
transects and scenarios are provided in Appendix B (online
supplementary material). In general terms, protected areas
were preserved (with an additional 100-m buffer zone in the
SEDG scenario), so no changes were allowed. Urban areas
expanded mostly into agricultural land-use types, with
locations varying according to the scenarios. Agricultural land
tended to decrease with, aside from the conversion to urban
land, some conversion to biofuels (woody biofuels could also
be grown on land converted from forest) and some to forest.
Surplus land (mostly from agriculture) was considered to be
abandoned. Cells were preferably changed in clusters corre-
sponding to the original BIOPRESS polygons, i.e. areas which
had a common history of change. This common history is
likely to be the result of other factors, such as land-ownership,
which also affect future changes.
The results are presented as maps for the various
scenarios. For transect UK1, we also plotted the total area
per CORINE level 3 class for the six most abundant classes in
2000 (i.e. in the baseline data).
LUC allocation
1. Identify protected areas & 100 m buffer zones
BAMBU
No change within protected
area
As GRAS + avoid allocating urban
areas & intensive agriculture next
to protected areas
As BAMBU + extend
protection to 100 m buffer
zone
GRAS SEDG
landAllocate urban
BAMBU
2.
Create new settlements or
expand small ones by converting
(extensive) arable land
Expand small
settlements by
converting arable land
Expand or fill in large
settlements by converting
(intensive) arable land
SEDGGRAS
3. Allocate agricultural land (arable, liquid/non-woody bio-fuels & grass)
Identify unfavourable land, e.g. steep terrain, river banks, & historic trends
BAMBU
Convert arable land &
grassland to liquid & non-
woody bio-fuels, preferably
near industrial zones
Convert arable land to liquid
& non-woody bio-fuels,
preferably near industrial
zones
Convert (intensive) arable
land to liquid & non-woody
bio-fuels, preferably near
industrial zones
GRAS SEDG
bio-fuels)woody(incl.forestsAllocate4.
BAMBU
Convert forest & grassland
to woody bio-fuels
Convert grassland to woody
bio-fuels & forest
Convert arable & grassland
to woody bio-fuels & forest,
expand protected forests
SEDGGRAS
5. Allocate abandoned land (former arable & grassland)
BAMBU
Convert remaining arable &
grassland to abandoned land
Convert remaining arable &
grassland to abandoned land,
esp. around protected areas
Convert remaining arable &
grassland to abandoned land,
esp. around protected areas
GRAS SEDG
End
Fig. 2 – Flow chart illustrating the steps in the land-use change allocation method. Blue, dotted arrows indicate surplus land
which is re-allocated outside the initial category. (For interpretation of the references to colour in this figure legend, the
reader is referred to the web version of the article.)
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8272
3. Results
Fig. 3 shows an example of land-cover projections for transect
UK1 (Kennet Valley Alderwoods) according to the three
ALARM scenarios. Urban areas (categories 1.1.2–1.4.2) increase
by similar amounts in all scenarios (24–27 ha), but locations
are scattered with GRAS, whereas with SEDG they are close to
existing urban centres. The GRAS scenario has the lowest
conversion to biofuel crops (cat. 2.5.1–2.5.3; 120 ha) and the
most abandonment of pasture land (cat. 2.3.2; �302 ha), while
the SEDG scenario is the opposite (235 ha biofuels, �91 ha
pastures). Results for the BAMBU scenario are generally
somewhere in between, though often closer to GRAS scenario
results. Abandoned land is mostly situated along the Kennet
River, which runs the length of the transect. These areas were
considered more difficult to access, making them less
profitable for agriculture. The historical data (Fig. 1), which
shows some conversion from pasture to forest along the
Table 3 – Example of land-cover allocation rules and local decisions (in italics) for the three ALARM scenarios (GRAS,BAMBU, SEDG) in transect UK1. The results are shown in Fig. 3.
GRAS BAMBU SEDG
1. Protected Current protected
areas maintained
Current protected
areas maintained
Current protected
areas maintained
Expand protection
with 100 m buffer
2. Urban Dispersed urban
sprawl, new urban
centres can appear
Compact city
development and
peri-urbanisation
around existing
smaller centres
Compact city
development and
peri-urbanisation
around existing cities
Small new settlements,
some expansion of
existing small town;
existing industrial sites
expand, new site near
small town
Urban expansion on
agricultural land around
existing towns; new
industrial sites on edges
of towns; bypass extended
to industrial estate; urban
green in arable field in
town; new leisure facility
next to village
Urban expansion on
agricultural land around
existing towns; new
industrial sites on edges
of towns; bypass extended
to industrial estate; urban
green in arable field in town;
new leisure facility next
to village
3. Agriculture Agriculture shifts to
most favourable
areas
Limited shifts in
agriculture to most
favourable areas
Extensification—little
change in agriculture
distribution
First allocate crops, then
bio-energy crops, then
grassland; less intensive
agriculture categories
disappear first
First allocate crops, then
bio-energy crops, then
grassland; less intensive
agriculture categories
reduced by 50%; liquid
biofuels located on arable
land near towns/industry
First allocate crops, then
bio-energy crops, then
grassland; less intensive
agriculture categories are
maintained; liquid biofuels
located on arable land near
towns/industry
4. Forest General support for
afforestation
General support for
afforestation
General support for
afforestation
Existing patches expand
or contract around the
periphery. Some conversion
of woodland to woody
bio-energy crops; bio-energy
crops also expand on wet
grasslands and arable land
along river
Existing patches expand or
contract around the periphery;
expansion of smaller patches;
woody biofuels on grasslands
and arable land along river
Existing patches expand or
contract around the periphery;
expansion of smaller patches;
woody biofuels on grasslands
and arable land along river
5. Abandoned Abandoned land goes
into succession
Abandoned land goes into
succession
Abandoned land goes into
succession
Abandoned grasslands
mainly along the river
(too wet or remote) and
around nature reserve
Abandoned grasslands mainly
along the river (too wet or remote)
Very limited abandonment of
grasslands along the river
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8 273
Kennet River between 1950 and 1990, supports this reasoning.
Details of the surfaces gained/lost per category and maps for
the all transects can be found in Appendix B.
The evolution over time of the total area occupied by the
main land-cover categories in transect UK1 (Fig. 4) shows
the largest differences between scenarios for the three
categories of agricultural land. For non-irrigated arable land
(2.1.1), GRAS shows a slight increase, similar to the 1950–
1990 period, whereas BAMBU shows a stabilising situation
and SEDG a continuation of the 1990–2000 trend. For
pastures (2.3.1), GRAS again shows a similar trend to
1950–1990 (marked decrease), while SEDG results in a much
smaller decrease. For agricultural land with natural areas
(2.4.3), SEDG shows a continuation of the 1990–2000 trend
(slight increase), while GRAS results in the disappearance of
that land-cover category. Results for BAMBU are intermedi-
ate between the other two scenarios, but often closer to
GRAS. For urban fabric (1.1.2) and forests (3.1.1, 3.1.3), the
future changes are much smaller than in the past 50 years
and there is little difference between scenarios in terms of
total area.
4. Discussion
Our results are plausible according to scenario terminology,
i.e. they may be radical but are ultimately possible (Rounsevell
and Metzger, 2010), in the opinion of the authors. In fact, the
projected changes are no more radical than those observed in
the past (Fig. 4). These results raise general policy questions,
Fig. 3 – Transect UK1 (Kennet Valley Alderwoods). Current (2000) land-cover and projections for 2030 according to three
scenarios. The categories correspond to CORINE land-cover level 3 (1.x.x artificial surfaces, 2.x.x agricultural areas, 3.x.x
forest and semi-natural areas, 5.x.x water bodies; see Appendix B for full nomenclature), except for the new categories:
2.1.4 (abandoned arable land), 2.3.2 (abandoned pastures), 2.5.1 (liquid biofuels), 2.5.2 (non-woody biofuels) and 2.5.3
(woody biofuels). The black lines are the perimeters of the protected areas.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8274
such as the impact of a biofuel-oriented ‘‘sustainable’’
scenario. However, this work is mainly policy-relevant
through the method’s potential for exploring local-level
policies by means of stakeholder dialogue.
4.1. Land-use projections
In our example transect (UK1-Kennet Valley Alderwoods),
there is little difference between the scenarios in the quantity
of new urban land-use by 2030. However, in accordance with
the scenario storylines, the new urban areas vary in location
from scattered new settlements (GRAS) to compact city
development (SEDG). The land-use categories which show
the most differences between scenarios are arable and
grassland categories (2.x.x). These also happen to be the most
interesting categories in terms of biodiversity and ecosystem
services, as agricultural land-uses are more heterogeneous in
this respect than urban or forest types.
The GRAS scenario shows a small increase in intensively
cultivated land (2.1.1) and some conversion to biofuel
production (mostly liquid biofuels, often intensive monocul-
tures). Simultaneously, extensive agricultural land-covers (e.g.
2.4.3) decrease dramatically. Pastures (2.3.1) also decrease
quite strongly in the GRAS projection, as a lot of pasture-land
is abandoned. This will eventually revert to (semi-)natural
vegetation types, such as scrub or forest, which will favour
some species (e.g. forest dwellers) but be detrimental to others
(Falcucci et al., 2007). This scenario therefore shows two
contrary trends: intensification of agriculture and a potential
increase in natural vegetation. Consequently, the land-cover
type most likely to decrease drastically in the long term is
open, semi-natural land, such as pastures with low grazing
intensities.
The high uptake of biofuel production inherent to the SEDG
scenario (Reginster et al., 2010; Spangenberg et al., in press)
raises an interesting paradox in terms of natural habitat
quality. Although this scenario is supposedly more ‘‘sustain-
able’’, large areas of biofuel monocultures may in fact be less
favourable for biodiversity, either directly through land-use
conversion or indirectly through invasive species than the
abandoned agricultural land and pastures found in the GRAS
scenario (Barney and DiTomaso, 2010). However, some biofuel
types such as short rotation forestry stands seem more
biodiversity-friendly than others (Giordano and Meriggi, 2009;
Rowe et al., 2009). In general though, this scenario does show a
shift from intensive to more extensive agriculture, which is
associated with valued semi-natural habitats, favourable for
many species, of plants and birds in particular (Falcucci et al.,
2007; MacDonald et al., 2000; Uematsu et al., 2010; Wolff et al.,
2001).
Fig. 4 – Area change (ha) in transect UK1 between 1950 and 2030 for six of the most common CORINE level 3 categories.
Scenarios GRAS, BAMBU and SEDG are represented by full, dashed and dotted lines, respectively.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8 275
The BAMBU scenario gave intermediate results between
the two extremes of GRAS and SEDG for both urban and
agricultural land-uses. It is interesting to note that the SEDG
scenario results, for agriculture at least, seem closer to a
continuation of recent trends (1990–2000) than BAMBU, which
is supposed to represent ‘‘business as usual’’. This may be a
result of more environmentally conscious ‘‘SEDG-like’’ poli-
cies in the more recent time step. On the other hand, the GRAS
scenario gives similar trends to the earlier time-step (1950–
1990). BAMBU, which gives intermediate results, therefore
shows a continuation of long-term trends.
The small changes observed in urban and forest land-use
types for all scenarios come from the ALARM scenarios and
the quantities of change in the large-scale ALARM projections.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8276
If these changes are small, the downscaled ones will be small
too. For discontinuous urban fabric (1.1.2, Fig. 4), this small
change represents a continuation of the levelling-off of growth
which already started in the 1990–2000 period. For broad-
leaved forests (3.1.1), previous changes have been more
variable (slow increase between 1950 and 1990, followed by
a sharp decrease), so the absence of change in the scenarios
may simply reflect the absence of a clear trend.
4.2. New downscaling method
Downscaling of land-use change projections often addresses
the spatial resolution and ignores the thematic disaggre-
gation, which makes the results unsuitable for many ecologi-
cal purposes. However, using a finer thematic resolution can
be difficult, as it will highlight any disparities in the original
data from different time steps or countries (e.g. CORINE vs.
PELCOM classification) (Schmit et al., 2006). This is generally
overcome by grouping the land-cover types into broader
categories, at the expense of the thematic resolution. In
contrast, the downscaling and allocation method described
here gives plausible, scenario-specific projections when a finer
resolution is necessary. Another advantage of this method is
that it incorporates an element of informed, but also partly
arbitrary, decision-making. This mimics local decision-mak-
ing by planners, farmers and other agents, such as switching
from food to biofuel crop production. This is partly driven by
market rules and partly by less predictable events, such as
land-ownership changing through succession or local atti-
tudes (Geoghegan et al., 2001; Liverman and Roman Cuesta,
2008). This method therefore results in projections which are
more representative of real life in terms of patterns, though
exact locations may vary. It has greater flexibility than large-
scale statistical downscaling and allocation, which cannot
take into account some human aspects of decision-making or
local scale socio-economic factors, which may affect land-
cover patterns significantly (Gellrich et al., 2007; Gellrich and
Zimmermann, 2007; Rickebusch et al., 2007). It can also use
ancillary knowledge which is only available at the local scale,
though we did not include any external experts in the present
study, which focussed on method development. Local expert
knowledge can prove a useful complement to large-scale
scientific knowledge in understanding land-cover patterns
and changes (Chalmers and Fabricius, 2007). Finally, the user
can check that the results reflect local conditions throughout
the process.
While the downscaling method was developed as an
academic desk study, it could be used as a participatory
planning tool. Land-use scenarios act as a stimulus for critical
thinking, providing a heuristic and rhetorical guide in the
planning process, both for organizational learning and for
option searching (Xiang and Clarke, 2003). While the thematic
downscaling of existing scenarios provides boundary condi-
tions for change, regional stakeholders can help interpret the
impacts of these changes for the region, thus helping planners
to minimise potential conflicts and maximise opportunities.
For example, if a scenario indicates a large increase in bio-
energy production or forest expansion, local expertise can be
used to identify the most likely regions of conversion based on
suitability and assess potential consequences for biodiversity
targets. This could lead to additional protection policies for
biodiversity or incentives for stimulating conversion in parts
of the region. In this way the method presented here allows for
strategic conversations (Van der Heijden, 2005) about possible
future changes to be framed in a credible scenario framework
and to be graphically illustrated and communicated in maps.
From a methodological point of view, different stake-
holders’ perspectives could also provide a basis for assigning
probabilities of change for each grid cell, using the level of
agreement between stakeholders as a proxy. Furthermore, it
would be possible to explore the differences in preference
between stakeholder groups (e.g. planners and land-owners)
or to get them to negotiate the most optimal compromise.
Additionally, stakeholder dialogue could be used to refine the
allocation rules (Table 3 and Appendix B) and give them more
transparency. Both sides benefit from this type of approach, as
decision-makers are then more likely to find the results
credible and relevant, therefore useful (Cash et al., 2003;
Olsson and Andersson, 2007).
Every land-cover change analysis method has its limita-
tions and is more suited to specific uses (Lambin, 1997).
Despite its relatively fine thematic resolution, CLC remains
unable to detect very small, key ecological features. This is
why it is important to add local expert knowledge when
applying this method in future, for instance to point out
which agricultural areas within the landscape contain the
most patches of other habitats, which might influence the
allocation. The method presented here could also be used to
apply scenarios to a smaller field study site, if a finer land-
cover map were available. This method is definitely most
suitable for small spatial extents, up to the landscape level,
as it would be unpractical to select and change manually
more than a few hundred points. It is therefore not easily
applicable as such to areas much larger than the transects
used here (3000 grid cells). One possibility would be to
combine this with other downscaling methods to focus on
areas of particular interest. Alternatively, the rule-based
method could be automated, in part at least. This could
include selecting areas with a particular land-cover history
or certain spatial attributes, such as distance to urban areas.
Finally, it would be interesting to explore the possibilities of
machine learning, whereby a computer could automate the
rules based on the responses given by experts or stake-
holders.
5. Conclusions
The qualitative, rule-based method for downscaling land-use
change scenarios described here provides a fine spatial
resolution and a detailed thematic disaggregation for small
areas up to the landscape level. It gives results which reflect
local conditions more closely than automated statistical
methods and provides opportunities for direct stakeholder
involvement in the downscaling process. It is therefore
particularly relevant for stakeholders and policy-makers
working at the local level. To overcome the current limitations
in the size of the study region, it would be useful to automate
parts of the rule base, while maintaining the flexibility of the
approach.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8 277
Acknowledgements
This research was funded by the EU in the 6th framework
project ‘‘COCONUT - Understanding effects of land use
changes on ecosystems to halt loss of biodiversity’’ (SSPI-
CT-2006-044343). We would like to thank Dr. France Gerard
from the Centre for Ecology and Hydrology for providing the
BIOPRESS dataset. Our thanks also go to Prof. David Sugden
from the University of Edinburgh and two anonymous
reviewers for their helpful comments.
Appendix A. Supplementary data
Supplementary data associated with this article can
be found, in the online version, at doi:10.1016/j.envsci.
2010.11.003.
r e f e r e n c e s
Barney, J.N., DiTomaso, J.M., 2010. Bioclimatic predictions ofhabitat suitability for the biofuel switchgrass in NorthAmerica under current and future climate scenarios.Biomass Bioenergy 34, 124–133.
Busch, G., 2006. Future European agricultural landscapes—Whatcan we learn from existing quantitative land use scenariostudies? Agric. Ecosyst. Environ. 114, 121–140.
Buttner, G., Feranec, J., Jaffrain, G., 2002. Corine land coverupdate 2000—Technical guidelines, Technical Report 89.European Environment Agency, Copenhagen, p. 56.
Cash, D.W., Clark, W.C., Alcock, F., Dickson, N.M., Eckley, N.,Guston, D.H., Jager, J., Mitchell, R.B., 2003. Knowledgesystems for sustainable development. Proc. Natl. Acad. Sci.U.S.A. 100, 8086–8091.
Chakir, R., 2009. Spatial downscaling of agricultural land-usedata: an econometric approach using cross entropy. LandEcon. 85, 238–251.
Chalmers, N., Fabricius, C., 2007. Expert and generalist localknowledge about land-cover change on south Africa’s wildcoast: can local ecological knowledge add value to science?Ecol. Soc. 12.
Conway, T.M., 2009. The impact of class resolution in land usechange models. Comput. Environ. Urban Syst. 33, 269–277.
de Chazal, J., Rounsevell, M.D.A., 2009. Land-use and climatechange within assessments of biodiversity change: areview. Global Environ. Change-Hum. Policy Dimens. 19,306–315.
Dendoncker, N., Bogaert, P., Rounsevell, M., 2006. A statisticalmethod to downscale aggregated land use data andscenarios. J. Land Use Sci. 1, 63–82.
Dendoncker, N., Schmit, C., Rounsevell, M., 2008. Exploringspatial data uncertainties in land-use change scenarios. Int.J. Geogr. Inf. Sci. 22, 1013–1030.
Ewert, F., Rounsevell, M.D.A., Reginster, I., Metzger, M.J.,Leemans, R., 2005. Future scenarios of European agriculturalland use: I. Estimating changes in crop productivity. Agric.Ecosyst. Environ. 107, 101–116.
Fahrig, L., 2003. Effects of habitat fragmentation on biodiversity.Annu. Rev. Ecol. Evol. Syst. 34, 487–515.
Falcucci, A., Maiorano, L., Boitani, L., 2007. Changes in land-use/land-cover patterns in Italy and their implications forbiodiversity conservation. Landsc. Ecol. 22, 617–631.
Gellrich, M., Baur, P., Koch, B., Zimmermann, N.E., 2007.Agricultural land abandonment and natural forest re-growth in the Swiss mountains: a spatially expliciteconomic analysis. Agric. Ecosyst. Environ. 118, 93–108.
Gellrich, M., Zimmermann, N.E., 2007. Investigating theregional-scale pattern of agricultural land abandonment inthe Swiss mountains: a spatial statistical modellingapproach. Landsc. Urban Plann. 79, 65–76.
Geoghegan, J., Villar, S.C., Klepeis, P., Mendoza, P.M., Ogneva-Himmelberger, Y., Chowdhury, R.R., Turner, B.L., Vance, C.,2001. Modeling tropical deforestation in the southernYucatan peninsular region: comparing survey and satellitedata. Agric. Ecosyst. Environ. 85, 25–46.
Gerard, F., Petit, S., Smith, G., Thomson, A., Brown, N.,Manchester, S., Wadsworth, R., Bugar, G., Halada, L., Bezak,P., Boltiziar, M., De badts, E., Halabuk, A., Mojses, M.,Petrovic, F., Gregor, M., Hazeu, G., Mucher, C.A., Wachowicz,M., Huitu, H., Tuominen, S., Kohler, R., Olschofsky, K., Ziese,H., Kolar, J., Sustera, J., Luque, S., Pino, J., Pons, X., Roda, F.,Roscher, M., Feranec, J., 2010. Land cover change in Europebetween 1950 and 2000 determined employing aerialphotography. Prog. Phys. Geogr. 34, 183–205.
Giordano, M., Meriggi, A., 2009. Use by small mammals of short-rotation plantations in relation to their structure andisolation. Hystrix 20, 127–135.
Jakeman, A.J., Letcher, R.A., Norton, J.P., 2006. Ten iterativesteps in development and evaluation of environmentalmodels. Environ. Modell. Softw. 21, 602–614.
Kohler, R., Olschofsky, K., Gerard, F., 2006. Land Cover Change inEurope from the 1950’ies to 2000: Aerial PhotoInterpretation and Derived Statistics from 59 SamplesDistributed Across Europe. World Forestry, Hamburg, p. 364.
Lambin, E.F., 1997. Modelling and monitoring land-cover changeprocesses in tropical regions. Prog. Phys. Geogr. 21, 375–393.
Lambin, E.F., Turner, B.L., Geist, H.J., Agbola, S.B., Angelsen, A.,Bruce, J.W., Coomes, O.T., Dirzo, R., Fischer, G., Folke, C.,George, P.S., Homewood, K., Imbernon, J., Leemans, R., Li,X., Moran, E.F., Mortimore, M., Ramakrishnan, P.S.,Richards, J.F., Skanes, H., Steffen, W., Stone, G.D., Svedin,U., Veldkamp, T.A., Vogel, C., Xu, J., 2001. The causes ofland-use and land-cover change: moving beyond the myths.Global Environ. Change-Hum. Policy Dimens. 11, 261–269.
Liverman, D.M., Roman Cuesta, R.M., 2008. Human interactionswith the Earth system: people and pixels revisited. EarthSurf. Process. Landf. 33, 1458–1471.
MacDonald, D., Crabtree, J.R., Wiesinger, G., Dax, T., Stamou, N.,Fleury, P., Lazpita, J.G., Gibon, A., 2000. Agriculturalabandonment in mountain areas of Europe: environmentalconsequences and policy response. J. Environ. Manage. 59,47–69.
Metzger, M.J., Bunce, R.G.H., Jongman, R.H.G., Mucher, C.A.,Watkins, J.W., 2005. A climatic stratification of theenvironment of Europe. Global Ecol. Biogeogr. 14, 549–563.
Nakicenovic, N., Swart, R., 2000. Emissions Scenarios: A SpecialReport of Working Group III of the Intergovernmental Panelon Climate Change. Cambridge University Press,Cambridge, UK, p. 570.
Neumann, K., Herold, M., Hartley, A., Schmullius, C., 2007.Comparative assessment of CORINE2000 and GLC2000:Spatial analysis of land cover data for Europe. Int. J. Appl.Earth Obs. Geoinf. 9, 425–437.
Olsson, J.A., Andersson, L., 2007. Possibilities and problems withthe use of models as a communication tool in waterresource management. Water Resour. Manage. 21, 97–110.
Reginster, I., Rounsevell, M., Butler, A., Dendoncker, N., 2010.Land use change scenarios for Europe. In: Settele, J., Penev,L.D., Georgiev, T.A., Grabaum, R., Grobelnik, V., Hammen,V., Klotz, S., Kotarac, M., Kuhn, I. (Eds.), Atlas ofBiodiversity Risk. Pensoft Publishers, Sofia/Moscow, p. 300.
Reidsma, P., Tekelenburg, T., van den Berg, M., Alkemade, R.,2006. Impacts of land-use change on biodiversity: anassessment of agricultural biodiversity in the EuropeanUnion. Agric. Ecosyst. Environ. 114, 86–102.
e n v i r o n m e n t a l s c i e n c e & p o l i c y 1 4 ( 2 0 1 1 ) 2 6 8 – 2 7 8278
Rickebusch, S., Gellrich, M., Lischke, H., Guisan, A., Zimmermann,N.E., 2007. Combining probabilistic land-use change and treepopulation dynamics modelling to simulate responses inmountain forests. Ecol. Model. 209, 157–168.
Rounsevell, M.D.A., Ewert, F., Reginster, I., Leemans, R., Carter,T.R., 2005. Future scenarios of European agricultural landuse II. Projecting changes in cropland and grassland. Agric.Ecosyst. Environ. 107, 117–135.
Rounsevell, M.D.A., Metzger, M.J., 2010. Developing qualitativescenario storylines for environmental change assessment.Wiley Interdis. Rev.: Climate Change 1, 606–619.
Rounsevell, M.D.A., Reginster, I., Araujo, M.B., Carter, T.R.,Dendoncker, N., Ewert, F., House, J.I., Kankaanpaa, S.,Leemans, R., Metzger, M.J., Schmit, C., Smith, P., Tuck, G.,2006. A coherent set of future land use change scenarios forEurope. Agric. Ecosyst. Environ. 114, 57–68.
Rowe, R.L., Street, N.R., Taylor, G., 2009. Identifying potentialenvironmental impacts of large-scale deployment ofdedicated bioenergy crops in the UK. Renew. Sust. EnergyRev. 13, 260–279.
Schmit, C., Rounsevell, M.D.A., La Jeunesse, I., 2006. Thelimitations of spatial land use data in environmentalanalysis. Environ. Sci. Policy 9, 174–188.
Settele, J., Hammen, V., Hulme, P., Karlson, U., Klotz, S., Kotarac,M., Kunin, W., Marion, G., O’Connor, M., Petanidou, T.,Peterson, K., Potts, S., Pritchard, H., Pysek, P., Rounsevell,M., Spangenberg, J., Steffan-Dewenter, I., Sykes, M., Vighi,M., Zobel, M., Kuhn, I., 2005. ALARM: Assessing LArge-scaleenvironmental Risks for biodiversity with tested Methods.Gaia 14, 69–72.
Spangenberg, J.H., 2007. Integrated scenarios for assessingbiodiversity risks. Sustain. Dev. 15, 343–356.
Spangenberg, J.H., Carter, T.R., Fronzek, S., Jaeger, J., Jylha, K.,Kuhn, I., Omann, I., Paul, A., Reginster, I., Rounsevell,M.D.A., Schweiger, O., Stocker, A., Sykes, M., Settele, J., inpress. Scenarios for investigating risks to biodiversity. Glob.Ecol. Biogeogr.
Thomson, A.G., Manchester, S.J., Swetnam, R.D., Smith, G.M.,Wadsworth, R.A., Petit, S., Gerard, F.F., 2007. The use ofdigital aerial photography and CORINE-derivedmethodology for monitoring recent and historic changes inland cover near UK Natura 2000 sites for the BIOPRESSproject. Int. J. Remote Sens. 28, 5397–5426.
Uematsu, Y., Koga, T., Mitsuhashi, H., Ushimaru, A., 2010.Abandonment and intensified use of agricultural landdecrease habitats of rare herbs in semi-natural grasslands.Agric. Ecosyst. Environ. 135, 304–309.
Van der Heijden, K., 2005. Scenarios: The Art of StrategicConversation, 2nd ed. John Wiley & Sons, Chichester, UK.
Verburg, P.H., Eickhout, B., van Meijl, H., 2008. A multi-scale,multi-model approach for analyzing the future dynamics ofEuropean land use. Ann. Reg. Sci. 42, 57–77.
Verburg, P.H., Schulp, C.J.E., Witte, N., Veldkamp, A., 2006.Downscaling of land use change scenarios to assess thedynamics of European landscapes. Agric. Ecosyst. Environ.114, 39–56.
Verburg, P.H., van de Steeg, J., Veldkamp, A., Willemen, L., 2009.From land cover change to land function dynamics: a major
challenge to improve land characterization. J. Environ.Manage. 90, 1327–1335.
Waser, L.T., Schwarz, M., 2006. Comparison of large-area landcover products with national forest inventories and CORINEland cover in the European Alps. Int. J. Appl. Earth Obs.Geoinf. 8, 196–207.
Wolff, A., Paul, J.-P., Martin, J.-L., Bretagnolle, V., 2001. Thebenefits of extensive agriculture to birds: the case of thelittle bustard. J. Appl. Ecol. 38, 963–975.
Xiang, W.-N., Clarke, K.C., 2003. The use of scenarios in land-useplanning. Environ. Plan. B: Plan. Des. 30, 885–909.
Sophie Rickebusch is a post-doctoral researcher at the Centre forthe study of Environmental Change and Sustainability (CECS),University of Edinburgh. An ecologist by training, her currentresearch concerns various aspects of land-use change modelling.
Marc Metzger is a senior research fellow at the Centre for the studyof Environmental Change and Sustainability (CECS), University ofEdinburgh. His research interests include global change foresightanalysis and the development of assessment techniques for globalchange impacts, across scales.
Guangcai Xu is a doctoral student majoring in land resourcemanagement at the College of Resources Science and Technology,Beijing Normal University. His current research interest is focusedon regional environmental change and human adaptation, partic-ularly in China’s northern dryland area.
Ioannis Vogiatzakis is a research fellow at the Centre for Agri-Environmental Research (CAER) at the University of Reading. He isa landscape ecologist and biogeographer with research interests invegetation and habitat modelling, particularly in Mediterraneanislands and mountain environments.
Simon G. Potts is a principal research fellow in the School ofAgriculture, Policy and Development at the University of Reading.His interests include developing evidence-based tools to underpinthe management of protected areas, agro-ecosystems and associ-ated ecosystem services.
Maria Teresa Stirpe is a research assistant at the Centre for Agri-Environmental Research (CAER) at the University of Reading. Herresearch interests include the effects of land cover change onspecies and habitat distribution.
Mark Rounsevell is a professor of Rural Economy and Sustainabil-ity at the University of Edinburgh and director of the University’sCentre for the study of Environmental Change and Sustainability(CECS). His research focuses on the effects of environmentalchange on rural and urban landscapes with particular emphasison land-use change modelling. Model applications include theexploration of alternative futures of climate and other environ-mental change drivers and the response of individuals and societyto these changes. He was a lead author to Working Group II of theIntergovernmental Panel on Climate Change (IPCC).