MT

Ma

b

c

a

ARRA

KESILMG

1

ae2shelv(stlor

1h

Ecological Indicators 24 (2013) 273–283

Contents lists available at SciVerse ScienceDirect

Ecological Indicators

journa l homepage: www.e lsev ier .com/ locate /eco l ind

apping landscape services: A case study in a multifunctional rural landscape inhe Netherlands

.M.C. Gulickxa,∗, P.H. Verburgb, J.J. Stoorvogela, K. Koka, A. Veldkampc

Soil Geography and Landscape Group, Wageningen University, PO Box 47, 6700 AA Wageningen, The NetherlandsInstitute for Environmental Studies (IVM), VU University Amsterdam, De Boelelaan 1087, 1081 HV Amsterdam, The NetherlandsFaculty of Geo-Information Science and Earth Observation (ITC), University of Twente, PO Box 6, 7500 AA Enschede, The Netherlands

r t i c l e i n f o

rticle history:eceived 28 December 2011eceived in revised form 29 June 2012ccepted 2 July 2012

eywords:cosystem servicespatial characteristicsndicatorsandscape functionsultifunctionalityIS

a b s t r a c t

The wide variety of landscape services, e.g. food production, water quality, and recreation, necessitatesthe use of a wide range of data sources for their identification. Subsequently, an array of approaches isrequired to analyse and map differ different landscape services, which we have explored in this study.Approaches to identify and map four landscape services are illustrated for the municipalities Deurne andAsten in province Noord-Brabant, The Netherlands: wetland habitat, forest recreation, land-based animalhusbandry, and recreation for hikers. The landscape services were identified through ground observa-tions at 389 locations. Spatial indicators were used to identify and map the landscape services. Basedon the ground observations, correlations between the landscape services and spatial characteristics (e.g.elevation, soil, road-type) were calculated within a neighbourhood with a radius of 0 m, 50 m, and 100 m.These correlations identified several site-specific indicators to map the landscape services. The accuracyof the landscape service maps created was assessed. The indicators proved to be adequately reliable for

forest recreation and reasonably reliable for land-based animal husbandry and recreation for hikers. Onlylandscape service map forest recreation was shown to be highly accurate. The four landscape servicesrarely coincide, but within a 1 km radius it is apparent that some occur closer together. The approachthat we have used is applicable for a wide range of different services and establishes a fundamental basisfor determining their spatial variation. As such, it should provide vital information for policy makers andspatial planners.

. Introduction

The importance of landscape services, provided by both naturalnd cultural landscapes, is increasingly recognised (e.g. Costanzat al., 1997; MA, 2005; de Groot, 2006; Termorshuizen and Opdam,009; Verburg et al., 2009). Landscapes are spatial social-ecologicalystems that deliver a wide range of functions, which are valued byumans in terms of economic, sociocultural, and ecological ben-fits (DeFries et al., 2004; Termorshuizen and Opdam, 2009). Aandscape service is defined here as ‘the goods and services pro-ided by a landscape to satisfy human needs, directly or indirectly’Termorshuizen and Opdam, 2009). We prefer the term landscapeervices over ecosystem services, as it infers pattern-process rela-ionships, unites scientific disciplines, and is better understood by

ocal practitioners (Termorshuizen and Opdam, 2009). Examplesf landscape services include food production, pollination, wateregulation, and provision of recreation.

∗ Corresponding author. Tel.: +31 317 482947; fax: +31 317 419000.E-mail address: monique.gulickx@wur.nl (M.M.C. Gulickx).

470-160X/$ – see front matter © 2012 Elsevier Ltd. All rights reserved.ttp://dx.doi.org/10.1016/j.ecolind.2012.07.005

© 2012 Elsevier Ltd. All rights reserved.

Increasing attention is paid, both by policy makers and scien-tists, to the multifunctionality (Fry, 2001; Holmes, 2006; Wilson,2008) and the potential synergies and conflicts that may arise.Policy makers and spatial planners are gradually directing theirpolicies and plans to provide and strengthen desired landscapeservices. To support the establishment of these policies and plans,geographical maps of existing and desired services are requiredto identify where services border each other or coincide and, thus,lead to possible synergies or conflicts. In this way, they may be usedto determine optimal solutions. Hence, it is necessary to developmethods and tools to quantify and map the different services acrossthe landscape.

The spatial distribution of intended landscape services that arerelated to the intended land use (e.g. food and fibre production)are often documented. However, the spatial distribution of land-scape services that are often an unintended consequence of landmanagement (e.g. provision of aesthetic beauty), are commonly

unknown. Additionally, they may be unrelated to a single land-cover or land-use type, which makes them more difficult to quantifyand map. It is postulated that landscape analyses based on land-cover and land-use are inadequate for landscape characterisation

2 ical Indicators 24 (2013) 273–283

ocHatisoecottlti

ratEsospae

asswsmsaoTldl

2

2

((opiPtloiatq

2

B

ing landscape services were identified using ground observations,sometimes complemented with information from governmentaldatabases or management strategies (Table 1). In addition, the spa-tial characteristics (Table 2) were assembled at a radius of 0, 50, and

74 M.M.C. Gulickx et al. / Ecolog

f such unintended services, since these approaches are specifi-ally related to the intended use of the land (Verburg et al., 2009).ence, common observation techniques, available land cover mapsnd spatial datasets, are insufficient for quantifying and mappinghese landscape services (Verburg et al., 2009). Consequently, var-ous spatial attributes, mainly biophysical, but also economic andocial, are used as indicators to quantify and map the spatial extentf landscape services (e.g. Gimona and van der Horst, 2007; Egoht al., 2008; Willemen et al., 2008; Kienast et al., 2009). Yet, indi-ators related to landscape services are often unknown or basedn general assumptions. Identifying suitable indicators is essen-ial for the improvement of landscape service maps. Therefore,he quantification of relations between site-specific attributes andandscape services are required in order to develop reliable indica-ors. Yet, site-specific indicators for landscape services are hardlynvestigated.

The vast array of landscape services is delivered across a greatange of temporal and spatial scales. Examples of services thatpply to different temporal scales are carbon sequestration (long-erm carbon storage) and seasonal recreation (short-term visits).xamples of services that apply to different spatial scales are waterupply (up to many km2) and cultural heritage, such as monumentsf architecture (as small as m2). Therefore, the development of atandard procedure to quantify and map landscape services is ham-ered by the fact that the appropriate spatial scales differs greatlymongst landscape services (de Groot and Hein, 2007; Pérez-Sobat al., 2008).

The objective of this study is to develop an approach to identifynd map various landscape services, by using indicators and con-idering spatial scales. Correlations between observed landscapeervices and spatial characteristics of the surrounding landscapeere analysed to ascertain site-specific indicators for landscape

ervices. These indicators were extrapolated into landscape serviceaps. The methodology and results are illustrated for four land-

cape services (i.e. wetland habitat, forest recreation, land-basednimal husbandry, and recreation for hikers) in the municipalitiesf Deurne and Asten, province of Noord-Brabant, The Netherlands.his case study aimed to obtain insights into the relations betweenandscape services and the surrounding landscape. The indicatorserived are specific to this area, but highlight linkages between

andscape services and their surroundings.

. Data and methods

.1. Study area

The study area comprised the municipalities of Deurne120 km2; 5 villages; 31.496 inhabitants; May 2009) and Asten72 km2; 3 villages; 16.398 inhabitants; May 2009) in the provincef Noord-Brabant, The Netherlands (Fig. 1). Both municipalities areart of De Peel region (approximately 600 km2), which is known for

ts intensive livestock production and nature reserve ‘De Grooteeel’ (peat-bog that has remained partly untouched by peat cut-ing). This area has to deal with various conflicting services in theandscape. For example, intensive animal husbandry has an impactn the environment, such as odour emission, which has a negativempact on recreation, such as farm camping. As a result, the nationalnd regional authority has assigned this region as a ‘reconstruc-ion area’ with high priority, in order to improve the environmentaluality of the rural area (Provincie Noord-Brabant, 2005).

.2. General design of methodology

At first, point observations of landscape services were made.ased on relations between the occurrence of landscape services

Fig. 1. Study area comprising municipalities Asten and Deurne. At the top on theright, the location of the study area (black mark) in The Netherlands is shown.

and the spatial characteristics of these locations, an extrapola-tion of these services to the whole study area was conducted. Themethodology consists of four components: (1) point observations oflandscape services; (2) point observations of spatial characteristics;(3) correlation analysis and selection of indicators; and (4) extrap-olation of indicators for mapping landscape services (Fig. 2). Thefour components are described in the paragraphs below. First, wedescribed the sampling method that was used to obtain point datafor the observation of landscape services and the spatial character-istics. The study area was divided into grid cells of 1 km2. Withineach grid cell, two points were selected approximately 500 m apart.This structured sample design provided an equal distribution ofdata points, resulting in a total of 389 points. Per data point, exist-

Fig. 2. Overview of the overall methodology.

M.M.C. Gulickx et al. / Ecological Indicators 24 (2013) 273–283 275

Table 1Landscape services and the expected data sources that are required to identify the landscape service. Services in bold are further described in this paper.

Landscape service Service category Map Governmental Database Management strategy Fieldwork

Land cover Routes ERDa GIABb Observe Counts

Residential Carrier XIndustrial production Provision XOutdoor sport Information XFruit and nut production Provision XGreenhouse food production Provision XForest habitat Habitat XWetland habitat Habitat XWater storage Regulatory XWater supply Regulatory X XEnergy conversion Carrier X XHobby gardening Information X X XCrop production Provision X X X XOvernight tourism Information X X XForest recreation Information X X X X XWetland recreation Information X X X X XRecreation for hikers Information X X XRecreation for cyclists Information X X XRecreation for horse riders Information X X XNon-land-based animal husbandry Provision X XLand-based animal husbandry Provision X XHorse boarding Provision X X XHobby farming Information X X XDitch bank protection Habitat X X X XWading bird protection in agricultural land Habitat X X XWading bird habitat in agricultural land Habitat X X

a

b

1F

2

tes

TL

ERD: environmental registration database (StraMis, 2009).GIAB: agricultural assessment database.

00 m to ascertain the neighbourhoods of the landscape service.ield observations were carried out from June to August 2009.

.3. Point observations of landscape services

Landscape services vary greatly as they, for instance, differ inheir properties (de Groot and Hein, 2007). Consequently, differ-nt methods and data sources are required to identify landscapeervices (Willemen et al., 2008). In general, we can differentiate

able 2ist of included spatial characteristics and used data sources, divided into point observati

Spatial characteristics

At data point Soil typeGround water table

Distance to Unpaved roadRural roadProvincial roadHighwayNatural areaCity/villageCultural heritage (monuments)Industrial areaGreenhouseRecreational area/element

Neighbourhood ReliefDitchPondSolitaire treeTree lineHedgerowBushCultural heritageOpennessHilliness

a Soil map: Digitised soil map of The Netherlands at scale 1:50,000 with PAWN-units (db TOP10-SE: topographical map spatial edition (vector), including land use classificatioc CHW Brabant: cultural historical valuable (monumental buildings), Atlas Province Nod AHN: Dutch digital elevation map, spatial resolution 5 m × 5 m.

between landscape services with a one-to-one relation toland-cover; those, which require one data source and are there-fore easy to identify, and other landscape services which requiremultiple data sources and are more laborious to identify.

A list of 25 landscape services present in the study area and

the potential data sources to identify the service was composed(Table 1). To account for diversity of landscape services, five cat-egories (de Groot, 2006) are included: regulatory services (e.g.flood control), habitat services (e.g. provision of natural habitat),

ons, distance to, and neighbourhood (occurrence within a radius of 50 and 100 m).

Field observation Database

Soil mapa (2006)Soil map (2006)

X TOP10-SEb (2006)X TOP10-SE (2006)X TOP10-SE (2006)X TOP10-SE (2006)X TOP10-SE (2006)X TOP10-SE (2006)X CHW Brabantc (2006)X TOP10-SE (2006)X TOP10-SE (2006)X TOP10-SE (2006)

X AHNd (2002)XXX TOP10-SE (2006)X Google Earth (2009)X TOP10-SE (2006)XX CHW Brabant (2006)

Calculated (Weitkamp et al., 2011)X AHN (2002)

e Vries, 2008).n of TDN (Topographical Service Netherlands), scale 1:10,000.ord-Brabant.

2 ical In

pr

opoCtil

fsa

2

rhliGreasa

2

wsfoOhiEattTcoatff

2

smtpooi

2

cittt

76 M.M.C. Gulickx et al. / Ecolog

rovision services (e.g. food production), information services (e.g.ecreation), and carrier services (e.g. habitation).

Broad categories of landscape services bring about a wider setf required data sources to identify the service. For example, foodroduction is a very broad category that contains different typesf landscape services, and as such, a diverse set of data sources.onversely, the subcategory land-based animal husbandry (con-aining mainly milk production) is more specified, and as a result,ncludes less diversity in the required data sources. The 25 selectedandscape services are therefore specified explicitly.

We opted to present the methodology by describing four dif-erent landscape services with different requirements (i.e. dataources): wetland habitat, forest recreation, recreation for hikers,nd land-based animal husbandry.

.3.1. Wetland habitatWetland habitat in the study area is of great importance to the

egion for both nature conservation and historical value. Wetlandsarbour a great variety of flora (e.g. peat moss Sphagnum magel-

anicum, Bog Rosemary Andromeda polifolia, and Sundews Droserantermedia), and fauna, including rare birds (e.g. Black-neckedrebe Podiceps nigricollis and Nightjar Caprimulgus europaeus), and

are butterflies and dragon flies (e.g. Large Chequered Skipper Het-ropterus morpheus and White-faced Darter Leucorrhinia dubia). Inddition, historical traces of peat extraction, such as big lakes andmall peat pits, are still visible. Wetland habitat was identified usingland-cover map (TOP10-SE, 2006).

.3.2. Forest recreationThe area contains several fragments of forested areas. Forests

ere predominantly planted between 1840 and 1900 to preventand drifting and to provide wood (Bont de, 1993). Some naturalorests started to grow on the drier and more nutrient-rich soilsf the wetland areas. These are dominated by birch Betula trees.ver the last few decades, recreational use of the forested areasas increased. Forest recreation is defined as recreational activities

n a forest larger than 2 hectares. A land-cover map (TOP10 Spatialdition, 2006) was used to determine the location of the forestedreas. Within these forested areas, recreational activity was ascer-ained using simple indicators, namely, the presence of walkingrails, cycling paths, horse riding trails, picnic tables, and car parks.hese indicators were derived from management plans, walking,ycling and horse riding routes, and from field observations. Inrder to identify the actual service, it is preferred to quantify themount of visitors to the forested areas, which is unfortunately veryime consuming. Instead, we enquired with the land owners of theorested areas to deduce whether these areas are used by peopleor recreational purposes.

.3.3. Recreation for hikersRecreation for hikers is defined as (perceived) attractive land-

capes suitable for leisure walking activity. We used a hiking routeap (‘knooppuntenroute’ network of hiking routes, 2008) to iden-

ify recreation for hikers. The route is designed to pass importantoints of interest, along attractive landscapes, and where possiblen good quality roads. This hiking route map is the most sold typef hiking routes by the tourist information centre, and therefore, its expected that they are actually used by recreational hikers.

.3.4. Land-based animal husbandryLivestock production has intensified rapidly in the study area,

orrespondingly to other parts of the Netherlands. This has resulted

n outbreaks of various infectious diseases amongst livestock, andriggering a renovation plan to improve the environmental situa-ion of livestock production. Land-based husbandry is defined ashe production of food and goods (e.g. milk and wool) by farms

dicators 24 (2013) 273–283

that depend on the land quality (i.e. they use their own landfor fodder production). Land-based husbandry is an importantsource of income in the region. The environmental RegistrationDatabase (StraMis), which details farm types (e.g. land-based, non-land-based, horticulture) and their location, was used to identifyland-based animal husbandry.

2.4. Point observations of spatial characteristics

Several spatial characteristics were identified to analyse thespatial indicators of each landscape service (Table 2). For the col-lection of spatial characteristics, both field observations and spatialdatabases were used (Table 2). This predominantly comprises ofmaps and data sources from 2006, with the exception of the ele-vation map (AHN, 2002). The openness was calculated using theprocedure proposed by Weitkamp et al. (2011).

2.5. Correlation analysis and indicator selection

In total, five data points were excluded from data analyses,because the ground observation was not in agreement with the spa-tial databases. For instance, the land was leased out and the user(the type of farm) of an arable field was not retraceable. Therefore,a total of 384 data points were included in the analyses. Statisticalanalyses were calculated in SPSS Statistics 17.

Several spatial characteristics (i.e. ditch, pond, solitaire tree,tree line, hedgerow, bush, cultural heritage) have binary variables(present = 1; absent = 0). The relation between the landscape ser-vices and the binomial spatial characteristics within a 0, 50, and100-metre radius, and correlations between landscape services wascalculated using Spearman’s Rho. Cultural heritage was also cal-culated within a 500 m radius, considering cultural heritage doesnot have to be visible to have an influence. Correlations betweenlandscape services and spatial characteristics with a continuousnumeral system (i.e. openness, elevation, relief, and distance to spa-tial characteristics) were calculated for a 0, 50, and 100-m radiususing Pearson’s r. In The Netherlands, wetland is a well-mappedland-cover type, and therefore, land-cover is considered as thespatial determinant for wetland habitat. Due to this one-to-onerelation with land-cover, further calculations for assessing corre-lations between wetland habitat and spatial characteristics werenot applied, considering these correlations are not necessary formapping wetland habitat.

The identified correlations between landscape services wereused as indicators to map the service. For each service, thecorrelation between the set of indicators and the services was cal-culated using logistic regression. The goodness of fit of the logisticregression was measured by means of the Receiver Operating Char-acteristic (ROC) curve (Pontius and Schneider, 2001; Verburg et al.,2004), which involves plotting each pair of true positive and falsepositive proportions for every possible decision threshold between0 and 1. A ROC curve value of 0.5 indicates that the model is com-pletely random and a value of 1 indicates perfect discrimination.

Logistic regression assumes that the variables are independent.Therefore, we tested the variables for their independency, i.e. formulticollinearity (Variance Inflation Factors (VIF) and tolerancetest) and spatial autocorrelation (Moran’s I).

To evaluate spatial synergies between landscape services,correlations between the location of services were calculated(Spearman’s Rho). In addition, within a radius of 1 km, the occur-

rence of other landscape services, and the distance between thedifferent services were assessed. A Kruskal–Wallis test was used tocalculate differences between the distances to the different land-scape services.

ical In

2

fLf(tta

3

ot4Aw

3s

3

(aad

3

oa(tnspawcprGtfr(PlrtgcrlwTfasf

tu

M.M.C. Gulickx et al. / Ecolog

.6. Mapping landscape services

Wetland habitat was mapped by extracting land-cover wetlandrom the land-cover map (TOP10 Spatial Edition) using ArcGIS 9.3.and-based animal husbandry, forest recreation, and recreationor hikers were mapped using the fitted logistic regression modelArcGIS 9.3). The goodness of fit of the maps was tested by a two-by-wo contingency table (cross-validation) using the observed data ofhe landscape services. This resulted in an overall, a producer’s, anduser’s accuracy.

. Results

The landscape service wetland habitat was present at 8% (N = 32)f the analysed data points, forest recreation at 8% (which is 70% ofhe data points with forest habitat; N = 29), recreation for hikers at1% (N = 157), and land-based animal husbandry at 52% (N = 200).t 7% of the analysed data points more than one landscape serviceas provided.

.1. Correlations with spatial characteristics and landscapeervices maps

.1.1. Wetland habitatWetland habitat was mapped using land-cover type wetland

Fig. 3). Validation of the wetland habitat map shows an overallccuracy of 0.96 (Table 4), with a producer’s accuracy of 0.74 anduser’s accuracy of 0.82. Considering that the accuracy is not 1.0emonstrates that the land-cover map is not 100% accurate.

.1.2. Forest recreationThe occurrence of forest recreation depends on the presence

f the land-cover forest. We found forested areas without recre-tional activities and forested areas with recreational activitiesi.e. landscape service forest recreation). Several spatial charac-eristics explained the presence of forest recreation, specifically aegative correlation with elevation, and positive correlations withoil type, ground water table, and relief (Table 3). However, com-aring forested areas with the landscape service forest recreationnd without this service (i.e. forested areas where no recreationas observed), no correlation was found with soil type (Sand

over on peat on sand, r = −0.24, P < 0.09; Earthy topsoil on deepeat, r = 0.22, P < 0.13; ‘Enk’ earth soil, r = −0.13, P < 0.38; Drift sand,= 0.15, P < 0.30) and ground water table (GWT-I, r = 0.22, P < 0.13;WT-VI, r = −0.26, P < 0.07; GWT VII, r = 0.17, P < 0.24). This shows

hat soil type and ground water table explain the occurrence oforested areas, but not the occurrence of landscape service forestecreation. However, for elevation (r = −0.391, P < 0.01) and reliefwithin 50 m radius: r = 0.29, P < 0.04; within 100 m radius: r = 0.32,< 0.02) a correlation was found between forested areas with the

andscape service forest recreation. This is in agreement with lessecreation in the forested areas of the wetlands, considering thathe wetland forests are found in higher, flatter areas. In addition, theround water level in the wetlands was higher (for which no signifi-ance was found, nonetheless, GWT-VI does show a negative trend:= −0.26, P < 0.07), resulting in less accessible forests in the wet-and. The most significant spatial characteristic was unpaved paths,

hich was positively correlated with forest recreation (Table 3).his makes a forest accessible for recreation. When consideringorests with no recreation in combination with unpaved paths,

strong negative correlation was found (r = −0.48, P < 0.00). Thishows that the presence of unpaved paths is indeed important for

orest recreation.

Relief is not included as an indicator, because of its high correla-ion to forest (VIF of 9). It is evident that the spatial characteristicsnpaved paths and land-cover forest are important factors, and

dicators 24 (2013) 273–283 277

therefore, used as indicators of the service (Table 4). The ROCvalue indicates that forest recreation is adequately explained bythe designated indicators (Table 4). Initially, elevation was alsoincluded as an indicator, however, the ROC value showed thatincluding elevation explained forest recreation less well (ROC valueof 0.81). Therefore, elevation was not included as an indicator forforest recreation. The resulting map is shown in Fig. 3. Validationof the forest recreation map shows an overall accuracy of 0.93(Table 4), with a producer’s accuracy of 0.83 and a user’s accuracyof 0.67.

3.1.3. Recreation for hikersUnderstandably, paths to walk on are crucial for recreation for

hikers, however, not all paths are equally attractive. Therefore, dif-ferent types of paths in combination with tree lines have beenassessed. Both rural roads and unpaved paths are positively cor-related with recreation for hikers (Table 3). However, unpavedpaths without tree lines are not correlated with recreation forhikers (Table 3), hence, assumedly tree lines are essential. Con-versely, there was a positive correlation found for rural roadswithout tree lines within 100 m. Then again, a positive trend wasfound between recreation for hikers and rural roads with treelines (r = 0.09, P < 0.06). In general, there was a positive correlationbetween paths and tree lines.

Landscape elements (i.e. ditches, ponds, solitaire tree lines,hedgerows, and bushes) are positively correlated with recreationfor hikers within a radius of 50 and 100 m (Table 3). Separately,only the landscape elements solitaire trees, tree lines, and ditchesare positively correlated within 100 m (Table 3). It is not a surprisethat ditches are positively correlated, considering the high densityof ditches throughout the study area. In addition, no sufficient mapof ditches was available for this study area, therefore, ditches werenot included as a determinant of recreation for hikers.

An unexpected result is the positive correlation between recre-ation for hikers and short distances to industry (Table 3). However,there was no correlation between recreation for hikers and indus-try within a radius of 50 metres (r = 0.00, P < 0.99), or within 100 m(r = 0.00, P < 0.99). Therefore, industry was not taken into accountfor mapping recreation for hikers.

Cultural heritage was positively correlated with short distancesto recreation for hikers (Table 3). Likewise, there was a positivecorrelation between a high density of hiking routes and culturalheritage (r = 0.50, P < 0.00). However, cultural heritage was not cor-related with recreation for hikers within 50 m (r = 0.01, P < 0.89), norwithin 100 m (r = 0.07, P < 0.16). Cultural heritage seems to have apositive influence on recreation for hikers. Presumably, due to fewcultural heritage locations within 50 m (N = 3) and 100 m (N = 12)from walking recreation, no direct correlation with a defined dis-tance to cultural heritage could be recognised, and therefore, is notconsidered as a determinant for recreation for hikers.

The selected indicators for mapping the occurrence of recre-ation for hikers are: unpaved paths with solitary trees or tree lineswithin 100 m and rural roads with solitary trees within 100 m(Table 4). The ROC value indicates that recreation for hikers is par-tially explained by the designated indicators (Table 4). The resultingmap is shown in Fig. 3. Validation of the recreation for hikers mapshows an overall accuracy of 0.56 (Table 4), a producer’s accuracyof 0.55 and a user’s accuracy of 0.56.

3.1.4. Land-based animal husbandryLand-based animal husbandry has a negative correlation with

relief and a positive correlation with openness (Table 3), which can

be explained by the fact that level and open terrain has benefitsfor land cultivation. These spatial characteristics do not explainland-based animal husbandry explicitly, but rather agriculturalactivities in general. Soil type is another spatial characteristic that

278 M.M.C. Gulickx et al. / Ecological Indicators 24 (2013) 273–283

reatio

aswvwPspcae

Fig. 3. Landscape service maps: wetland habitat; forest rec

lso has positive correlations with other agricultural landscapeervices. Land-based animal husbandry was positively correlatedith sand-cover-on-peat-on-sand, slightly-loamy-fine-sand, and

ery-loamy-fine-sand (Table 3). Non-land-based animal husbandryas positively correlated with slightly-loamy-fine-sand (r = 0.12,< 0.02), provision of tillage crops was positively correlated withlightly-loamy-fine-sand (r = 0.145, P < 0.00) and sand-cover-on-

eat-on-sand (r = 0.10, P < 0.04), and greenhouse was positivelyorrelated with very-loamy-fine-sand (r = 0.16, P < 0.00). As therere differences in the relations between soil type and the differ-nt agricultural landscape services, soil type can be used as an

n; recreation for hikers; and land-based animal husbandry.

indicator in combination with other spatial characteristics that areonly applicable with land-based animal husbandry.

Short distances to nature area, city, and industry are nega-tively correlated with land-based animal husbandry (Table 3). Alsofor other agricultural landscape services, a negative correlationwith short distances to nature areas was found, specifically, non-land-based animal husbandry (r = 0.20, P < 0.00), provision of tillage

crops (r = 0.17, P < 0.00), and greenhouse (r = 0.10, P < 0.05). How-ever, no correlation was found with short distances to either village(non-land-based animal husbandry: r = 0.01, P < 0.90; and provi-sion of tillage crops: r = 0.03, P < 0.54), or industry (non-land-based

M.M.C. Gulickx et al. / Ecological Indicators 24 (2013) 273–283 279

Table 3Correlations between landscape services and spatial characteristics. Correlations between landscape services and spatial characteristics are calculated using Pearson’s r(normal distributed data) and Spearman’s rho (non-normal distributed data) in SPSS Statistics 17.

Landscape service

Spatial characteristics Forest recreation Recreation for hikers Land-based animal husbandry

At data pointElevation −0.12* −0.13** −0.06Ground water tablea

I (<20 to <50 cm-gl) 0.23*** −0.11* −0.09II (<40 to <50–80 cm-gl) −0.00 −0.12* −0.05III (<40 to <80–120 cm-gl) −0.08 −0.06 −0.07IV (<40 to <80–120 cm-gl) −0.06 0.07 0.10*V (<40 to <120 cm-gl) −0.07 0.02 0.01VI (<40–80 to <120 cm-gl) −0.15** 0.03 0.11*VII (>80 to >516 cm-gl) 0.13* 0.03 −0.06

Soil typeSand cover on peat on sand −0.13* 0.04 0.18***Slightly loamy fine sand −0.01 0.09 0.12*Earthy topsoil on deep peat 0.10 −0.16*** −0.15**Earthy topsoil on peat on sand −0.01 −0.15*** −0.18***‘Enk’ earth soil −0.10* 0.04 −0.02Very loamy fine sand −0.05 0.09 0.10*Drift sand 0.26*** 0.03 −0.15**

Decreasing distance toRoads

Unpaved path 0.31*** 0.09 −0.11*Rural road −0.18*** 0.10 0.00Highway and Provincial road 0.14** 0.06 −0.04

Industrial area −0.01 0.11* −0.12*City/village −0.01 0.07 −0.11*Greenhouse −0.13** 0.04 0.01Cultural heritage −0.05 0.13** −0.06Natural area – −0.04 −0.24***

Forest recreation Recreation for hikers Land-based animal husbandry

50 m 100 m 50 m 100 m 50 m 100 m

NeighbourhoodRelief 0.19*** 0.19*** 0.02 0.06 −0.31*** −0.28***Openness – 0.33*** – 0.04 – 0.37***Land elementsb – – 0.12* 0.18*** 0.33*** 0.40***

Ditch – – 0.08 0.11* 0.35*** 0.43***Pond – – 0.02 0.04 0.06 −0.03Solitaire tree – – 0.06 0.11* 0.03 0.12*Tree line – – 0.10 0.15** 0.08 0.20***Hedgerow – – 0.08 0.037 −0.08 −0.07Bush – – 0.05 0.087 −0.03 0.08

RoadsUnpaved path 0.32*** 0.29*** 0.14** 0.19*** −0.09 −0.07

Without tree line – – 0.07 0.04 −0.18*** −0.12*With tree line 0.12* 0.19*** 0.07 0.04

Rural road −0.18*** −0.24*** 0.11* 0.19*** −0.05 0.15**Without tree line – – 0.09 0.11* 0.00 0.07With tree line – 0.05 0.09 −0.06 0.10

Provincial road 0.00 −0.05 −0.03 −0.01 −0.07 −0.09Highway −0.04 0.06 0.02 0.00 −0.03 0.04

T and aof the

arlw

wowPhb1(g

he significance level is indicated with * for p < 0.05, ** for p < 0.01, *** for p < 0.001,a Groundwater table is derived from the Policy Analysis for Water-managementb Land elements is aggregation of ditch, pond, solitaire tree, hedgerow, and bush.

nimal husbandry: r = 0.04, P < 0.40; provision of tillage crops:= 0.039, P < 0.44; and greenhouse: r = −0.05, P < 0.32). In contrast toand-based animal husbandry, greenhouses have a positive relation

ith villages (r = −0.10, P < 0.05).A positive correlation with all landscape elements assembled

as found within both a 50 and 100 m radius (Table 3). Forther agricultural landscape services no correlation was foundithin a 50 m radius (non-land-based animal husbandry: r = 0.07,< 0.17; provision of tillage crops: r = 0.04, P < 0.45; and green-ouse: r = −0.06, P < 0.20). In addition, the positive correlation

etween solitaire trees and land-based animal husbandry within a00 m radius was not found for non-land-based animal husbandryr = −0.01, P < 0.89), provision of tillage crops (r = 0.09, P < 0.07), andreenhouse (r = −0.63, P < 0.21). These results show that there are

re shown in bold.Netherlands. Groundwater level is expressed in cm below ground level (gl).

differences between spatial characteristics and various agriculturallandscape services.

Notably, a negative correlation was found with unpaved pathswithout tree lines within both 50 and 100-m radius, even thoughno correlation was found for unpaved paths in total (Table 3). Thiscan be explained by the fact that unpaved paths without tree linesoften occur in forested areas where no land-based animal hus-bandry occurs. When excluding unpaved paths without tree linesin forested areas, no correlation with unpaved paths without treelines was found (r = −0.00, P < 0.96).

Land-based animal husbandry occurred mainly within theland cover grassland, indicating the importance of grassland forthis service. Grassland was indeed positively correlated (r = 0.49,P < 0.00), however, it can provide many services. As such, we

280 M.M.C. Gulickx et al. / Ecological Indicators 24 (2013) 273–283

Table 4Indicators (spatial characteristics) used to map the landscape services, logistic regression of the indicators with Cox & Snell R2 (significance in brackets), goodness of fit ofthe logistic regression (ROC curve), and the overall accuracy of the landscape service map (contingency table).

Landscape service Indicators Log. regression ROC Map accuracy

Wetland habitat Land-cover wetland – – 0.96

Forest recreation Land-cover forest, and 0.30 (0.00) 0.98 0.93Unpaved path within 200 m

Recreation for hikers Land elementsa within 100 m, and 0.12 (0.00) 0.68 0.56Rural road within 100 m, orUnpaved path within 100 m

Land-based animal husbandry Industry not within 100 m, or 0.21 (0.00) 0.78 0.64Villages not within 100 m, andSolitaire trees within 100 m on sand-cover-on-peat-on-sand orslightly-loamy-fine-sand, or very-loamy-fine-sand, orRural roads within 100 m on sand-cover-on-peat-on-sand,slightly-loamy-fine-sand, or very-loamy-fine-sand, and

nd,-fine-

hsowahs(bt

bhnlsoab

atooliVaa

3

hschlbot(hl1s

other regions. However, relations between landscape services and

Grassland on sand-cover-on-peat-on-saslightly-loamy-fine-sand, or very-loamy

a Tree lines and solitaire trees.

ave calculated the correlation with grassland located on theoil types sand-cover-on-peat-on-sand, slightly-loamy-fine-sand,r very-loamy-fine-sand (soil types that are positively correlatedith land-based animal husbandry; Table 3), which still showedpositive correlation (r = 0.41, P < 0.00). Non-land-based animal

usbandry was also positively correlated with slightly-loamy-fine-and, however, not with grassland on slightly-loamy-fine-sandr = 0.06, P < 0.09). Considering the importance of grassland for land-ased animal husbandry, this land cover type, in combination withhe three soil types, was included as an indicator for the service.

The objective of this study is to map the occurrence of land-ased animal husbandry, and not agriculture in general. Hence, weave only included correlations with spatial characteristics that areot correlated with other agricultural landscape services (i.e. tree

ines and solitary trees within 100 m on sand-cover-on-peat-on-and, slightly-loamy-fine-sand, or very-loamy-fine-sand, grasslandn the same 3 soil types, and by excluding distances to naturereas). These correlations are assumed to be more related to land-ased animal husbandry and less to agriculture in general.

The selected indicators to map land-based animal husbandryre: no villages within 100 m; no industry within 100 m; soli-aire trees within 100 m; rural road within 100 m; and grasslandn either sand-cover-on-peat-on-sand, slightly-loamy-fine-sand,r very-loamy-fine-sand (Table 4). The ROC value indicates thatand-based animal husbandry is reasonably explained by the des-gnated indicators (Table 4). The resulting map is shown in Fig. 3.alidation of the land-based animal husbandry map shows an over-ll accuracy of 0.64 (Table 4), with a producer’s accuracy of 0.57 anduser’s accuracy of 0.54.

.2. Correlations between landscape services

Land-based animal husbandry, forest recreation, and wetlandabitat have opposing requirements. Therefore, no overlap of theseervices was found. Land-based animal husbandry has a negativeorrelation with forest recreation (r = −0.20, P < 0.00) and wetlandabitat (r = −0.15, P < 0.00). In addition, forest recreation and wet-

and habitat do not occur within a radius of 550 m around landased animal husbandry (Fig. 4d). This indicates that they do notccur close to each other either. Similarly, a negative correla-ion was found between recreation for hikers and wetland habitatr = −0.10, P < 0.04). This was partly due to the fact that the selected

iking routes only cross a fraction of the wetlands, but also because

arge parts of the wetland habitat are inaccessible. However, withinkm radius, recreation for hikers occurs regularly, and even at the

ame location (N = 3; Fig. 4c), showing that these two services can

sand

occur closely together. Recreation for hikers does occur at the samelocation as land-based animal husbandry and forest recreation, yet,no positive correlations are found. Within a distance of 1 km radiusit is apparent that recreation for hikers occurs regularly at a shortdistance with both forest recreation (Fig. 4b) and land-based animalhusbandry (Fig. 4d).

The assessment of the occurrence of different landscape ser-vices within a 1 km radius shows that some landscape functions dooccur together within this range (Fig. 4), whereas they did not occurtogether within a range of 100 m. In addition, wetland habitat didnot occur with forest recreation and land-based animal husbandrywithin a 100-m radius. The assessment of a 1 km radius shows thatforest recreation and land-based animal husbandry do not occurwithin a range of 680 and 570 m, respectively.

3.3. Multicollinearity and spatial autocorrelation

The Variance Inflation Factors (VIF) and the tolerance testshowed no evidence for multicollinearity between the variablesused as indicators for the landscape services (VIF ranged between1.01 and 1.73, tolerance ranged between 0.58 and 1.00; VIF morethan 5 and tolerance less than 0.2 are considered to be a causefor concern for multicollinearity (Menard, 2001). However, thevariables relief and forest did show multicollinearity (VIF of 9),therefore, relief was not included as an indicator for forest recre-ation.

The spatial autocorrelation analysis of the spatial indicators offorest recreation has indicated a weak positive dependence onthe geographical space (Moran’s I = 0.13, P < 0.00), which can beexplained by the fact that forests are concentrated in patches acrossthe study area. A random spatial pattern was found for the indi-cators of both recreation for hikers (Moran’s I = 0.06, P < 0.10) andland-based animal husbandry (Moran’s I = 0.01, P < 0.10).

4. Discussion

This paper is set out to develop an approach to identify and maplandscape services. We have created four landscape service maps.The obtained indicators are site-specific, however, they indicatethat linkages between landscape services and physical propertiesof the environment exist. Hence, our approach can be applied in

spatial indicators are likely to differ by region depending on theenvironmental and socio-economic context. In the following sec-tions we discuss the strength and weaknesses of the methods andresults of this study.

M.M.C. Gulickx et al. / Ecological Indicators 24 (2013) 273–283 281

F t; (B)

4

teaaiha

sEBo2eYaafs

ig. 4. Error bars of the mean distances of landscape services to: (A) wetland habita

.1. Overall methodology

This study shows that a wide variety of data sources are neededo identify and map landscape services, as indicated by Willement al. (2008). The necessity for these various data sources can bescribed to the vast variety of landscape services. As a consequence,standardised approach to identify and map landscape services is

nfeasible, and perhaps even wrong. The methodology presentedere takes this variety into account, providing a robust frameworknd a flexible way of assessing landscape services.

In general, the potential of the landscape to provide land-cape services was mapped (e.g. Gimona and van der Horst, 2007;goh et al., 2008; Willemen et al., 2008; Kienast et al., 2009; vanerkel and Verburg, 2011). The potential landscape services areften based on proxies (e.g. Egoh et al., 2008; Willemen et al.,008), on generalised relations with land cover and use (Burkhardt al., 2011), or on expert knowledge (Kienast et al., 2009; Haines-oung et al., 2012). Some of these studies are based on general

ssumptions and not on site-specific quantified relations, which isdrawback of these approaches (de Groot et al., 2010). A valuable

eature of our methodology is that the indicators of the landscapeervices are based on site-specific relations. These are necessary

forest recreation; (C) recreation for hikers; (D) land-based animal husbandry.

to map the actual landscape service. The actual service requiresspecific information of the area that is often not available. In thisstudy we used ground observations to obtain information on theactual distribution of landscape services and based our maps onsite-specific correlations accordingly. However, for forest recre-ation and recreation for hikers, it would have been ideal to includethe number of visitors, which gives the most accurate depictionof the actual service. Unfortunately, it is very time consuming toacquire such specific information. As an alternative, proxies suchas the location of frequently used hiking trails were used as indica-tors and checked with experts of the area (e.g. land owners), whichprovided a valuable substitute for counting visitors.

4.2. Analysing spatial scales

Our study has shown that the spatial scale of indicators of land-scape services differ, which is also recognised by other researchers(de Groot and Hein, 2007; Pérez-Soba et al., 2008). We have found

that 19% of the correlations between landscape services and spatialcharacteristics are different for the 50 and 100-m radius. This indi-cates that differences are found even at a relatively small spatialscale. The largest scale of 100 m radius in this study is presumably

2 ical In

nswa

etrsise

4

bihstttvalatasltvhtrauifrrcufiaatamtawdc2wdisfauii

82 M.M.C. Gulickx et al. / Ecolog

ot adequate for the recreation for hikers, considering the land-cape – as far as the eye can see – has an influence on this service,hich can easily range beyond 100 m. Therefore, in case of recre-

tion for hikers, a larger scale would be advisable.In addition, distances between landscape services are differ-

nt at various scales. Correlations between landscape services athe same location can differ from the assessment within a 1 kmadius. However, various studies of landscape services do not con-ider different spatial scales. Bearing in mind that the spatial scales important for landscape services and the fact that the effectivepatial scales of most landscape services are still uncertain, it isssential that future studies include spatial scale differences.

.3. Strength of correlations and validation

The efficiency of mapping landscape services can be improvedy recognising indicators for landscape services. Several spatial

ndicators for landscape services have been found, although mostave weak correlations (Table 3), which is also found in othertudies (e.g. Chan et al., 2006; Egoh et al., 2008). The results ofhe logistic regression show that combining spatial characteris-ics improves the strength of correlations (Table 4). Presumably,he surroundings of the studied landscape services have so muchariability that they cannot be explained by individual spatial char-cteristics. For instance, hiking routes pass as many attractiveandscapes as possible, however, it is impossible to avoid all lessttractive sites. The indicators found for forest recreation provedo be adequately reliable (Table 4). The indicators for land-basednimal husbandry and recreation for hikers explain the landscapeervices reasonably well. However, the validation of the createdandscape service maps, which are based on the indicators, showhat only forest recreation is highly accurate. The landscape ser-ice map recreation for hikers and land-based animal husbandryave an accuracy of only 60%, which can be partly explained byhe high variability of their location, as stated above. In addition,ecreation for hikers is established using hiking routes that aressumed to be used by hikers, however we did not consider these of these routes to make the demand more explicit. By includ-

ng the demand of the hiking routes, indicators for recreationor hikers are likely to become more accurate, and are thereforeecommended. However, we have carefully selected the hikingoutes that are currently promoted by the tourist informationentre and most frequently sold, therewith assuming a frequentse of these routes. For land-based animal husbandry it is dif-cult to determine correlated spatial characteristics, consideringgricultural landscape services have numerous similar spatial char-cteristics. Distinction between agricultural landscape services isherefore difficult. Yet, several distinct indicators for land-basednimal husbandry are found, but these alone are not enough toake a highly accurate map. Additional data (e.g. the environmen-

al permit database that includes the type and size of agriculturalctivities) could be used to increase the accuracy. Unfortunately,e were not allowed to duplicate spatial data from this databaseue to privacy issues. Inaccessibility and the lack of data is aritical constraint in landscape service research (Verburg et al.,009). Collaboration with governing bodies and other institutionsith landscape service interest will improve the availability ofata, either by gaining access to existing databases or by involv-

ng them in the collection of the necessary data. Additionally, dataources are harboured at different organisations and within dif-erent databases, which makes the collection of data extremely

rduous. Collaboration with all organisations that might haveseful information for the analysis of the landscape services of

nterest is essential and needs to be further developed in order tomprove the accuracy of landscape service maps.

dicators 24 (2013) 273–283

4.4. Applicability of landscape service maps

Opposing to the definition as we presented it, a landscape ser-vice is more often defined as ‘the capacity of the landscape toprovide goods and services that satisfy human needs, directly orindirectly’ (MA, 2005; de Groot, 2006; Hein et al., 2006; Syrbe et al.,2007; Willemen et al., 2008; Kienast et al., 2009; Verburg et al.,2009; Posthumus et al., 2010). In this paper, we did not considerthe capacity, instead, we looked at the actual presence of a service.By regarding the capacity to provide landscape services, a more in-depth representation of the potential benefits can be obtained, asthe actual supply of landscape services can rapidly change due to,for instance, change in human demand or depletion of supply (deGroot and Hein, 2007). However, measuring or even defining thecapacity for a landscape services has proven to be difficult, whichis also recognised by de Groot and Hein (2007). The assessment ofthe capacity is complicated further through human technology thatcan increase our capability to adjust the landscape to our desires.The assessment of the actual presence of a landscape service, i.e. thelandscape provides the related goods and services, has proven tobe difficult due to insufficient data, which is also discussed above.However, the proposed methodology can be used to show differ-ent gradations of suitability (Willemen et al., 2010). For instance,when all indicators are present, the location is highly suitable, butwhen only half of the indicators are present the location is moder-ately suitable. These suitability maps can be of great value for policymakers and spatial planners.

5. Conclusions

To analyse and map various landscape services different datasources and approaches are required, therefore, standardisationis not possible. Instead, this study provided and tested a robustframework and a flexible approach to analyse and map landscapeservices. The results show that the effective spatial scales and pat-terns of landscape services differ, which is important for assessingindicators to map these services and for analysing the multifunc-tionality of a landscape. The landscape service maps provide policymakers and spatial planners insight on actual landscape services,which they can incorporate in their decision making.

Acknowledgements

We are grateful to the many experts who contributed to thiswork, especially Gerd Weitkamp for his support with the opennessmodel, Gerard Heuvelink and Bram van Putten for their statisticalsupport. We also thank Richard Smithers for his valuable commentsand two anonymous reviewers for their constructive remarks.

References

AHN, 2002. Actueel Hoogtebestand Nederland (Current Elevation database theNetherlands). Rijkswaterstaat, Delft, the Netherlands. http://www.ahn.nl.

Bont de, C.H.M., 1993. ‘. . .Al het merkwaardige in bonte afwisseling. . .’ Een his-torische geografie van Midden- en Oost-Brabant. Stichting Brabants Heem,Waalre.

Burkhard, B., Kroll, F., Nedkov, S., Müller, F., 2011. Mapping ecosystem service supply,demand and budgets. Ecol. Indic. 21, 17–29.

Chan, K.M.A., Shaw, M.R., Cameron, D.R., Underwood, E.C., Daily, G.C., 2006. Conser-vation planning for ecosystem services. PLoS Biol. 4, e379.

CHW Brabant, 2006. Map Cultural Historical Value, Digital Atlas ProvinceNoord-Brabant. [Online] http://www.brabant.nl/kaarten/culturele-kaarten/cultuurhistorische-waardenkaart-2010.aspx.

Costanza, R., d’Arge, R., de Groot, R., Farber, S., Grasso, M., Hannon, B., Limburg, K.,Naeem, S., O’Neill, R.V., Paruelo, J., Raskin, R.G., Sutton, P., van den Belt, M., 1997.

The value of the world’s ecosystem services and natural capital. Ecol. Econ. 25,3–15.

de Groot, R., 2006. Function-analysis and valuation as a tool to assess land use con-flicts in planning for sustainable, multi-functional landscapes. Landsc. UrbanPlann. 75, 175–186.

ical In

d

d

D

d

E

F

G

GH

H

H

K

M

M

P

M.M.C. Gulickx et al. / Ecolog

e Groot, R., Hein, L., 2007. Concept and valuation of landscape functions at differentscales. Multifunct. Land Use, 15–36.

e Groot, R.S., Alkemade, R., Braat, L., Hein, L., Willemen, L., 2010. Challenges inintegrating the concept of ecosystem services and values in landscape planning,management and decision making. Ecol. Complex. 7, 260–272.

eFries, R.S., Foley, J.A., Asner, G.P., 2004. Land-use choices: balancing human needsand ecosystem function. Front. Ecol. Environ. 2, 249–257.

e Vries, F., 2008. Nationaal Hydrologisch Instrumentarium NHI. Modelrapportage,deelrapport Bodem. Alterra, Wageningen, The Netherlands.

goh, B., Reyers, B., Rouget, M., Richardson, D.M., Le Maitre, D.C., van Jaarsveld,A.S., 2008. Mapping ecosystem services for planning and management. Agric.Ecosyst. Environ. 127, 135–140.

ry, G.L.A., 2001. Multifunctional landscapes—towards transdisciplinary research.Landsc. Urban Plann. 57, 159–168.

imona, A., van der Horst, D., 2007. Mapping hotspots of multiple landscape func-tions: a case study on farmland afforestation in Scotland. Landsc. Ecol. 22,1255–1264.

oogle Earth 5.1, 2009. Satellite images, 51◦24′N 5◦48′E. http://earth.google.com.aines-Young, R., Potschin, M., Kienast, F., 2012. Indicators of ecosystem service

potential at European scales: mapping marginal changes and trade-offs. Ecol.Indic. 21, 39–53.

ein, L., van Koppen, K., de Groot, R.S., van Ierland, E.C., 2006. Spatial scales,stakeholders and the valuation of ecosystem services. Ecol. Econ. 57,209–228.

olmes, J., 2006. Impulses towards a multifunctional transition in rural Australia:gaps in the research agenda. J. Rural Stud. 22, 142–160.

ienast, F., Bolliger, J., Potschin, M., de Groot, R., Verburg, P., Heller, I., Wascher, D.,Haines-Young, R., 2009. Assessing landscape functions with broad-scale envi-ronmental data: insights gained from a prototype development for Europe.Environ. Manage. 44, 1099–1120.

A, 2005. Millenium Ecosystem Assessment. Ecosystems and Human Well-Being:Biodiversity Synthesis. Island Press, Washington, DC.

enard, S., 2001. Applied Logistic Regression Analysis. Sage University Paper, Sage,Thousand Oaks, USA.

érez-Soba, M., Petit, S., Jones, L., Bertrand, N., Briquel, V., Omodei-Zorini, L., Contini,C., Helming, K., Farrington, J.H., Mossello, M.T., Wascher, D., Kienast, F., Groot, R.,2008. Land use functions—a multifunctionality approach to assess the impact

dicators 24 (2013) 273–283 283

of land use changes on land use sustainability. Sustain. Impact Assess. Land UseChanges, 375–404.

Pontius, R.G., Schneider, L.C., 2001. Land-cover change model validation by an ROCmethod for the Ipswich watershed, Massachusetts, USA. Agriculture. Ecosyst.Environ. 85, 239–248.

Posthumus, H., Rouquette, J.R., Morris, J., Gowing, D.J.G., Hess, T.M., 2010. A frame-work for the assessment of ecosystem goods and services; a case study onlowland floodplains in England. Ecol. Econ. 69, 1510–1523.

Provincie Noord-Brabant, 2005. Reconstructieplan/Milieueffectrapport De Peel.’sHertogenbosch.

StraMis, 2009. Environmental Registration Database. Roxit B.V., Zwolle, TheNetherlands.

Syrbe, R.U., Bastian, O., Roder, M., James, P., 2007. A framework for monitor-ing landscape functions: The Saxon Academy Landscape Monitoring Approach(SALMA), exemplified by soil investigations in the Kleine Spree Floodplain (Sax-ony, Germany). Landsc. Urban Plann. 79, 190–199.

Termorshuizen, J.W., Opdam, P.F.M., 2009. Landscape services as a bridgebetween landscape ecology and sustainable development. Landsc. Ecol. 24,1037–1052.

TOP10-SE, 2006. Topographic Map Spatial Edition, scale 1:10,000 (Top10 vector).Topografische Dienst Kadaster, Apeldoorn, The Netherlands.

van Berkel, D.B., Verburg, P.H., 2011. Sensitising rural policy: assessing spatial vari-ation in rural development options for Europe. Land Use Policy 28, 447–459.

Verburg, P., Schot, P., Dijst, M., Veldkamp, A., 2004. Land use change modelling:current practice and research priorities. GeoJournal 61, 309–324.

Verburg, P.H., van de Steeg, J., Veldkamp, A., Willemen, L., 2009. From land coverchange to land function dynamics: a major challenge to improve land charac-terization. J. Environ. Manage. 90, 1327–1335.

Weitkamp, G., Bregt, A., Van Lammeren, R., 2011. Measuring visible space to assesslandscape openness. Landsc. Res. 36, 127–150.

Willemen, L., Hein, L., van Mensvoort, M.E.F., Verburg, P.H., 2010. Space for peo-ple, plants, and livestock? Quantifying interactions among multiple landscape

functions in a Dutch rural region. Ecol. Indic. 10, 62–73.

Willemen, L., Verburg, P.H., Hein, L., Mensvoort, M.E.F., 2008. Spatial characterizationof landscape functions. Landsc. Urban Plann. 88, 34–43.

Wilson, G.A., 2008. From ‘weak’ to ‘strong’ multifunctionality: conceptualising farm-level multifunctional transitional pathways. J. Rural Stud. 24, 367–383.