Schwarz Urban Indicators 2010

Landscape and Urban Planning 96 (2010) 2947

Contents lists available at ScienceDirect

Landscape and Urban Planning

journa l homepage: www.e lsev ier .com/ l

Urban ch

Nina SchUFZ Helmhol y, Perm

a r t i c l

Article history:Received 17 MReceived in re18 December 2Accepted 21 JaAvailable onlin

Keywords:TypologyCluster analysUrban auditCORINE land cLandscape meEurope

in urbimpoorm,e avarelatban fformpeanrelampa

1. Introdu

1.1. Urban form between compactness and sprawl

Urban form reveals the relationship between a single city andits rural hinterland (Grimm et al., 2008) as well as the impactof human a(Alberti, 200et al., 2007)andWegenurban spraidealised, E(Dieleman a

Urban sareas by thor semi-detandWegenment is conwhich aimstowardsmosuch planni

The comhigh densitoften foundthe compac

Tel.: +49 0E-mail add

andminimising the use of undeveloped land. The compact city as avision of urban planning is characterised by a high density of usage,short travel distances and a higher quality of life (Jenks et al., 1996).For designing policies of urban planning towards a more compactdevelopment, measuring urban form can provide a variety of infor-

0169-2046/$ doi:10.1016/j.ctions on the environment within and around a city5; Environmental Protection Agency, EPA, 2001;Weng. This also relates to transportation patterns (Dielemaner, 2004). An ongoing debate distinguishes between thewl often found in North American cities versus theuropean compact city as two opposite urban formsnd Wegener, 2004; Frenkel and Ashkenazi, 2008).prawl is the large expansion of cities into surroundinge creation of new low-density suburbs with detachedached housing and large commercial strips (Dielemaner, 2004; Schneider andWoodcock, 2008). This develop-tested by groups such as the Smart Growth Movement,at managing sprawl and steering urban developmentre compactness (see Jabareen, 2006, for an overview onng concepts).pact city is the opposite urban form characterised by

ies and relatively shorter distances. European cities areto be compact (Guerois and Pumain, 2008). However,t city is also an objective of urban planning (Burton,

341 235 1970; fax: +49 0 341 235 1939.ress: [email protected].

mation (Alberti, 1999; Batty, 2008; EPA, 2001). The idea of thecompact city was integrated into the concept of sustainable urbanform (Jabareen, 2006), which includes compactness amongst otheraims suchas sustainable transport andadiversity of potential activ-ities within a neighbourhood.

Ananalysis ofurban formreveals theproblemsandchallengesofurban development. From a policy point of view, this is necessaryto identify areas with a high need of policy intervention and todetermine the diversity of urban developments. In the following,the relevance of urban form in European policies is highlighted.

1.2. Urban form in Europe

Analysing urban form is crucial for Europe because four outof ve European citizens live in urban areas (European Commis-sion, EC, 2006a). Furthermore, European cities have very diversehistories of urban development, especially since World War II,with the iron curtain separating capitalist and communist urbandevelopment, and the economic transition after 1990. Today, spa-tial planning still lies in the sovereignty of the European Unionmember states. Therefore, the method and extent of how spa-tial planning in Europe is coordinated vertically and horizontallyamong various levels of government, sectoral policies, and gov-

see front matter 2010 Elsevier B.V. All rights reserved.landurbplan.2010.01.007form revisitedSelecting indicators for

warz

tz Centre for Environmental Research, Department of Computational Landscape Ecolog

e i n f o

arch 2009vised form009nuary 2010e 4 March 2010

is

overtrics

a b s t r a c t

Four out of ve European citizens lifea city inuences daily life and is anUrban planning can inuence urban fof the numerous indicators which arto landscape metrics and population-analysis identify the most relevant urpean cities according to their urbanResults indicate that researchers, Eurofor analysing urban form due to strongof differences in urban form when cowhole of Europe.

ction 2002)ocate / landurbplan

aracterising European cities

oserstrasse 15, 04318 Leipzig, Germany

an areas, and urban form like the density or compactness ofrtant factor for both quality of life and environmental impact.but due to practicality needs to focus on a few indicators outilable. The present study analyses urban form with respected indicators for 231 European cities. Correlations and factororm indicators. Furthermore, a cluster analysis groups Euro-. Signicant differences between the clusters are presented.administration and urban planners can select few indicators

tionships between single indicators. But they should be awarering European cities or working on planning policies for the

2010 Elsevier B.V. All rights reserved.

is meant to accommodate urban development while

30 N. Schwarz / Landscape and Urban Planning 96 (2010) 2947

ernmental and non-governmental organisations is organised verydifferently (Albrechts et al., 2003).

Accordingly, European cities and their spatial developmenthavebeen receiving much attention. Starting with the Aalborg Char-ter of Europ(European SEuropean cilanduse andspectivewaSpatial Planin the eldof the Eurotheir enviroronment (explicit aimthe Leipzigritorial Agediversity ofEuropean pthe Europeaour develop

Few combeen condutial Planninin Europe (resembles tprovide a dKasanko et2006) analydispersion.Europe, NorAccording tthe densestEurope arelower densern or centstructures,than in othdevelopmeland cover.yse gradienvery diverseurban form

1.3. Denit

The deerature. Whmeasure ur(Herold et aeconomic aand Ashkenmore, the qof urban for2008; Huanmainly refegiven pointalters urban

For thisis used. Accstructure aof populatioresult of a vnomic andpast (Batty

Researchers and practitioners who endeavour to quantify theurban form of a single city or a whole range of cities can choosefrom numerous indicators. At least two strands of discussionwith respect to measuring urban form are distinguished: land-

metrelopnerate fotion-sed inthe

e botommcus ir ofat areomp

dscama

Kasaies oment08)rldwan ldimues f95; Tio-eentlydy ong sx ofed 7urba, andsocii (20ss froe, declustpulatcludil and

ms a

overities, butConsethoet ofroadors erath

publanaly005gle ohioniricabyean Cities and Towns Towards Sustainability in 1994ustainable Cities and Towns Campaign, ESCTC, 1994),ties and towns are recognised as key players in effectivedevelopment. The European Spatial Development Per-

s adopted in 1999 (EC, 1999), followed by the Europeanning Observation Network as a scientic communityof territorial development in 2002. The Commissionpean Communities stressed the inuence of cities onnment in the Thematic Strategy on the Urban Envi-EC, 2006a) and adopted the Cohesion policy with theof managing urban sprawl (EC, 2006b). Furthermore,Charter on Sustainable European Cities and the Ter-

nda (European Union, EU, 2007a, 2007b) highlight theurban development in Europe and its importance forolicy. Lately, the Council of Europe (CoE, 2008) adoptednUrban Charter II, declaring . . . thatwemust organisement around different types of urban form [. . .].parative analyses regarding urban form in Europe havected so far. A typology elaborated by the European Spa-g Observation Network regarding urban-rural relationsESPON, 2005) based upon demography and land usehe typology found here in some respects, but does notetailed and distributed view on urban form in Europe.al. (2006; see also European Environment Agency, EEA,sed 15 European cities over time to characterise theirThey qualitatively clustered these cities into Southernthern or central Europe, andWestern or central Europe.o their ndings, cities in Southern Europe are still todayand most compact cities. Cities in Northern or central

, according to Kasanko et al. (2006), characterised byities than Southern or Western European cities. North-ral European cities have looser, discontinuous urbanand the amount of built-up area per person is higherer regions. Guerois and Pumain (2008) analysed thent of 54 cities in Western Europe, also using CORINEThe authors used the density of built-up areas to anal-ts from the city centre to the periphery. In sum, theseanalyses reveal differing concepts of and indicators for

.

ion of and indicators for urban form

nitions of urban form vary to a great extent in the lit-ile some authors solely rely on land use/land cover toban form in terms of the physical structure of a cityl., 2002; Huang et al., 2007), others also include socio-spects such as population number or density (Frenkelazi, 2008; Kasanko et al., 2006; Tsai, 2005). Further-uestion of whether the sheer size of a city is one aspectm (e.g. Tsai, 2005) or an independent indicator (Batty,g et al., 2007) is still open. However, urban form itself isrred to as a property of a city and therefore static for ain time, while urban growth is a dynamic process thatform.

paper, the broadest denition possible of urban formordingly, urban form here encompasses the physical

nd size of the urban fabric as well as the distributionn within the area. Urban form of a specic city is the

ariety of inuences, including site and topography, eco-demographic development and planning efforts in theand Longley, 1994).

scapeas devThe gecompuPopuladiscussity oranalysmost cThe fonumbeand thto be c

(1) Lansumbycitop(20wourbtalval19

(2) SocrecstuusimiparofitytoTsanesizofpo(intia

1.4. Ai

Thepean cstudiestively.two msmall syse a bindicatupon aa fewcators(Tsai, 2the junble fasto empEuropetors.ics and socio-economic indicators. Landscape metricsed by landscape ecologists identify landscape forms.l approach is to analyse maps of land use or cover torm parameters such as fragmentation or edge density.related indicators for measuring urban form are alsothe literature, e.g. populationnumber, populationden-administrative area of the city. Finally, some studiesh types of indicators in parallel. In the following, theon studies for the different approaches are described.

s on indicators that can possibly be applied to a largecities (excluding the indicators by Galster et al., 2001)applicable for a city as awhole, allowing different cities

ared.

pe metrics. Frequently used urban form indicators arerised in Table 1. Recent publications include the studynko et al. (2006) who analysed the form of 15 Europeanver several decades to describe their land use devel-with a focus on dispersion. Schneider and Woodcock

quantied urban sprawl for 25 metropolitan regionside, while Herold et al. (2002) described changes inand uses for Santa Barbara, CA. Indicators on the frac-ension identify gradients within a city and aggregatedor clustering (Longley and Mesev, 2000; Mesev et al.,homas et al., 2008).conomic indicators. Socio-economic indicators used inpublished studies are summarised in Table 2. In her

f 25 English cities, Burton (2002) analysed compactnessocio-economic indicators in the categories of density,uses, and intensication. Huang et al. (2007) com-7 cities around the world regarding ve dimensionsn form: complexity, centrality, compactness, poros-density. The authors related these landscape metrics

o-economic indicators such as the GDP per capita.05) used four indicators for distinguishing compact-m sprawl in his analysis of US metropolitan areas:nsity, the degree of equal distribution, and the degreeering. These indicators use data on employment andion. Finally, Tratalos et al. (2007) linked urban formng socio-economic indicators) with biodiversity poten-ecosystem services for ve cities in the UK.

nd organisation of the study

all aim of this study is to characterise and classify Euro-according to their urban form. It builds upon existinggreatly extends the number of cities analysed quantita-idering the large number of indicators in the literature,ds of choosing indicators are feasible: either to select aindicators based upon ex ante assumptions or to anal-range of indicators and determine themost appropriatempirically. Most of the studies on urban form builder small set of indicators chosen beforehand. In only

ications was the appropriateness of the selected indi-sed by checking relationships to alternative indicators

). However this is exactly what could help to thin outf indicators and exclude indicators in a comprehensi-

. Accordingly, the methodological aim of this study islly reduce the number of indicators for urban form inanalysing the relationships among urban form indica-

N. Schwarz / Landscape and Urban Planning 96 (2010) 2947 31

Table 1Landscape metrics for urban form in the literature.

Indicator Measurement Interpretation Source Included

Size of continuous area [areacont]

Spatial extent of continuousarea [km2]

The absolute extent of continuous areaindicates the size of the dense sealedurban area.

Input for various indicators. Yes

Size of discontinuous area[area discont]

Spatial extent of discontinuousarea [km2]

The absolute extent of discontinuousarea indicates the size of the less densesealed urban area.

Related to area cont. Yes

Size of total area [area total] Spatial extent of whole city[km2]

The absolute extent of the city areaindicates the size of city in itsboundaries.


Size of sealed urban area [areaurban]

Spatial extent of sealed urbanarea [km2]

The absolute extent of urban areaindicates the size of sealed urban areain the city.

Schneider and Woodcock (2008) Yes

Area weighted mean patchfractal dimension [AWMPFD][MPFD]

Area weighted mean patchfractal dimension

AWMPFD indicates the raggedness ofthe urban boundary. It approaches 1 forsimple forms and 2 for complex forms.

Herold et al. (2002) Yes

AWMPFD =i=Ni=1

2 ln0.25pi/ ln si

N si

i=Ni=1

si

Huang et al. (2007)

si and pi are the area andperimeter of patch i, and N isthe total number of patchesMPFD: without weighting ofareas

Area weighted mean shapeindex [AWMSI][MSI]

Area weighted mean shapeindex

AWMSI indicates the regularity of thepatches. It equals 1 for circular featuresor square cells and increases withirregularity.

Huang et al. (2007) Yes

AWMSI =

i=Ni=1

pi/4

si

N si

i=Ni=1

si

si and pi are the area andperimeter of patch i, and N isthe total number of patchesMSI: without weighting ofareas

Centrality index [centrality] Centrality index =N1i=1

Di/N1

R =

n1i=1

Di/N1

S/

The centrality index indicates theaverage distance of sealed urbanpatches with respect to the largestsealed urban patch.


Di is the distance of centroid ofpatch i to centroid of thelargest patch, N is the totalnumber of patches, R is theradius of a circle with area of s,and s is summarization area ofall patches

Compactness index [CI] Compactness index The CI indicates compactness and ishigher for more regular landscapeswith a low number of patches.


CI =

iPi/pi

N2=

i2

si//pi

N2

si and pi are the area andperimeter of patch i, Pi is theperimeter of a circle with thearea of si and N is the totalnumber of patches

Compactness index of thelargest patch [CILP]

Compactness index of thelargest patch

The CILP indicates compactness and ishigher for a more regular patch.


CILP = 2

s/p

s and p are the area andperimeter of largest patch


Table 1(Continued )


Share of continuous/residentialland [cont/resid]

Percentage of continuousresidential area over allresidential area

The share cont/resid rangesfrom 0 to 100% and indicatesthe density of residential areas.

Burton (2002) Yes*

Kasanko et al. (2006)Share of continuous/urban land

[cont/urban]Percentage of continuousresidential area over all urbanarea

The share cont/urban rangesfrom 0 to 100% and indicatesthe density of residential areas.

It is similar to cont/resid. Yes

Edge density [ED] Sum of all urban edgelengths/total sealed urban area[km/km2]

ED accounts for the length ofedge relative to the area of thepatch. A high ED indicates aragged patch.


Median patch size [MDPS] Median size of sealed urbanpatches [km2]

The MDPS indicates theaverage size of sealed urbanpatches.

Helpful for interpreting PSSD. Yes

Mean perimeter-area ratio[MPAR]

Sum of each sealed urbanpatches perimeter/arearatio/NP [km/km2]

The MPAR indicates theaverage complexity of sealedurban patches.

Related to other complexity indicators Yes

Mean patch edge [MPE] TE/NP [km] The MPE indicates the averagecomplexity of sealed urbanpatches.

Helpful for interpreting ED. Yes

Mean patch size [MPS] Mean size of sealed urbanpatches [km2]

The MPS indicates the averagesize of sealed urban patches.

Helpful for interpreting PSSD. Yes

Number of patches [NP] Number of sealed urbanpatches in city.

The NP indicates compactness. Input for various indicators. Yes

[Number of districts SCD 1/2] Number of sub-city districts inthe city as reported by cities.

The number of districts SCDindicates the administrativestructure of a city.


Patch size coefcient ofvariance [PSCV]

PSSD/MPS The higher the PSCV, the largerare the differences in patchsize between the single sealedurban patches.

Related to PSSD. Yes

Patch size standard deviation[PSSD]

Deviation from mean in patchsize for sealed urban patches[km2]

The higher the PSSD in km2,the larger are the differences inpatch size between the singlesealed urban patches.


Porosity [ROS] Ratio of open space The ROS ranges from 0 to 100%.The higher the ratio of openspace, the less area isurbanised.


ROS = SS 100%s is the summarisation of allnon-urban area, s is the totalarea

Total edge [TE] Sum of perimeters of all sealedurban patches [km]

The TE indicates the length ofall borders of sealed urbanpatches and measurescomplexity.


Share of sealed urban area[urban/area]

Percentage of sealed urbanarea of total land area

The share urban/area rangesfrom 0 to 100%. The higher theshare urban/area, the moresurfaces in the city areurbanised.


Herold et al. (2002):commercial & industrialarea/total land area; high andlow density area/total land area

Kasanko et al. (2006)

Schneider and Woodcock (2008)Contagion index Sum over all patches of all

classes of two probabilities(randomly, conditional) ofpatch (class i) being adjacent topatch (class j)

The contagion index rangesfrom 0 to 100% and gives theprobability of a patch of theclass i to be adjacent to a patchof the class j, indicating lowversus highly fragmentedlandscapes.

Herold et al. (2002) No**

Density of building Number of buildings perhectare

The number of buildings perhectare indicates the density ofsealed urban area.

Tratalos et al. (2007) No

No data


Table 1(Continued )


Fractal dimension D = 1log10R log10[

N(R)4

]

Using this denition, fractal dimension

lustere city.lues cd form

Mesev et al. (1995) No

Proportiondetached/

n of dehous

ucture

* residential: C d usethe form of seaColumn IndicColumn Sour d expthe indicator i

The papof the meth3 describessection 5 pr

2. Method

The metform indicaing steps. (cities, socioprepared (Surban formthese indicaFig. 1b). Allsoftware pa

2.1. Data

Data prothis study.and social sdevelopmefor Regionamented witofces of 2(50,000 up250,000 inhing all capitof cities. Dare reportezone, and (3encompasslarger urbaoften includthermore, fdened towith other

258 citieperiod from284 cities, 2fore includethe map in

The spatboundariesing in Urba

01 dic a

io-ec3/20availors osed asub

ing inationa of tnt) pd coe aian eis pan Eugs: al arearesoeredaturb

d in

lectio

pro2 + log10p(R )log10Rrefers to the cthroughout thand 2, with vamore clustere

D: Fractal dimensionR mean distanceN (R) cumulative, spatiallyexplicit population

semi-detachedProportion of houses that aredetached or semi-detached

The proportiosemi-detachedsettlement str

LC111 and CLC112. ** The contagion index refers to the relationship between all lanled urban patches.ator: Abbreviations in brackets are used in subsequent tables.ce: If the cell does not contain a literature citation, the indicator has not been usen the analysis is given.

er is organised as follows. Section 2 gives an overviewodology including the data sets and analysis. Sectionthe results, which are discussed in section 4. Finally,ovides a summary and conclusions.

ology

hodology for this study builds upon the review of urbantors as described above and consists of the follow-1) Appropriate data on the spatial delineation of the-economic indicators and land coverwere acquired andection 2.1, Fig. 1a). (2) A minimal set of indicators ofwas identied (Section 2.2, Fig. 1b). (3) Building upontors, citiesweregroupedandcharacterised (Section2.3,of the statistical procedures were computed using theckage SPSS version 14.

vided by the Urban Audit initiative form the basis ofThe Urban Audit initiative aims at providing economictatistical information for analysingpan-Europeanurbannt.UrbanAuditwas initiatedby theDirectorate-Generall Policy at the European Commission and is imple-h the support of Eurostat and the national statistical7 European countries. Urban Audit selected mediumto 250,000 inhabitants) and large-sized (more thanabitants) cities fromeach participating country, includ-al cities and aiming at a broad geographical dispersionata for around 300 indicators on three spatial levels

The 20econom

Socof 200latestindicatwere uues forregardacteris

Datronmethe langrammEurope1995)cover iheadinnaturawith a2000 wing thesealedmarise

2.2. Se

The

d for each city: (1) the city level, (2) the larger urban) sub-city districts (Eurostat, 2004, 2007). The city leveles the administrative boundaries of the city, while then zone reects the broader, functional urban region,ing an employment area or commuting distances. Fur-or both London and Paris, a so-called kernel has beenease the comparison of these two metropolitan areascities in Europe.s participated in the 2003/2004 Urban Audit reporting1999 to 2002. 26 Turkish cities joined in 2006. Of these31provideddata relevant for this study andwere there-d in the analysis (see the list of cities in the Annex andFig. 2).ial delineation of the cities refers to their administrative. Data on administrative boundaries for cities participat-n Audit were downloaded from the Eurostat website.

form is givemain proceeral, the aimby extractinvidual variathe inuenThere are sprincipal axproduces oritate the intwas estimaby changinrotation dofactors to bfactors, sowere extracing of populationD lies between 1lose to 2 indicatings.

No data

tached andes indicates.


No data

classes. It was excluded from this study as the focus here is solely on

licitly in the reviewed literature. In this case, the reason for entering

ata set was used because it corresponds to the socio-nd land cover data.onomic data from the Urban Audit reporting period04 for the years 19992002 were used, matching theable land cover data for Europe in the year 2000. Thebtained from Eurostat are listed in Table 3. Indicatorst the city level, in addition to population and area val--city districts. Additionally, socio-economic indicatorscome, education, et cetera were included for the char-of European cities (Tables 3 and 4).he CORINE (COoRdination of INformation on the Envi-rogramme of the European Commission were used forver information of European cities. The CORINE pro-ms at providing harmonised data on the state of thenvironment. The CORINE Land Cover (CLC) project (EEA,rt of the CORINE programme and covers data on landrope. Land cover is classied into 44 classes under verticial surfaces, agricultural areas, forest and semi-s,wetlands, andwater bodies. For this study, raster datalution of 100 per 100 meters for land cover in Europeobtained from the European Environment Agency. Dur-a preparation, urban land covers were summarised toan patches. All steps of data preparation are sum-Fig. 1a.

n of a minimal set of indicators

cedure for selecting the minimal indicators of urban

n in the upper part of Fig. 1b. The factor analysis is thedure for determining minimal indicators (A4). In gen-of a factor analysis is to reduce the available variables

g common factors that inuence all variables. The indi-bles load onto one or more of these factors, allowing

ce of all of the factors on the variables to be estimated.everal extraction methods leading to the factors. Theis factoringwas chosenas theextractionmethod,whichthogonal (and therefore uncorrelated) factors. To facil-erpretation of the extracted factors, a Varimax rotationted. It maximises the variance of the extracted factorsg the factor loadings for the variables, but due to thees not change the factors themselves. The number ofe extracted was determined by the Eigenvalues of thethat only factors with an Eigenvalue greater than oneted.


Table 2Socio-economic indicators for urban form in the literature.


Index of Dissimilarity inpopulation distribution[diss2]

Index of Dissimilarity The diss2 measures thedistribution of populationacross districts compared torespective area of the district.A diss2 close to 1 indicateslarge differences in populationdensity between districts, adiss2 close to 0 indicates aneven distribution of populationacross districts.

Tsai (2005) (the author callsthis index Gini although theGini index has anotherdenition)

Yes

ID = 0.5N

i=1

| Xi YiN number of sub-city districts(Urban Audit SCD level 2)Xi proportion of land insub-city district iYi proportion of population insub-city district i

Dwelling number [dwell] Number of dwellings for wholecity

The dwell indicates the totalsize of the population.

Related to pop. Yes

Gini coefcient of populationdistribution [Gini2]

Gini =

Ni

Nj

yiyj2N2 y

The Gini2 compares populationnumbers of all districtswithout accounting for thearea of the districts. It rangesfrom 0 to 1, the closer the Gini2is to 1, the more unevenness.

Tsai (2005) Yes

N number of sub-city districts(Urban Audit SCD level 2)y mean of population densityin all sub-city districts

Household number [hh] Number of households forwhole city

The hh indicates the total sizeof the population.

Related to pop. Yes

Density of housing [hh/area] Number of households per area[1/km2]

The number of hh/areaindicates density within thewhole city.

Burton (2002) Yes

Burton (2002): percentage oftotal housing stock made up ofhigher density dwellings(terraces, ats,conversions)/lower densitydwellings (detached andsemidetached)/represented bysmall dwellings (1-3rooms)/large dwellings (7 ormore rooms)

Tratalos et al. (2007)

Density of housing in urbanland [hh/urban]

Households per sealed urbanarea [1/km2]

The number of hh/urbanindicates density in the sealedurban area.

Burton (2002) Yes

Population number [pop] Number of inhabitants forwhole city

The pop indicates the total sizeof the population.

Tsai (2005) Yes

Density of population [popdens]

Total population/area [1/km2] The pop dens indicatespopulation density for thewhole city.


Tratalos et al. (2007)Tsai (2005)

Density of population in urbanland [pop dens urban]

Population per km2 of urbanland [1/km2]

The pop dens urban indicatespopulation density in thesealed urban area.

Burton (2002) Yes

Burton (2002): +population perkm2 of residential urban land

Kasanko et al. (2006)

Schneider and Woodcock(2008)

Sealed urban area per person[urban/capita]

Sealed urban area/person[m2/person]

The amount of urban/capitaindicates the urban surfacerelated to population number.

Kasanko et al. (2006) Yes

Car availability [cars] Vehicles/1000 inhabitants Car availability indicateswelfare and transportstructure.

Huang et al. (2007) Yes#

GDP per capita [GDP/capita] GDP per capita [D ] The GDP/capita indicateseconomic welfare.

Huang et al. (2007) Yes#


Table 2(Continued )


Proportion higher education[prop high education]

Proportion of population withhigher education [%]

The prop high education indicatesthe level of education in thepopulation.

Tratalos et al. (2007) Yes*,#

Originally in Tratalos et al. (2007):Proportion of residents classied insocial group AB (more afuent andprofessionally qualied sectors ofsociety in UK national census)

IT availability [PC in hh] [wwwin hh]

Share of households with PC andInternet availability [%]

IT availability indicates availabilityof current infrastructure.

Huang et al. (2007) Yes**,#

Originally in Huang et al. (2007):Telephone lines/1000 people

Density of addresses Number of addresses per area The number of addresses perhectare indicates density.


No dataDensity of buildings with

addressesNumber of buildings with one ormore associated addresses per area

The number of buildings perhectare indicates density.


No dataMix of use Provision of facilities The provision of facilities indicates

the supply of residents with avariety of infrastructure andamenities.

Burton (2002) No

- number of key facilities(newsagent, restaurant or caf,takeaway, food store, bank orbuilding society, chemist, doctorssurgery) for every 1000 residents

No data

- number of newsagents for every10,000 residentsHorizontal mix of uses:- percentage of postcode sectorscontaining fewer than two keyfacilities/four or more keyfacilities/six or more keyfacilities/all seven key facilities- variation in the number offacilities per postcode sector- overall provision and spread ofkey facilities: variation in thenumber of facilities per postcodesector divided by the averagenumber of facilities per sectorVertical mix of uses- living over the shop: area of retailspace that includesaccommodation, as percentage oftotal retail space- mixed commercial or residentialuses: number of purpose-built atsin commercial buildings, as apercentage of all purpose-builtats

Moran coefcient Moran =

N

Ni=1

Nj

Wij (XiX)(XjX)(N

i=1

Nj=1

Wij

)(XiX)2

The Moran coefcient ranges from1 to +1, with a value close to +1indicating that high-densitydistricts are closely clustered, avalue close to zero meaningrandom scattering and a valueclose to 1 representing achessboard pattern ofdevelopment.

Tsai (2005) No

N number of sub-city districts No dataXi population or employment insub-city district iXj population or employment insub-city district jX mean of population oremploymentWij weighting (=distance) betweensub-city district i and j

* Education level instead of social groups.** PC and Internet availability instead of telephone lines.# Indicators for socio-economic characterisation of cities. Column Indicator: Abbreviations in brackets are used in subsequent tables. Column Source: If the cell does

not contain a literature citation, the indicator has not been used explicitly in the reviewed literature. In this case, the reason for entering the indicator in the analysis is given.


Fig. 1. (a)Methodology: Data preparation. P1. Urban Audit administrative boundaries of cities were revised. Distinct polygons for the same cityweremerged if these featureswere separated by small water bodies, roads et cetera. Polygons with holes were lled up. P2. Administrative boundaries were used to cut out CORINE land cover raster data.P3. These raster data were converted into features. P4. The CLC classes mentioned in the note were merged into a single class of sealed urban patches. This procedure ofsummarising urban land covers into a single category for analysis is in line with other research on urban form (Huang et al., 2007; Galster et al., 2001). P5. Landscape metricsfor this urban class were computed using the open source software Patch Analyst (Rempel, 2008). Patch Analyst is an extension for the ArcGIS Software and provides themost common landscape metrics for landscape analysis. Landscape metrics were computed using the class level procedure and the joined sealed urban area as input. Classlevel procedure implies referring to the sealed urban patches only and not to all patches in the city. P6. Additionally, ratios like the share of sealed urban patches comparedto the size of the whole city as well as Index of Dissimilarity and Gini-coefcient were computed. The latter two were computed for Urban Audit SCD level 2 (sub-districtspossibly created for UrbanAudit to bemore comparable). (b)Methodology: Analysis. A1. Linear correlations among (1) population-related and (2) landscapemetrics aswell as(3) between population-related and landscape metrics were computed to reveal similarities between indicators. Spearmans Rho rs was used as correlation measure becausenon-linear relationships between indicators were identied. A2. To reduce the number of indicators entering the factor analysis, indicators of urban form were omittedfrom further analysis, if they show strong correlations (absolute correlations >.8) to other indicators in the data set. A3. All variables were normalised (Z-transformation), sothat the mean for each indicator equals 0 and the standard deviation equals 1. A4. A factor analysis was computed to identify underlying patterns of relationships betweenindicators. A5. Indicators which best represent the extracted factors are included into the minimal indicator set for urban form.

Table 3Data out of Urban Audit used for this study.

Related to indicator Name in Urban Audit Urban Audit code Level No. of valid cases(total: N=231)

Population-related indicatorsArea total Total land area (km2) according to cadastral register en5003i City 209Dwell Number of dwellings sa1001i City 197hh Total Number of Households de3003i City 219Pop Total Resident Population de1001i City 228Pop Total Resident Population de1001i SCD 2 197Popdens Population density total resident population per square km en5101i City 208Needed for diss2 Total land area (km2) according to cadastral register en5003i SCD 2 124

Socio-economic indicators for characterising citiesCars Number of registered cars per 1000 population tt1057i City 181GDP/capita GDP per head ec2001i City 208

PC in hh Proportion of households with a PC it1001i City 63Prop high education Proportion of the resident population qualied at levels 56 ISCED te2022i City 196www in hh Percentage of households with Internet access at home it1005i City 68

See Tables 1 and 2 for abbreviations of indicators.


Fig. 2. Map of nistraEurostat. Eur

2.3. The Ch

Theprocminimal seand consistanalysis.

A6. In orform, a cluswas compuapplied witsure. The variance exmined the noptimal numgain of addisolution wi

A7. Toone-way anindicator separameterswere used aous aspectstheGDPpercar ownershand Interne

3. Results

3.1. Reduct

Correlatshown in tthe landsca

e td lanatedmons andEuropean cities analysed in this study with respective cluster type. Sources: AdmioGeographics and ESRI for the administrative boundaries of countries.

aracterisation of European cities

edure of characterising European cities according to thet of indicators is displayed in the lower part of Fig. 1bs of the two steps (A6) cluster analysis and (A7) variance

der to compare European cities according to their urban

the samtors anelabortions ametricter analysis with the identied minimal indicator setted. The hierarchical clustering (Ward procedure) wash the squared Euclidian distance as the distance mea-elbow-criterion, which focuses on the percentage ofplained as a function of the number of clusters, deter-umber of clusters. According to this rule of thumb theber of clusters is the number after which the marginal

ng onemore cluster drops sharply. For this analysis, theth eight clusters was chosen.

characterise the cities belonging to each cluster, aalysis of variance was computed, using the minimalt as dependent variables and the clusters as group. Additionally, the following socio-economic indicatorss dependent variables to gain an impression of the vari-of humanwelfare (Huang et al., 2007) in these clusters:capita, proportionof populationwithhigher education,ip per 1000 inhabitants, PC availability in households,t availability in households.

ion of indicators due to correlations

ions among the population-related indicators arehe lower left part of Table 5, and correlations amongpe metrics are presented in the upper right part of

3.1.1. CorreRegardin

lowing indi(1) The sizewith size ofinate the sidiscontinuoto leave theThe ratio ofareawas debecause it isof a ratio. (3residentialof the contof the contshare of opratio of theperimeter-ato themediof the ratio.the shape cop< .001). AWto other tharea, the Mfore, these utotal edgep< .001), shber of sealetive boundaries and location of Urban Audit cities downloaded from

able. Correlations between population-related indica-dscapemetrics are depicted in Table 6. These results arein the following paragraphs, differentiated into correla-g the landscape metrics, among the population-relatedbetween these two.lations among the landscape metricsg correlations among the landscape metrics, the fol-

catorsofurban formwereomitted fromfurther analysis.of urban areawas excluded due to its close relationshipthe discontinuous urban fabric. The choicewas to elim-ze of urban area because it is the sum of the size of theus and continuous urban fabric, and it was consistenttwo more detailed measurements in the analysis. (2)the share of the continuous urban fabric to the urbanleted in favour of the size of the continuous urban fabricmore straightforward to use the absolute value instead) The ratio of the share of continuous urban fabric to thearea was removed due to its relationship with the sizeinuous urban fabric. The same reason as for the shareinuous urban fabric to the urban area applies. (4) Theen space was excluded because it is the inverse of theshare of the urban area to the total area. (5) The meanrea ratio was omitted because of its close relationshipanpatch size, in order to keep the absolute value instead(6) The two metrics, AWMPFD and AWMSI, measuringmplexity and irregularity arehighly correlated (rs = .89,MPFD was omitted because it has lower correlations

e indices compared to AWMSI. Without weighting thePFD and MSI indicators are not correlated, and there-nweighted indicators were no longer analysed. (7) The

and number of patches are highly correlated (rs = .83,owing the geometrical relationship between the num-d urban patches and the sum of all the perimeters of


Table 4Descriptive results for urban form indicators.

Indicator N Min Max Mean SD

Landscape metricsArea contArea discoArea totalArea urbanAWMPFDAWMSICentralityCICILPCont/residCONT/urbED [km/kmMDPS [kmMPAR [kmMPE [km]MPFDMPS [km2

MSINPNumber dNumber dPSCVPSSD [km2

ROS [%]TE [km]Urban/are

Population-rArea totaldiss2Dwell [100Gini2hh [1000]hh/area [nhh/urbanPop [1000Popdens (Popdens [Popdens uUrban/cap

Socio-econoCars [no. pGDP/capitPC in hh [%Prop highwww in h

SD: standard d

all the sealefurther anaedge. (8) Threlation witbecause praarea than w(9) A closethe patch siomitted froto understathe number

3.1.2. CorreRegardin

cators, thefrom furthethe populadwellings, ationship iswho foundthe househ[km2] 231 0.0unt [km2] 231 0.0[km2] 231 18.8[km2] 231 7.3

231 1.2231 1.5230 0.4231 0.0231 0.1

[%] 231 0.0an [%] 231 0.0

2] 231 0.12] 231 0.0/km2] 231 1.5

231 2.0231 1.2

] 231 0.3231 1.4231 1.0

istricts SCD 1 18 2istricts SCD 2 220 3

231 0.0] 231 0.0

231 5.7231 23.9a [%] 231 2.0

elated indicators(UA) [km2] 209 23.4

124 0.00] 197 15.0

197 0.0219 11.8

o. per km2] 219 22.6[no. per km2] 219 764.5] 228 30.8UA) [no. per km2] 205 44.9no. per km2] 228 45.0rban [no. per km2] 228 1800.2ita [m2 per capita] 228 40.4

mic indicators for characterising citieser 1000 inhabitants] 181 186.2a [D per capita] 208 2513.0] 63 8.6education [%] 196 4.8h [%] 68 1.3

eviation. SCD: Sub-city district in Urban Audit. (UA): value as reported by cities in Urban

d urban patches. Only the latter indicator was kept forlysis because it is even easier to measure than the totalemeanpatch edgewas excludedbecause of its high cor-h the mean patch size. The mean patch size was keptctitioners are more used to working with the surfaceith the perimeter of a patch, e.g. when describing it.relationship exists between the mean patch size andze standard deviation (rs = .85, p< .001). The latter wasm further analysis because themean patch size is easiernd. (10) The compactness index was omitted instead ofof patches because the latter is easier to measure.

lations among the population-related indicatorsg the correlations among the population-related indi-following indicators of urban form were omitted

r analysis. (1) The three population-related indicators,tion number, number of households and number ofre closely related with rs > .98 (p< .001). This close rela-in accordance with ndings by Tratalos et al. (2007),a high (also Spearman rank) correlation of .97 betweenold density and population density for ve UK cities.

Accordinglywere omittber is moreThe same hholds per tarea. (2) Thnumber divreported pothe absoluttion densityarea was omarea per caare presum

3.1.3. Correpopulation-

Only twbetween th(1) The indcities in therelationship89.3 6.6 12.4923.6 52.6 76.6

2324.8 248.9 277.71093.6 77.7 96.1

1.4 1.3 0.011.8 3.6 1.3

273.1 8.6 27.30.6 0.1 0.10.8 0.3 0.2

100.0 14.4 21.981.0 10.6 16.22.4 1.0 0.4

44.9 0.8 3.742.2 21.8 7.259.6 9.9 6.61.3 1.3 0.0

44.9 4.2 4.92.4 1.6 0.2

124.0 26.7 23.337 13.3 10.2

624 23.5 45.7785.9 322.6 134.7132.2 11.9 12.398.0 61.9 21.0

1512.9 206.1 183.1

94.3 38.1 21.0

2317.0 253.8 278.90.7 0.4 0.1

1869.9 180.6 224.10.4 0.2 0.1

3016.0 190.2 290.010 290.4 968.2 1158.312 931.3 2305.4 1338.6

7172.1 414.7 636.615 246.9 2049.0 1769.020 515.3 2242.6 2684.224 745.3 5430.8 3273.4

555.5 226.3 91.1

698.6 395.8 105.069 875.0 22 389.7 13 118.7

67.0 37.2 15.032.3 16.1 5.263.0 29.2 14.0

Audit. See Tables 1 and 2 for abbreviations of indicators.

, the number of dwellings and number of householdsed from further analysis because the population num-often used in the literature and is available more often.olds for the two indicators, the proportion of house-otal area and the proportion of households per urbane computed population density (reported populationided by area of the city) was favoured instead of thepulation density for Urban Audit because more data one population numbers were available than for popula-. (3) The indicator of the population density per urbanitted because of its inverse relationship to the urban

pita. The latter was retained because per capita valuesably easier to communicate to practitioners.

lations among the landscape metrics andrelated indicatorso indicators were excluded because of the correlationse landscape metrics and population-related indicators.icator of the area of the city (reported) as reported byUrban Audit data was excluded because it has a tight(rs = .97, p< .001) to the landscape metric area of the

N.Schw

arz/Landscape

andUrban

Planning96 (2010) 2947

39

Table 5Correlations among socio-demographic indicators and landscape metrics respectively.

Area total Areaurban

Urban/area

Discont Cont Cont/urban

Cont/resid

ROS AWMSI MSI MPAR MPFD AWMPFD TE ED MPE MPS NP MDPS PSCOV PSSD CILP CI Centrality

1.00 .64*** .37*** .57*** .26** .05 .03 .37*** .37*** .13 .04 .05 .31*** .80***.45*** .18** .18** .76*** .01 .58*** .13 .36*** .77*** .24*** Areatotal

1.00 .40*** .91*** .35*** .13 .12 .40*** .65*** .04 .23*** .15 .35*** .88*** .16 .29*** .40*** .56*** .21** .66*** .74*** .31*** .56*** .04 Areaurban

1.00 .37*** .19** .01 .04 1.00*** .29*** .09 .36*** .25*** .00 .06 .76*** .60*** .74*** .27*** .25*** .03 .72*** .08 .28*** .27*** Urban/area1.00 .11 .33*** .34*** .37*** .68*** .03 .19** .14* .39*** .82*** .18** .31*** .39*** .50*** .16* .61*** .70*** .34*** .50*** .03 Discont

Area total (UA) 1.00 1.00 .85*** .85*** .19*** .19** .05 .11 .05 .06 .26*** .01 .12 .20** .16* .13* .23*** .30*** .10 .16* .00 Conthh .50*** 1.00 1.00 1.0*** .01 .10 .04 .04 .03 .08 .16**.09 .02 .00 .10 .09 .07 .05 .05 .10 .00 Cont/urbanDwell .51*** .99*** 1.00 1.00 .04 .11 .05 .03 .02 .00 .14* .10 .05 .02 .07 .08 .06 .07 .05 .07 .02 Cont/residPop .49*** .99*** .99*** 1.00 1.00 .29*** .09 .36*** .25*** .00 .06 .76*** .60*** .74*** .27*** .25*** .03 .72*** .08 .28*** .27*** ROSPopdens (UA) .34*** .60*** .58*** .58*** 1.00 1.00 .21** .29*** .35*** .89*** .67*** .34*** .34*** .26*** .36*** .29*** .66*** .59*** .45*** .36*** .01 AWMSIPopdens .30*** .60*** .58*** .58*** .97*** 1.00 1.00 .35*** .03 .28*** .04 .16 .62*** .36*** .37*** .43*** .38*** .18** .14 .34*** .14 MSIPopdens Urban .15 .28*** .32*** .36*** .55*** .57*** 1.00 1.00 .87*** .22*** .15 .31*** .11 .03 .18** .86*** .36*** .23*** .03 .15 .03 MPARUrban/capita .15 .28*** .32*** .36*** .55*** .57*** 1.00*** 1.00 1.00 .38*** .16 .33*** .11 .07 .19** .74*** .33*** .12 .14 .17** .03 MPFDhh/area .23*** .64*** .63*** .63*** .95*** .98*** .50*** .50*** 1.00 1.00 .54*** .26*** .12 .06 .37*** .24*** .56*** .22*** .48*** .39*** .07 AWMPFDhh/urban .10 .45*** .46*** .46*** .63*** .65*** .91*** .91*** .65*** 1.00 1.00 .06 .01 .01 .82*** .16 .78*** .40*** .44*** .83*** .16 TEdiss2 .45*** .00 .00 .04 .38*** .29** .17 .17 .35*** .31*** 1.00 1.00 .29*** .30*** .09 .22*** .11 .31*** .02 .10 .17** EDGini2 .09 .02 .06 .01 .10 .11 .10 .10 .05 .03 .09 1.00 1.00 .93*** .53*** .22*** .30*** .73*** .00 .52*** .26*** MPE

Area total(UA)

hh Dwell Pop Popdens(UA)

Popdens Popdensurban

Urban/capitahh/area hh/urbandiss2 Gini2 1.00 .48*** .08 .20** .85*** .09 .49*** .26*** MPS

1.00 .22*** .81***.07 .36***1.00*** .27*** NP1.00 .41***.15 .04 .19** .03 MDPS

1.00 .27*** .38*** .80*** .18** PSCOV1.00 .12 .07 .17** PSSD

1.00 .38*** .37*** CILP1.00 .26*** CI

1.00 Centrality

* p


Table

6Cor

rela

tion

sbe

twee

nla

ndsc

apem

etrics

and

pop

ula

tion

-rel

ated

indices

(Spea

rman

sRho)

.

Are

ato

tal

Are

aurb

anUrb

an/

area

Disco

nt

Con

tCon

t/urb

anCon

t/re

sid

ROS

AW

MSI

MSI

MPA

RM

PFD

AW

MPF

DTE

EDM

PEM

PSNP

MDPS

PSCOV

PSSD

CIL

PCI

Cen

tral

ity

Are

ato

tal

(UA)

.97*

**.6

2***

.30

***

.55*

**.3

6***

.08

.10

.30*

**.3

6***

.12

.06

.07

.30*

**.7

9***

.39

***

.16

.15

.74*

**.0

0.5

5***

.14

.36

***

.74*

**.2

3***

hh

.45*

**.8

8***

.55*

**.7

4***

.49*

**.0

7.0

7.

55**

*.5

6***

.05

.24*

**.1

0.2

3***

.70*

**.2

5***

.39*

**.5

3***

.38*

**.

20**

.53*

**.7

8***

.24

***

.37*

**.

07Dw

ell

.48*

**.8

8***

.49*

**.7

3***

.52*

**.1

5*.1

5*.

49**

*.5

6***

.08

.29*

**.1

7.2

6***

.73*

**.2

2**

.31*

**.4

6***

.44*

**.

27**

*.6

0***

.74*

**.

20**

.42

***

.02

Pop

.44*

**.8

6***

.50*

**.6

9***

.54*

**.1

4*.1

4*.

50**

*.5

4***

.05

.25*

**.1

2.2

4***

.68*

**.2

1**

.34*

**.4

8***

.37*

**.

22**

*.5

2***

.73*

**.

25**

* .3

6***

.06

Pop

den

s(U

A)

.34

***

.35*

**.8

6***

.29*

**.2

8***

.11

.08

.86

***

.28*

**.0

0.3

7***

.24*

**.0

0.0

3.6

0***

.52*

**.6

8***

.24

***

.31*

**.0

7.6

8***

.02

.26*

**.

32**

*

Pop

den

s.

39**

*.3

1***

.88*

**.2

3**

.33*

**.2

0**

.17*

.88

***

.23*

**.0

6.3

2***

.20*

*.

02.

02.6

4***

.54*

**.6

8***

.30

***

.23*

**.0

0.6

4***

.04

.31*

**.

31**

*

Pop

den

surb

an.

22**

.10

.18*

*.

24**

*.4

1***

.48*

**.4

7***

.18

**.

05.

03.0

7.0

0.

08.

23**

*.0

3.1

5.1

9**

.25

***

.10

.12

.11

.02

.26*

**.

24**

*

Urb

an/c

apita

.22*

*.1

0.

18**

.24*

**.

41**

*.

48**

*.

47**

*.1

8**

.05

.03

.07

.00

.08

.23*

**.

03.

15.

19**

.25*

**.1

0.1

2.

11.0

2.

26**

*.2

4***

hh/a

rea

.32

***

.38*

**.9

0***

.31*

**.2

9***

.11

.08

.90

***

.27*

**.0

1.3

1***

.17

.01

.06

.64*

**.5

4***

.69*

**.

23**

* .2

5***

.06

.67*

**.0

5.2

4***

.26

***

hh/u

rban

.19

**.0

4.3

2***

.08

.36*

**.3

7***

.36*

**.

32**

*.0

2.

08.1

0.

01.

07.

13.1

4.2

2**

.30*

**.

21**

.12

.05

.24*

**.0

4.2

2**

.20

**

diss2

.37*

**.1

2.

37**

*.0

8.2

0*.1

8*.2

0*.3

7***

.02

.15

.18

.04

.11

.30*

**.

22.

19.

25**

.35*

**.1

9.2

1.

18.

37**

* .3

7***

.09

Gin

i2.1

2.0

4.

06.

05.0

7.0

6.0

6.0

6.

06.

07.

05.

04.

08.0

4.

17.

09.

05.1

0.

02.0

6.0

0.2

1**

.09

.30*

**

*p

Schwarz Urban Indicators 2010

Documents

Transcript of Schwarz Urban Indicators 2010