Measuring urban sprawl: how can we deal with it?

1 IntroductionOver the past two decades urban sprawl has become a subject of particular interestamong planners and policy makers. Critics of sprawl all over the world are con-cerned by its many alleged negative impacts, such as a lack of scale economies,which reduces the level of public services in the suburbs and weakens the economicbase of central cities; increased energy consumption through encouragement ofthe use of private vehicles, causing traffic congestion and air pollution; and irrever-sible damage to ecosystems, caused by scattered and fragmented urban developmentin open lands (Brueckner, 2000; Burchell et al, 1998; Downs, 1998; Ewing, 1997; Johnson,2001).

Not all planners, however, agree that sprawl has to be `dealt with' or restrained.Some consider it to be inevitable, harmless, or even positive. The main alleged positiveeffect of sprawl is that, being an outcome of free-market decisions made by householdsand firms, it maximizes the overall welfare of society (Gordon and Richardson, 1997).It also provides easy access to open space, both in one's own home and in the country-side, and relatively short commuting times for most of those who both live and workin the suburbs (Downs, 1998). Furthermore, sprawling patterns are usually associatedwith high income rates; hence, they have lower crime rates than do large cities.Residents in sprawling small communities have greater influence on governmentdecisions than do residents in larger cities, being better able to pass zoning regulations,and other regulations that exclude undesirable development in their neighborhoods(Burchell et al, 1998).

Common agreement exists as to its vast impact on urban landscape in Westerncountries (Hartshorn and Muller, 1992), as does a consensus on its ambiguity and itslack of an accurate definition and measures (Ewing et al, 2002; Galster et al, 2001).In other words, we know sprawl is significant, but we are not yet sure what it isexactly or how to measure it. These questions are, of course, crucial to future studiesthat will attempt to analyze the impact of sprawl on the urban landscape (Torrensand Alberti, 2000).

Measuring urban sprawl: how can we deal with it?

Amnon Frenkel, Maya AshkenaziFaculty of Architecture and Town Planning and Center for Urban and Regional Studies,TechnionöIsrael Institute of Technology, Haifa, Israel; e-mail: [email protected],maya [email protected] 25 November 2005; in revised form 21 August 2006; published online 1 October 2007

Environment and Planning B: Planning and Design 2008, volume 35, pages 56 ^ 79

Abstract. Measuring urban sprawl is a controversial topic among scholars who investigate the urbanlandscape. This study attempts to measure sprawl from a landscape perspective. The measures andindices used are derived from various research disciplines, such as urban research, ecologicalresearch, and fractal geometry. The examination was based on an urban land-use survey performedin seventy-eight urban settlements in Israel over the course of fifteen years. Measures of sprawl werecalculated at each settlement and were then weighted into one integrated sprawl index through factoranalysis, thus enabling a description of sprawl rates and their dynamics over a time period ofapproximately two decades. The results reveal that urban sprawl is a multidimensional phenomenonthat is best quantified by various measures.

DOI:10.1068/b32155

On the basis of the literature it is clear that urban sprawl is a complex phenomenonthat, besides not being defined unequivocally, is difficult to quantify and measureaccurately. Hence, moving from `sprawl' to compact' form is more likely to be adirection in a continuum rather than across fixed and measurable categories. Thusour purpose in this paper is to address the question of how urban sprawl can bemeasured. Several studies in recent years have tried to deal with this question bysuggesting a variety of urban sprawl measures (Galster et al, 2001; Malpezzi andWen-Kai, 2001; Torrens and Alberti, 2000). Although some of these measures weretested in empirical studies, many of them still remain theoretical or tested only in thecontext of the ecological research discipline, but not applied in urban research.

This paper introduces empirical results obtained by implementing various meas-ures of sprawl that are suggested in the literature, and investigates the most efficientway to use them to measure the phenomenon. Further, we identify the relationshipsamong the various sprawl measures and the changes in their values over the periodinvestigated, as obtained from a large sample of urban settlements. Both the static andthe dynamic development patterns of sprawl in these urban settlements are described.

Notwithstanding the copious discussion on sprawl in the literature, the absence ofempirical evidence in support of the various positions is still clear. Therefore, empiricalstudies, like the current one, are important in light of the ideological and practicalargument that exists between those who oppose and those who side with this phenom-enon. Empirical studies contribute to our understanding of the essence of sprawl,bringing quantitative knowledge into the discussion and suggesting possible solutions(Batty et al, 1999). Without this knowledge, ideological and practical discussion onurban sprawl and the effectiveness of a growth-management policy remains only inthe conceptual and speculative realm (Torrens and Alberti, 2000). In the next sectionwe present various measures of sprawl gathered from a literature review. Section 3defines the framework of this study; the data and the region investigated are describedin section 4. The main empirical results are contained in section 5. Finally in section 6we present the conclusions.

2 Measures of sprawlThe sprawl phenomenon stimulates valuable dispute among scholars and urban plan-ners, since there is no acceptable uniform definition of its characteristics (Johnson,2001). Many researchers point to the lack of a single, clear, accurate, full, and quanti-tative definition of urban sprawl as the major problem relating to this issue (Burchellet al, 1998; Ewing et al, 2002; Galster et al, 2001; Hadly, 2000; Johnson, 2001; Malpezzi,1999). The most common definition of sprawl was given by Ewing (1994; 1997), whoargued that we still lacked a clear definition of the term ùrban sprawl' with which wecould work, and suggested five prominent characteristics of sprawl: (1) a scattered anddiscontinuous pattern of development, which leaves nonused spaces in the built-up area;(2) development of residential areas with low densities, which creates extensive expansionof single dwelling units with their own private courts and causes an absence of publicopen space; (3) commercial strip development alongside the main transportation axes;(4) segregation of land uses, which separates residential areas from other urban landuses and removes urban functions from each other; (5) low accessibility and highdependency on private vehicles, mainly because of land-use segregation.

There are several patterns or types of sprawl. The phenomenon of urban sprawl, oftenreferred to as suburbanization, is defined as a movement of residents outside big cities tothe suburbs (to places with only residential land use). This movement started as far backas at the end of the industrial era, and it has continued since throughout the world,but especially in Western countries (Belser, 1960; Gans, 1967; Harvey and Clark, 1965;

Measuring urban sprawl 57

Jackson, 1985; Mills and Hamilton, 1989; Real Estate Research Corporation, 1974). Morerecent patterns of sprawl are the American edge city' (Garreau, 1991), which features thesuburbanization of residents, employment, and retail trade, and the European functionalurban region; these patterns represent the disurbanization phase in the cyclic model of thedevelopment process of metropolitan regions (Klaassen et al, 1981; Van Den Berg et al,1982).

In fact, sprawl has been conceptualized in recent studies as a multidimensionalphenomenon that requires a different set of measures for each dimension (Ewing et al,2002; Galster et al, 2001; Torrens and Alberti, 2000). Most sprawl measures suggestedin the literature can be divided into five major groups: growth rates, density, spatialgeometry, accessibility, and aesthetic measures.

2.1 Growth ratesIn terms of growth rates, urban sprawl is defined as a condition in which populationgrowth rates in the suburbs are higher than inside the central city (Jackson, 1985).Another popular growth-rate measure is the `sprawl index' (SI) or `sprawl quotient',defined as the ratio between the growth rate of built-up areas and the populationgrowth rate in those areas. A quotient higher than 1 implies urban sprawl (Hadly,2000; Weitz, 2000).

2.2 DensityThis is the most popular sprawl measure (Galster et al, 2001) and, some will argue,the one that best represents the phenomenon (Maret, 2002). There are various typesof density, as well as many ways and scales to measure it (Burton, 2000; Chin, 2002;Churchman, 1999). Density is defined as the ratio between the amount of a certainurban activity and the area within which it takes place. Urban activity can be definedas the amount of residential units, number of residents, or employees (Razin andRosentraub, 2000).

In terms of density, sprawl is defined as a condition in which density is relativelylow or decreases during a certain time period. Another popular density measurethat quantifies the latter definition is the density gradient, which is the constant inthe negative exponential model (Alperovich, 1995; Batty and Longley, 1994). Over thepast few decades density gradients have been falling constantly in developed as well asdeveloping countries (Ingram, 1998) which well proves the universality of urban sprawl.

2.3 Spatial geometryThis constitutes the largest group of measures. There are numerous geometricmeasures, most of which have been adopted from ecological research (McGarigal andMarks, 1995; Turner, 1989) or from fractal geometry (Batty and Kim, 1992; Battyand Longley, 1994; Herold and Menz, 2001; Torrens and Alberti, 2000). Geometric orecological measures quantify two main characteristics of the urban landscape: config-uration and composition. Configuration refers to the geometry of the urban built-uparea, and composition refers to its level of heterogeneity. An urban area will beconsidered sprawling as long as its geometric configuration is irregular, scattered, andfragmented, and its land-use composition more homogenous and segregated.

Some common measures in this category are leapfrog or continuity measures(Galster et al, 2001), measure of circularity (Gibbs, 1961), fractal dimension, andmean patch size, M (Batty and Kim, 1992; Batty and Longley, 1994; Benguigui, 1995;Herold and Menz, 2001; Torrens and Alberti, 2000)öall of these quantify the level ofscatter and fragmentation of the urban landscape. The percentage of different landuses quantifies the level of heterogeneity of the landscape (Frenkel, 2004a; McGarigaland Marks, 1995).

58 A Frenkel, M Ashkenazi

2.4 AccessibilitySprawl is defined as a condition of poor accessibility, followed by the massive use ofprivate vehicles (Ewing, 1994; 1997; Ewing et al, 2002), or, as Al Gore very simply putit À gallon of gas can be used up just driving to get a gallon of milk.'' (1) Accessibilitycan be quantified by measuring road lengths, road areas, and the traveling times ofhouseholds (Hadly, 2000). Another way to measure accessibility is to calculate the fractaldimensions of road networks (Benguigui, 1995). Some ecological measures are useful formeasuring accessibility, such as the `mean proximity index' (Gustafson, 1998; Torrens andAlberti, 2000). Another group of accessibility measures is used in transportation models,such as the gravity or logit models (Torrens and Alberti, 2000).

2.5 Aesthetic measuresSprawl is often considered a boring, homogeneous form of development (Fulton, 1996;Gordon and Richardson, 1997). Being subjective by definition, it is difficult to measureand quantify the aesthetics of sprawl. However, several recent studies have attemptedto define archetypes of urban development or sprawl, such as residential sprawl orstrip-mall sprawl, and to compare various landscapes with those archetypes. It seemsthat much work is still needed in this area (Torrens and Alberti, 2000).

3 Methodology3.1 Research frameworkThis study is a first attempt to measure urban sprawl in Israel and its dynamics overthe past two decades from a landscape perspective. Urban sprawl in Israel startedtwo decades ago, but has not yet been empirically measured or characterized. Majorprocesses that influenced sprawl in Israel were the rise in the standard of living andin the residential floor area per person, consumer preference for low-density single-family housing in the suburbs, and the arrival of one million immigrants from theformer USSR during the 1990söall of which led to a massive transformation ofagricultural land into urban land uses all over the country, and perhaps to sprawl-likepatterns of development (Gonen, 1995; 1996; Schiffman, 1999). Scholars and plannersin Israel are concerned with the negative impacts of sprawl and its leading to wastefulland consumption, especially given Israel's unique condition as a small country withlimited land resources, but at the same time with relatively high population growthrate, in contrast to other Western countries (Frenkel, 2004b). Hence, measuringsprawl in Israel is crucial to a better management of its land resource.

Two major issues were investigated within the framework of this study. The firstconcerns the questions: how can we measure urban sprawl and is urban sprawla measurable phenomenon? The second issue concerns the question: does urbandevelopment in Israel comply with the definition of sprawl as we know it?

Based on a review of the literature, we hypothesize that sprawl is a complexphenomenon that cannot be measured by only one or two measures, as is often donein many urban studies. Hence, we assume that its various dimensions are independentand not significantly correlated with one another. In order to examine this hypothesis,measures of sprawl were defined, calculated, and compared, enabling the descriptionof sprawl rates and its dynamics over the time period investigated. The results of thisstudy provide a rather comprehensive and useful database on urban sprawl in Israel, aswell as a better understanding of what sprawl is and how it can be measured from alandscape perspective.

(1) Quote from a speech by Al Gore during his campaign for the US presidency, January 1999(http://www.greenclips.com/00issues/139.htm).


3.2 The sampleIn order to answer the investigating questions, a sample of seventy-eight settlementswas included in the land-use field survey, representing 67% of all Jewish urbansettlements in Israel.(2) The sample covers about 4.76 million residents, or 78% of allresidents in urban settlements in the country and 71% of the entire Israeli population(CBS, 2004). The urban built-up areas that were mapped in this survey add up to624:5 km2, constituting 55% of all built-up areas in Israel (Frenkel, 2004b). The sampleincludes settlements from all six districts and all four metropolitan regions of Israel, aswell as settlements of different metropolitan functionality. It includes all four metro-politan core cities in Israel (Jerusalem, Tel-Aviv, Haifa, and Beer Sheva), 96% of allurban settlements located within the inner and middle rings of the four metropolitanregions in Israel as defined by Israel's Central Bureau of Statistics (CBS), and 72% ofthe edge cities and towns in peripheral regions. Settlement sizes vary from small townsof 2000 ^ 5000 residents to medium-size towns of up to 100 000 residents and largetowns with more than 100 000 residents. The sample contains almost all Jewish urbansettlements above 20 000 residents, half of the urban settlements with 10 000 ^ 20 000residents, and one fourth of the urban settlements with fewer than 10 000 residents.The sample, therefore, well represents all Jewish urban settlements from all perspectives:spatial location, metropolitan functionality, and size.

3.3 Unit of investigationResearchers tend to measure urban sprawl either on a metropolitan scale, as mostlyused in the US (Ewing et al, 2002; Razin and Rosentraub, 2000; Wolman et al, 2004), oron smaller scales of towns or neighborhoods, mostly used in Europe (Batty and Longley,1994; Burton, 1996; 2000; Hasse and Lathrop, 2003).We chose to investigate sprawl on atown scale. This decision was based on the availability of the data used in our study, aswell as on the investigation perspective. The selection of this investigation unit enabled usto obtain a large database, sufficient for examining statistical differences in the sprawlvariables between the selected urban settlements.(3) Likewise we were able to test theexistence of this phenomenon in a large variety of types of urban settlements.

In order to examine the spatial evolution of sprawl in Israel we used the availablebasic survey from the mid-1980s carried out by the Israeli CBS as a starting point totest the dynamic of the sprawl phenomenon in time and space dimensions (CBS, 1997).Accordingly we adopted the same unit of investigation as that employed in the CBSsurvey and also used it in our field survey in 2002. In this way we were able to test thevariability of land-use patterns and urban configurations on a suitable time scale, inorder to compute quantitative measures of sprawl, and to compare the values obtainedfor a variety of settlements, over the course of two decades. In many cities in oursample the municipal boundaries expanded during the period investigated. Thus the

(2) Arab settlements were not included in the sample because of their unregulated development ofland uses. The development of their physical pattern did not result from controlled planning, butwas constrained by historical causes connected to the lack of statutory planning, landownershippatterns, and social norms referring to land development accepted by Arab society and expressedin a multigenerational building patterns, which justify a separate investigation. Young Arab house-holds traditionally tend to settle near their parents' households, whether in the same house or in anattached unit. This tendency leads to a situation in which there is no active housing market in theArab society, which regards land as family property and not offered for sale (since land is kept forthe next generation). This tradition causes an extensive leapfrog pattern of development, motivatedby absolutely different reasons from those that characterize sprawl pattern in Western countriesand in the Jewish settlements in Israel as well. We were also not able to include rural settlementsbecause of the lack of information regarding their land-use composition during the time periodinvestigated. However, we did include small urban settlements, which are semirural in character.(3)Unlike the USA, which has hundreds of metropolitan regions, Israel has only four metropolitan regions.


unit determined for the analysis was that of the 2002 boundaries, which included inmany cases the expanded built-up areas beyond the municipal boundaries of the 1980s.

Furthermore, in this study we set out to define the measures of sprawl that wereimplemented from a landscape perspective. Accordingly we examined the developmentof the spatial structure of the urban built-up area and identified whether this patternpresented sprawling or compact development during the period investigated. In thiscontext the study methodology was derived, to a great extent, from ecological researchthat examined landscape structure. Thus, some of the sprawl measures employed inthis study (see section 3.4) describe the aggregate configuration of the urban landscapeand its scatter. Such indices can easily be mapped and measured on continuous urbanbuilt-up areas at a town scale. Another group of indices employed in this studydescribe the segregated structure of the urban fabric and its land-use composition.This is also easy to implement on a town scale, because of the availability of dataand the ability to compare large numbers of observations (municipalities).

Hence, the unit of investigation was defined as the urban built-up area inside themunicipal border of an urban settlement (figure 1). The boundaries of the built-upareas of each of the urban settlements selected were marked on city maps scale1 :10 000 ^ 1 :12 000 and were verified through aerial photography and field surveys.Open unused land was excluded from the total area within the jurisdiction of a city.The final built-up areas, therefore, included only the areas that were in use for variouspurposes within the jurisdiction of the locality.

We divided each built-up area into two major groups: the central built-up area andthe leapfrog areas (see figure 1). The central built-up area contains most of the resi-dential land and other built-up land uses, as well as inner public open spaces or naturalopen spaces, agricultural lands, and inner unused lands surrounded by built-up areas.

Municipal borderCentral built-up

area borderResidentialIndustryInstitutionsMixed land use

(CBD)Tourism and

recreationOpen spaceAgricultural landSpecial usesUnused land

1 : 120 000

Figure 1. Scale unitöthe urban built-up area (central area and leapfrog areas), example of theCity of Rishon Letzion (200 000 residents). CBD denotes the central business district.


Leapfrog areas are built-up land use areas that are located separately at a distancefrom the central built-up area, but which have a functional linkage with it (residential,industrial, or institutional). We excluded from this category special leapfrog areas thatwere identified through the field survey, such as built-up land-use areas that do nothave any direct functional linkage with the urban settlement (regional installations,interchanges, interurban highways, or army camps).

3.4 Urban sprawl variablesTorrens and Alberti (2000) suggested an analogy between open space or natural lands(used as the investigated unit in ecological research) and urban built-up areas as apractical method for quantifying sprawl, using measures adopted' from the ecologicaldiscipline. We adopt this perspective, and therefore refer to polygons of different urbanland uses as `patches'. Accordingly we define the urban sprawl by using two basic charac-teristics of the urban built-up area: configuration' refers to the shape of the urban fabricand composition' refers to the mixture of land uses within the built-up area (see table 1).

As some researchers have suggested, moving from `sprawl' to `nonsprawl' form ismore likely to be a direction in a continuum rather than across fixed measurablecategories (Johnson, 2001; Pendall, 1999). Based on this assumption, we define directionsin a continuum for each sprawl measure implemented in our study (table 1). Forexample, relatively low density means sprawl, whereas high density means nonsprawlor compact development. With respect to land-use composition, urban sprawl isdefined as a homogenous development pattern, characterized by a poor diversity ofmixed land uses (in particular, trade and services) in the city and at the neighborhoodlevel (Fulton, 1996). Built-up areas with a high mixture land uses are regarded ascompact and sustainable (Burton, 2000; Jenks et al, 1996), whereas a high percentage

Table 1. Characteristics, dimensions, and indices of the built-up area.

Characteristics Dimensions Variables Indices Directionof impacton levelof sprawl

Configuration density population density Dg i Ðgross population ÿdensity

Dn i Ðnet population density ÿscatter irregularity of the Fi Ðfractal dimension �

shape of the central Si Ðshape index �built-up areaboundary

fragmentation Ig i Ðgross leapfrog index �In i Ðnet leapfrog index �Mij Ðmean patch size ÿ

Composition mixture of land-use segregation a NA b

land uses land-use composition Ui 1 Ðresidential area �(percentage of each Ui 2 Ðindustrial area ÿland-use category) Ui 3 Ðpublic institutions ÿ

land-use areaUi 4 Ðmixed land use ÿUi 5 Ðtourism and recrea- ÿ

tion areaUi 6 Ðspecial land uses c ÿ

aNot measured in this study. bNot applicable. c Sport centers, cemetery, urban interchanges,bus and railway stations.


of residential land use in an urban area means that it is homogenous and nonmixed,thus sprawling. Another way of looking at this aspect is the balance that exists betweenthe population level and the number of jobs (Ewing et al, 2002). A nonbalanced situationis regarded as sprawl and is characterized by a high rate of residential area amongthe land uses of the built-up area. A complete explanation of all sprawl measures usedin our study and the direction of their impact on sprawl is presented in the appendix.

3.5 The integrated sprawl indexIn order to compare levels of sprawl among settlements, a method of weighting allthirteen measures to produce a single integrated sprawl index is suggested. Severalmethods of weighting measures of sprawl have recently been discussed, such as Z-scorescaling, factor analysis, and cluster analysis (Ewing et al, 2002; Galster et al, 2001;Wolman et al, 2004). Still, most of the sprawl studies generally focus on one or two typesof measures, usually taken from the same research discipline. Studies in which populationsizes and densities are implemented generally have no reference to geometric or ecologicalmeasures (Ewing et al, 2002; Galster et al, 2001; Shoshany and Goldshleger, 2002), andvice versa (Herold and Menz, 2001). Studies that combine density and scatter dimensionsof sprawl usually lack the third dimension, land-use composition (Benguigui et al, 1998).Our study suggests a weighted combination of the three dimensions, or sets of com-pletely different measures of sprawl, adopted from different disciplines: population-densitymeasures, geometry and ecological indices, and land-use composition.

Factor analysis was chosen as a method of weighting all thirteen measures toproduce one integrated sprawl index. We found it suitable as a data-reduction methodin cases where some of the variables are linearly correlated with one another (Kim andMuller, 1978). We assume that dimensions of sprawl are independent, and are notsignificantly correlated with each other. Therefore we first separated our thirteensprawl measure into two groups (configuration and composition characteristics of theurban landscape), representing the three conceptual dimensions of sprawl: density,scatter, and mixture of land uses (see table 1). Factor analysis was first performedfor all sprawl measures, for t1 (year 2002), divided into these two major groups ofmeasures, where each group contained 6 ^ 7 measures.

In the next step we computed a weighted sprawl score for each of two dimensions (4)

(density and scatter and mixture of land uses), according to the percentage of theexplained variance obtained by the variables included in each of the two dimensions.We then computed a final weighted sprawl index as a weighted average of the two sprawlscores, based on the explained variance of each one. For convenience, the weightedsprawl index was normalized to a positive Z score, with the most compact settlementreceiving a score of zero and the most sprawling settlement a score of 524.7.We call thisscale the integrated sprawl index for t1 . In order to describe the dynamics of sprawl duringthe period investigated, we also computed an integrated sprawl index for t0 (mid-1980s).For the sake of consistency, which allowed us to compare the two integrated indices (at t1and at t0 ), we transformed all sprawl measures for t0 to Z scores comparable to those fort1 , using the same factor loadings received for t1 and did not run a separate factor analysisfor t0 (a similar methodology was implemented in a study by Ewing et al, 2002).(5)

(4) Density and scatter were united into one sprawl dimension for convenience.(5)We did this by subtracting the averages of sprawl measures at t1 from measures at t0 anddividing them by standard deviations of sprawl measures at t1 . We then multiplied each vector Zscores (meaning a row of thirteen normalized sprawl measures for each settlement at t0 ) by thecomponent score-coefficient matrices obtained in the factor analysis procedure for t1 , resulting intwo sprawl scores for each dimension at t0 . The scores were weighted by the percentage explainedvariance and were normalized to a positive Z score, which represents an integrated sprawl indexfor t0 which is consistent and comparable with that for t1 .


4 The challenge of using sprawl measuresAverage results of each group of sprawl configuration measures show that most urbansettlements in Israel over the last two decades have become denser, their geometricalshapes more regular and compact, the leapfrog areas smaller, and the mean patch sizesof built-up areas larger. However, land-use composition did not dramatically changeduring the time period investigated (table 2).

4.1 Density measuresAs density is the most popular sprawl measure, we will elaborate on the density resultsobtained in our study. The average gross and net population densities have risen by13.9% and 9.9%, respectively, in Israeli urban settlements over the period investigated.In terms of density, sprawl is defined as a state in which density is relatively low and/orbecoming lower during a certain time period. On the basis of these two categories,we divided Israeli urban settlements into four major groups (see figure 2). The borderlines that distinguish each of the four groups indicate the average gross-densitymeasure of the whole sample in 2002 (x axis) and the zero growth rate of the grossdensity during the period investigated (y axis).

It is clear from figure 2 that the growth rate in population density between t0 andt1 in most of the settlements is above zero. However, the gross density in most of thesesettlements at the end of the time period investigated is below the average of the wholesample. This means that most of the settlements in which density increased are stillsprawling in relative terms.

One group of settlements is definitely sprawling (group 4), as its density is relativelylow and becoming even lower over time. Most settlements in this group are semiruralor peripheral settlements. In contrast, another group is definitely compact (group 1), asits density is relatively high and becoming higher over time. Most settlements in thisgroup are relatively large cities, where denser development is usually expected becauseof higher land rents.

Table 2. Sprawl measures (average values) for seventy-eight urban settlements at t0 and t1 .

Dimension Sprawl measure Mid-1980s (t0 )a 2002 (t1 )

a Percentagechange(t0 ± t1 )

Configuration gross population density 6 676 (3 432) 7 609 (3 710) 13.9(density and (pop/km2)scatter) net population density 11 201 (5 349) 12 243 (5 525) 9.9

(pop/km2)fractal dimension 1.287 (0.038) 1.273 (0.036) ÿ1.1shape index 2.581 (0.867) 2.424 (0.817) ÿ6.1gross leapfrog index (%) 6.5 (10.5) 5.2 (9.3) ÿ20.0net leapfrog index (%) 3.5 (8.5) 2.8 (8.3) ÿ20.0mean patch size (ha) 44.1 (26.3) 52.9 (23.0) 20.1

Composition residential area (%) 66.7 (16.9) 67.9 (14.7) 1.8(mixture of industrial area (%) 11.9 (11.6) 12.4 (9.6) 4.2land uses) public institutions land-use 4.9 (5.1) 4.7 (4.5) ÿ4.1

area (%)mixed land use retails (%) 3.3 (3.3) 2.6 (2.3) 39.4tourism and recreation 1.6 (4.6) 1.5 (4.3) ÿ6.3

area (%)special land uses (%) 4.0 (6.3) 4.2 (4.7) 5.0

a Standard deviation is given in parentheses.


Group 2 and 3 are variations of sprawl or compactness. Group 2, the biggest groupin our sample, contains about half of the urban settlements. Although density hasrisen in this group, we consider the urban areas in the group to be sprawling: some ofthe settlements in this group are semirural, suburban, or peripheral settlements withaverage densities lower than 5000 residents per km2. During the 1990s much rural landwithin these settlements was given over to residential use. Former agricultural farmswere divided into smaller parcels, enabling a massive development of detached residen-tial units, purchased by middle-class and upper-class households. Thus, although averagedensities in this group of settlements have risen, their development pattern of suburbandetached houses is characteristic of sprawl. Another part of this group consists ofsettlements with average densities higher than 5000 residents per km2. These are biggerand older cities, where many new immigrants were absorbed during the 1990s either invacant apartments in older neighborhoods or in new densely built neighborhoods.

Group 3, which contains only four urban settlements, is the smallest group in oursample. Average densities in this group are relatively high, but are becoming lower overtime. Negative density gradients in this group can be explained either by low rates ofpopulation growth, owing to population aging, or by high growth rates of nonresidentialareas. Either way, density decline in this group can be considered sprawl.

To conclude, although the growth rate of built-up areas over the period investigatedwas high (26%) and almost all additional built-up areas came at the expense of openand rural lands, the population growth rate was higher (44%);(6) thus, the average grossdensity of the sample increased by 13.9%. This positive density growth rate may haveresulted from the high capacity of cities and smaller towns, where part of the newpopulation could have been absorbed in an ìnfill' manner inside the existing urbanbuilt-up areas. This high capacity could have resulted from a high amount of vacantdwellings in older cities, where new immigrants were usually absorbed, or, alterna-tively, from the high amount of available land inside smaller quasirural settlements,where agricultural land was transformed into urban land and filled in by high amountsof lower-density detached single-family residential units. Thus, average densities in those

Cha

ngein

grossdensity

from

t 0to

t 1

1.0

0.5

0.0

ÿ0.5

ÿ1.05 10 15 20

Gross density at t1 (�1000)

Group 2

Group 4 Group 3

Group 1

Average density � 7609

Figure 2. Change in gross urban density between t0 and t1 versus gross urban density at t1 .

(6) The settlement sample size adds up to 3.3 million residents on 493:8 km2 at t0 , and to 4.8 millionresidents on 624:9 km2 at t1 .


settlements increased, although the development pattern in this case was rathersprawling because of the high percentage of detached residential units. Our conclusion,therefore, is that density cannot be a sole parameter of sprawl, and that further indicesare needed in order to quantify this phenomenon.

4.2 Scatter measuresShape-index and fractal measures show the level of irregularity of the perimeter of thecentral built-up area. Shape-index and fractal values have slightly decreased on averageover time. Thus the level of irregularity of the urban form in most of the settlementshas not changed much and there has been a slight tendency toward filling the form in acompact manner. Still, in twenty-nine urban settlements we found that shape-index andfractal values have increased, pointing to a sprawl configuration. In addition, shape-index and fractal measures were found to be higher and to increase over time (thusmore sprawling) in mountainous topography, as has similarly been found in otherstudies (O'Neill et al, 1988).

As for leapfrog measures, we found that most settlements (70 ^ 80%) did nothave leapfrog residential areas at all. About 50% of the settlements had nonresidentialleapfrog areas that were reasonably located outside the central built-up area, such asindustrial areas. We also found that, on average, leapfrog index values decreasedduring the period investigated (by 20%), thereby proving the assumption that leapfrogdevelopment tended to be filled as time passed; hence, sprawl may be considered atemporary condition (Brueckner, 2000; Peiser, 1989).

Although most of the settlements in our sample are not developed in a scatteredleapfrog manner, leapfrog areas were found to be significant in sixteen settlements, withsums of up to 11% of the entire urban built-up area, and to be increasing over time.

The mean patch size, M, of built-up areas increased by about 20% during the timeperiod investigated. As with the result for the leapfrog measures, this proves that mosturban settlements in Israel are developed in a nonscattered way. However, M decreasedin twelve of the seventy-eight settlements, thus development in those settlementswas more scattered and sprawling. The results imply that urban development tends tobe more scattered and fragmented in old, semirural settlements, in which rural parcelswere divided and transformed into residential parcels over the course of decades. Anopposite pattern seems to characterize preplanned new settlements, in which mostof the residential areas were developed simultaneously. Further investigation is neededin order to confirm this assumption.

4.3 Land-use compositionThe proportion of residential areas did not dramatically change over time. About 67%of the urban built-up areas of the settlements are residential areas, 12% are industrialareas, and each of the other land-use categories constitutes about 1 ^ 6% of the urbanbuilt-up area. Most of the new built-up areas were developed on open, rural landsoutside the central built-up areas. During the period investigated, the inner unusedareas in the built-up areas of the seventy-eight urban settlements constricted by only2.2% (from 27:2 km2 at t0 to 26:6 km2 at t1 ). This finding confirms the assumption thatit is easier to build on vacant lands than to develop in an ìnfill' manner inside existingneighborhoods (Ewing, 1994). We also noticed a rise in industrial areas on the fringesof settlements, as opposed to a decline in commercial areas inside them. We explainthis finding by the rise of commercial malls within industrial areas in Israel since the1990s, as well as the parallel decline in commercial activity inside old neighborhoods.

We also found that residential areas in all the settlements exceeded 50% of thetotal built-up area, and that there was little variance among settlements in regard tothis parameter. This finding proves the essential difference between open lands and


urban lands. Open natural lands, where ecological measures of land-use compositionare implemented, are much more heterogeneous and diversified than urban built-upareas. Although variation in land-use composition exists between different groups ofsettlements (Frenkel, 2004a), the fact that urban landscapes are less diversified thannatural landscapes necessitates the use of other composition measures in future studies.

On the basis of these results, especially those referring to the settlement configura-tion characteristic, we may conclude that most urban settlements in Israel have becomeless sprawling and more compact during the past two decades. This relatively compactpattern of development uniquely characterizes Israel, in contrast to North Americancountries, and is more similar to the West European urban pattern. It may result fromlimited land resources causing a more regulative planning policy and denser, morecompact forms of urban development. But does it mean that urban sprawl does notexist at all in Israel? We will discuss this question in the next section.

5 Level of sprawl in Israeli urban settlements5.1 Sprawl indexThe level of sprawl was obtained by computing the integrated sprawl index for eachof the urban settlements in the sample through factor analysis. Factor analysis wasemployed for the two dimensions of sprawl examined in this study: density and scattercorrespond to the configuration', and the mixture of land uses to the composition', ofthe urban landscape. The results of the factor analysis are presented in table 3.

The analysis of each group produced three main factors, with a total explainedvariance of 70 ^ 80%. Each factor is a linear combination of all measures in the group,with one or two dominant measures that define its unique ìdentity'.(7) Each factorreceived a positive or negative sign according to its relative contribution to sprawl, basedon the previously defined continuum (see the appendix). The signs were given to the entirelinear combination of measures in each factor, not only to its dominant measures.

Although the M loading coefficient is lower than 0.5, it is interesting to introducethis measure in regard to its interaction with shape and fractal measures. We realizedthat the more irregular the shape of the central built-up area (higher values of shape orfractal indices), the more fragmented it becomes (lower M values). To the best of ourknowledge, correlations between shape indices and patch indices have so far beenexamined only in ecological research (McGarigal and Marks, 1995); and, since theirresults vary, they can be interpreted in both ways. Hence, further investigation on thissubject is needed also in urban studies.

The composition dimension also has three factors: the residential ^ industrial factor,the commerce and services factor, and the leisure factor. Commercial and service areasare positively correlated with each other, as are areas of leisure (tourism and specialuses). Factor analysis was found to be efficient in this case, not only as a data reductionmethod, but also as a method of identifying major interactions between differentsprawl measures. Not all of these interactions could have been identified by simplelinear correlations.

On the basis of a linear correlation between the computed integrated sprawl indexand each of the sprawl measures, we found a relatively higher dominance of density,shape ^ fractal, residential, commercial, and industrial land-use measures when measuringurban sprawl. This is in contrast to the leapfrog and mean patch size measures, whichare apparently less important when measuring sprawl on a town scale.

(7) Dominant measures were defined as those with an absolute value of the component coefficientgreater than 0.5. For example, a factor whose absolute values of gross and net density coefficients weregreater than 0.5 and whose absolute values of all other measure coefficients were smaller than 0.5was identified as the `density factor'.


Additionally, the residential land-use percentage is highly correlated with the land-use composition score (R 2

p � 0:939, with a significance of 99%). This means that theresidential land-use measure represents well, by itself, the urban land-use composition.What is lacking is a better measure of the level of fragmentation and segregation ofthe landscape, ascertained by employing other geometric measures (lacking in thisstudy), such as accessibility, proximity, and contagion indices of land-use polygons orpatches.(8) Further development of these sprawl measures is needed in future studies.

5.2 Ranking urban settlements according to the level of sprawlBased on the integrated sprawl index value (Zi ) at t0 and t1 , each of the settlements inthe sample was ranked on a relative `sprawl scale'. The most sprawling settlementreceived the highest Zi value, and the most compact settlement received the lowestvalue. We then divided the sample into four `sprawl clusters' (see also figure 3):. cluster 1: `highly compact', 0 < Zi < 200;. cluster 2: compact', 200 < Zi < 300;. cluster 3: `sprawling', 300 < Zi < 350;. cluster 4: `highly sprawling' Zi > 350.

Table 3. Factor analysis results at t1 , major sprawl indices, and factor loading.

Urban built- Sprawl Major factor a Dominant Compo- Direction Percent- Cumula-up area dimen- measures nent of impact age lativecharacteristic sion in each score on sprawl variance percent-

factor b (� or ÿ) c age ofvariance

Configuration density irregularity of F 0.962 � 34.9and the central S 0.918scatter built-up area M ÿ0.469

perimeterpopulation Dg 0.961 ÿ 25.1density Dn 0.971

leapfrog Ig 0.923 � 23.3 83.3In 0.933

Composition mixture residential and U1 ÿ0.838 ÿ 33.7of land industrial U2 0.925uses land use,

weightedindex

commerce and U3 0.842 ÿ 20.3institutions U4 0.651land use,weightedindex

leisure land use U5 0.869 ÿ 18.4 72.3

(tourism and U IN6 0.560

special uses),weightedindex

aMajor factors were defined by eigenvalues >1. b For a detailed definition of each sprawlmeasure see appendix A. c Each factor received a sign (positive or negative) according to itspositive or negative contribution to the level of sprawl. This determination was made in thecomputation of the integrated sprawl index (see section 3.5).

(8)We know part of the level of fragmentation by the M measure, but only for all built-up areasand without further specification of each land use.


The sprawl scale at t0 and t1 provides us with a relative measure of comparisonamong settlements, as well as an examination of the sprawl dynamics of each settle-ment over a period of two decades. In general the integrated sprawl-index levels rangefrom 12.8 ^ 631.9, with an average sprawl level of 310.3 at t0 , to a range of 0 ^ 524.6 andan average sprawl level of 286.8 at t1 . This means that urban settlements became a littlemore compact during the 1980s and 1990s, a finding that is in contrast to the generalbelief that most settlements in Israel developed in a sprawl-like manner during thosedecades.We also found that most settlements (72%) did not change their sprawl clusterdramatically over time.

The highly compact settlements are usually large, denser cities, while highlysprawling settlements are usually small and semirural. Generally there is a high andnegative correlation between the level of sprawl and the size of a settlement, as has

600

500

400

300

200

100

0

600

500

400

300

200

100

0

Zi

Zi

Cluster 1:highly compact

Cluster 2:compact

Cluster 3:sprawling

Cluster 4:highly sprawling

(a)

(b)

Figure 3. Rank of urban settlements in descending order, by integrated sprawl index at (a) t1 , and(b) t0 .


been found in similar studies (Ewing et al, 2002). Some settlements, however dense,are still sprawling because of other measures (shape and fractal indices). We found that75% of the sample population resided in more compact settlements (sprawl clusters1 ^ 2), which encompassed only 67% of the total built-up area of the settlements in thesample. Thus, sprawling urban settlements are less efficient in using land than arecompact urban settlements.

Another major finding is that the growth rates of the population and built-up areasin sprawling settlements (sprawl clusters 3 ^ 4) over the time period investigated wererelatively high: 52.5% and 29.7%, as opposed to 39.9% and 24.5%, respectively, incompact settlements (sprawl clusters 1 ^ 2) (see table 4). This means that sprawlingsettlements have a higher capacity and are apparently more attractive to new residentsthan are compact settlements. it seems that there is a higher consumer preference forresiding in smaller, more sprawling settlements. Although their density has increaseddramatically over the past two decades, sprawling settlements are not expected totransform into compact, dense cities. We assume that sprawling settlements are morelikely to reach the saturation phase as soon as all their vacant lands are filled withdetached houses. Then, other new sprawling settlements might be developed in order tomeet these consumer preferences.

Inevitably, we did not find a high correlation between the level of sprawl and theaverage population density. For example, average density in sprawl cluster 3 is less thanin sprawl cluster 4. Similarly, average density in sprawl cluster 2 is higher than insprawl cluster 1 (table 4). Hence, even though a density measure is prevalent in urbanresearch in general and in sprawl research in particular, it still cannot substitute for theintegrated sprawl index, which encompasses other characteristics of the sprawl phe-nomenon, such as geometric scattering, fragmentation, and land-use composition.Additionally, during the period investigated, density increased more in sprawl clustersthan in compact clusters: 17.6% and 12.3%, respectively (table 4). This finding points to

Table 4. Population and urban built-up area growth rates during the period investigated, accordingto sprawl cluster. Values lying between columns are weighted means.

Variable Sprawl cluster at t1 Total

cluster 1, cluster 2, cluster 3, cluster 4,highly compact sprawling highlycompact sprawling

Number of settlements 15 23 18 22 78Population size at t1 1 461 100 1 598 300 562 300 1 130 497 4 752 197Built-up area at t1 (km

2) 189.3 186.4 87.4 161.9 625.0Population increase from 299 422 573 154 177 162 405 642 1 455 380

t0 to t1Increase in built-up area 26.2 47.8 19.0 38.2 131.1

from t0 to t1 (km2)Population growth rate 25.8 55.9 46.0 56.0

from t0 to t1 (%) 39.9 52.5 44.1Built-up area growth rate 16.1 34.5 27.7 30.8

from t0 to t1 (%) 24.5 29.7 26.6Population density at t0 7 124 7 395 5 629 5 858

7 249 5 777 6 676Population density at t1 7 720 8 572 6 435 6 983

8 143 6 791 7 604Density growth rate 8.4 15.9 14.3 19.2

from t0 to t1 (%) 12.3 17.6 13.9


the uniqueness and complexity of the urban sprawl phenomenon. Hence, as opposedto the accepted assumption in many studies (eg Torrens and Alberti, 2000), urban sprawlis not necessarily expressed in terms of a decreasing tendency in density with time.

In order to illustrate the advantage of using the integrated sprawl index, land-usemaps for six different settlements are presented in figure 4. The maps serve as examplesof the diversity of urban settlements in our sample and of the level of complexity of thesprawl phenomenon. Three of the settlements present the compact pattern of the built-up area and were classified in cluster 1 (at t1 ), based on our integrated sprawl index.

Variable

Integrated sprawl indexPopulation size (000)Built-up area (km2)Gross population densitySimple sprawl index

Compact pattern Sprawl pattern

Bat Yam

0133.97.6

176001.07

Ramla

129.662.88.6

76000.87

Bet She'an

106.815.94.8

33001.02

Jerusalem

432.7680.451.1

10 5000.85

Pardes Hanna

39228.88.5

3 4000.67

Bet Shemesh

366.353.47.6

7 0000.54

1 : 260 000

1 : 72 000

1 : 110 000

1 : 110 000

1 : 110 000

1 : 165 000

Compact pattern Sprawling pattern

Municipalborder

Central built-uparea border

ResidentialIndustryInstitutionsMixed land use(CBD)

Tourism andrecreation

Open spaceAgriculturalland

Special usesUnused land

Figure 4. Six examples of land-use maps of cities at t1 . CBD denotes the central business district.


The other three examples present a sprawl pattern and belong to cluster 4 (at t1 ). Eachof the two groups contains large and medium-sized cities, as well as small towns, onthe basis of population size or the size of the built-up area.

Bet She'an is an example of a small town with a low population density that, at thesame time, exhibits a compact pattern. An opposite example is Jerusalem, the largestcity in Israel, which has a relatively high population density but presents a sprawlpattern. However, the most valuable finding is the comparison between the classifica-tion result based on the integrated sprawl index developed in this study and theclassification that would have resulted from employing the popular SI used in manyother studies (eg Hadly, 2000; Weitz, 2000). Towns such as Bat Yam and Bet She'anthat definitely present compact patterns (in configuration and composition dimen-sions) are sprawling according to the simple SI (1.07 and 102, respectively). The reasonis that Bat Yam is an aging city with a declining built-up area that has been losing itsresidents. Therefore, the growth rate of its built-up area is higher than that of itspopulation, leading to SI > 1ösprawl pattern. An opposite example is presented byJerusalem and Pardes Hanna, which definitely present sprawling patterns based on theintegrated sprawl index, but should be classified as compact according to the SI,with a value <1 (0.85 and 0.67, respectively). These two settlements benefited fromhigh immigration rates, which led to a population growth rate higher than thebuilt-up area growth rate. Our findings strengthen the assumption, suggested inrecent studies (Galster et al, 2001; Malpezzi and Wen-Kai, 2001; Torrens and Alberti,2000), that sprawl is rather the description of a relative condition than a fixed,measurable category.

5.3 Sprawl dimensionsOur last examination was performed on the two sprawl dimensions: configurationand composition. The purpose was to see whether these two dimensions are inter-linked, and to assess their contribution to the general level of sprawl and sprawlchange over time.

As previously mentioned, Israeli urban settlements became more compact overtime. By examining each sprawl dimension, we found that the configuration scoredecreased and the composition score slightly increased.(9) This means that settlementstook on a more compact configuration (denser, more regular, and less scattered), butalso a more sprawling, less mixed, land-use composition. This finding accords with theearlier findings on each of the sprawl measures.

Urban sprawl was found to be a multidimensional phenomenon, as illustrated inthe three-dimensional diagram in figure 5. This diagram presents the value of theintegrated sprawl index of every settlement as a function of the two separate sprawlcharacteristics, configuration and composition. As can be seen, some settlementsdemonstrate either a sprawling or nonsprawling pattern in terms of both character-istics (Bat Yam versus Qiryat Tiv'on), while other settlements show a level of sprawlaffected by one of the sprawl characteristics (Bene Beraq versus Be'er Ya'aqov).

We found that most of the settlements presented a sprawl or compact pattern ofdevelopment with respect to both characteristics. However, an examination of theeffect of the integrated configuration and the composition sprawl indices separatelyshowed that the composition characteristic had a greater impact on the final integratedsprawl index. With respect to the change over time in the pattern of development, theconfiguration characteristic was found to be more dominant through its impact onthe change in the value of the integrated sprawl index. In 41% of the settlements the

(9) Average configuration score at t0 was 0.233, and 0 at t1 , with a standard deviation of 0.587;the average composition score at t0 was ÿ0:063, and 0 at t1 , with a standard deviation of 0.599.


level of sprawl changed over time (positively or negatively) as a result of the indicesincluded in the configuration characteristic; in 40% of the settlements, the changeresulted from indices belonging to both characteristics. Change was brought about byindices belonging to the composition characteristic in only 19% of the settlements.We also found that settlements with maximum population growth and maximumbuilt-up area growth rates during the period investigated exhibited a sprawl patternof development with respect either to both characteristics or to only the configurationcharacteristic.

6 ConclusionsUrban sprawl is still a controversial issue among scholars, who argue over its impactas well as the way it should be measured. Hence, employing public policy in order torestrain the phenomenon is hampered, in particular, by the lack of empirical evi-dence. This lack affects the ability to convince the authorities to adopt such policies.

500

400

300

200

100

0

Sprawlindex

ÿ10

1ÿ1

0

1

Sprawl configuration index Sprawl

mixed

land-u

seind

ex

Settlement

Bat YamQiryat Tiv'onBene Beraq

Be'er Ya'aqov

Ranking score

configurationcharacteristic

2781

77

compositioncharacteristic

75536

10

integratedsprawl index

1782

70

Sprawl pattern

compact in terms of both characteristicssprawling in terms of both characteristicscompact in terms of the configuration

characteristicsprawling in terms of the configuration

characteristic

Figure 5. Three-dimensional diagram of ranking urban settlements, according to level of sprawl.The sprawl configuration index and sprawl mixed land-use index are the aggregated indicesresulting from the factor analysis. All values are those at t1 .


Questions about exactly what sprawl is, how it affects the urban environment, and howit should be measured remain unanswered. Efforts have been undertaken recently todeal with these issues. This study focused on the measurable question: how can thevarious aspects and characteristics of sprawl be measured and what are the indices thatshould be implemented empirically in a unit of investigation at the town scale?

The integrated sprawl index introduced in this paper is an unusual combination,making use of sprawl measures from different disciplines: urban studies, fractal geom-etry, and ecological research. We note, however, that there are some measures that aremore effective in measuring sprawl on a municipal scale (eg density, shape/fractal,residential, commercial, and industrial land-use composition) and other measuresthat are less effective or less relevant (eg leapfrog, mean patch size, other built-upland uses). The latter group seems to be more effective in measuring sprawl on aregional or metropolitan scale. For example, the appearance of new scattered settle-ments in a region will be considered leapfrogging development, whereas this index isnot noticeable at a town scale.

Urban land-use composition is less heterogeneous than is open land-use composi-tion, because of the dominance of residential uses within the urban built-up area.Hence, a residential land-use measure represents well the urban land-use composition,obviating the need to compute the percentages of all other land uses. The integratedsprawl index according to land-use mixture suggested by this study represents the land-use composition and the diversity or the equilibration that exists between residentialland use and all other land uses in the built-up area of the urban municipality.However, the spatial distribution of land uses as defined by the level of segregationand access between residential and other land uses is another aspect of mixed land usethat was not tested in this study. Therefore, further investigation of sprawl, in terms ofland-use mixture, on the municipal scale is still needed; for example, distances, traveltime, and the accessibility of different land uses; geometry of the spatial distribution ofland uses; and the population distribution at different distances from these land uses.An urban landscape that is characterized by a high level of land-use mixture, as well asby heterogeneity of its spatial land-use distribution is considered to be compact. On theother hand, an area in which land uses are segregated and distant from one another isconsidered to be sprawling.

The various sprawl measures used in this study show clearly that urban sprawl is aphenomenon that should be described and quantified by a combination of severalmeasures. Each group of measures represents different features or characteristics ofthis phenomenon and does not necessarily depend on other dimensions. We especiallyfound differences between the two characteristics of sprawl and their effect on theurban landscape pattern. The configuration characteristic of sprawl is linked to accel-erated urban growth and increased land consumption more than is the compositioncharacteristic. Thus, sprawl that resulted from a scattered, less-dense configuration ofthe built-up area is more responsible for the waste of land than is sprawl that emanatedfrom a homogeneous land-use pattern. On the other hand, sprawl that resulted fromthe composition characteristic is tied to the socioeconomic level of the population.The latter is characterized by a high level of car-based transportation systems and theincreased use of private vehicles for work travel. This finding leads to the hypothesisthat different sprawl patterns have different impacts on the urban form and should bestudied in the future.

Since urban sprawl appears to be a multidimensional phenomenon, we hypothesizethat its implications probably emerge from different urban patterns of development.Apparently, this complexity is linked to the disagreements that exist between scholarsand planners on this issue. Our finding implies that different sprawl patterns have


diverse implications for urban form that should be investigated. Some settlements,especially quasirural ones, were found to be more sprawling than others. This factimplies that sprawl rates may be higher in rural settlements than in urban settlements.Therefore, we highly recommend continuing the investigation of rural sectors, as thismight be more relevant to sprawl and its impacts on land consumption.

Higher sprawl rates were found to be significantly correlated with higher popula-tion and land-consumption growth rates. This finding implies a higher consumerpreference for residing in more sprawling patterns, meaning that sprawling settlementsare probably more attractive to new residents who seek housing improvement than arecompact settlements. A definite conclusion on this matter requires further investigationof consumer preferences and the alleged positive impacts of sprawling patterns per-ceived by consumers. We believe that this possible consumer preference for sprawlingpatterns, along with the lack of available land in Israel, fully justifies attempts toregulate and restrain sprawl in Israel.

ReferencesAlperovich G, 1995, ``The effectiveness of spine urban density functions: an empirical investigation''

Urban Studies 32 1537 ^ 1548BattyM,KimKS,1992,` Form follows function: reformulating urban population density functions''

Urban Studies 29 1043 ^ 1070BattyM, Longley P,1994FractalCities: AGeometryof FormandFunction (Academic Press, London)Batty M, Xie Y, Zhanli S, 1999, ` The dynamics of urban sprawl'',WP 15, Centre for Advanced

Spatial Analysis, University College London, http://www.casa.ucl.ac.uk/sprawl.pdfBelser K, 1960, ` Urban dispersal in perspective'', in The Nature and Control of Urban Dispersal

Ed. E A Engelbert, California Chapter of the American Institute of Planners, Berkeley, CA,pp 1 ^ 18

Benguigui L, 1995, À fractal analysis of the public transportation system of Paris'' Environmentand Planning A 27 1147 ^ 1161

Benguigui L,MarinovM, CzamznskiD,1998CityGrowth as a Leap-frogging Process: AnApplicationto theTel-Aviv Metropolis Center for Urban and Regional Studies,TechnionöIsrael Institute ofTechnology, Haifa

Brueckner JK, 2000,` Urban sprawl: diagnosis and remedies'' InternationalRegional ScienceReview23 160 ^ 171

Burchell RW, Shad N A, Listokin D, Phillips H, Downs A, Seskin S, Davis J, Moore T, Helton D,Gall M, 1998, ` The cost of sprawl-revisited'', report 39, Transit Cooperative ResearchProgram (TCRP) Transportation Research Board, National Research Council,Washington,DC, pp 83 ^ 125

Burton E, 1996, ` The potential of the compact city for promoting social equity'', in The CompactCityöASustainable Urban FormEdsM Jenks, E Burton, KWilliams (Spon, London) pp 19 ^ 29

Burton E, 2000, ` The compact city: just or just compact? A preliminary analysis'' Urban Studies37 1969 ^ 2001

CBS, 1997 Localities in Israel, Spatial Data publication number 1103, Central Bureau of Statistics,Jerusalem (in Hebrew)

CBS, 2004, ` Localities and population by municipal status and district'', in Statistical Annual ofIsrael Number 55, Central Bureau of Statistics, Jerusalem (in Hebrew)

Chin N, 2002, ` Unearthing the roots of urban sprawl: a critical analysis of form, function andmethodology'',WP 47, Centre for Advanced Spatial Analysis, University College London,www.casa.ucl.ac.uk

Churchman A, 1999, ` Disentangling the concept of density'' Journal of Planning Literature13 389 ^ 411

Downs A, 1998, ``How America's cities are growing: the big picture'' Brookings Review 16 8 ^ 12Ewing R, 1994, ` Characteristics, causes, and effects of sprawl: a literature review'' Environmental

and Urban Issues 21 1 ^ 15Ewing R,1997, ` Is Los Angeles-style sprawl desirable? ''Journal of the American Planning Association

63 107 ^ 126Ewing R, Pendall R, Chen D, 2002, ` Measuring sprawl and its impact'', volume 1, technical report,

Smart Growth America,Washington, DC, http://www.smartgrowthamerica.org


Frenkel A, 2004a, ` Land-use patterns in the classification of cities: the Israeli case'' Environmentand Planning B: Planning and Design 31 711 ^ 730

Frenkel A, 2004b, À land consumption model: its application to Israel's development'' Journal ofthe American Association 70 453 ^ 470

FultonW, 1996 The New Urbanism (Lincoln Institute of Land Policy, Cambridge, MA)Galster G, Hanson R, Ratcliffe M R,Wolman H, Coleman S, Freihage J, 2001, ` Wrestling sprawl

to the ground: defining and measuring an elusive concept Housing Policy Debate 12 681 ^ 717Gans H J,1967The Levittowners:Ways of Life and Politics in a New Suburban Community (Pantheon

Books, NewYork)Garreau J, 1991Edge City: Life on the New Frontier (Doubleday, NewYork)Gibbs J P, 1961, À method for comparing the spatial shapes of urban units'', in Urban Research

Methods Ed. J P Gibbs (D Van Nostrand, Princeton, NJ) pp 99 ^ 106Gonen A, 1995, ` Recent middle-class spread to suburbia'', in Between City and Suburb: Urban

Residential Patterns and Processes in Israel Ed. A Gonen (Avebury, Aldershot, Hants)pp 114 ^ 138

Gonen A, 1996, ` Changing urban residential structure in Israel'', inTheMosaic of Israeli GeographyEdsYGradus, G Lipshitz, The Negev Center for Regional Development (Ben-GurionUniversity of the Negev Press, Beer Sheva) pp 55 ^ 62

Gordon P, Richardson H W, 1997, Àre compact cities a desirable planning goal? '' Journal of theAmerican Planning Association 63 95 ^ 106

Gustafson E J,1998,` Quantifying landscape spatial pattern: what is the state of the art? ''Ecosystems1 143 ^ 156

Hadly C C, 2000, ` Urban sprawl: indicator, causes, and solutions'', paper prepared for theBloomington Environmental Commission, http://www.city.bloomington.in.us/planning/boardcomm/ec/reports/sprawl.html.p

Hartshorn TA, Muller P O, 1992, ` The suburban downtown and urban economic developmenttoday'', in Sources of Metropolitan Growth Eds E S Mills, J F McDonald, Center for UrbanPolicy Research, Rutgers University, New Brunswick, NJ, pp 147 ^ 158

Harvey R O, ClarkWAV, 1965, ` The nature and economics of urban sprawl'' Land Economics41 1 ^ 9

Hasse J E, LathropRG, 2003,` Land resource impact indicators ofurban sprawl''AppliedGeography23 159 ^ 175

Herold M, Menz G, 2001, ` Landscape metric signatures (LMS) to improve urban land useinformation derived from remotely sensed data'', inADecadeofTrans-EuropeanRemoteSensingCooperation Ed. M F Buchroithner (A A Balkema, Rotterdam) pp 251 ^ 256

Ingram G K, 1998, ` Patterns of metropolitan development: what have we learned? Urban Studies35 1019 ^ 1035

Jackson KT, 1985 Crabgrass Frontier: The Suburbanization of the United States (Oxford UniversityPress, Oxford)

JenksM, Burton E,Williams K,1996The Compact CityöASustainableUrban Form (Spon, London)JohnsonMP, 2001, ` Environmental impacts of urban sprawl: a survey of the literature and proposed

research agenda Environment and Planning A 33 717 ^ 735Kim J O, Mueller C W, 1978, ` Introduction to factor analysis: what it is and how to do it'', Paper 13

Quantitative Applications in the Social Sciences Series (Sage, London)Klaassen L H, MolleW T N, Paelinck J H P, 1981The Dynamics of Urban Development (Gower,

Aldershot, Hants)McGarigal K, Marks B, 1995 FRAGSTATS: Spatial Pattern Analysis Program for Quantifying

Landscape Structure TR PNW-GTR-351, USDA Forest Service,Washington, DCMalpezzi S, 1999, ` Estimates of the measurements and determinants of urban sprawl in US

metropolitan areas'',Wisconsin ^Madison CULERWP 99-06, Center for Urban LandEconomics Research, University of Wisconsin, Madison

Malpezzi S,Wen Kai G, 2001Measuring ` Sprawl'': Alternative Measures of Urban Form in U.S.Metropolitan Areas Center for Urban Land Economics Research, University of Wisconsin,Madison,WI

Maret I, 2002 `Ùnderstanding the diversity of sprawl and the need for targeted policies'', paperpresented at the 5th Congress on Environment and Planning, Oxford, 23 ^ 27 September,available from 922 North Lopez Street, New Orleans

Mills E S, Hamilton BW, 1989, ``Trends in sizes and structure of urban areas'', in Urban Economics4th edition, Eds E S Mills, B W Hamilton (Scott, Foresman and Company, Glenview, IL)pp 68 ^ 83


O'Neill R V, Krummel J R, Gardner Sugihara G, Jackson B, DeAngelis D L, Milne B T, TurnerM G, Zygmunt B, Christensen S W, Dale V H, Graham R L, 1988, ` Indices of landscapepattern'' Landscape Ecology 1 153 ^ 162

Peiser R B, 1989, ` Density and urban sprawl'' Land Economics 65 193 ^ 203Pendall R, 1999, ` Do land-use controls cause sprawl?'' Environment and Planning B: Planning and

Design 26 555 ^ 571Razin E, Rosentraub M, 2000, Àre fragmentation and sprawl interlinked? North American

Evidence'' Urban Affairs Review 35 821 ^ 836Real Estate Research Corporation, 1974, ``The costs of sprawl: detailed cost analysis'' (Government

Printing Office,Washington, DC)Schiffman I, 1999, ` The limits of ideology: the first for open space in Israel'', inHandbook of Global

Environment Policy and Administration Eds D L Soden, B S Steel (Marcel Decker, NewYork)pp 559 ^ 578

Shoshany M, Goldshleger N, 2002, ` Land-use and population density in Israelö1950 to 1990:analysis of regional and local trends'' Land Use Policy 19 123 ^ 133

Torrens P M, Alberti M, 2000, ` Measuring sprawl'',WP 27, Centre for Advanced Spatial Analysis,University College London, http://www.casa.ucl.ac.uk

TurnerMG,1989, ` Landscape ecology: the effect of pattern on process''Annual Review of EcologicalSystems 20 171 ^ 197

Van Den Berg L, Drewett R, Klaassen L H, Rossi A,Vijverberg C H T, 1982, ` Stages of urbandevelopment'', in A Study of Growth and DeclineVolume 1, Eds LVan Den Berg, R Drewett,L H Klaassen, A Rossi, C H T Vijverberg (Pergamon Press, NewYork) pp 24 ^ 45

Weitz J, 2000,``State programs that bust sprawl andmeasures that can be used in diagnosing sprawl'',paper presented at the 2000 APA National Planning Conference, http://www.asu.edu/caed/proceedings00/WEITZ/weitz.htm

Wolman H, Galster G, Hanson R, Ratcliffe M, Furdell K, Sarzynski A, 2004, ` The fundamentalchallenge in measuring sprawl: which land should be considered? ''The Professional Geographer57 94 ^ 105


AppendixSprawl measures that were operationalized in this researchEach measure was operationalized both at t0 (mid-1980s) and at t1 (2002).

Nomenclature:i a particular urban settlement in the research sample.j type of land use (10) ( j � 1, . . . , 10)ö j � 1 residential (11), j � 2 industry, j � 3

institutions, j � 4 mixed land use (CBD), j � 5 tourism and recreation, j � 6special uses, j � 7 malls (12), j � 8 open space, j � 9 unused land, j � 10agricultural land;

Pi number of residents in urban settlement i ;A c

i central built-up area of urban settlement i ;A u

i urban built-up area of settlement i, including land uses 1, 2, 3, 4, 5, 6(in), 7,8(in), 9(in), 10(in);

Ri residential area of settlement i (land use 1);Li perimeter of central built-up area of settlement i ;A out

i leapfrog areas in settlement i, including land-use categories: j out1ÿ7 , except forland-use category j out6 ;(13)

R outi residential area outside the central built-up area of settlement i ;

ai j area of land use j in urban settlement i ( j � 1, . . . , 10);ni j number of polygons of land use j in urban settlement i ( j � 1, . . . , 10).(14)

Given this nomenclature, we define each one of the sprawl measures operationalizedin this study as follows:

Density indicesGross density,

Dgi �Pi

A ui

.

Net density,

Dni �Pi

Ri

.

Impact on sprawl: negative (higher density means lower sprawl).

Irregularity of shape indices (15)

Shape index,

Si �Li

2�pA ci �1=2

.

(10)Here j refers to each of the specified land-use categories within a settlement's jurisdiction;j in refers only to land-use categories located within the central built-up area, while j out refers only tothose located outside the central built-up area.(11) This category includes the residential parcels and the attached neighborhood infrastructure(public institutions, neighborhood civic centers, open spaces, public gardens, and local roads).(12) At t1 (2002) land use 2 (industry) was summed with land use 7 (malls), unless specified otherwise.At t0 (mid-1980s), there were no malls in Israel, except for the city of Ramat-Gan; hence, land use 7is irrelevant in that time period.(13) Special areas outside the central built-up area are usually not functionally correlated with theurban settlement, such as cemeteries, airports, and military bases. Therefore, they were not includedin the urban-sprawl measures, unless otherwise specified.(14)Minimum size of the polygon of specific land use j was 3 ha for all uses, except for unused land,for which the minimal size of the polygon was 5 ha.(15) In this group of indices, the units of measurements were m for Li and m2 for Ai .


Fractal dimension,

Fi �2 lnLi

lnA ci

.

Impact on sprawl: positive.

Fragmentation indicesGross leapfrog index,

Igi �A out

i

A ui

.

Net leapfrog index,

Ini �R out

i

Ri

.

Impact on sprawl: positive.Mean patch size,

Mij �X7j� 1

ai jni j

.

Impact on sprawl: negative.

Land-use composition indicesSix land-use categories were used to identify the land-use composition of the built-uparea of settlement i. The measures were calculated as a percentage of land use j fromthe entire urban built-up area (A u

i ).

Ui j �ai jA u

i

, j � 1, . . . , 6 .

Impact on sprawl: residential useöpositive; all other built-up usesönegative.

ß 2007 a Pion publication printed in Great Britain


Conditions of use. This article may be downloaded from the E&P website for personal researchby members of subscribing organisations. This PDF may not be placed on any website (or otheronline distribution system) without permission of the publisher.

Date post:	08-Dec-2016
Category:	Documents
Upload:	maya
View:	218 times
Download:	4 times

Measuring urban sprawl: how can we deal with it?

Documents