You are on page 1of 11

Journal of Environmental Management 115 (2013) 87e97

Contents lists available at SciVerse ScienceDirect

Journal of Environmental Management


journal homepage: www.elsevier.com/locate/jenvman

Urban sprawl and fragmentation in Latin America: A dynamic quantification and


characterization of spatial patterns
Luis Inostroza a, *, Rolf Baur b,1, Elmar Csaplovics a
a
Institute of Photogrammetry and Remote Sensing, TU Dresden, Helmholtzstrasse 10, Dresden Saxony D-O1062, Germany
b
CIPSEM Centre for International Postgraduate Studies of Environmental Management, DIU/TU Dresden, Zellescher Weg 41c, 01062 Dresden, Germany

a r t i c l e i n f o a b s t r a c t

Article history: South America is one of the most urbanized continents in the world, where almost 84% of the total
Received 31 March 2012 population lives in cities, more urbanized than North America (82%) and Europe (73%). Spatial dynamics,
Received in revised form their structure, main features, land consumption rates, spatial arrangement, fragmentation degrees and
31 October 2012
comparability, remain mostly unknown for most Latin American cities. Using satellite imagery the main
Accepted 7 November 2012
Available online 12 December 2012
parameters of sprawl are quantified for 10 Latin American cities over a period of 20 years by monitoring
growth patterns and identifying spatial metrics to characterize urban development and sprawling
features measured with GIS tools. This quantification contributes to a better understanding of urban form
Keywords:
Urban sprawl
in Latin America. A pervasive spatial expansion has been observed, where most of the studied cities are
Latin America expanding at fast rates with falling densities trend. Although important differences in the rates of land
GIS metrics consumption and densities exist, there is an underlying fragmentation trend towards increasing sprawl.
Spatial development These trends of spatial discontinuity may eventually be intensified by further economic development.
Urban Sprawl/Latin America/GIS metrics/spatial development.
Ó 2012 Elsevier Ltd. All rights reserved.

1. Introduction highest in Bolivia by 84%, and in Paraguay by 99% (UN, 2009). The
impact on the land and the surrounding ecosystems will depend on
With almost 84% of the total population living in cities, South how urban development would take place in responding to these
America is one of the most urbanized continents in the world, more trends. Urban form, intentionally driven by urban planning, could
urbanized than North America (82%) and the European Union make a difference in limiting the undesirable effects of urban
countries (73%) (UN, 2009). Certainly the destiny of the subconti- expansion e such as watershed degradation, wetland destruction,
nent will be led by urban development and driven by urban and agricultural land consumption (Kasanko et al., 2005a; Schwarz,
systems. However besides a few studies analysing urban sprawl 2010). However it has to be properly defined by adequate spatial
among some of the bigger South American cities (Huang et al., indices and an appropriate scale under integrated land use and
2007; Schneider and Woodcock, 2008; Angel et al., 2010b), there resource management perspective.
is a lack of analytical studies addressing urban spatial development To give a better understanding of the spatial development of
in this continent. urban systems a differentiation between expansion, sprawl and
South America plays a relevant role in the global urban context: fragmentation seems to be necessary, because these terms are not
in 1950 three cities were counted among the 30 most populated equivalent in the literature (Galster et al., 2001). Although sprawl
cities of the world. By 2005 the continent counted four cities in the and expansion refer both to the spatial process of size increase in
ranking and by 2025 it will have five (UN, 2009). cities, a sprawling process implies in most cases a specific expan-
Cities are growing in population and size; and this expansion is sion pattern, identified with the model of suburban growth. This
a common urban challenge. By 2050 urban population will increase determines a certain urban form opposite to the compact city
by 34% on the South American continent; growth rates will be model (Galster et al., 2001; UN, 2011).
Within this debate between critics of urban sprawl and their
opponents, the critics of the compact city do not argue on matters
of definitions alone, but also take into consideration the potential
* Corresponding author. Tel.: þ49 351 463 33680.
impacts of different patterns of urban expansion in certain devel-
E-mail addresses: inostroza@urbansprawlproject.com (L. Inostroza), rolf.baur@
tu-dresden.de (R. Baur), Elmar.Csaplovics@tu-dresden.de (E. Csaplovics). oping countries where most urban growth is still to come (UN,
1
Tel.: þ49 351 497 99 11. 2011).

0301-4797/$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.jenvman.2012.11.007
88 L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97

The compact city has already been on the political agenda for radial direction around the city centre or in a linear direction along
Europe (CEC, 1990, 1996; Kasanko et al., 2005b). It is an objective of the road network, on the urban fringe, at the edge or close by an
urban planning (Burton, 2002) to accommodate urban develop- existing urban area or along the highways: new developments will
ment while minimising the use of undeveloped land and identi- tend to follow naturally a city edge. This is a gravitational force
fying areas with a high need of “policy” intervention (Schwarz, dragging urban development to the city core. Benefits arise from
2010). In this spatial perception the gaps in the urban fabric being part of the infrastructure network. As a result, urban cores
suggest a lack of efficiency in the urban development due to inor- follow an “oil stain” pattern of expansion. Such continuous
ganic growing processes (Batty et al., 2003). This sets the impor- expansion over the edges is a physical behaviour present in all cities
tance of analysing urban sprawl over other urban forms (Galster despite the enormous variations in shapes (Andersson et al., 2002).
et al., 2001; UN, 2011). It is possible to distinguish two spatial patterns: infill and axial
Urban sprawl has been defined as growth by the creation of new development, understood in all new development taking place in
low-density suburbs with detached or semi-detached housing and adjacency with the pre-existing urban tissue where infilling makes
large commercial strips (Schneider and Woodcock, 2008; Schwarz, formerly sprawling areas more compact. Axial growth follows the
2010). It has been described as the physical pattern of low-density main development axis, and is strongly determined by infrastruc-
expansion of large urban areas, under market conditions, mainly ture. Infill might change depending on the criteria used to define it.
into the surrounding agricultural areas. Sprawl is the leading edge Infill has been defined as vacant parcels or remaining open
of urban growth and implies little planning control of land subdi- spaces, in already urbanized areas which might or might not
vision and urban development. It is patchy, scattered and strung include patches located at a certain distance from the existing
out, with a tendency for discontinuity, and it leap-frogs over areas urban tissue (Angel et al., 2005, 2010b). In this study, a spatially
leaving agricultural or open space enclaves behind. Sprawling cities explicit threshold of adjacency which determines infill patterns in
are the opposite of compact cities, with empty spaces that indicate new urban development is proposed.
the inefficiencies in development and highlighting the conse- At the same time, a second spatial behaviour can be observed:
quences of uncontrolled expansion (EEA, 2006). a tendency to escape from the city core and to spread urban
Under this perspective urban sprawl has been understood as development over hinterlands where urban development as spatial
a specific pattern of urban expansion; typifying an increasing trend pattern rises from discontinuity of the urban tissue. Here we find
of undesirable effects. It has been mostly studied as a matter of leapfrogging development or isolated patches, a process driven by
dismissing densities in the spatial development of cities (Galster centrifugal forces from the city. All this behaviour is mainly spatial
et al., 2001; Angel et al., 2010a), or as a matter of differences in (physical): cities, at certain scale, behave according to similar rules
the extent of urban development (Burchfield et al., 2006). following some common patterns that lie beyond social or cultural
Nevertheless, urban sprawl remains as a diffuse and elusive context.
concept. There is no general agreement on what sprawl means, or The quantification of the sprawl might provide certain facts
its measurement empirically and the comparison of its occurrence towards a better understanding of spatial urban development,
across different cities (Galster et al., 2001). Despite increasing determining under spatial metrics, whether urban expansion
awareness of the need for rigorous definition and systematic trends towards sprawl and fragmentation. It must be determined
measuring of sprawl (Galster et al., 2001; Tsai, 2005; Angel et al., what development patterns and trends are most common and
2005), the term is used ambiguously and is equivalent to expan- where different patterns are taking place. This is the first step to
sion (Galster et al., 2001; Angel et al., 2005). But at the same time develop parameters to monitor city trends. By using common
the term is strongly controversial, with no neutral point about spatial classification criteria, the degree of fragmentation of urban
sprawl: there are foes (EEA, 2002; Kasanko et al., 2005a, 2005b) and sprawl was quantified, characterizing and comparing the differ-
friends (Bruegmann, 2005; Neuman, 2005; Echeñique et al., 2012). ences in the Latin American urban development scenario in the
On the contrary expansion is a neutral definition which depicts the general expansive processes.
increases of cities in size and surface into surrounding land areas
without qualitative implications. As cities expand as a result of
demographic and economic growth further expansion is an 2. Methodology
unavoidable fact. What is aimed to be addressed in this study is
whether the specifics patterns of urban expansion are leading The approach towards the investigation of urban expansion and
towards sprawl and fragmentation. If these trends are clear, what sprawling processes constitutes a big challenge in Latin-America,
similarities and differences will be observed along the continent considering the enormous lack of information and measurement
looking at population dynamics and spatial arrangements? of urban phenomena. Thus the first aim is to quantify the urban
The concept of fragmentation has been largely used in Land- development process in a simple way that allows comparisons and
scape Ecology and Urban Planning (Angel et al., 2010b; Schneider further applications beyond the scope of this study, like informing
and Woodcock, 2008; Schwarz, 2010). Fragmentation reflects the policy options for instance.
morphological e spatial in this paper - efficiency of the urban The aim is to address the problem with the use of appropriate
expansion process (Batty et al., 2003) and has been normally analytical tools. A platform of categories, concepts and method-
understood as static metric and measured as the morphological ology for comparing urban sprawl and fragmentation across the
property of patches. Studies about urban form refer to fragmenta- continent is proposed to quantify the processes of urban sprawl and
tion as one or more landscape metrics: edge density, mean patch fragmentation in ten large Latin American cities, by using simple
size, etc. (Schwarz, 2010), where fragmentation can be understood well known metrics, readily available.
as the spatial pattern of discontinuity. A dynamic approach towards sprawl and fragmentation is
In a dynamic way, by assessing how new development is taking applied, considering it as an on-going process that shows trends in
place in relation to the existing urban tissue, two types of spatial spatial behaviour with specific characteristics, topology and rela-
behaviour of urban growth patterns can be distinguished. The first tionship with the pre-existing urban tissue e core city, spatial
being, new urban development taking place along the boundaries structure, infrastructure, land use, etc.
of already urbanized areas with the implication of “how growth The method defines a spatially explicit threshold of adjacency in
attracts further growth”. It usually advances in an asymmetric order to determine compactness and fragmentation, with general
L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97 89

criteria but specific spatial application which allows comparisons of water supply or sewerage. To support the determination of city’s
despite the differences between cities. edge, an analysis with the ARCWORLD imagery tool of ArcGIS10Ó,
based on GeoEye IKONOS 1 m resolution imagery was done. The
2.1. Data sources and selection of cities delimitation of the core areas was finally cross tested by 3 different
GIS analysts’ experts, with an overall delta error of less than 2% of
To face the aims of this study a complete spatial data base for 10 the total perimeter.
Latin American cities was developed using LANDSAT imagery as the One aim of the applied method of segmentation before the pixel
main source. The images are covering a time window of 20 years, classification, was to include suburbanization into low density
one image circa 1990 (T1) and one circa 2010 (T2) (Table 1). urban fabric from the core area. The suburbanization process is
An unsupervised land cover classification routine was done for mainly a land use change, the transformation of agricultural land,
the 20 images (see 2.2 below) (Fig. 1). This classification was and it implies the subdivision into smaller plots or new suburban
complemented with vector information (neighbourhoods, infra- rural neighbourhoods with or without minimal infrastructure
structure, and census), field information, and other ancillary vector (roads, equipment, etc) taking place around the edge of the cities.
and raster data collected and obtained during visits of the cities, This kind of urban development has not been normally considered
complemented by interviews with local professionals of urban into the quantification of urban sprawl (Schneider et al., 2009;
planning, architecture, real estate market, infrastructure develop- Taubenböck et al., 2012). However to improve the sprawl quanti-
ment and other backgrounds, aiming to improve the classification fication it is necessary to properly include these new morphological
accuracy. typologies of urbanization. A second aim of the edge delineation
was to improve the image classification by running separated
2.2. Spatial approach, and land cover classification criteria routines for fringes and cores. The derived land cover imagery
performed a high level of accuracy: an average above 89% accuracy
Administrative definitions are widely used in spatial analysis, with ground field points collected during our visits to the cities.
but they introduce distortions and are not directly comparable for Four main classes were defined: built-up or sealed surfaces
different cities and countries. Thus, a spatially explicit approach (core and fringe), open spaces (only in the core area), geographical
was used to define the geographical area to be included in the constraints including all areas where urban development could not
analysis: the MOLAND method (Kasanko et al., 2005a; EEA, 2002) take place such as sea, rivers, lakes, mountains, etc., and develop-
which defines a fringe area around the city core. This approach is able area including either agricultural or natural land representing
both spatially explicit and dynamic as it varies according to the city all areas susceptible to receive further urban development. This
size and time. classification criterion was aimed to focus on land cover changes
The city core (area) corresponds to the urban continuous fabric over time, estimating the amount of new land transformed from
and it was determined for each city based on the on-screen non-urban to urban cover/uses.
segmentation process. The city core is an approximation of the Image classification of urban open space (golf clubs, cemeteries,
true physical city boundaries and is an alternative definition of sport fields, playgrounds, etc.) is a challenge, as these land use types
urban footprint (Angel et al., 2005; Taubenböck et al., 2012), have a spectral performance similar to natural open space.
although the aim is similar: to obtain a delineation of the physical Attempting to reduce the amount of fuzzy open space in the cities
man-made properties of cities. edge, most of these urban areas were identified during the
Morphological continuity of the urban tissue was the criterion segmentation process, the separation of core from fringe areas.
to find the city edge and to separate cores from their fringes; where Thus, all open space within the core area is considered as urban.
that continuity finishes, whether by a geographical barrier, However, with LANDSAT imagery as main source it was hardly
a sudden or smooth interruption of the urban fabric, the city core possible to estimate the capture of open space in the fringe without
will find its edge. This was determined by using spectral behaviour, controversial assumptions. Free space being used for new urban
shapes and textures of the LANDSAT images in T1 and T2. The development in the fringe gets lost in mixed pixel spectral signa-
continuity of the urban fabric is following roads and infrastructure tures or is being confused with natural/agricultural land; thus very
which are important vectors leading to the expansion of cities; this high resolution imagery or high quality of ancillary data (owner-
fact has been tested by random checks during the field excursions ship data sets for instance) is needed to identify such areas. Even
by following certain transects from the very centre towards the though there are some general assumptions in the reviewed liter-
hinterland. For some cities this determination was complemented ature (e.g. Angel et al., 2005), they are all indirect mechanisms. For
by aerial photography and other ancillary data, such as the coverage this study direct measuring of urban phenomena was preferred.

Table 1
Selected cities and respective satellite imagery. The first a priori selection of cities tried to cover all capitals for the whole continent. This preliminary selection was adjusted and
several possibilities were explored, assessing both availability and quality of imagery for the two times period. The final selection of cities with the dates of the imagery used in
the analysis is shown in this table.

ciudades Pais Coordinates Image T1 Image T2 Projections

Latitud Longitud Date Fecha UTM, m


1 Asunción Paraguay 25 160 0000 S 57 400 0000 W 03.03.1987 29.01.2010 WGS 21 S
2 Bogotá Colombia 4 350 56.5700 N 74 040 51.3000 W 22.03.1988 22.01.2010 WGS 18 N
3 Brasilia Brasil 15 460 000 S 47 550 000 W 24.07.1988 18.07.2009 WGS 23 S
4 Buenos Aires Argentina 34 360 1400 S 58 220 5400 W 25.11.1988 18.10.2009 WGS 21 S
5 Córdoba Argentina 18 280 0000 N 69 540 00"W 22.06.1985 12.09.2009 WGS 20 S
6 La Paz Bolivia 16 300 0000 S 68 090 0000 W 02.08.1987 11.04.2010 WGS 19 S
7 Lima Perú 12 030 0000 S 77 030 0000 W 05.03.1987 02.04.2009 WGS 18 S
8 Montevideo Uruguay 34 530 100 S 56 100 5500 W 20.02.1985 17.08.2009 WGS 21 S
9 Santa Cruz de la Sierra Bolivia 10 300 0000 N 66 550 0000 W 11.01.1985 04.06.2010 WGS 20 S
10 Santiago Chile 33 260 1600 S 70 390 0100 W 11.10.1986 19.03.2010 WGS 19 S
90 L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97

Fig. 1. Flow diagram of the image processing and land cover classification protocol. The land cover classification protocol has followed three main steps: imagery pre-processing,
imagery processing and GIS modelling. The classification procedure followed the method proposed by Angel et al. (2005), to allow comparability of the results.

Thus, the classification focuses on the quantification of built up all built-up pixels, district population has been assigned to those
areas e sealed surfaces e in the fringe, leaving the consideration of built-up pixels, considering zero population value for “non-urban”
open spaces being captured by urban development for further pixels, such as agricultural land. This spatial assignation of pop-
research. ulation is methodologically more robust than the administrative
density or another modelling starting from already estimated
2.3. Population densities densities.
Densities were calculated estimating the amount of population
Official census data has been used as the main source for ana- living in the specific spatial unit and then aggregated in order to
lysing population growth and densities. However, census data are estimate averages for larger areas. The measure of density used
indexed to administrative boundaries, which in most cases are not corresponds to the net built-up density, which is the ratio of pop-
comparable among countries e administrative denominations vary ulation living in all built-up areas into the core, as against all built-
across the continent with enormous differences in size and scope. up area into the fringe. While not including open spaces in this
The use of census data without taking into consideration its spatial calculation, trends of net density changes between T1 and T2 are
implications might produce misunderstandings about population clear, in terms of determining how much more or less dense new
and density dynamics. Population dynamics are clear on a certain developments are (Table 2).
scale, according to administrative levels, but within those areas at
a smaller scale the spatial process of populations, such as de-
2.4. Metrics for measuring sprawl
concentration, loss of density, suburbanization and others, might
be hidden.
Four quantifiable indicators are used: built-up area, density,
Accumulated statistics for city areas or for the whole urban
spatial configuration and speed. While the first three metrics are
regions suggest that large urban regions are becoming more pros-
commonly used in similar studies (Angel et al., 2005, 2010a;
perous while not being affected by demographic changes. They do
not consider developments on a smaller scale (Kroll and Kabisch,
2011). Table 2
For the purposes of this study, the aim was to get a realistic Dates of census, dates of imagery acquisition and respective methods of adjustment.
estimation of densities in the specific spatial arrangement where When official estimations were not available for the times of the imagery, census
data were interpolated/extrapolated to match T1 and T2 (Table 2) times using linear
the new urban development is taking place, being capable of
or exponential demographic models.
characterizing net density trends and answering the questions if
new built-up areas are more or less dense than existing ones, giving City Census Imagery Method Census Imagery Method
overall sprawl trends of net cities densities. The objective was, to 80’s T1 00’s T1
arrive at the real extension of urban developments with their Asunción 1982 1987 INTRAPOL 2002 2010 EXTRAPOL
respective population data, where normal census districts are too Bogotá 1985 1988 OFICIAL 2006 2010 OFICIAL
big in the periphery to compare with inner city census districts. Brasilia 1991 1988 EXTRAPOL 2010 2009 CENSUS
Buenos Aires 1991 1988 EXTRAPOL 2010 2009 CENSUS
The methodological problem of addressing the shared pop-
Còrdoba 1991 1985 EXTRAPOL 2010 2009 CENSUS
ulation in periphery districts e shared population between urban La Paz 1992 1987 EXTRAPOL 2001 2010 OFICIAL
and rural areas e emerges from the fuzzy definition of “urban” used Lima 1981 1987 INTRAPOL 2007 2009 OFICIAL
by census approaches. To overcome this issue, population data have Montevideo 1985 1985 CENSUS 2004 2009 EXTRAPOL
been assigned only to built-up areas in the periphery districts by Santa Cruz de 1992 1985 EXTRAPOL 2001 2010 OFICIAL
la Sierra
using the smallest available spatial unit (blocks, neighbourhoods, Santiago 1982 1986 INTRAPOL 2002 2010 OFICIAL
etc.) of census data bases. As the classification method determined
L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97 91

Schneider and Woodcock, 2008; Schwarz, 2010), speed of the maximum feasible optimization of the urban form possible to
fastest vector, is a new proposed metric. By assessing the trends achieve in T2. Morphologically, it reflects the maximum compact-
across these metrics between T1 and T2 the approach leads to ness a city can achieve within the period investigated. It corre-
a dynamic measure of sprawl and fragmentation. sponds to the negative buffer from the fringe towards the core area
Built-up area corresponds to the total amount of sealed surfaces by using the same value of fringe thickness (W in MOLAND defi-
and is used to measure land cover changes from non-urban to nition, Kasanko et al., 2005a). All new urban fabric into that area
urban over the period analysed. Urbanized area corresponds to the was classified as infill.
core area, including open space, plus built-up into the fringe. Axial: all new urban fabric physically connected with, but
Density is measured as population per area and it is a key metric beyond the infill area was classified as axial. This kind of develop-
of sprawl. A net built-up density (built-up in core þ built-up in ment typically follows the main infrastructure axes.
fringe) was used. This metric does not consider open spaces. When Isolated: these are all patches discontinuously spread in the
comparing net density, T1 and T2 indicates the average trend in the fringe (Fig. 3).
land consumption, whether new developments (changes T1eT2) The share of infill, axial and isolated built up area between T1
are more or less dense than pre-existing ones (net built-up and T2 shows the relative spatial efficiency of new urban fabric.
density in T1) and how the average net density is changing (trend This configuration of new urban fabric is showing spatially explicit
T1eT2). trends of urban expansion by their relationship with the pre-
Spatial configuration corresponds to a dynamic assessment of existing city. The overall proportion of those patterns between T1
fragmentation, differing from the common measure calculated by and T2 shows the fragmentation trend: a city with higher propor-
typical landscape ecology metrics. To identify trends of fragmen- tion of infill patches is increasing its compactness while others with
tation all new urban developments between T1 and T2 were clas- higher proportion of isolated patches are decreasing. This trend is
sified into three categories: infill, axial and isolated (Fig. 2), easy to calculate and clear while giving a fast and spatially explicit
according to its spatial relationship with the pre-existing urban assessment of new urban fabric in a certain window of time.
tissue. Speed has been measured with two different metrics: first being
Infill: corresponds to new build-up pixels within the infilling defined as the rate of land conversion per capita per year. In addi-
area defined for each city. This infilling area represents the tion, the expansion speed of the fastest vector in meters per year

Fig. 2. Spatial configuration of urban space: a dynamic characterization (T1eT2) of fragmentation of Santiago de Chile by classifying all new development into the three proposed
categories: infill, axial and isolated. The red line shows the proposed spatial threshold of adjacency for this city.
92 L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97

Fig. 3. Core areas (Km2) and built-up areas (Km2) in fringes: sizes and increases. The ratio of built-up area is moving towards fringes: The average of built-up area in the fringe was
8% in T1 and 12% in T2. All cities have experienced an increase. Brasilia for instance had 9% of all the built-up area in the fringe during the middle 80s, and that percentage has more
than doubled up to 23% nowadays (Fig. 4).

was measured. This is a new proposed metric, easy to calculate and 2010a; Schwarz, 2010), 98,4 inhabitants per hectare in T1 and 101
self evident. It shows the dynamics of cities in the context of inhabitants per hectare in T2. Even though two opposite trends can
persistent urban expansion and has been calculated by measuring be observed: in four cities (Asuncion, Santa Cruz, Lima and Bogotá)
the changes in length of the longest vectors of core areas expansion densities are increasing while in the rest six cities (Brasilia,
in T1 and T2 starting from the city centre (CBD). Córdoba, Montevideo, Buenos Aires, Santiago and La Paz) densities
are decreasing (Fig. 6). While decreasing in density, these cities are
3. Results and discussion in a clear trend of sprawl and consuming more land than before.
Density trend of Bogotá however differs from all the rest of the
On average in the 10 studied areas, city cores account for 92% of continent, first because its initial density is already too high for the
the urbanized area. This high proportion of cores over the urban south American context, almost two and a half times bigger than
spatial structure is subject to change: cities like Brasilia or Santiago the average density in T1 (213.8 for Bogotá and 85.5 for the rest nine
are loosing weight of the city core (80%) (Figs. 3 and 5). cities) and even three and a third times in T2 (271.9 and 81.1
With the exceptions of Montevideo, Asuncion and Buenos Aires, respectively). Secondly because Bogotá’s trend of increasing
the amount of built-up area growing in the fringe between T1 and T2 density is an outlier: the city has increase its net built-up density by
has been bigger than the pre-existing built-up area (Fig. 3). The built- 58 inhabitants per hectare between T1 and T2, which is more than
up fringes are growing at an average rate of 11% per year, thus, much twofold the second increasing density value, Santa Cruz with 22
faster than the core areas (2.9%) or the urbanized areas (3.1%) (Fig. 4). inhabitants per hectare increase between T1 and T2.
Looking at densities, they are still high in comparison with To improve the estimations in the continent, not considering
densities in developed countries (Kansako et al., 2005a; Angel et al., Bogotá in the calculations a trend of dismissing densities appears:

Fig. 4. Changes in the percentage of built-up area in the fringe over all built-up area. All cities have increased their respective amount of built-up areas in the fringes between T1 and
T2. However, the percentage of increase differs between the study areas on a wide range; it is for: Brasilia 711%, La Paz 536% and Santa Cruz 356%; whereas the lowest in Montevideo
with 44%.
L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97 93

ten cities, Bogotá consumed just 0.4 m2 while Brasilia 7.9 m2.
However, the speed of the fastest vector does not match per capita
land consumption (Fig. 7).
Under this persistent urban expansion scenario the way cities
will expand could make a big difference (Fig. 8). If the average land
consumption continues at the same annual rate as it has during the
past twenty years (Fig. 7) the urban core area of major cities in
South America will have doubled by 2035. This would be even more
dramatic if densities persist to decline at similar rates as now.

3.1. Spatial configuration and trends

Although all studied cities are filling their infill areas, which is
expected behaviour in the pervasive expansion scenario, the
growth into axial and isolated patterns is more prominent. This
demonstrates the trend of urban development towards increasing
sprawl: growing by adding small, low density and patchy areas of
newly developed land located further from the established core,
instead of growing by adding plots at the edge or adjacent to the
Fig. 5. Changes in Km2 of built-up surface in the fringe (Buenos Aires not included: urban core (in the infill area) aiming to reduce fragmentation and
158 km2 Built-up in fringe T1 and 258 km2 Built-up in fringe T2).
increase compactness (Fig. 10).
On average, about 30% of all new development took place within
a total average decrease of 2.4% and a 0.13% decrease per year. This the infill area, 49% was axial and 21% isolated. The highest infilling
downward trend coincides with that found in other studies (Angel city was Cordoba, with 62% of infill, 28% axial and 10% isolated
et al., 2010a). development. La Paz shows the lowest total infilling rate with
Net built-up density as an average might be hiding spatial trends percentages of 13%, 86% and 1% for infill, axial and isolated
of changes across the urban tissue. This decrease might show respectively (Fig. 9). However the overall efficiency in the use of
a spatial divergence process which is changing the population infill is led by the increase of the core area, with a correlation of 0.7,
distribution within the urban structure. To look at these processes followed by surface of geographical constraints with 0.5. Higher
an analysis at a smaller scale is necessary. On the other hand, cities infilling rates are found in both: faster growing cities and cities with
like Bogotá, Lima, Asuncion and Santa Cruz are still in a densifica- more geographical constraints. Population growth was not statis-
tion process. It is interesting to note that this densification process tically significant while explaining infilling trends.
is taking place in the cities with both highest and lowest density
(Fig. 6). Bogotá, the city with the highest density, is (Fig. 7) showing 3.2. Discussion
an increase of net density of 27% during the study period. In the
case of Lima (12% of net density increase) this densification process Three aspects of the present study are discussed: the estimation
might be a response to the serious land availability restrictions and of population in the periphery, and the quantification of open space
despite the decreasing density of some very central districts. being captured by urban development are related to the method-
Asuncion has increased its low density by 18% responding to an ological approach; the links between sprawl and possible driving
important migration processes from the rural area (Rodriguez and factors focus on potential implications of the results.
Busso, 2009). Santa Cruz shows the highest increase of 42%. On the Population and densities have been measured for specific spatial
contrary Brasilia is loosing net density at an average of 31%. units facing the methodological constrains of spatial estimations
The amount of the per capita per year land consumption shows between different sizes of census units. The gross density for the
big differences: with an average of 3 m2 per capita per year in the whole “administrative city” has been normally accepted as the

Fig. 6. Net built-up densities of population per ha in T1 and changes T1eT2. Six cities are decreasing their net built-up density at an average of 15.6%, in some cases like Brasilia and
Montevideo the trend of decreasing densities is taking place in considerable amounts between T1 and T2 (31% and 18% respectively).
94 L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97

Fig. 7. Land consumption (m2 per year) and speed of fastest vector (m/year).

Fig. 8. Buenos Aires and Bogotá: sprawl v/s compact. Spatial patterns of urban development differ highly along the continent. It can be seen easily that Buenos Aires is more
sprawled and fragmented than Bogotá, which is more compact. Buenos Aires core area is 6 times larger than that of Bogotá (2103 km2 vs. 361 Km2), while its population is two times
less (12.9 million vs 7.6 million).

Fig. 9. Relative spatial efficiency: percentage of infill, axial and isolated of new built-up areas. Despite infilling cities, Bogotá and Córdoba, the sprawling trend is prevalent due to the
higher amount of axial and isolated patterns. At the same time there is an underlying trend of fragmentation reflected in the high average of isolated patterns, 21% for the ten cities.
L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97 95

Fig. 10. Spatial configuration of all cities, from above left: Asunción, Bogotá, Brasilia, Buenos Aires, Córdoba, La Paz, Lima, Montevideo and Santa Cruz, (Santiago, in Fig. 3).

measure of the cities’ population density, despite differences in the districts which can produce underestimated population density
size of administrative areas and the size of the real city. When (Chomitz et al., 2005). This could result in bias when assessing the
estimations for smaller spatial units are used in the periphery they dynamics of density. A smaller level of desegregation for population
have to face the problem of population shared between urban and data (blocks, smaller districts) was needed to better spatially
rural administrative units, besides the larger size of suburban address population and density information. This has been the
96 L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97

methodological key in order to properly assess densities in new


urban developments; it strongly depends on the data availability
and common frame for comparisons. This research attempted to
estimate densities on a smaller possible scale by assigning pop-
ulation to the built-up area; here, more detailed work should be
done in future research. Average city density estimations and big
census districts in the periphery are concealing the real perfor-
mance of the sprawling process and to face that, studies on the
basis of small scale spatial entities are needed.
Bogotá depicts a specific case. Density values of Bogotá were
outliers in the continent, showing strong differences with the rest
of the studied cities. Two considerations can be taken to the
discussion: first the reliability of the population data, which still
remains in estimations based on population models and not relying
in direct measuring techniques as is normally obtained in the other
countries. The second consideration is the possible link with high Fig. 12. Motorization rate versus percentage of infill areas (relative spatial efficiency).
levels of violence, as was suggested in many interviews. This
connection between lower rates of urban expansion with high
levels of violence and terrorism was proposed by Glaeser and IKONOS 1 m resolution imagery (ARCWORLD imagery tool of
Shapiro (2002), suggesting violence as a strong driving factor of ArcGIS10Ó) have been used in this research. With more accurate
compact development. Whether the failure of population models estimation of urban open space in the fringe, the impact of urban
or violence are driving the high densities found in Bogotá, needs sprawl would be certainly bigger than it was addressed in this
small scale studies to be determined. research.
To achieve a better understanding of urban sprawl it is necessary Urban sprawl as a specific spatial development pattern is led by
to have a closer approximation to the specific spatial behaviour of economic dynamics of both the city itself and the country, while the
every city, looking also at the density configuration within the factor of population is losing its importance (Kasanko et al., 2005a).
urban structure. The aim of a compact development will be difficult Urban sprawl is strongly connected to economic indicators such as
to achieve if we cannot adjust those differences within the inner the GDP (Angel et al., 2005; Huang et al., 2007). The correlation
city spatial structure. between GDP per capita and amount of land consumption in
The second aspect to be discussed is the quantification of open hectares per year in all studied cities is 0.77 (Fig. 11). Thus, as the
space being captured by urban development. In this research GDP increases, further sprawling patterns might be expected.
a method has been developed to catch suburbanization in the Issues of transportation are at the root of many discussions of
periphery based on the core edge delimitation prior to running the urban sprawl and urban density (Angel et al., 2005; Huang et al.,
land cover classification. Beyond the core edge, direct measures of 2007; UN, 2011). Regarding this link, in the studied cities higher
open space in the fringe are not possible with LANDSAT imagery as levels of fragmentation were correlated with higher motorization
a main source of information. Appropriate ancillary data (cadastre rates. The low efficiency in the use of the infill area correlates with
or similar together with a very high resolution imagery) or land use the motorization rate in vehicles per 1000 inhabitants (Fig. 12).
maps are necessary to catch this kind of urban development within Vehicles availability is playing an important role in the sprawling
the fringe. Here, only built-up pixels were measured, which is in trends which Latin American cities are facing.
fact an underestimation of urban development because it is not
including all land being transformed into urban use, such as play- 4. Conclusions
grounds, cemeteries, sport facilities and other land uses that cannot
be classified by a pure pixel-based approach. Though, there are The aim of the study was to improve the understanding of urban
some possible estimations (e.g. Angel et al., 2005) direct measuring sprawl in Latin America. The general urban expansion scenario in
during random field samples, experts’ interviews and GeoEye the continent was addressed by using certain metrics, some of
them new, like the speed of the fastest vector (Fig. 7).
Cities will expand, which is unavoidable, but how to avoid
sprawl and manage the urban development in the highest compact
way is a matter of urban planning. Morphologically, Latin American
cities tend to concentrate most of their built-up area in the core in
an average of over 92% of the total urbanized area. Population
densities are still high if we compare with cities in other latitudes
(Kansako et al., 2005a; Huang et al., 2007; Angel et al., 2010a). The
lack of maturity in the Latin American economies explain this
surprising fact, however if we compare the actual average of built-
up area in the fringe (12%) with cities in developed countries (27%)
(Angel et al., 2005; Kasanko et al., 2005a; Schwarz, 2010) an
important increase in the share of isolated patches is expected: at
least twofold compared to what it was during the last two decades.
As urban systems evolve and the national economies reach stronger
performance, cities will certainly expand more and faster. The
dynamic assessment of fragmentation also shows a trend towards
less compactness: out of the overall new urban development, only
Fig. 11. GDP per capita and land consumption in hectares per year. Such patterns one third is infilling, helping to reduce the fragmentation of the
emerge due to economic performance of countries and cities. periphery; two thirds are of a sprawled pattern.
L. Inostroza et al. / Journal of Environmental Management 115 (2013) 87e97 97

In the current scenario of constant urban expansion, the ques- Echenique, M.H., Hargreaves, A.J., Mitchell, G., Namdeo, A., 2012. Growing cities
sustainably. Journal of the American Planning Association 78 (2), 121e137.
tion for the urban planning is what the acceptable degrees of
EEA (European Environment Agency), 2002. Towards an Urban Atlas: Assessment of
sprawl and fragmentation are. Urban expansion will inevitably take Spatial Data on 25 European Cities and Urban Areas, Copenhagen, p. 128. <http://
place, most cities in Latin America will double their size by 2035, www.eea.europa.eu/publications/environmental_issue_report_2002_30>.
and thus there should be concern about its spatial configuration, to EEA (European Environment Agency), 2006. Urban Sprawl in Europe. The Ignored
Challenge, Copenhagen, p. 60.
have an appropriate estimation of the amount of land which will be Galster, G., Hanson, R., Ratcliffe, M.R., Wolman, H., Coleman, S., Freihage, J., 2001.
needed by urban development to implement an adequate spatial Wrestling sprawl to the ground: defining and measuring an elusive concept.
strategy that can be achieved with the lowest cost and highest Housing Policy Debate 12 (4), 681e717. http://www.fanniemaefoundation.org/
programs/hpd/pdf/HPD_1204_galster.pdf.
efficiency possible. The way cities expand matters. Glaeser, E.L., Shapiro, J.M., 2002. Cities and warfare: the impact of terrorism on
urban form. Journal of Urban Economics, Elsevier 51 (2), 205e224 (March).
Acknowledgements Huang, J., Lu, X.X., Sellers, J.M., 2007. A global comparative analysis of urban form:
applying spatial metrics and remote sensing. Landscape and Urban Planning 82
(4), 1847197.
The main findings of this article were results of a Lincoln Insti- Kasanko, M., Barredo, J., Lavalle, C., McCormick, N., Demicheli, L., Sagris, V.,
tute 2010e2011 research project (project reference: LRB082510). Brezger, A., 2005a. Are European cities becoming dispersed? A comparative
analysis of 15 European urban areas. Landscape and Urban Planning 77, 111e
Lincoln Institute provided the LANDSAT imagery for T2. The authors 130.
would like to thank three anonymous reviewers for comments on Kasanko, M., Barredo, J., Lavalle, C., Sagris, V., Genovese, E., 2005b. Towards urban
earlier versions of this paper. un-sustainability in Europe? An indicator-based analysis. 45th Congress of the
European Regional Science Association. Land Use and Water Management in
a Sustainable Network Society, Amsterdam, 23e27 August 2005.
References Kroll, F., Kabisch, N., 2011. The relation of diverging urban growth processes and
demographic change along an urbanerural gradient. Population, Space and
Andersson, C., Lindgren, K., Rasmussen, S., White, R., 2002. Urban growth simula- Place 18, 260e276. http://dx.doi.org/10.1002/psp.653.
tion from “First Principles”. Physical Review E 66 (2), 026204. Neuman, M., 2005. The compact city fallacy. Journal of Planning Education and
Angel S., Parent J., Civco D., Blei, A.M., 2010a. The Persistent Decline in Urban Research 25 (1), 11e26.
Densities: Global and Historical Evidence of ‘Sprawl’. Lincoln Institute of Land Rodríguez, J., Busso, G., 2009. Migración Interna y Desarrollo en América Latina
Policy Working Paper. Cambridge, MA. p. 151. entre 1980 y 2005. Un Estudio Comparativo con Perspectiva Regional Basado en
Angel, S., Parent, J., Civco, D., 2010b. The Fragmentation of Urban Footprints: Global Siete Países. Comisión Económica para América Latina y el Caribe (CEPAL),
Evidence of Urban Sprawl 1990-2000. Lincoln Institute of Land Policy Working Santiago de Chile.
Paper. Cambridge, MA, p. 104. Schneider, A., Woodcock, C., 2008. Compact, dispersed, fragmented, extensive? A
Angel, S., Sheppard, S.C., Civco, D.L., 2005. The Dynamics of Global Urban Expansion. comparison of urban growth in twenty-five global cities using remotely
Transport and Urban Development Department, The World Bank, Washington sensed data, pattern metrics and census information. Urban Studies, 659e
D.C, p. 205. 692.
Burchfield, M., Overman, H.G., Puga, D., Turner, M.A., 2006. Causes of sprawl: Schneider, A., Friedl, M.A., Potere, D., 2009. A new map of global urban extent from
a portrait from space. Quarterly Journal of Economics 121 (2), 587e633. MODIS satellite data. Environmental Research Letters. 4 article 044003.
Burton, E., 2002. Measuring urban compactness in UK towns and cities. Environ- Schwarz, N., 2010. Urban form revisited e Selecting indicators for characterising
ment and Planning B 29 (2), 219e250. European cities. Landscape and Urban Planning 96, 29e47.
Bruegmann, R., 2005. Sprawl: A Compact History. The University of Chicago Press, p. Taubenböck, H., Esch, T., Felbier, A., Wiesner, M., Roth, A., Dech, S., 2012. Monitoring
302. urbanization in mega cities from space. Remote Sensing of Environment 117,
Batty, M., Besussi, E., Chin, N., 2003. Traffic, Urban Growth and Suburban Sprawl. 162e176.
Working paper N 70, Centre for Advanced Spatial Analysis (UCL), London, UK. Tsai, Y., 2005. Quantifying urban form: compactness versus ‘Sprawl’. Urban Studies
CEC, 1990. Commission of the European Communities. Green Paper on the Urban 42 (1), 141e161. England, January 2005.
Environment. Communication from the Commission to the Council and United Nations (UN), 2009. Department of Economic and Social Affairs, Population
Parliament. COM 1990/218 final. Division. World Urbanization Prospects: The 2009 Revision. http://esa.un.org/
CEC, 1996. Commission of the European Communities. Expert Group on Urban unpd/wup/index.htm (accessed 14.07.11.).
Environment. European Sustainable Cities Report. Brussels. United Nations (UN), 2011. Population, Distribution, Urbanization, Internal Migra-
Chomitz, K. M., Buys, P., Thomas T.S., 2005. Quantifying the Rural-Urban Gradient in tion and Development: An Internal Perspective. Department of Economic and
Latin America and the Caribbean. World Bank Policy Research Working Paper Social Affairs, Population Division. http://esa.un.org/unpd/wup/index.htm
3634, June 2005. (accessed 14.07.11.).

You might also like