See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/309897308

Accessibility to urban green spaces in Chilean cities using adaptive thresholds

Article  in  Journal of Transport Geography · December 2016

DOI: 10.1016/j.jtrangeo.2016.10.012

CITATIONS READS

29 758

4 authors:

Carolina Alejandra Rojas Antonio Páez

Pontificia Universidad Católica de Chile McMaster University
88 PUBLICATIONS   487 CITATIONS    165 PUBLICATIONS   5,058 CITATIONS   

SEE PROFILE SEE PROFILE

Olga Barbosa Juan Antonio Carrasco

Universidad Austral de Chile University of Concepción
61 PUBLICATIONS   1,358 CITATIONS    64 PUBLICATIONS   1,993 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Breaks in W View project

Overweight and Obesity in Canada - PhD Thesis View project

All content following this page was uploaded by Carolina Alejandra Rojas on 17 April 2018.

The user has requested enhancement of the downloaded file.

 Journal of Transport Geography 57 (2016) 227–240

Contents lists available at ScienceDirect

Journal of Transport Geography

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

Accessibility to urban green spaces in Chilean cities using

adaptive thresholds
Carolina Rojas a,⁎, Antonio Páez b, Olga Barbosa c, Juan Carrasco d
a
Department of Geography, University of Concepción, Centre of Urban Sustainable Development (CEDEUS), Chile
b
School of Geography and Earth Sciences, McMaster University, Canada
c
Institute of Environmental Science and Evolution, Austral University of Chile, Centre of Urban Sustainable Development (CEDEUS), Chile
d
Department of Civil Engineering, Faculty of Engineering, University of Concepción, Centre of Urban Sustainable Development (CEDEUS), Chile

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

Article history: Open spaces in cities, such as green spaces and parks, are key elements of urban landscapes and urban sustain-
Received 3 August 2015 ability. Open spaces improve the quality of the environment, quality of life, and promote social interaction and
Received in revised form 28 October 2016 inclusion. There is evidence of positive impacts on health. This research analyses the potential accessibility to
Accepted 31 October 2016
open spaces in two Chilean cities, Valdivia and Temuco. Analysis of accessibility is based on patterns of daily mo-
Available online xxxx
bility by travelers while considering their trips and transport modes, in particular walking. Statistical analysis of
Keywords:
trip length forms the basis of an adaptive threshold for the calculation of a cumulative potential measure of ac-
Urban green spaces cessibility. The analysis shows that, for these two cities, variations in accessibility tend to be driven by age and
Accessibility gender, and less by income. There are also variations between the two cities, with Temuco showing better levels
Mobility of accessibility, as a consequence of residents generally undertaking longer walking trips.
Walk © 2016 Elsevier Ltd. All rights reserved.
Chilean
Cities

1. Introduction
Urbanization in Chile in the past 30 years has been characterized by
growing numbers of households living in peripheral locations. This is
true for social housing, and to a lesser extent also for condominiums
preferred by higher income households. Much of this growth has been
in low-density development and, as a consequence, transportation has
become an increasingly essential mediating factor for accessibility to
opportunities for daily life activities. Transportation can increase or re-
duce the spatial segregation for those living in the periphery of urban
centers. Poor transportation conditions can lead to spatial polarization
in which lower income groups in peripheral areas face challenges to ac-
cess quality services and facilities (Sabatini and Brain, 2008). It is for this
reason that accessibility is considered an important indicator of social
equity (Lucas, 2012).
Among urban services, green spaces are recognized as key elements
that improve urban landscapes and enhance urban sustainability. Urban
green spaces enhance the quality of life of urban residents, by improving
their physical, social, and psychological well-being (Fuller and Irvine,
2010; Keniger et al., 2013). Open green spaces are thought to promote
⁎ Corresponding author at: Department of Geography, University of Concepción, Centre
of Urban Sustainable Development (CEDEUS), Calle Victoria S/N Barrio Universitario,
Concepción, Chile.
E-mail addresses: crojasq@udec.cl (C. Rojas), paezha@mcmaster.ca (A. Páez),
olga.barbosa@uach.cl (O. Barbosa), j.carrasco@udec.cl (J. Carrasco).
social interaction and inclusion, and can support biodiversity and pro-
vide important ecosystem services as well (Bolund and Hunhammar,
1999; Chiesura, 2004; Wolch et al., 2014), This includes the provision
of habitat for urban biodiversity (Bolund and Hunhammar, 1999;
Tratalos et al., 2007), hydrological regulation (Bradshaw et al., 2007;
Ryan et al., 2010), carbon sequestration (Nowak and Crane, 2002),
and attenuation of urban heat island effects (Arnfield, 2003),
among others.
A number of benefits associated with accessibility to urban green
spaces have been researched in the past. For instance, accessibility to
green spaces increases the probability of residents visiting, and there-
fore contribute to increased physical activity (Krenichyn., 2004;
Roemmich et al., 2006). The benefits, on the other hand, are not only
to physical health, since access to green spaces offers opportunities to
reduce stress and mental fatigue (Burgess et al., 1988; Chiesura, 2004;
Van den Berg et al., 2010). Lower access to urban green spaces, in con-
trast, correlates with a higher risk of being overweight or obese, as
well as with poorer self-perceived health, and higher mortality
risks (Mitchell and Popham, 2007; Mitchell and Popham, 2008).
Moreover, recent evidence indicates that the loss of human-nature
interactions can have profound negative effects on society and, in a
city, these interactions occur mainly in open spaces. Other green fea-
tures of the city, such as private or domestic gardens, are not neces-
sarily good substitutes (Barbosa et al., 2007). Given the above, it is
not a coincidence that green spaces are of interest for public health
reasons (Dai, 2011).

http://dx.doi.org/10.1016/j.jtrangeo.2016.10.012
0966-6923/© 2016 Elsevier Ltd. All rights reserved.
228 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240
Given the potential benefits of green spaces, and the fact that their
mechanism of delivery is through accessibility pathways (Keniger et
al., 2013), accessibility to green spaces is recognized as an environmen-
tal justice issue (Wolch et al., 2014). From this perspective, accessibility
refers to the potential for exercising the right to use from particular
places, at particular times, and by specific people. Critical factors that
are relevant (see Handley et al., 2003) include physical constraints
(such as distance from home and the ability to move), as well as
socio-economic and demographic factors.
Previous research has focused primarily on two aspects of accessibil-
ity, namely, distance to green spaces and area available at that distance.
In some cases, accessibility is defined in terms of only one of these fac-
tors. For example, the European Environment Agency (EEA) recom-
mends that residents have access to green space within 15 min
walking distance of their place of residence (Stanners and Bourdeau,
1995), but does not specify the amount of green space to be reached.
More frequently, accessibility is defined as a combination of proximity
and area. Coles and Bussey (2000), for example, assess the ability to
reach at least 2 ha of green space with a 5–10 min walk (approximately
100–400 m). Van Herzele and Wiedemann (2003) propose a 5-min
walk (equivalent approximately to 400 m) to reach 1–10 ha of green
space. The measure of accessibility used by Handley et al. (2003) is
based on 2 ha of green space within 300 m from home or 20 ha within
2 km. Richardson et al. (2010) considered both walking and cycling and
defined a radius of 1300 m using a buffer around a usable area of 5 km2
of green space. The government agency English Nature, in Handley et al.
(2003), recommends a provision of at least 2 ha of green space within
300 m from the place of residence. Using floating catchment areas, Dai
(2011) determined the potential accessibility to green spaces based on
a 10 to 30 min bandwidth for walking distances. According to this anal-
ysis, accessibility within a radius of 15 min ranges between 0.49 and
18.31 acres/1000 people in a case study of Atlanta.
An underlying assumption in the analysis of accessibility is that
shorter distances to green spaces (assuming a minimum surface area)
increase accessibility – and consequently the likelihood that people
will use the area (e.g., Coles and Bussey, 2000; Peschardt and
Stigsdotter, 2013). Nevertheless, as discussed by Páez et al. (2012), al-
though these accessibility measures consider the spatial characteristics
of locations, they often fail to account for observed variations in individ-
ual mobility patterns. In other words, use of a fixed buffer (e.g., 5, 10,
15 min walks), while perhaps appropriate for normative analysis, is in-
adequate to understand accessibility for different segments of the pop-
ulation. By assuming that accessibility is determined by location only,
the characteristics of the potential users of urban services are ignored.
The assumption that accessibility depends on location only is, therefore,
unrealistic, given various well-documented variations in mobility by
different individuals (e.g., Morency et al., 2011; Reyes et al., 2014). As
a consequence, a key question, given the impact of access to green
spaces, is whether variations in patterns of mobility can exacerbate dis-
parities in accessibility to various urban services, including parks and
open spaces.
There is evidence of differential accessibility in several regions in the
world. The question of accessibility to urban green spaces has been
studied in US cities, for instance with a focus on race and ethnicity
(Dai, 2011) and in the UK in relation to geodemographic variables that
describe social diversity (Barbosa et al., 2007). Similarly, there is a rising
concern surrounding rapidly growing Asian cities, where deficient ac-
cessibility to green areas has been detected (Wolch et al., 2014; Yin
and Xu, 2009). In contrast, there is a dearth of studies in developing
countries. This is certainly the case in Chile, where rapid urban growth
has been accompanied by an uneven distribution of parks and urban
green areas (Reyes and Figueroa, 2010). The OECD (2013), for instance,
found that Chile's urban regions tend to have a relatively low supply of
green spaces: on average, just 4.15 m2 per capita. Recent studies about
the accessibility to urban green spaces in Santiago de Chile, found that
there is an unequal distribution in surface and accessibility to parks,
and that these inequalities are directly correlated with income in
three case studies of Santiago (Reyes and Figueroa, 2010) as well as
the different opportunities to spatial interactions in parks (Krellenberg
et al., 2014).
In this context, the objective of this research is to investigate the
levels of walking accessibility to urban green spaces in two medium
size Chilean cities, namely Temuco and Valdivia. Selection of these cities
was based on data availability. Data have become available, in turn, due
to a new interest in green spaces from a policy and planning perspective
in these cities. This is a timely endeavor, given that Chile is in the process
of implementing a new National Urban Policy (CNDU, 2015). The study,
thus, is well aligned with the objectives of recent urban policies, and can
provide valuable information to planners as they work towards more
equitable access and increased connectivity and accessibility of urban
public spaces including the green areas. The research is motivated by
concerns with equity, understood as justice or fairness of treatment,
which implies a proportional distribution of resources between areas,
depending on the needs and preferences of the same (Tsou et al.,
2005). This is of particular importance in countries like Chile, where
there are marked levels of socio-economic inequality. In such cases,
the distribution of open green spaces can contribute to a more just dis-
tribution of ecosystem services (Pauchard and Barbosa, 2013).
For the analysis, we follow the approach used by Reyes et al. (2014)
in Montreal, which is based on the use of adaptive bandwidths. Whereas
the focus of Reyes et al. (2014) was on children, our interest is in age,
gender, and income. These are all factors known to influence accessibil-
ity (Barbosa et al., 2007; Reyes and Figueroa, 2010; Dai, 2011; Reyes et
al., 2014). The crux of the method is the revealed behavior of travelers,
and the ease with which different individuals can overcome the dis-
tance between two places and, thus, exercise their right to use urban
spaces.
2. Study context
Temuco and Valdivia are two cities characterized by socio-spatial
segregation between traditional city centers and more recent peripheral
growth. These issues have historically been compounded by limited
urban planning. Although both cities have open green spaces with
high ecological value (CEDEUS, 2013a; CEDEUS, 2013b; Pauchard and
Barbosa, 2013), the available supply in each city is quite different.
Whereas Temuco is identified as one of the cities in Chile with the
best provision of urban green areas (with 10.98 m2 per capita), green
space is more limited in Valdivia with around 6 m2 per capita according
to a recent study (OECD, 2013).
2.1. Green spaces
We collected data regarding green spaces in the cities of Temuco and
Valdivia from two different sources. Data for Temuco were digitalized
and provided to us by the GIS Office of the Regional Government of
Araucanía (2015), and were verified by the first author through photo-
interpretation. Data regarding green spaces in Valdivia, on the other
hand, were collected by Silva et al. (2015).
In Temuco green spaces are defined as urban spaces predominantly
occupied or destined to be occupied by trees, shrubs or plants, and that
provide public leisure and recreation space. These are: parks, squares
with vegetation, and median strip. Most of them have maintenance
from the government but to improve this classification we also include
spaces with vegetation but without maintenance, such as wetland and
floodplain rivers.
In Valdivia, green spaces are parcels with permeable surfaces and
open access to the public, with predominant use for vegetation, includ-
ing grasses, trees, shrubs, and/or plants. These spaces can be used freely
for leisure and recreation activities. There are a number of different
spaces that fulfill these characteristics, including municipal parks or
plazas, public gardens attached to public buildings, sports fields,
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240 229

Table 1 2.3. Travel survey

Green spaces in Temuco and Valdivia: descriptive statistics.
Source: prepared by the author based on green spaces data from Regional Government of
Araucanía (2015) and Silva et al. (2015).
Data regarding mobility patterns were collected as part of the Temu-
co and Valdivia Origin-Destination Surveys of 2013. These are the most
Min Max Mean STD Total Distance Mean distance from recent and complete mobility databases used for transportation plan-
(Km2) (m) household (m)
ning in these cities. These data include information regarding travel be-
Temuco 0,00 2,97 0,01 0,10 10,42 3867,31 545 havior and trips of participants (e.g., number of trips, purpose, location,
Valdivia 0,00 0,46 0,01 0,03 6,05 2737,00 155
and transport mode), as well as details about individual travelers (e.g.,
age, gender, income, size of household, the number of vehicles, and oc-
cupation). Geocoded origins and destinations are available for 25,225
cemeteries, and spaces with the status of vacant lots that are used by the and 18,218 trips for Temuco and Valdivia, respectively. A total of
community in similar ways to municipal green spaces. For the purpose 10,073 residents in 3002 households were surveyed in Temuco, and
of this paper, we classify all these as green spaces and make no distinc- 6931 residents in 2105 households were surveyed in Valdivia. Descrip-
tion between them. tive statistics are shown in Table 2.

2.2. Urban green spaces in Temuco and Valdivia 2.4. Travel in Valdivia and Temuco

Temuco is located in the Araucanía Region in the South of Chile and
has a population of 311,873 (SECTRA, 2013). There are 1153 open green
spaces in the city, with a total area of 10.42 km2. The average size of
urban green areas in Temuco is 0.01 km2 and the mean distance be-
tween these is 3867 m (Table 1). For the analysis of accessibility to
green spaces, we include the flood plain of Cautín River in the southern
part of the city. This is the largest green space in the region with an area
of 5.29 km2. Temuco has a total of 9 squares and around 278 small
squares, which are distributed in the north and west of the city in
areas of recent urban growth (Fig. 1).
Valdivia is also located in the South of Chile and has a population of
161,304 (SECTRA, 2013). There are 696 open spaces with a total area of
6.05 km2 (Fig. 2). The average size of urban green spaces is 0.01 km2 and
the mean distance between these is 2737 m (Table 1). The largest green
area is located in the east and corresponds to a system of wetlands with
a total surface of 3.08 km2.
The majority of motorized trips in Temuco are performed for the
purpose of traveling to work and school. The most popular transport
modes are walking (7872 trips - 31%), transit: bus (5978 trips - 24%)
and car: driver (4229 trips - 17%). bicycle (614 trips - 2%) and multi-
modal (intermodal trips in any modes) (1774 - 7%). Meanwhile Taxi
(176 – 1%) and private bus (376 - 1.5%) have a much lower modal
share (Fig. 3).
In Valdivia, the predominant purpose for travel is traveling to work
and school. The most common modalities of travel are by car: driver
(4344 trips - 24%), walking (4265 trips - 23%), and transit: bus (2760
trips - 15%). Bicycle (453 - 3%) taxi (327 – 2%) and private bus (64 -
0.4%) have much lower transport mode (Fig. 3).
There are substantial differences in average trip lengths by transport
mode. In Temuco, average trip lengths are longer when a combination
of modes is used (multimodal: 5926 m), followed by transit:bus, with
an average length of 4226 m, 3402 m in private bus, and shared taxi

Fig. 1. Urban green spaces of Temuco. Prepared by the author based on green spaces data from Regional Government of Araucanía (2015). (For interpretation of the references to color in
this figure legend, the reader is referred to the web version of this article.)
230 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240
Fig. 2. Urban green spaces of Valdivia. Prepared by the author based on green spaces data from Silva et al. (2015). (For interpretation of the references to color in this figure legend, the
reader is referred to the web version of this article.)
with 3337 m. The average walking distance, in contrast, is only approx-
imately 815 m (Fig. 4). In Valdivia, average multimodal trips tend to be
longer than travel by any single mode (12,289 m). Trips by transit:bus
are on average 3586 m, and by car:driver 2487 m. Again, walking dis-
tances tend to be much shorter, with an average of 690 m (Fig. 4).
Traditionally, walking has been an important mode of transportation
in Chilean medium sized cities. While this is still the case in Temuco,
Valdivia (with a population a fraction the size) is seen an increase in
the use of cars.The two cities thus provide somewhat different perspec-
tives in terms of provision of green areas in growing regions.
3. Methods
3.1. Regression analysis for adaptive bandwidths
The analysis in this paper is based on the use of adaptive bandwidths
for walking trips. Adaptive bandwidths are obtained from a regression
model of trip length (see Páez et al., 2010).
To capture geographical variations in mobility, we use the spatial ex-
pansion method of Casetti (1972). The expansion method is a technique
used to generate models with spatially-varying coefficients. The expan-
sion method is not the only technique available for obtaining models
with spatially-varying coefficients. Alternatives include multilevel
models (Duncan and Jones, 2000; Páez et al., 2007), Bayesian models
(Best et al., 2005; Wheeler et al., 2014), and geographically weighted re-
gression, or GWR for short (Brunsdon et al., 1996; Páez et al., 2002).
Multilevel models can in fact be seen as a type of expanded model,
using random components. Bayesian models also operate by introduc-
ing random components, and multilevel models could be seen as
models with uninformative priors. For this paper, we prefer to expand
the models using deterministic expansions, since they are
straightforward to implement. Random coefficients would be able to
provide standard deviations on point estimates, but we feel that the
gains in information do not exceed their cost in the present case. GWR
provides local point estimates, but we prefer to use the expansion ap-
proach given concerns regarding the ability of GWR to retrieve a true
data generating process (see Páez et al., 2011; and Wheeler and
Tiefelsdorf, 2005).
It is worthwhile to note that the expansion method has been effec-
tively used in several studies in Canadian cities, such as Páez et al.
(2010) to identify food deserts in Montreal, Roorda et al. (2010) in the
study of trip generation, and Morency et al. (2011) in their investigation
of distance traveled by vulnerable population groups in Toronto, Hamil-
ton and Montreal. Of particular interest from the perspective of the
present investigation is the research of accessibility to green spaces by
children in Montreal (Reyes et al., 2014).
The expansion method operates by expanding the coefficients of an
initial model as functions of the coordinates of the observations (in the
present case, the UTM coordinates of the place of residence). The ex-
panded (or final) model thus includes location parameters. As a simple
example of this approach, consider an initial model with a constant
term (β0), one explanatory variable and corresponding parameter (Xi,
β1i), and error term (εii):
Y i ¼ β0 þ β1i X i þ εi ð1Þ
Now assume that the coefficient β1i is a function of the geographic
coordinates ui and vi (for easting and northing), represented as:
β1i ¼ f ðui ; vi Þ ð2Þ
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240 231

Table 2 The final model can be used to obtain estimates of walking trip
Travel surveys Temuco and Valdivia. length. Estimated walking trip distances are obtained for in-sample res-
Source: prepared by the author based on travel sample and expanded travel survey data.
The source of a city's income is National Socioeconomic Characterization of Minister of So-
idents. These distances, in turn, can be used to calculate accessibility
cial Development (Encuesta de Caracterización Socioeconómica Nacional (CASEN), measures to urban green spaces, as discussed next.
Ministerio de Desarrollo Social).
3.2. Accessibility indicator
Variable Sample Temuco City Sample Valdivia City
Temuco Valdivia
The accessibility indicator used for the analysis is a cumulative op-
Trips 25,225 1,008,087 18,208 561,830
Residents 10,073 311,873 6931 161,304 portunities measure, defined as follows:
Average age 35 – 37 –
X
Average 209,000 CLP 705.271CLP 240,000 CLP 563,611 CLP Api ¼ Hj ð5Þ
income (315 USD in (1056 USD in (360 USD in (848 UDS in p
^
dijbd
2016) 2016) 2016) 2016) i

Households 3002 92,569 2105 49,126

Size 4 3.37 4 3.28
household
Mean 1 0.53 1 0.6
number of
vehicles
Eq. (2) can be written as a linear form (or other polynomial expan-
sion as desired) of the coordinates, with parameters γ:
β1i ¼ γ0 þ γ1 ui þ γ2 vi ð3Þ
The final model can be obtained replacing Eq. (3) into Eq. (1):
Y i ¼ β0 þ ðγ0 þ γ1 ui þ γ 2 vi ÞX i þ εi β0 þ γ0 X i þ γ 1 ui X i þ γ 2 vi X i þ εi ð4Þ
The expanded parameters are substituted into the initial model to
yield the terminal model.
In this research, the independent variable is the natural logarithm of
trip length (Yi = log(Li)), where Li (i.e., trip length) was calculated as the
Euclidean distance from origin (households) to destinations; this trans-
formation is normally used to ensure that the model predicts positive
trip length (Reyes et al., 2014). Implementation of the expansion meth-
od was supported by statistical analysis of trip length and expanded pa-
rameters for the following variables: age, age square (to account for
potentially non-linear relationship with Li), gender and income level,
and the expansion is a combination of a quadratic trend surface and dis-
tance from Central Business District (CBD) (Figs. 1 and 2).
For modeling trip lengths, low-frequency modes were removed
from the databases, namely taxi, private bus, school bus, passenger
ferry (case of Valdivia). In addition, the analysis of accessibility is cen-
tered on the place of residence, and for this reason, we select only
home-based trips. After processing the databases, a total of 10,441
trips are available for Temuco, and 7586 trips are available for Valdivia.
Variables used for analysis are shown in Tables 3 and 4.
In Eq. (5), Api is the walking accessibility to urban green spaces by in-
dividual type p at location i, and Hj is the area of rasterized green spaces
at j, which in this case, is equivalent to 625 m2, (25 m ∗ 25 m cells). This
cell size gives a reasonable resolution for parks, and was previously used
by Reyes et al. (2014).
The summation in Eq. (5) includes all green raster spaces j that are at
a distance d from i, as long as d is less than bandwidth d ^p . The band-
ij ij i
width is obtained in an adaptive fashion based on the estimated values
of walking trip lengths for individual type p at location i – according to
the spatially expanded regression modeling approach described in the
preceding section. In this way, the levels of accessibility depend on typ-
ical mobility patterns observed in the sample. Individual accessibility by
traveler p can be obtained based on gender and age by using appropriate
values for the variables in the regression model. Estimated accessibility
by walking using adaptive bandwidths can be compared with a norma-
tive definition of accessibility suggested by English Nature in Handley et
al. (2003) that uses a fixed bandwidth of 300 m. Note that norms vary
across jurisdictions, since, as reported by Smoyer-Tomic et al. (2004, p.
292), in Edmonton the distance to parks is expected to be b800 m.
4. Results and discussion: accessibility to urban green spaces
The results of estimating regression models for trip length are shown
in Tables 5 and 6. Overall the models explain 42% and 29% of the vari-
ance in Temuco and Valdivia, respectively. This values are comparable
to those reported in previous research (e.g., Morency et al., 2011,
Table 1). A number of variables are significant at conventional levels
of significance. As a general remark, we note that the results indicate
the presence of spatial trends for several, but not all variables. The sig-
nificance of these trends is assessed in the usual way, by means of p-
values or t-scores. This results help us confirm our choice of modeling
approach with respect, in particular, to GWR, a method that implicitly
assumes a spatial trend for all coefficients (although see the more

Fig. 3. Modal split in Temuco and Valdivia. Source: prepared by the author based on Temuco and Valdivia Origin-Destination Survey of 2013.
232 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240

Fig. 4. Mean distance (m) in transport modes of Temuco and Valdivia. Source: prepared by the author based on Temuco and Valdivia Origin-Destination Survey of 2013.

cumbersome and less researched mixed GWR approach provided by
Fotheringham et al., 2002).
The coefficients for age in the case of Temuco indicate a non-linear
effect whereby trip length declines with increasing age. In Valdivia,
the effect of age is similar, but the reduction in mobility with increasing
age is less pronounced. These effects are modified by location through
the spatially expanded coefficients. For instance, the effect of age is
modified by distance from the CBD in such a fashion that, for all resi-
dents, the further away the live from the CBD, the longer their trips
tend to be. This agrees with the evidence available from other metropol-
itan regions (e.g., Morency et al., 2011; Moniruzzaman et al., 2015).
In both cities, there is a contextual effect for gender, whereby
females have shorter trip lengths on average. In Temuco the effect
the trend is of decreasing trip lengths in the north-east direction,
whereas in Valdivia the effect tends to decline towards the north of
the region.
Other variables related to trip length in these cities are driver license
availability (with a spatial trend), and occupation status. Working at
home is, as anticipated associated with shorter trips, as well as being
in the category “other occupation” (housewife). In Valdivia, in addition,
being a student is generally associated with longer trips. For estimation,
walking was set as the reference mode. Positive coefficients for other
modes indicate that they all tend to be associated in average with longer
trips than walking. In the case of income, only mid-level income was
significant in both cities. This is in agreement with other studies
where (other things being equal) the mobility of high and low-income

Fig. 5. Estimated walking distance in meters. Prepared by the author based on results of expansion method presented in Tables 5 & 6.
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240 233

Fig. 6. Estimated walking distance in meters. Prepared by the author based on results of expansion method presented in Tables 5 & 6.

Fig. 7. Estimated accessibility in sq.km. Prepared by the author based on results of expansion method reported in Tables 5 & 6.
234 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240

Fig. 8. Estimated accessibility in sq.km. Prepared by the author based on results of expansion method reported in Tables 5 & 6.

Fig. 9. Accessibility differences male and female (estimated accessibility male - estimated accessibility female). Prepared by the author based on results of expansion method reported in
Tables 5 & 6.
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240 235

Fig. 10. Accessibility differences male and female (estimated accessibility male - estimated accessibility female). Prepared by the author based on results of expansion method reported in
Tables 5 & 6.

Fig. 11. Estimated accessibility in sq.km in a band distance of 300 m. Prepared by the author based on fixed radio of 300 m.
236 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240

Fig. 12. Estimated accessibility in sq.km in a band distance of 300 m. Prepared by the author based on fixed radio of 300 m.

individuals tends to be similar (e.g., Moniruzzaman et al., 2015). This Given the presence of several spatial trends, the most effective way
could potentially be a relation of availability of means to travel and of parsing the results of the model is by mapping the estimates of trip
the need to travel. length. The spatial pattern indicates shorter trips in the central parts

Table 3
List of variables and descriptive statistics Temuco.

Dimensions Variable Descriptions Min Max Mean STD

Location Distance to CBD (m) Distance in meters 115.13 16,394.79 4077.30 3240.16
u (easting coordinate) Coordinate based on false origin 13.19 16.36 13.19 0.04
v (northing coordinate) Coordinate based on false origin 0.00 10.59 4.54 2.12
Age and gender Age (years) Age 0.00 93.00 36.01 20.07
Female Yes = 1 0.00 1.00 0.54 0.50
Size household 1 person Yes = 1 0.00 1.00 0.03 0.17
2 people Yes = 1 0.00 1.00 0.14 0.35
3 people Yes = 1 0.00 1.00 0.22 0.41
4 people or more than Yes = 1 0.00 1.00 0.61 0.49
Occupation Employment Yes = 1 0.00 1.00 0.34 0.47
Student Yes = 1 0.00 1.00 0.29 0.45
Retired Yes = 1 0.00 1.00 0.08 0.27
At Home Yes = 1 0.00 1.00 0.08 0.27
Unemployed Yes = 1 0.00 1.00 0.02 0.15
Not-Work Yes = 1 0.00 1.00 0.03 0.17
Other Occupation Yes = 1 0.00 1.00 0.14 0.35
Income Low-Income b$480 US (yes = 1) 0.00 1.00 0.17 0.38
Mid-level Income $480 US–$1300 US (yes = 1) 0.00 1.00 0.15 0.36
High-level Income N$1300 US (yes = 1) 0.00 1.00 0.11 0.32
Driver license Driver License Yes = 1 0.00 1.00 0.33 0.47
Transportation mode Car: driver Yes = 1 0.00 1.00 0.17 0.37
Car: passenger Yes = 1 0.00 1.00 0.10 0.31
Shared taxi Yes = 1 0.00 1.00 0.09 0.29
Transit: bus Yes = 1 0.00 1.00 0.27 0.44
Walk Yes = 1 0.00 1.00 0.34 0.47
Bicycle Yes = 1 0.00 1.00 0.03 0.16
Car ownership 0 Car Yes = 1 0.00 1.00 0.34 0.47
1 Car Yes = 1 0.00 1.00 0.50 0.50
2 Cars Yes = 1 0.00 1.00 0.29 0.45
3 or more than Cars Yes = 1 0.00 1.00 0.15 0.36
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240 237

Table 4
List of variables and descriptive statistics Valdivia.

Dimensions Variable Descriptions Min Max Mean STD

Location Distance to CBD (m) Distance in meters 105.07 14,586.12 3500.82 2301.90
u (easting coordinate) Coordinate based on false origin 0.00 17.67 14.66 2.96
v (northing coordinate) Coordinate based on false origin 0.00 9.14 4.30 1.71
Age and Gender Age (years) Age 0.00 98.00 36.89 20.01
Female Yes = 1 0.00 1.00 0.52 0.50
Size Household 1 person Yes = 1 0.00 1.00 0.03 0.16
2 people Yes = 1 0.00 1.00 0.15 0.36
3 people Yes = 1 0.00 1.00 0.22 0.42
4 people or more than Yes = 1 0.00 1.00 0.60 0.49
Occupation Employment Yes = 1 0.00 1.00 0.35 0.48
Student Yes = 1 0.00 1.00 0.31 0.46
Retired Yes = 1 0.00 1.00 0.08 0.28
At Home Yes = 1 0.00 1.00 0.09 0.28
Unemployed Yes = 1 0.00 1.00 0.02 0.15
Not-work Yes = 1 0.00 0.00 0.00 0.00
Other occupation Yes = 1 0.00 1.00 0.13 0.33
Income Low-income b$480 US (yes = 1) 0.00 1.00 0.72 0.45
Mid-level income $480 US–$1300 US (yes = 1) 0.00 1.00 0.23 0.42
High-level income N$1300 US (yes = 1) 0.00 1.00 0.05 0.22
Driver license Driver license Yes = 1 0.00 1.00 0.37 0.48
Transportation mode Car: driver Yes = 1 0.00 1.00 0.23 0.42
Car: passenger Yes = 1 0.00 1.00 0.16 0.37
Shared taxi Yes = 1 0.00 1.00 0.14 0.35
Transit: bus Yes = 1 0.00 1.00 0.17 0.38
Walk Yes = 1 0.00 1.00 0.26 0.44
Bicycle Yes = 1 0.00 1.00 0.03 0.17
Car ownership 0 Car Yes = 1 0.00 1.00 0.45 0.50
1 Car Yes = 1 0.00 1.00 0.40 0.49
2 Cars Yes = 1 0.00 1.00 0.13 0.33
3 or more than cars Yes = 1 0.00 1.00 0.03 0.17

of the city, and longer trips in the periphery, as seen in Figs. 5 and 6. The (health centers, education facilities, and government institutions)
model results indicate an estimated mean walking trip length of 472 m (Figs. 5 and 6).
– with a maximum of 886 m - in Temuco; and an estimated mean walk- The next step in the analysis is to calculate accessibility. This we do
ing trip length of 237 m – with a maximum of 936 m – in Valdivia. As using Eq. (5), and the estimates of trip length after the models presented
seen in the figures, walking distance tends to be b300 m in areas sur- in Tables 5 and 6. The variables used for the calculation of trip length
rounding the central business district where most services are located were the attributes of respondents in the sample (travel survey). In
this way, trip length is calculated using the observed attributes of re-
Table 5
spondents. However, since we are interested in estimating walking ac-
Regression model of trip length with spatially expanded coefficient Temuco. cessibility, the mode of transport is set to walking. That is, the variable
is set to 1 for this mode and 0 for all other modes for all respondents.
Variable Estimate p-Value
Estimated walking accessibility for Temuco's sample is on average
Constant Const 555,526 0.000 0.13 km2 with a maximum of 0.72 km2, whereas in Valdivia walking ac-
Age Age 2,738,620,624 0.000
cessibility sample tends to be somewhat lower with 0.05km2 with a
*CBD –225,284,599 0.000
*u2 7,294,233 0.000 maximum value of 0.59 km2. The total provision of green spaces in Te-
*u 79,985,525 0.000 muco is greater than Valdivia (10 km2 against 6 km2) and in addition,
*uv –5,717,713 0.000 there are important differences in the geographic distribution of green
*v 208,124,958 0.000 spaces in both cities. The location of green spaces in Valdivia tends to
2
*v 2,940,184 0.000
Age2 –721,252,143 0.000
be more sprawling, and the spaces larger in size. In addition, in Valdivia,
*CBD 59,218,853 0.000 the average distance from place of residence to green space is 545 m,
*u2 –1,903,459 0.000 which is much higher than in Temuco, where this distance is 155 m
*u –20,251,346 0.000 (see locations in Figs. 1 and 2). The results indicate that walking acces-
*v –51,383,099 0.000
2 sibility tends to be higher in Temuco. This is a consequence of a tenden-
*v –2,083,086 0.000
Gender Female (yes = 1) cy towards longer walking trips, as well as a shorter distance between
*u –0.02849 0.068 green spaces and places of residence, especially in the city center.
*v –0.15799 0.002 The results of our accessibility calculations are presented in Figs. 7
Occupation At home –0.14391 0.000 and 8. The map for Temuco shows the estimated values of accessibility
Others –0.23133 0.000
in square kilometers range from 0 (white color) to 1.3 km2 (intense
Income Mid-level income
*u 2
–0.06441 0.000 green), with a mean of 0.13 km2. It bears noting that 0.005 km2 repre-
*v2 0.5073 0.000 sents the minimum standard of park size associated with benefits for
Driver license (yes = 1) mental health (Dadvand et al., 2016). The first color class in the map
Driver license *uv –0.15176 0.000
(white) represents accessibility values of zero. These can be found to
Transportation Mode Car: driver 209,186 0.000
Car: passenger 207,031 0.000 the south of the city center (CBD), where the levels of accessibility are
Shared taxi 242,346 0.000 extremely low. The next color classification in the map represents the
Transit: bus 243,292 0.000 adequate levels of accessibility, meaning that residents have access to
Bicycle 171,763 0.000 at least 5000 m2 of green spaces. Locations with this level of accessibility
R2 = 0.42, σ = −0.376, n = 10,441 are close to the city center where green spaces, especially small parks,
238 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240

Table 6 demographic or socio-economic attribute included in the model. Con-

Regression model of trip length with spatially expanded coefficient Valdivia. sider, for example, the case of gender. The results of the model indicate
Variable Estimate p-Value that in both cities females tend to travel shorter distances than males. As
Constant Const 503,522 0.000
a consequence of lower mobility, accessibility levels of females tend to
Trend surface Distance to CBD 0.85467 0.001 be lower relative to those of males (0.08 km2 to 0.10 km2). In compari-
Age Age 2,088,147 0.000 son, mean accessibility in Valdivia is approximately 0.05 km2, both for
*CBD –1,467,678 0.000 males and females. These summary values, however, mask important
*u2 1,384,743 0.000
spatial variations in accessibility by gender. To illustrate this, Figs. 9
*u –2,825,075 0.000
*uv –1,355,054 0.002 and 10 shows the differences in accessibility between genders. In
*v these maps, negative values and zero in white color indicate locations
*v2 1,714,594 0.003 where females have higher accessibility than males, and positive values
Age2 –3,127,023 0.000 show areas where males have higher accessibility.
*CBD 2,127,079 0.000
*u2 –212,301 0.000
In addition to accessibility calculated using revealed behavior to de-
*u 4,323,786 0.000 rive adaptive bandwidths, it is interesting to calculate accessibility using
*uv 2,003,027 0.005 a fixed ratio as it is commonly done in the literature. Páez et al. (2012)
*v2 –2,507,501 0.007 discuss the conceptual differences between a normative approach
Gender Female (yes = 1)
(based on revealed behavior) and a normative approach (based on a
*u 250,878 0.000
*v –829,562 0.000 regulatory framework). As noted before, English Nature recommends
*uv 524,166 0.000 a maximum distance of 300 m to a park at least of 0.02 km2 in area
*u2 –155,118 0.000 (Handley et al., 2003). This value, in fact, is similar to the mean estimat-
Occupation Students 0.15192 0.001 ed walking distance in both Temuco and Valdivia. It is to be expected,
At home –0.34788 0
however, that the use of a fix bandwidth (say, of 300 m) will underesti-
Others –0.15001 0.009
Income Mid-level income mate accessibility in places where people are more mobile, and will
*uv 0.72535 0.007 overestimate it in places where mobility is more limited. Figs. 11 and
*v2 –253,499 0.010 12 shows in zeros the areas were accessibility for people do not satisfy
Driver License (yes = 1) –819,897 0.000
the norm (300 m). If we consider that estimated trip lengths represent
*CBD 467,224 0.000
*u 809,164 0.000
the actual mobility patterns and willingness to travel, we can see that a
Driver License *uv –272,309 0.012 fixed bandwidth tends to overestimate accessibility in Temuco around
*u2 –315,675 0.000 the city center, while it tends to substantially underestimate accessibil-
*v 1,096,934 0.000 ity in more peripheral parts of the city. In the case of Valdivia, there are
*v2 –50,591 0.001
important regional differences with the highest values in the north part
Transportation mode Car: driver 186,901 0.000
Car: passenger 188,511 0.000 of the city (see Figs. 11 and 12).
Shared taxi 237,029 0.000 In summary, the analysis reveals that accessibility varies significant-
Transit: bus 243,923 0.000 ly between the two cities and also within each city. In Temuco, people
Bicycle 184,648 0.000
living in the city center enjoy the greatest levels of accessibility to
Size household 3 people 0.08696 0.026
green space, with a mean accessibility of 0.13 km2 to green areas, in ex-
R2 = 0.29, s = 2.603, n = 7585 cess of 0.005 km2 (Dadvand et al., 2016). Residents of Valdivia, mean-
while, with generally shorter walking trips, have generally lower
and plazas are situated. The next color class in the maps represents ac- levels of accessibility. The analysis also reveals that there are significant
cessibility between 0.005 km2 and 0.02 km2. This is observed in places differences in accessibility by gender, attributable to the lower mobility
where walking trips tend to be longer, and also around Municipal Stadi- of females.
um Park and Ñielol Hill (Fig. 1). Locations with higher accessibility
(N0.02 km2) are immediately north of the flood plain of Cautin River 5. Conclusions
in Temuco to the southwest and northeast. In contrast, accessibility
levels are more modest in the south of the city where green spaces are In this paper, we have explored the question of accessibility to green
more scarce. Despite the presence of a large park, walking trips there spaces in two Chilean cities. The focus of the paper has been on the ap-
tend to be shorter, thus reducing the accessibility of potential users plication of a method to calculate adaptive bandwidths (see Páez et al.,
(Figs. 7 and 8). 2010) and accessibility by walking.
In the case of Valdivia, the estimated values of accessibility range be- This paper makes a number of contributions. First, it contributes to a
tween 0 and 0.34 km2, with a mean value of 0.05 km2. Accessibility growing literature on accessibility to green spaces in developing coun-
levels for Valdivia are shown in Fig. 8 where it can be seen that accessi- tries. A dearth of knowledge on this topic, as well as large socio-eco-
bility N0.005 km2 are found around the city center, whereas higher nomic disparities, make it difficult to generate guidelines for urban
levels of accessibility with values of up to 0.02 km2 are seen in proximity policy (ONU-Habitat, 2012). Secondly, this research is a prerequisite
to Isla Teja across the river to the north, as well as in the neighborhoods for further examinations of the positive contributions of green spaces
nearest to the wetlands areas south of the CBD (see Fig. 2). Isla Teja is an to the well-being of the population. Third, by looking at Chilean cities,
interesting case, since it is a residential sub-center of the CBD, character- we present a case study of medium size cities that have not previously
ized by mixed uses such as commercial uses and campus university and been reported in the literature. In the context of Chile, both cities are
schools, with large landmark green spaces, such as the Botanical Garden experiencing rapid peripheral growth with low-density residential
and the Saval Park (see Fig. 2). The cluster of green spaces around the areas, which, as the analysis suggests could lead to lower levels of acces-
wetlands, paired with relatively long walking trips (see Fig. 6), lead to sibility for more segments of the population. Also, most of the Chilean
higher levels of accessibility (N 0.02 km2). In contrast, accessibility levels cities are experimenting a phenomenon of housing densification and
are lower to the east outside of city center because of more limited sup- segregation without green areas. And finally, this research presents a
ply of green spaces there. case for investigating accessibility differentials by demographic and
The results of the analysis indicate important spatial variations in ac- socio-economic groups, using walking and transport mode. Comparison
cessibility to green spaces in these cities. In addition, a model-based ap- with a fixed bandwidth, of the type commonly used in the literature,
proach allows us to explore differentials in accessibility by any relevant shows that using revealed behavior can provide a more realistic picture
 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240
of accessibility levels. An important conclusion is to highlight the limita-
tions of measuring accessibility to green spaces using city-wide aver-
ages or even indicators that are insensitive to mobility differentials by
location or socio-economic and demographic profile. Mobility, as our
analysis demonstrates, strongly influences the potential of the public
to reach those spaces.
In the present paper, we concentrated on accessibility by walking,
which provides a characterization of social conditions associated with
residents. In fact, distance or walking time from home has been
shown in many studies as the single most important precondition for
use of green spaces (Van Herzele and Wiedemann, 2003; Giles-Corti
et al., 2005). On the other hand, quantity (surface) and quality of
those green spaces are of major importance. A recent study by
Shanahan et al. (2015), used dose–response modeling, to explore the
relationship between “nature-dose” (percentage of green space per
neighborhood, minutes exposed to environment and naturalness of
green space) and health measures. These authors highlight the need
for taking more systematic approaches to understanding how interven-
tions in the urban environment could be directed to improve the health
and well-being of urban dwellers, for which a closer collaboration
among health professional, social scientist, urban planners and ecolo-
gists is needed. In this context, we believe our paper contributes not
only by providing evidence about accessibility to green spaces in cities
in a developing country, but also about how this access space is distrib-
uted among different members of the society, in a region where this
kind of information is fragmented, not systematically measured, or is
simply lacking.
Studies of this kind can help to inform policymaking in other Latin
American countries that like Chile have a predominantly urbanized
population. For example, a U.N. report shows that half of 16 cities in
Latin America, exceeded the World Health Organization recommenda-
tion of having between 9 and 11 m2 of green space per inhabitant, how-
ever the lack of a uniform and consistent methodology, the uneven
distribution and size of them, and the lack of quality measures, prevents
making accurate conclusions that can derive on sound guidelines and
recommendations (ONU-Habitat, 2012). In the case of Chile, it is rele-
vant due to the urban inheritance law (approved 1975) strongly focused
on increasing housing in response to a long-term housing deficit, over
other aspects of urban design such as the provision of sufficient public
green spaces. Although parks, plazas and open spaces with vegetation
are defined as a space with public use, there does not exist a real polit-
ical will to encourage an increase in the green areas of cities, especially
in mid-size cities, as most of them have tight budgets to maintain these
areas.
More concretely, the present analysis suggests a number of opportu-
nities to reduce existing socio-economic inequalities in accessibility in
urban areas by transforming existing (informal) spaces into formal
green spaces (Barbosa and Villagra, 2015). This can be done in a
targeted fashion, based on the results of this analysis by providing so-
called “cues to care” (Nassauer, 1995a, 1995b). Social inequities impact
not only economic opportunities but also the environmental quality of
surrounding landscapes (Rozzi, 2013; Barbosa and Villagra, 2015). In-
creased accessibility to green spaces, the main providers of ecosystem
services on cities, can help to ameliorate some of those inequities
(Pauchard and Barbosa, 2013).
In closing, it is worthwhile to note that there are a number of ave-
nues that could be followed to extend the research presented in this
paper. First, future research could incorporate information about the
frequency of use and activities conducted in green spaces. This type of
data could also enhance our understanding of other factors that might
influence the willingness and actual visits to certain green spaces,
such as local exhibitions, festivals and other recreational activities,
that may attract urban residents from locations beyond walking dis-
tance. Secondly, the analysis could be complemented by examining
other modes of transportation. In addition to the alignment with
urban planning goals (e.g., Municipalidad de Valdivia, 2015),
239
accessibility by bicycle is of particular interest as this mode, like walking,
is a complement to other physical activities that could enhance public
health. This is relevant given the increase in the prevalence of obesity
in the developing world, including Chile (Kain et al., 2002). Finally, in-
formation on health indicators would prove valuable to explore the
question of whether multi-modal accessibility to green spaces does
translate in better health outcomes. These are questions for future
research.
References
Arnfield, A.J., 2003. Two decades of urban climate research: a review of turbulence, ex-
change of energy and water, and the urban heat island. Int. J. Climatol. 23, 1–26.
Barbosa, O., Villagra, P., May Jr., R.H., 2015. Socio-ecological studies in urban and rural eco-
systems in Chile. In: Rozzi, R., Chapin., S.F., Callicott, J.B., Pickett, S.T.A., Power, M.E.,
Armesto, J.J. (Eds.), Earth Stewardship: Linking Ecology and Ethics in Theory and
Praxis. Springer Verlag, Berlin-Heidelberg, pp. 297–311.
Barbosa, O., Tratalos, J., Armsworth, P., Davies, R., Fuller, R., Johnson, P., Gaston, K., 2007.
Who benefits from access to green space? A case study from Sheffield, UK. Landsc.
Urban Plan. 83 (2–3), 187–195.
Best, N., Richardson., S., Thomson., A., 2005. A comparison of Bayesian spatial models for
disease mapping. Stat. Methods Med. Res. 14 (1), 35–59.
Bolund, P., Hunhammar, S., 1999. Ecosystem services in urban areas. Ecol. Econ. 29 (2),
293–301.
Bradshaw, C., Sodhi, N., Peh, K., Brook, B., 2007. Global evidence that deforestation am-
plifies flood risk and severity in the developing world. Glob. Chang. Biol. 13,
2379–2395.
Brunsdon, C., Fotheringham, A.S., Charlton, M.E., 1996. Geographically weighted regres-
sion: a method for exploring spatial nonstationarity. Geogr. Anal. 28, 281–298.
Burgess, J., Harrison, C., Limb, M., 1988. People, parks and the urban green: a study of pop-
ular meanings and values for open spaces in the City. Urban Stud. 25 (6), 455–473.
Casetti, E., 1972. Generating models by the expansion method: applications to geograph-
ical research. Geogr. Anal. 4, 81–91.
CEDEUS (Centro de Desarrollo Urbano Sustentable), 2013a. Síntesis Principales
Resultados Conversatorio Urbanos 2013. Padre las Casas. Centro de Desarrollo Urbano
Sustentable, Temuco, Chile Online at. http://www.cedeus.cl.
CEDEUS (Centro de Desarrollo Urbano Sustentable), 2013b. Síntesis Principales
Resultados Conversatorios Urbanos 2013. Centro de Desarrollo Urbano Sustentable,
Valdivia, Chile Online at. http://www.cedeus.cl.
Chiesura, A., 2004. The role of urban parks for the sustainable city. Landsc. Urban Plan. 68
(1), 129–138.
CNDU (Consejo Nacional de Desarrollo Urbano), 2015. Política Nacional Urbana. Consejo
Nacional de Desarrollo Urbano, Santiago, Chile Online. http://cndu.gob.cl/wp-
content/uploads/2014/10/L4-Politica-Nacional-Urbana.pdf.
Coles, R.W., Bussey, S.C., 2000. Urban forest landscapes in the UK - progressing the social
agenda. Landsc. Urban Plan. 52 (2–3), 181–188.
Dadvand, P., Bartoll, X., Basagaña, X., Dalmau-Bueno, A., Martinez, D., Ambros, A., Cirach,
M., Triguero-Mas, M., Gascon, M., Borrel, C., Nieuwenhuijsen, M., 2016. Green spaces
and general health: roles of mental health status, social support, and physical activity.
Environ. Int. 91, 161–167.
Dai, D., 2011. Racial/ethnic and socioeconomic disparities in urban green space accessibil-
ity: where to intervene? Landsc. Urban Plan. 102 (4), 234–244.
Duncan, C., Jones, K., 2000. Using multilevel models to model heterogeneity: potential
and pitfalls. Geogr. Anal. 32, 279–305.
Fotheringham, A.S., Brunsdon, C., Charlton, M., 2002. Geographically weighted regression:
The analysis of spatially varying relationships. Wiley, Chichester.
Fuller, R., Irvine, K., 2010. Interactions between people and nature in urban environments.
In: Gaston, K. (Ed.), Urban Ecology. Cambridge University Press, Cambridge,
pp. 134–171.
Giles-Corti, B., Broomhall, M., Knuiman, M., Collins, C., Douglas, K., Ng, K., Donovan, R.J.,
2005. Increasing walking: how important is distance to, attractiveness, and size of
public open space? Am. J. Prev. Med. 28 (2 Suppl. 2), 169–176.
Handley, J., Pauleit, S., Slinn, P., Lindley, S., Barber, L., Jones, C., 2003. Accessible natural
green space standards in towns and cities: a review and toolkit for their implemen-
tation. London, U.K.: English Nature Research Reports (526) Online. http://
publications.naturalengland.org.uk/file/82004.
Kain, J., Uauy, R., Vio, F., Albala, C., 2002. Original communications-trends in overweight
and obesity prevalence in Chilean children: comparison of three definitions. Eur.
J. Clin. Nutr. 56, 200–204.
Keniger, L., Gaston, K., Irvine, K., Fuller, R., 2013. What are the benefits of interacting with
Nature ? Int. J. Environ. Res. Public Health 10, 913–935.
Krellenberg, K., Welz, J., Reyes-Päcke, S., 2014. Urban green areas and their potential for
social interaction - a case study of a socio-economically mixed neighbourhood in San-
tiago de Chile. Habitat Int. 44, 11–21.
Krenichyn., K., 2004. Women and physical activity in an urban park: enrichment and sup-
port through and ethic of care. J. Environ. Psychol. 24, 117–130.
Lucas, K., 2012. Transport and social exclusion: where are we now? Transp. Policy 20,
105–113.
Mitchell, R., Popham, F., 2007. Greenspace, urbanity and health: relationships in England.
J. Epidemiol. Community Health 61 (8), 681–683.
Mitchell, R., Popham, F., 2008. Effect of exposure to natural environment on health in-
equalities: an observational population study. Lancet 372 (9650), 1655–1660.
240 C. Rojas et al. / Journal of Transport Geography 57 (2016) 227–240
Moniruzzaman, M., Páez, A., Scott, D., Morency, C., 2015. Trip generation of seniors and
the geography of walking in Montreal. Environ. Plan. A 47 (4), 957–976.
Morency, C., Páez, A., Roorda, M., Mercado, R., Farber, S., 2011. Distance traveled in three
Canadian cities: spatial analysis from the perspective of vulnerable population seg-
ments. J. Transp. Geogr. 19 (1), 39–50.
Municipalidad de Valdivia, 2015. Plan de Acción Capital Sostenible. Valdivia, Chile. Online
at. http://www.iadb.org/es/temas/ciudades-emergentes-y-sostenibles/plan-de-
accion-de-valdivia-chile,19095.html.
Nassauer, J., 1995a. Culture and changing landscape structure. Landsc. Ecol. 10, 229–237.
Nassauer, J., 1995b. Messy ecosystems, orderly frames. Landsc. J. 14, 161–170.
Nowak, D., Crane, D., 2002. Carbon storage and sequestration by urban trees in the USA.
Environ. Pollut. 116, 381–389.
OECD (Organization for Economic Co-operation and Development), 2013n. OECD Urban
Policy Reviews, the Chilean Urban System and its Challenges. OECD, Paris, France On-
line at. http://www.oecd-ilibrary.org/urban-rural-and-regional-development/oecd-
urban-policy-reviews-chile-2013/the-chilean-urban-system-and-its-challenges_
9789264191808-4-en.
ONU-Habitat, 2012. Estado de las ciudades de América Latina y el Caribe, 2012. Rumbo a
una nueva transición urbana, Online. http://www.onuhabitat.org.
Páez, A., Uchida, T., Miyamoto, K., 2002. A general framework for estimation and inference
of geographically weighted regression models: 1. Location-specific kernel band-
widths and a test for locational heterogeneity. Environ. Plan. A 34, 733–754.
Páez, A., Scott, D., Potoglou, D., Kanaroglou, P., Newbold, K.B., 2007. Elderly mobility: de-
mographic and spatial analysis of trip making in the Hamilton CMA, Canada. Urban
Stud. 44, 123–146.
Páez, A., Farber, S., Wheeler, D., 2011. A simulation-based study of geographically weight-
ed regression as a method for investigating spatially varying relationships. Environ.
Plan. A 43 (12), 2992–3010.
Páez, A., Scott, D., Morency, C., 2012. Measuring accessibility: positive and nor-mative
implementations of various accessibility indicators. J. Transp. Geogr. 25, 141–153.
Pauchard, A., Barbosa, O., 2013. Regional assessment of Latin America: rapid urban devel-
opment and social economic inequity threaten biodiversity hotspots, chapter 28. In:
Elmqvist, T., Fragkias, M., Goodness, J. (Eds.), Urbanization, Biodiversity and Ecosys-
tem Services: Challenges and Opportunities. Springer, Berlin-Heidelberg,
pp. 589–608.
Peschardt, K., Stigsdotter, U., 2013. Associations between park characteristics and per-
ceived restorativeness of small public urban green spaces. Landsc. Urban Plan. 112
(1), 26–39.
Reyes, S., Figueroa, I., 2010. Distribución, superficie y accesibilidad de las áreas verdes en
Santiago de Chile. Revista latinoamericana de estudios urbanos regionales EURE (San-
tiago) 36 (109), 89–110.
Reyes, M., Páez, A., Morency, C., 2014. Walking accessibility to urban parks by children: a
case study of Montreal. Landsc. Urban Plan. 125, 38–47.
Richardson, E., Pearce, J., Mitchell, R., Day, P., Kingham, S., 2010. The association between
green space and cause-specific mortality in urban New Zealand: an ecological analy-
sis of green space utility. BMC Public Health 10:240 Online. http://www.
biomedcentral.com/1471-2458/10/240.
Roemmich, J., Epstein, L., Raja, S., Yin, L., Robinson, J., Winiewicz, D., 2006. Association of
access to parks and recreational facilities with the physical activity of young children.
Prev. Med. 43, 437–441.
View publication stats
Roorda, M.J., Páez, A., Morency, C., Mercado, R., Farber, S., 2010. Trip generation of vulner-
able populations in three Canadian cities: a spatial ordered probit approach. Trans-
portation 37 (3), 525–548.
Rozzi, R., 2013. Biocultural ethics: from biocultural homogenization toward biocultural
conservation, chapter 2. In: Rozzi, R., Pickett, S.T.A., Palmer, C., Armesto, J.J.,
Callicott, J.B. (Eds.), Linking Ecology and Ethics for a Changing World: Values, Philos-
ophy, and Action. Springer, Berlin-Heidelberg, pp. 9–32.
Ryan, J., McAlpine, C., Ludwig, J., 2010. Integrated vegetation designs for enhancing water
retention and recycling in agroecosystems. Landsc. Ecol. 25, 1277–1288.
Sabatini, F., Brain, I., 2008. La segregación, los guetos y la integración social urbana: Mitos
y claves. Revista latinoamericana de estudios urbanos regionales EURE (Santiago) 34,
5–26.
SECTRA (Secretaria de Planificación de Transporte), 2013. Transport Origin-Destination
Survey. Secretaria de Planificación de Transporte, Ministerio de Transporte y
Telecomunicaciones, Gobierno de Chile, Santiago, Chile Online at. http://www.
sectra.gob.cl.
Shanahan, D., Fuller, R., Bush, R., Lin, B., Gaston, K., 2015. The health benefits of urban na-
ture: how much do we need? Bioscience 65 (5), 476–485.
Silva, C., García, C., Estay, S., Barbosa, O., 2015. Bird richness and abundance in response to
urban form in a Latin American City: Valdivia, Chile as a case study. PLoS One 10 (9),
e0138120.
Smoyer-Tomic, K., Hewko, J., Hodgson, J., 2004. Spatial accessibility and equity of play-
grounds in Edmonton, Canada. Can. Geogr. 48 (3), 287–302.
Stanners, D., Bourdeau, P., 1995. The urban environment. In: Stanners, D., Bourdeau, P.
(Eds.), Europe's Environment: The Dobris Assessment. Copenhagen, European Envi-
ronment Agency Online. http://www.eea.europa.eu/publications/92-826-5409-5.
Tratalos, J., Fuller, R., Philip, W., Davies, R., Gaston, K., 2007. Urban form, biodiversity po-
tential and ecosystem services. Landsc. Urban Plan. 83, 308–317.
Tsou, K., Hung, Y., Chang, Y., 2005. An accessibility-based integrated measure of relative
spatial equity in urban public facilities. Cities 22 (6), 424–435.
Van den Berg, A., Maas, J., Verheij, R., Groenewegen, P., 2010. Green space as a buffer be-
tween stressful life events and health. Soc. Sci. Med. 70 (8), 1203–1210.
Van Herzele, A., Wiedemann, T., 2003. A monitoring tool for the provision of accessible
and attractive urban green spaces. Landsc. Urban Plan. 63 (2), 109–126.
Wheeler, D., Tiefelsdorf, M., 2005. Multicollinearity and correlation among local regres-
sion coefficients in geographically weighted regression. J. Geogr. Syst. 7, 161–187.
Wheeler, D., Páez, A., Spinney, J., Waller, L.A., 2014. A Bayesian approach to hedonic price
analysis. Pap. Reg. Sci. 93, 663–683.
Wolch, J., Byrne, J., Newell, J., 2014. Urban green space, public health, and environmental
justice: the challenge of making cities “just green enough.”. Landsc. Urban Plan. 125,
234–244.
Yin, H., Xu, J., 2009. Spatial accessibility and equity of parks in Shangai. Urban Stud. 6,
71–76.
Páez, A., Mercado, R., Farber, S., Morency, C., Roorda, M., 2010. Relative accessibility dep-
rivation indicators for urban settings: definitions and application to food deserts in
Montreal. Urban Stud. 47 (7), 1415–1438.