Urban Forestry & Urban Greening 29 (2018) 58–67
Contents lists available at ScienceDirect
Urban Forestry & Urban Greening journal homepage: www.elsevier.com/locate/ufug
Original article
Spatial disparities in neighborhood public tree coverage: Do modes of transportation matter? Haoluan Wanga, Feng Qiub, a b
T
⁎
Department of Agricultural and Resource Economics, University of Maryland, College Park, MD 20742, USA Department of Resource Economics and Environmental Sociology, University of Alberta, Edmonton, Alberta, T6G 2H1, Canada
A R T I C L E I N F O
A B S T R A C T
Keywords: Resident modes of transportation Public tree coverage Spatial regression model Edmonton Sustainable community development
Urban green space has various environmental and ecological benefits, and uneven access to such amenities has drawn substantial attention from policy makers in developing sustainable community planning. In this study, we illustrate the spatial distribution of publicly owned and maintained trees in Edmonton, Canada and assess neighborhoods’ heterogeneous tree availability by using the container approach. Through spatial regression models, we further investigate the association of neighborhood public tree availability with socio-economic status (SES). We contribute to the existing literature by taking resident modes of transportation into consideration, in addition to many other commonly examined SES such as household income and ethnicity. Another unique contribution of this study is that we distinguish trees planted on different location types (i.e., boulevard, park, and buffer areas) when exploring the unequal coverage across neighborhoods and among different SES groups. Key results include: (1) a general examination without differentiating location types can lead to misleading results and thus provide inappropriate policy recommendations; (2) resident modes of transportation is a critical factor associated with a neighborhood’s public tree coverage; and (3) there exists evident spatial dependence on public tree availability between neighborhoods. The results from this study provide important information to better understand the issue and to allocate public resource (such as tree coverage) more efficiently and effectively to support sustainable community development.
1. Introduction Growing numbers of studies have indicated a variety of benefits that urban green space can bring to cityscapes and residents’ quality of life. From the environmental and ecological perspective, urban greenery helps purify airborne pollutants (Jim and Chen, 2008; Nowak et al., 2006), filtrate storm-water runoff (Pataki et al., 2011; Zhang et al., 2012), and mitigate urban heat islands (Jansson et al., 2007; Onishi et al., 2010). In addition to the benefits to the city environment, urban green space also improves human mental and physical health, as many experimental studies have shown (e.g., Barton and Pretty, 2010; Sugiyama et al., 2008; van den Berg et al., 2015; van den Berg et al., 2016). Furthermore, urban green space also contributes to reducing crime rates through a number of channels, including the promotion of physical activities and encouragement of social interactions (Donovan and Prestemon, 2012; Fleming et al., 2016; Troy et al., 2012). Consequently, access to green space in an urban setting has drawn substantial attention from policy makers in developing sustainable growth and planning strategies (Chen and Chang, 2015; Pham et al., 2012; Shan,
2014). In response to urban environmental justice, another body of research has investigated the unequal access to urban green space and/or tree canopy that represents environmental disparities among groups with varying socio-economic status (SES). Typically, urban residents with deprived SES have poor access and/or lower coverage of green space in their communities. For example, racial/ethnic minorities, such as black and Hispanic populations, have a smaller amount of tree canopy and less green space in their neighborhoods (Dai, 2011; Zhou and Kim, 2013). Prior studies (Krafft and Fryd, 2016; Landry and Chakraborty, 2009; Zhou and Kim, 2013) have also found that household income is positively associated with access to green space and the availability of tree canopy in the community. One potential reason is that residents with high incomes possess the greater social capital to influence community policies (e.g., park development plans) in their neighborhoods (Luck et al., 2009; Merse et al., 2009). Another possible explanation is that residents with high incomes are able and willing to pay more for environmental quality/goods. Meanwhile, mixed outcomes have been reported with respect to other SES groups. For
⁎ Corresponding author at: 509 General Services Building, Department of Resource Economics and Environmental Sociology, University of Alberta, Edmonton, Alberta, T6G 2H1, Canada. E-mail addresses:
[email protected] (H. Wang),
[email protected] (F. Qiu).
https://doi.org/10.1016/j.ufug.2017.11.001 Received 28 March 2017; Received in revised form 31 October 2017; Accepted 2 November 2017 Available online 10 November 2017 1618-8667/ © 2017 Elsevier GmbH. All rights reserved.
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
access to urban green space and the growing disparities in urban green space accessibility among different socio-economic groups, differentiating the type of urban green space has important policy implications In the literature, there exists much variation in findings between studies investigating different components of urban green space. For example, Lotfi and Koohsari (2011) and Scopelliti et al. (2016) examined the proximity to urban parks and residents’ experience of these facilities, indicating that high-income residents have greater access to urban parks and thus a higher level of well-being from the experience. Pham et al. (2012) and Pham et al. (2013) constructed various urban vegetation indicators (e.g., trees/shrubs, lawn, and total vegetation) and found that the availability of street and backyard vegetation is positively related to residents with university education but negatively associated with the presence of recent immigrants. Another stream of studies specifically focuses on the association of tree coverage and accessibility with residents’ socio-economic conditions. For example, Zhou and Kim (2013) chose tree canopy cover as an indicator of urban tree spatial distribution using aerial photography analysis and showed that racial/ethnic minorities have less tree canopy in their neighborhoods. Landry and Chakraborty (2009) focused on street trees in residential areas, and their results indicated that neighborhoods with a higher proportion of African-Americans and renters have a lower rate of tree cover. Chuang et al. (2017) looked at urban tree canopy and found that stable-wealthy neighborhoods are more likely to have more, and more consistent, tree canopy cover than other neighborhood types.
instance, renters tend to have fewer incentives to maintain neighborhood landscapes, trees in particular (Perkins et al., 2004). Previous studies have confirmed a negative relationship between the proportion of renter-occupied properties and residential tree canopy cover (Landry and Chakraborty, 2009; Li et al., 2015). However, Pham et al. (2013) found that communities with higher renter populations have more street and backyard vegetation. Although both tree canopy coverage and/or accessibility have been widely studied (e.g., Krafft and Fryd, 2016; Landry and Chakraborty, 2009; Zhou and Kim, 2013), research specifically focusing on tree count in an urban setting is scarce, partially due to unavailable or limited data sets (Mills et al., 2016). Additionally, though a variety of neighborhood SESs have been investigated in relation to tree coverage, the associations of various modes of transportation with tree availability are generally neglected in the literature. For example, Dai (2011) included car ownership in the analysis and found that households without cars are usually in neighborhoods with limited greenery. However, the current literature has barely incorporated other commuting modes such as walking and bicycling in the analysis of association with tree coverage. Much evidence has indicated that residents with better access to green environments tend to walk or bike more (Giles-Corti et al., 2005; Takano et al., 2002). Meanwhile, visiting green space on foot or by bike has helped maintain urban dwellers’ physical and mental health and promoted harmonious community development, from both social and environmental perspectives (Beyer et al., 2014; Sugiyama et al., 2008; van den Berg et al., 2016). Therefore, an investigation of the associations between resident modes of transportation and neighborhood tree coverage will provide useful information for policy makers whose aim is to design and evaluate programs that promote sustainable development from multiple perspectives. Another distinction between this study and existing research is that we take into account the role of different location types when conducting the association analysis, while prior studies only focused on tree coverage in a general sense or relied on one specific location type for analysis (e.g., Kendal et al., 2012; Lotfi and Koohsari, 2011). Limited previous research has compared the association results across different location types. As we will show in the results section, a general analysis without differentiation can result in misleading conclusions. An exploration of heterogeneous associations between neighborhood tree availability and SES by location type will thus offer more tailored information regarding the reasons and explanations behind certain inequality in community greenery access. Taking advantage of a unique dataset that includes detailed individual tree information (such as location, planted date, and type of location) across the entire city, we first spatially depict the distribution and coverage of publicly owned and maintained trees at the neighborhood level in Edmonton, Canada. We then investigate the associations between public tree availability in terms of tree count per resident and neighborhood SES using both the classic linear regression model and spatial lag model, with the latter one controlling for the potential spatial autocorrelation issue. We use the word “tree count” to represent “tree count per resident” in the rest of the paper, unless otherwise specified. We further divide tree coverage by location type to compare to the general model and to show the heterogeneous effects associated with diverse SES groups.
2.2. Tree coverage measurement In the literature, different metrics have been proposed to analyze the accessibility and/or availability of urban green space, among which geographic information systems (GIS) are most widely used. Based on road networks, distances between neighborhoods and urban greenery are commonly measured (Kessel et al., 2009; Lotfi and Koohsari, 2011). Other studies have used the coverage method, also referred to as the “container approach,” to count the presence of trees within a particular aggregated geographic unit such as a neighborhood (Krafft and Fryd, 2016; Landry and Chakraborty, 2009). More advanced research has taken into account the effect of transportation systems through social networks (Chen and Chang, 2015; Dai, 2011) and adopted visualization techniques such as Google Street View (Li et al., 2015) to more accurately assess the green space availability at a micro level. Using satellite imagery of tree cover often fails to provide information on tree diversity, density, and its vertical structure. Therefore, a more recent study utilized Forest Inventory and Analysis inventory data with detailed tree counts and tree species information to examine the relationship between greenness and socioeconomic status information (Mills et al., 2016). Previous studies used the proximity approach (e.g., Euclidean distance or nearest distance through road networks) to quantify the accessibility to urban green space, primarily based on the geometric centroids of census zones and targeted green space parcels (Dai, 2011; Lotfi and Koohsari, 2011). However, such a method has an evident flaw when measuring clustered and high-density data such as tree coverage, as the number of trees in a neighborhood is usually substantial. Using the proximity method to calculate the nearest tree or tree cluster does not reflect the richness of the greenery. Alternatively, a “container approach” (Maroko et al., 2009; Mills et al., 2016) has been adopted to analyze tree coverage by relying on the count numbers. This study followed this practice and utilized the container approach. However, one can possibly argue that tree count in a geographic unit may not reflect residents’ actual tree availability. For example, there are neighborhoods with large numbers of trees but no population. We thus modified the tree count per resident in the empirical analysis to solve this problem. Such a concept has been used in a recent study in a
2. Literature review 2.1. Urban green space and tree coverage According to previous studies (Dai, 2011; Kessel et al., 2009), urban green space can be a very broad concept that includes parks, recreational facilities, public gardens, greenways, cemeteries, and historical preservations. Although the accessibility of urban green space in general has been widely studied, recent research tends to focus on various green space types separately. Given the multiple factors that influence 59
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
expressed as follows:
different setting where the amount of urban green in a certain buffer area per capita is calculated (Wüstemann et al., 2017).
K
yi = α +
2.3. Modes of transportation
∑
x ik βk + εi
k=1
where yi is neighborhood i errorimage ’s tree count per resident, xik is a SES factor associated with neighborhood i, and βk is the corresponding coefficient to be estimated. Based on prior studies, these SES variables include the percentage of the child population aged 18 and younger (Children), the percentage of the senior population aged 65 and older (Senior), the median household income (Median Income), the percentage of minority groups (Minority, which refers to immigrants who are mainly South Asian, Chinese, European, black, and Latin American), the percentage of residents as renters (Renter), the percentage of private car access (Car, which refers to individuals who use a car, truck, or van as their primary commuting transportation, including passengers and drivers). We additionally include the percentage of residents who choose to walk (Walk) and bicycle (Bicycle) as the primary daily mode of transportation. εi is an unobserved i.i.d. error term. Note that all modes of transportation are based on residents who are older than 15 and employed, and resident modes of transportation may not necessarily represent residents’ preferred modes; the observed data reveal residents’ choices of commuting options given constrains such as availability and costs. Another concern is the correlation between unemployment and transportation modes. We checked the correlation between resident modes of transportation and unemployment rate at the neighborhood level. The correlation analysis shows that the two are not highly correlated with each other. However, the existence of the spatial autocorrelation could likely violate the assumption of the linear regression model and thus lead to biased coefficient estimation. Such a concern stems from at least two reasons. First, closer neighborhoods tend to have more characteristics in common than distant ones, which implicitly induces spatial autocorrelation (Wang et al., 2016). Second, the location of urban green space tends to be spatially clustered (Pham et al., 2012). To control for the potential spatial dependence, we adopt a spatial lag model (also referred to as an SAR model) in the following equation:
The transportation mode plays an essential role in defining the accessibility of urban green space as it more realistically captures residents’ actual accessibility. Taking this research interest into account, a growing number of literature focuses on how to incorporate various transit options, such as walking and public transportation, into the assessment of green space accessibility (Chen and Chang, 2015; Comber et al., 2008; Fan et al., 2017; Shan, 2014). However, limited research has been conducted to directly investigate the relationship between various modes of transportation and urban space accessibility/availability. For example, Dai (2011) analyzed household car ownership at the census tract level and found that households without cars are usually in neighborhoods with limited urban green space. The current literature has barely incorporated other commuting modes such as walking and bicycling in the analysis of association with tree coverage, although such research interest has be expressed in other fields such as urban food access (McKenzie, 2014; Wang and Qiu, 2016). As discussed previously, an investigation of the associations between resident modes of transportation and neighborhood public tree coverage could provide more insights for policy makers to better design and evaluate programs that promote sustainable urban development. 3. Data and methods 3.1. Study area and data The study area is Edmonton, Canada, where proactive policies with respect to the improvement of urban green space and neighborhood sustainable development are being undertaken. Edmonton is the capital city of the province of Alberta, covering about 684 km2 with a population of approximately 0.88 million in 2014. As the fifth largest city in Canada in terms of population, the city is taking active efforts to create sustainability and resilience in the region with a strong emphasis on residents’ connection to nature and biodiversity. The tree dataset was obtained from the City of Edmonton’s Open Data Catalogue, Trees – Species (Map View), which is maintained by the Urban Forestry Unit (City of Edmonton, 2011b). In addition to the geographic location of each publicly owned and maintained tree, the dataset also contains other tree information such as planted date and location types (e.g., parks, boulevards, buffer areas, and other public areas such as center medians, service lanes, schools, and libraries). The neighborhood SES information was extracted from the National Household Survey (Statistics Canada, 2011). The survey includes a wide range of neighborhood information such as total population, population by different age groups, ethnicity, median income, living status, and primary transportation modes for daily commute that is all aggregated at the neighborhood level. In addition, the City of Edmonton provides the boundary shapefile of 388 defined neighborhoods that can also be retrieved from the Open Data Catalogue. By connecting neighborhood SES information to neighborhood boundaries, we identified 245 residential neighborhoods with demographic information. The other 143 are non-residential neighborhoods (mainly industrial areas) that have no residents or census data and, thus, are excluded from the empirical analysis.
K
J
∑
x ik βk + ρ ∑ wij yj + εi
k=1
j=1
where the spatial lag,
∑
yi = α +
J
wij yj is the weighted average of the nearby
j=1
neighborhoods’ tree count per resident, and ρ is the spatial autoregressive coefficient that measures the strength of spatial dependence. Following Chuang et al. (2017), Dai (2011), and Pham et al. (2012), we used the queen continuity weight (i.e. immediately adjacent neighborhoods) as the spatial weight matrix because residential neighborhoods in Edmonton generally share similar size and shape. For each model (i.e., general, boulevard, buffer, and park), we presented the Lagrange Multiplier and Robust Lagrange Multiplier tests in Table 3, based on which we decided whether or not the spatial dependence occurs at the dependent variable and that a spatial lag model should be adopted. 4. Results 4.1. Descriptive results
3.2. Methods
Fig. 1 shows the spatial distribution of public tree coverage in Edmonton. It is evident that almost all residential neighborhoods have public tree coverage. With respect to the tree count per resident, the high coverage is concentrated in three areas that include mature neighborhoods (i.e., neighborhoods in the center of the city that were built up before the 1970s), the university area, and relatively wealthy neighborhoods in the southwestern region of the city. In terms of tree
Using the ordinary least squares (OLS) technique, the classic linear regression model is a common practice in the literature to examine the associations between the availability of urban green space or tree coverage and neighborhood SES (Dai, 2011; Li et al., 2015; Zhou and Kim, 2013). Specifically, the linear regression model using OLS can be 60
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
Fig. 1. Tree count per resident distribution at the neighborhood level.
possessing approximately two trees per person. The average tree count per resident in boulevards and parks at the neighborhood level is 0.09 and 0.11, respectively. Fig. 2 provides some details about the spatial distribution of trees planted in different locations, which is also aggregated at the neighborhood level. For trees planted along boulevards, neighborhoods in the city core have the highest per-resident tree count. Such an outcome is expected, as most mature neighborhoods that have large roads with decent planning and maintenance of street trees are located in this region. In contrast, the distribution of tree count in the buffer areas shows the opposite pattern as most of these zones are
location type as shown in Fig. 2, most of the trees have been planted in parks and along boulevards, comprising, respectively, 45% and 40% of all trees. About 12% of the trees are located in buffer areas (i.e., zones between commercial and residential districts in the city), with the rest in other landscapes (e.g., public areas such as center medians, service lanes, schools, and libraries) representing only 3% of the total. Table 1 lists the descriptive statistics of the dependent and independent variables for the regression models at the neighborhood level. The residential neighborhoods in Edmonton have an average of 0.36 trees per resident, with some neighborhoods well covered and 61
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
Fig. 2. Tree count per resident distributions by location type.
neighborhoods with highest count (i.e., No. 6, 7, and 9) are mature neighborhoods located in the center of the city, and three neighborhoods with the lowest count (i.e., No. 1, 2, and 3) are in the city periphery. Table 2 compares the neighborhood SES of these ten selected neighborhoods with the city average. In general, neighborhoods with the highest tree count have lower rates of child population and minority groups, higher median income, and higher private car usage than the city mean. The five neighborhoods with the lowest tree count have a smaller percentage of minority and renter groups as well as lower rates of residents choosing to walk or bicycle in their primary daily commute than the city average.
located in the north and south of the city and have heavy industry and commercial activities. For trees located in parks, however, no particular pattern is observed, with the neighborhoods having the highest number spread out across the city. 4.2. Container approach results Further analysis based on the container approach leads to the identification of neighborhoods with the highest and lowest per-resident public tree count. Fig. 3 shows the locations of five selected neighborhoods for each category. As expected, three of the five 62
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
children do not have better public environmental amenities in terms of greenery in buffers and parks. This may be an area that deserves future policy and community efforts. Similar findings are also identified to the senior population. In the general model, there shows an insignificant relationship between the percentage of seniors and public tree count per resident. However, in the disaggregated case, neighborhoods with a higher rate of senior population are found to have higher public tree coverage in parks and buffer areas and lower in boulevards. Second, the minority group’s relationship with the availability of boulevard trees is negative and significant. This is again because this type of tree is clustered in mature neighborhoods and the southwestern part of the city (Fig. 2), with the majority of residents being non-immigrant. As an increasing body of new immigrants is residing in the southern region of the city, where new residential areas and large roads are being built up, the lack of street trees in those neighborhoods deserves further policy efforts. Preference may be given to street trees when considering enhancing greenery investments, especially those related to public funds, in these neighborhoods. In addition, we find that neighborhoods with a higher rate of renters tend to have lower public tree availability in buffer areas. This outcome is somewhat contrary to Pham et al. (2013), who found that residents in Montreal possess more residential greenery due to the city’s distinctive pattern of housing tenure in areas with a high percentage of renters. In the case of Edmonton, as most renters are clustering in the city core and the areas surrounding the university, which are the two regions with the highest tree coverage on boulevards due to historical development strategies of the city but lowest public tree coverage in buffer areas (Fig. 2). Therefore, such a result is not unreasonable. Third, the significant and positive coefficient of Walk in the buffer model indicates that neighborhoods with more residents choosing walking as their primary mode of transportation tend to have more tree coverage. Residents with better access to a green environment tend to walk and stay outside more (Giles-Corti et al., 2005; Takano et al., 2002), and our results can provide evidence to support this point of view. For example, walking maps in Edmonton show that most walking and bicycle trails in the city are located in mature and university neighborhoods along the North Saskatchewan River as well as relatively wealthy neighborhoods in the southwestern region of the city (City of Edmonton, 2017). Finally, in terms of spatial dependence, the adoption of the spatial lag model increases the goodness of fit of the models. Second, the spatial autoregressive coefficient, ρ, is positive and significant in all four models when different dependent variables are selected. The results confirm strong spatial dependence on the public tree availability, regardless of the location types. Such a finding shall provide important information for policy makers and community leaders to design more effective programs with respect to public investment when taking spatial interactions into consideration, which will be discussed in more detail in the next section.
Table 1 Descriptive statistics for dependent and independent variables (N = 245). Variablesa
Median
Tree Count per Resident (No.) All Location Types 0.29 Boulevard 0.09 Buffer 0.03 Park 0.11 Children (%) 23.43 Seniors (%) 10.92 Median Income (1000 CAD 35.71 $) Minority (%) 23.54 Renter (%) 9.41 Car (%) 42.00 Walk (%) 1.22 Bicycle (%) 0.00
Mean
Std. Dev.
Minimum
Maximum
0.36 0.15 0.05 0.15 22.81 12.39 37.52
0.27 0.17 0.06 0.18 5.52 7.11 9.02
0.00 0.00 0.00 0.00 4.88 1.10 3.23
1.98 0.81 0.28 1.71 35.05 43.27 65.22
24.01 12.16 41.76 1.89 0.48
9.16 11.06 7.68 2.56 0.98
0.91 0.00 9.32 0.00 0.00
56.60 64.52 68.64 19.71 8.10
a We investigated the potential multicollinearity problem among independent variables and found that the coefficients of the correlation matrix are all relatively small.
4.3. Regression results Based on the Lagrange Multiplier and Robust Lagrange Multiplier tests (Table 3), we conclude that there exists evident spatial autocorrelation in the dependent variables. In detail, a larger Lagrange multiplier statistic and a lower P-value indicate a higher level of spatial dependence and a higher significance level, respectively. The simple Lagrange Multiplier (lag) specifically tests for a missing spatially lagged dependent variable, and the Robust Lagrange Multiplier (lag) further tests for lagged dependence in the possible presence of error dependence. As we can see from Table 3, both the simple and robust tests of Lagrange Multiplier (lag) are significant, indicating strong presence of spatial dependence in the lagged dependent variable. Therefore, we present results from both the regular regression models and from spatial lag models which take into account the spatial effects. Results from the general regression model using the aggregate tree count as dependent variable are presented in Table 4. Similar to other studies in North America reporting that high-income neighborhoods tend to possess higher tree coverage (e.g., Landry and Chakraborty, 2009; Zhou and Kim, 2013), we also discover a positive and significant relationship between neighborhood median income and public tree availability in Edmonton. In terms of resident modes of transportation, however, the coefficient estimates of all three variables (i.e., Car, Walk, and Bicycle) turn out to be statistically insignificant. There also shows an insignificant association of the rates of minority and renter groups with public tree count per resident. But before we jump to a conclusion and make policy recommendations based on the above results, it is necessary to investigate more details by separately assessing the associations by location type. To further explore specific associations when public trees are planted in different locations, we re-estimate the regression models by location type and present the results in Table 5. Unlike the results from the general model presented in Table 4, the associations between neighborhood SES and public tree count show some apparent distinctions, which shall be especially important for appropriate policy recommendations. First, unlike the insignificant association indicated by the general model, neighborhoods with a higher rate of child population have significantly less public trees planted in Boulevards. The otherwise misleading insignificant relationship is a consequence of aggregating all trees planted in different locations and expecting them to have identical relationship with neighborhood SES. In Edmonton, most of the neighborhoods with large child populations are located farther away from the city center (i.e., the city core and university areas), the areas with major roads and highways and thus plenty of trees along boulevards. The results also indicate that in Edmonton, neighborhoods with more
5. Discussion and policy implications Realizing the multi-dimensional benefits, such as improving resident mental and physical wellbeing and mitigating the heat-island effects, of planting more trees, the City of Edmonton advocated the maintenance of urban forests by continuing to invest and expand the city’s tree inventory, and adopted the “no net loss” approach to the replacement of public trees. At the neighborhood level, the Neighborhood Park Development Program was established in 2015 to provide assistance and support for green space development (City of Edmonton, 2015). Projects that aim to improve amenities in neighborhood parks, trees in particular, are being undertaken. Results from this study have particular relevance for this program. First, the visualized tree distributions by location type, and the identified five neighborhoods with lowest public tree availability in terms of tree count, in particular, can help pinpoint hotspots for decision makers to allocate 63
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
Fig. 3. Identified neighborhoods with highest and lowest tree count per resident.
urban housing literature, numerous studies have found evidence that individuals and households pay extra premium (called implicit price) for neighborhood environmental amenities including green space/tree coverage (Donovan and Butry, 2010; Hussain et al., 2014; Luttik, 2000), which is reflected in the housing prices they are paying. The willingness to pay of course will depend on individual/household characteristics such as how often and how much they are able to enjoy the amenities. From a policy perspective, this indicates that promoting a more healthy and active lifestyle could potentially contribute to community environmental improvement. In our tree planting case,
public resources to where it was most needed. Second, our results indicate that improving the availability of publicly owned and maintained trees can make multi-dimensional contributions to a sustainable living and community development. According to the regression analysis, neighborhoods with more residents choosing walking as a commuting mode are positively associated with public tree availability on boulevard and in buffer areas. Individuals who prefer walking are more likely to enjoy the environmental amenities, and potentially are more willing to pay to improve the neighborhood environmental quality such as planting more trees. In 64
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
Table 2 Statistics summary for neighborhoods with five highest and five lowest tree count per resident (N = 10). Neighborhoodsa
Tree Count Per Resident (No.)
Children (%)
Seniors (%)
Median Income (1000 CAD$)
Minority (%)
Renter (%)
Car (%)
Walk (%)
Bicycle (%)
1 2 3 4 5 6 7 8 9 10 City Average
0 0.02 0.03 0.05 0.06 1.09 1.11 1.68 1.96 1.98 0.36
17.62 19.50 26.13 28.49 25.67 19.19 16.92 14.67 13.04 12.64 22.81
15.52 15.35 10.81 5.72 17.62 27.64 11.44 9.78 38.41 6.32 12.39
37.42 30.64 36.44 37.58 39.36 34.12 37.36 53.32 32.97 45.76 37.52
6.24 8.28 16.53 22.87 32.65 20.21 9.58 17.51 9.46 20.15 24.01
3.57 6.21 3.31 8.08 16.33 12.12 11.71 6.78 20.81 29.91 12.16
52.11 34.85 38.29 48.73 37.16 34.55 49.75 59.78 27.54 47.13 41.76
0.89 0.00 0.00 0.00 1.22 4.23 0.00 4.52 0.00 11.60 1.89
0.00 0.00 0.00 0.00 0.00 1.92 0.00 0.00 0.00 0.00 0.48
a Neighborhoods 1–5 with lowest tree count: Westview Village, Evergreen, Rural North East South Sturgeon, Crawford Plains, and Westbrook Estates; Neighborhoods 6–10 with highest tree count: Woodcroft, Bellevue, Cloverdale, Virginia Park, and Rossdale.
resulting from increased property values and retail sales. In 2011, the City of Edmonton launched an important environmental strategic plan, The Way We Green, to call for strong action to protect the environment across the city, and to promote sustainable community development (City of Edmonton, 2011a). Our findings provide evidence and support that the environmental improvement as part of community development will further contribute to sustainable development in a broad context including urban infrastructure, social equity, and municipal governance. It is important for urban planners and policy makers to take into consideration these positive externalities from urban greenery improvement projects such as public trees when conducting benefit-cost analysis. Finally, the results from this study also indicate strong spatial effects in public tree availability between neighborhoods. Positive spatial effects imply that areas with low (or high) public tree coverage tend to cluster together in certain parts of the city (as shown in Figs. 1 and 2). These regions might be labeled as environmentally unfavorable hot spots that deserve additional investigation and policy attention. Given the positive spatial effects found in this study, improved public tree availability in one of these clustered unfavorable neighborhoods could potentially encourage more public trees in the nearby areas, thus improving the overall public tree coverage in the entire hot spots. One feasible policy intervention is to encourage coordinated planning and thus share some common resources among neighboring communities that are in lack of public tree coverage.
Table 3 Lagrange multiplier tests for spatial dependence. Lagrange multiplier (lag) General Statistic (P-value) Robust statistic (P-value)
***
46.175 (0.000) 10.226*** (0.001)
Boulevard ***
87.465 (0.000) 12.948*** (0.000)
Buffer
Park ***
52.270 (0.000) 4.965** (0.026)
5.749** (0.017) 4.895** (0.027)
***, **, and * indicate the coefficient is significant respectively at 1%, 5%, and 10% level. Table 4 Regression results (N = 245): General. Variables
Tree count per resident OLS
Spatial Lag
ρ
−1.356 (0.560) 0.232 (0.377) 0.007*** (0.002) −0.226 (0.192) −0.022 (0.266) 0.350 (0.342) 0.825 (0.983) 0.087 (1.922) 0.272 (0.297) –
R2 Pseudo R2
0.163 –
Children Senior Median Income Minority Renter Car Walk Bicycle Constant
**
−0.715 (0.497) 0.375 (0.490) 0.004** (0.002) −0.185 (0.163) −0.188 (0.228) 0.297 (0.289) 0.647 (0.831) −0.935 (1.641) 0.029 (0.258) 0.619*** (0.149) – 0.386
6. Conclusions Urban green space has various environmental and ecological benefits, and unequal availability of such amenities has drawn growing attention from policy makers in developing sustainable community planning. Expanding community green space can benefit from working with other relevant projects such as encouraging an active lifestyle and reducing the urban carbon footprint, as they mutually affect each other. Collaborations (e.g., coordinated planning) among relevant stakeholders and decision makers can help design more effective and efficient strategies with respect to the improvement of urban greenery such as public tree maintenance and thus allocate public resources more appropriately. In this study, we depict the spatial distribution of publicly owned and maintained trees in Edmonton, Canada and assess neighborhoods’ heterogeneous tree availability by using the container approach. Through the classic linear regression and spatial regression models, we further investigate the association of public tree availability with neighborhood SES. We find that neighborhoods with higher median income have higher public tree count planted in parks, but not along boulevards. With respect to resident modes of transportation, neighborhoods with more residents choosing walking as their primary mode of transportation tend to have more public tree availability in buffer
*** and ** indicate the coefficient is significant respectively at 1% and 5% level. Standard errors are in parentheses.
encouraging residents to walk more through, for example, Walkability Strategy Project (Stantec, 2010) could increase residents’ demand for, and thus willingness to pay for investment on neighborhood tree availability (i.e., through property taxes and fees for specific programs/ projects). On the other hand, more tree count further brings many other benefits to the community and the city such as better air quality, improved resident health and reduced risk of diseases through increasing physical activity, reduced congestion and energy consumption by replacing driving with walking and biking, and increased tax revenues 65
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
Table 5 Regression results (N = 245): Type of location. Variables
Tree count per resident Boulevard OLS
Spatial Lag
ρ
−1.391 (0.334) −0.601*** (0.225) 0.002 (0.001) −0.338*** (0.115) −0.171 (0.158) 0.061 (0.204) −0.364 (0.586) 2.346** (1.146) 0.551*** (0.177) –
R2 Pseudo R2
0.225 –
Children Senior Median Income Minority Renter Car Walk Bicycle Constant
Buffer
***
−0.711 (0.306) −0.349* (0.186) 0.001 (0.001) −0.219** (0.094) −0.180 (0.125) 0.112 (0.161) −0.134 (0.464) 0.732 (0.976) 0.248 (0.156) 0.642*** (0.149) – 0.509 **
OLS **
0.310 (0.120) 0.289*** (0.081) −0.001* (0.001) 0.021 (0.041) −0.036 (0.057) 0.071 (0.073) 0.307 (0.210) 0.543 (0.411) −0.062 (0.063) – 0.091 –
Park Spatial Lag
OLS
Spatial Lag
0.057 (0.121) 0.097 (0.084) −0.001 (0.000) −0.010 (0.037) −0.103** (0.052) 0.042 (0.064) 0.328* (0.183) 0.081 (0.374) −0.007 (0.057) 0.911*** (0.219) – 0.301
−0.290 (0.389) 0.495* (0.262) 0.006*** (0.002) 0.114 (0.134) 0.179 (0.184) 0.189 (0.237) 0.865 (0.682) −2.958** (1.334) −0.194 (0.206) –
−0.146 (0.375) 0.528** (0.250) 0.004** (0.002) 0.063 (0.129) 0.080 (0.180) 0.117 (0.229) 0.505 (0.667) −2.132 (1.314) −0.183 (0.197) 0.557*** (0.212) – 0.159
0.113 –
***, **, and * indicate the coefficient is significant respectively at 1%, 5%, and 10% level. Standard errors are in parentheses.
areas. Finally, we find strong spatial dependence on public tree availability, regardless of the location types. Despite the distinct contributions of the study, there are limitations that remain the subject of future research. One limitation is the exclusion of the 143 non-residential neighborhoods that potentially ignores the “border” effect. For example, residents may live in a neighborhood devoid of trees but their street might border a non-residential area with abundant trees. Future studies may find it helpful to adopt an approach measuring the actual tree availability taking into consideration of the edge effects. In addition, future studies may also want to incorporate privately-owned trees together with the publicly-owned ones and to compare the potential disparities in association with neighborhoods SES, if there are any.
Oregon. Landsc. Urban Plann. 94 (2), 77–83. Donovan, G., Prestemon, J., 2012. The effect of trees on crime in Portland, Oregon. Environ. Behav. 44 (1), 3–30. Fan, P., Xu, L., Yue, W., Chen, J., 2017. Accessibility of public urban green space in an urban periphery: the case of Shanghai. Landsc. Urban Plann. 165, 177–192. Fleming, C.M., Manning, M., Ambrey, C.L., 2016. Crime, greenspace and life satisfaction: an evaluation of the New Zealand experience. Landsc. Urban Plann. 149, 1–10. Giles-Corti, B., Broomhall, M.H., Knuiman, M., et al., 2005. Increasing walking-how important is distance to, attractiveness, and size of public open space? Am. J. Prev. Med. 28, 169–176. Hussain, M.R.M., Tukiman, I., Zen, I.H., Shahli, F.M., 2014. The impact of landscape design on house prices and values in residential development in urban areas. APCBEE Procedia 10, 316–320. Jansson, C., Jansson, P.E., Gustafsson, D., 2007. Near surface climate in an urban vegetated park and its surroundings. Theor. Appl. Climatol. 89 (3), 185–193. Jim, C.Y., Chen, W., 2008. Assessing the ecosystem service of air pollutant removal by urban trees in Guangzhou (China). J. Environ. Manag. 88, 665–676. Kendal, D., Williams, N.S.G., Williams, K.J.H., 2012. Drivers of diversity and tree cover in gardens, parks and streetscapes in an Australian city. Urban For. Urban Green. 11, 257–265. Kessel, A., Green, J., Pinder, R., Wilkinson, P., Grundy, C., Lachowycz, K., 2009. Multidisciplinary research in public health: a case study of research on access to green space. Public Health 123 (1), 32–38. Krafft, J., Fryd, O., 2016. Spatiotemporal patterns of tree canopy cover and socioeconomics in Melbourne. Urban For. Urban Green. 15, 45–52. Landry, S., Chakraborty, J., 2009. Street trees and equity: evaluating the spatial distribution of an urban amenity. Environ. Plann. A 41 (11), 2651–2670. Li, X., Zhang, C., Li, W., Kuzovkina, Y.A., Weiner, D., 2015. Who lives in greener neighborhoods? The distribution of street greenery and its association with residents’ socioeconomic conditions in Hartford, Connecticut, USA. Urban For. Urban Green. 14, 751–759. Lotfi, S., Koohsari, M.J., 2011. Proximity to neighborhood public open space across different socio-economic status areas in Metropolitan Tehran. Environ. Justice 4 (3), 179–184. Luck, G., Smallbone, L., O’Brien, R., 2009. Socio-economics and vegetation change in urban ecosystems: patterns in space and time. Ecosystems 12, 604–620. Luttik, J., 2000. The value of trees, water and open space as reflected by house prices in the Netherlands. Landsc. Urban Plann. 48 (3–4), 161–167. Maroko, A.R., Maantay, J.A., Sohler, N.L., Grady, K.L., Arno, P.S., 2009. The complexities of measuring access to parks and physical activity sites in New York City: a quantitative and qualitative approach. Int. J. Health Geogr. 8 (1). McKenzie, B.S., 2014. Access to supermarkets among poorer neighborhoods: a comparison of time and distance measures. Urban Geogr. 35 (1), 133–151. Merse, C.L., Buckley, G.L., Boone, C.G., 2009. Street trees and urban renewal: a Baltimore case study. Geogr. Bull. 50, 65–81. Mills, J.R., Cunningham, P., Donovan, G.H., 2016. Urban forests and social inequality in the Pacific Northwest. Urban For. Urban Green. 16, 188–196. Nowak, D.J., Crane, D.E., Stevens, J.C., 2006. Air pollution removal by urban trees and shrubs in the United States. Urban For. Urban Green. 4 (3–4), 115–123.
References Barton, J., Pretty, J., 2010. What is the best dose of nature and green exercise for improving mental health? A multi-study analysis. Environ. Sci. Technol. 44 (10), 3947–3955. Beyer, K.M.M., Kaltenbach, A., Szabo, A., Bogar, S., Nieto, F.J., Malecki, K.M., 2014. Exposure to neighborhood green space and mental health: evidence from the survey of the health of wisconsin. Int. J. Environ. Res. Public Health 11 (3), 3453–3472. Chen, J., Chang, Z., 2015. Rethinking urban green space accessibility: evaluating and optimizing public transportation system through social network analysis in megacities. Landsc. Urban Plann. 143, 150–159. Chuang, W., Boone, C.G., Locke, D.H., Grove, J.M., Whitmer, A., Buckley, G., Zhang, S., 2017. Tree canopy change and neighborhood stability: a comparative analysis of Washington D.C. and Baltimore, MD. Urban For. Urban Green. 27, 363–372. City of Edmonton, 2011a. The Way We Green: The City of Edmonton’s Environmental Strategic Plan. Available at: http://www.edmonton.ca/city_government/documents/ PDF/TheWayWeGreen-approved.pdf. City of Edmonton, 2011b. Trees −Species (Map View). Available at https://data. edmonton.ca/Environmental-Services/Trees-Species-Map-View-/cggb-hzzm. City of Edmonton, 2015. Neighborhood Park Development Program. Available at http:// www.edmonton.ca/programs_services/documents/PDF/NPDP_Community_Manual. pdf. City of Edmonton, 2017. Bike Maps. Available at https://www.edmonton.ca/ transportation/cycling_walking/bicycle-map.aspx. Comber, A., Brunsdon, C., Green, E., 2008. Using a GIS-based network analysis to determine urban green space accessibility for different ethnic and religious groups. Landsc. Urban Plann. 86 (1), 103–114. Dai, D., 2011. Racial/ethnic and socioeconomic disparities in urban green space accessibility: where to intervene? Landsc. Urban Plann. 102, 234–244. Donovan, G.H., Butry, D.T., 2010. Trees in the city: valuing street trees in Portland,
66
Urban Forestry & Urban Greening 29 (2018) 58–67
H. Wang, F. Qiu
greenness with physical and mental health: do walking, social coherence and local social interaction explain the relationships? J. Epidemiol. Community Health 62, e9. Takano, T., Nakamura, K., Watanabe, M., 2002. Urban residential environments and senior citizens’ longevity in megacity areas: the importance of walkable green spaces. J. Epidemiol. Commun. Health 56, 913–918. Troy, A., Grove, J.M., O’Neil-Dunne, J., 2012. The relationship between tree canopy and crime rates across an urban-rural gradient in the greater Baltimore region. Landsc. Urban Plann. 106, 262–270. van den Berg, M., Wendel-Vos, W., van Poppel, M., Kemper, H., van Mechelen, W., Maas, J., 2015. Health benefits of green spaces in the living environment: a systematic reveiw of epidemiological studies. Urban For. Urban Green. 14 (4), 806–816. van den Berg, M., van Poppel, M., van Kamp, I., et al., 2016. Visiting green space is associated with mental health and vitality: a cross-sectional study in four European cities. Health Place 38, 8–15. Wang, H., Qiu, F., 2016. Fresh food access revisited. Cities 51, 64–73. Wang, H., Tao, L., Qiu, F., Lu, W., 2016. The role of socio-economic status and spatial effects on fresh food access: two case studies in Canada. Appl. Geogr. 67, 27–38. Wüstemann, H., Kalisch, D., Kolbe, J., 2017. Access to urban green space and environmental inequalities in Germany. Landsc. Urban Plann. 164, 124–131. Zhang, B., Xie, G., Zhang, C., Zhang, J., 2012. The economic benefits of rainwater-runoff reduction by urban green spaces: a case study in Beijing, China. J. Environ. Manag. 100, 65–71. Zhou, X., Kim, J., 2013. Social disparities in tree canopy and park accessibility: a case study of six cities in Illinois using GIS and remote sensing. Urban For. Urban Green. 12, 88–97.
Onishi, A., Cao, X., Ito, T., Shi, F., Imura, H., 2010. Evaluating the potential for urban heat-island mitigation by greening parking lots. Urban For. Urban Green. 9, 323–332. Pataki, D., Carreiro, M., Cherrier, J., Grulke, N., Jennings, V., Pincetl, S., Pouyat, R., Whitlow, T., Zipperer, W., 2011. Coupling biogeochemical cycles in urban environments: ecosystem services, green solutions, and misconceptions. Front. Ecol. Environ. 9 (1), 27–36. Perkins, H., Heynen, N., Wilson, J., 2004. Inequitable access to urban reforestation: the impact of urban political economy on housing tenure and urban forests. Cities 21 (4), 291–299. Pham, T., Apparicio, P., Séguin, A., Landry, S., Gagnon, M., 2012. Spatial distribution of vegetation in Montreal: an uneven distribution or environmental inequity? Landsc. Urban Plann. 107, 214–224. Pham, T.T.H., Apparicio, P., Landry, S., Séguin, A.M., Gagnon, M., 2013. Predictors of the distribution of street and backyard vegetation in Montreal, Canada. Urban For. Urban Green. 12 (1), 18–27. Scopelliti, M., Carrus, G., Adinolfi, C., Suarez, G., Colangelo, G., Lafortezza, R., Panno, A., Sanesi, G., 2016. Staying in touch with nature and well-being in different income groups: the experience of urban parks in Bogotá. Landsc. Urban Plann. 148, 139–148. Shan, X., 2014. The socio-demographic and spatial dynamics of green space use in Guangzhou, China. Appl. Geogr. 51, 26–34. Stantec, 2010. Proposed Walkability Strategy. Available at https://www.edmonton.ca/ transportation/PDF/WalkabilityStrategy200909.pdf. Statistics Canada, 2011. National Household Survey (NHS). Available at https://www12. statcan.gc.ca/nhs-enm/index-eng.cfm. Sugiyama, T., Leslie, E., Giles-Corti, B., Owen, N., 2008. Associations of neighborhood
67