Urban Ecosystems https://doi.org/10.1007/s11252-018-0764-8
People or place? An exploration of social and ecological drivers of urban forest species composition James W. N. Steenberg 1
# Springer Science+Business Media, LLC, part of Springer Nature 2018
Abstract Urban forests have garnered increasing attention as providers of an array of beneficial ecosystem services. However, urban forest ecosystems are highly complex and heterogeneous systems whose structure are shaped by interacting social and ecological processes. Approaches to reliably identify and differentiate these processes could be valuable for addressing complexity and reducing uncertainty in decision-making in urban forestry. The purpose of this study is to identify and quantify a range of social and ecological drivers of urban forest species composition, distribution, and diversity. This was done using hierarchical cluster analysis and discriminant analysis with empirical plot data describing the tree species composition in Toronto, Canada. Tree density and imperviousness were by far the most influential drivers of species composition. Increasing imperviousness saw not just reduced tree density but a decline in native species abundance. Additionally, single-detached housing, homeownership, and income were closely associated and explained higher tree densities and abundances of native species. However, income had a lower than expected influence on urban forest species composition given its importance in canopy cover research. Continuous forest patches were highly distinct compared to the remainder of the urban landscape, which highlights the ecological distinctiveness of residual forests within cities and lends support to their conservation. Increasing the understanding of social and ecological drivers of tree species composition, distribution, and diversity within cities is an integral part of urban forest ecosystem classification, which can be a valuable decision-support tool for ecosystem-based management in urban forestry. Keywords Urban forest; ecosystem classification . Species composition . Diversity . Social-ecological system
Introduction Urban forests have garnered increasing attention as providers of an array of beneficial ecosystem services (Nowak and Dwyer 2007; Duinker et al. 2015). Environmental organizations and municipal governments are also realizing the value of assessing and managing the urban forest resource through the development of comprehensive urban forest management programs (Ordóñez and Duinker 2013). However, both researchers and practitioners alike acknowledge the high degree of complexity of urban forest ecosystems compared to their hinterland counterparts, as well as the management challenges associated with this complexity. Urban forests are multi-scaled and highly heterogeneous systems whose structure are shaped by interacting social and ecological * James W. N. Steenberg
[email protected] 1
School for Resource and Environmental Studies, Dalhousie University, 6100 University Avenue, Halifax, NS B3H 4R2, Canada
processes (Avolio et al. 2015; Steenberg et al. 2015). Approaches to reliably identify and differentiate these processes are valuable for addressing complexity and reducing uncertainty in decisionmaking and management in urban forestry. There has been increasing research attention to socioeconomic, built, and biophysical drivers urban forest structure and amenities (e.g., Heynen and Lindsey 2003; Grove et al. 2006; Troy et al. 2007). It is well established that affluence is strongly associated with the extent of canopy cover and the supply of urban forest ecosystem services (Schwarz et al. 2015). However, much of the existing research in this area is restricted to satellite-derived canopy cover data where urban forest structure is reduced to the presence or absence of canopy. There remains a high level of uncertainty around both social and ecological drivers of other metrics of urban forest structure, like tree species composition, distribution, and diversity (Bourne and Conway 2014; Conway and Bourne 2013; Nitoslawski et al. 2016). Moreover, many of the studies that reinforce the strong association between affluence – especially income – and canopy cover are derived from larger American cities with strong income inequalities. This association
Urban Ecosyst
is shown to be weaker or more nuanced in other regions. For instance, Kendal et al. (2012) found that education level was a better predictor of canopy than income in Ballarat, Australia, while Conway and Bourne (2013) found that housing age and ethnocultural background where the better predictors of canopy, stem density, and species diversity. The degree to which drivers of species composition are understood and documented varies across the urban landscape. Species composition in forested parks and undeveloped residual forests is perhaps better understood than in more developed areas, as it is largely driven by well-documented biophysical conditions, such as climate, soils, topography, and successional processes (Burns and Honkala 1990; Hargrove and Hoffman 2005; McKenney et al. 2007). However, the predictability of these processes is frequently interrupted to varying degrees by introduced naturalized and invasive plant species that are typical of the urban environment and outcompete native species (Call and Nilsen 2003; Ordóñez and Duinker 2012). In most developed parts of cities, urban forest structure is less predictable and characterized by novel species assemblages and ecosystems with a mix of native species, cultivated varieties and crosses, and introduced non-native species (McPherson et al. 1997; Ordóñez and Duinker 2012). These urban tree species assemblages consist of both trees that are planted by a variety of stakeholders and trees that are ingrown (i.e., germinated from seed) and naturalized (Nitoslawski et al. 2016; Nitoslawski et al. 2017). Arguably, the decision-making process around tree species selection for municipal tree plantings is a more understood and tractable one because the process is governed entirely by municipal urban forestry departments. Yet, in most cities municipal tree planting is limited to street trees in the public right of way and public parks. Despite the high level of attention to the importance of street trees in the literature and in municipal urban forestry, they represent a small proportion of the urban forest. Tree species composition, distribution, and diversity in the majority of the urban forest in a given city is dictated by a far less predictable mix of resident decisions and ecological processes. Land use is a commonly used framework for understanding and characterizing the structure and function of urban forests across entire cities (Nowak et al. 1996; Steenberg et al. 2013). However, land-use categories generalize or omit key social and ecological gradients that are vital for understanding urban forest structure, especially in residential landscapes that comprise the majority of most North American urban forests (Kenney and Idziak 2000). Tree planting and tree-species preferences in residential land uses are influenced by income, education, ethnocultural background, and a host of other social processes (Fraser and Kenney 2000; Greene et al. 2011). Conway (2016) found that tree planting decisions were highly influenced by species-specific maintenance requirements and aesthetics. Avolio et al. (2015) found that local environmental factors influenced tree species preferences (e.g., hot climates
and shade trees), while residential landscaping practices in general can be influenced by the practices of neighbours in what is often termed the neighbourhood effect (Zmyslony and Gagnon 2000; Grove et al. 2006). Understanding and differentiating between socialecological drivers of urban forest species composition is more than an interesting academic research endeavour, this understanding also contributes valuable knowledge for management purposes. Characterizations of tree species composition and the distribution of different species is valuable for assessing spatial variability in urban forest vulnerability to species- and genusspecific insects and pathogens (Laćan and McBride 2008; Greene and Millward 2016; Steenberg et al. 2017a). Moreover, tree species vulnerability to climate change due to rising temperatures and the attendant shifts in species ranges will be highly variable across species (McKenney et al. 2007; Woodall et al. 2010). The propensity of invasive tree species to colonize some areas over others, especially residual forests and parks with an abundance of native species, is useful knowledge for biodiversity conservation and targeted species removals (Toni and Duinker 2015). Enhancing the predictability of tree species composition and distribution in the urban landscape by identifying broad social and ecological patterns that can be easily measured can help practitioners address these latter management endeavours (Steenberg et al. 2015). Such information is additionally valuable because it reduces the need for city tree inventories to guide decision-making, as such datasets are expensive to collect, difficult to measure (i.e., access to private property), and quickly become dated (Roman et al. 2013). The purpose of this study is to identify and quantify a range of social and ecological drivers of urban forest species composition, distribution, and diversity. The research objectives are to: 1) analyze social and ecological correlates of tree species distribution; 2) identify meaningful tree species assemblages in the urban forest; and 3) test the relative capacity of key social and ecological variables for differentiating between the species assemblages. To meet these research objectives, hierarchical cluster analysis and discriminant analysis are applied to empirical plot data describing the tree species composition in Toronto, Canada. The overarching goal is to understand and communicate the relative roles of various socialecological variables in driving urban forest structure, but also to contribute to the on-going research of urban forest ecosystem classification (Steenberg et al. 2015).
Methods Study area This research was conducted in Toronto, Canada (Fig. 1). Toronto is situated in the Deciduous Forest Region and Mixedwood Plains Ecoregion, which have the highest tree
Urban Ecosyst
Fig. 1 The City of Toronto, Canada, showing the location of the 407 plots used in the study and the extent of urban forest canopy cover and wooded areas
species diversity but are the most deforested and developed of their respective classifications in Canada (Ontario Ministry of Natural Resources 2012). Residual and undeveloped forests are characterized by sugar maple, white ash, red oak, white pine, and eastern hemlock (see Table 1 for scientific names). The City of Toronto recently conducted an assessment of its urban forest (City of Toronto 2010). Typical urban species included Norway maple, sugar maple, eastern white cedar, Manitoba maple, and green ash. The city also has an extensive system of ravines and river valleys, which are typically forested. However, the city is at the centre of a large regional urban area (i.e., Greater Toronto Area) and there is very little residual undeveloped forest within the municipal boundaries. Toronto’s climate is continental, with cold winters and humid, hot summers. The mean annual precipitation is 834 mm and the mean annual temperature is 9.2o C, with mean January and July temperatures of −4.2o C and of 22.2o C, respectively (Environment Canada 2015). Toronto has a population of 2,615,060 and population density of 4150 people per km2 (Statistics Canada 2016).
Data and processing Urban tree species composition was derived from 407 plots 0.04 ha in size that were randomly sampled from within the
municipal boundaries of Toronto. Field data were collected by the City of Toronto in 2008 following the i-Tree Eco measurement protocols as part of a broader assessment of the city’s urban forest (City of Toronto 2010). A total of 2674 trees and 119 species were measured within the 407 sample plots. In the subsequent analysis, only species that comprised greater than 1% of the entire sample were included (McGarigal et al. 2000), totaling 24 species (Table 1). Rare species may act as indicator species for a given site type or environmental condition in more naturalized forests (Cao et al. 2001; Matthews et al. 2011). However, this omission was seen as a necessary measure given the much higher tree species richness typical of cities and likelihood of highly anomalous introduced species (McKinney 2008). While the relative abundance of morecommon species is known to follow dominant environmental gradients in naturalized forests, there remains uncertainty around the possible influence of rare species (Cao et al. 2001). This uncertainty is an important consideration in examining social-ecological systems like the urban forest, where these analytical tools have not been extensively tested. Several ecosystem component variables were used to describe social and ecological processes known to influence urban forest structure and function to varying degrees (Table 2). Land use is both a well-documented driver of urban forest structure and a frequently used approach to landscape
Urban Ecosyst Table 1 Common names, abbreviations, scientific names, and abundance of tree species that comprise greater than 1% of all measured trees
Common name
Abbreviation
Scientific name
Count (percent)
American basswooda,b
AB AP BC CC CA
Tilia americana Pinus nigra Prunus serotina Prunus virginiana Malus spp.
36 (1.3) 36 (1.3) 61 (2.3) 52 (1.9) 60 (2.2)
Green asha,b Honeylocustb Ironwooda Lawson cypressb Manitoba maplec,d Norway mapleb,c Norway spruceb Red pinea Siberian elmc Sugar maplea,b Trembling aspena White asha,b White bircha White cedara,b White elma,b White oaka White pinea
EB GA HL IW LC MM NM NS RP SE SM TA WA WB WC WE WO WP
Rhamnus cathartica Fraxinus pennsylvanica Gleditsia triacanthos Ostrya virginiana Chamaecyparis lawsoniana Acer negundo Acer platanoides Picea abies Pinus resinosa Ulmus pumila Acer saccharum Populus tremuloides Fraxinus americana Betula papyrifera Thuja occidentalis Ulmus americana Quercus alba Pinus strobus
42 (1.6) 95 (3.6) 40 (1.5) 82 (3.1) 41 (1.5) 135 (5.0) 177 (6.6) 32 (1.2) 27 (1.0) 75 (2.8) 263 (9.8) 60 (2.2) 142 (5.3) 38 (1.4) 430 (16.1) 40 (1.5) 26 (1.0) 35 (1.3)
White sprucea,b
WS
Picea glauca
91 (3.4)
Austrian pineb Black cherrya Chokecherrya,b Crabappleb European buckthornc
a
Species that is native to the region (Burns and Honkala 1990; Farrar 1995; City of Toronto 2010)
b
Species that is frequently planted in cities (Burns and Honkala 1990; Farrar 1995; City of Toronto 2010)
c
Non-native naturalized/invasive species (Burns and Honkala 1990; Farrar 1995; City of Toronto 2010)
d There
is debate as to whether Manitoba maple is native to the Toronto region. It is highly tolerant of urban conditions, grows readily from seed, and frequently colonizes disturbed sites (Farrar 1995; Foster and Sandberg 2004)
classification in urban forest planning and management (Nowak et al. 2004; Steenberg et al. 2015). Six coarse landuse categories were included in the study: commercial/industrial, open green space, residual forest, institutional, residential, and utilities and transportation. These categories were mutually exclusive and assigned to individual plots in the field. Several variables describe the intensity of the built environment and are all highly influential on tree and forest establishment and growth (Trowbridge and Bassuk 2004; Jutras et al. 2010; Lu et al. 2010; Steenberg et al. 2017a). These variables include impervious surface cover, population density, and building intensity. Additionally, housing type and age influence the structure of urban forest ecosystems. Older, single-detached housing is commonly associated with higher canopy cover and a greater abundance of mature trees while newer housing is often associated with small trees or an absence of trees (Troy et al. 2007; Nitoslawski et al. 2017). Both older housing and newer housing variables were therefore included.
Several socioeconomic variables were included to represent social drivers and correlates of urban forest structure and canopy cover distribution. Urban forest canopy cover is consistently higher in more affluent urban areas, as has been shown by research in many North American cities (e.g., Schwarz et al. 2015). Homeowners have also been found to be more likely to plant and maintain trees on their properties compared to renters (Greene et al. 2011; Ko et al. 2015; Steenberg et al. 2017b). Lastly, biophysical variables were included to represent likely ecological drivers of urban tree species composition. Slope and topography are two frequently used variables in ecological classification, as they influence species composition through drainage, exposure, microclimate, and light availability (Klijn and Udo de Haes 1994; Bailey 2009). Arguably, the greatest sources of heterogeneity in overall urban forest structure and composition are the extent, pattern, and type of residual forests within a city. To capture this effect using a continuous variable (i.e., in addition to land use), tree density (stems/ha) was included in the
Urban Ecosyst Table 2
Ecosystem component variable descriptions and descriptive statistics
Variable
Description
Mean Standard deviation (Count*) (Percent*)
Impervious PopDensity Owned Before1946
Impervious surface cover (%)a Population density (people/km2)b Percent of owner-occupied private dwellings (%)b Percent of dwellings constructed before 1946 (%)b
51.3 3867 69.1 13.1
35.4 5123 29.8 23.1
2001to2006
Percent of dwellings constructed between 2001 and 2006 (%)b
9.9 47.4 72,718 2.5 0.1
22.7 35.3 38,426 3.7 6.4
19.0
9.5
164.3 77* 208* 32* 35* 27* 29*
330.1 18.9* 51.1* 7.9* 8.6* 6.6* 7.1*
SingleDetached Income MeanSlope TPI
Percent of dwellings that are single-detached (%)b Median family income ($ CAD)b Mean slope gradient (%)c Topographic position index; negative values indicate valleys/depressions and positive values indicate ridges/hilltopsc BuildIntensity Intensity of the built area (%), estimated as the mean ratio of building footprint (m2) to parcel area (m2)d Density Tree density (stems/ha)e Commercial Land use dummy variable; commercial and industrial land usese Residential Land use dummy variable; residential land usese OpenGreenSpace Land use dummy variable; parks, protected areas, and green space without forest cover land usese ResidualForest Land use dummy variable; parks, protected areas, and green space with forest cover land usese Transport Land use dummy variable; transportation and utility land usese Institutional Land use dummy variable; institutional land usese a
Land cover data derived from 2007 QuickBird satellite imagery (0.6-m resolution)
b
Statistics Canada 2006 census data
c
Ontario Ministry of Natural Resources digital elevation model (DEM; 10-m resolution)
d
City of Toronto cadastral data
e
City of Toronto 2008 i-Tree Eco plots
analyses. Soils data would have been valuable to include in the study, but soil surveys of sufficient detail have not been completed for Toronto.
Analysis There were three stages of analysis. The first stage was to examine relationships between individual tree species and social-ecological drivers of change, which was accomplished through a correlation analysis. The correlation analysis applied Spearman’s Rho to evaluate the relationship between the ecosystem component variables (excluding land use) and the abundances and basal areas of the top five species. The relationship between land use and individual tree species was considered separately through the application of canonical correlation analysis, which is described below. The second stage of analysis applied cluster analysis to identify prominent urban forest species assemblages reflecting the urban environment, as well as to perform dimension reduction to further examine drivers of species composition across the city. Cluster analysis refers to a collection of multivariate techniques used to explore the relationships among ecological variables by grouping, or clustering, similar data points into classes (Jongman et al. 1995; McGarigal et al. 2000). This
study used a hierarchical cluster analysis with Ward’s method. Starting with n clusters, Ward’s method hierarchically groups data points into increasingly larger classes that maximize within-class homogeneity and between-class heterogeneity in multivariate data space based on squared Euclidean distance. Hierarchical clustering techniques are used more frequently than non-hierarchical ones (e.g., k-means), as these techniques are better suited for examining ecological relationships within and between resulting classes (McGarigal et al. 2000). Ward’s method in particular is used for ecosystem classification research in natural landscapes (McNab et al. 1999; Mora and Iverson 2002; Hargrove and Hoffman 2005; Maxwell et al. 2014) and more recently in urban landscapes (Steenberg et al. 2015). The hierarchical nature of the cluster analysis approach enabled the classification of both prominent tree species assemblages and the sub-classes within these assemblages. The cluster analysis results were therefore assessed at two levels, which were identified at natural breaks in the dendrogram (Zhou et al. 2003; Steenberg et al. 2015). Variables (i.e., tree species abundance) were standardized for the analysis using their maximum value so that all values ranged between 0 and 1 (Jongman et al. 1995). A total of 131 (32.2%) plots had no trees located within plot boundaries. These plots were
Urban Ecosyst
consequently omitted from the cluster analysis, but were grouped together as an additional class to be used in the subsequent analysis. Urban landscapes and urban forest ecosystems are highly heterogeneous and heavily built-up areas without tree cover are prominent in cities. From a planning and management perspective, an understanding of the drivers of areas without trees – not just tree species composition – is valuable. A second exploratory cluster analysis was conducted using tree species basal area in place of abundance, though only main species assemblages and not sub-classes were identified. The third and final stage of analysis applied discriminant analysis to examine the relative influence of the 12 ecosystem component variables in explaining the urban tree species assemblages. Discriminant analysis, also called discriminant function analysis or canonical analysis of discriminance, is a multivariate method that uses several independent variables and a categorical dependent variable with two or more groups. The test examines multiple weighted linear combinations of the independent variables, called discriminant functions, that are intended to maximize the ratio of among-group to withingroup variance and best predict group membership for individual observations. Like cluster analysis, discriminant analysis is frequently used in ecological research for classification and for identifying key environmental gradients that explain ecological variability (McNab et al. 1999; Hamann and Wang 2006; Bailey 2009; Womack and Carter 2011). There are two primary uses of discriminant analysis. When group membership (e.g., ecosystem type) is not known a priori, discriminant analysis is used for classification to predict membership using a set of independent variables known to be important in defining groups. When group membership is known a priori, discriminant analysis is used to identify which independent variables are most effective for discriminating between groups. This study employs the latter approach, as the overarching research objective is to identify the social-ecological divers that best explain variability in urban forest species composition, and because cluster analysis was used a priori to define the dependent grouping variable (i.e., species assemblages). Cluster analysis is often used prior to discriminant analysis where known ecosystem types and species assemblages do not exist or are not well established in the literature (McGarigal et al. 2000; Matthews et al. 2011), as is the case with urban forests. Lastly, a canonical correlation analysis was conducted to examine the relationship between individual tree species abundance and land use, given the prominence of land use as an approach to landscape classification in urban forest research. Canonical correlation analysis is a multivariate analysis technique used to describe the relationship between two or more datasets, typically species abundance data and ecosystem component data (e.g., soils, climate) in ecological research. It creates weighted linear combinations of variables
within each respective dataset so that the correlation between combinations across datasets is maximized (McGarigal et al. 2000). Canonical correlation analysis was used here in place of discriminant analysis because of the higher number of zeros (i.e., tree species absent in a given plot) in the individual species abundance data that can negatively influence the validity of the discriminant analysis results.
Results The correlation analysis demonstrated that some ecosystem components are strong and consistent correlates of tree species distribution, while other components tended to be more variable (Tables 3 and 4). Tree density exhibited a strong positive correlation with each of the five most abundant species in the study area for both species abundance and basal area. Conversely, all species, with the exception of white cedar, had a significant negative correlation with impervious surface cover. There was more variability in the relationship between species and the other ecosystem component variables. White cedar had a relationship with housing, and was positively correlated with homeownership and the proportion of singledetached dwellings. In contrast, sugar maple appeared to be more associated with urban density and topography, being negatively associated with population density and built area intensity and positively associated with ravine terrain (i.e., steep slopes and valley topography). Manitoba maple was similarly associated with ravine terrain, but also had a negative correlation with newer housing. Norway maple was positively correlated with topographic position index, indicating its propensity to grow on hills and ridges. White ash exhibited similarities to sugar maple, having a negative association with urban density and a positive one with steeper terrain. It was expected that median family income would have more and stronger correlations given its near universal association with canopy cover, but was only found to have a significant positive correlation with sugar maple and Norway maple. Lastly, while some of these aforementioned relationships were statistically significant, the correlation was sometimes low (e.g., 0.10) and the validity of the social-ecological relationship should be tested with further empirical research. The hierarchical cluster analysis was used with species abundance to classify species assemblages and sub-classes within at two levels based on natural breaks in the dendrogram (Zhou et al. 2003; Steenberg et al. 2015). Six larger species assemblages were first identified (Fig. 2), which includes the plots with no trees for subsequent use in the discriminant analyses (Table 5). A total of 22 clusters were then identified in the dendrogram, yielding 23 sub-classes including the plots with no trees. These sub-classes were summarized for descriptive purposes and to get a sense of the within-class variability of the six species assemblages (Table 6). However, only the
Urban Ecosyst Table 3 Correlation analysis of ecosystem component variables with the abundance of the five most abundant tree species in the study area using Spearman’s Rho with p-values in brackets. Significant correlations at α = 0.05 are in bold
Variable
White Cedar
Sugar maple
Norway maple
White ash
Manitoba maple
Impervious PopDensity
−0.09 (0.08) 0.06 (0.23)
−0.31 (0.00) −0.14 (0.01)
−0.15 (0.00) 0.09 (0.06)
−0.29 (0.00) −0.12 (0.01)
−0.18 (0.00) 0.06 (0.20)
Owned Before1946
0.15 (0.00) 0.06 (0.22)
0.04 (0.40) −0.06 (0.21)
0.07 (0.13) 0.04 (0.45)
0.06 (0.27) −0.05 (0.37)
0.02 (0.64) 0.09 (0.09)
2001to2006
−0.05 (0.36)
0.02 (0.74)
−0.09 (0.07)
0.06 (0.24)
−0.10 (0.04)
SingleDetached
0.21 (0.00)
0.10 (0.05)
0.09 (0.07)
0.08 (0.11)
0.04 (0.49)
Income MeanSlope
0.09 (0.06) 0.04 (0.44)
0.12 (0.02) 0.22 (0.00)
0.13 (0.01) 0.04 (0.49)
0.00 (0.99) 0.22 (0.00)
0.07 (0.19) 0.18 (0.00)
TPI BuiltIntensity
0.02 (0.63) 0.01 (0.80)
−0.10 (0.05) −0.22 (0.00)
0.12 (0.01) 0.03 (0.60)
−0.09 (0.08) −0.23 (0.00)
−0.11 (0.02) −0.00 (0.94)
Density
0.41 (0.00)
0.36 (0.00)
0.37 (0.00)
0.33 (0.00)
0.38 (0.00)
six larger classes were used in the subsequent analysis, both to meet discriminant analysis assumptions and to provide clear and parsimonious results. Of the six classes, Class 1 (native broadleaves) consisted primarily of native broadleaves and a higher density and species richness, with a prominence of sugar maple, ironwood, and white oak. Class 2 (white cedar) was dominated by white cedar, with an abundance of Siberian elm in sub-class 3. Class 3 (mixed broadleaves) had more internal variability in species dominance and density, with the most consistent feature being white ash. Class 3 also had some sub-classes dominated by comparatively rare species, such as trembling aspen and white birch. Class 4 (Norway maple-conifers) had lower density and species richness. While Norway maple was more abundant overall, Austrian pine was consistently found in all sub-classes. Class 5 (mixed low density) had by far the most plots and sub-classes with a considerable amount of internal heterogeneity, being dominated by white cedar and sugar maple, with white spruce. The most consistent theme was low density and species richness, and a number of sub-classes were typically dominated by a single, less-common species, such as white spruce (sub-class 15), crabapple species (sub-class 16), and
Table 4 Correlation analysis of ecosystem component variables with the basal area of the five most abundant tree species in the study area using Spearman’s Rho with p-values in brackets. Significant correlations at α = 0.05 are in bold
honeylocust (sub-class 17). Class 6 (non-forested) was the second largest and was comprised of the plots with no trees. The hierarchical cluster analysis using species basal area identified eight species assemblages (Fig. 3; Table 7). There were similarities with the species assemblages resulting from the cluster analysis using species abundance. Both had a mixed-broadleaves class dominated by sugar maple and including American basswood and ironwood. Both had a very extensive (i.e., percent area) species assemblage with low density/basal area and high imperviousness. However, the basal area cluster analysis also produced some novel assemblages, such as Class 2 (street trees) with smaller trees, lower density, and species typically planted in streets and Class 1 (mixed conifers) that had a high density and basal area of both native and non-native conifers. Discriminant analysis revealed several insights into the social and ecological drivers of urban forest species composition (Table 8; Fig. 4). The first three orthogonal discriminant functions were statistically significant (α = 0.05) and cumulatively explained 94.3% of the variance in the ecosystem component variables. Plotting the discriminant functions illustrates the data grouping for each class of species assemblage in data
Variable
White Cedar
Sugar maple
Norway maple
White ash
Manitoba maple
Impervious PopDensity Owned Before1946 2001to2006 SingleDetached Income MeanSlope TPI BuiltIntensity Density
−0.09 (0.07) 0.06 (0.24) 0.14 (0.01) 0.06 (0.22) −0.05 (0.36) 0.21 (0.00) 0.09 (0.07) 0.03 (0.52) 0.02 (0.71) 0.01 (0.88) 0.41 (0.00)
−0.31 (0.00) −0.13 (0.01) 0.04 (0.41) −0.06 (0.21) 0.02 (0.73) 0.1 (0.05) 0.12 (0.02) 0.22 (0.00) −0.09 (0.06) −0.22 (0.00) 0.36 (0.00)
−0.13 (0.01) 0.11 (0.03) 0.07 (0.15) 0.05 (0.32) −0.09 (0.07) 0.09 (0.08) 0.14 (0.01) 0.02 (0.66) 0.14 (0.01) 0.04 (0.47) 0.35 (0.00)
−0.29 (0.00) −0.12 (0.02) 0.05 (0.28) −0.05 (0.37) 0.06 (0.27) 0.08 (0.11) 0.00 (0.99) 0.21 (0.00) −0.08 (0.10) −0.23 (0.00) 0.33 (0.00)
−0.17 (0.00) 0.07 (0.15) 0.03 (0.52) 0.09 (0.07) −0.10 (0.04) 0.04 (0.45) 0.08 (0.11) 0.18 (0.00) −0.11 (0.03) 0.01 (0.91) 0.37 (0.00)
Urban Ecosyst
Fig. 2 Dendrogram showing the results of the hierarchical cluster analysis of 407 plots and 24 tree species to classify urban forest species assemblages using species abundance. Numbers on the left indicate the sub-classes and dashed boxes indicate the main classes
space (Fig. 4). Classes 5 (mixed low density), 6 (non-forested), and 4 (Norway maple-conifers) show more internal homogeneity, though there is overlap between classes 4 and 5. Also, clear gradients of density and imperviousness are evident. Classes 1 (native broadleaves), 2 (white cedar), and 3 (mixed broadleaves) span a considerable range of mid-tohigher densities, with Class 3 and its mix of native and nonnative broadleaves tended towards higher imperviousness. Discriminant function 1 (eigenvalue = 1.20, variance explained = 72.3%) showed a clear gradient from highly impervious areas with few trees to high tree densities and abundant open green spaces (Table 8). Discriminant function 2
Table 5
(eigenvalue = 0.25, variance explained = 15.1%) illustrated a gradient from older housing and higher population densities to newer housing with higher tree densities and open green spaces, though as shown it explained less variance than function 1. The least influential discriminant function 3 (eigenvalue = 0.12, variance explained = 3.9%) appeared to show a gradient from older, owned housing towards higher income and a higher topographic position index (i.e., ridge and hilltop terrain). The classification accuracy of the discriminant analysis (Fig. 5) can be used as metric of model performance, though recall the purpose of this study was not predictive classification but rather an examination of the social-ecological discriminating variables. Relating these results back to the earlier analysis of individual tree species with the ecosystem component variables, canonical correlation analysis of individual tree species with land use provided some additional insight (Table 9). Only the first two canonical correlations were significant at α = 0.05. Canonical correlation 1 (eigenvalue = 0.78, correlation = 0.66) showed a strong negative correlation of all but residual forest land uses with some tree species that are typically associated with wooded areas and not maintained and planted areas (e.g., trembling aspen). However, there was no apparent trend around native species versus introduced species or broadleaves versus conifers. Canonical correlation 2 (eigenvalue = 0.33, correlation = 0.50) again provided little insight except for the apparent correlation between commercial and industrial land uses with the absence of trees (i.e., no trees = 1). The cumulative variance explained by both canonical correlations was only 12.7%.
Discussion The study of social and ecological drivers of ecosystem conditions – especially urban ecosystems – has been an evolving and increasingly important research focus for several decades (Stearns and Montag 1974). Brady et al. (1979) focused on integrating social and urban processes into ecology through the development of a hierarchical urban ecosystem
The six classes identified as urban forest species assemblages using species abundance
Class
Area (%)
Mean Species Richness
Density (stems/ha)
Imper-vious (%)
Species 1 (%)
Species 2 (%)
Species 3 (%)
1 – Native broadleaves 2 – White cedar 3 – Mixed broadleaves 4 – Norway maple-conifers 5 – Mixed low density 6 – Non-forested
1.2 3.0 4.9 5.9 52.8 32.2
8.4 5.0 6.1 3.7 2.5 0.0
1210 779 771 288 135 0
11.2 29.7 13.6 33.9 49.5 66.9
SM (40.1) WC (73.0) WA (21.0) NM (33.0) WC (17.2) None
IW (24.3) SE (16.4) MM (14.6) LC (18.2) SM (15.0) None
AB (9.0) NM (2.6) GA (9.4) AP (15.8) WS (10.0) None
Urban Ecosyst Table 6
The 23 sub-classes of urban forest species assemblages using species abundance
Sub-class
Area (%)
Mean Species Richness
Density (stems ha−1)
Impervious (%)
Species 1 (%)
Species 2 (%)
Species 3 (%)
1–1
0.7
7.3
1425
17.7
SM (51.3)
IW (25.6)
AB (6.3)
1–2 2–1
0.5 1.0
10.0 4.0
888 725
1.5 22.0
IW (21.0) SE (44.7)
WO (19.4) WC (42.1)
AB (16.1) NM (6.1)
2–2
2.0
5.5
806
33.5
WC (88.0)
CC (3.4)
SE (2.6)
3–1
0.2
7.0
2525
0.0
TA (49.5)
WA (23.2)
BC (19.8)
3–2 3–3
0.7 1.2
6.3 5.8
1125 440
0.0 27.6
WA (51.0) WA (25.3)
SM (24.0) WB (21.3)
WC (9.6) SM (14.7)
3–4 3–5
1.5 0.5
5.0 9.0
450 763
17.2 15.0
MM (63.3) CC (42.9)
CC (7.1) WA (31.4)
WE (6.1) WE (5.7)
3–6 4–1
0.7 0.5
6.0 2.5
1033 275
0.0 50.0
GA (38.9) HL (68.8)
WP (20.4) AP (31.1)
EB (14.8) BS (13.6)
4–2
3.7
2.9
118
35.5
AP (53.2)
WC (14.9)
NM (12.8)
4–3
1.7
5.7
654
26.0
NM (43.2)
LC (26.0)
MM (6.2)
5–1 5–2 5–3 5–4
0.2 0.5 0.2 5.2
3.0 3.0 3.0 2.7
1375 475 550 112
0.0 46.5 0.0 61.3
NS (49.1) WS (94.4) CA (86.4) HL (40.8)
RP (47.3) NM (2.8) WS (9.1) SM (15.5)
WS (3.6) WB (2.8) RP (4.5) WC (8.5)
5–5 5–6 5–7
1.2 3.7 2.2
8.6 3.5 3.2
850 230 147
8.2 36.5 39.1
SM (32.7) SM (53.5) CC (40.0)
IW (22.1) WC (13.3) GA (22.9)
WA (14.2) WB (10.8) CA (11.4)
5–8 5–9
3.7 35.9
2.9 2.1
122 89
38.9 52.9
NM (68.6) WC (34.3)
MM (9.8) NM (10.2)
WC (3.9) WS (9.8)
6–1
32.2
0.0
0
66.9
None
None
None
classification system. Dorney et al. (1984) described the effects of housing and land use on woody plant communities by making comparisons to non-urban ecosystems of similar structure. Zipperer et al. (1997) aimed to expand ecological research in cities beyond publicly-owned land and investigated patch dynamics across the entire urban landscape. An early research theme was often the detrimental effect of urbanization (e.g., built environments, intensive anthropogenic land uses, and environmental pollution) on ecosystem structure and function. However, given global demographic trends towards a largely urban population, there was a shift in ecological thinking towards research that integrated urbanization and social processes (i.e., humans as ecosystem components) to gain a better understanding of urban ecosystem structure and function (McDonnell and Pickett 1990). These changes have been framed as a shift in focus on the ecology in the city towards the ecology of the city (Grimm et al. 2000). This was coupled with a growing focus on the beneficial ecosystem services provided by urban ecosystems and the possibility for humans to influence their supply through management (Nowak and Dwyer 2007). This study aims to contribute to the this rapidly growing body of knowledge in the context of sustainable urban forest ecosystem management.
The findings of this study highlight several important social-ecological drivers of urban tree species composition, distribution, and diversity. First, tree density and imperviousness were by far the most influential drivers, which is a somewhat intuitive result as built-up areas restrict tree establishment while higher-density forest stands are favourable growing conditions for many of the tree species analyzed. However, decoupling these two dominant drivers and examining the other social-ecological relationships beneath reveals some important trends. One of these underlying trends was the distinctness of continuous forest patches compared to the remainder of the urban landscape. This result was exemplified by the composition of the Class 1 (native broadleaves), being exclusively situated in continuously forested areas. Furthermore, Class 1 exhibited the highest tree density and species richness, lowest imperviousness, and was characterized by mid-to-late successional native tree species typical of the ecozone (i.e., sugar maple, ironwood, basswood, white oak; Farrar 1995; Ontario Ministry of Natural Resources 2012). These findings reinforce the ecological distinctiveness of residual forests within cities and lends support to their conservation, especially given the smaller area, as native tree species and forest patches are important contributors to overall
Urban Ecosyst Table 8 Discriminant function loadings for the ecosystem component variables
Fig. 3 Dendrogram showing the results of the hierarchical cluster analysis of 407 plots and 24 tree species to classify urban forest species assemblages using basal area
urban biodiversity (McKinney 2008; Nitoslawski et al. 2016). These findings also highlight the need to move beyond canopy cover alone when characterizing urban forests and establishing management priorities (Kenney et al. 2011; Conway and Bourne 2013). The income and housing variables yielded several notable relationships with individual tree species and species assemblages. Single-detached housing, homeownership, and income were closely associated and explained higher tree densities and abundances of native species. These variables all tend to be associated with affluence and highlight the positive association that is also found between canopy cover and
Table 7
Variable
Function 1
Function 2
Function 3
Impervious
0.33
0.47
0.09
PopDensity
−0.10
−0.20
0.10
Owned Before1946
−0.01 0.04
−0.02 −0.15
−0.46 −0.37
2001to2006
0.13
0.25
−0.20
SingleDetached Income
−0.17 −0.05
0.02 −0.16
0.15 0.58
MeanSlope TPI
−0.03 0.24
0.14 −0.14
0.11 0.21
BuiltIntensity
0.08
−0.04
−0.07
Density
−0.88
0.41
0.24
Commercial OpenGreenSpace Transport
0.13 0.23 0.27
0.49 0.33 0.52
0.28 0.24 0.09
Institutional
0.20
0.37
−0.08
ResidualForest
0.36
0.14
−0.67
affluence (Grove et al. 2014). However, what is also interesting about the socioeconomic variables was the lower than expected influence of income urban forest species composition. It is well established in the literature focusing on socioeconomic drivers of canopy cover distribution that the positive association between income and canopy cover is among the most generalizable relationships across cities (Schwarz et al. 2015). This study suggests that when understanding species composition, distribution, and diversity – both within residual forests and across the entire urban landscape – housing attributes and especially biophysical conditions (e.g., slope and topography) appeared more influential than income. This strong relationship between housing and urban forest structure is also supported by other Canadian studies (e.g., Conway and Bourne 2013; Pham et al. 2013). From a management perspective, characterizing the distribution of urban forest
The eight classes identified as urban forest species assemblages using basal area
Class
Area (%)
Basal Area (m2/ha)
Density (stems/ha)
Imper-vious (%)
Species 1 (BA)
Species 2 (BA)
Species 3 (BA)
1 – Mixed conifers 2 – Street trees 3 – Native broadleaves 4 – Green ash 5 – White birch 6 – Non-native broadleaves 7 – Mixed broadleaves 8 – Non-forested
2.5 3.2 1.7 1.7 2.5 2.2 54.1 32.2
128.8 41.1 190.7 80.8 58.2 99.6 48.8 0.0
480 138 711 275 265 556 208 0
24.1 56.3 14.6 31.0 45.3 18.8 46.4 66.9
WS (28.2) HL (20.0) SM (100.0) GA (65.3) WB (26.1) CA (21.0) NM (7.5) None
RP (17.3) AP (3.0) AB (17.5) EB (4.0) MM (4.7) SE (11.2) SM (2.8) None
NS (15.4) NM (2.6) IW (17.2) SM (3.5) WA (4.0 MM (8.8) WA (2.0) None
Urban Ecosyst Fig. 4 Discriminant function scores for the discriminant analysis of the ecosystem component variables for the six classes of urban forest species assemblages
amenities and affluence is important for addressing environmental justice (Heynen et al. 2006). However, with regard to biodiversity management and urban forest resilience, it is important recognize and not discount the role of the built environment and biophysical conditions of the city. Regarding the built environment, increasing imperviousness saw not just reduced tree density but a decline in native species abundance. For instance, Class 3 (mixed broadleaves) and Class 2 (white cedar) with higher impervious cover saw a greater abundance of non-native and hardy naturalized species like Manitoba maple and Siberian elm. It is, however, worthwhile to consider the nuances in the concept of nativeness in urban forests. White cedar, a species native to the region, was the most abundant tree species in the city and the dominant feature of one of the six species assemblages (i.e., Class 2). However, much of this white cedar population is comprised of
Fig. 5 Classification accuracy of discriminant analysis of the ecosystem component variables for the six species assemblages
small trees planted in hedges and along fencelines on residential properties (City of Toronto 2010). Given their location and likely point of origin from nurseries or retail garden centres, their contribution to biodiversity is speculative due to a lack of genetic diversity. Regarding non-native urban tree species, some common examples like Austrian pine are not naturalized and their distribution is restricted to sites where they are planted, while others like Norway maple include both planted/maintained street trees and dense colonization of residual forests (Martin and Marks 2006). The distribution and mix of native and non-native species across the urban landscape is not as simple as residual forests versus built-up areas, but is dependent on a number of these above characteristics. Future research on social-ecological drivers of urban forest structure and biodiversity might differentiate between genetic provenances, cultivated varieties, and the degree to which non-native species are naturalized. An overarching question of this study is whether urban forest species composition is predictable using coarse environmental variables as it is in hinterland forests, or is it unpredictably heterogeneous and driven predominantly by social processes (e.g., landowner tree planting preferences). This question has important implications for how urban forest practitioners can classify the urban landscape for the purposes of resource assessment and strategic planning. Ecosystem classification, or ecological land classification, is one approach to landscape classification that has a long history in natural resource management (Omernik 1987; Klijn and Udo de Haes 1994). Through the identification of key biophysical components and processes and their subsequent organization into relatively homogenous management units, ecosystem classification helps to inform decision-making in an ecosystem-based management context (Grumbine 1994; Steenberg et al. 2015). However, the degree to which the predictability of ecosystem
Urban Ecosyst Table 9 Standardized canonical correlation coefficients for the tree species and land use variables used in the analysis
Variable
Canonical Correlation 1
Canonical Correlation 2
No trees
0.145
0.858
White cedar Sugar maple
0.090 −0.390
−0.241 0.069
Norway maple
0.014
−0.127
White ash Manitoba maple
−0.090 −0.313
0.160 0.006
Green ash
0.019
0.076
White spruce Ironwood
0.005 −0.353
−0.109 0.145
Siberian elm Black cherry
0.042 0.355
0.031 −0.127
Crabapple
−0.341
0.065
Trembling aspen Chockecherry
−0.502 −0.089
0.133 −0.023
European buckthorn Lawson cypress
−0.034 0.003
0.105 −0.049
Honeylocust White elm White birch Austrian pine
0.084 −0.270 −0.075 0.004
0.115 0.227 −0.053 0.035
American basswood White pine
0.058 −0.257
−0.106 0.142
Tree species
Norway spruce
−0.284
−1.237
Red pine White oak Land use Commercial OpenGreenSpace
0.036 0.221
1.314 −0.189
0.131 0.061
0.866 0.397
Transport Institutional ResidualForest
0.075 0.033 −0.956
0.607 0.466 0.386
conditions (e.g., species composition) is eroded by the greater role of social processes in shaping ecosystem conditions in cities is unknown. Future research on social and ecological drivers of urban forest ecosystem structure at the household scale is needed to fully characterize these processes. Additionally, research that investigates socio-ecological legacy effects (e.g., Boone et al. 2010) on ecosystem structure using empirical field data not just canopy cover data would also be valuable. Recent urban forest research has documented several social drivers of urban forest structure and function. Perhaps the most studied is personal preferences and behaviours of residents around tree planting practices, which is important since the majority of urban trees are typically situated on privately owned residential properties (Kenney and Idziak 2000). Studies have shown that residents’ tree planting practices are
influenced by socioeconomic status and cultural background (Fraser and Kenney 2000; Greene et al. 2011), perceived tree aesthetics and maintenance requirements (Conway 2016), local climate (Avolio et al. 2015), and the landscaping practices of neighbours (e.g., ‘keeping up with the Joneses’; Grove et al. 2006; Warren et al. 2008). The species selection process for municipal tree plantings is a similar example, though municipal species palettes tend to be better documented and more uniform across a given city (Conway and Vander Vecht 2015). Another social driver of urban forest species composition that is arguably less understood and documented is nursery and retail garden centre practices. Little research exists that focuses on how the popularity of certain species, commercial availability, and stocking practices of nurseries and garden centres influence tree planting decisions and ultimately urban forest species composition and diversity (Pincetl et al. 2013;
Urban Ecosyst
Conway and Vander Vecht 2015). It is most likely that urban forest structure is driven by a scale-dependent and spatiallyvariable mix of social and ecological drivers. Traditional approaches to forest ecosystem classification and ecosystembased management that integrate only biophysical ecosystem components and processes will likely be ineffective in the urban landscape. Conversely, most of the classification schemes currently applied in urban areas for the purposes of planning (e.g., land use, zoning, planning districts, neighbourhoods) tend to ignore key biophysical processes and would not be ideal for urban forestry. Both the results of this study and the urban forestry literature (e.g., Steenberg et al. 2015) point to limitations of using land use for these purposes despite the prominence of doing so, as both social and ecological drivers of urban forest structure and function are omitted. For ecosystem-based management in urban forestry to be relevant, any classification scheme should integrate both social and ecological processes. Such a decision-support tool would be additionally beneficial given the high cost and challenges of maintaining urban tree inventories for such purposes (Roman et al. 2013). As global populations become increasingly urban and reliant upon urban forest ecosystem services, building the repertoire of robust, science-based management tools for sustainable urban forest management can contribute to the establishment of resilient cities. Acknowledgements Funding for this research was provided by the Killam Trusts at Dalhousie University. Thank you to Dr. Peter Duinker and Dr. Christopher Greene at Dalhousie University for their review of the draft manuscript.
References Avolio ML, Pataki DE, Pincetl S, Gillespie TW, Jenerette GD, McCarthy HR (2015) Understanding preferences for tree attributes: the relative effects of socio-economic and local environmental factors. Urban Ecosyst 18:73–86 Bailey RG (2009) Ecosystem geography: from ecoregions to sites, 2nd edn. Springer, New York Boone CG, Cadenasso ML, Grove JM, Schwarz K, Buckley GL (2010) Landscape, vegetation characteristics, and group identity in an urban and suburban watershed: why the 60s matter. Urban Ecosyst 13: 255–271 Bourne KS, Conway TM (2014) The influence of land use type and municipal context on urban tree species diversity. Urban Ecosyst 17:329–348 Brady RF, Tobias T, Eagles PJF, Ohrner R, Micak J, Veale B, Dorney RS (1979) A typology for the urban ecosystem and its relationship to larger biogeographical landscape units. Urban Ecol 4:11–28 Burns RM, Honkala BH (1990) Silvics of North America: 1. Conifers; 2. Hardwoods. Washington, DC, United States Department of Agriculture Forest Service Call LJ, Nilsen ET (2003) Analysis of spatial patterns and spatial association between the invasive tree-of-heaven (Ailanthus altissima) and the native black locust (Robinia pseudoacacia). Am Midl Nat 150: 1–14
Cao Y, Larsen DP, Thorne RSJ (2001) Rare species in multivariate analysis for bioassessment: some considerations. J N Am Benthol Soc 20:144–153 City of Toronto (2010) Every tree counts: A portrait of Toronto’s urban forest. http://www1.toronto.ca/staticfiles/City%20Of%20Toronto/ Parks%20Forestry%20&%20Recreation/Urban%20Forestry/Files/ pdf/E/every_tree_counts.pdf. Accessed 12 June 2017 Conway TM (2016) Tending their urban forest: residents’ motivations for tree planting and removal. Urban For Urban Greening 17:23–32 Conway TM, Bourne KS (2013) A comparison of neighborhood characteristics related to canopy cover, stem density and species richness in an urban forest. Landsc Urban Plan 113:10–18 Conway TM, Vander Vecht J (2015) Growing a diverse urban forest: species selection decisions by practitioners planting and supplying trees. Landsc Urban Plan 138:1–10 Dorney JR, Guntenspergen GR, Keough JR, Stearns F (1984) Composition and structure of an urban woody plant community. Urban Ecol 8(1–2):69–90 Duinker PN, Ordóñez C, Steenberg JWN, Miller KH, Toni SA, Nitoslawski SA (2015) Trees in Canadian cities: indispensable life form for urban sustainability. Sustainability 7:7379–7396 Environment Canada (2015) Canadian climate normal: 1981–2010 climate normals & averages. http://climate.weather.gc.ca/climate_ normals/index_e.html. Accessed 12 June 2017 Farrar JL (1995) Trees in Canada. Markham, Canadian Forest Service and Fitzhenry & Whiteside Ltd Foster J, Sandberg LA (2004) Friends or foe? Invasive species and public green spaces in Toronto. Geogr Rev 94:178–198 Fraser ED, Kenney WA (2000) Cultural background and landscape history as factors affecting perceptions of the urban forest. J Arboric 26: 106–113 Greene CS, Millward AA (2016) The legacy of past tree planting decisions for a city confronting emerald ash borer (Agrilus planipennis) invasion. Front Ecol Evol 4:1–12 Greene CS, Millward AA, Ceh B (2011) Who is likely to plant a tree? The use of public socio-demographic data to characterize client participants in a private urban forestation program. Urban For Urban Green 10:29–38 Grimm NB, Grove JM, Pickett STA, Redman CL (2000) Integrated approaches to long-term studies of urban ecological systems. Bioscience 50:571–584 Grove JM, Troy AR, O’Neil-Dunne JPM, Burch WR, Cadenasso ML, Pickett STA (2006) Characterization of households and its implications for the vegetation of urban ecosystems. Ecosystems 9:578–597 Grove JM, Locke DH, O’Neil-Dunne JPM (2014) An ecology of prestige in new York City: examining the relationships among population density, socio-economic status, group identity, and residential canopy cover. Environ Manag 54:402–419 Grumbine RE (1994) What is ecosystem management? Conserv Biol 8: 27–38 Hamann A, Wang T (2006) Potential effects of climate change on ecosystem and tree species distribution in British Columbia. Ecol 87: 2773–2786 Hargrove WW, Hoffman FM (2005) Potential of multivariate quantitative methods for delineation and visualization of ecoregions. Environ Manag 34:S39–S60 Heynen NC, Lindsey G (2003) Correlates of urban forest canopy cover: implications for local public works. Public Works Manage Policy 8: 33–47 Heynen N, Perkins HA, Roy P (2006) The political ecology of uneven urban green space the impact of political economy on race and ethnicity in producing environmental inequality in Milwaukee. Urban Aff Rev 42:3–25 Jongman RHG, Ter Braak CJF, Van Tongeren OFR (1995) Data analysis in community and landscape ecology. Cambridge University Press, Cambridge
Urban Ecosyst Jutras P, Prasher SO, Mehuys GR (2010) Appraisal of key biotic parameters affecting street tree growth. J Arboric 36:1–10 Kendal D, Williams NSG, Williams KJH (2012) Drivers of diversity and tree cover in gardens, parks and streetscapes in an Australian city. Urban For Urban Green 11:257–265 Kenney WA, Idziak C (2000) The state of Canada’s municipal forests – 1996 to 1998. For Chron 76:231–235 Kenney WA, Van Wassenaer PJ, Satel AL (2011) Criteria and indicators for strategic urban forest planning and management. Arboricult Urban For 37:108–117 Klijn F, Udo de Haes HA (1994) A hierarchical approach to ecosystems and its implications for ecological land classification. Landsc Ecol 9: 89–104 Ko Y, Lee JH, McPherson EG, Roman LA (2015) Factors affecting longterm mortality of residential shade trees: evidence from Sacramento, California. Urban For Urban Green 14:500–507 Laćan I, McBride JR (2008) Pest vulnerability matrix (PVM): a graphic model for assessing the interaction between tree species diversity and urban forest susceptibility to insects and diseases. Urban For Urban Green 7:291–300 Lu JWT, Svenden ES, Campbell LK, Greenfeld J, Braden J, King KL, Falxa-Raymond N (2010) Biological, social, and urban design factors affecting young tree mortality in new York City. Cities Environ (CATE) 3:1–15 Martin PH, Marks PL (2006) Intact forests provide only weak resistance to a shade-tolerant invasive Norway maple (Acer platanoides L.). J Ecol 94:1070–1079 Matthews ER, Peet RK, Weakley AS (2011) Classification and description of alluvial plant communities of the piedmont region, North Carolina, USA. Appl Veg Sci 14:485–505 Maxwell RS, Taylor AH, Skinner CN, Safford HD, Isaacs RE, Airey C, Young AB (2014) Landscape-scale modeling of reference period forest conditions and fire behavior on heavily logged lands. Ecosphere 5:1–28 McDonnell MJ, Pickett STA (1990) Ecosystem structure and function along urban-rural gradients: an unexploited opportunity for ecology. Ecol 71:1232–1237 McGarigal K, Cushman S, Stafford SG (2000) Multivariate statistics for wildlife and ecology research. Springer, New York McKenney DW, Pedlar JH, Lawrence K, Campbell K, Hutchinson MF (2007) Potential impacts of climate change on the distribution of north American trees. Bioscience 57:939–948 McKinney ML (2008) Effects of urbanization on species richness: a review of plants and animals. Urban Ecosyst 11:161–176 McNab WH, Browning SA, Simon SA, Fouts PE (1999) An unconventional approach to ecosystem unit classification in western North Carolina, USA. For Ecol Manag 114:405–420 McPherson EG, Nowak D, Heisler G, Grimmond S, Souch C, Grant R, Rowntree R (1997) Quantifying urban forest structure, function, and value: the Chicago urban Forest climate project. Urban Ecosyst 1: 49–61 Mora F, Iverson L (2002) A spatially constrained ecological classification: rationale, methodology, and implementation. Plant Ecol 158: 153–169 Nitoslawski SA, Duinker PN, Bush PG (2016) A review of drivers of tree diversity in suburban areas: research needs for north American cities. Environ Rev 24:471–483 Nitoslawski SA, Steenberg JWN, Duinker PN, Bush PG (2017) Assessing the influence of location attributes on urban forest species composition in suburban neighbourhoods. Urban For Urban Greening 27:187-195 Nowak DJ, Dwyer JF (2007) Understanding the benefits and costs of urban forest ecosystems. In: Kuser JE (ed) Urban and community forestry in the northeast. New Brunswick, Springer, pp 25–46
Nowak DJ, Rowntree RA, McPherson EG, Sisinni SM, Kerkmann ER, Stevens JC (1996) Measuring and analyzing urban tree cover. Landsc Urban Plan 36:49–57 Nowak DJ, Kuroda M, Crane DE (2004) Tree mortality rates and tree population projections in Baltimore, Maryland, USA. Urban For Urban Green 2:139–147 Omernik JM (1987) Ecoregions of the conterminous United States. Ann Assoc Am Geogr 77:118–125 Ontario Ministry of Natural Resources (2012) Ontario’s forest regions: Deciduous forest. http://www.mnr.gov.on.ca/en/Business/Forests/ 2ColumnSubPage/STDPROD_091289.html. Accessed 12 June 2017 Ordóñez C, Duinker PN (2012) Ecological integrity in urban forests. Urban Ecosyst 15:863–877 Ordóñez C, Duinker PN (2013) An analysis of urban forest management plans in Canada: implications for urban forest management. Landsc Urban Plan 116:36–47 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 12:18–27 Pincetl S, Prabhu SS, Gillespie TW, Jenerette GD, Pataki DE (2013) The evolution of tree nursery offerings in Los Angeles County over the last 110 years. Landsc Urban Plan 118:10–17 Roman LA, McPherson EG, Scharenbroch BC, Bartens J (2013) Identifying common practices and challenges for local urban tree monitoring programs across the United States. Arboric Urban For 39:292–299 Schwarz K, Fragkias M, Boone CG, Zhou W, McHale M, Grove JM, O’Neil-Dunne J, McFadden JP, Buckley GL, Childers D, Ogden L, Pincetl S, Pataki D, Whitmer A, Cadenasso ML (2015) Trees grow on money: urban tree canopy cover and environmental justice. PLoS One 10:e0122051 Statistics Canada (2016) Census profile – Toronto. https://www12. statcan.gc.ca/census-recensement/2011/dp-pd/prof/details/page. cfm?Lang=E&Geo1=CSD&Code1=3520005&Geo2= PR&Code2=35&Data=Count&SearchText=Toronto&SearchType= Begins&SearchPR=01&B1=All&GeoLevel=PR&GeoCode= 3520005. Accessed 12 June 2017 Stearns F, Montag T (1974) The urban ecosystem: a holistic approach. Stroudsburg, Dowden, Hutchinson, and Ross. Steenberg JWN, Duinker PN, Charles JD (2013) The neighbourhood approach to urban forest management: the case of Halifax, Canada. Landsc Urban Plan 117:135–144 Steenberg JWN, Millward AA, Duinker PN, Nowak DJ, Robinson PJ (2015) Neighbourhood-scale urban forest ecosystem classification. J Environ Manag 163:134–145 Steenberg JWN, Millward AA, Nowak DJ, Robinson PJ (2017a) A conceptual framework of urban forest ecosystem vulnerability. Environ Rerv 25:115–126 Steenberg JWN, Robinson PJ, Millward AA (2017b) The influence of building renovation and rental housing on urban trees. J Environ Plan Manage 61:553-567 Toni SA, Duinker PN (2015) A framework for urban–woodland naturalization in Canada. Environ Rev 23:321–336 Trowbridge PJ, Bassuk NL (2004) Trees in the urban landscape. Wiley, Hoboken Troy AR, Grove JM, O’Neil-Dunne JP, Pickett ST, Cadenasso ML (2007) Predicting opportunities for greening and patterns of vegetation on private urban lands. Environ Manag 40:394–412 Warren PS, Lerman SB, Charney ND (2008) Plants of a feather: spatial autocorrelation of gardening practices in suburban neighbourhoods. Biol Conserv 141:3–4 Womack B, Carter R (2011) Landscape scale forest community classification in the Horseback Mountain range of the Talladega National Forest, Alabama. Nat Areas J 31:51–64
Urban Ecosyst Woodall CW, Nowak DJ, Liknes GC, Westfall JA (2010) Assessing the potential for urban trees to facilitate forest tree migration in the eastern United States. For Ecol Manag 259:1447–1454 Zhou Y, Narumalani S, Waltman WJ, Waltman SW, Palecki MA (2003) A GIS-based spatial pattern analysis model for eco-region mapping and characterization. Int J Geogr Inf Sci 17:445–462
Zipperer WC, Sissinni SM, Pouyat RV (1997) Urban tree cover: an ecological perspective. Urban Ecosyst 1:229–246 Zmyslony J, Gagnon D (2000) Path analysis of spatial predictors of frontyard landscape in an anthropogenic environment. Landsc Ecol 15: 357–371