UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Assessing and analyzing urban tree condition using airborne remote sensing Plowright, Andrew Alexander 2016

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2016_september_plowright_andrew.pdf [ 4.26MB ]
Metadata
JSON: 24-1.0306910.json
JSON-LD: 24-1.0306910-ld.json
RDF/XML (Pretty): 24-1.0306910-rdf.xml
RDF/JSON: 24-1.0306910-rdf.json
Turtle: 24-1.0306910-turtle.txt
N-Triples: 24-1.0306910-rdf-ntriples.txt
Original Record: 24-1.0306910-source.json
Full Text
24-1.0306910-fulltext.txt
Citation
24-1.0306910.ris

Full Text

ASSESSING AND ANALYZING URBAN TREE CONDITION USING AIRBORNE REMOTE SENSING by  Andrew Alexander Plowright     A thesis submitted in partial fulfillment of the requirements for the degree of MASTER OF SCIENCE  in  THE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES  (Forestry)      THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)  July 2016 © Andrew Alexander Plowright 2016 ii  Abstract Though urban forests play a key role in the sustainable development of cities, the environmental stressors faced by urban trees are numerous. City managers rarely have access to information on the condition of urban trees, which impedes both their management and the study of environmental factors that affect their vitality, such as landscape imperviousness. The research presented in this thesis aims to bridge this gap in information. The capacity of airborne light detection and ranging (LiDAR) and hyperspectral imagery to evaluate tree condition in the city of Surrey, British Columbia, Canada, is explored, and its relationship with impervious land cover is investigated.  LiDAR was used to estimate two indicators of tree condition: crown density and tree height. Tree heights estimated by LiDAR were well correlated with field measurements (Pearson’s r = 0.927, p < 0.001), while the coefficient of variation of return height was able to predict crown density with an r2 = 0.617 for trees over 8 m. The heights of 1,914 trees across the city were then estimated using LiDAR. To account for the effects of age, species-specific height models were fitted using planting dates recorded by city authorities. The residuals from these models were then used as indicators of tree condition.  A 1.0 m resolution classified land cover map (accuracy of 88.6%) was produced for the city, from which landscape imperviousness was then derived. When aggregated to broad-scale 0.5 km2 spatial units, negative relationships (r2 between 0.292 and 0.753) were found between height model residuals and land cover imperviousness. However, this relationship was found to be non-significant when examined at the individual tree level. iii  We conclude that imperviousness does not appear to be a significant driver of tree height variation, with negative broad-scale relationships likely due to correlations with other unmeasured environmental variables associated with the urban-rural gradient. Hyperspectral and LiDAR data proved to be a powerful tool for mapping land cover and imperviousness. The results of this research show that airborne remote sensing is a promising tool for assessing the condition of urban trees and studying the environmental factors that impact their development.      iv  Preface This thesis is the combination of two scientific papers written for peer-review of which I am the lead author. I was responsible for defining research objectives, developing methodology and analyzing data, in addition to writing and editing on both manuscripts. Conceptual development, project oversight and editorial assistance was provided by Dr. Nicholas Coops. This project was undertaken in partnership with the City of Surrey. Neal Aven, manager of Surrey’s Urban Forestry and Environmental Programs, contributed all the data analyzed in this thesis, as well as technical advice and editorial assistance. Drs. Stephen Sheppard and Bianca Eskelson were both involved in project development and provided editorial assistance and statistical advice. Curtis Chance assisted in field work and data processing. Publications arising from this thesis include:  Chapter 3: Plowright, A. A., Coops, N. C., Eskelson, B. N. I., Sheppard, S. R. J., Aven, N. W. (2016). Assessing urban tree condition using airborne light detection and ranging.  Chapter 4: Plowright, A. A., Coops, N. C., Chance, C. M., Eskelson, B. N. I., Sheppard, S. R. J., Aven, N. W. (2016).  Multi-scale analysis of relationship between imperviousness and urban tree height using airborne remote sensing.  v  Table of Contents Abstract ....................................................................................................................................................................................... ii Preface ......................................................................................................................................................................................... iv Table of Contents ...................................................................................................................................................................... v List of Tables ............................................................................................................................................................................ vii List of Figures ......................................................................................................................................................................... viii List of Equations ....................................................................................................................................................................... x Glossary ...................................................................................................................................................................................... xi Acknowledgments ................................................................................................................................................................. xii 1. Introduction ..................................................................................................................................................................... 1 1.1. Benefits and challenges of managing urban trees .................................................................................. 1 1.2. The need for information on tree condition ............................................................................................. 2 1.3. Remote sensing in urban forestry ................................................................................................................. 3 1.3.1. Remote sensing of tree condition indicators .................................................................................. 5 1.3.2. Remote sensing of impervious surfaces ............................................................................................ 9 1.4. Research objectives ........................................................................................................................................... 10 2. Data & study site .......................................................................................................................................................... 12 2.1. Study site ............................................................................................................................................................... 12 2.2. Tree database ....................................................................................................................................................... 13 2.3. Remotely sensed data ....................................................................................................................................... 14 3. Detecting and assessing the condition of urban trees using airborne LiDAR ..................................... 18 3.1. Introduction ......................................................................................................................................................... 18 3.1.1. Surveying urban tree condition .......................................................................................................... 18 3.1.2. Trends in individual tree extraction from LiDAR data ............................................................. 18 3.1.3. Extracting trees in the urban landscape ......................................................................................... 20 3.1.4. Chapter objectives .................................................................................................................................... 20 3.2. Methods .................................................................................................................................................................. 21 3.2.1. Field data ...................................................................................................................................................... 21 3.2.2. Canopy Height Model .............................................................................................................................. 22 3.2.3. Locating trees ............................................................................................................................................. 23 3.2.4. Outlining tree crowns ............................................................................................................................. 26 3.2.5. Estimating crown density ..................................................................................................................... 27 3.2.6. Tree height as an indicator of condition ......................................................................................... 30 3.3. Results..................................................................................................................................................................... 30 3.3.1. Treetop location and tree height estimation ................................................................................. 30 3.3.2. Tree crown outlines ................................................................................................................................. 32 3.3.3. Crown density estimation ..................................................................................................................... 34 3.4. Discussion ............................................................................................................................................................. 36 3.4.1. Accuracy of treetop location algorithm ........................................................................................... 36 3.4.2. Performance of crown outline methods ......................................................................................... 37 3.4.3. Potential use of LiDAR metrics for assessing urban tree condition .................................... 38 3.4.4. Use of a raster-based canopy height model ................................................................................... 40 4. Examining the relationship between urban tree condition and landscape imperviousness ....... 42 4.1. Introduction ......................................................................................................................................................... 42 4.1.1. Environmental challenges to urban tree development ............................................................ 42 4.1.2. Multi-scale analysis .................................................................................................................................. 42 vi  4.1.3. Chapter objectives .................................................................................................................................... 43 4.2. Methods .................................................................................................................................................................. 44 4.2.1. Trees of interest ........................................................................................................................................ 44 4.2.2. Field measurements ................................................................................................................................ 44 4.2.3. LiDAR data pre-processing ................................................................................................................... 45 4.2.4. Treetop detection and height estimation ....................................................................................... 45 4.2.5. Land cover classification and estimation of imperviousness ................................................. 46 4.2.6. Multi-scale analysis of relationships between imperviousness and tree height ............ 50 4.3. Results..................................................................................................................................................................... 52 4.3.1. Land cover classification ....................................................................................................................... 52 4.3.2. Derivation of height model residuals ............................................................................................... 53 4.3.3. Relationships between imperviousness and height model residuals ................................. 54 4.4. Discussion ............................................................................................................................................................. 57 4.4.1. Remote sensing of landscape imperviousness ............................................................................. 57 4.4.2. Analysis of relationships between imperviousness and tree height ................................... 58 5. Conclusions..................................................................................................................................................................... 61 5.1. Key findings .......................................................................................................................................................... 61 5.1.1. Using LiDAR and GIS data to estimate tree condition indicators ......................................... 61 5.1.2. Relationship between tree condition and imperviousness..................................................... 62 5.2. Implications for urban tree management ................................................................................................ 62 5.3. Limitations ............................................................................................................................................................ 64 5.4. Recommendations for future research ..................................................................................................... 66 References ................................................................................................................................................................................ 68 Appendix 1 ............................................................................................................................................................................... 84 Appendix 2 ............................................................................................................................................................................... 86    vii  List of Tables Table 1-1. Summary of studies using remote sensing technology to assess indicators of tree condition. ........................................................................................................................................................................... 8 Table 2-1. Flight parameters for LiDAR and hyperspectral imagery acquisition. ...................................... 16 Table 3-1. Number of reference trees per height class. ......................................................................................... 29 Table 4-1. Number of trees of interest per species retained for analysis. ..................................................... 44 Table 4-2. Number of training points, producer’s and user’s accuracies, and average imperviousness of each land cover class. Imperviousness percentages are referenced from (Hodgson & Bresnahan, 2004). ........................................................................................................................................................ 53 Table 4-3. Coefficients for Chapman-Richards height models and model residual statistics. ............... 54 Table 4-4. Spatial correlation structures and p-values of GLS models relating height model residuals and imperviousness for three circular area size ratios. ............................................................................... 55     viii  List of Figures Figure 2-1. Location of the city of Surrey within the Greater Vancouver regional district. ................... 13 Figure 2-2. Subset of four reflectance bands of a hyperspectral image of a suburban street in the city of Surrey. These bands roughly correspond to blue (473.3 nm), green (530.7 nm), red (693.0 nm) and near infrared (1008.1 nm) ...................................................................................................... 15 Figure 2-3. Orthographic photo of a suburban street in the city of Surrey and a LiDAR point cloud of the same area. LiDAR returns are coloured according to elevation. ...................................................... 17 Figure 3-1. Location of trees measured in the field within the city of Surrey. ............................................. 22 Figure 3-2. Variably-sized moving focal window: a) a moving focal window scans the canopy height model; b) window size varies according to the canopy height; c) cells are tagged as tree tops (red) if they have the highest value within the window. ............................................................................. 24 Figure 3-3. Process for locating the treetop of an entry in the city’s urban tree database. The white dotted line represents the approximate coordinates of the tree as registered within the GIS. A variable window filter (VWF) is used to locate all potential treetops within a 20 m2 area centered on these coordinates. Using growth curve formulas, upper and lower height thresholds are computed according to the age of the tree. Finally, the closest treetop within these thresholds is selected. .................................................................................................................................... 26 Figure 3-4. Watershed segmentation: A) a CHM of two neighboring trees; B) the CHM is inverted so both trees resemble basins; C) the basins are filled with water from the bottom up; D) borders between the two tree crowns are drawn where the basins connect. ..................................................... 27 Figure 3-5. Rationale for using a resized crown outline when estimating crown density with the coefficient of variation of LiDAR point heights. Differentiating the vertical distribution of points between a sparse and a dense crown may be difficult when using a subset that extends to the edges of the tree (gray area). When the subset is limited to the apex of the tree and the interior of its crown (crosshatched area), a large amount of the returns being reflected from the exterior of the tree are excluded. Most of the remaining returns from a dense tree will occur near its apex, while a sparse tree, having allowed more returns to reach the interior of its crown, will have a more even distribution of return heights along its vertical profile. ................. 29 Figure 3-6. Relationship between tree heights measured in the field and tree heights measured from the LiDAR data. Four outliers whose difference between field and LiDAR height was SD > 3 from the mean were removed. ............................................................................................................................ 31 Figure 3-7. Chapman-Richards height model fitted to the field trees. Many trees share the same planting year, resulting in clustering along the x-axis. This model is specific to the field-measured trees, and the significance of its residuals may only be considered relative to the other trees in the dataset. ......................................................................................................................................... 32 Figure 3-8. Examples of automatically generated (dotted lines) and manually delineated (solid lines) tree crown outlines. A) An accurate watershed segmentation outline (A = 0.92) B) A watershed segmentation error of commission: a neighboring tree or shrub has been included in the outline (A = 0.42). C) An accurate circular outline (A = 0.88). D) An inaccurate circular outline caused by an off-center tree apex (A = 0.39). ................................................................................... 33 ix  Figure 3-9. Ordinary least-squares linear regression models of field-measured crown density versus LiDAR metrics across multiple age classes. Left column: percentage of non-ground LiDAR returns. Right column: coefficient of variation of LiDAR heights. Based on r2 values, the coefficient of variation yielded the strongest relationships with crown density. Predictive power was highest for trees over 8 m. As tree height is lowered, the significance of relationships is diminished. ..................................................................................................................................... 35 Figure 3-10. A comparison of tree condition indicators for 40 field trees around Kildare Drive, in Surrey, British Columbia, Canada. Gray points represent trees that are not western redcedars. A) Tree height model residuals for western redcedar. B) Estimated crown density. ..................... 36 Figure 3-11. Off-nadir LiDAR pulses may have unobstructed paths to the ground below an isolated tree in an open area (left). These pulses can be intercepted by neighbouring objects, reducing the number of returns to reach the ground below a tree in clustered surroundings (right). ...... 39 Figure 4-1. A) Classified land cover map of the city of Surrey. B) Imperviousness map derived from land cover. C) Locations of all trees of interest. Note that some trees may not be visible due to overlapping points. D) Average height model residual per 0.5 km2 hexagonal cell. ........................ 49 Figure 4-2. Four sample trees used in the tree-level analysis representing a range of species, ages, height model residuals and levels of imperviousness. First column: orthographic photos. Second column: imperviousness measured within three circular areas of varying sizes centered on each tree (radii measuring 1.2, 2.0 and 3.0 times that of the crown). Dark pixels represent areas that are highly impervious (buildings, pavement), whereas light pixels are more pervious (grass, bare ground, shrubs, tree canopy). Pixels corresponding to the tree on which the areas are centered are excluded. Third column: tree information, including height model residuals and mean imperviousness within each circular area. .................................................................................. 51 Figure 4-3. Average height model residual versus average imperviousness for hexagonal grid cell sizes of 0.25, 0.5 and 1.0 km2, and minimum tree counts per cell of 20 and 30. Points with positive values on the y-axes have positive average height model residuals, i.e.: trees within these cells are on average higher for their age, and vice versa. Square points represent highly influential points. Models in which influential points have been removed are represented by dotted regression lines and italicized r2 and p-values. Significant relationships were found for 0.5 km2 units. ................................................................................................................................................................. 56    x  List of Equations Equation 3-1 ............................................................................................................................................................................ 23 Equation 3-2 ............................................................................................................................................................................ 24 Equation 3-3 ............................................................................................................................................................................ 30    xi  Glossary A A segmentation assessment value between two outlines. A = 1 indicates a perfect match, while any other case corresponds to A < 1. Chapman-Richards A function commonly used to describe a cumulative growth curve. CHM Canopy height model. A rasterized surface derived from raw LiDAR data by recording the difference in elevation between the highest classified vegetation return and the underlying DEM within each cell of a continuous grid. CV Coefficient of variation. A measure of dispersion calculated as the ratio of the standard deviation of a set of data values to its mean. DEM Digital elevation model. A rasterized surface in which each cell value is equal to the terrain’s elevation above sea level at that location. GIS Geographical information system. A system in which spatial data is stored, manipulated and analyzed. GPS Global positioning system. A space-based navigation system accessible through mobile receivers. Hyperspectral imagery In image in which reflectance is captured across several narrow, continuous spectral bands. Imperviousness The quality of a material that prevents the passage of water. In an urban context, impermeable surfaces include buildings, roads, parking lots and any other feature which prevents water from infiltrating into the underlying soil. LiDAR Light detection and ranging. A remote sensing technology that uses pulses of infrared light to acquire precise three-dimensional spatial coordinates. Q The ratio between the areas of an automated outline and its corresponding reference outline. Random forest An ensemble learning technique that constructs multiple decision trees to classify observations based on multiple variables. SD A measure of dispersion of a set of data values calculated as the square root of its variance. VWF Variable window filter. A treetop detection algorithm that uses a moving focal window to scan the canopy for local canopy height maxima. xii  Acknowledgments Funding for this project was provided by an NSERC Discovery grant to Dr. Coops and Engage and Engage+ collaboration grants with Surrey City Energy. The City of Surrey provided the technical support and the high-quality data which was the foundation for this research. I would like to thank the organizations that provided financial assistance through various awards and scholarships. These include NSERC, the Fond de recherche du Québec – Nature et technologies, the Canadian Cartographic Association, Esri Canada, the Canadian TREE Fund, the Canadian Council on Ecological Areas and the Canadian Wildlife Federation. Fellowships granted by the University of British Columbia include the VanDusen Graduate Scholarship, the Mary and David Macaree Fellowship and the James Robert Thompson Fellowship. I am grateful to my supervisory committee, Dr. Stephen Sheppard, Dr. Bianca Eskelson and Neal Aven, for their involvement in this project. The members of the Integrated Remote Sensing Studio provided support and company for which I am very appreciative. In particular, I would like to thank Mitchell Vartanian, Trevor Liu and Curtis Chance for their assistance in lab and fieldwork, and Gregory Rickbeil and Dr. Piotr Tompalski, for their advice on statistical analysis and data processing.  Lastly, I would like to express my deep gratitude to Dr. Nicholas Coops for his excellent supervision. His expertise, constant support and motivation, and generous amount of time and effort made this project possible. Dr. Coops’ outstanding work ethic has created a truly exceptional environment in which to learn and conduct research, and I feel extremely fortunate to have benefited from his guidance.   xiii  Dedicated to my loving family Far away but always close at heart   1   1. Introduction 1.1. Benefits and challenges of managing urban trees Recognition of urban trees’ key role in sustainable urban development has led to the recent emergence of the discipline of urban forestry. In the past decades, scientific research has revealed a broad range of social, environmental, and economic benefits derived from urban trees. Trees absorb airborne pollutants, diminish flood hazards, and improve local water quality (Nowak, Crane, & Stevens, 2006; Xiao & McPherson, 2002). Urban forests mitigate the effects of heat islands, allay heat-stress related illnesses by providing shade, and can reduce energy consumption associated with cooling buildings (McPherson & Simpson, 2003; Rosenfeld, Akbari, Romm, & Pomerantz, 1998). Citizens who enjoy regular contact with green spaces experience an improved sense of well-being and community cohesion (Herzog & Strevey, 2008; Peckham, Duinker, & Ordóñez, 2013). Appreciation for urban trees is reflected by the higher property values in vegetated neighborhoods, which in turn increase city income through tax revenues (Anderson & Cordell, 1985; Payton, Lindsey, Wilson, Ottensmann, & Man, 2008). Though the benefits derived from urban forests are numerous, improperly managed trees can lead to serious property damage or personal injury (Smiley, Fraedrich, & Fengler, 2007). The challenges associated with the management of trees in urban areas are diverse. Cities possess distinct microclimates, hydrological systems and soil compositions, as well as higher levels of air and soil pollution than exurban areas  (Pickett et al., 2011). These environmental characteristics are largely caused by human activities and imprints on the landscape, a phenomenon commonly referred to as “urbanization” (Hahs & McDonnell, 2006). Trees in urban ecosystems are exposed to higher levels of atmospheric deposition and concentrations of airborne pollutants (Gregg, Jones, & 2  Dawson, 2003; Lovett et al., 2000a). Trenching operations and the application of de-icing salts impact urban trees, which can also sustain accidental mechanical damage from vehicles and construction equipment (Baines, 1994; Malthus & Younger, 2000). A particular concern in urban areas is the widespread presence of impervious surfaces: landscape features such as roads, buildings and parking lots which prevent water from infiltrating into the underlying soil (Arnold Jr. & Gibbons, 1996). These surfaces have significant impacts on urban ecological systems. The construction of impervious features can compact underlying soil and prevent roots from spreading (Jim, 1993, 1998). Pavement is associated with reduced soil aeration and increased concentrations of heavy metals, soluble salts and soil acidity (Morgenroth & Buchan, 2009; Pouyat, McDonnell, & Pickett, 1995). Highly impervious landscapes are also largely responsible for the well-documented heat island effect (Arnfield, 2003). Previous studies have shown that solar heat absorbed and then re-radiated by impervious surfaces can affect the development of urban plants and trees (Kjelgren & Montague, 1998; Mueller & Day, 2005) 1.2. The need for information on tree condition To meet these challenges and reap the full benefits of urban forests, it is incumbent on city authorities to implement comprehensive tree management strategies. While cities have managed trees and other forms of green spaces for centuries, recent enthusiasm for expanding urban forests may outpace a city’s capacity to support them (Konijnendijk, Ricard, Kenney, & Randrup, 2006). While high-profile initiatives such as Toronto’s goal of doubling its tree canopy, and New York City’s One Million Trees project receive considerable public attention, not all such undertakings are successful (City of Toronto, 2008; Locke et al., 2010). For instance, the city of Gatineau, Quebec, planted 183,000 trees over four years as part of a city greening project. Three years later, the survival rate of these newly planted trees was between 5% and 14%, an outcome mainly attributed to a lack of proper monitoring and maintenance (St-Pierre, 2011). 3  Urban forests require intensive management, with expenditures towards urban trees per acre far surpassing those in forested lands outside of cities (McPherson, 1993). Yet, despite the considerable investment required by urban forests, cities often lack comprehensive information on tree condition, without which authorities are inhibited in their ability to efficiently organize resources. McPherson (1993) identified the rarity of effective tree condition monitoring programs as a key issue in urban forestry more than two decades ago. While urban tree monitoring pilot projects have been developed by the U.S. Department of Agriculture, Forest Service (USFS), these programs rely on small numbers of sampling plots across multiple cities, and were meant to analyze broad-scale trends as opposed to providing actionable information to local managers (Cumming, Twardus, & Nowak, 2008). Furthermore, sampling plot approaches may not lend themselves well to urban areas, which are spatially heterogeneous, contain a wide range of tree species and age classes, and whose constant land use changes render permanent plots impractical (McPherson, 1993). Though uncommon, exhaustive tree censuses have been performed in cities such as Canberra, Australia (Brack, 2006). The resulting information has allowed city managers to anticipate the location, timing and cost of tree maintenance treatments (Banks, Brack, & James, 1999). Many non-profit organizations and municipal agencies have recognized the potential benefits of these programs, though limited staff, funding, and data management capacities remain frequently cited challenges to their implementation (Roman, McPherson, Scharenbroch, & Bartens, 2013). 1.3. Remote sensing in urban forestry With logistical and financial limitations making field-based assessments nonviable for many cities, remote sensing has the potential to supply significant information for evaluating tree condition and studying its environmental drivers. By providing efficient and repeatable means of acquiring 4  quantitative and spatially explicit data, air- and space-borne sensors have been noted for their potential for studying vegetation over large areas (Xie, Sha, & Yu, 2008). Remote sensing technologies are generally divided into two types of sensors: active and passive (Campbell & Wynne, 2011). Passive sensors detect naturally occurring energy, which is either reflected or emitted from terrestrial features. The most common form of passive remote sensing involves measuring the intensity of radiation within one or more wavelength bands of the electromagnetic spectrum. Images are created when scanning is performed using two-dimensional arrays of detectors. Sensors measuring reflected energy within three to ten reflectance bands and have been in use for decades (Kramer, 2002). These range from digital cameras to sophisticated sensors aboard the Landsat, QuickBird or Spot satellites. Recent advances in data storage and component fabrication, however, have led to the development of sensors capable of acquiring measurements within hundreds of narrow contiguous reflectance bands. These instruments, known as “hyperspectral” sensors, offer a greater potential to detect the unique spectral properties of various terrestrial features (Eismann, 2012). In contrast with passive sensors, active remote sensing systems emit their own energy, which is directed to the area of interest and then reflected back to the sensor. While potentially more complex and costly, active sensors offer the key advantage of being independent of solar illumination. Commonly used active remote sensing technologies include radar systems, which detect and image targets using emitted radio waves. (Campbell & Wynne, 2011). Analogous to radar, light detection and ranging (LiDAR) sensors calculate the distance between the instrument and terrestrial features using reflected pulses of infrared light (Baltsavias, 1999). By emitting hundreds of thousands of pulses per second, LiDAR systems can scan large swaths of terrain when mounted on aircraft. 5  Several urban forestry and ecology studies have employed various combinations of LiDAR data, multispectral and hyperspectral imagery and geospatial data. Vegetation indices derived from multispectral imagery have been used to assess the relationship between vegetation and the urban heat island effect (Weng, Lu, & Schubring, 2004). Remotely sensed lawn characteristics have been used to model fertilization practices on urban residential properties (Zhou, Troy, & Grove, 2008). When integrated with spectral imagery, several studies have noted LiDAR’s utility for mapping urban canopy cover in dense urban areas where the shadowing effects of buildings can be problematic (MacFaden, O’Neil-Dunne, Royar, Lu, & Rundle, 2012; O’Neil-Dunne, MacFaden, Royar, & Pelletier, 2013). LiDAR’s ability to measure three-dimensional structures was leveraged by Tooke, Coops, & Voogt (2009) to model the effects of urban trees on solar radiation. Other applications of LiDAR and high resolution imagery include mapping urban tree species (C. Zhang & Qiu, 2012) and estimating citywide carbon storage (Schreyer, Tigges, Lakes, & Churkina, 2014). Recently, the fusion of LiDAR and spectral imagery has gained particular attention for its potential utility in detecting and mapping invasive plant species (Chance, Coops, Crosby, & Aven, 2016; Singh, Davis, & Meentemeyer, 2015). 1.3.1. Remote sensing of tree condition indicators Previous research has also explored the potential of remote sensing to assess tree condition, a term which is complex and multifaceted in itself. While early definitions of healthy trees and forests were formulated in terms of management objectives, a recent focus on ecological processes has shifted the understanding of tree health towards ecosystem-based definitions (Kolb, Wagner, & Covington, 1994). These incorporate a variety of factors, such as productivity, resistance to pests and disease, and the ability to recover from catastrophic events. Furthermore, the notion of health is heavily dependant on scale, ranging from the perspective of a single tree to that of forests, landscapes and global ecosystems (Trumbore, Brando, & Hartmann, 2015). Although definitions of tree condition 6  may vary, common threads include the capacity of a tree to survive, grow and reproduce. Dobbertin (2005) notes that hypothetical optimal tree vitality cannot be determined, and so the condition of a tree is instead understood in terms of its reactions to stress-inducing agents. Since vitality cannot be measured directly, Stone, Coops, & Culvenor (2000) define the condition of a tree as the culmination of the symptoms of environmental stress. Indicators of tree condition vary considerably in their ease of measurement.  Drought-stressed trees exhibit a variety of cellular, biochemical and molecular changes, several of which can only be observed using laboratory equipment (Dobbertin, 2005). While the yellowing of senescent leaves can be observed visually, subtler variations in pigment content can be detected using spectral sensors (Blackburn, 2007). At a broader scale, sub-optimal environmental conditions will impact the architecture of a tree over time. Various crown metrics, such as its diameter, density and volume, are frequently used as indicators of tree condition (Zarnoch, Bechtold, & Stolte, 2004). When compared against a reference for its age, the structural metrics of a tree can be used to assess its reaction to environmental stressors (Dobbertin, 2005). For instance, the commonly used “site index” measurement describes the growing potential of an area based on the height of its trees relative to their age (Clutter, Fortson, Pienaar, Brister, & Bailey, 1983).  The capacity of remote sensing technologies to measure these indicators is dependant on the resolution and the type of information recorded by the system. The high spectral and spatial resolutions of handheld spectroradiometers allow spectral indices to be derived from combinations of reflectance bands, which can then be used to predict the leaf cover and crown colour of individual trees (Barry, Stone, & Mohammed, 2008). These same indices can be upscaled to multiple trees over large areas when obtained from air- or space-borne sensors (Coops, Stone, Culvenor, Chisholm, & Merton, 2003; Goodwin, Coops, & Stone, 2005; Sampson, Zarco-Tejada, Mohammed, Miller, & Noland, 2003; Somers et al., 2010; Stone et al., 2000). 7  Given that LiDAR sensors record three-dimensional spatial coordinates instead of reflectance, they are not adapted to detecting spectral manifestations of tree stress. They can, however, measure the size and structure of an individual tree. A large number of studies have used airborne LiDAR scanners to estimate tree metrics such as stem height, crown width or diameter at breast height (Sorin C. Popescu, Wynne, & Nelson, 2003; Schardt, Ziegler, Wimmer, & Wack, 2002; Suárez, Ontiveros, Smith, & Snape, 2005; Yu, Hyyppä, Vastaranta, Holopainen, & Viitala, 2011). Although the majority of these studies have focused on forest inventorying as opposed to condition assessment, Leckie et al. (2003) noted the potential of high-density LiDAR data for measuring health attributes such as crown closure, crown size or canopy gaps. Kato et al. (2009) reconstructed three-dimensional crown models from a LiDAR point cloud using a surface wrapping technique, from which crown width, crown volume and live crown base were derived. Most research into remotely sensed tree condition assessments have focused on natural or managed forests. In contrast, urban applications are relatively scarce, despite the considerable demand for comprehensive information on trees in urban areas. Notably, several existing studies have attempted to estimate composite indices of tree condition from the analysis of airborne imagery and photographs (Malthus & Younger, 2000). These indices can be obtained from a principal component analysis of various field-measured condition indicators, such as needle loss, foliage colour, number of dead limbs, crown metrics, and trunk and foliage conditions (Nowak & McBride, 1991).  8  Table 1-1. Summary of studies using remote sensing technology to assess indicators of tree condition. Study Urban/ exurban Remote sensing technology Measured variables/analysis technique Condition indicators Nowak & McBride (1991) Urban  Color infrared aerial photographs Microdensitometric analysis Composite condition index derived from field measurements of:  Needle loss  Foliage color  Percent large dead limbs  Percent small dead limbs  Percent natural crown pruning  Dead crown ratio  Crown shape  Foliage condition  Trunk condition Malthus & Younger (2000) Urban  Airborne multispectral imagery (1.0 x 1.0 m)  Red edge position (REP)  Green normalized difference vegetation Index (gNDVI) Composite condition indices derived from field measurements of:  Leaf colour  Crown density  Crown die-back  Leaf size  Number of small dead limbs  Foliage colour  Leaf chlorosis  Leaf necrosis Goodwin et al. (2005) Exurban  Airborne multispectral imagery (0.5 x 0.5 cm) Spectral endmembers for:  Sunlit canopy  Soil  Shadow  Crown transparency  Crown colour  Crown volume Blackburn (2007) Exurban  Image-replicating imaging spectrometer Wavelet decomposition analysis  Pigment concentrations 9  Study Urban/ exurban Remote sensing technology Measured variables/analysis technique Condition indicators Barry et al. (2008) Exurban  Hand-held spectroradiometer  Lower red edge slope (REls)  Total red edge slope (RET)  Red edge position (REP)  Modified chlorophyll absorption ratio index 2 (MCARI2)  Modified triangular vegetation index 2 (MTVI2)  Red-green index (RGI)  Anthocyanin reflectance index (ARI)  Crown colour  Leaf cover Kato et al. (2009) Urban  Airborne LiDAR (9 returns/m2) Three-dimensional tree crown reconstruction  Crown width  Live crown base  Crown volume Somers et al. (2010) Exurban  Hyperion satellite hyperspectral sensor (30 m x 30 m)  Landsat TM satellite multispectral sensor (30 m x 30 m) Spectral mixture analysis  Defoliation Oshio et al. (2013) Urban  Airborne LiDAR (unspecified density) Voxel structure analysis  Foliage distribution  1.3.2. Remote sensing of impervious surfaces As one of the most defining characteristics of urban areas, impervious surfaces have been the focus of substantial research (Arnfield, 2003; Kjelgren & Montague, 1998; Mueller & Day, 2005; Pouyat et al., 1995). The scale of these studies range from a single parcel of land (Morgenroth & Buchan, 10  2009) to large urban transects (Weng, Hu, & Lu, 2008). Consequently, maps of impervious areas have become valuable sources of information. Previous research has integrated remotely sensed and other forms of geospatial data to derive these maps at the citywide level (Weng, 2012). Analysis techniques are generally selected according to the spatial resolution of the chosen sensor; these range from sub-pixel analysis for medium- to coarse-scale imagery (Weng et al., 2008) to object-oriented algorithms for high-resolution sensors (Hu & Weng, 2011; Miller, Nelson, & Hess, 2009). A frequently used approach for high-resolution imagery is image classification, by which pixels are classified into discrete land cover classes from which imperviousness can be assessed (Weng, 2012). While land cover classification is most often applied to multispectral imagery, certain classifiers are able to incorporate different data types from multiple sources. For instance, spectral information can be used in conjunction with airborne LiDAR data to produce accurate classifications of urban areas (Huang & Zhu, 2013) and map imperviousness in urban land parcels (Hodgson, Jensen, Tullis, Riordan, & Archer, 2003).  1.4. Research objectives The overall research objective of this thesis was to explore the use of airborne remote sensing in the assessment and the analysis of urban tree condition. Using high-resolution LiDAR data and hyperspectral imagery, the condition of urban trees was assessed in the city of Surrey, British Columbia, Canada, followed by an examination of the relationship between condition and impervious surfaces. The anticipated outcomes of this thesis were the development of new methodologies for LiDAR in urban forestry applications, the complementing of existing tree inventories with information on condition, and informing city managers and planners on the effects of imperviousness on urban trees. To meet these goals, three sub-objectives were outlined: 11  1. Develop a method for obtaining precise tree locations using LiDAR data and linking those locations to an existing geographic information system (GIS) maintained by the city. 2. Test the viability and the effectiveness of measuring two indicators of tree condition using LiDAR and ancillary geospatial data: crown density and height relative to the age and species of a tree. 3. Investigate the relationship between tree condition and the degree of landscape imperviousness at multiple spatial scales. Chapter 2 will introduce the city of Surrey, its geography, its climate and its urban characteristics. The geodatabase containing the tree inventory of the City will be presented, and technical details on the LiDAR data and hyperspectral imagery used in the thesis will be provided. Chapter 3 will focus on the first two research objectives. A method for locating trees using LiDAR and GIS data will be developed. The capacity for using LiDAR point cloud metrics to estimate field-measured crown density will be tested, and the potential for using height measurements as an indicator for tree condition by accounting for age and species will be explored. Chapter 4 will apply the methods developed in Chapter 3 to assess the condition of 1,914 trees across the city of Surrey. A land cover map for the city will be generated from the LiDAR and hyperspectral data, from which landscape imperviousness estimates will be derived. The relationship between imperviousness and tree condition will then be examined at multiple spatial scales. Finally, Chapter 5 will discuss key findings, conclusions and limitations of the research. Recommendations for future research will also be made.  12  2. Data & study site 2.1. Study site The city of Surrey is located in the Greater Vancouver regional district, in the Canadian province of British Columbia (Figure 2-1). It is one of the fastest growing cities in Canada, with an 18.6% increase in population between 2006 and 2011 (Statistics Canada, 2011). The area’s climate is typical of the Northwestern coast of North America:  cool, rainy winters and mild, sunny summers with average daily temperatures ranging between 3.4°C in December to 18.2°C in August. The landscape is characterized by gently rolling hills attaining a maximum elevation of 134 m, dissected by wide, flat-bottomed deltaic valleys. The area’s uplands are comprised of glacial deposits while its valleys consist of fluvial sediments and peat (Clague, Luternauer, & Hebda, 1983). Most of the lowlands within city boundaries are protected under British Columbia’s Agricultural Land Reserve, which restricts most urban development to the city’s less fertile hills and plateaus.  13   Figure 2-1. Location of the city of Surrey within the Greater Vancouver regional district. 2.2. Tree database The city of Surrey manages over 100,000 trees across its 316.4 km2 area, with 3,500 to 5,000 additional trees being planted every year. All trees directly managed by the city are planted on city property. The species, subspecies, planting date and approximate geographic coordinates of each tree are recorded within a comprehensive GIS database. Tree coordinates are either recorded by field crews using global positioning system (GPS) receivers, or, for trees predating the database’s creation, located through interpretation of aerial orthophotos. Due to inaccuracies inherent to mobile GPS units and the variable methods used for recording coordinates, certain entries contain 14  positional errors, while others may be out-of-date or supply locations for trees that have been removed. 2.3. Remotely sensed data In April and May of 2013, airborne LiDAR data and hyperspectral imagery were acquired over the city of Surrey under leaf-off conditions by Airborne Imaging (Calgary, Alberta, Canada). Table 2-1 summarizes flight parameters for both types of data. Hyperspectral imagery was acquired using a CASI-1500 sensor. ITRES Research (Calgary, Alberta, Canada) applied radiometric corrections to convert raw data to values representing spectral radiance for each band. Geometric corrections were performed by georeferencing the image to the LiDAR data and a ground-based GPS system. Finally, atmospheric conditions and topographic and bi-directional effects were corrected using the ATCOR-4 procedure (Richter & Schlapfer, 2016). The final imagery consisted of 72 spectral bands of reflectance imagery covering the 363-1051 nm spectral range, with a spectral resolution of 9.6 nm and a spatial resolution of 1.0 m2 (Appendix 1). 15   Figure 2-2. Subset of four reflectance bands of a hyperspectral image of a suburban street in the city of Surrey. These bands roughly correspond to blue (473.3 nm), green (530.7 nm), red (693.0 nm) and near infrared (1008.1 nm)   16   Table 2-1. Flight parameters for LiDAR and hyperspectral imagery acquisition. Flight parameter LiDAR Hyperspectral imagery Flight height above ground level 1000 m 2100 m Flight line spacing 344 m 975 m Ground swath width 688 m 1500 m Overlap 50% 35%  LiDAR data was acquired using a Leica ALS70-HP discrete return system, with up to four discrete returns per pulse with wavelengths of 1064 nm. The pulse rate was 500 KHz, which resulted in an average point density of 25 points per square meter. Before being delivered by the contractor, all LiDAR returns were classified into land cover types (i.e.: ground, building or vegetation) using TerraScan software (TerraSolid Ltd., Helsinki). A 1.0 m2 rasterized digital elevation model (DEM) interpolated from classified ground returns was also supplied. 17   Figure 2-3. Orthographic photo of a suburban street in the city of Surrey and a LiDAR point cloud of the same area. LiDAR returns are coloured according to elevation.   18   3. Detecting and assessing the condition of urban trees using airborne LiDAR 3.1. Introduction 3.1.1. Surveying urban tree condition Urban trees represent a considerable financial investment on behalf of city administrations, and require a significant commitment in terms of maintenance and care (McPherson, 1993). To effectively allocate resources devoted to tree management, city authorities require information on the condition of urban forests (Banks et al., 1999). However, sampling plot schemes used to assess tree condition in exurban forests are poorly adapted to the urban landscape’s complex and heterogeneous environment, while thorough ground surveys are impractical due to staff and funding limitations (Roman et al., 2013). With its capacity to acquire quantitative and spatially explicit information over broad scales, remote sensing may provide cities with the means of efficiently and repeatedly performing exhaustive assessments of tree condition. Though previous research has used spectral imagery for this purpose, few studies have explored the potential of airborne LiDAR to measure meaningful indicators of tree condition (Table 1-1). 3.1.2. Trends in individual tree extraction from LiDAR data The ability of LiDAR to reliably acquire three-dimensional measurements of terrain has made it an appealing tool for forest resource management (Hudak, Evans, & Smith, 2009), particularly due to the capacity of pulses to penetrate tree canopies. Forest attributes are generally derived from airborne LiDAR data using either an area-based approach or through individual tree detection. In the area-based method, forest stand characteristics are estimated using predictors such as the 19  density and the vertical distribution of LiDAR returns from within the forest canopy (Næsset, 2002). With the increasingly widespread availability of high-density LiDAR data, however, many recent studies have instead focused on detecting, delineating and extracting metrics for individual trees. Automated tree detection algorithms can either be applied directly to a point cloud or to a rasterized canopy height model (CHM) derived from raw LiDAR data (Jakubowski, Li, Guo, & Kelly, 2013). Falkowski et al. (2006) used spatial wavelet analysis to extract location, height and crown diameters from a mixed conifer forest. Several methods rely on the integration of LiDAR and other sources of remotely-sensed data. Approaches such as valley following and object-oriented classification have been applied to coregistered LiDAR data and multispectral imagery to detect and isolate individual trees in coniferous forests (Leckie et al., 2003; Suárez et al., 2005). Kaartinen et al. (2012) compared a variety of approaches to tree detection, and found that several outperformed manual processing of LiDAR data in terms of accuracy.  While tree locations can be used to estimate the height and number of trees in an area, additional steps are required to extract information such as crown width or volume. Various methods have been developed to automatically outline the horizontal extent of tree crowns from LiDAR data in a process referred to as crown segmentation. Brandtberg et al. (2003) created a space-scale structure from a CHM using a Gaussian filter at multiple scales. Trees that were represented as “blobs” in this structure were detected and outlined. Kato et al. (2009) modelled the three-dimensional space occupied by individual trees using a process referred to as “wrapped surface reconstruction”. A k-means clustering algorithm was used by Morsdorf et al. (2002) to segment a pine forest canopy. Lee et al. (2010) applied a region growing procedure to a set of seed points drawn directly from the LiDAR point cloud. The resulting clusters were then agglomerated using an adaptive clustering technique. 20  3.1.3. Extracting trees in the urban landscape Though some recent efforts have been made to develop algorithms suitable for detecting and outlining trees in urban areas (C. Zhang, Zhou, & Qiu, 2015), most research on the use of automatic tree extraction techniques have focused on natural or semi-natural forest structures. Previous studies have shown optimal results in homogenous or near-homogenous forest stands (S.C. Popescu & Wynne, 2004; Schardt et al., 2002). These are not, however, the conditions generally found in urban landscapes. The city of Surrey, for instance, manages over two hundred species of trees on city property, including deciduous and coniferous taxa and domestic and imported varieties. In terms of age, these trees range from newly planted juveniles to decades-old mature trees. Considerable spatial variation exists in the city’s urban forest. Young street trees, which are generally spaced at regular intervals, often have large gaps between each other, while older street trees may form continuous canopies. Trees planted in parkland may be placed in dense clusters or in dispersed patterns. This broad range of ages, species and spatial configurations makes calibrating automated tree detection and segmentation routines difficult. 3.1.4. Chapter objectives The research presented in this chapter seeks to overcome the challenges posed by the urban forest’s complex structure by leveraging the city’s existing tree inventory. An automated method for locating and outlining urban trees using LiDAR and ancillary GIS data collected by the city is developed and tested. LiDAR’s capacity to estimate two dendrological metrics, tree height and crown density, will then be examined. Methods for using these metrics as indicators of urban tree condition are presented and discussed. Our analysis focuses on western redcedar trees (Thuja plicata) located on city property. Western redcedar is the official tree of the Canadian province of British Columbia, and possesses unique 21  cultural significance in the region. Furthermore, the species is of particular concern to city managers due to its preference for moist soils, and consequently its susceptibility to poor watering conditions (Stewart, 1984). 3.2. Methods 3.2.1. Field data A ground survey was undertaken of 169 western redcedar trees, all of which had corresponding entries in the city’s database. The focus of this chapter was on detecting trees whose growth is affected by environmental stressors as opposed to competition with neighboring trees, and so the sample was restricted to free-growing trees. The criteria for defining a free-growing tree was that its crown not be in contact with the crown of any trees in its surroundings. Trees planted in parks, parking lots and along roadsides were included. In March, 2015, an initial four sampling sites were visited. For logistical reasons, only sites with high numbers of western redcedar were considered. Accompanying a surge in urban development, a substantial portion of the city’s western redcedar were planted in the late 1990s and early 2000s, and consequently, the four initial sites contained mostly trees planted within this period. To ensure that the sample represented a sufficiently broad range of ages, three additional sampling sites were subsequently visited, containing trees under 15 years and over 25 years. The combined sample from the seven sites contained 169 free-growing western redcedars with ages ranging from 11 to 33 years old (Figure 3-1). Surrey’s urban trees are approximately seven years old when planted. Tree heights were measured using a LTI TruPulse 360 laser rangefinder. Crown density was measured using methods described in Schomaker et al. (2007). Crown densities ranged from 25% to 95%, while heights ranged from 2.1 m to 14.5 m. 22   Figure 3-1. Location of trees measured in the field within the city of Surrey. 3.2.2. Canopy Height Model A canopy height model (CHM) was produced using FUSION software (McGaughey, 2014). This product is derived from the raw LiDAR data by recording the difference in elevation between the 23  highest classified vegetation return and the underlying DEM within each cell of a continuous grid. The maximum above-ground height of vegetation is indicated by the pixel value of the CHM at any given location. The selection of an appropriate cell size is an important parameter: large cells reduce variation within the CHM, while small cells can create gaps within the canopy and increase the volume of data.  Chen, Baldocchi, Gong, & Kelly (2006) recommend calculating an optimal cell size based on the first-return density of the LiDAR data (Equation 3-1), where 𝜆 is returns/m2, and 𝑑 is cell size. 𝑑 = √1𝜆 Equation 3-1 Using a density of 25 returns/m2, this equation yielded a cell size of 0.2 m. In practice, however, return density is not uniform, and so a conservative cell size of 0.5 m was selected to account for areas where flight swaths did not overlap. By computing the return count in areas free of water, it was found that approximately 99.94% of cells contained at least one return when using this resolution.  3.2.3. Locating trees For information on tree condition to be extracted from the LiDAR data, the location and spatial extent of the trees of interest are required. This study employed the popular and well-researched variable window filter (VWF)  method for detecting trees (S.C. Popescu, Wynne, & Nelson, 2002; S.C. Popescu & Wynne, 2004). A VWF uses a moving focal window to scan the canopy for local canopy height maxima, which often correspond to treetops. When applied to a CHM, cells are tagged as treetops if their values are the highest within the window. Errors of commission occur when the window size is too small, causing branches, canopy protrusions, or other non-tree local maxima to 24  be incorrectly tagged as treetops. Conversely, an overly large window may pass over treetops that are closely clustered together, causing errors of omission. A VWF minimizes both types of error by adjusting the size of the window according to the height of the cell on which it is centered (Figure 3-2). This is based on the relationship between tree height and crown width: pixels with high values are likely to belong to tall trees with correspondingly wide crowns, and vice versa.  Figure 3-2. Variably-sized moving focal window: a) a moving focal window scans the canopy height model; b) window size varies according to the canopy height; c) cells are tagged as tree tops (red) if they have the highest value within the window. To parameterize the VWF, a formula relating tree height and crown width is required. Crown outlines for a randomly selected subset of 115 surveyed trees were manually delineated through visual interpretation of the CHM. Using a quadratic model to regress the average crown diameter of these reference outlines with their height produced Equation 3-2, which yielded an adjusted r2 value of 0.5958. The equation is then used to derive the diameter of the search window ?̂?𝑊𝑖𝑛𝑑𝑜𝑤 when centered over any given CHM cell with a height value of HCell.  ?̂?𝑊𝑖𝑛𝑑𝑜𝑤 =  0.69906 + 0.66511 × 𝐻𝐶𝑒𝑙𝑙 − 0.01052 × 𝐻𝐶𝑒𝑙𝑙2 Equation 3-2 While Surrey’s GIS database contains the approximate locations of all the trees maintained by the city, coordinates may be offset from the trees’ actual locations, and some trees may be missing 25  entirely. The following process (Figure 3-3) uses a VWF to complement the city’s GIS data and locate trees with increased precision.  A 20 m2 square search area, approximately the size of the largest possible tree crown for western redcedar, is centered on the coordinates of a tree as recorded in the database. To prevent the detection of false treetops from shrubs or other low-lying non-tree objects, all CHM pixels with a height value less than 1 m are masked out within the search area. Then, the VWF is applied to locate all potential treetops within the immediate vicinity of the tree.  The potential treetops identified by the VWF are then filtered according to height. Commonly used growth curves for coastal western redcedar developed by Kurucz (1985) provide estimates of the height of a tree over time for any given site index. Site indices, which are expressed as the height of a dominant tree in meters at age 50, measure the growing potential of a forested site, with an elevated site index indicating optimal growing conditions and vice versa (Clutter et al., 1983). Site indices of 40 and 20 were selected as upper and lower thresholds, which correspond to the range found within the field sample and are considered to represent the range of growing conditions found in the city. Using Site Tools software (BC Ministry of Forests, 2014), maximum and minimum values HTreetopMax and HTreetopMin were computed for each tree according to its age. Treetops outside of these thresholds were removed, and the remaining treetop closest to the coordinates of the tree was then selected. If no treetops within these thresholds are found, the tree is reported as missing. 26   Figure 3-3. Process for locating the treetop of an entry in the city’s urban tree database. The white dotted line represents the approximate coordinates of the tree as registered within the GIS. A variable window filter (VWF) is used to locate all potential treetops within a 20 m2 area centered on these coordinates. Using growth curve formulas, upper and lower height thresholds are computed according to the age of the tree. Finally, the closest treetop within these thresholds is selected. 3.2.4. Outlining tree crowns Once the coordinates of a given tree have been located, an outline of its crown’s horizontal extent is needed. An image processing technique known as watershed segmentation, originally developed for delineating drainage basins (Beucher & Lantejoul, 1979), can be used to segment tree crowns due to the morphological similarity between canopies and topographical terrain models (Figure 3-4). To avoid over-segmentation, wherein branches or other protrusions in the canopy are assigned spurious segments, the process is constrained by the pre-defined locations of treetops as detected by the VWF (Chen et al., 2006; Schardt et al., 2002; J. Zhang, Sohn, & Brédif, 2014). 27   Figure 3-4. Watershed segmentation: A) a CHM of two neighboring trees; B) the CHM is inverted so both trees resemble basins; C) the basins are filled with water from the bottom up; D) borders between the two tree crowns are drawn where the basins connect. A second automated outlining approach, wherein the crown’s horizontal size and shape are approximated using a circle, was also tested. The average crown width of each tree is computed in a process described by McGaughey (2014). This approach determines the crown radius along 16 equally spaced radial profiles centered on a the apex of a tree. Along each profile, the radius is given by the distance between the apex and the first CHM pixel that is either A) a local minima within the profile or B) less than 50% of the height of the tree. A circular outline is then generated using this average of these radii. Both automated outlining approaches were compared. Brandtberg et al. (2003) developed a fuzzy method for assessing the similarity between two corresponding tree crown outlines. This approach, which uses spatial weighting to assign greater importance to the center of an outline than its periphery, produces a segmentation assessment value (A) between two outlines, with an A = 1 indicating a perfect match, and any other case an A < 1. The ratio between the areas of automated outlines and their corresponding reference outlines (Q) was also computed. 3.2.5. Estimating crown density In a technique similar to the use of a cookie-cutter, LiDAR point cloud subsets corresponding to individual trees were created according to the boundaries of the crown outlines. The vertical measurements of the points in these subset were then normalized according to the DEM.  28  Two metrics were tested as potential predictors of crown density. Previous attempts at estimating the foliar density of forests at an area-based level have used the ratio of LiDAR returns below a certain height threshold over the total number of returns (Lovell, Jupp, Culvenor, & Coops, 2003). The reasoning behind this metric is that denser canopies are more likely to intercept LiDAR pulses, preventing a larger portion of them from reaching the ground. Here, the percentage of returns with a height above ground higher than 0.5 m within the normalized LiDAR point cloud subset of each tree will be referred to as “percentage of non-ground returns”. Other studies have used the coefficient of variation (CV) of the height of LiDAR returns to estimate forest attributes (Næsset, 2002).  A large CV indicates a wide dispersion of LiDAR returns along the vertical profile, suggesting increased penetration of the canopy. While this may hold true for an area of continuous forest, certain considerations should be made for an individual tree. For instance, in an open area where the laser path is unobstructed by neighboring objects, the conical shape of the western redcedar may intercept LiDAR pulses evenly along the vertical profile, regardless of the crown’s density (Figure 3-5). To correct for this, the size of the outline used to subset the LiDAR points was halved, so that a large portion of the returns from the crown’s exterior are excluded. The coefficient of variation of the height of the returns within this subset will be referred to as “CV of return heights”. 29   Figure 3-5. Rationale for using a resized crown outline when estimating crown density with the coefficient of variation of LiDAR point heights. Differentiating the vertical distribution of points between a sparse and a dense crown may be difficult when using a subset that extends to the edges of the tree (gray area). When the subset is limited to the apex of the tree and the interior of its crown (crosshatched area), a large amount of the returns being reflected from the exterior of the tree are excluded. Most of the remaining returns from a dense tree will occur near its apex, while a sparse tree, having allowed more returns to reach the interior of its crown, will have a more even distribution of return heights along its vertical profile. To remove the potential confounding effects of tree height, the reference trees were separated into 3 m equal interval height classes spanning the range of tree heights as measured in the field (Table 3-1). Due to the small number of trees in the highest class (14 trees between 11 m and 14 m) it was merged with the preceding class to ensure that all classes had a minimum of 30 trees. Table 3-1. Number of reference trees per height class. Height class interval Number of trees [2 m, 5m) 34 [5 m, 8 m) 76 [8 m, 14 m) 59 30  3.2.6. Tree height as an indicator of condition The highest point within each LiDAR point subset was used as an estimate of the height of the tree. the height to be an informative measure of the condition of a tree, it needs to be compared to a reference of growth for that species. With the availability of planting dates from Surrey’s GIS database, it is possible to fit growth models to trees whose heights have been accurately measured using LiDAR. The residual of the height of a tree against this model can then be used as an indicator of its condition (Dobbertin, 2005). Trees with positive residuals are growing at an optimum rate, while trees with negative residuals may be experiencing some sort of stress that is impeding growth. Based on methods described in Fekedulegn et al. (1999), a Chapman-Richards model was fit to the field-measured trees using the minpack.lm package for R statistical software (Elzhov, Mullen, Spiess, & Bolker, 2015). Chapman-Richards models are commonly used to describe non-linear relationships between tree metrics (Liu & Li, 2003). The model’s general formula is given by Equation 3-3, where Hi is the height of a given tree i as measured using LiDAR, agei is its age according to the city’s database, εi is the model’s residual and 𝛽1, 𝛽2 and 𝛽3 are estimated coefficients: 𝐻𝑖 = 𝛽1 × (1 − 𝑒𝛽2×𝑎𝑔𝑒𝑖)𝛽3 + 𝜀𝑖 Equation 3-3 3.3. Results 3.3.1. Treetop location and tree height estimation Of the 169 reference trees measured in the field, the automated tree detection algorithm reported one as missing. Figure 3-6 shows the relationship between tree heights measured in the field and heights measured from the LiDAR data, which had a Pearson correlation coefficient of 31  r = 0.927, p < 0.001. Trees for which the difference between LiDAR- and field-based height measurements was more than three standard deviations from the mean were selected for visual inspection. The inspection revealed that these four outliers had been incorrectly identified by the algorithm and were removed. The remaining LiDAR heights were, on average, 0.72 m lower than those measured in the field, with a standard deviation of 1.09 m. Figure 3-7 shows the Chapman-Richards growth model fitted to the 169 field trees using their LiDAR-based height measurements.  Figure 3-6. Relationship between tree heights measured in the field and tree heights measured from the LiDAR data. Four outliers whose difference between field and LiDAR height was SD > 3 from the mean were removed.  32   Figure 3-7. Chapman-Richards height model fitted to the field trees. Many trees share the same planting year, resulting in clustering along the x-axis. This model is specific to the field-measured trees, and the significance of its residuals may only be considered relative to the other trees in the dataset. 3.3.2. Tree crown outlines On average, the ratio between the reference and the watershed segmentation outline areas (Q) was Q = 1.06, and Q = 1.01 for the circular outlines. Manually delineated reference outlines produced a segmentation assessment value (A) of A = 0.71 when compared to the watershed segmentation outlines and A = 0.63 when compared to the circular outlines. Figure 3-8 shows examples of various segmentation assessment values A. The standard deviation of A for watershed segmentation outlines was 0.21, with 19% of the outlines being nearly identical to their references (A > 0.9), and 23% of the outlines being substantially different (A < 0.5). In comparison, the standard deviation for circular outlines was 0.14, with no outlines at A > 0.9, but only 17% with A < 0.5. 33   Figure 3-8. Examples of automatically generated (dotted lines) and manually delineated (solid lines) tree crown outlines. A) An accurate watershed segmentation outline (A = 0.92) B) A watershed segmentation error of commission: a neighboring tree or shrub has been included in the outline (A = 0.42). C) An accurate circular outline (A = 0.88). D) An inaccurate circular outline caused by an off-center tree apex (A = 0.39). 34  3.3.3. Crown density estimation Figure 3-9 shows ordinary least-squares linear regression models with field-measured crown density as the response variable and the percentage of non-ground LiDAR returns and the CV of return heights as separate predictor variables. For each of the three height classes, CV of return heights was a better predictor based on r2 values. Both predictor variables performed best in the highest height class (8 m to 14 m) and poorest in the smallest height class (2 m to 5 m). With an r2 = 0.617, predictions were best for the 8 m to 14 m height class using the CV of return heights. In order to demonstrate how the estimates of tree condition could be applied over the urban landscape, Figure 3-10 shows a comparison between tree height residuals and crown density for 40 of the field-measured trees in a subset of the residential area of Surrey. The map shows some apparent clumping in the spatial arrangement of tree condition indicators. Investigation of the drivers of these spatial patterns is the logical subsequent analysis which will follow in future work. 35   Figure 3-9. Ordinary least-squares linear regression models of field-measured crown density versus LiDAR metrics across multiple age classes. Left column: percentage of non-ground LiDAR returns. Right column: coefficient of variation of LiDAR heights. Based on r2 values, the coefficient of variation yielded the strongest relationships with crown density. Predictive power was highest for trees over 8 m. As tree height is lowered, the significance of relationships is diminished. 36   Figure 3-10. A comparison of tree condition indicators for 40 field trees around Kildare Drive, in Surrey, British Columbia, Canada. Gray points represent trees that are not western redcedars. A) Tree height model residuals for western redcedar. B) Estimated crown density. 3.4. Discussion 3.4.1. Accuracy of treetop location algorithm For an automated algorithm to accurately estimate crown density from LiDAR data at an individual tree basis, the precise location of that tree is required. It should not be expected that city GIS databases, when available, will contain tree coordinates that meet this level of precision. Therefore, it is important to incorporate steps that will account for this potential inaccuracy, both by correcting positional errors and by skipping trees that may have been removed or incorrectly recorded.  Here, a VWF was applied to identify all potential treetops in the vicinity of the coordinates of a tree entry in the city’s GIS database. When used in forestry research, tree extraction routines such as the VWF are generally applied over large areas using a single set of parameters. These parameters are 37  calibrated according to the forest’s species and age composition. This task is challenging in urban forests, however, whose highly variable structure make parameterizing automated algorithms difficult. Surrey’s existing GIS data can be used to address this issue. Although we focus here on a single species (Thuja plicata), species-specific equations for deriving search window sizes (such as Equation 3-2) can be developed and applied when extending this approach to other types of trees. Once a series of potential treetops have been identified, the treetop matching the tree entry must be selected. Using growth curves to compute a range of likely heights based on the age of a tree, and then using this range to filter potential treetops identified by the VWF was found to be a reliable way of locating trees within the LiDAR data and avoiding the incorrect selection of neighboring trees or false treetops such as buildings or power lines. The efficacy of this method was demonstrated by the highly significant correlation between height estimates obtained from the LiDAR data and heights measured in the field (r = 0.927, p < 0.001). 3.4.2. Performance of crown outline methods In comparison with previous methodologically similar studies (Brandtberg et al., 2003), both automated approaches for generating tree crown outlines produced acceptable results. The average ratio between the area of the circular outlines and their corresponding references (Q = 1.02) suggests this is an accurate method for measuring crown diameter. This approach produced a lower average segmentation assessment value (A = 0.63) than the watershed segmentation algorithm (A = 0.71), which is likely due to a mismatch between the circles and the sinuated contour of the tree’s fringe. It should be noted that western redcedars generally have a circular shape, and that the quality of a circular outline will likely be reduced for trees with irregularly shaped crowns, and particularly those with off-center apexes.  38  The standard deviation (SD) of A was highest for watershed segmentation outlines (SD = 0.21) which indicates a wider variation in the quality of the outlines. This is explained by this method’s reliance on accurate prior treetop detection. When all treetops are successfully detected within the vicinity of the targeted tree, watershed segmentation has the potential to produce highly accurate outlines, as demonstrated by the 19% of outlines with an A > 0.9. Conversely, watershed segmentation can oversegment a single tree when spurious treetops are mistakenly identified within its crown, or include multiple tree crowns in a single outline when treetops are missed. Although the watershed segmentation algorithm outperformed the circular outlines with respect to the average A, its higher incidence of unreliable outlines (23% of watershed segmentation outlines with A < 0.5, compared to 17% for circular outlines), has important implications when attempting to extract LiDAR metrics. For instance, an outline that erroneously includes neighboring trees will result in a highly inaccurate LiDAR-based estimate of crown density. Furthermore, Figure 3-5 illustrates the rationale for using only the horizontal center of the crown of a tree for density estimation. For this reason, the circular outlines were used to estimate crown density in lieu of those generated using watershed segmentation. 3.4.3. Potential use of LiDAR metrics for assessing urban tree condition While previous studies have used the ratio of LiDAR returns intercepted over a certain height (usually between 0.5 m to 2 m) to measure foliar density over large areas, this method does not appear to be applicable at the individual tree level. The percentage of returns above 0.5 m was a poor predictor of crown density for all tree height classes. Potential reasons include varying live crown ratios (the ratio of the vertical length of the crown to the full height of the tree) or interference from surrounding objects. Figure 3-11 illustrates how neighboring trees or buildings could potentially affect the number of returns to reach the ground under a tree. 39   Figure 3-11. Off-nadir LiDAR pulses may have unobstructed paths to the ground below an isolated tree in an open area (left). These pulses can be intercepted by neighbouring objects, reducing the number of returns to reach the ground below a tree in clustered surroundings (right). The coefficient of variation of LiDAR return heights is a stronger predictor of density, particularly for trees in the 8 m to 14 m height class (r2 = 0.617). Taller trees intercept a greater number of LiDAR returns, which may reduce variability and improve the relationship between the two variables. According to estimates made by applying the treetop detection algorithm citywide, approximately 69% of Surrey’s urban western redcedars are taller than 8 m, suggesting that the coefficient of variation of return heights may be a useful measure of crown density for a large portion of Surrey’s urban trees. It should be noted, however, that the coefficient of variation is influenced by the number of LiDAR returns per tree and so if this metrics is applied in future work, transformations that account for variable LiDAR return density levels should be applied. Previous studies have established LiDAR as a reliable tool for measuring individual tree heights (Hyyppä & Inkinen, 1999). LiDAR measurements tend to underestimate tree height, however, as pulses have little chance of hitting the apex of a tree. Tree growth occurring between LiDAR and field measurements may have also contributed to this effect.  These underestimations are generally consistent, as they are here by an average of 0.72 m, which allows them to potentially be corrected 40  using empirically-derived offsets (Schardt et al., 2002). The correlation between tree heights estimated by LiDAR and heights measured in the field demonstrate both the functionality of the treetop detection method and the accuracy of LiDAR height estimates. By extension, these accurate height estimates support this approach’s potential for assessing urban tree condition through residuals derived from fitted growth models (Figure 3-6).  It should be noted that these two indicators of tree condition show no obvious correlation (Figure 3-10). This demonstrates the multi-faceted nature of tree condition: the numerous manifestations of tree stress (leaf discoloration, variations in height and crown structure, leaf loss, etc.), are reflective of the wide range of environmental factors that can affect trees (Dobbertin, 2005). Few indicators of tree condition are associated with specific environmental factors, and so multiple indicators are generally combined into composite tree condition indices (Nowak & McBride, 1991; Zarnoch et al., 2004). Not all indicators are correlative, however, and so multiple indices are often generated when large numbers of indicators are recorded (Malthus & Younger, 2000). Finally, there exists potential for further data integration. Here, we demonstrate a method for accurately detecting individual trees and associating them to tree entries in the city’s GIS database. Once the exact tree location has been determined using LiDAR, coregistered imagery can be used to extract spectral information which, in combination with structural information gathered from LiDAR, can be used to produce composite indices of overall tree condition.     3.4.4. Use of a raster-based canopy height model As the availability of high-density LiDAR datasets rises and processing techniques develop, an increasing number of studies have attempted to perform both treetop detection and crown delineation directly on LiDAR point clouds, as opposed to rasterized surfaces such as a CHM (Li, Guo, Jakubowski, & Kelly, 2012; C. Zhang & Qiu, 2012).  A common motive for this approach are the 41  spatial errors that can be introduced when interpolating a point cloud to a grid (S. Smith, Holland, & Longley, 2004). Here, complex interpolation techniques were eschewed for a descriptive procedure by which CHM values were equal to the height above ground of the highest classified vegetation point within each cell. While the high point density of the dataset minimized concerns over artificial gaps in the CHM, this approach may still affect the results of the analysis. For instance, the spatial coordinates of the tree apexes are rounded to the nearest cell center, and the extent of some tree crowns may appear exaggerated. Tree height estimates derived from the CHM, however, will be identical to those taken from the raw LiDAR cloud, while density estimates, themselves calculated from the directly from the point cloud, are unaffected.   42  4. Examining the relationship between urban tree condition and landscape imperviousness 4.1. Introduction 4.1.1. Environmental challenges to urban tree development The urban environment presents a multitude of challenges to the development of urban trees. Anthropogenic impacts on the landscape create distinct climatic conditions, hydrological systems, soil compositions and levels of air- and soil-borne pollutants (Pickett et al., 2011). From a physiographic perspective, urban landscapes are perhaps best known for their extensive impervious surfaces: artificial structures such as roads, parking lots and buildings that prevent water infiltration (Arnold Jr. & Gibbons, 1996). Though the effects of imperviousness on urban trees have been the focus of several studies, these have been conducted using relatively small numbers of plots or sample trees (Kjelgren & Montague, 1998; Pouyat et al., 1995; Viswanathan, Volder, Watson, & Aitkenhead-Peterson, 2011). As previously discussed, remote sensing offers the possibility of acquiring information on a wide variety of tree characteristics. Chapter 3 explored methods for detecting and assessing urban tree condition using airborne LiDAR and ancillary GIS data. Though the application of these methods may produce results that are useful in their own right, they also open avenues for spatially explicit analyses of potential drivers of urban tree condition, such as landscape imperviousness. 4.1.2. Multi-scale analysis While remote sensing provides the benefit of acquiring large quantities of data over wide areas, it also facilitates the analysis of patterns and processes at multiple scales. Scale is a fundamental issue in the study of a wide range of disciplines, including urban ecology, where processes may occur at 43  multiple levels of organizational complexity (Niemelä, 1999).  For instance, tree health and development indicators can be considered at scales ranging from the microscopic (i.e.: mycorrhizal fungi formation, chlorophyll concentrations, etc.), to the ecosystem level (i.e.: species diversity, forest productivity, resilience to pests, etc.) (Kurth et al., 2015; Rapport, Costanza, & McMichael, 1998). The selection of a single level of analysis is often arbitrary, and can create mismatching results between studies (Wiens, 1989). While examining phenomena at broad scales often requires details to be simplified or aggregated, doing so without consideration for linkages between scales can lead to the loss of essential information (Levin, 1992). Therefore, multi-scale analysis is key to identifying variables of interest and understanding the functional relationships between different levels of organization. 4.1.3. Chapter objectives The research presented in this section uses methods developed in Chapter 3 to assess the condition of trees across the city of Surrey. The relationship between tree condition and landscape imperviousness is then examined. Height estimates for 1,914 city trees are derived from LiDAR measurements, while a fusion of LiDAR data and hyperspectral imagery is used to assess citywide levels of impervious land cover. By using municipal tree planting records to control for age and species, the variation in tree height explained by impervious land cover is investigated.  While previous research has revealed localized relationships between impervious surfaces and tree development, this chapter examines this phenomenon at multiple scales: local effects are considered at an individual tree level, while broad-scale effects are tested by aggregating individual trees to spatial units of varying sizes. The results of this chapter have implications on the understanding of urban environmental impacts on trees and can help city authorities integrate knowledge of landscape imperviousness into the management of their urban tree stock. The 44  potential and limitations of using airborne LiDAR data and hyperspectral imagery to estimate landscape imperviousness are discussed, which can inform future broad-scale or multiscale urban ecology studies. 4.2. Methods 4.2.1. Trees of interest The five most commonly planted coniferous species in Surrey were selected (Table 4-1). Because deciduous trees were under leaf-off conditions at the time of acquisition, they were not included in the sample. As with the previous chapter, the goal was to examine the impact of imperviousness on urban tree development, as opposed to the well-documented effects of competition from neighboring trees. Therefore, only unclustered, free-growing trees were considered. Table 4-1 shows the number of trees per species that were retained for analysis, all of which have corresponding entries in the city’s GIS database.  Table 4-1. Number of trees of interest per species retained for analysis. Species name (latin) Species name (common) Number of free-growing trees Thuja plicata Western redcedar 1017 Pseudotsuga menziesii Douglas fir 583 Sequoiadendron giganteum Giant sequoia 161 Cupressus nootkatensis Yellow cedar 78 Picea abies Norway Spruce 75  4.2.2. Field measurements To adjust for any bias in LiDAR height estimations, sample trees for all five species were measured in the field. These trees were selected using a stratified random sampling process based on tree age 45  for Pseudotsuga menziesii, Sequoiadendron giganteum, Cupressus nootkatensis and Picea abies. Each species was divided into five-year strata, from which an equal number of trees were drawn for a total of 33 trees per species. The sample trees were measured in the field using a LTI TruPulse laser rangefinder. These trees were then combined with the sample of 169 Thuja plicata, selected and measured in Chapter 3. 4.2.3. LiDAR data pre-processing Using FUSION software (McGaughey, 2014), the raw LiDAR data was gridded to produce multiple rasterized data layers. A series of descriptive statistics were also computed using a 1.0 m2 grid, including the maximum, minimum, standard deviation, interquartile distance, kurtosis and skewedness of the height above ground of LiDAR returns within each cell. The percentage of returns higher than height break of 0.5 m above ground was also computed. 4.2.4. Treetop detection and height estimation Although the approximate locations for all trees of interest were available from the city’s GIS, this database has positional inaccuracies inherent to the methods used to record coordinates. The method presented in Section 3.2.3 was used to obtain accurate tree coordinates and height measurements. Growth curves developed by the BC Ministry of Forests for the five species of interest were used to parameterize the automated treetop detection algorithm (Bruce, 1981; Goudie, 1984; Kurucz, 1985). Due to the small chance of pulses hitting the apex of a tree, LiDAR measurements tend to underestimate tree height, and empirically-derived offsets are required to correct the data (Hyyppä & Inkinen, 1999; Schardt et al., 2002). Using field measurements, height estimates were modelled 46  using ordinary least squares regression, with the height of the CHM at the detected treetop’s location as the predictor variable.  As presented in Section 3.2.6, Chapman-Richards equations (Equation 3-3) were fit for each of the five tree species using the minpack.lm package for R statistical software (Elzhov et al., 2015). The resulting models describe the average height of free-growing trees at a given age for each species within the city. The residuals of these models will be used as indicators of tree condition for this chapter, with positive residuals representing trees that have grown to greater heights for their age, while more stunted trees are represented by negative residuals (Dobbertin, 2005). 4.2.5. Land cover classification and estimation of imperviousness To establish the degree of impervious land cover surrounding each tree, a land cover map was produced for the city of Surrey using a random forest classifier (Breiman, 2001). Random forest is a widely used ensemble classification technique that is well-suited for remote sensing applications due to its accuracy, processing speed, insensitivity to overtraining and its ability to handle high-dimensional and multi-source data (Belgiu & Drăguţ, 2016).  To train the classifier, training data were collected through a random sampling scheme, stratified by spectral reflectance. To ensure that training data would be representative of the landscape’s full spectral variability, an initial unsupervised k-means classification of a Landsat-7 scene (date: March 25th, 2012, path: 47, row: 26) of the city was performed. An even number of randomly selected points within each k-means class were generated, and then assigned to one of seven land cover classes (Table 4-2) based on visual interpretation of orthorectified aerial photos. This process was iterated until a minimum of 90 points for each land cover class were identified. A total of 1,823 points were identified through visual interpretation. Of these, three-quarters were randomly selected to train a random forest classifier and the remaining quarter to perform an independent 47  assessment of classification’s accuracy. Training and validation points were not visited in the field due to the fact that the spatial accuracy of mobile GPS receivers is lower than the spatial resolution of the classified data and that accessibility issues would prevent the verification of points on private land, within water bodies, on top of buildings or within densely wooded areas. Predictor variables for the classifier represented three types of landscape information: spectral, textural and vertical, all of which were derived from the hyperspectral imagery and LiDAR data (Huang & Zhu, 2013). Spectral information was supplied by the surface reflectance values of the hyperspectral imagery. Textural information was produced by using the extended morphological profile method (Benediktsson, Palmason, & Sveinsson, 2005), by which circular structuring elements with 2.0, 4.0, 6.0 and 8.0 m diameters were used to create morphological profiles of the first three principal components of the hyperspectral data. Finally, vertical information consisted of the aforementioned gridded descriptive statistics of LiDAR return heights. Because the maximum above ground height of LiDAR returns is an important discriminant between low-lying ground cover (pavement, grass, bare earth, etc.) and raised landscape features (buildings, trees, etc.), an additional filtering step was taken to remove the confounding effects of electrical power lines. A mask was created by selecting pixels for which the percentage of LiDAR returns above 0.5 m was less than 30.0%. Since land cover at these locations is generally flat except for sparse overhanging features such as power lines, lampposts and branches, the maximum above ground height of LiDAR returns for these pixels was lowered to ground level. Although random forest classifiers are capable of handling high-dimensional data, it is preferable to select variables that are uncorrelated (Millard & Richardson, 2013) and are capable of discriminating between classes of interest (Fauvel & Benediktsson, 2008). An initial reduction in dimensionality was performed by removing variables that were found to be highly correlated using 48  Pearson’s product-moment correlation coefficient. Then, a classification using the remaining uncorrelated variables was performed using the randomForest R package (Liaw & Wiener, 2002). Finally, estimates of variable importance from the random forest classifier were used to remove variables with low predictive power. The reduced set of predictor variables was used to produce a city-wide land cover map (Figure 4-1). Although random forest classifier algorithms provide a built-in accuracy measurement known as the “out-of-bag” (OOB) error, Millard & Richardson (2015) found that in certain cases the OOB error measurement inflated the accuracy of the classification. Therefore, an independent accuracy assessment was performed by withholding a quarter of the training points to be used for validation. Land cover classes were selected according to those used by Hodgson et al. (2003), who synthesized class-specific published runoff coefficients for estimating landscape imperviousness. These coefficients ranged from 10.0 to 100.0%, with higher percentages associated with impermeable classes, and lower percentages with more porous and absorptive categories. A map of imperviousness was derived, in which each cell was assigned the imperviousness coefficient associated with its land cover class (Figure 4-1). It should be noted that this map and its associated values are not equivalent or comparable to binary impervious maps used in other studies, in which cell values indicate either the presence or absence of highly impervious features such as pavement or buildings (Mallin, Williams, Esham, & Lowe, 2000).  49   Figure 4-1. A) Classified land cover map of the city of Surrey. B) Imperviousness map derived from land cover. C) Locations of all trees of interest. Note that some trees may not be visible due to overlapping points. D) Average height model residual per 0.5 km2 hexagonal cell.  50  4.2.6. Multi-scale analysis of relationships between imperviousness and tree height Two distinct scale-specific methodologies were used to analyze the relationship between tree height and landscape imperviousness. The first approach examined the relationship at the individual tree level by measuring the average degree of imperviousness within circular areas centered on each of the 1,914 sample trees. Since many of the documented impacts of impervious surfaces on trees are effectuated through changes in underlying soil conditions, these circular areas were made to be coterminous with potential root spread extents. Previous studies have shown that root spread is proportional to crown width, and so the sizes of the circular areas were calculated as a function of the crown diameter of each tree (Gilman, 1988; H. G. Smith, 1964). Ratios of circular area to crown diameter of 1.2, 2.0 and 3.0 were tested (Figure 4-2). Crown diameters were computed from the CHM using a process described and implemented in FUSION software by McGaughey (2014), wherein crown radius measurements are taken along 16 equally spaced radial profiles centered on the apex of the tree, and then averaged. Within each of these circular areas, impervious map pixels were subset, and pixels corresponding to the tree crown itself were removed. The mean value of the remaining pixels was assigned to each circular area as its average degree of imperviousness. To relate imperviousness to height model residuals, generalized least squares (GLS) models with a maximum-likelihood iterator were fit for each of the circular area size ratios using the nlme package in R (Pinheiro, Bates, DebRoy, & Sarkar, 2013). Significant spatial autocorrelation was revealed by graphing the GLS model residuals against their spatial coordinates and performing a Moran’s I test. Experimental variograms showed exponential correlation structures between the variance of the GLS model residuals and the distance between trees.  A second series of models were then fitted by incorporating these spatial correlation structures to account for autocorrelation, the ranges and nuggets of which are given in Table 4-4 (Zuur, Ieno, Walker, Savliev, & Smith, 2009). 51   Figure 4-2. Four sample trees used in the tree-level analysis representing a range of species, ages, height model residuals and levels of imperviousness. First column: orthographic photos. Second column: imperviousness measured within three circular areas of varying sizes centered on each tree (radii measuring 1.2, 2.0 and 3.0 times that of the crown). Dark pixels represent areas that are highly impervious (buildings, pavement), whereas light pixels are more pervious (grass, bare ground, shrubs, tree canopy). Pixels corresponding to the tree on which the areas are centered are excluded. Third column: tree information, including height model residuals and mean imperviousness within each circular area. 52  Due to the proximity of certain trees to each other, broadening the scale of a tree-by-tree analysis could result in overlapping areas. To avoid measuring imperviousness of the same area multiple times, a second approach was devised, referred to as the “broad-scale analysis”. The city’s extent was divided into a continuous hexagonal grid. Since hexagons have shorter perimeters than squares of equal area, hexagonal sampling grids have been recommended due to potential reductions in edge effects (Rempel & Kushneriuk, 2003). Cell sizes of 0.25 km2 (approximate size of an average city park), 0.5 km2 and 1.0 km2 (approximate size of a second-order block in Surrey’s road grid) were used. Each hexagonal cell’s average imperviousness was defined as the mean pixel value of the impervious map within the bounds of that cell (Figure 4-1). Trees within the cell are aggregated and their mean height model residual was recorded. An ordinary least-squares regression model was fit with hexagonal grid cells as the measurement unit, average imperviousness as the predictor variable and average height model residual as the response. To ensure that a sufficient number of trees were present to characterize each cell, a minimum number of trees per cell was applied: thresholds of 20 and 30 trees were tested. Highly influential data points were present in four of the six resulting models. Following investigation, the corresponding cells revealed no anomalies in measurement. However, given their potential for influencing results, regressions were calculated both with and without these cells. 4.3. Results 4.3.1. Land cover classification Overall accuracy, measured as the percentage of correctly classified reference points, was 88.6%, with a kappa value of 85.5%. Producer’s and user’s accuracies for each land cover class are reported in Table 4-2. A total of 72 bands of spectral imagery, 27 textural layers, and 7 LiDAR return height metrics were considered as predictor variables for the classification. Following the 53  process of dimensionality reduction, by which correlated and non-predictive variables were removed, only three spectral bands and four LiDAR return height metrics were used for the final classification. These were 9.6 nm-wide bands centered on wavelengths 463.7, 731.1 and 1008.1 nm, and the maximum height, standard deviation, skewedness, and percent of returns above ground of the LiDAR returns. Although the textural metrics were uncorrelated, they were found to have low predictive power and were discarded. Table 4-2. Number of training points, producer’s and user’s accuracies, and average imperviousness of each land cover class. Imperviousness percentages are referenced from (Hodgson & Bresnahan, 2004). Land cover class Number of training points Producer’s classification accuracy User’s classification accuracy Average imperviousness Coniferous trees 259 96.6% 86.2% 10% Deciduous trees 154 80.4% 97.4% 10% Grass 319 87.8% 89.0% 33% Bare earth 91 61.5% 76.2% 60% Paved 482 95.5% 81.7% 99% Building 353 96.7% 97.8% 99% Water 165 70.8% 91.9% 100%  4.3.2. Derivation of height model residuals LiDAR tree height estimates for all trees were on average 0.9 m lower than heights measured in the field. Species-specific average underestimations ranged from -0.5 m for Cupressus nootkatensis to -1.5 m for Picea abies. When modelled, LiDAR predicted field-measured height values with r2 coefficients ranging between 0.95 for Thuja plicata to 0.99 for Pseudotsuga menziesii. Residuals for all species-specific height estimation models were homoscedastic and normally distributed. 54  Table 4-3 lists coefficients for Chapman-Richards height models that were fit for each species. Residuals from these models were heteroscedastic, as variability in height increases with age. Statistics for height model residuals are reported in Table 4-3. Table 4-3. Coefficients for Chapman-Richards height models and model residual statistics. Species name (latin) Height model coefficients Height model residuals 𝜷𝟏 𝜷𝟐 𝜷𝟑 Minimum Maximum Std. dev. Thuja plicata 44.04345 -0.01235 1.19197 -8.12 12.59 2.29 Pseudotsuga menziesii 56.53141 -0.01587 1.14315 -8.65 11.61 2.25 Sequoiadendron giganteum 470.30136 -0.00038 0.80231 -2.95 7.54 1.82 Cupressus nootkatensis 278.79818 -0.00146 0.96824 -3.35 4.62 1.70 Picea abies 47.98686 -0.02356 1.82269 -5.71 8.37 2.31  4.3.3. Relationships between imperviousness and height model residuals For the individual tree scale, three generalized least squares (GLS) models were fit to assess the relationship between height model residuals and the imperviousness surrounding each tree, as averaged within circular areas of various diameters. Once spatial autocorrelation was accounted for, relationships were non-significant for all three circular area size ratios. The p-values of the models, both before and after inclusion of spatial autocorrelation structures, are given in Table 4-4 for each circular area size.   55  Table 4-4. Spatial correlation structures and p-values of GLS models relating height model residuals and imperviousness for three circular area size ratios. Ratio of circular area radius to tree crown radius Spatial correlation structure p-values Nugget Range Without spatial correlation structure With spatial correlation structure 1.2 0.108 32.972 0.108 32.972 2.0 0.102 31.903 0.102 31.903 3.0 0.103 32.120 <0.001 0.12  For the broad-scale analysis, the average height model residual was regressed against average imperviousness within hexagonal cells of varying sizes. After removing a highly influential point for 0.25 km2 cells, relationships were non-significant at the 0.05 level (Figure 4-3). At the 0.5 km2 cell scale, p-values were 0.025 and 0.079 (r2 of 0.292 and 0.336) when using a minimum tree threshold of 20 and 30 respectively, though when a highly influential point was removed, p-values were lowered to <0.001 and 0.002 (r2 of 0.570 and 0.753). For 1.0 km2 cells, p-values were 0.340 when using a minimum tree threshold of 20, and 0.084 when the threshold was set to 30. Significant relationships at the 0.5 km2 scale were negative, with regression coefficients ranging from -2.46 to -2.36, i.e.: every 1% increase in average imperviousness was associated with a decrease in average height model residuals of approximately -0.02 m. 56   Figure 4-3. Average height model residual versus average imperviousness for hexagonal grid cell sizes of 0.25, 0.5 and 1.0 km2, and minimum tree counts per cell of 20 and 30. Points with positive values on the y-axes have positive average height model residuals, i.e.: trees within these cells are on average higher for their age, and vice versa. Square points represent highly influential points. Models in which influential points have been removed are represented by dotted regression lines and italicized r2 and p-values. Significant relationships were found for 0.5 km2 units. 57  4.4. Discussion 4.4.1. Remote sensing of landscape imperviousness The “importance” metric generated by the random forest classifier can provide insight into the usefulness of remotely sensed variables for discriminating between land cover classes (Millard & Richardson, 2013). Here, four of the seven variables ranked as most important in the land cover classification were derived from the LiDAR data. This reflects the particular utility of LiDAR in urban areas, where several types of land cover are characterized by distinct vertical structures which cannot be captured by spectral imagery. Furthermore, LiDAR is not affected by shadows or variations in light, a common issue for high-resolution spectral imagery (MacFaden et al., 2012). The three remaining variables were hyperspectral bands, which can differentiate between land cover types with similar three-dimensional structures but with distinctive spectral signatures. It is notable that only three of the 72 available bands were required to achieve acceptable classification accuracy. This indicates a high degree of spectral separability between classes; classes with subtler spectral variation would likely require more bands (DeBacker, Kempeneers, Debruyn, & Scheunders, 2005). None of the textural metrics were rated as important for this classification. While previous studies have successfully used the extended morphological profiles to classify urban areas (Benediktsson et al., 2005; Huang & Zhu, 2013), the land cover classes used here may not be characterized by sufficiently distinct textural signatures for the technique to be applied effectively.  Having created and validated a random forest classifier, a map of imperviousness was then derived from a classified land cover map of the city. Some limitations to this approach should be acknowledged. While estimating imperviousness based on land cover-specific coefficients has precedent (Hodgson et al., 2003; Sleavin, Civco, Prisloe, & Gianotti, 2000), this approach fails to capture impervious variability within each land cover class. Furthermore, inaccurately classified 58  land cover pixels may propagate errors when used to derive the imperviousness map.  As noted by Hodgson et al. (2003), however, aggregating pixels to larger spatial units (as was done here) results in a higher accuracy than for the original unit of analysis.  4.4.2. Analysis of relationships between imperviousness and tree height Before interpreting the results of the broad-scale analysis, certain statistical considerations should be taken into account. First, aggregating trees and averaging imperviousness within gridded cells can introduce the well-documented modifiable areal unit problem (Jelinski & Wu, 1996). Here, the potential bias caused by arbitrary cell boundaries is addressed by repeating the analysis using cells of varying sizes. Because of the hexagonal shape of the cells, their boundaries at different sizes do not overlap. Second, the effects of influential points may be difficult to interpret. Investigation of the cells corresponding to these points did not reveal any errors in data collection, and several trees in each outlying cell were measured in the field. However, these cells are influential, as the significance levels and coefficients of determination of the models in which they were present changed considerably when they were removed. With these considerations in mind, we conclude that, after accounting for age, there exists a negative relationship between surface imperviousness and tree height at a broad scale in the city of Surrey for 0.5 km2 spatial units. Once influential points were removed, p-values were between <0.001 and 0.002 and r2 ranged from 0.570 to 0.753, depending on the minimum number of trees per cell (Figure 4-3). While height model residuals are negatively correlated with average imperviousness within 0.5 km2 spatial units, this relationship does not persist when imperviousness is measured within the immediate vicinity of individual trees. The variation of these imperviousness measurements reflect the range of urban tree planting conditions: trees with high values are found on roadsides, in 59  medians, or adjacent to buildings, while trees with low values are generally found in parks (Figure 4-2). After accounting for spatial autocorrelation, none of the three models fit for the individual tree level analysis yielded significant relationships. The substantial difference in p-values between models before and after incorporating spatial correlation structures underlines the importance of testing for autocorrelation in urban studies (Table 4-4).  To help explain the scale-dependent nature of this relationship, the mechanisms by which imperviousness affects tree development must be considered. As previously discussed, earlier studies examining the effects of impervious surfaces have focused on localized effects, such as soil compaction, uptake of mineral nutrients and microclimatic conditions (Jim, 1993; Kjelgren & Montague, 1998; Kozlowski, 1999; Mueller & Day, 2005; Viswanathan et al., 2011). The spatial scale of these effects is restricted to the direct vicinity of the trees, and so they would be captured by the individual tree analysis. The lack of correlation found here does not necessarily dispute the existence of these localized effects, but suggests that their importance as factors in explaining tree height variation is low within the city of Surrey. Having failed to detect a clear relationship at the individual tree level, it could be expected that similar results would be obtained at a broader scale. At the landscape level, however, the degree of surface imperviousness is a measure of the wider phenomenon of urbanization, for which the commonly used urban-rural gradient correlates with a wide range of environmental characteristics and processes (Hahs & McDonnell, 2006). For instance, tree communities in highly developed urban areas have been shown to have lower fungal and microarthropod populations, poorer leaf litter quality and increased levels of air and soil pollution than their rural counterparts (McDonnell et al., 1997). Given the lack of a corresponding localized relationship between imperviousness and height variation, it is likely that the broad-scale analysis has revealed the effects of these co-varying environmental factors. It should be noted, however, that the negative relationship found here is at 60  odds with the results of some previous studies. Some of the environmental factors found in highly urbanized areas have been shown to be beneficial to plant growth. For instance, urban forests receive higher levels of atmospheric deposition of inorganic nitrogen (a major plant nutrient) as well as faster litter decomposition and nitrification rates than areas at the rural end of the gradient (Lovett et al., 2000b; McDonnell et al., 1997). In a controlled experiment, Gregg et al. (2003) found that cottonwood clones planted in urban areas grew to twice the biomass of those planted in rural sites, an outcome to which lower urban ozone exposures was attributed. However, this study was performed on a broader scale (~100 km) than our analysis, where the maximum distance between two trees was approximately 20 km. Within the spatial scope of Surrey, our results suggest that the benefits of environmental factors associated with high degrees of imperviousness are outweighed by the drawbacks.   61  5. Conclusions 5.1. Key findings The overarching goal of this research was to assess and analyze urban tree condition using airborne remote sensing technology. In Chapter 3, using LiDAR and ancillary geospatial data, a method for acquiring and linking tree locations to an existing urban tree inventory was developed and tested. The capacity of LIDAR technology to assess two indicators of tree condition, height and crown density, were then investigated. Using the methods developed in Chapter 3, a condition assessment based on tree height and age was performed for trees across the city. In Chapter 4, the relationship between this indicator of tree condition and imperviousness was examined. 5.1.1. Using LiDAR and GIS data to estimate tree condition indicators Trees were located using a variably sized search window, which detected local maxima from a canopy height model derived from the LiDAR data. Using the age and species of tree entries in the city’s GIS database, corresponding treetops could then be identified and linked to the inventory. This method met the first research objective of this thesis: to locate urban trees using LiDAR and GIS data. With an accuracy of 97%, the approach proved to be highly successful, and allowed metrics derived from the remote sensing data to be associated with existing information contained within the city’s GIS. Accordingly, species-specific models describing the relationship between age and height were fit using the planting dates recorded in the GIS and the tree heights extracted from the LiDAR. The residuals of these models could then be used as indicators of tree condition: positive residuals represent a tree that is particularly tall for its age, and vice versa. In addition to height model residuals, LiDAR’s capacity to estimate crown density was also considered. Two LiDAR point cloud metrics were tested as potential predictors of field-measured 62  crown density. The percentage of non-ground returns, while conceptually sound, did not predict crown density reliably. The coefficient of variation of LiDAR returns proved to be a successful predictor for trees above 8 m (r2 = 0.617), although results were poorer for smaller trees. With regards to the second research objective, LiDAR and ancillary GIS data provided effective means for assessing tree height relative to age. Given the poor results for smaller trees, the viability of the method for estimating crown density presented here is limited. 5.1.2. Relationship between tree condition and imperviousness The third research objective was to investigate the relationship between tree condition and the degree of landscape imperviousness at multiple spatial scales. Landscape imperviousness estimates were derived from a classified land cover map. This map was produced by using LiDAR data and hyperspectral imagery as inputs for a random forest classification technique. The classification yielded an accuracy of 88.6%, using four LiDAR metrics and three hyperspectral bands as explanatory variables. The residuals of species-specific height models accounting for age were used as indicators of tree condition for 1,914 trees across the city. When analyzed at a broad-scale, a negative relationship between the degree of landscape imperviousness and average tree height was found for spatial units of 0.5 km2 in size (r2 between 0.292 and 0.753). However, this relationship did not persist at the scale of individual trees and their direct vicinities. 5.2. Implications for urban tree management This research demonstrates that high-density, discrete return airborne LiDAR is a promising tool for evaluating the condition of individual free-growing urban trees. Despite some limitations (see Section 5.3), there is potential for measuring crown density by analyzing individual tree point 63  clouds. Establishing a connection between LiDAR and existing GIS datasets (which record tree age and species) opens analytical opportunities such as using the residuals of growth models to assess tree condition. These indicators are important signs of a tree’s condition, and can be used to identify struggling urban trees. Comprehensive tree condition assessments can, in turn, allow city managers to efficiently organize resources and prioritize maintenance prescriptions, removals and replacements. When integrated with hyperspectral imagery, LiDAR provided several important explanatory variables for mapping land cover across the city of Surrey. Classified land cover maps are important data sources for urban planning, and can provide systematic estimates of landscape imperviousness at the citywide level. The results presented here underline the complex and multi-scaled relationship between urban tree development and imperviousness. Previous studies have established that trees respond to various soil and microclimatic conditions caused by impervious surfaces. Here, although a broad-scale statistical relationship between tree height model residuals and imperviousness was found at 0.5 km2, a corresponding relationship at the individual tree level was absent. Given that the previously documented effects of imperviousness occur within the tree’s immediate vicinity, the lack of an individual tree level relationship indicates that the overall impacts of these effects, as measured in this study, are minimal. The broad-scale relationship found here is likely due to other environmental factors associated with urbanization. These could potentially include variations in air pollution levels, fungal and microarthropod populations, leaf litter quality and decomposition rates, ozone exposure, atmospheric deposition, soil compaction or exposure to mechanical damage. Studying the relationship between manmade environments and tree condition is key to the complex task of managing urban forests. This research demonstrates how this endeavor is greatly 64  facilitated by modern remote sensing technology, which allows spatially comprehensive analyses to be performed at multiple scales. The resulting information is valuable to managers, urban planners, and all city authorities charged with promoting the vigor of urban trees.  The implications for urban tree management are summarized below:  LiDAR can accurately measure tree height, though offsets derived from empirical field measurements are required.  For height to be an informative metric of tree condition, tree age and species must also be taken into account. Therefore, data fusion with existing planting records is recommended.  Though potential exists, further research is required before LiDAR can produce accurate estimates of crown density for all tree sizes (see section 5.3).  LiDAR data and hyperspectral imagery are shown to be valuable for mapping urban land cover and estimating landscape imperviousness.  Within the city of Surrey, landscape imperviousness appears to have little overall impact on coniferous tree height, though its influence on other indicators of tree condition remains unmeasured.  Research into broad-scale environmental factors associated with imperviousness which may affect tree development is warranted. 5.3. Limitations Some limitations to this research should be acknowledged. Firstly, all analyses were restricted to free-growing trees. Trees in dense clusters present challenges for treetop detection and crown 65  delineation, while competition between trees can affect tree growth independently of environmental stressors. Furthermore, ancillary data such as tree age and species are required for tree height and crown density to be meaningful indicators of condition. The lack of availability of this data would significantly reduce LiDAR’s ability to produce information in this regard. Although an effort was made to include multiple species, the automated treetop detection approach was limited by the need for pre-existing growth curves for Surrey’s ecoregion (see Section 3.2.3). Applying this technique to exotic species without documented growth models may be problematic. Furthermore, the five selected species have natural habitats that extend from lowland valleys to mountainsides. Given that bottomland species tend to be most tolerant of compacted urban soils (Day, Seiler, & Persaud, 2000), it would be advisable that future studies take natural habitat into account when selecting species. To extrapolate results to a wider range of urban trees, deciduous, imported and exotic species should also be included. In Chapter 4, only one metric of tree condition (height model residuals) was used when analyzing relationships with imperviousness. As detailed in Section 1.3.1, the notion of tree condition is complex and multi-faceted, and can manifest through a wide variety of indicators. While it may be ideal to assess the development and condition of a tree by using multiple metrics, airborne sensors are often limited in their ability to measure them. As this study demonstrates, however, there is potential for assessing additional structural metrics such as crown density using LiDAR, while spectral metrics can be measured using multispectral imagery. Finally, the impervious map produced in this project gives no information regarding the surface directly beneath the crown of the tree. Although LiDAR pulses are capable of penetrating tree canopies, hyperspectral sensors can only capture the reflectance of the upper surface of the crown. Furthermore, while most city-managed trees are planted during the construction phase of new 66  roads or housing projects, in some cases, redevelopments may change the degree of surrounding imperviousness over a tree’s lifetime. Without data covering an extended period of time, these potential changes cannot be accounted for using the methodology presented here. 5.4. Recommendations for future research This research has underlined the need for more sophisticated techniques for measuring structural tree metrics using LiDAR data. Although the approach developed here was acceptable for trees over 8 m, methods for accurately assessing trees below this height class are required. This need is particularly acute for juvenile trees, which often require additional care to ensure their survival. Developing and testing new descriptors of three-dimensional LiDAR return distribution may offer new possibilities for measuring the structure and density of tree crowns remotely. With the urban ecosystem’s many complexities, multiple environmental factors can influence the development and condition of urban trees. Although this research does not suggest that imperviousness is a major determinant of tree condition in the city of Surrey, a relationship between the two variables was found at a broad scale. As previously discussed, this is likely due to unmeasured relationships with other factors associated with urbanization. Further research into the drivers of urban tree condition is warranted. Site conditions such as water availability, competition from other trees and solar irradiation should be considered in future studies. Additional tree species, potentially including deciduous and non-native species, should also be included. Although the remotely sensed measurement of several tree condition indicators remains unfeasible (i.e.: biochemical changes, phytohormone levels, etc.), previous research has demonstrated that air- and space-borne spectral sensors are capable of detecting macroscopic manifestations of tree stress, such as leaf yellowing and defoliation. By combining structural indicators measured using 67  LiDAR with spectral indicators extracted from imagery, composite indices of tree condition can be created. These indices may represent a more complete assessment of the many facets of the health and vitality of a tree, and could open new avenues for understanding the drivers of urban tree condition.   68  References Anderson, L. M., & Cordell, H. K. (1985). Residential property values improved by landscaping trees. Southern Journal of Applied Forestry, 9(3), 162–166. Arnfield, J. A. (2003). Two decades of urban climate research: A review of turbulence, exchanges of energy and water, and the urban heat island. International Journal of Climatology, 23(1), 1–26. doi:10.1002/joc.859 Arnold Jr., C. L., & Gibbons, C. J. (1996). Impervious surface coverage: the emergence of a key environmental indicator. Journal of the American Planning Association, 62(2), 243–258. Baines, C. (1994). Trenching and street trees. Arboricultural Journal, 18(3), 231–236. Baltsavias, E. P. (1999). Airborne laser scanning: basic relations and formulas. ISPRS Journal of Photogrammetry and Remote Sensing, 54(2), 199–214. Banks, J. C., Brack, C. L., & James, R. N. (1999). Modelling changes in dimensions , health status , and arboricultural implications for urban trees. Urban Ecosystems, 3, 35–43. Barry, K. M., Stone, C., & Mohammed, C. L. (2008). Crown-scale evaluation of spectral indices for defoliated and discoloured eucalypts. International Journal of Remote Sensing, 29(1), 47–69. doi:10.1080/01431160701281056 BC Ministry of Forests. (2014). Site Tools 4.0. North Vancouver, BC: BC Ministry of Forests, Lands and Natural Resource Operations. Belgiu, M., & Drăguţ, L. (2016). Random forest in remote sensing: A review of applications and future directions. ISPRS Journal of Photogrammetry and Remote Sensing, 114, 24–31. 69  doi:10.1016/j.isprsjprs.2016.01.011 Benediktsson, J. A., Palmason, J. A., & Sveinsson, J. R. (2005). Classification of hyperspectral data from urban areas based on extended morphological profiles. IEEE Transactions on Geoscience and Remote Sensing, 43(3), 480–491. doi:10.1109/TGRS.2004.842478 Beucher, S., & Lantejoul, C. (1979). Use of watersheds in contour detection. In International Workshop on Image Processing: Real-time Edge and Motion Detection/Estimation. Rennes, France. Blackburn, G. A. (2007). Wavelet decomposition of hyperspectral data: a novel approach to quantifying pigment concentrations in vegetation. International Journal of Remote Sensing, 28(12), 2831–2855. doi:10.1080/01431160600928625 Brack, C. L. (2006). Updating urban forest inventories: An example of the DISMUT model. Urban Forestry and Urban Greening, 5(4), 189–194. doi:10.1016/j.ufug.2006.09.001 Brandtberg, T., Warner, T. A., Landenberger, R. E., & McGraw, J. B. (2003). Detection and analysis of individual leaf-off tree crowns in small footprint, high sampling density lidar data from the eastern deciduous forest in North America. Remote Sensing of Environment, 85(3), 290–303. doi:10.1016/S0034-4257(03)00008-7 Breiman, L. (2001). Random forests. Machine Learning, 45(1), 5–32. Bruce, D. (1981). Consistent height-growth and growth-rate estimates for remeasured plots. Forest Sciencei, 27(4), 711–725. Campbell, J. B., & Wynne, R. H. (2011). Introduction to Remote Sensing (Fifth.). Guilford Press. 70  Chance, C. M., Coops, N. C., Crosby, K., & Aven, N. (2016). Spectral Wavelength Selection and Detection of Two Invasive Plant Species in an Urban Area. Canadian Journal of Remote Sensing, 40(1), 27–40. doi:10.1080/07038992.2016.1143330 Chen, Q., Baldocchi, D., Gong, P., & Kelly, M. (2006). Isolating Individual Trees in a Savanna Woodland Using Small Footprint Lidar Data. Photogrammetric Engineering & Remote Sensing, 72(8), 923–932. doi:10.14358/PERS.72.8.923 City of Toronto. (2008). Ahead of the Storm: Preparing Toronto for Climate Change. Toronto, Ontario. Clague, J. J., Luternauer, J. L., & Hebda, R. J. (1983). Sedimentary environments and postglacial history of the Fraser Delta and lower Fraser Valley, British Columbia. Canadian Journal of Earth Sciences, 20, 1314–1326. doi:10.1139/e83-116 Clutter, J. L., Fortson, J. C., Pienaar, L. V., Brister, G. H., & Bailey, R. L. (1983). Timber Management: A Quantitative Approach. New York, NY: John Wiley & Sons. Coops, N. C., Stone, C., Culvenor, D. S., Chisholm, L. A., & Merton, R. N. (2003). Chlorophyll content in eucalypt vegetation at the leaf and canopy scales as derived from high resolution spectral data. Tree Physiology, 23, 23–31. doi:10.1093/treephys/23.1.23 Cumming, A. B., Twardus, D. B., & Nowak, D. J. (2008). Urban forest health monitoring: Large scale assessments in the United States. Arboriculture and Urban Forestry, 34(6), 341–346. Day, S. D., Seiler, J. R., & Persaud, N. (2000). A comparison of root growth dynamics of silver maple and flowering dogwood in compacted soil at differing soil water contents. Tree Physiology, 20(4), 257–263. DeBacker, S., Kempeneers, P., Debruyn, W., & Scheunders, P. (2005). A Band Selection Technique for 71  Spectral Classification. IEEE Geoscience and Remote Sensing Letters, 2(3), 319–323. doi:10.1109/LGRS.2005.848511 Dobbertin, M. (2005). Tree growth as indicator of tree vitality and of tree reaction to environmental stress: A review. European Journal of Forest Research, 124(4), 319–333. doi:10.1007/s10342-005-0085-3 Eismann, M. T. (2012). Hyperspectral Remote Sensing. SPIE Press. Elzhov, T. V., Mullen, K. M., Spiess, A.-N., & Bolker, B. (2015). R interface to the Levenberg-Marquardt nonlinear least-squares algorithm found in MINPACK, plus support for bounds. R Package Version 1.1-8. Falkowski, M. J., Smith, A. M. S., Hudak, A. T., Gessler, P. E., & Vierling, L. A. (2006). Automated estimation of individual conifer tree height and crown diameter via two-dimensional spatial wavelet analysis of lidar data. Canadian Journal of Remote Sensing, 32(2), 153–161. Fauvel, M., & Benediktsson, J. (2008). Spectral and Spatial Classification of Hyperspectral Data Using SVMs and Morphological Profile. IEEE Transactions on Geoscience and Remote Sensing, 46(11), 3804–3814. Fekedulegn, D., Siurtain, M. P. Mac, & Colbert, J. J. (1999). Parameter Estimation of Nonlinear Growth Models in Forestry. Silva Fennica, 33(4), 327–336. Gilman, E. F. (1988). Tree Root Spread in Relation to Branch Dripline and Harvestable Root. HortScience, 23(2), 351–353. Goodwin, N., Coops, N. C., & Stone, C. (2005). Assessing plantation canopy condition from airborne imagery using spectral mixture analysis and fractional abundances. International Journal of 72  Applied Earth Observation and Geoinformation, 7(1), 11–28. Goudie, J. W. (1984). Height growth and site index curves for lodgepole pine and white spruce interim managed stand yield tables for lodgepole pine in British Columbia. Victoria, British Columbia. Gregg, J. W., Jones, C. G., & Dawson, T. E. (2003). Urbanization effects on tree growth in the vicinity of New York City. Nature, 424, 183–187. doi:10.1038/nature01776.1. Hahs, A. K., & McDonnell, M. J. (2006). Selecting independent measures to quantify Melbourne’s urban-rural gradient. Landscape and Urban Planning, 78(4), 435–448. doi:10.1016/j.landurbplan.2005.12.005 Herzog, T. R., & Strevey, S. J. (2008). Contact With Nature, Sense of Humor, and Psychological Well-Being. Environment and Behavior, 40(6), 747–776. doi:10.1177/0013916507308524 Hodgson, M. E., Jensen, J. R., Tullis, J. a, Riordan, K. D., & Archer, C. M. (2003). Synergistic Use of Lidar and Color Aerial Photography for Mapping Urban Parcel Imperviousness. Photogrammetric Engineering & Remote Sensing, 69(9), 973–980. Hu, X., & Weng, Q. (2011). Impervious surface area extraction from IKONOS imagery using an object-based fuzzy method. Geocarto International, 26(1), 3–20. doi:10.1080/10106049.2010.535616 Huang, R., & Zhu, J. (2013). Using Random Forest To Integrate Lidar Data and Hyperspectral Imagery for Land Cover Classification. In Geoscience and Remote Sensing Symposium (IGARSS) (pp. 3978–3981). Hudak, A. T., Evans, J. S., & Smith, A. M. S. (2009). LiDAR Utility for Natural Resource Managers. Remote Sensing. doi:10.3390/rs1040934 73  Hyyppä, J., & Inkinen, M. (1999). Detecting and estimating attributes for single trees using laser scanner. The Photogrammetric Journal of Finland, 16(2). Jakubowski, M., Li, W., Guo, Q., & Kelly, M. (2013). Delineating Individual Trees from Lidar Data: A Comparison of Vector- and Raster-based Segmentation Approaches. Remote Sensing, 5(9), 4163–4186. doi:10.3390/rs5094163 Jelinski, D. E., & Wu, J. (1996). The modifiable areal unit problem and implications for landscape ecology. Landscape Ecology, 11(3), 129–140. doi:10.1007/BF02447512 Jim, C. Y. (1993). Soil compaction as a constraint to tree growth in tropical & subtropical urban habitats. Environmental Conservation, 20(1), 35–49. Jim, C. Y. (1998). Physical and chemical properties of a Hong Kong roadside soil in relation to urban tree growth. Urban Ecosystems, 2, 171–181. Kaartinen, H., Hyyppä, J., Yu, X., Vastaranta, M., Hyyppä, H., Kukko, A., … Wu, J.-C. (2012). An International Comparison of Individual Tree Detection and Extraction Using Airborne Laser Scanning. Remote Sensing, 4(12), 950–974. doi:10.3390/rs4040950 Kato, A., Moskal, L. M., Schiess, P., Swanson, M. E., Calhoun, D., & Stuetzle, W. (2009). Capturing tree crown formation through implicit surface reconstruction using airborne lidar data. Remote Sensing of Environment, 113(6), 1148–1162. doi:10.1016/j.rse.2009.02.010 Kjelgren, R., & Montague, T. (1998). Urban tree transpiration over turf and asphalt surfaces. Atmospheric Environment, 32(1), 35–41. doi:10.1016/S1352-2310(97)00177-5 Kolb, T. E., Wagner, M. R., & Covington, W. W. (1994). Concepts of forest health: Utilitarian and ecosystem perspectives. Journal of Forestry, 92, 10–15. 74  Konijnendijk, C. C., Ricard, R. M., Kenney, A., & Randrup, T. B. (2006). Defining urban forestry – A comparative perspective of North America and Europe. Urban Forestry & Urban Greening, 4(3-4), 93–103. doi:10.1016/j.ufug.2005.11.003 Kozlowski, T. T. (1999). Soil Compaction and Growth of Woody Plants. Scandinavian Journal of Forest Research, 14(6), 596–619. doi:10.1080/02827589950154087 Kramer, H. J. (2002). Observation of the Earth and Its Environment (4th ed.). Springer-Verlag Berlin Heidelberg. doi:10.1007/978-3-642-56294-5 Kurth, F., Feldhahn, L., Bönn, M., Herrmann, S., Buscot, F., & Tarkka, M. T. (2015). Large scale transcriptome analysis reveals interplay between development of forest trees and a beneficial mycorrhiza helper bacterium. BMC Genomics, 16, 658. doi:10.1186/s12864-015-1856-y Kurucz, J. F. (1985). Metric SI tables for red cedar stands. Nanaimo, BC. Leckie, D. G., Gougeon, F., Hill, D., Quinn, R., Armstrong, L., & Shreenan, R. (2003). Combined high-density lidar and multispectral imagery for individual tree crown analysis. Canadian Journal of Remote Sensing, 29(5), 633–649. Lee, H., Slatton, K. C., Roth, B. E., & Cropper, W. P. (2010). Adaptive clustering of airborne LiDAR data to segment individual tree crowns in managed pine forests. International Journal of Remote Sensing, 31(1), 117–139. doi:10.1080/01431160902882561 Levin, S. A. (1992). The Problem of Pattern and Scale in Ecology: The Robert H. MacArthur Award Lecture. Ecology, 73(6), 1943–1967. Li, W., Guo, Q., Jakubowski, M. K., & Kelly, M. (2012). A New Method for Segmenting Individual Trees from the Lidar Point Cloud. Photogrammetric Engineering & Remote Sensing, 78(1), 75–84. 75  doi:10.14358/PERS.78.1.75 Liaw, A., & Wiener, M. (2002). Classification and Regression by randomForest. R News, 2(3), 18–22. Liu, Z., & Li, F. (2003). The generalized Chapman-Richards function and applications to tree and stand growth. Journal of Forestry Research, 14(1), 19–26. Locke, D. H., Grove, J. M., Lu, J. W. T., Troy, A., Neil-dunne, J. P. M. O., & Beck, B. D. (2010). Prioritizing Preferable Locations for Increasing Urban Tree Canopy in New York City. Cities and the Environment, 3(1), 1–18. Lovell, J., Jupp, D. L., Culvenor, D. S., & Coops, N. C. (2003). Using airborne and ground-based ranging lidar to measure canopy structure in Australian forests. Canadian Journal of Remote Sensing, 29(5), 607–622. Lovett, G. M., Traynow, M. M., Pouyat, R. V., Carreiro, M. M., Zhu, W.-X., & Baxter, J. W. (2000a). Atmospheric deposition to oak forests along an urban-rural gradient. Environmental Science & Technology, 34(20), 4294–4300. doi:10.1021/es001077q Lovett, G. M., Traynow, M. M., Pouyat, R. V., Carreiro, M. M., Zhu, W.-X., & Baxter, J. W. (2000b). Atmospheric deposition to oak forests along an urban-rural gradient. Environmental Science & Technology, 34(20), 4294–4300. doi:10.1021/es001077q MacFaden, S. W., O’Neil-Dunne, J. P. M., Royar, A. R., Lu, J. W. T., & Rundle, A. G. (2012). High-resolution tree canopy mapping for New York City using LIDAR and object-based image analysis. Journal of Applied Remote Sensing, 6(1), 063567. doi:10.1117/1.JRS.6.063567 Mallin, M. A., Williams, K. E., Esham, E. C., & Lowe, R. P. (2000). Effect of Human Development on Bacteriological Water Quality in Coastal Watersheds. Ecological Applications, 10(4), 1047–76  1056. doi:10.1890/1051-0761(2000)010[1047:EOHDOB]2.0.CO;2 Malthus, T. J., & Younger, C. J. (2000). Remotely sensing stress in street trees using high spatial resolution imagery. In Proceedings of the 2nd International Conference on Geospatial Information in Agriculture and Forestry. Lake Buena Vista, Florida. McDonnell, M. J., Pickett, S. T. A., Groffman, P., Bohlen, P., Pouyat, R. V., Zipperer, W. C., … Medley, K. (1997). Ecosystem processes along an urban-to-rural gradient. Urban Ecosystems, 1(1), 21–36. doi:10.1007/978-0-387-73412-5_18 McGaughey, R. J. (2014). FUSION/LDV: Software for LIDAR data analysis and visualization. United States Department of Agriculture, Forest Service. McPherson, E. G. (1993). Monitoring urban forest health. Environmental Monitoring and Assessment, 26(2-3), 165–174. McPherson, E. G., & Simpson, J. R. (2003). Potential energy savings in buildings by an urban tree planting programme in California. Urban Forestry & Urban Greening, 2(2), 73–86. doi:10.1078/1618-8667-00025 Millard, K., & Richardson, M. (2013). Wetland mapping with LiDAR derivatives, SAR polarimetric decompositions, and LiDAR-SAR fusion using a random forest classifier. Canadian Journal of Remote Sensing, 39(4), 290–307. doi:10.5589/m13-038 Millard, K., & Richardson, M. (2015). On the Importance of Training Data Sample Selection in Random Forest Image Classification: A Case Study in Peatland Ecosystem Mapping. Remote Sensing, 7(7), 8489–8515. doi:10.3390/rs70708489 Miller, J. E., Nelson, S. a. C., & Hess, G. R. (2009). An Object Extraction Approach for Impervious 77  Surface Classification with Very-High-Resolution Imagery. The Professional Geographer, 61(February 2015), 250–264. doi:10.1080/00330120902742920 Morgenroth, J., & Buchan, G. D. (2009). Soil moisture and aeration beneath pervious and impervious pavements. Arboriculture and Urban Forestry, 35(3), 135–141. Morsdorf, F., Meier, E., Allg, B., & Daniel, N. (2002). Clustering in airborne laser scanning raw data for segmentation of single trees. International Archives of the Photogrammetry, Remote Sensing and Spatial Information Sciences, 34, 27–33. Mueller, E. C., & Day, T. A. (2005). The effect of urban ground cover on microclimate, growth and leaf gas exchange of oleander in Phoenix, Arizona. International Journal of Biometeorology, 49(4), 244–255. doi:10.1007/s00484-004-0235-1 Næsset, E. (2002). Predicting forest stand characteristics with airborne scanning laser using a practical two-stage procedure and field data. Remote Sensing of Environment, 80(1), 88–99. doi:10.1016/S0034-4257(01)00290-5 Niemelä, J. (1999). Is there a need for a theory of urban ecology? Urban Ecosystems, 3(1), 57–65. doi:10.1023/a:1008817325994 Nowak, D. J., Crane, D. E., & Stevens, J. C. (2006). Air pollution removal by urban trees and shrubs in the United States. Urban Forestry & Urban Greening, 4(3-4), 115–123. doi:10.1016/j.ufug.2006.01.007 Nowak, D. J., & McBride, J. R. (1991). Testing microdensitometric ability to determine Monterey pine urban tree stress. Photogrammetric Engineering & Remote Sensing, 59(1), 89–91. O’Neil-Dunne, J. P. M., MacFaden, S. W., Royar, A. R., & Pelletier, K. C. (2013). An object-based system 78  for LiDAR data fusion and feature extraction. Geocarto International, 28(3), 227–242. doi:10.1080/10106049.2012.689015 Oshio, H., Asawa, T., Hoyano, A., & Miyasaka, S. (2013). Accuracy of external form of individual trees acquired by high-resolution airborne LiDAR. In Joint Urban Remote Sensing Event 2013 (pp. 103–106). doi:10.1109/JURSE#.2013.6550676 Payton, S., Lindsey, G., Wilson, J., Ottensmann, J. R., & Man, J. (2008). Valuing the benefits of the urban forest: a spatial hedonic approach. Journal of Environmental Planning and Management, 51(6), 717–736. doi:10.1080/09640560802423509 Peckham, S. C., Duinker, P. N., & Ordóñez, C. (2013). Urban forest values in Canada: Views of citizens in Calgary and Halifax. Urban Forestry & Urban Greening, 12(2), 154–162. doi:10.1016/j.ufug.2013.01.001 Pickett, S. T. A., Cadenasso, M. L., Grove, J. M., Boone, C. G., Groffman, P. M., Irwin, E., … Warren, P. (2011). Urban ecological systems: scientific foundations and a decade of progress. Journal of Environmental Management, 92(3), 331–62. doi:10.1016/j.jenvman.2010.08.022 Pinheiro, J., Bates, D., DebRoy, S., & Sarkar, D. (2013). nlme: Linear and nonlinear mixed effects models. Popescu, S. C., & Wynne, R. H. (2004). Seeing the trees in the forest: using lidar and multispectral data fusion with local filtering and variable window size for estimating tree height. Photogrammetric Engineering and Remote, 70(5), 589–604. Popescu, S. C., Wynne, R. H., & Nelson, R. F. (2003). Measuring individual tree crown diameter with lidar and assessing its influence on estimating forest volume and biomass. Canadian Journal of 79  Remote Sensing, 29(5), 564–577. doi:10.5589/m03-027 Popescu, S. C., Wynne, R. H., & Nelson, R. H. (2002). Estimating plot-level tree heights with LIDAR: Local filtering with a canopy-height based variable window size. Computers and Electronics in Agriculture, 37(1-3), 71−95. Pouyat, R. V., McDonnell, M. J., & Pickett, S. T. a. (1995). Soil Characteristics of Oak Stands along an Urban-Rural Land-Use Gradient. Journal of Environment Quality, 24(3), 516–526. doi:10.2134/jeq1995.00472425002400030019x Rapport, D. J., Costanza, R., & McMichael, A. J. (1998). Assessing ecosystem health. Trends in Ecology and Evolution, 13(10), 397–402. doi:10.1016/S0169-5347(98)01449-9 Rempel, R. S., & Kushneriuk, R. S. (2003). The influence of sampling scheme and interpolation method on the power to detect spatial effects of forest birds in Ontario (Canada). Landscape Ecology, 18(8), 741–757. doi:10.1023/B:LAND.0000014469.30984.24 Richter, R., & Schlapfer, D. (2016). ATCOR-4 User Guide. Roman, L. A., McPherson, E. G., Scharenbroch, B. C., & Bartens, J. (2013). Identifying common practices and challenges for local urban tree monitoring programs across the United States. Arboriculture and Urban Forestry, 39(6), 292–299. Rosenfeld, A. H., Akbari, H., Romm, J. J., & Pomerantz, M. (1998). Cool communities: strategies for heat island mitigation and smog reduction. Energy and Buildings, 28(1), 51–62. Sampson, P. H., Zarco-Tejada, P. J., Mohammed, G. H., Miller, J. R., & Noland, T. L. (2003). Hyperspectral remote sensing of forest condition: Estimating chlorophyll content in tolerant hardwoods. Forest Science, 49(3), 381–391. 80  Schardt, M., Ziegler, M., Wimmer, A., & Wack, R. (2002). Assessment of forest parameters by means of laser scanning. International Archives of Photogrammetry Remote Sensing and Spatial Information Sciences, 34(3/A), 302–309. Schomaker, M. E., Zarnoch, S. J., Bechtold, W. A., Latelle, D. J., Burkman, W. G., & Cox, S. M. (2007). Crown-Condition Classification : A Guide to Data Collection and Analysis. Schreyer, J., Tigges, J., Lakes, T., & Churkina, G. (2014). Using Airborne LiDAR and QuickBird Data for Modelling Urban Tree Carbon Storage and Its Distribution—A Case Study of Berlin. Remote Sensing, 6(11), 10636–10655. doi:10.3390/rs61110636 Singh, K. K., Davis, A. J., & Meentemeyer, R. K. (2015). Detecting understory plant invasion in urban forests using LiDAR. International Journal of Applied Earth Observation and Geoinformation, 38, 267–279. doi:10.1016/j.jag.2015.01.012 Sleavin, W. J., Civco, D. L., Prisloe, S., & Gianotti, L. (2000). Measuring impervious surfaces for non-point source pollution modeling. Proceedings of the ASPRS 2000 Annual Conference, 22–26. Smiley, E., Fraedrich, B., & Fengler, P. (2007). Hazard tree inspection, evaluation and management. In J. E. Kuser (Ed.), Urban and Community Forestry in the Northeast (pp. 277–294). Springer Netherlands. doi:10.1007/978-1-4020-4289-8 Smith, H. G. (1964). Root Spread Can Be Estimated From Crown Width of Douglas Fir, Lodgepole Pine, and Other British Columbia Tree Species. The Forestry Chronicle, 40(4), 456–473. Smith, S., Holland, D., & Longley, P. (2004). The importance of understanding error in LiDAR digital elevation models. In 20th ISPRS conference. Istanbul, Turkey (pp. 12–34). Istanbul, Turkey. Somers, B., Verbesselt, J., Ampe, E. M., Sims, N., Verstraeten, W. W., & Coppin, P. (2010). Spectral 81  mixture analysis to monitor defoliation in mixed-aged Eucalyptus globulus Labill plantations in southern Australia using Landsat5-TM and EO-1Hyperion data. International Journal of Applied Earth Observation and Geoinformation, 12(4), 270–277. doi:10.1016/j.jag.2010.03.005 Statistics Canada. (2011). Surrey, British Columbia (Code 5915004) and British Columbia (Code 59) (table). Census Profile. 2011 Census. Ottawa. doi:98-316-XWE Stewart, H. (1984). Cedar: Tree of Life to the Northwest Coast Indians. Vancouver, British Columbia: Douglas & McIntyre. Stone, C., Coops, N., & Culvenor, D. (2000). Conceptual Development of a Eucalypt Canopy Condition Index Using High Resolution Spatial and Spectral Remote Sensing Imagery. Journal of Sustainable Forestry, 11(4), 23–45. doi:10.1300/J091v11n04_02 St-Pierre, S. (2011). Évaluation et optimisation des gains environnementaux découlant dʼun programme municipal de plantation dʼarbres. Université de Sherbrooke. Suárez, J. C., Ontiveros, C., Smith, S., & Snape, S. (2005). The Use of Airborne LiDAR and Aerial Photography in the Estimation of Individual Tree Heights in Forestry. Computers & Geosciences, 31(2), 253–262. Tooke, T. R., Coops, N. C., & Voogt, J. (2009). Assessment of urban tree shade using fused LIDAR and high spatial resolution imagery. In Urban Remote Sensing Event, 2009 Joint (pp. 1–6). IEEE. Trumbore, S., Brando, P., & Hartmann, H. (2015). Forest health and global change. Science, 349(6250), 814–818. doi:10.1126/science.aac6759 Viswanathan, B., Volder, A., Watson, W. T., & Aitkenhead-Peterson, J. A. (2011). Impervious and pervious pavements increase soil CO2 concentrations and reduce root production of American 82  sweetgum (Liquidambar styraciflua). Urban Forestry & Urban Greening, 10(2), 133–139. doi:10.1016/j.ufug.2011.01.001 Weng, Q. (2012). Remote sensing of impervious surfaces in the urban areas: Requirements, methods, and trends. Remote Sensing of Environment, 117, 34–49. doi:10.1016/j.rse.2011.02.030 Weng, Q., Hu, X., & Lu, D. (2008). Extracting impervious surfaces from medium spatial resolution multispectral and hyperspectral imagery: a comparison. International Journal of Remote Sensing, 29(11), 3209–3232. doi:10.1080/01431160701469024 Weng, Q., Lu, D., & Schubring, J. (2004). Estimation of land surface temperature–vegetation abundance relationship for urban heat island studies. Remote Sensing of Environment, 89(4), 467–483. doi:10.1016/j.rse.2003.11.005 Wiens, J. A. (1989). Spatial scaling in ecology. Functional Ecology, 3(4), 385–397. doi:10.2307/2389612 Xiao, Q., & McPherson, E. G. (2002). Rainfall interception by Santa Monica’s municipal urban forest. Urban Ecosystems, 6(4), 291–302. Xie, Y., Sha, Z., & Yu, M. (2008). Remote sensing imagery in vegetation mapping: a review. Journal of Plant Ecology, 1(1), 9–23. doi:10.1093/jpe/rtm005 Yu, X., Hyyppä, J., Vastaranta, M., Holopainen, M., & Viitala, R. (2011). Predicting individual tree attributes from airborne laser point clouds based on the random forests technique. ISPRS Journal of Photogrammetry and Remote Sensing, 66(1), 28–37. doi:10.1016/j.isprsjprs.2010.08.003 83  Zarnoch, S. J., Bechtold, W. A., & Stolte, K. W. (2004). Using crown condition variables as indicators of forest health. Canadian Journal of Forest Research, 34(5), 1057–1070. doi:10.1139/X03-277 Zhang, C., & Qiu, F. (2012). Mapping individual tree species in an urban forest using airborne lidar data and hyperspectral imagery. Photogrammetric Engineering & Remote Sensing, 78(10), 1079–1087. doi:10.14358/PERS.78.10.1079 Zhang, C., Zhou, Y., & Qiu, F. (2015). Individual Tree Segmentation from LiDAR Point Clouds for Urban Forest Inventory. Remote Sensing, 7(6), 7892–7913. doi:10.3390/rs70607892 Zhang, J., Sohn, G., & Brédif, M. (2014). A hybrid framework for single tree detection from airborne laser scanning data: A case study in temperate mature coniferous forests in Ontario, Canada. ISPRS Journal of Photogrammetry and Remote Sensing, 98, 44–57. doi:10.1016/j.isprsjprs.2014.08.007 Zhou, W., Troy, A., & Grove, M. (2008). Modeling residential lawn fertilization practices: Integrating high resolution remote sensing with socioeconomic data. Environmental Management, 41(5), 742–752. doi:10.1007/s00267-007-9032-z Zuur, A. F., Ieno, E. N., Walker, N. J., Savliev, A. A., & Smith, G. M. (2009). Mixed effects models and extensions in ecology with R. New York, NY: Springer.    84  Appendix 1 Spectral channel numbers, band centres and band widths for the hyperspectral imagery of the city of Surrey, British Columbia, Canada. Imagery was acquired by Airborne Imaging (Calgary, Alberta, Canada), between May 2nd and May 5th, 2013. Spectral channel number Band centre (nm) Band width (nm) Spectral channel number Band centre (nm) Band width (nm) 1 367.6 9.6 37 712.1 9.5 2 377.2 9.6 38 721.6 9.5 3 386.8 9.7 39 731.1 9.6 4 396.5 9.6 40 740.7 9.5 5 406.1 9.6 41 750.2 9.5 6 415.7 9.6 42 759.7 9.6 7 425.3 9.6 43 769.3 9.5 8 434.9 9.6 44 778.8 9.5 9 444.5 9.6 45 788.3 9.6 10 454.1 9.6 46 797.9 9.5 11 463.7 9.6 47 807.4 9.5 12 473.3 9.6 48 816.9 9.6 13 482.9 9.5 49 826.5 9.5 14 492.4 9.6 50 836 9.5 15 502.0 9.6 51 845.5 9.6 16 511.6 9.6 52 855.1 9.5 17 521.2 9.5 53 864.6 9.6 18 530.7 9.6 54 874.2 9.5 19 540.3 9.6 55 883.7 9.6 20 549.9 9.5 56 893.3 9.5 85  Spectral channel number Band centre (nm) Band width (nm) Spectral channel number Band centre (nm) Band width (nm) 21 559.4 9.6 57 902.8 9.6 22 569.0 9.5 58 912.4 9.5 23 578.5 9.6 59 921.9 9.6 24 588.1 9.5 60 931.5 9.6 25 597.6 9.6 61 941.1 9.5 26 607.2 9.5 62 950.6 9.6 27 616.7 9.5 63 960.2 9.6 28 626.2 9.6 64 969.8 9.6 29 635.8 9.5 65 979.4 9.5 30 645.3 9.6 66 988.9 9.6 31 654.9 9.5 67 998.5 9.6 32 664.4 9.5 68 1008.1 9.6 33 673.9 9.6 69 1017.7 9.6 34 683.5 9.5 70 1027.3 9.6 35 693.0 9.5 71 1036.9 9.6 36 702.5 9.6 72 1046.5 9.6    86  Appendix 2 Summary of results for City of Surrey managers: Linking LiDAR data to existing tree inventories  Before any useful information can be extracted at an individual tree basis, tree locations within the LiDAR data must be matched to existing entries in the city’s tree inventory.  A variable filter window (VWF) was found to be an effective way of locating potential treetops in the LiDAR data.  Using LiDAR to assess tree condition  The vertical distribution of LiDAR returns within a tree crown was found to be an adequate predictor of crown density for trees taller than 8 m.  LiDAR can obtain accurate tree height estimates. When paired with tree ages obtained from planting records, height models can be computed for specific tree species.  The residuals of these models indicate the difference between the actual height of a given tree and the expected height for its age. This information can be used as an indicator of a tree’s condition.  Remote sensing of imperviousness and land cover  A classified land cover map with an accuracy of 88.6% was created using a combination of LiDAR data and hyperspectral imagery.  Land cover classes were: deciduous trees, coniferous trees, buildings, pavement, grass, bare ground and water.  LiDAR supplied four of the seven key variables used in the classification. This reflects LiDAR’s utility in characterizing the complex vertical structure of the urban environment.  A map of imperviousness was derived from the land cover map by assigning an imperviousness coefficient to each land cover class, with deciduous and coniferous trees being least impervious (10%) and pavement and buildings being most impervious (99%).    87  Relationship between tree height and imperviousness  Height model residuals were computed for 1,914 trees drawn from the five most commonly planted coniferous species in the city.  At an individual tree level, imperviousness had no discernable relationship with the trees’ height.  At a broad spatial scale (0.5 km2), there was a significant relationship between imperviousness and tree height. Without a similar relationship at the individual tree level, this may be due to correlation with other environmental variables associated with urban development, such as pollution or microclimatic variations.  These results suggest that imperviousness has a limited impact on tree height within the city of Surrey.  Data products The following data products will be delivered to City of Surrey personnel.  Classified land cover map  Imperviousness map  Height model residuals for free growing Thuja plicata, Pseudotsuga menziesii, Sequoiadendron giganteum, Cupressus nootkatensis and Picea abies  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0306910/manifest

Comment

Related Items