Modelling tree canopy cover and evaluating the driving factors based on remotely sensed data and machine learning

Quantifying urban tree cover is important to ensure sustainable urban ecosystem. This study calculates urban percent tree cover (PTC) for Bursa city, Turkey from Sentinel-2 data and evaluates the driving factors of PTC using an Artificial Neural Network-Multi Layer Perception (ANN-MLP) approach. For the PTC calculation, a Regression Tree (RT) analysis was performed using several vegetation indices (NDVI, LAI, fCOVER, MSAVI2, and MCARI) to improve accuracy. Socio-economic, topographic, and biophysical variables were incorporated into the ANN-MLP approach to evaluate the factors that drive urban PTC. A PTC prediction map was generated with an accuracy of 0.95 and a coefficient of determination of 0.87. The ANN-MLP training process yielded a correlation coefficient value of 0.71 and an R-square of 0.82 was achieved between the predicted ANN-MLP and observed tree cover maps. A priority tree cover map was generated considering statistical relationships between the factors and the ANN-MLP prediction map in addition to visual interpretations at the urban scale. Results demonstrate that, unlike other urban forms, PTC has a statistically negative relationship with the gross dwelling density (R2 =0.31). Topographic variables including slope and DEM were positively correlated with PTC with the R2 value of 0.80 and 0.72 respectively. The integration of remote sensing data with vegetation indices and driving factors yielded accurate prediction for identifying and evaluating the variability in the urban PTC.     

Land use, land cover changes and coastal lagoon surface reduction associated with urban growth in northwest Mexico

Coastal land use and land cover changes, emphasizing the alterations of coastal lagoons, were assessed in northwest Mexico using satellite imagery processing. Supervised classifications of a Landsat series (1973–1997) and the coefficients Kappa (K) and Tau (τ), were used to assess the area and verify the accuracy of the classification of six informational classes (urban area, aquatic systems, mangrove, agriculture, natural vegetation, and aquaculture). Pixel-by-pixel change detection among dates was evaluated using the Kappa Index of Agreement (KIA). Besides the overall estimation of the aquatic systems class, variations in the three lagoons present in the study area were analyzed individually. Measures of agreement between the classification and reference data indicate that the accuracy for the classification ranked from moderate to high (K = 0.76 ± 0.07; τ = 0.77 ± 0.06). From 1973 to 1997 urban area has doubled, growing to the north and the northeast, extending mainly over natural vegetation and agricultural land. La Escopama and El Sabalo, two of the lagoons studied, reduced their size to less than half that estimated in 1973, but the main estuarine system in the study area, Estero de Urias - El Infiernillo, has maintained its area without noticeable changes. However, the surrounding landscape in Estero de Urias - Infiernillo is changing from natural vegetation and agriculture to urban land use. Consequently, to limit as much as possible changes in the area to natural causes, some management measures must be considered to design urban development plans and to recover and preserve the natural areas, on a broad scale rather than a local spatial scale. This revised version was published online in August 2006 with corrections to the Cover Date.     

Urban permeability for birds: An approach combining mobbing-call experiments and circuit theory

The urban matrix was recently shown to be a mosaic of heterogeneous dispersal habitats. We conducted a playback experiment of mobbing calls to examine the probabilities of forest birds to cross a distance of 50 m over urban matrix with different land-cover types in an urban area. We treated the reciprocal of the crossing probabilities as a movement resistance for forest birds. We drew resistance surfaces based on the land-cover maps of urban Sapporo. We applied a circuit theory to examine the relative role of a detour route consisting of a riparian corridor and urban matrix for dispersing forest bird individuals from continuous forest to an isolated green space in the midst of an urban area. Our results showed that wood cover had the highest crossing probability, while open land (grassland and pavement) had the lowest probabilities. Buildings and water surface displayed an intermediate probability. Resistance surfaces and flow maps at 25- and 50-m resolutions were very similar and suggested that dispersing individuals are likely to use the intervening building areas that dominate the urban matrix rather than detour through riparian corridors. Our results showed the useful combination of experimental approaches and circuit theory, and the importance of the spatial configuration of corridors, as well as the composition and management of dispersal habitats, to landscape connectivity.     

Mapping and analyzing landscape patterns

Landscapes were mapped as clusters of 2 or 3 land cover^** types, based on their pattern within the clusters and tendency for a single type to dominate. These landscapes, called Landscape Pattern Types (LPTs), were combined with other earth surface feature data in a Geographic Information System (GIS) to test their utility as analysis units. Road segment density increased significantly as residential and urbanized land cover components increased from absent, to present as patch, to present as matrix (i.e., the dominant land cover type). Stream segment density was significantly lower in LPTs with an urbanized or residential matrix than in LPTs with either a forest or agriculture matrix, suggesting an inverse relationship between stream network density and the prevalence of human development other than agriculture in the landscape. The ratio of average forest patch size to total forest in the LPT unit decreased as agriculture replaced forest, then increased as residential and urban components dominated. Wetland fractal dimension increased as agriculture and residential land cover components of LPTs increased. Comparison of LPT and LUDA land cover area statistics in ecoregions suggested that land cover data alone does not provide information as to its spatial arrangement.     

Studying the spatial evolutionary behavior of urban forest patches from the perspective of pattern-process relationships

Urban forest dynamics can influence the provision of ecosystem services provision. Considerable research has been conducted to understand how these dynamics respond to urbanization, from individual patches to entire landscapes. However, most of these are cross-sectional studies based on landscape metrics, and research using a process-based perspective in this context is scarce. In this study, we present a “pattern-process” analytical framework to quantify the evolutionary behavior of urban forest patches. We combine this framework with land cover classification data based on high-resolution remote sensing images (< 1 m) from 2002, 2013, and 2019 to detect the dynamic characteristics of four processes of forest patches in Beijing urban areas. These dynamic characteristics include: size distribution, aggregation and fragmentation, transfer, and self-stabilization. The results showed that 1) the average size of the patches in the study area is increasing, and patches larger than 50 m² have a more positive influence on the process of patch structure evolution, 2) patch fragmentation shifts with the direction of urban sprawl, 3) transfer between urban forest and bare land is increasing, and 4) urban forest network construction positively enhances the stability of patches. This framework can provide a useful basis for understanding the spatial and temporal evolution of urban forest landscapes during urban development and contribute to the sustainable management of urban forests.     

Evaluating the potential for urban heat-island mitigation by greening parking lots

Artificial urban land uses such as commercial and residential buildings, roads, and parking lots covered by impervious surfaces can contribute to the formation of urban heat islands (UHIs), whereas vegetation such as trees, grass, and shrubs can mitigate UHIs. Considering the increasing area of parking lots with little vegetation cover in Nagoya, Japan, this study evaluated the potential for UHI mitigation of greening parking lots in Nagoya. The relationships between land surface temperature (LST) and land use/land cover (LULC) in different seasons were analyzed using multivariate linear regression models. Potential UHI mitigation was then simulated for two scenarios: (1) grass is planted on the surface of each parking lot with coverage from 10 to 100% at an interval of 10% and (2) parking lots are covered by 30% trees and 70% grass. The results show that different LULC types play different roles in different seasons and times. On average, both scenarios slightly reduced the LST for the whole study area in spring or summer. However, for an individual parking lot, the maximum LST decrease was 7.26 °C in summer. This research can help us understand the roles of vegetation cover and provide practical guidelines for planning parking lots to mitigate UHIs.     

Generating confidence intervals for composition-based landscape indexes

Many landscape indexes with ecological relevance have been proposed, including diversity indexes, dominance, fractal dimension, and patch size distribution. Classified land cover data in a geographic information system (GIS) are frequently used to calculate these indexes. However, a lack of methods for quantifying uncertainty in these measures makes it difficult to test hypothesized relations among landscape indexes and ecological processes. One source of uncertainty in landscape indexes is classification error in land cover data, which can be reported in the form of an error matrix. Some researchers have used error matrices to adjust extent estimates derived from classified land cover data. Because landscape diversity indexes depend only on landscape composition – the extent of each cover in a landscape – adjusted extent estimates may be used to calculate diversity indexes. We used a bootstrap procedure to extend this approach and generate confidence intervals for diversity indexes. Bootstrapping is a technique that allows one to estimate sample variability by resampling from the empirical probability distribution defined by a single sample. Using the empirical distribution defined by an error matrix, we generated a bootstrap sample of error matrixes. The sample of error matrixes was used to generate a sample of adjusted diversity indexes from which estimated confidence intervals for the diversity indexes were calculated. We also note that present methods for accuracy assessment are not sufficient for quantifying the uncertainty in landscape indexes that are sensitive to the size, shape, and spatial arrangement of patches. More information about the spatial structure of error is needed to calculate uncertainty for these indexes. Alternative approaches should be considered, including combining traditional accuracy assessments with other probability data generated during the classification procedure.     

Do land cover changes shape sensitivity to forest fires in peri-urban areas?

The present study assesses the spatial distribution of selected land cover classes at two years (1975 and 2000) in a Mediterranean urban area (Athens, Greece) to test the hypothesis that land cover changes determine an increase in the sensitivity of landscape to forest fires on a regional scale. While urban and agricultural areas increased, although with different rates of growth, forests and semi-natural areas decreased in the study area. These changes are reflected in a significant increase of vegetation sensitivity to forest fires measured by the forest fire risk (FR) index developed in the framework of MEDALUS project. The cover classes which contributed the most to the increase of the FR index were crop mosaic, mixed agricultural-natural areas and discontinuous, low-density settlements. Results of the present study indicate that the transformation of the fringe landscape towards low-quality agricultural and pasture areas and fragmented forest patches is potentially detrimental for environmental quality and the ecological fragility of land.     

Mapping vegetation functional types in urban areas with WorldView-2 imagery: Integrating object-based classification with phenology

Mapping urban vegetation is a prerequisite to accurately understanding landscape patterns and ecological services provided by urban vegetation. However, the uncertainties in fine-scale vegetation biodiversity mapping still exist in capturing vegetation functional types efficiently at fine scale. To facilitate the application of fine-scale vegetation spatial configuration used for urban landscape planning and ecosystem service valuation, we present an approach integrating object-based classification with vegetation phenology for fine-scale vegetation functional type mapping in compact city of Beijing, China. The phenological information derived from two WorldView-2 imagery scenes, acquired on 14 September 2012 and 26 November 2012, was used to aid in the classification of tree functional types and grass. Then we further compared the approach to that of using only one WorldView imagery. We found WorldView-2 imagery can be successfully applied to map functional types of urban vegetation with its high spatial resolution and relatively high spectral resolution. The application of the vegetation phenology into classification greatly improved the overall accuracy of classification from 82.3% to 91.1%. In particular, the accuracies of vegetation types was improved by from 10% to 13.26%. The approach integrating vegetation phenology with high-resolution remote sensed images provides an efficient tool to incorporate multi-temporal data into fine-scale urban classification.     

Land Use – Land Cover Conversion, Regeneration and Degradation in the High Elevation Bolivian Andes

Regional land-cover change affects biodiversity, hydrology, and biogeochemical cycles at local, watershed, and landscape scales. Developing countries are experiencing rapid land cover change, but assessment is often restricted by limited financial resources, accessibility, and historical data. The assessment of regional land cover patterns is often the first step in developing conservation and management plans. This study used remotely sensed land cover and topographic data (Landsat and Shuttle Radar Topography Mission), supervised classification techniques, and spectral mixture analysis to characterize current landscape patterns and quantify land cover change from 1985 to 2003 in the Altiplano (2535–4671 m) and Intermediate Valley (Mountain) (1491–4623 m) physiographic zones in the Southeastern Bolivian Andes. Current land cover was mapped into six classes with an overall accuracy of 88% using traditional classification techniques and limited field data. The land cover change analysis showed that extensive deforestation, desertification, and agricultural expansion at a regional scale occurred in the last 20 years (17.3% of the Mountain Zone and 7.2% of the Altiplano). Spectral mixture analysis (SMA) indicated that communal rangeland degradation has also occurred, with increases in soil and non-photosynthetic vegetation fractions in most cover classes. SMA also identified local areas with intensive management activities that are changing differently from the overall region (e.g., localized areas of increased green vegetation). This indicates that actions of local communities, governments, and environmental managers can moderate the potentially severe future changes implied by the results of this study.     

Predicting land cover change and avian community responses in rapidly urbanizing environments

We used an integrated modeling approach to simulate future land cover and predict the effects of future urban development and land cover on avian diversity in the Central Puget Sound region of Washington State, USA. We parameterized and applied a land cover change model (LCCM) that used output from a microsimulation model of urban development, UrbanSim, and biophysical site and landscape characteristics to simulate land cover 28 years into the future. We used 1991, 1995, and 1999 Landsat TM-derived land cover data and three different spatial partitions of our study area to develop six different estimations of the LCCM. We validated model simulations with 2002 land cover. We combined UrbanSim land use outputs and LCCM simulations to predict changes in avian species richness. Results indicate that landscape composition and configuration were important in explaining land cover change as well as avian species response to landscape change. Over the next 28 years, urban land cover was predicted to increase at the expense of agriculture and deciduous and mixed lowland forests. Land cover changes were predicted to reduce the total number of avian species, with losses primarily in native forest specialists and gains in common synanthropic species such as the American Crow (Corvus brachyrhynchos). The integrated modeling framework we present has potential applications in urban and natural resource planning and management and in assessing of the effects of policies on land development, land cover, and avian biodiversity.     

Urban vegetation phenology analysis using high spatio-temporal NDVI time series

Most of current products can partially reach the requirement of high spatial and temporal resolution needed in urban applications. Fortunately, the new generation of satellite in a form of constellation, e.g. Europe’s Sentinel-2, China’s HJ-1A/B and GF-1/6, is expected to provide more frequent observations (<1 week) with a higher spatial resolution (<30 m). Consequently, a proper method should be selected to construct high spatio-temporal time-series NDVI and to derive phenological features for urban applications. In this study, a high spatio-temporal NDVI product for urban scale vegetation time series is conducted based on HJ-1A/B data. Three related issues, i.e. the optimal filter, time series decomposition, phenological features derivation are addressed. In addition, the effect of spatial and temporal resolution on the phenological features extraction is also discussed according to the comparison between the derived NDVI product and that extracted from MODIS. The results show that the Savitzky-Golay (S-G) filter is the best filter for the reconstruction of HJ NDVI time series. There is some difference for phenology derivation using “season” and “season + trend” depending on the absence/presence of breakpoints in the curve. The spatial details of phenological features can be built by the high-spatial time-series NDVI, showing a great potential in urban applications. Compared with the MODIS NDVI time series, HJ NDVI time series can get more detail information than overall phenological features because of its high spatio-temporal resolution.     

Microclimates in a desert city were related to land use and vegetation index

A heterogeneous patchwork mosaic of soil, vegetation, and built surfaces that result from a variety of urban land uses cause urban microclimates within cities. We studied the seasonal relationships of land use, urban plant cover and microclimate in Phoenix, Arizona, USA, metropolitan. Early morning (0500 HR) and afternoon (1500 HR) near-surface temperatures and humidities were measured along multiple transects in this desert city and outlying areas during June and December 1999. A Landsat thematic mapper normalized differential vegetation index (NDVI) image was used to quantify spatial patterns of plant density. Land use had the most pronounced effect on microclimate during the early morning hours of summer. Agricultural and residential land uses had the highest relative humidities, dew point temperatures, and NDVI, and the lowest air temperatures. Commercial and industrial land uses had highest temperatures and lowest NDVI. Temperatures were generally negatively correlated to NDVI, while humidity and dew point temperatures were generally positively correlated to NDVI. Distance from the urban core did not affect NDVI but had a significant negative effect on adjusted air temperature. In addition, a historical comparison of land use, NDVI and microclimate data collected during 1976 and again during 1999 along two transects revealed overall decreases in NDVI and relative increases in air temperature indicative of urban expansion. These findings show that microclimates in this desert city are caused by more than just variations in plant cover, and are likely an interactive effect of vegetation density and other non-vegetative urban surfaces.     

GeoAI to implement an individual tree inventory: Framework and application of heat mitigation

Individual Tree Inventory (ITI) is critical for urban planning, including urban heat mitigation. However, an ITI is usually incomplete and costly due to data collection challenges in the dynamic urban landscape. This research developed a methodical GeoAI framework to build a comprehensive ITI and quantify tree species cooling on rising urban heat.The object detection Faster R-CNN model with Inception ResNet V2 was implemented to detect individual trees canopy and seven tree species (Callery pear, Chinese elm, English elm, Mugga ironbark, Plane tree, Spotted gum and White cedar). The land surface temperature (LST) was derived from Landsat 8 surface reflectance imagery. Two models for ITI were further developed for spatial and statistical analysis. Firstly, an ‘Individual tree-based model’ stores the attributes of tree species and its vertical configuration obtained from LiDAR, along with its tree canopy area and surface temperature. Secondly, the ‘LST zone-based model’ stores tree canopy cover and building areas in each zone unit. Pearson correlation, global linear regression, and geographically weighted regression (GWR) were applied to establish the relationship between tree attributes, building areas (explanatory variables) with local temperature (dependent variable). Results showed that English elm has the highest cooling and least by Mugga ironbark in the study area. GWR results demonstrate that 94% of the LST was explained by tree height and tree canopy area. The LST zone-based model showed that 85% of the LST was explained by the percentage of tree species and buildings. Maps of the local R² and coefficients of the independent variables provide spatially explicit information on the cooling of different tree species compared to building areas. The implemented GeoAI approach provides important insights to urban planners and government to monitor urban trees with the enhanced Individual Tree Inventory and strategies mitigation plan to reduce the impact of climate change and global warming.     

Land use dynamics within an urbanizing non-metropolitan county in New York State (USA)

Land use/land cover data for fifteen minor civil divisions (MCDs) in Ulster County, New York (USA) were interpreted from 1968 and 1985 aerial photographs. These data were combined with ancillary physiographic and demographic data as raster layers within a computerized geographic information system (GIS). Class to class changes in land use/land cover were quantified for a study area approximately 30 kilometers by 50 kilometers. The relationships between the land use/land cover variables and the ancillary variables were modeled in a series of weighted least squares regressions employing data spatially aggregated by general soil map unit (N = 44).Between 1968 and 1985, nearly one-third of the study area changed to another land use/land cover class. Land in the urban class increased from 6.7% to 17.8% of the study area, while the forest class declined from 65.0% to 55.2%, and the agriculture class declined from 12.7% to 8.9%. Gains and losses in the remaining five major (Level I) land use/land cover classes were relatively small. Land use changes primarily involved the conversion of land from the forest, agriculture, and vacant classes to the urban class, and from the agriculture class to the forest and vacant classes. Variables accounting for the variance in the land use/land cover class proportions of the soil units were population density, highway proximity, distance to urban centers, mean elevation, mean slope gradient, and soil suitability for farming and for urban development.     

The impact of habitat characteristics on bird presence and the implications for wildlife management in the environs of Ottawa,Canada

Urban forestry is increasingly vital for both wildlife conservation and human use, despite frequent conflicts between these functions. A fundamental task in urban habitat and recreation forestry is the identification of those habitat characteristics important for animal species and the evaluation of these within the geographies of human presence, urban proximity and land cover variation and change. This paper examines the habitat characteristics for birds in urban built, green and greenbelt areas of Ottawa, Canada, and an area of continuous Ontario forest, to determine the effects of vegetation density and patch size, and human presence on bird presence. Bird presence was measured by point counts, and land cover was mapped using field observation and aerial photographs (1955 and 1999). At the species level, the pre-dominantly forest birds were affected by human presence and were primarily associated with tree stands in the greenbelt and continuous forest. In dense urban areas there were larger numbers of a few ‘generalist’ species. Both forested and urban (residential/commercial) environments increased in area between 1955 and 1999, creating the two types of land cover favouring the largest number of birds, while the less habituated grass/farmed areas declined in area. More informed bird conservation and recreation management will depend on paying greater attention to vegetation cover combinations with urban development.     

Development potential of sky gardens in the compact city of Hong Kong

Sky gardens have been actively studied and installed in different cities. Their development potential in compact developing cities has received little attention. Using remote sensing and GIS techniques, this study evaluates the vegetation configuration and development potential of sky gardens in urban Hong Kong, their underlying location, land use and building factors, and future planning and implementation concerns. Existing sky garden area is limited with sparse vegetation cover and low biomass. Existing podium gardens exceed roof gardens by about nine times. District development age has little effect on existing and potential sky gardens. Old towns have higher potential roof and podium gardens than new towns in most land uses. The effect of land use on potential sky gardens varies greatly by districts. Buildings with 10–20 floors have higher potential roof gardens in most districts. Building area is the main determinant of potential roof garden, and population density of potential podium gardens. Three scenarios of realization, namely minimum (20%), medium (50%) and maximum (80%), are adopted to project sky garden provision in individual districts. The projection extends to the contribution of new sky gardens to urban greening in terms of green cover and greening rate in districts. The challenges include susceptibility to typhoon damage in high-rise exposed sites, aggressive weed invasion, lack of roof-slab loading data in old buildings, and poor building maintenance. The opportunities include affordability of the new technology, enabling government policy, and establishment of scientific and research foundation in the local context. The development strategy could aim squarely at stringent technical standards and contractor skill requirement, and programme prioritization based on research findings. The study provides useful hints, approaches and recommendations for a systematic sky-garden action plan in Hong Kong and other similar compact cities.     

Riparian influences on stream fish assemblage structure in urbanizing streams

We assessed the influence of land cover at multiple spatial extents on fish assemblage integrity, and the degree to which riparian forests can mitigate the negative effects of catchment urbanization on stream fish assemblages. Riparian cover (urban, forest, and agriculture) was determined within 30 m buffers at longitudinal distances of 200 m, 1 km, and the entire network upstream of 59 non-nested fish sampling locations. Catchment and riparian land cover within the upstream network were highly correlated, so we were unable to distinguish between those variables. Most fish assemblage variables were related to % forest and % urban land cover, with the strongest relations at the largest spatial extent of land cover (catchment), followed by riparian land cover in the 1-km and 200-m reach, respectively. For fish variables related to urban land cover in the catchment, we asked whether the influence of riparian land cover on fish assemblages was dependent on the amount of urban development in the catchment. Several fish assemblage metrics (endemic richness, endemic:cosmopolitan abundance, insectivorous cyprinid richness and abundance, and fluvial specialist richness) were all best predicted by single variable models with % urban land cover. However, endemic:cosmopolitan richness, cosmopolitan abundance, and lentic tolerant abundance were related to % forest cover in the 1-km stream reach, but only in streams that had <15% catchment urban land cover. In these cases, catchment urbanization overwhelmed the potential mitigating effects of riparian forests on stream fishes. Together, these results suggest that catchment land cover is an important driver of fish assemblages in urbanizing catchments, and riparian forests are important but not sufficient for protecting stream ecosystems from the impacts of high levels of urbanization.     

90 years of forest cover change in an urbanizing watershed: spatial and temporal dynamics

Landscape structure in the Eastern US experienced great changes in the last century with the expansion of forest cover into abandoned agricultural land and the clearing of secondary forest cover for urban development. In this paper, the spatial and temporal patterns of forest cover from 1914 to 2004 in the Gwynns Falls watershed in Baltimore, Maryland were quantified from historic maps and aerial photographs. Using a database of forest patches from six times—1914, 1938, 1957, 1971, 1999, and 2004—we found that forest cover changed, both temporally and spatially. While total forest area remained essentially constant, turnover in forest cover was very substantial. Less than 20% of initial forest cover remained unchanged. Forest cover became increasingly fragmented as the number, size, shape, and spatial distribution of forest patches within the watershed changed greatly. Forest patch change was also analyzed within 3-km distance bands extending from the urban core to the more suburban end of the watershed. This analysis showed that, over time, the location of high rates of forest cover change shifted from urban to suburban bands which coincides with the spatial shift of urbanization. Forest cover tended to be more stable in and near the urban center, whereas forest cover changed more in areas where urbanization was still in process. These results may have critical implications for the ecological functioning of forest patches and underscore the need to integrate multi-temporal data layers to investigate the spatial pattern of forest cover and the temporal variations of that spatial pattern.     

Spatial patterns of a subtropical,coastal urban forest: Implications for land tenure,hurricanes, and invasives

Spatial patterns of tree structure and composition were studied to assess the effects of land tenure, management regimes, and the environment on a coastal, subtropical urban forest. A total of 229 plots in remnant natural areas, private residential, public non-residential, and private non-residential land tenures were analyzed in a 1273 km² study area encompassing the urbanized portion of Miami-Dade County, USA. Statistical mixed models of structure, composition, location, and land tenure data were used to analyze spatial patterns across the study area. A total of 1200 trees were measured of which 593 trees (49%) were located in residential areas, 67 (6%) in public non-residential areas, 135 trees (11%) in private non-residential areas, and 405 (34%) in remnant, natural areas. A total of 107 different tree species belonging to 90 genera were sampled. Basal area in residential land tenures increased towards the coast while private residential land tenures and natural areas had higher species diversity than non-residential areas. Tree height, crown light exposure, and crown area might indicate the effects of past hurricane impacts on urban forest structure. Land tenure, soil types, and urban morphology influenced composition and structure. Broadleaf evergreen trees are the most common growth form, followed by broadleaf deciduous, palms, and conifers. Exotic tree species originated mainly from Asia and 15% of all trees measured were considered exotic-highly invasive species. We discuss the use of these results as an ecological basis for management and resilience towards hurricane damage and identifying occurrence of invasive, exotic trees.