Urban Ecosystems

, Volume 11, Issue 4, pp 409–422

Estimates of air pollution mitigation with green plants and green roofs using the UFORE model


    • Centre for Social Innovation
  • Brad Bass
    • Adaptation and Impacts Research Group (AIRG), Atmospheric and Climate Science Directorate (ACSD)Institute for Environmental Studies University of Toronto

DOI: 10.1007/s11252-008-0054-y

Cite this article as:
Currie, B.A. & Bass, B. Urban Ecosyst (2008) 11: 409. doi:10.1007/s11252-008-0054-y


The purpose of this study was to investigate the effect of green roofs and green walls on air pollution in urban Toronto. The research looked at the synergistic effects on air pollution mitigation of different combinations of vegetation by manipulating quantities of trees, shrubs, green roofs and green walls in the study area. The effects of these manipulations were simulated with the Urban Forest Effects (UFORE) model developed by the USDA Forest Service Northeastern Regional Station. While UFORE contains several modules, Module—D quantifies the levels of air pollution for contaminants such as NO2, S02, CO, PM10 and ozone as well as hourly pollution removal rates and the economic value of pollutant removal. Six vegetation scenarios were developed within the Toronto study area to compare different subsets of vegetation and their effect on air contaminants. Results of the study indicate that grass on roofs (extensive green roofs) could augment the effect of trees and shrubs in air pollution mitigation, placing shrubs on a roof (intensive green roofs) would have a more significant impact. By extension, a 10–20% increase in the surface area for green roofs on downtown buildings would contribute significantly to the social, financial and environmental health of all citizens.


Green roofAir qualityAir contaminantsTreesShrubsUFORE ModelUrbanPlanningSmart growthBenefitsDensity


It is well known that trees, shrubs and other natural vegetation in urban areas affect air contaminant levels, and by extension, air quality and the overall experience of health and well-being of humans living in urban areas (Nowak et al. 1998). Quantifying the contribution made by green walls and green roofs on air contaminant levels within an urban neighbourhood however, is a relatively new application within the emerging discipline of green roof study. In general, green plants affect air pollutants by taking up gaseous pollutants primarily through leaf stomates. Once inside the plant, these gases react with water to form acids and other chemicals (Baldocchi et al. 1987). Green plants can also intercept particulate matter as wind currents blow them into contact with sticky plant surfaces (Bidwell and Fraser 1972). Some of these particulates can be absorbed into the plant while others simply adhere to the surface. Vegetation can be a temporary site for particulates as they can be re-suspended into the atmosphere by winds or washed off by rain water to the soil beneath (Wesely 1989).

This study was part of a larger policy exercise to assess how municipalities could integrate green roofs, green walls and other vegetation to meet energy and air quality targets at the neighbourhood or community scale. The community scale was chosen as this allows comparison of different scenarios that go beyond the scale of a building, that may not be evident when aggregated up to the whole city. For example, at the community scale, the impact of urban density on tree growth, the impacts of greening flat roofs versus every roof and the impact of installing extensive versus intensive roofs can be separated or integrated while also accounting for the effects of tree placement and specific building orientation. Beyond the integration question, recent policy changes in the Province of Ontario on growth management, could lead to significantly higher residential densities in many municipalities (Winfield 2005). Higher densities do not always accommodate the same amount of tree planting and growth. The results of this study can also be used to estimate what the loss of trees will mean in terms of air quality, and to what degree this loss can be mitigated through green roofs and green walls.

Toronto and air pollution

It is well documented that air pollution can aggravate existing breathing and heart problems to such an extent that medical treatment is necessary (Ontario Medical Association 2001). Of particular concern is asthma, which currently affects about 12% of children and 6% of adults in Canada (Yaffe 2004). Children are the most vulnerable, and Toronto-based hospitalization data reveal that children account for the largest number of asthma-related hospital admissions. While public health responses to a predicted or sudden peak in air pollution levels can be planned at the municipal level, a bigger concern is the chronic, long-term effect of air pollution on residents in urban areas.

A study by Toronto Public Health Unit (2000) estimated that exposure to five common smog-related air pollutants contributes to over 1,000 premature deaths and about 5,500 hospitalizations each year in Toronto. One of the major components of smog is ground level ozone, a gas that is created when oxides of nitrogen (NOx) and volatile organic compounds (VOCs) mix with the atmosphere in sunlight. The Ontario Medical Association (OMA) reports that each day in Canadian cities we are exposed to a chemical “soup” that contains several poisons, particularly, ground level ozone and particulate matter, otherwise known as acidic water droplets (Ontario Medical Association 2001). The OMA estimates that air pollution costs Ontario more than one billion dollars per year from hospital admissions, emergency room visits and absenteeism. The extreme consequence of inhaling smog and particulates is sudden death, but more common health-related consequences include breathing difficulties, cardiac exacerbations and asthma. The effects are most noticeable immediately after air pollution levels peak, especially in hot summer temperatures.

Vegetation and air pollution

It is well known that trees, shrubs and other natural vegetation affect urban air contaminant levels, and, by extension, air quality and the overall experience of health and well-being of humans living in urban areas (Bass 2001; Bass and Baskaran 2001; Cheney and Rosenzweig 2003; Chiotti et al. 2002). In response to urban environmental problems some authors have studied the effects of vegetation, particularly trees, on cooling ambient urban air, shading buildings and absorbing gaseous air contaminants (Akbari et al. 2001; Bass and Baskaran 2001; Bass 2001). Research in Los Angeles measured the effects of tree planting and re-roofing in lighter colours, on ambient temperatures and air pollution. The results confirmed that combining trees with cool roofs could lower the ambient temperature in Los Angeles by 3°C and cool the air around buildings (Akbari et al. 2001). Smog formation occurs when nitrogen oxides (NOx) react with volatile organic compounds that are released from the incomplete combustion of fossil fuels and is accelerated at higher ambient temperatures (Chiotti and Urquizo 1999). Although the chemistry behind smog formation is complex, it is caused by a photochemical reaction that is accelerated under higher temperatures. Cooler air reduces the reaction rate and hence the formation of smog and is also more comfortable hence reducing the need for air conditioning.

Akbari (2002) calculated that daytime temperature reductions would decrease reliance on air conditioning and reduce emissions of NOx from coal fired electricity plants resulting in an estimated 10% reduction in smog precursors or a reduction of 350 tons of NOx per day (Akbari et al. 2001). Los Angeles has a smog offset trading mark that trades NOx at $3,000 US per ton. Multiplying by 0.5 kg/MWH to get 0.15 c/kWh converts the 350 tons/day of avoided “equivalent” NOx into approximately one million US dollars per day to a city like Los Angeles (Akbari et al. 2001).

Other researchers reported that air pollution levels are reduced when wind blown particulates (PM2.5 and PM10) stick to the leaves and stems of plants (Hosker and Lindberg 1982). Similarly, gaseous air pollutants can be dissolved or sequestered, particularly carbon dioxide, through stomata on plant leaves (McPherson et al. 1994; McPherson et al. 1998; Nowak et al. 2000; Nowak and Crane 1998; Nowak and Dwyer 2001). Johnson et al. estimated that 2,000 m2 of un-mowed grass on a roof could remove as much as 4,000 kg of particulates in their leaves and stems (Johnson and Newton 1996). Peck cites German research, currently not available in English, that suggests that 1 m2 of uncut grass on a roof would create enough oxygen to meet the needs of one human over 1 year (Minke and Witter 1982). More recently, Tan and Sia (2005) sampled roof temperatures, roof glare and other air quality parameters both pre- and post-green roof installation in Singapore. Using light sensors, mini-volume aerosol samplers, particle counters, an aethalometer for black carbon mass concentration and a weather station, they reported that acidic gaseous pollutants, glare, ambient green roof surface temperatures and black carbon mass (or soot) levels all dropped significantly after the installation of the green roof. These results lend further support to the increased use of urban vegetation to improve urban air quality (Tan and Sia 2005).


The urban forest effects model (UFORE) provided a field collection tool to guide researchers in the collection and measurement of plot features such as buildings, amount of cement, tar, impervious material, soil, rock, duff/mulch, herbaceous, grass, wild grass, water, shrubs and other ground cover (Bass 2001). Other plot features were recorded as point items including: trees, shrubs, telephone poles, light standards, traffic signs, sewer grates, fire hydrants and other above ground point utilities; or as polygons with each vertex recorded: shrub beds, grass, wild grass, soil, duff/mulch, herbaceous (excluding grass and shrubs), water, buildings, asphalt, cement, rock, wood and other impervious material.

A geographic study area known as Midtown was selected within the Greater City of Toronto. Midtown is constituted by parts of Ward 22 (St. Pauls), Ward 27 (Toronto central-Rosedale) and Ward 20 (Trinity Spadina) and bounded by Spadina Avenue in the west, Bloor Street in the south, Eglinton Avenue in the north and the Don Valley ravine, Bayview Avenue, Moore Street, Frobisher Street and Chaplin Street in the east, as indicated by the yellow square (Fig. 1).
Fig. 1

The ward boundary map of Toronto

Urban forest health in this neighbourhood had been previously investigated by Kenney (2001), in a study which quantified the environmental role of Toronto’s urban forest in the Greater City of Toronto. Kenney’s study provided criteria data from 72 randomly selected on-the-ground study plots within the Midtown neighbourhood. Criteria data were adapted from these plot data that had been collected in accordance with the requirements of the UFORE (Urban Forest Effects) Model field collection tool (Nowak et al. 1998). This model formed the basis for the investigation of the effect of vegetation, particularly green roofs, on air pollution in an urban setting.

Each plot was circular with a radius of 11.287 m and provided a total surface area of 400 m2 or 0.04 ha per plot (Fig. 2). The total area of the Midtown neighbourhood was approximately 1,216 ha within the City of Toronto. Plots were selected from land-use types by randomly selecting points from a 50 × 50 m grid, overlaid on a GIS-based map of Midtown (Map Library, University of Toronto), using ArcView GIS 3.1. Colour orthophotos of the area were analyzed using Arc View GIS 3.1 to calculate plot details as required. Each orthophoto was examined separately at a scale of approximately 1:5000. Within each plot, a forest surveyor’s transit was utilized to determine the UTM (Universal Transverse Mercator) co-ordinates of each feature within the plot relative to a GPS-established plot center.
Fig. 2

Sample circular plot

A method for plot classification within Midtown was developed by the United States Department of Agriculture (USDA) Forest Service (Nowak et al. 1998). Midtown was stratified into eight land-use classes: low, medium and high residential; commercial; industrial; institutional; unclassified, and open areas, including parks, ravines, cemeteries, transportation corridors and golf courses. These categories were derived from GIS data obtained from CanMap ® Streetfiles V2.0 from DMTI Spatial 2000.

On each of the 72 plots, the following additional information was recorded:
  • Land use

  • Plot tree cover (%)

  • Ground cover (%)

  • Building information (wall material, roof material, building height in meters)

  • Shrub information (species, height [meters], percent missing, and percent of coverage of the plot)

  • Tree information including species, diameter at breast height (dbh) taken at 1.37 m, total height, bole height—height to base of live crown, crown width, missing crown, health of tree and distance to buildings.

The UFORE computer model used these measured field data inputs as well as local hourly meteorological data and air pollutant concentration measurements (collected from Environment Canada 1998) to quantify Midtown neighbourhood—specific vegetation effects on urban air pollutant concentrations. There were four UFORE modules available, but only module D was used in this research. UFORE—D, the dry deposition of air pollution, quantifies the hourly amount of pollution removed by the urban vegetation and the associated per cent improvement in air quality through out a year. Pollution removal is calculated for 03, S02, N02, C0 and PM10. UFORE calculations are based on vegetation cover data, hourly weather data and pollution concentration data. This hourly weather data was collected and collated from three Environment Canada weather sites: Toronto’s Pearson International Airport; Buttonville Airport, Richmond Hill: and, Toronto Island Airport down town Toronto. The research used 1 year of hourly pollution data from Environment Canada due to the logistics of converting hourly data to a UFORE compatible format.

Pollution deposition

Hourly pollution concentrations in parts per million (ppm), for gaseous pollutants over the city of Toronto, were obtained from the Province of Ontario’s Ministry of the Environment (six monitors at three sites). Hourly parts per million values were converted to micrograms per cubic meter based on measured atmospheric temperature and pressure. Average daily concentrations of PM10 (micrograms per cubic meter) were averaged across three sites. Missing hourly meteorological or pollution-concentration data were estimated using the monthly average for that specific hour. For example, O3 concentrations were not measured during winter months and existing O3 concentration data were extrapolated to missing months based on an average Canadian O3 concentration monthly pattern. Average hourly pollutant flux (grams per square meter of canopy coverage) among the pollutant monitor sites was multiplied by Midtown’s grass coverage (square meter) to estimate total hourly pollutant removal across Midtown. Bounds of total removal of O3, NO2, SO2, and PM10 were estimated using the typical range of published tree and shrub in-leaf dry deposition velocities (Lovett 1994).

To approximate boundary-layer heights in the study area, mixing-height measurements were used. Daily morning and afternoon mixing heights were interpolated to produce hourly values using a program from the US EPA (1995). Minimum boundary-layer heights were set to 150 m during the night and 250 m during the day based on estimated minimum boundary layer heights in cities. Heights of buildings that would have green roofs in the scenario analysis were estimated to be around two stories—or 15 m in height—a similar height to many trees in the Midtown study area. Hourly mixing heights (meters) were used in conjunction with pollution concentrations (micrograms per cubic meter) to calculate the amount of pollution within the mixing layer (micrograms per square meter). This extrapolation from ground-layer concentration to total pollution within the boundary layer assumes a well-mixed boundary layer, which is common during the day when unstable conditions prevail (Colbeck and Harrison 1985). The amount of pollution in the air was contrasted with the amount removed by the vegetation on an hourly basis to calculate the relative effect of vegetation in reducing local pollution concentrations.

The ability of individual vegetation (trees, shrubs and grass) to remove pollutants was estimated for each diameter class yielding an estimate of pollution removal by individual trees, shrubs and grass based on leaf surface area, which is the major surface for pollutant removal. Particle collection and gaseous deposition on deciduous trees in winter assumed a surface-area index for bark of 1.7 m2 of bark per m2 of ground surface covered by the tree crown (Whittaker and Woodwell 1967). To limit deposition estimates to periods of dry deposition, deposition velocities were set to zero during periods of precipitation.

Scenario development

Seven scenarios were created that representing different levels of natural vegetation within the Midtown Toronto study area. These varying amounts of natural vegetation were created by manipulating the number of trees, shrubs and grass species within the 72 study plots in Midtown. UFORE-D was used to quantify the impact of varying urban vegetation on air pollutant levels.

Scenario 1


This scenario was based on the reductions in pollutants provided by existing trees and shrubs in Midtown.

Scenario 2

Green Walls

This scenario examined the effect on air pollutant reductions in Midtown when existing trees and shrubs were removed and vertical “hedges” or walls of Juniper1 species were added within 3 m of residential (medium and low) houses.

Scenario 3

No big trees

This scenario examined the effect on air pollutant reductions in Midtown when all big trees with a diameter-at-breast-height greater than 22 cm were removed and was considered as a potential smart growth scenario.

Scenario 4

No trees

This scenario examined the effect on air pollutant reduction in Midtown when all trees are removed, and the existing shrubs were augmented with shrubs or intensive green roofs on flat roof surfaces (represented 20% of Midtown roofs in total) such as commercial, high residential and institutional buildings. The surface area for shrub roofs was derived by adding the total eligible roof surface areas of buildings that would typically quality for an intensive green roof. Hence the accumulated flat roof surface areas from commercial, institutional, and high residential buildings located on the 72 plots added up to be approximately 20% of the total surface area available in the Midtown study area. UFORE calculated the removal effect of trees, shrubs and grass separately. The combined effect is a linear combination of all three. Results from this scenario can also be derived from the baseline scenario by subtracting the impact of the trees.

Scenario 5

Trees off Buildings

This scenario examined the effect on air pollutant reduction in Midtown when trees that provided shade to buildings (within 3–5 m) were removed as occurs in many urban areas with higher densities.

Scenario 6

Trees Low Residential

This scenario examined the effect on air pollutant reduction in Midtown when baseline trees and shrubs were augmented with grass on flat roof surfaces (represented 20% of Midtown roofs in total) such as commercial, high residential and institutional buildings. The surface area for grass roofs was derived by adding the total eligible roof surface areas of buildings that would typically quality for an extensive green roof. Hence, the accumulated flat roof surface areas from commercial, institutional, and high residential buildings located on the 72 plots to be approximately 20% of the total surface area available in the Midtown study area.

Scenario 7

Grass roofs

This scenario examined the effect on air pollutant reduction in Midtown when baseline trees and shrubs were augmented with grass on all available roof surface areas across Midtown.


Results of the study are presented in Figs. 3, 4, 5, 6 and 7 below. In Figs. 3, 4, 5 and 6, the histograms illustrate the amount of a particular air contaminant that was removed—nitrogen dioxide, sulphur dioxide, ozone and particulate matter—for each scenario. In Fig. 7, the histogram reflects the US dollar monetary value of the overall reduction of the four major air contaminants considered in this study. The UFORE model illustrates that trees and shrubs remove air contaminants more effectively than green roofs or green walls. Further, when trees are compared with shrubs, trees exceed shrubs in their ability to reduce pollutants. This result is expected based on the number of functioning leaf units that provide maximal surface area in contact with air and particulates. However, in the No Trees scenario, which utilizes intensive green roofs composed of shrubs, the shrubs make up for some of the lost pollutant removal, and in the case of PM10, existing shrubs and intensive green roofs almost equal the removal that occurred in the baseline scenario.
Fig. 3

Total NO2 removal (Mg) by trees, shrubs and grass in Midtown per Annum

Fig. 4

Total O3 removal (Mg) by trees, shrubs and grass in Midtown per Annum

Fig. 5

Total PM10 removal (Mg) by trees, shrubs and grass in Midtown per Annum

Fig. 6

Total SO2 removal (Mg) by trees, shrubs and grass in Midtown per Annum

Fig. 7

Total pollution removal value (US$) by trees, shrubs and grass in Midtown per Annum

Despite the strong performance of trees and shrubs, it is neither practical nor plausible to seed most elevated roof surfaces with these heavy, tap and fibrous rooted species. The UFORE model illustrated that when grassy species were added to a mixture of vegetation groupings, they too contribute to air contaminant reductions. Adding grass resulted in significant reductions, but smaller with respect to shrubs and trees, across all four air contaminant levels measured by the UFORE model. Green walls also had a small impact on pollutant removal. Generally, their impact was slightly less than the extensive roofs in the 20% coverage scenario, except for PM10, where they removed a higher level of pollutant than the extensive green roof scenario. The monetized benefit of reducing air contaminants across the seven scenarios favours trees, but shrubs and intensive roofs can make up more than 50% of the monetary benefits in the most extreme high density scenario where sufficient space for trees does not exist. Green walls had a higher value than the 20% extensive green roof coverage but not as high as the complete roof coverage scenario.


Trees are the most important vegetation strategy for removing all pollutants at the community scale. However, shrubs, green walls and green roofs can complement, and in some cases, almost equal the capacity of existing trees, specifically in Midtown. Extensive green roofs can play a small, but important complementary role, but intensive green roofs can play a much more significant role in terms of improvements to air quality. Due to the added expense of structural loading requirements for intensive green roofs,, it is unlikely that this technology can be widely implemented in existing urban areas. In addition, although the intensive green roofs were almost as effective as the baseline for the removal of PM10, their performance was not as good for the other pollutants and hence the total value of this scenario is still much lower than the baseline.

The benefits of extensive green roof coverage did not increase in a linear fashion. The benefits of 100% coverage are not five times the benefits of the 20% coverage. Although a linear increase in pollutant removal was not expected, the removal rate may have also been affected by the orientation of the existing urban forest with respect to each building in the sample. The existing orientation would result in some of the roofs being shaded, and this would reduce their impacts on benefits such as pollutant removal and energy conservation.

The green wall scenario may be more promising than initially suggested by the outcomes of the UFORE modelling exercise. Although they did not replace the capacity of the trees, even though these walls were used on every building in that scenario, they did yield a slightly higher benefit than the 20% coverage of extensive green roofs. The cost of creating green walls may be lower than the 20% green roof option, and should be afforded serious consideration for improving air quality. In addition, a separate analysis revealed that the green walls scenario had a more significant impact on energy consumption than the baseline. Green walls could also have a larger impact on energy consumption than green roofs, but the exact difference depends on building orientation, the roof-to-building envelope ratio and the specific design technicalities associated with the green roof and the insulation in each building roof.

From a policy perspective, extensive green roofs and green walls will complement existing urban vegetation, but cannot replace a widespread removal of urban trees that might occur with an outbreak of disease, or a lack of planning for trees that might occur in newer, higher density developments. Extending the results to a smart or compact growth policy means that higher densities should still be planned to accommodate the addition of trees to avoid a further reduction in air quality. However, with widespread implementation, green roofs and green walls can compensate for some reduction, but not a complete removal of urban trees. In Midtown, combining Scenario 6, 20% coverage with extensive green roofs, with some coverage of green walls would provide an impact approximately equal to 20% of the trees.


In this study, grass was chosen as a proxy unit for green roofs because it is a known quantity in seed mixtures for green roof planting. The UFORE model was able to predict leaf area index and evapotranspiration rates for grass in its calculation of air pollution values, so that green roofs could be compared with other vegetation scenarios involving trees and shrubs. Similar data were not available for other vegetation such as sedums that are also typically used on extensive green roofs. In the grass roof scenario, data was estimated based on a predicted Leaf Area Index (LAI) of 3 for most grass species (Kenney 2001). The UFORE model was the only model available in North America that estimated a vegetation effect on air contaminants. As a result, the accuracy with which the UFORE model calculated outputs in air contaminants is unknown. Shrub coverage was used as a proxy for intensive green roofs. The UFORE model is programmed with data including the leaf area index for many individual species of tree and shrub data. Local leaf-on and leaf-off dates were given to the model so that deciduous-tree transpiration and related pollution deposition were limited to the Toronto (Canada) in-leaf period.

The plant selection was limited by the UFORE database. Another limitation is in the economic analysis, specifically the dollar values associated with the removal of pollutants over the year. The monetary values are based on American externality values that have been derived from the work done by (Murray et al. 1994) in New York State’s energy department. These values incorporate the perceived cost to society of pollution emissions based on predicted air pollution consequences to health and the environment. Exact monetary values, even using Murray’s assumptions, would most likely differ between countries with different government priorities for funding health care.


This study demonstrates that green roofs improve air quality and by extension public health safety and thereby a perceived improvement in quality of life in urban settings. Clearly, trees had the largest impact on pollutant removal, but shrubs and grass made important contributions to air quality in this case study. In the case of PM10, shrubs were shown to be almost equivalent to trees in the baseline in terms of air pollutant removal. The results demonstrate the degree to which green roofs and green walls can be used in populated urban areas to supplement existing vegetation and improve air quality when installed in sufficient quantities.

Information derived from these model outputs can however, be applied to the development of policy across several levels of municipal, provincial and federal governments. As many jurisdictions begin to consider or implement policies that support compact development and reduced urban sprawl, it is clear that planning for trees and shrubs is essential to maintaining local air quality. If a green roof policy were being developed to improve air quality, it would need to target a large number or aggregation of roofs in order to bring about a significant air quality impact. Air quality improvements such as reduced amounts of particulate matter, ozone, nitrogen dioxide, and sulphur dioxide take place during daylight hours and during the in-leaf season—hence, the selection of coniferous or evergreen species would serve best to improve air pollution levels year round in Toronto. Junipers would serve to trap particulates with sticky plant surfaces and work year round—when not covered with snow on roofs—to reduce air contaminants through their functioning stomates.

These results might also be useful to disciplines such as urban public health and regional planners, watershed managers, neighbourhoods, parks and recreation departments, architects and urban forest and landscape management practitioners. A commitment to primary research will contribute to a growing body of Canadian evidence that will further assist green roof study, species choices and performance comparisons and contribute to a continuous quality improvement cycle in green roof research. It is hoped that this study will also provide rationale for ongoing research on the environmental benefits of green roofs as well as support the development of municipal policy to support the installation and proliferation of green roofs.


Walls of Juniper trees were chosen to represent a green wall as UFORE is able to estimate the impacts of this green wall on energy consumption, which was utilized in a parallel study. Although vines can be selected for green walls, UFORE does not simulate the impacts of vines on energy consumption. Vines would not have a significantly different impact on air quality as their LAI is similar to that of the Juniper species selected for the green walls.


Copyright information

© Springer Science+Business Media, LLC 2008