Abstract

The Advanced Research version of the Weather Research and Forecasting Model (ARW) coupled with an urban canopy model is used to investigate the potential of vegetative (green) roof technology to mitigate against ongoing climate warming and continued urban sprawl for a day representing average summer conditions in late-twenty-first-century Chicago, Illinois. Effects related particularly to human health hazards resulting from excessive heat and high pollution concentrations are emphasized. Continued expansion of the urban environment over the next century is shown to lead to an expansion of the warming signal across the metropolitan region. Widespread adoption of vegetative rooftops, through increased albedo and evapotranspiration, reduces temperatures in the urban environment by as much as 3°C, an effect similar to the simpler but less appealing alternative of employing painted or other reflective rooftop structures (e.g., white roofs). A significant limitation to the green roof approach for the case studied is that the increase in moisture resulting from transpiration leads to only marginal cooling when apparent temperatures are considered. An additional complication arises in that the reduced temperatures alter the lake-breeze circulation, potentially reducing circulation of pollutants into the city core, but also reducing natural cooling in the most urbanized areas during the climatologically warmest hours. Future work that evaluates these impacts over a broader range of synoptic settings, documents changes in the planetary boundary layer structure and attendant pollution, and considers the multiple-day dependence of these effects is needed.

1. Motivation

The roughness and radiative properties of surfaces in urban environments are often substantially different from those of outlying regions. These differences can affect local weather and climate within the urban canopy. One common example is the urban heat island, in which the reduction in vegetative surfaces within the urban area decreases the amount of cooling achieved by evaporation (e.g., Rosenzweig et al. 2006a). Likewise, cases of heat stress and stroke increase during periods of hot weather (McMichael et al. 2003) as a result of the added heat retention by urban building materials (Ackerman 1985 and references therein; Kosatsky 2005). Owing to climate warming, heat waves will likely increase in frequency and intensity, and urban populations will experience greater mortality from this effect (McMichael et al. 2003; Parry et al. 2007). Also, urban air quality is often poorer than that of the surrounding countryside owing to the high population density and associated activity (e.g., Lyons and Olsson 1973; Lyons and Cole 1976), and so an understanding of these various interactions is important for assessing impacts on human health.

Chicago, Illinois, is a major U.S. urban area and is an exemplar of the problems and opportunities facing the population in the twenty-first century. The Chicago metropolitan area has experienced dramatic growth over the past few decades, resulting in significant changes in the urban landscape with a greater usage of pavement, concrete, and other heat-retaining materials relative to natural spaces (Acevedo 1999). The continually changing landscape suggests the need for accurate land cover1 classification for use in mesoscale models to gain a reliable representation of the urban canopy. Land cover classification and parameterization have long been an important part of atmospheric modeling but have not been included at the level of detail needed to account fully for these urban effects. The urban canopy model (hereinafter UCM; Kusaka and Kimura 2004) allows for the details of the urban canopy to be taken into account by providing a means for higher-resolution land cover and land use data to be used in the model energy exchanges rather than the typical slab representation that treats the radiative properties of urban portions of the grid uniformly.

Here, by using the Advanced Research version of the Weather Research and Forecasting (WRF) Model (ARW; Michalakes et al. 2001, 2004; Wicker and Skamarock 2002; Skamarock et al. 2005; Klemp et al. 2007) coupled with the UCM, the effects of the urban environment are modeled. This is done to gain insight into the effectiveness of mitigation strategies that can be used to counter late-twenty-first-century business-as-usual climate-warming scenarios (Parry et al. 2007) as well as the effects of continued urban sprawl. The need for this is made clear by Lynn et al. (2007), who found that the Chicago area could see an increase in the average summer temperature of approximately 5°C by 2080 as a result of anthropogenic climatic change. At this level of warming, an especially hot summer day now can be considered as a proxy for a future “average” day and can thus be used to examine more closely the effects of the urban canopy in this environment.

On 15 July 2006, observed daytime temperatures in the Chicago metropolitan area ranged from 32° to 35°C, 4°–7°C above the July daily average maximum temperature. As such, this day is used to represent conditions analogous to that of an average late-twenty-first-century summer day in the business-as-usual scenario. An additional component of business-as-usual is the ongoing growth of the urban area, and therefore the effects of continued urban sprawl are also examined. The effects of strategies to mitigate the potential negative impacts of continued warming in this context are investigated. Given the increasing popularity (Fischetti 2008) and multiple positive benefits (Getter and Rowe 2006) of green roof technology, this study focuses on green roofs as the primary mitigation technique.

2. Background

Chicago, like many major U.S. metropolitan areas, continues to experience dramatic growth in both population and developed land (Acevedo 1999). By 1970, with a population of just over 6.5 million, the Chicago metropolitan area covered approximately four counties with 800 mi2 of developed land. By 1990, the population increased further to over 6.7 million and the city continued to spread with close to 1000 mi2 of developed land (Auch et al. 2004). According to the U.S. Environmental Protection Agency (EPA; Environmental Protection Agency 2007), by 2007 the population of 7 million encompassed six counties. This growth and spread in population and accompanying changes in land use can have profound impacts on native ecosystems, water and air quality, local climate and weather, natural resource availability, and other important sustainability factors (Acevedo 1999).

In July 1995, the Chicago area experienced an intense heat wave that resulted in the highest number (approximately 500) of heat wave–related deaths in the history of the United States (Whitman et al. 1997). An unseasonably warm and humid air mass affected the region for a period of five days, with highs over 37°C and dewpoints over 24°C across much of the area for at least two days (Livezey and Tinker 1996). The oppressive high and low temperatures in the city (minimum temperatures were as high as 28°C) were in part due to the abundance of heat-retaining materials and the lack of natural space to aid in cooling the air, especially in the most dense urban areas. Given this situation and in the context of expected continued climate warming and urban sprawl as detailed above, it is desirable to model meteorological conditions while specifically taking into account these effects.

Until recently, atmospheric models of all scales used a standard slab model to simulate an urban area. This technique treats portions of the domain classified as urban as a flat surface with increased roughness length and decreased albedo to approximate urban structure, an approach that often overgeneralizes the influence of urban canyons (Kusaka et al. 2001). The ARW, coupled with the UCM, allows for a more detailed simulation of the specific characteristics of the urban environment. Chen et al. (2004) and Tewari et al. (2006) have found a significant impact on weather variables in the urban region when implementing the UCM, including a considerably more detailed thermal structure across the urban domain. Such details are essential to evaluate the thermal and other climate impacts of continued large-scale warming and urban expansion on the “streetscape” where individuals live and work.

Given ongoing climatic change, adaptation strategies are of obvious interest as a means for reducing the overall negative impact of ongoing warming (Smoyer and Rainham 2001). One important technique, which is only beginning to be explored from a climate perspective, is the use of green roofs. Such roofs are increasingly being used in major cities across the United States (Fischetti 2008). Aside from the obvious improvement to the living environment through increased green space, a green roof provides cooling through evapotranspiration in the summer and insulation during the winter, thereby reducing energy costs. Chicago is the leading city in using green roof technology with a total of more than 530 000 ft2 of green roofs across the metropolitan area in 2008 (approximately 8%–16%; see industry survey online at http://www.greenroofs.org/resources/GRHC_Industry_Survey_Report_2008.pdf).

At the Chicago City Hall, 20 300 ft2 (53%) of the roof is green. The City of Chicago reports a year-round average rooftop temperature reduction of nearly 4°C (7°F) with a peak reduction of approximately 22°C (40°F) in the summer (Table 1), with attendant energy savings (see online at http://www.artic.edu/webspaces/greeninitiatives/greenroofs/ and links therein). These findings suggest that there may be substantial impacts on surrounding air temperatures as well.

Table 1.

Temperature observations for the planted and paved portions of the Chicago City Hall roof and the Cook County black tar roof. Observations were taken on 9 August 2001 at 1345 central daylight time (see description online at http://www.artic.edu/webspaces/greeninitiatives/greenroofs/ and related links).

Temperature observations for the planted and paved portions of the Chicago City Hall roof and the Cook County black tar roof. Observations were taken on 9 August 2001 at 1345 central daylight time (see description online at http://www.artic.edu/webspaces/greeninitiatives/greenroofs/ and related links).
Temperature observations for the planted and paved portions of the Chicago City Hall roof and the Cook County black tar roof. Observations were taken on 9 August 2001 at 1345 central daylight time (see description online at http://www.artic.edu/webspaces/greeninitiatives/greenroofs/ and related links).

Recent studies undertaken in the New York City area have also examined green roofs as a possible mitigation strategy for energy conservation, human health, and air quality (Rosenzweig et al. 2006b). Results indicate as much as a 0.8°C reduction in average surface temperatures could be achieved with 50% green roof coverage across the city. This level of cooling could help to minimize the negative effects of the urban heat island, including heat stress and increased heat-related mortality, and could lead to improved air quality and reduced energy consumption for air conditioning.

3. Data and methods

Standard surface reports from three stations located in the main urban portions of the Chicago area were chosen for ground-truth comparisons. These include Chicago—O’Hare airport (ORD), Chicago—Midway airport (MDW), and Northerly Island (CGX), formerly known as Meigs Field (Fig. 1). These sites were chosen specifically to capture variations between near-shore locations (CGX) and locations that are farther inland (ORD and MDW). In addition, temperatures from four other stations (DeKalb, Waukegan, Aurora, and DuPage, Illinois; see Fig. 1) were used to gain a more spatially expansive observation network for reference when using satellite-derived temperatures. In the discussions that follow, ORD and CGX will be the primary sites referenced because MDW correlates highly with ORD and the sites to the west.

Fig. 1.

Observation station locations used in the study. Shaded areas indicate counties included in the Chicago metropolitan region as defined by the U.S. EPA.

Fig. 1.

Observation station locations used in the study. Shaded areas indicate counties included in the Chicago metropolitan region as defined by the U.S. EPA.

The National Centers for Environmental Prediction–National Center for Atmospheric Research reanalysis data for 15 July 2006 are used to initialize the model and to provide the synoptic-scale scenario in which the urban environment evolves through the course of the day. As noted previously, temperatures were 4°–7°C above the July average, in keeping with Lynn et al. (2007) that such conditions could represent an “average” summer day in the late twenty-first century. Thus, this case is used here as a proxy for a business-as-usual climate-warming scenario, understanding that this is one of many synoptic patterns that might be experienced in the future. High pressure centered over the region kept prevailing winds relatively light on this day (Fig. 2), allowing an afternoon lake breeze to form and to advance inland, resulting in cooling along the lakeshore. Such breezes are a characteristic climatic feature along the western shores of Lake Michigan (e.g., Atkinson 1981), and hence any potential changes in this feature resulting from alterations to the built environment are of interest and are also assessed in this study. No precipitation was reported, but scattered cumuli formed along the lake-breeze frontal boundary.

Fig. 2.

Sea level pressure analysis for 1200 UTC 15 Jul 2006 (obtained from the National Weather Service Daily Weather Map archive).

Fig. 2.

Sea level pressure analysis for 1200 UTC 15 Jul 2006 (obtained from the National Weather Service Daily Weather Map archive).

High-resolution land cover data were obtained from the U.S. Geological Survey (USGS) 2001 National Land Cover Database (referred to hereinafter as NLCD 2001). These data are at 30-m resolution and contain detailed classifications including four subcategories of urban land use classes in addition to vegetative and other categories (Fig. 3). These high-resolution data were remapped and interpolated to the model grid, resulting in more detailed and accurate land use, specifically with respect to the developed and built-up areas within the UCM.

Fig. 3.

NLCD 2001 land cover classification for Chicago and surrounding areas.

Fig. 3.

NLCD 2001 land cover classification for Chicago and surrounding areas.

Satellite imagery obtained from the Landsat-5 Thematic Mapper (TM) satellite was used to derive surface skin temperatures across the Chicago area on 15 July 2006 (Fig. 4). Data from band 6 of this image were ingested into a geographic information system to be converted to temperature values. The conversion from radiance values to radiant temperatures requires knowledge of the emissivity across the domain to better address temperature variations of varying land cover types. Emissivity values are typically assigned based on previous studies, but as much as 1.5°C in derived temperature error can result from unaccounted differences in emissivity (Roth et al. 1989). Despite this, the sensitivity to this aspect was explored by assigning commonly used emissivity values (Oke 1987; Ishida and Kawashima 2002; Venkateswarlu et al. 2004; Stathopoulou and Cartalis 2007) chosen to correspond to each land cover type, and then skin temperatures were derived accordingly. Results were slightly warmer than skin temperatures calculated assuming uniform emissivity for all surfaces. Verification, however, is hindered without ground-truth skin temperatures. As a result, the simpler scheme that obtains these temperatures by assuming uniform emissivity is used in all analyses because it sufficiently illustrates the small-scale thermal variations across the urban area (Fig. 5).

Fig. 4.

Landsat-5 TM composite image acquired at 1630 UTC 15 Jul 2006. The image combines multiple bands in false color to illustrate the spatial and spectral extent of the satellite data. Data from band 6 are used to derive surface skin temperatures (shown in Fig. 5).

Fig. 4.

Landsat-5 TM composite image acquired at 1630 UTC 15 Jul 2006. The image combines multiple bands in false color to illustrate the spatial and spectral extent of the satellite data. Data from band 6 are used to derive surface skin temperatures (shown in Fig. 5).

Fig. 5.

Surface skin temperatures (bluest shade is 21°C, and reddest shade is 51°C) derived from band 6 of the Landsat-5 TM satellite at 120 m × 120 m resolution. The image was acquired for 1630 UTC 15 Jul 2006 (Fig. 4). The boxed area represents a point of focus for one area of intense thermal variations south of ORD. Scattered cumuli are also apparent in white.

Fig. 5.

Surface skin temperatures (bluest shade is 21°C, and reddest shade is 51°C) derived from band 6 of the Landsat-5 TM satellite at 120 m × 120 m resolution. The image was acquired for 1630 UTC 15 Jul 2006 (Fig. 4). The boxed area represents a point of focus for one area of intense thermal variations south of ORD. Scattered cumuli are also apparent in white.

The coupled WRF–UCM developed by Kusaka et al. (2001) and later revised by Kusaka and Kimura (2004) allows for the influences of urban environments to be accounted for in the model. The UCM is a single-layer canopy model, meaning that exchanges between the base of the atmospheric model (WRF) occur only at the top of the canyons and roofs (Masson 2006). The single-layer scheme takes into account important attributes of the urban canyon environment including canyon orientations, diurnal change of solar azimuth angles, and shadowing effects from buildings (Kusaka et al. 2001). Variables such as moisture availability, heat capacity, albedo, emissivity, and building heights are adjustable to represent a specific urban area whose size is determined by the atmospheric model grid spacing. Then, the temperature, wind, and humidity within the canopy as well as skin temperature and fluxes from all surfaces are resolved. The temperature resolved by the UCM is defined as the air temperature at the top of the canopy that is exchanged with the lowest level of the atmosphere, encompassing all factors within the urban canopy itself. Therefore, the UCM produces a temperature that provides a measure of the overall impact of the canopy but does not specifically resolve conditions experienced by a hypothetical person at “street level.” Lynn et al. (2009) present such a method in the context of their study, which focused on the percent reflective roof coverage and albedo of that coverage, but that study did not include canyon structure (such as building shadowing effects), nor did it investigate potential effects of additional moistening of the environment from increased vegetation, as detailed below.

The model is run for a single domain of 332 × 379 grid points, using a 1-km spacing that encompasses the Chicago area. Specific physics schemes and parameterizations used in the model configuration include the WRF Single-Moment (WSM) 3-class simple ice microphysics, Rapid Radiative Transfer Model longwave radiation, Dudhia shortwave radiation, Monin–Obukhov surface layer, Mellor–Yamada–Janjic boundary layer, Kain–Fritsch cumulus, and the “Noah” land surface model (LSM) surface physics. The Noah LSM is the land surface option used to couple and employ the UCM. Six 24-h simulations are run for the period 0600 UTC 15–16 July 2006.

The first model run (hereinafter called Basic) is a standard model configuration without coupling the UCM. This uses the standard slab-model land use classification (one category for all urban land use) provided by the terrestrial 30-s data. This means that, for this run, all radiative properties and urban geometries remain uniform across the urban portions of the grid.

The second run (hereinafter called Urban) employs the UCM as well as the high-resolution land cover data from the NLCD 2001 to gain a more comprehensive representation of the thermal structures within the urban canopy. The four urban categories of the NLCD 2001 are reclassified to the three detailed urban categories (Table 2) that are used in the UCM (Tewari et al. 2007). Urban categories 21 and 22 from the NLCD are remapped to category 31 of the UCM, and categories 23 and 24 from the NLCD are remapped to 32 and 33 of the UCM, respectively. This new land cover dataset is then ingested and interpolated to the model grid (Fig. 6). All grid points covered by this new dataset use the UCM urban classification (31, 32, and 33), indicating grid points that will be resolved by the UCM, whereas those not covered use the default land cover categories.

Table 2.

Description of the four NLCD 2001 and three UCM urban land categories.

Description of the four NLCD 2001 and three UCM urban land categories.
Description of the four NLCD 2001 and three UCM urban land categories.
Fig. 6.

New urban classification used by the UCM, resulting from interpolation of remapped NLCD data to the model grid. Shown are low-intensity residential (light gray), high-intensity residential (dark gray), and industrial/commercial (black).

Fig. 6.

New urban classification used by the UCM, resulting from interpolation of remapped NLCD data to the model grid. Shown are low-intensity residential (light gray), high-intensity residential (dark gray), and industrial/commercial (black).

A third run was performed to examine the impact of vegetative roofs on the overall thermal structure in the urban canopy (hereinafter called GreenRef). To simulate the effect of green roofs in the UCM, an equivalent albedo, as defined by Rosenzweig et al. (2006b), that includes the impacts of both evaporative cooling and shortwave reflection from a vegetative surface is used. In that study, vegetative rooftops were found to have an emissivity value between 0.9 and 1.0 and an equivalent albedo between 0.7 and 0.85, similar to that of white paint (Rosenzweig et al. 2006b). In accord with this finding, the albedo for all rooftops across the urban domain is changed to 0.8 to simulate the properties of a green roof. By applying this increased albedo to every rooftop at every urban grid point, a maximized impact of green roof technology is explored. Obviously, depending on the extent of actual green roof technology adoption, real impacts will be less than those reported here.

While the GreenRef run uses an equivalent albedo to simulate green roofs, this method is equivalent to a rooftop with increased reflectance properties as opposed to an actual vegetative covering. The additional moisture added to the environment by the increased vegetation is neglected. In physical terms, it is the process of evapotranspiration that results in this equivalent albedo. In this way, resultant increases in humidity and apparent temperature,2 defined in Steadman (1979) as what the temperature feels like to a typical human given the combination of dry-bulb temperature and vapor pressure, could impact human health by negatively affecting the body’s ability to cool itself naturally. As such, another model run (hereinafter called GreenVeg) that directly accounts for such changes is included. The albedo of the rooftops is set to 0.18, representative of actual vegetative reflectance, and the moisture availability of the roofs is increased to 0.25.

The fifth run (hereinafter called Sprawl) takes a business-as-usual approach to urban sprawl by expanding the urban domain based on current growth projections. Wang and Wen (2002) found the projected growth rate in spatial coverage in the Chicago area from 2000 to 2020 was about 16%. This growth rate was applied to the core metropolitan area to estimate the potential urban growth in 2100. Five categories from the NLCD were chosen to be reclassified to urban categories based on the likelihood of these areas being developed: Barren land (31), scrub/shrub (52), grasslands (71), pasture/hay (81), and cultivated crops (82). Forested and wetland regions were not reclassified because there are currently numerous parks and areas that are preserved or protected, and it is assumed that these areas will remain outside of consideration for development. The five nonurban categories noted above were all reclassified to the low-intensity residential (31) category of the UCM, resulting in a new land cover dataset with expanded urban growth. This is then interpolated to the model grid, and a larger urban domain is created for use in the Sprawl case (Fig. 7).

Fig. 7.

As in Fig. 6, but representing an urban sprawl scenario in the late-twenty-first century.

Fig. 7.

As in Fig. 6, but representing an urban sprawl scenario in the late-twenty-first century.

The final run (hereinafter called GSprawl) simulates the adoption of green roof technology across the domain of the Sprawl case to examine maximal mitigation of the combined impact of urban sprawl and climatic change. The albedo parameters within the UCM are altered again as with the GreenRef run to represent the equivalent albedo of green roofs.

4. Results

Owing to a mismatch between the model and reanalysis grid resolutions, there are temperature errors evident in the first few hours of the Basic and Urban simulations. Skamarock (2004) shows through an analysis of kinetic energy spectra that the majority of mesoscale model spinup is accomplished within the first 6 h. As such, analysis of all model runs in this paper is begun at 1200 UTC.

a. Basic

The Basic run provides a reference for standard model behavior to ensure that the overall weather characteristics are adequately captured. At ORD, a general cold bias of about 4°C exists for the Basic run through midday, slowly lowering to approximately 1°C toward evening (Fig. 8). The run simulates a lake-breeze feature at CGX with temperatures cooling slightly at 1900–0100 UTC. This agrees with observations but the model feature is not as robust and is somewhat late (1900 UTC as compared with the observed 1600 UTC). Simulated temperatures remain near 31°C during this time while observations indicate more cooling. While some quantitative discrepancies exist between observations and model output, the simulation demonstrates the ability to simulate the primary characteristics of a typical summer day in Chicago in the late twenty-first century.

Fig. 8.

Temperatures from the Urban run in comparison with the Basic run and observations at (top) ORD and (bottom) CGX.

Fig. 8.

Temperatures from the Urban run in comparison with the Basic run and observations at (top) ORD and (bottom) CGX.

b. Urban

At ORD, the Urban run warms faster than the Basic run but remains 2°–3°C cooler until about 1700 UTC. Temperatures peak near observations just at or slightly above 35°C between 2000 and 2200 UTC (Fig. 8), but there is a delay in the evolution of the model temperature trends. Temperatures perform much better relative to the Basic run, however, indicating that the added structure of the UCM environment has a large impact on the daytime heating that occurs within the urban canopy. During the nighttime hours at ORD, the Urban run remains up to 4°C warmer than observations toward the end of the simulation. This warm bias may in part be due to additional heat retention within the urban canopy. A delay of 3–4 h in the lake-breeze formation is seen again at CGX and also continues to produce a weaker lake breeze in comparison with observations (Fig. 8).

To understand this result, lake observations and model output were examined to determine whether the model was reproducing an adequate representation of the lake environment. Based on observations from the southern buoy in Lake Michigan (buoy 45007) and confirmed by high-resolution water surface temperature data from the Landsat image, the lake temperatures on this day varied from 21° to 23°C as compared with a model lake temperature of about 16°C. Such biases often exist when initializing mesoscale models with coarse-resolution reanalysis data in the Great Lakes region, since water temperatures lack sufficient detail in such datasets. To explore the impact of this cold bias, the Urban case was rerun with model lake temperatures increased to 21°C. The impact of this change is minimal relative to the original results from the Urban run (not shown), likely owing to the generally southwesterly synoptic flow on that day, which mostly kept the lake breeze and its effects pinned to the shoreline and directly offshore.

Despite these difficulties, overall the Urban run performed qualitatively well in consideration of the objectives of this study. A comparison is performed between the Urban and Basic runs to quantify the differences in temperatures simulated when adding urban structure. This difference or offset is calculated as the sum of the ratios of Urban temperature relative to the Basic temperature normalized over total grid points and is found to be 1.0033, or approximately 1°C overall warming across the urban domain. Although this average difference is relatively small, the root-mean-square difference (RMSD) of 3.3°C suggests high spatial variability across the urban landscape resulting from the added urban structure using the UCM. A time series analysis of the offset and RMSD illustrates the variability across the urban landscape, especially in the afternoon hours when temperatures from the UCM appear to vary greatly relative to the standard slab model (Fig. 9). Such spatial variations are clearly seen within the actual urban landscape based upon the satellite-derived temperatures (Fig. 5). For example, one area of hot and cool spots in close proximity from Fig. 5 indicates a difference in skin temperatures of 27°C, with the center of the hot spot reaching 51°C owing to the varying land cover and associated building materials.

Fig. 9.

(top) RMSD and (bottom) bias between the Urban and Basic runs across urban portions of the grid.

Fig. 9.

(top) RMSD and (bottom) bias between the Urban and Basic runs across urban portions of the grid.

c. GreenRef

Implementation of green roofs across the urban domain results in a substantial cooling effect across the urban landscape. Temperatures at ORD between 1900 and 2300 UTC were 2°–3°C cooler than the Urban run where temperatures peak just above 35°C (Fig. 10), and during the late evening and night hours there was about a 1°C cooling effect. Directly along the lakeshore at CGX, the effect of the green roofs was 1°–2°C of cooling leading up to 1900 UTC, at which time air temperatures from the GreenRef run begin to exceed the Urban run (Fig. 10). This results from the modification of the lake breeze associated with green roof implementation, a difference that is clearly seen when the spatial distribution of temperature differences between the Urban and GreenRef runs is examined (Fig. 11).

Fig. 10.

Temperatures from the GreenVeg (denoted “Green2” in the key) run in comparison with the Urban and GreenRef (denoted “Green”) runs at (top) ORD and (bottom) CGX.

Fig. 10.

Temperatures from the GreenVeg (denoted “Green2” in the key) run in comparison with the Urban and GreenRef (denoted “Green”) runs at (top) ORD and (bottom) CGX.

Fig. 11.

Temperature difference between the Urban and GreenRef runs at 2100 UTC (negative/positive values indicate that the GreenRef run is cooler/warmer than the Urban run).

Fig. 11.

Temperature difference between the Urban and GreenRef runs at 2100 UTC (negative/positive values indicate that the GreenRef run is cooler/warmer than the Urban run).

The substantial direct thermal impact of this mitigation strategy is readily apparent across the urban landscape during the afternoon hours, with a cooling of over 3°C indicated in the core urban center and from 1° to 5°C of warming along the lakeshore associated with the weakening of the lake breeze (Fig. 11). The reduction in lake-breeze penetration is also evident in the wind fields (Fig. 12), which show the lake breeze pinned to the shoreline. In fact, a land breeze is apparent north of the city core owing to the reduced differential heating from green roof coverage and the prevailing synoptic flow from the west. The net impact, then, is for a more widespread cooling across the city but a loss of some of the natural cooling along the lakeshore associated with the lake-breeze circulation. This may be of considerable importance when considering the net effect on energy usage because the loss of natural cooling along the highly urbanized lakeshore areas would likely result in the need for increased use of air conditioning.

Fig. 12.

Wind field for the (top) Urban and (bottom) GreenRef runs at 2100 UTC. Length of barb is such that a 5 m s−1 wind speed will exactly reach the tail of the adjacent vector.

Fig. 12.

Wind field for the (top) Urban and (bottom) GreenRef runs at 2100 UTC. Length of barb is such that a 5 m s−1 wind speed will exactly reach the tail of the adjacent vector.

While the significant cooling over land is evident in this case, the impact of potential warming of the surface waters of the Great Lakes in the context of ongoing climatic warming was not considered. To address this aspect, we consider the lake-breeze index (LBI) defined by Lyons (1972) as a function of the 4-hPa isobar spacing and the difference between inland air temperature and surface water temperature. Using various LBI thresholds that were determined by Lyons (1972) to be indicative of conditions preventing lake-breeze formation, one can estimate the water temperature that must be reached to produce that state. In the GreenRef case, that of the weakest pressure gradient, water temperatures must increase to approximately 31°C to assure no lake breeze. Assuming an upper limit of water temperature increase to be approximately equal to that of the future projected daily average summer air temperature, 27°–30°C, it seems likely that lake-breeze formations will decrease in frequency rather than be wholly eliminated for these conditions.

d. GreenVeg

As noted previously, an additional effect of vegetative roofs is likely to be an increase in atmospheric moisture through transpiration. Apparent temperatures then become a concern given the impact that increased humidity has on human health. When this moistening is accounted for, overall cooling relative to the Urban run is similar but just slightly less than that of the GreenRef run (Fig. 10). As expected, however, dewpoints increase relative to the Urban run, and this increase in humidity appears to limit substantially the cooling effect of green roofs, with a reduction of at most 1°C (Fig. 13) based on apparent temperature. This suggests an important potential limitation to the amount of cooling that can be realized through full adoption of vegetative rooftops when considering human health impacts.

Fig. 13.

Apparent temperatures from the GreenVeg (Green2) and Urban runs at (top) ORD and (bottom) CGX.

Fig. 13.

Apparent temperatures from the GreenVeg (Green2) and Urban runs at (top) ORD and (bottom) CGX.

e. Sprawl

The Sprawl run addresses the impact of continued urban development over the course of the next century. Implementing a more spatially expansive urban environment results in an increase in temperatures across the newly urbanized portions of the grid by 1°–3°C during the afternoon hours (not shown). Not surprising, however, is that there was little impact on temperatures in areas that were already previously classified as urban as seen at ORD and CGX (Fig. 14) where temperatures differ only slightly. The primary impact of continued sprawl is the increased area of warmer temperatures as the urban area expands across previously less developed areas.

Fig. 14.

Temperatures from the Sprawl and Urban runs at (top) ORD and (bottom) CGX.

Fig. 14.

Temperatures from the Sprawl and Urban runs at (top) ORD and (bottom) CGX.

f. GSprawl

The substantial warming that would occur in the more spatially expansive urban environment again gives rise to the need for a possible mitigation strategy. As with the Urban and GreenRef cases, GSprawl is run with a modified roof albedo across the urban domain to represent 100% green roof coverage. A cooling of 2°–3°C occurs at ORD (Fig. 15), similar to the cooling achieved in the GreenRef case relative to the Urban scenario. In a similar way, temperatures at CGX behave as with the GreenRef case, producing temperatures greater than the Sprawl run between 2100 and 0100 UTC (Fig. 15). This counterintuitive result is again owing to the loss of cooling that would otherwise have been produced by the lake breeze. Across the newly urbanized portions of the domain, temperatures cool by approximately 3°–4°C (Fig. 16). As is also evident in the GreenRef run but here is enhanced, the wind fields indicate a substantial alteration to the lake-breeze circulation, effectively creating a land breeze that is especially pronounced north of the city core (Fig. 17). The reduction in the land–water thermal contrast owing to the extensive green roof coverage allows the synoptic flow from the west to prevail. Although not examined directly, one should expect that the effective cooling achieved by this approach will be lessened considerably by increases in moisture, as is found in GreenVeg.

Fig. 15.

Temperatures from the Sprawl and GSprawl runs at (top) ORD and (bottom) CGX.

Fig. 15.

Temperatures from the Sprawl and GSprawl runs at (top) ORD and (bottom) CGX.

Fig. 16.

As in Fig. 11, but between the Sprawl and GSprawl runs (negative/positive values indicate that the GSprawl run is cooler/warmer than the Sprawl run).

Fig. 16.

As in Fig. 11, but between the Sprawl and GSprawl runs (negative/positive values indicate that the GSprawl run is cooler/warmer than the Sprawl run).

Fig. 17.

As in Fig. 12, but for the (top) Sprawl and (bottom) GSprawl runs.

Fig. 17.

As in Fig. 12, but for the (top) Sprawl and (bottom) GSprawl runs.

5. Conclusions and future work

Significant warming occurs within the urban canopy, giving rise to the potential for negative impacts on human health. With the further expansion of the urban environment over the next century based on recent trends, these negative consequences likewise expand across the region. Vegetative rooftops, or green roofs, offer a potentially attractive mitigation strategy by increasing equivalent albedo and thereby reducing temperatures in the urban environment. Green roofs also provide a variety of additional benefits in addition to aesthetic improvements, including decreased or delayed storm runoff and reduced urban flooding, provision of new wildlife habitats, increased energy conservation, and reduced heating/cooling costs (Getter and Rowe 2006). Further, vegetation can filter out some air and water pollutants and thereby improve urban air and water quality (see information online at http://www.epa.gov/heatisland/mitigation/greenroofs.htm).

This makes such efforts an appealing strategy when compared with simpler white roofs (painted or other reflective rooftop structures), which simply increase the albedo of the rooftop and thereby reflect solar energy away from the building. Nonetheless, there are some significant limitations to the green roof approach. Most important, surface dewpoints increase with extensive use of green roofs owing to the increased moisture from transpiration. Although this occurs in conjunction with temperature decreases of as much as 3°C, apparent temperatures reveal only minimal reduction as a result of the increased moisture. This suggests that a more optimal strategy might include an appropriate mix of vegetative roofs and other roofs composed of light-colored materials (e.g., white roofs), which retain the radiative advantage and mitigate against increased humidity.

An additional complication is that the adoption of green roofs in either a static or expanding urban area promotes changes in the lake-breeze circulation. The lake breeze can capture pollutants in the boundary layer, including ozone (e.g., Lyons and Olsson 1973; Lyons and Cole 1976), and recirculate them through the metropolitan area (e.g., Harris and Kotamarthi 2005). In that sense, the reduced inland penetration of the lake breeze could potentially improve air quality, but this loss also results in a reduction in natural cooling along the lakeshore during the climatologically warmest hours of the day. This cooling routinely occurs over some of the most densely populated and urbanized portions of the city, perhaps leading to increased net energy demand, despite the broad-scale cooling across the urban domain. Analysis of future warming scenarios suggests that the lake-breeze frequency may be reduced but not eliminated, and so these impacts must be considered. Future work should examine synoptic scenarios in which inland lake-breeze penetration is more prominent (such as those with more southerly or weaker prevailing winds than the case studied here) to document further the effects of green roofs on the lake breeze.

Additional research may focus on how temperatures within the urban canopy respond to an extended period of high temperatures with and without green roofs since the impact on human health is typically delayed during a heat wave (Whitman et al. 1997). Another important consideration is the effect of extensive green roof coverage on the height of the planetary boundary layer (PBL), since in the simulations investigated here, this mitigation strategy led to lower heights. Such changes could have a negative impact on air quality since a lower PBL height would reduce the ability of pollutants to diffuse and thereby increase surface pollutant concentrations.

Anthropogenic heating is itself a significant factor in the urban environment (Ichinose et al. 1999; Sailor and Lu 2004), and these effects should be quantified to extend the results obtained here. This aspect was not available as an output of the WRF–UCM at the time of this research but is now available, and so such an extension of this research is feasible. Owing to the complexity of the meteorological and sociological factors relevant to the urban setting and the inherent uncertainties in many of these aspects, it was not possible to do a complete analysis of the effects of urban-heating mitigation strategies in the face of ongoing climatic change. A suite of such studies, however, tailored to the specific directions detailed here would contribute greatly to a deeper understanding of this issue.

Acknowledgments

Research presented here was supported in part by the University of Wisconsin—Milwaukee RGI 1 01-0011. We thank the anonymous reviewers for their detailed and carefully considered comments.

REFERENCES

REFERENCES
Acevedo
,
W.
,
1999
:
Analyzing land use change in urban environments.
U.S. Geological Survey Fact Sheet 188-99, 4 pp. [Available online at http://landcover.usgs.gov/urban/info/factsht.pdf]
.
Ackerman
,
B.
,
1985
:
Temporal march of the Chicago heat island.
J. Climate Appl. Meteor.
,
24
,
547
554
.
Atkinson
,
B. W.
,
1981
:
Meso-Scale Atmospheric Circulations.
Academic Press, 405 pp
.
Auch
,
R.
,
J.
Taylor
, and
W.
Acevedo
,
2004
:
Urban growth in American cities: Glimpses of U.S. urbanization.
U.S. Geological Survey Circular 1252, 60 pp. [Available online at http://pubs.usgs.gov/circ/2004/circ1252/pdf/circ1252.pdf]
.
Chen
,
F.
,
H.
Kusaka
,
M.
Tewari
,
J.-W.
Bao
, and
H.
Hirakuchi
,
2004
:
Utilizing the coupled WRF/LSM/urban modeling system with detailed urban classification to simulate the urban heat island phenomena over the greater Houston area.
Preprints, Fifth Conf. on Urban Environment, Vancouver, BC, Canada, Amer. Meteor. Soc., 9.11. [Available online at http://ams.confex.com/ams/AFAPURBBIO/techprogram/paper_79765.htm]
.
Environmental Protection Agency
,
2007
:
EPA urban heat island pilot project city profile: Chicago.
U.S. EPA Archived Web Site, 10 pp. [Available online at http://www.epa.gov/heatisland/pilot/archives/Chicago.pdf]
.
Fischetti
,
2008
:
Green roofs: Living cover.
Sci. Amer.
,
208
,
104
105
.
Getter
,
K. L.
, and
D. B.
Rowe
,
2006
:
The role of extensive green roofs in sustainable development.
HortScience
,
41
,
1276
1285
.
Harris
,
L.
, and
V. R.
Kotamarthi
,
2005
:
The characteristics of the Chicago lake breeze and its effects on trace particle transport: Results from an episodic event simulation.
J. Appl. Meteor.
,
44
,
1637
1654
.
Ichinose
,
T.
,
K.
Shimodozono
, and
K.
Hanaki
,
1999
:
Impact of anthropogenic heat on urban climate in Tokyo.
Atmos. Environ.
,
33
,
3897
3909
.
Ishida
,
T.
, and
S.
Kawashima
,
2002
:
Geostatistical analyses of Landsat Thematic Mapper–derived surface temperature during winter nights.
J. Appl. Meteor.
,
41
,
931
940
.
Klemp
,
J. B.
,
W. C.
Skamarock
, and
J.
Dudhia
,
2007
:
Conservative split-explicit time integration methods for the compressible nonhydrostatic equations.
Mon. Wea. Rev.
,
135
,
2897
2913
.
Kosatsky
,
T.
,
2005
:
The 2003 European heat waves.
Eur. Surveill.
,
10
,
48
49
.
Kusaka
,
H.
, and
F.
Kimura
,
2004
:
Coupling a single-layer urban canopy model with a simple atmospheric model: Impact of urban heat island simulation for an idealized case.
J. Meteor. Soc. Japan
,
82
,
67
80
.
Kusaka
,
H.
,
H.
Kondo
,
Y.
Kikegawa
, and
F.
Kimura
,
2001
:
A simple single-layer urban canopy model for atmospheric models: Comparison with multi-layer and slab models.
Bound.-Layer Meteor.
,
101
,
329
358
.
Livezey
,
R. E.
, and
R.
Tinker
,
1996
:
Some meteorological, climatological, and microclimatological considerations of the severe U.S. heat wave of mid-July 1995.
Bull. Amer. Metoer. Soc.
,
77
,
2043
2054
.
Lynn
,
B. H.
,
R.
Healy
, and
L. M.
Druyan
,
2007
:
An analysis of the potential for extreme temperature change based on observations and model simulations.
J. Climate
,
20
,
1539
1554
.
Lynn
,
B. H.
, and
Coauthors
,
2009
:
A modification to the NOAH LSM to simulate heat mitigation strategies in the New York City metropolitan area.
J. Appl. Meteor. Climatol.
,
48
,
199
216
.
Lyons
,
W. A.
,
1972
:
The climatology and prediction of the Chicago lake breeze.
J. Appl. Meteor.
,
11
,
1259
1270
.
Lyons
,
W. A.
, and
L. E.
Olsson
,
1973
:
Detailed mesometeorological studies of air pollution dispersion in the Chicago lake breeze.
Mon. Wea. Rev.
,
101
,
387
403
.
Lyons
,
W. A.
, and
H. S.
Cole
,
1976
:
Photochemical oxidant transport: Mesoscale lake breeze and synoptic-scale aspects.
J. Appl. Meteor.
,
15
,
733
743
.
Masson
,
V.
,
2006
:
Urban surface modeling and the mesoscale impact of cities.
Theor. Appl. Climatol.
,
84
,
35
45
.
McMichael
,
A. J.
,
D. H.
Campbell-Lendrum
,
C. F.
Corvalán
,
K. L.
Ebi
,
A.
Githeko
,
J. D.
Scheraga
, and
A.
Woodward
,
2003
:
Climate Change and Human Health—Risks and Responses.
World Health Organization, 322 pp
.
Michalakes
,
J.
,
S.
Chen
,
J.
Dudhia
,
L.
Hart
,
J.
Klemp
,
J.
Middlecoff
, and
W.
Skamarock
,
2001
:
Development of a next generation regional Weather Research and Forecast model: Developments in teracomputing.
Proceedings of the Ninth ECMWF Workshop on the Use of High Performance Computing in Meteorology, W. Zwieflhofer and N. Kreitz, Eds., World Scientific, 269–276
.
Michalakes
,
J.
,
J.
Dudhia
,
D.
Gill
,
T.
Henderson
,
J.
Klemp
,
W.
Skamarock
, and
W.
Wang
,
2004
:
The Weather Research and Forecast model: Software architecture and performance.
Proceedings of the 11th ECMWF Workshop on the Use of High Performance Computing In Meteorology, W. Zwieflhofer and G. Mozdzynski, Eds., World Scientific, 156–168
.
Oke
,
T. R.
,
1987
:
Boundary Layer Climates.
Methuen, 435 pp
.
Parry
,
M. L.
,
O. F.
Canziani
,
J. P.
Palutikof
,
P. J.
van der Linden
, and
C. E.
Hanson
,
Eds.
2007
:
Climate Change 2007: Impacts, Adaptation and Vulnerability.
Cambridge University Press, 976 pp
.
Rosenzweig
,
C.
,
W.
Solecki
,
L.
Parshall
,
S.
Gaffin
,
B.
Lynn
,
R.
Goldberg
,
J.
Cox
, and
S.
Hodges
,
2006a
:
Mitigating New York City’s heat island with urban forestry, living roofs, and light surfaces.
Preprints, Sixth Symp. on the Urban Environment, Atlanta, GA, Amer. Meteor. Soc., J3.2. [Available online at http://www.giss.nasa.gov/research/news/20060130/103341.pdf]
.
Rosenzweig
,
C.
,
S.
Gaffin
, and
L.
Parshall
,
Eds.
2006b
:
Green roofs in New York metropolitan region: Research report.
Columbia University Center for Climate Systems Research and NASA Goddard Institute for Space Studies, 59 pp
.
Roth
,
M.
,
T. R.
Oke
, and
W. J.
Emery
,
1989
:
Satellite-derived urban heat islands from three coastal cities and the utilization of such data in urban climatology.
Int. J. Remote Sens.
,
10
,
1699
1720
.
Sailor
,
D. J.
, and
L.
Lu
,
2004
:
A top-down methodology for developing diurnal and seasonal anthropogenic heating profiles for urban area.
Atmos. Environ.
,
38
,
2737
2748
.
Skamarock
,
W. C.
,
2004
:
Evaluating mesoscale NWP models using kinetic energy spectra.
Mon. Wea. Rev.
,
132
,
3019
3032
.
Skamarock
,
W. C.
,
J. B.
Klemp
,
J.
Dudhia
,
D. O.
Gill
,
D. M.
Barker
,
W.
Wang
, and
J. G.
Powers
,
2005
:
A description of the Advanced Research WRF version 2.
NCAR Tech. Note NCAR/TN-468+STR, 88 pp
.
Smoyer
,
K. E.
, and
D. B.
Rainham
,
2001
:
Beating the heat: Development and evaluation of Canadian hot weather health-response plan.
Environ. Health Perspect.
,
109
,
1241
1248
.
Stathopoulou
,
M.
, and
C.
Cartalis
,
2007
:
Daytime urban heat islands from Landsat ETM+ and Corine land cover data: An application to major cities in Greece.
Sol. Energy
,
81
,
358
368
.
Steadman
,
R. G.
,
1979
:
The assessment of sultriness. Part I: A temperature–humidity index based on human physiology and clothing science.
J. Appl. Meteor.
,
18
,
861
873
.
Tewari
,
M.
,
F.
Chen
, and
H.
Kusaka
,
2006
:
Implementation and evaluation of a single-layer urban canopy model in WRF/Noah.
Extended Abstracts, Seventh WRF Users’ Workshop, Boulder, CO, NCAR Mesoscale and Microscale Meteorology Division, 5 pp. [Available online at http://www.mmm.ucar.edu/wrf/users/workshops/WS2006/abstracts/Session05/5_6_Tewari.pdf]
.
Tewari
,
M.
,
F.
Chen
,
H.
Kusaka
, and
S.
Miao
,
2007
:
Coupled WRF/unified Noah/urban-canopy modeling system.
.
Venkateswarlu
,
C.
, and
Coauthors
,
2004
:
Digital analysis of thermal infrared imagery using temperature mapping.
Proc. Conf. on Information Technology: Coding and Computing (ITCC 2004), Vol. 2, Las Vegas, NV, IEEE Computer Society Task Force on Information Technology for Business Applications, 682–688
.
Wang
,
Y. Q.
, and
Y.
Wen
,
2002
:
Spatial diffusion modeling in simulation of suburban sprawl: A case study in the Chicago metropolitan region.
NASA Land-Cover and Land-Use Change Science Team Meeting Poster, 1 p. [Available online at ftp://delphi.geog.umd.edu/LCLUC/ScienceTeamMtg/2002/Poster_WangYQ2002.pdf]
.
Whitman
,
S.
, and
Coauthors
,
1997
:
Mortality in Chicago attributed to the July 1995 heat wave.
Amer. J. Public Health
,
82
,
151
158
.
Wicker
,
L. J.
, and
W. C.
Skamarock
,
2002
:
Time-splitting methods for elastic models using forward time schemes.
Mon. Wea. Rev.
,
130
,
2088
2097
.

Footnotes

Corresponding author address: Paul J. Roebber, Professor, Atmospheric Sciences Group, Dept. of Mathematical Sciences, University of Wisconsin—Milwaukee, 3200 North Cramer Ave., Milwaukee, WI 53211. Email: roebber@uwm.edu

1

Contrary to the usual meteorological practice but to be more correct, we refer to land cover as the general term that describes what is covering the ground (e.g., urban, forest, grassland, or water) and land use to describe for what the land is being used (e.g., high-density residential buildings or park).

2

Apparent temperature is used in this analysis because it takes into consideration humidity and is also commonly used and understood by the public as a measure of what the air feels like on a hot and humid day. Another means by which to measure this would be through the use of “effective temperature” as defined by Lynn et al. (2009), which takes into account a person’s radiative and thermal energy balance in addition to the ambient temperature.