Building and Environment 197 (2021) 107792

Contents lists available at ScienceDirect

Building and Environment

journal homepage: http://www.elsevier.com/locate/buildenv

The seasonal microclimate trends of a large scale extensive green roof

Lauren Smalls-Mantey, Franco Montalto *
Department of Civil Architectural and Environmental Engineering, Drexel University, 3141 Chestnut Street, Philadelphia, PA, 19104, USA

A R T I C L E I N F O A B S T R A C T

Keywords: At nearly 27,500 m2, the Jacob K. Javits Convention Center (JJCC) located in New York City, hosts one of the
Urban Heat Island largest extensive green roofs in the United States. This paper explores three years of fine scale microclimate data
Green roof collected at the JJCC green roof and its potential ability to reduce the urban heat island intensity (UHII). Surface
Microclimate, Urban sistainability, climate
energy fluxes and microclimate parameters on four different surfaces are analyzed before and after installation of
the southern section of the green roof, offering a unique before/after, test/control study. The results indicate that
the temperatures of the air above the green roof, and its exterior surface are different (e.g. lower) than those
measured above and on, respectively, the black roof that preceded it. Differences in the maximum daytime air
and surface temperature between the black and green roof were 1.80 ◦ C and 18.4 ◦ C, respectively. Installation of
the green roof increased evapotranspiration, modifying the roof’s surface energy balance, and reduced the
median summer nighttime UHII (compared to the pedestrian level station) by 0.91 ◦ C. Though microclimatic
conditions on two sections of the green roofs vary somewhat, the research findings generally support the
statement that green roofs are a moderately effective strategy for mitigating the UHI effect.

1. Introduction 7,8,9,10,11]. Such processes are mediated through (1) the shade pro­
vided by rooftop growing media and plants, (2) convective cooling
Urbanization often involves the replacement of natural landscapes brought about through the process of evapotranspiration (ET) [11] and
with impervious surfaces, such as roofs and roads. Because the thermal (3) reflectance of solar radiation by high albedo leaf surfaces [12].
properties of built and natural surfaces differ significantly, urbanization Research by Peng [13], for example, suggests that green roofs can reduce
leads to a modification of the urban energy balance, including elevation surface temperatures by 15–45 ◦ C and the near surface air temperature
of urban temperatures above those of nearby rural areas, a difference by 2–5 ◦ C. Speak [14] and Susca [15] have also reported cooling re­
measured in terms of the urban heat island intensity (UHII) [1]. New ductions in this range. Studies that compare a green roof to a traditional
York City’s UHII has been estimated as 4 ◦ C and 3 ◦ C in the summer/­ roof structure include Getter [16] who reported air temperature directly
autumn and winter/spring seasons, respectively [2]. Due to the greater above the green roof to be 5 ◦ C lower than on a white gravel roof, with
relative thermal mass of urban surfaces, and their tendency to release peak temperature differences reaching as much as 20 ◦ C in the summer.
heat more slowly than natural surfaces, the greatest UHII is typically In another study, Qin [17], reports the green roof reduced surface
observed between the late afternoon and night [1,3]. Prolonged frequent temperature and air temperatures by an average of 7.3 ◦ C and 0.5 ◦ C,
nighttime UHIIs are considered risks to human health, since they reduce respectively. Jim and Peng [18] reported a 0.7 ◦ C air temperature dif­
the amount of time that the body has to recover from daytime high ference between the original and green roof at 10 cm but did not observe
temperatures [4]. a significant effect at 1.6 m. Alvizuri et al. [5], used thermal imaging to
To bring about a more sustainable form of urban development, quantify the ability of a green roof on top of the New York City’s Jacob
especially in the context of climate change, urban designers are in search K. Javits Convention Center to provide buffer thermal fluxes through the
of strategies like green (e.g. vegetated) roofs that can alter the local roof, finding that its exterior surface was >16 ◦ C cooler than a black
microclimate so as to partially mitigate some of the urban heat island bitumen roof, and 5–10 ◦ C cooler than an adjacent sidewalk surface
(UHI) effect. Previous work has documented the ability of green roofs to during the warm months.
provide thermal buffering, to increase the energy efficiency of building Though the body of green roof research is growing, thermal studies
systems, and to reduce their exterior surface and air temperatures [1,5,6, often focus on a very limited spatial area [19], are short-term, and do not

* Corresponding author.
E-mail address: [email protected] (F. Montalto).

https://doi.org/10.1016/j.buildenv.2021.107792
Received 29 June 2020; Received in revised form 6 March 2021; Accepted 7 March 2021
Available online 12 March 2021
0360-1323/© 2021 Published by Elsevier Ltd.
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

compare pre- and post- green roof observations [20,21] with those of a flux analysis is published separately in Alvizuri et al. [5].
control. In addition, most green roof thermal research utilizes modeling In July of 2013, four climate stations 3 m high were positioned (1) on
and not monitoring results, and does not investigate the influence of the the newly completed, north extensive green roof, hereafter referred to as
green roof on street level air temperature, and does not report changes in the “north green roof” (NGR), (2) on the south roof which still had a
air temperature before and after installation. The present study fills black asphalt bitumen surface, referred to as the “black roof” (BR), (3)
some of these gaps by studying the microclimate and energy fluxes of a on the metal reflective roof of a separate JJCC building located imme­
large scale extensive green roof built on top of the Jacob K. Javits diately to the north of the NGR, termed the “metal roof” (MR) and (4) on
Convention Center (JJCC) in New York City (Fig. 1), in an attempt to a light pole mounted above the side walk to the east of the JGR on 11th
quantify its role in reducing the UHII. It extends the work of Alvizuri avenue at street level, referred to as the “street level station” (SLS).
et al. [5], who quantified the ability of this particular green roof to By June 2014, the BR had been converted to an extensive green roof
modify its microclimate. The research utilizes data gathered over three and is thereafter referred to as the “south green roof,” (SGR). In this
years from four weather stations located on and around the JJCC, paper the BR refers to the original southern roof which includes all dates
beginning before the green roof was installed. The discussion attempts to before June 2014. All climate stations were installed at a height of
further practitioner questions regarding whether partially greened roof approximately 3 m over the respective surface and monitored continu­
surfaces can alter the UHII and if the cooling provided by the green roof ously at five-minute intervals. Although the weather station measure­
influences street level air temperatures. ments were made only 3 m above the surface, a higher section of roof
existed at the JJCC and the monitored roofs were at a similar height to
2. Materials and methods many of the roof tops in the vicinity at the time of the analysis. Thus, the
observations recorded at the stations fall within the urban canopy layer.
The JJCC is located on the west side of Manhattan between 34th and The parameters measured by the climate station and the corresponding
40th streets and between 11th and 12th avenues (Fig. 1). The building equipment specifications are described in Table 1. The albedo and
itself is incised into a sloping terrain that is higher on the east side of the monitoring periods for each surface are presented in Table 2. Each of the
building. However, while the street elevations vary, the building is 4-component net radiometers were calibrated for the surface using the
separated from the street by an open space roughly equivalent to the manufacturers procedure.
street elevation on the west side of the building. In addition, the roof of To quantify the rate of ET from the JGR, three weighing lysimeters
the building contains complex shapes and towers. The green roof surface were installed on the SGR (Fig. 2). Each lysimeter consists of a square
is approximately 17 m above the street elevation on its east side, or an section of green roof, enclosed in a 0.372 m2 metal box that rests on a
average of 21 m above the average elevation of the perimeter sloping custom 0.372 m2 Rice Lake Roughdeck mild steel platform scale
terrain. The JJCC’s 27,316 m2 green roof was fully completed during the equipped with four mild steel load cells. The maximum capacity of the
Spring of 2014 and consists of a Xero Flor XF301 + XT extensive system. scale is 227 kg with a 0.02 kg resolution. Although physically isolated
In section, it consists of a pre-vegetated sedum mat installed on top of from the green roof, the lysimeter is surrounded by four sloped transi­
1.5–5 cm of growing medium, a retention fleece layer, a drainage layer tion sections of the green roof to minimize the boundary effects. The
and a root barrier, as described in detail in Alvizuri et al. [5]. metal box is tilted on the weighing scale at a similar slope as the green
roof to allow drainage from the lysimeter.
To quantify the UHII at the JJCC monitoring locations, a fifth
2.1. Description of monitoring system weather station was established at a reference site. This climate station,
identical in components and sampling frequency, was established in
Monitoring of the Javits Green Roof (JGR) was initiated in July 2013, Alley Pond Park (Queens, NY), a 2.7 km2 urban park containing one of
before the south section of the green roof had been completed, enabling the last sections of old growth forest in New York City. The climate
a comparison of the pre- and post- green roof installation on the south station was installed under the forest canopy (Fig. 3). This reference site
section. The full monitoring setup was designed to aid in quantifying the is a canopied weather station approximating pre-industrial era air
energy and water balance of the JGR and includes climate stations, temperature conditions.
weighing lysimeters, soil sensors, flumes, pressure transducers and an The monitoring instruments were used to compare air and surface
infrared camera. Only the climate stations (Fig. 1) and weighing ly­ temperatures, compute surface energy fluxes, and to measure the UHII
simeters (Fig. 2) are utilized in the analysis presented here. A thermal

Fig. 1. Climate station locations and monitored surfaces at the Jacob K. Javits Convention Center (Google, 2017).

2
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Fig. 2. JJCC weighing lysimeter located on the SGR. Top: Picture of lysimeter installed on the roof; Bottom: Cross-section drawing of lysimeter.

Table 1
Climate station equipment and sensor accuracy.
Measured Parameter Equipment Sensor Accuracy

Logger Campbell Scientific CR1000 Logged at 5 min

intervals
Climate Station
Precipitation Texas Electronics, Inc. Series 1.0% at 10 mm/h
525 Rainfall Sensor or less
Wind Speed and Direction Model 5103 - Young Wind Anemometer:
Sentry Anemometer ±0.5 m/s
Vane: ±5◦
Longwave Radiation (In/ Hukseflux NR01 4-Comp­ ±10%
Out) nemnt net-radiation sensor
Shortwave Radiation (In/
Out)
Air Temperature and Campbell Scientific CS215 Air Temperature:
Relative Humidity ±0.3 ◦ C
Relative Humidity:
±4%
Surface Temperature Hukseflux NR01 4-Comp­ ±20%
nemnt net-radiation sensor
Evapotranspiration Custom 0.372 m2 Lysimeter –
Fig. 3. Image of Alley Pond Park climate station.

at the four monitoring locations. These methodologies are described

Table 2
below.
Surface monitoring periods and albedo.
Surface Type Monitoring Period Shortwave Surface
Analyzed Albedoa 2.2. Statistical analysis of air and surface temperatures
Metal Roof (MR) Jul 2013–Dec 2016 0.39
Street Level Station (SLS) Jul 2013–Dec 2016 0.16 The relationships between the air and surface temperature for each
North Green Roof (NGR) Jul 2013–Dec 2016 0.22 surface were evaluated by averaging five-minute measurements over
Black Roof (BR) Aug 2013–Nov 2013 0.13 hourly intervals. The temperature measurements were not normally
South Green Roof (SGR) Jun 2014–Dec 2016 0.21 distributed, requiring the use of non-parametric statistical analyses [12]
Alley Pond Park (forest Jul 2014–Dec 2016 0.24
floor)
performed in R, version 3.3.1 (Mavericks Build). The Kruskall-Wallis test
(α = 0.05) was first used to test whether the distributions of air tem­
a
Calculated as the ratio of shortwave observations from the NR01 radiometer. perature differences were the same (e.g. the null hypothesis) for all
climate stations. After rejection of the null hypothesis (p < 0.05),

3
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

pairwise differences in air and surface temperature between the Unfortunately, QLE could not be computed for the NGR because there is
different monitoring locations were computed, and Dunn post hoc tests no lysimeter present on this section of the roof. Previously published
were used to determine which pairs were statistically similar or research presents the measured heat fluxes through the roof layers [5].
different. The heat accumulation term could not be directly measured in this study
but is included cumulatively with the RSE.
2.3. Computation of surface energy fluxes
2.4. Computation of the UHI intensity
The main energy fluxes of the surface energy balance over any sur­
face are represented as: The UHII was calculated as the air temperature difference between
and of the JJCC monitoring stations and the rural reference station,
Q* = QH + QC + QLE, (1) Alley Pond Park, and was represented as follows:
where Q* represents the net radiation, QH is the sensible heat flux, QLE is UHII = TJ – TAP, (6)
the latent heat flux and QC is the conductive heat flux. The sign
convention denotes non-radiative flux away from the surface as positive. where TJ represents any of the four climate station locations on and
The net radiation (Q*) was calculated from the four-component radiation around the Javits Center, and TAP, is the simultaneous temperature
measured at the climate station and is expressed as: measured at Alley Pond Park.

Q* = K↓– K↑+ L↓– L↑, (2)

3. Results and discussion
where K is the shortwave radiation (solar radiation) flux, L is the long­
wave (terrestrial) radiation flux and the arrows represent the flow of This study presents an examination of the microclimate and energy
energy toward (↓) or away (↑) from the surface. Q* is an upper bound on fluxes above four different surfaces on and around a large-scale green
the energy that is available to heat the air and ground or to evaporate roof, pre and post installation. The surfaces included a concrete side­
water [22]. walk, a traditional urban metal roof, a black roof, and the green roof.
The sensible heat flux (QH) characterizes the transfer of energy be­ The results serve as an exploration into how these kinds of urban sur­
tween the surface and the air along a vertical temperature gradient. This faces may contribute to the warming of the urban environment assuming
heat transfer process is through both conduction (air and surface the partitioning of energy into QLE leaves less energy available for QH,
transfer) and convection (upwards circulation of air). The transfer of reducing the transfer of thermal energy to the atmosphere.
sensible heat from the surface into the air or vice versa is driven by the
temperature gradient between the air and surface. Positive QH values 3.1. Analysis of air temperature
indicate heat transfer to the air, while negative values indicate the
opposite. A microclimate summary of the four monitoring sites and hourly
Ground heat flux (QC) (or conductive heat flux in the case of the green temperature anomalies between the respective monitored station and
roof), involves the transfer of heat to the surface from the subsurface NGR is presented in Fig. 4. The results of the non-parametric post hoc
through the process of conduction. QC is driven by a temperature analysis of all the monitored station pairs (Fig. 5) are displayed using the
gradient between the surface and subsurface. following color scheme: (1) shaded red (e.g. true) indicating a failure to
The latent heat flux (QLE) is the consequence of the phase change of reject the H0, meaning that the two observed air temperatures were
liquid water to vapor. The energy required for this phase change is similar and (2) shaded blue (e.g. false) indicating a rejection of H0,
termed the latent heat of vaporization and is used to convert depth meaning that the air temperatures of the pair were different.
averaged ET rates into energy flux equivalents as follows: As expected, the air temperature above the BR was significantly
higher than the air temperature above the NGR. The maximum daytime
QLE = λvET, (3)
air temperature difference between the black and green roof was
where is the latent heat of vaporization and ET is the evapotranspiration recorded as 1.80 ◦ C. This finding differs from the results of Jim and Peng
rate [LT− 1]. The ET was calculated using the mass changes from the [18], whodid not observe air temperature differences greater than 1 m
lysimeter based on equation (4): above the surface. The smallest temperature anomaly was observed
between the SGR and NGR, indicating the similarity of air temperature
∑x
ET = (
mi − mi+1
)i , (4) for the two green roofs. However, the SGR air temperature was consis­
i=1
Aρ tently slightly higher than the NGR, even during the summer months.
This observation could be attributed to orientation of the building
where ET is the evapotranspiration, mi and mi+1 are the mass of the relative to the cardinal directions (e.g. greater solar gain on the south
lysimeter at consecutive sampling intervals, x is the number of sampling side of the building), with related research ongoing.
intervals, Δt is the time between sampling intervals, ρ is the density of The MR station was located on a roof surface between two buildings
water and A is the surface area of the weighing lysimeter open to ET. To and the SLS station at street level to the east of the green roof. However,
minimize the potential for errors due to gravity-driven vertical discharge even though the strongest wind speed was observed at the MR station,
of water out of the bottom of the lysimeter, ET values were only the SLS and MR maintain similar air temperature profiles.
calculated more than 24 h after a precipitation event. Significance analysis indicates similar air temperature between all
To quantify the amount of energy available to heat the air and sur­ the stations during the winter and spring months, a period which also
face (after any ET had occurred), the ‘residual energy flux’ (RSE) is corresponds to the dormancy of the sedum vegetation. In summer and
defined as follows: autumn, air temperature anomalies relative to the NGR were observed to
RSE = Q*-QLE = QH + QC. (5) be significantly different, with only two exceptions involving the com­
parison of NGR to SGR for two summer months (7/2014, 5/2015).
Specifically, the RSE represents the energy available for QH and QC, Interestingly, although the SLS and MR present different surface and
the two processes that can bring about a local temperature change. environmental conditions, their air temperatures do not differ signifi­
Because no ET measurements were performed within 24 h of a rain event cantly for all months except 12/2013.
and no there is no ponding on the surface, we assume negligible latent The differences displayed in Fig. 4 highlight the potential importance
heat flux (QLE ~ 0) on non-vegetated surfaces (i.e. SLS, BR, MR). of changes in energy fluxes (i.e. evapotranspirative cooling) on air

4
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Fig. 4. Left: Microclimate summary and air temperature anomaly profiles. Right: Monthly median air temperature differences from NGR. Note: The dotted and solid
line denote the beginning end of the SGR installation, respectively.

Fig. 5. Air temperature post-hoc test results of surface pairs per month. True = fail to reject H0, False = reject H0. The dotted and solid line denote the beginning end
of the SGR installation, respectively.

temperature. The results presented in Figs. 4 and 5 suggest that the surface of the SGR and NGR differed by a maximum of 3.81 ◦ C and
cooled air found immediately over the green roof may not reach street 2.88 ◦ C, in spring and autumn, respectively. Compared to the other
level. surface temperature anomalies these differences are relatively small.
While diurnal surface temperature trends displayed by the NGR and
SGR roof surfaces were similar, profile comparisons with the other
3.2. Analysis of surface temperature
surfaces revealed differences. On a typical green roof (i.e. NGR/SGR)
during the day radiant energy is absorbed into the surface during the day
The average hourly surface temperatures at each location versus the
and released as the sun sets. The black roof maintains a higher surface
surface temperature of the NGR, per season is displayed in Fig. 6. To
temperature and experiences wider temperature fluctuations than the
facilitate visualization of diurnal variations, the observations are
NGR only returning to similar surface temperatures during the late night
colored based on the hour of the day. During the day, the surface of the
to early morning time period. This supports the findings of Alvizuri et al.
BR was consistently higher than the surface of the NGR, especially in the
[5], who concluded that the NGR is a better thermal regulator than the
afternoon, and it rarely fell below the surface temperature of the NGR at
BR. It also upholds the observations in the previous section which
any time of the day. Unlike the other surfaces, the BR surface remained
showed higher air temperatures above the BR than above the NGR. We
warmer for a longer portion of the the day. After the completion of the
note, however, that higher air temperature anomalies are not always
SGR, the surface temperature on the two green roof surfaces were
indicative of higher surface temperature anomalies.
similar, with only a few summer daytime periods as exceptions.
Throughout all seasons the MR nearly always maintains a lower
In general, during the summer, maximum daytime surface temper­
surface temperature than the NGR. Differences in surface temperatures
atures differed from the NGR as follows (positive values are exceed­
between these two surfaces were greatest in the cooler months. In
ances): BR, 12.07 ◦ C (9/2013); SLS, 9.53 ◦ C; MR, − 22.15 ◦ C. Notably the

5
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Fig. 6. Hourly surface temperature comparisons between the NGR and other monitored surfaces; solid line = 1:1 line; broken line = (right) 1:0.5 (left) 1:2. The colors
represent the hour of day. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.)

contrast, during all seasons, the SLS only revealed a higher surface in Hein [37].
temperature than the NGR during the early morning to afternoon hours.
Recalling the previous section, the air temperature profiles of the SLS
3.3. Analysis of the residual energy flux
and MR are statistically similar, but the MR consistently maintains a
lower surface temperature than the SLS and green roofs. This observa­
The RSE of the BR, SGR, MR and SLS are shown in Fig. 7. Note that
tion can potentially be attributed to the reflective properties of the MR
the measurements on the south roof were conducted pre- and post-green
which reradiates the incoming radiant energy as discussed further
roof installation. The observations displayed thus represent the same
below. It should also be noted that this building serves as a connector or
seasons, but not necessarily the same days. Measurements of QLE were
“link” between two buildings but is not thermally regulated inside. The
not calculated in the winter due to the disconnction of the lysimter
SLS exhibits the most dynamic surface temperature profile compared to
during the cold months.
the NGR.
The conversion of the BR to the SGR contributed to a 50%, 35% and
In summary, the surface temperature on the BR was nearly always
56% decrease in the median daytime RSE, for the spring, summer and
greater than the NGR. The median and maximum daytime autumn
autumn season, respectively. The highest RSE during the day was dis­
surface temperature differences reached 3.26 ◦ C and 18.4 ◦ C respec­
played by the SLS, followed by the MR and SGR. The RSE of the BR could
tively. Gaffin [23] reported a daily average and maximum (peak sum­
only be computed for one season, due to the transition from the BR to
mer) BR to green roof surface temperature difference of 9.9 ◦ C and 33 ◦ C,
SGR. Although the UHII on the MR was statistically similar to the SLS,
respectively, on an extensive green roof in NYC. However, Gaffin’s
the RSE of the MR was 30%, 37% and 22% less than the SLS for the
measurements were made in the summer, while ours were during the
spring, summer and autumn seasons, respectively.
autumn, a potential explanation for this discrepancy. Hein [24] an Jim
The largest daytime air temperature was, unsurprisingly, observed
[18] observed a maximum surface temperature difference between the
on the BR. The lesser transfer of QH to the air from the two green roofs,
green roof and original surface of 18 ◦ C and 11 ◦ C, respectively. The
relative to the BR as displayed in Fig. 6, suggests that the green roofs do
maximum BR and NGR surface temperature difference measurements on
contribute less to warming of the environment.
the Javits roof was observed as 18.4 ◦ C, similar to observations reported
Installation of the green roof reduced the surface and air temperature

6
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Fig. 7. Residual energy flux (RSE) diurnal profile of the JJCC monitored locations per season.

on both sections of the JJCC roof. This finding is consistent with pre­ absorbs heat faster while a surface with a high negative rate of change
vious investigations into microclimatic conditions on the JJCC green denotes heat leaving the surface quickly. The black roof is not shown.
roofs conducted using an infrared camera [5]. The green roof decreased The magnitude of the roc of the air temperature is smaller than the
the RSE (QH + QC) by 68% in the summer, compared to the BR. Based on roc surface temperature. Air temperature roc varied most between the
equation (1) and the representation of RSE in equation (5), a higher QLE surfaces in the morning after sunrise and in the late afternoon during the
is attributed to lower RSE. The QLE at peak solar radiation (midday) summer. The air temperature roc of the green roofs begins to decrease
during the summer was 364 W m− 2. Similar values are reported in earlier than MR and SLS. Also of note is the slope of the line between
literature for other extensive green roofs, as detailed in Table 3 (adapted 16:00 and sunrise (~5:00) in each of the plots. As the seasons progresses
from Ref. [25]) even during the initial summer of the green roof. into cooler temperatures the slope gets smaller and closer to zero.
Increased partitioning of energy into QLE reduces the RSE, thereby Interestingly, the two green roofs, built about a year apart, maintain a
reducing the thermal transfer of energy in the air. The reduced energy similar roc temperature profile although the air temperature on the two
transfer into the atmosphere is expressed as lower air temperatures roofs are statistically different for a majority of the monitored months.
observed directly above the green roof, as observed on the Javits roof This finding supports the theory that the SGR maintains a slightly higher
between the BR and green roof surfaces. air temperature than the NGR due to its geographic orientation with
Large differences in the air temperature between the two sections of respect to solar exposure.
green roofs were not observed. However, during most of the monitoring The surfaces of the green roofs warm faster than MR and SLS, but the
period the SGR consistently displayed higher air temperatures than green roofs also cool faster after sunset. The MR releases heat at lower
NGR. This discrepancy could be due to one or more of the following: (1) roc than the green roofs. In the context of UHI formation a slower release
a slightly longer exposure to solar radiation (e.g. southern solar expo­ of heat is not a favorable. At night the slope of the line is nearly zero. The
sure), (2) differences in the density and maturity of vegetation on the SLS and MR air temperatures were significantly similar (section 3.1).
NGR and SGR (i.e. differences in green roof age), or (3) differences in the Interestingly, during the spring and summer at the SLS the surface
amount of water available for ET on the SGR. In support of the latter temperature is always greater than the corresponding air temperature,
explanations, air temperatures over the SGR differed most from the MR indicating a transfer of energy from the surface to the local environment.
and SLS locations during the sedum’s growing season, a potential indi­ During the spring and summer, the surface temperature at the SLS is
cator of the role that vegetation-mediated ET (QLE) in determining air nearly always greater than the air temperature. However, in autumn this
temperature. Precise differences in ET between the NGR and SGR could pattern changes. At lower air temperatures, the surface temperature is
not be computed, however, because QLE was only measured on the SGR. greater than the air for most of the day. At higher air temperatures this
The largest daytime air temperature was unsurprisingly observed on pattern is only exhibited during the day. These observations have led to
the BR. The lesser transfer of QH to the air from the two green roofs, the belief there is a tunnel under the SLS as the presence of a void would
relative to the BR as displayed in Fig. 7, suggests that the green roofs do change the surface/air dynamics of the surface.
contribute less to warming of the environment.

3.5. Analysis of the UHII

3.4. Analysis of surface/air temperature rate of change
Fig. 9 displays the diurnal variation in the UHII at each monitoring
As described previously, the UHII is typically greatest during the late location, by season. Only the data collected after completion of the SGR
afternoon/evening period with its intensity related to surface type and (e.g. between November 2014 to May 2016) are included in this anal­
energy exposure. Fig. 8 presents the diurnal rate of change (roc) of air ysis. Clear diurnal patterns are observed, with the lowest UHII observed
temperature (left) and surface temperature (right) of the four surfaces. at mid- day (12h) and the highest between sunset and sunrise.
The black roof is not shown. A surface with a high positive rate of change The statistical significance of the daytime and nighttime UHII per
season is displayed in Fig. 10. Boxes shaded red (H0 = true) indicate that
Table 3 the two observed UHII values were similar. Boxes shaded blue (H0 =
Latent heat from sedum extensive green roofs (Adapted after: Santamouris [25]). false) indicate that the UHII pairs were different. Regardless of time of
Reference Peak Solar Radiation (Wm− 2) QLE (Wm− 2) day, the NGR UHII was significantly lower than all the other surfaces. By
contrast, the UHII at the SGR was not significantly different from the MR
Feng [26] 900 600
Rezaei [27] and Berghahe [28] – 350 and SLS during two specific periods: (1) at night during the spring and
Marasco [29] – USPS: 439 autumn season and, (2) during winter daytime hours. The UHII of the
Columbia: 408 MR and SLS locations were generally similar, except during the winter
DiGiovanni [30] 914 310 season.
Current study 917 364
Maximum and median UHII measurements for daytime and

7
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Fig. 8. Rate of Change diurnal profile- Left: air temperature; Right: surface temperature; SS: sunset; SR: sunrise.

Fig. 9. UHII profile of all JJCC monitoring locations, per season. Reference site: Alley Pond Park.

hours.

4. Discussion

5. Conclusions

This study quantified the differences in observed air and surface

temperatures over various urban surfaces (including a green roof). Sig­
nificant air temperature variability was observed, even on two different
sections of the building’s green roof.
The addition of the NGR reduced the UHII (compared to the SLS) by
1.1 ◦ C and 0.91 ◦ C for the median summer daytime and nighttime hours,
respectively. The SGR reduced temperatures slightly less at 0.48 ◦ C and
0.44 ◦ C for the median summer daytime and nighttime hours, respec­
tively. This effect is attributed to the green roof’s role in partitioning
more incoming energy into QLE, which does not require a temperature
Fig. 10. Air temperature post-hoc test results- True = fail to reject H0, False =
reject H0. Shaded regions indicate statistical similarity among shaded roof
change, preventing the energy from being converted into QH and QC,
types, α = 0.05. which are fluxes driven by a vertical temperature gradient. Results show
that surfaces can vary diurnally and seasonally. When conducting
remote studies, care should be taken when analyzing results on land
nighttime periods are displayed per season in Table 4. Across all seasons
surface types (i.e. side walk, street, reflective roof)
and times of the day the max UHII on the NGR 0.88 ◦ C less than the SLS.
The green roof’s ability to reduce the local UHII suggests a role for
The NGR’s peak UHII is more than 1 ◦ C less than the SLS during the
this form of green infrastructure in climate change adaptation efforts.
spring daytime and nighttime hours, as well as the summer nighttime

8
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

Table 4
Median and maximum seasonal UHII observations. Reference site: Alley Pond Park.
UHII Period Maximum (◦ C) Median (◦ C)

NGR SGR MR SLS NGR SGR MR SLS

Spring (Daytime) 8.55 8.77 9.54 9.63 0.268 1.01 1.24 1.29
Spring (Nighttime) 6.04 6.24 6.87 7.13 1.03 1.70 1.70 1.80
Summer (Daytime) 6.15 6.96 6.71 7.13 1.20 1.68 2.25 2.29
Summer (Nighttime) 5.15 5.40 5.99 6.19 1.56 2.00 2.37 2.47
Autumn (Daytime) 5.46 5.58 5.78 6.04 0.347 1.02 1.38 1.25
Autumn (Nighttime) 4.71 5.43 5.28 5.41 0.801 1.38 1.43 1.44
Winter (Daytime) 1.88 3.07 2.66 2.72 − 0.191 0.734 0.610 0.749
Winter (Nighttime) 3.87 4.57 4.77 4.65 0.211 1.11 0.878 0.746

The NGR on the JJCC significantly reduced peak daytime air tempera­ [3] Ahmed Memon Rizwan, Leung YC. Dennis, L.I.U. Chunho, A review on the
generation, determination and mitigation of Urban Heat Island, J. Environ. Sci. 20
tures by 1.7 ◦ C. However, cooling was observed only on top of the roof
(1) (2008) 120–128.
and was not detectable at street level. Air temperature values at the SLS [4] Annie Hunter Block, Stephen J. Livesley, Nicholas SG. Williams, Responding to the
were similar before and after the SGR installation. Possible explanations urban heat island: a review of the potential of green infrastructure, Victorian
include the height of the building and/or the prevailing wind direction, Centre for Climate Change Adaptation Research Melbourne, 2012.
[5] J. Alvizuri, et al., Green Roof Thermal Buffering: insights derived from fixed and
which was towards the west side of the building and away from the portable monitoring equipment, Energy Build. 151 (2017) 455–468.
street level monitoring station. [6] Umberto Berardi, GAmirHosein haffarianHoseini, GhaffarianHoseini Ali, State-of-
Roofs represent approximately 30–35% of the total land area of the the-art analysis of the environmental benefits of green roofs, Appl. Energy 115
(2014) 411–428.
urban environment [31,32]. Because of their large footprint, many [7] Andrew M. Coutts, et al., Assessing practical measures to reduce urban heat: green
major cities are developing codes and ordinances that specify "greening" and cool roofs, Build. Environ. 70 (2013) 266–276.
of these spaces. For example, NYC has passed Local Laws 92 and 94 [8] Aikaterini Niachou, et al., Analysis of the green roof thermal properties and
investigation of its energy performance, Energy Build. 33 (7) (2001) 719–729.
which require new buildings and buildings undergoing “major [9] Adam Scherba, et al., Modeling impacts of roof reflectivity, integrated photovoltaic
constriction” to include either a roof with plants, solar panels and/or panels and green roof systems on sensible heat flux into the urban environment,
mini wind turbines. Data collected on the JJCC suggests that these or­ Build. Environ. 46 (2011) 2542–2551, 12.
[10] William D. Solecki, et al., Mitigation of the heat island effect in urban New Jersey,
dinances may be effective in reducing, though not eliminating, the UHII. Global Environ. Change B Environ. Hazards 6 (1) (2005) 39–49.
In this work, the cooling benefit could not be detected at street level, but [11] Gurdane Virk, et al., Microclimatic effects of green and cool roofs in London and
additional work with computational fluid dynamics models or dense their impacts on energy use for a typical office building, Energy Build. 88 (2015)
214–228.
sensor networks is needed to shed additional light on the three dimen­
[12] Nyuk Hien Wong, et al., Investigation of thermal benefits of rooftop garden in the
sional fluxes of air and energy within complex urban topographies such tropical environment, Build. Environ. 38 (2) (2003) 261–270.
as this one. [13] Lilliana LH. Peng, C.Y. Jim, Green-roof effects on neighborhood microclimate and
human thermal sensation, Energies 6 (2) (2013) 598–618.
[14] A.F. Speak, et al., Reduction of the urban cooling effects of an intensive green roof
Funding due to vegetation damage, Urban Climate 3 (2013) 40–55.
[15] T. Susca, S.R. Gaffin, G.R. Dell’Osso, Positive effects of vegetation: urban heat
This research was funded principally by the Jacob K. Javits island and green roofs, Environ. Pollut. 159 (8) (2011) 2119–2126.
[16] Kristin L. Getter, et al., Seasonal heat flux properties of an extensive green roof in a
Convention Center, with additional support by the National Science Midwestern US climate, Energy Build. 43 (12) (2011) 3548–3557.
Foundation through CAREER: Integrated Assessments of the Impacts of [17] Xiaosheng Qin, et al., A green roof test bed for stormwater management and
Decentralized Land Use and Water Management (CBET: 1150994), and reduction of urban heat island effect in Singapore, Br. J. Environ. Clim. Change 2
(4) (2012) 410–420.
the National Oceanic and Atmospheric Association (NOAA) through [18] Chi Yung Jim, LH Peng Lilliana, Weather effect on thermal and energy
Supporting Regional Implementation of Integrated Climate Resilience: performance of an extensive tropical green roof, Urban For. Urban Green. 11 (1)
Consortium for Climate Risks in the Urban Northeast (CCRUN) Phase II (2012) 73–85.
[19] Panagiotis Sismanidis, Iphigenia Keramitsoglou, Chris T. Kiranoudis, A satellite-
(NA15OAR4310147). based system for continuous monitoring of surface urban heat islands, Urban
Climate 14 (2015) 141–153.
Declaration of competing interest [20] Diana E. Bowler, et al., Urban greening to cool towns and cities: a systematic
review of the empirical evidence, Landscape Urban Planning97 3 (2010) 147–155.
[21] Corrie Clark, Brian Busiek, Adriaens Peter, Quantifying thermal impacts of green
The authors declare that they have no known competing financial infrastructure: review and gaps, Proc. Water Environ. Fed. 2 (2010) 69–77, 2010.
interests or personal relationships that could have appeared to influence [22] T.R. Oke, Boundary Layer Climates, second ed., Routledge, 1987.
the work reported in this paper. [23] S.R. Gaffin, et al., A temperature and seasonal energy analysis of green, white, and
black roofs, Center Climate Syst. Res. (2010). Columbia University, New York,
Technical Report.
Acknowledgments [24] Hien, Nyuk Wong, Puay Yok Tan, Yu Chen, Study of thermal performance of
extensive rooftop greenery systems in the tropical climate, Build. Environ. 42 (1)
(2007) 25–54.
The authors acknowledge the contributions of many members of [25] Mattheos Santamouris, Cooling the cities–a review of reflective and green roof
Drexel University’s Sustainable Water Resource Engineering lab in the mitigation technologies to fight heat island and improve comfort in urban
monitoring system installation and especially a critical review by Dr. environments, Sol. Energy 103 (2014) 682–703.
[26] Chi Feng, Qinglin Meng, Yufeng Zhang, Theoretical and experimental analysis of
Bita Alizadehtazi. the energy balance of extensive green roofs, Energy Build. 42 (6) (2010) 959–965.
[27] F. Rezaei, Evapotranspiration Rates from Extensive Green Roof Plant Species, M.S.
References Thesis, Agricultural and Biological Engineering, The Pennsylvania State University,
State College, PA, 2005.
[28] Robert Berghage, et al., Quantifying evaporation and transpirational water losses
[1] Eleftheria Alexandri, Phil Jones, Temperature decreases in an urban canyon due to
from green roofs and green roof media capacity for neutralizing acid rain, National
green walls and green roofs in diverse climates, Build. Environ. 43 (4) (2008)
Decentralized Water Resour. Capacity Dev. Project (2007).
480–493.
[29] Daniel E. Marasco, et al., Quantifying evapotranspiration from urban green roofs: a
[2] S.D. Gedzelman, et al., Mesoscale aspects of the urban heat island around New York
comparison of chamber measurements with commonly used predictive methods,
City, Theor. Appl. Climatol. 75 (1) (2003) 29–42.
Environ. Sci. Technol. 48 (2014) 10273–10281, 17.

9
L. Smalls-Mantey and F. Montalto Building and Environment 197 (2021) 107792

[30] Kimberly DiGiovanni, et al., Applicability of classical predictive equations for the [31] Urban roofscapes: using ‘wasted’ rooftop real estate to an ecological advantage,
estimation of evapotranspiration from urban green spaces: green roof results, Scientific American, Scientific American, 25 July 2008. www.scientificamerican.co
J. Hydrol. Eng. 18 (1) (2012) 99–107. m/article/urban-roofscapes-ecofriendly-rooftops/.
[32] Sadayuki Onmura, M. Matsumoto, S. Hokoi, Study on evaporative cooling effect of
roof lawn gardens, Energy Build. 33 (7) (2001) 653–666.

10