Urban heat mitigation by roof surface materials during the East Asian summer monsoon

The experimental site was located on the rooftop of the College of Agriculture and Life Sciences building of Seoul National University, Seoul, South Korea (37.457297°N, 126.948263°E). Data were collected from 20 May to 14 September (day of year (DOY): 140–257) in 2015. Summer is defined as the period from June to August (DOY 152–243).

We investigated three roof surface materials; Kentucky bluegrass (Poa pratensis; hereinafter called grass), Sedum sarmentosum (hereinafter called sedum), and a white-painted surface. The reference roof surface was an existing rooftop of the building that was painted green. On the rooftop, we measured the surface temperature of the reference roof in a 0.5 × 0.5 m plot. The other materials were laid horizontally on the reference roof. We planted grass in a 1.8 × 1.8 × 0.4 m planter that had been filled with a 0.15 m layer of organic soil on top of a 0.1 m layer of pearlite (artificial soil). This soil mix is commonly used in roof gardens (Sun et al 2014). We planted sedum in a 1.0 × 1.0 × 0.1 m planter, filled with a 0.05 m layer of organic soil on top of a 0.05 m layer of pearlite. High-albedo paint (ESU–100, NOROO Paint and Coatings Co.) was painted onto the surface of a 1.8 × 1.2 × 0.01 m aluminum panel.

A hyperspectroradiometer (FieldSpec 4 Wide-Res, ASD, Inc.) was used to measure albedo for four roof surface materials (white, grass, sedum, and reference) on DOY 173 and 257. The same instrument was used to measure the normalized difference vegetation index (NDVI) for grass and sedum on DOY 140, 181, and 223. For each of the four roof surface materials, spectrum measurements were performed through a reference panel (Spectralon, Labsphere) above the target surface and then again immediately after removing the panel; the two spectra were compared to calculate the albedo and reflectance of the surface. NDVI was calculated from the red reflectance (630–690 nm) and near-infrared reflectance (750–800 nm) (Tucker 1979). Type E thermocouples (CS220, Campbell Scientific) were attached underneath the surfaces of the white, grass, and sedum materials to measure the temperature beneath the surface of each roof surface material during DOY 174–243. Three infrared radiometers (SI–121, Apogee Instrument, Inc.) were installed at 0.8 m above the sedum roof, 0.9 m above the white roof, and 0.5 m above the grass roof to measure surface temperature during DOY 140–243.

There were heatwave alerts (days when a peak daily temperature exceeding 33 °C occurred on more than two consecutive days) in Seoul on DOY 218–220 in 2015. To compare the degree of required management for the different roof surface materials, we specifically analyzed data for three periods that included an irrigated period before the heatwave (DOY 146–152), a non-irrigated period before the heatwave (DOY 182–188), and a non-irrigated period after the heatwave (DOY 221–227). We watered the grass with 30 l per day during the irrigated period.

The daily air temperature, relative humidity, precipitation (for 1953–2015), and solar radiation (from 1973 to August 2015) were obtained from an observatory of the Korea Meteorological Administration (Seoul 37.571406°N, 126.965708°E). To obtain the sky clearness index (the ratio of actual daily total solar radiation at the land surface, R_sur, to daily total potential solar radiation at the top of the atmosphere, R_TOA), R_TOA was calculated as follows:

$$\begin{eqnarray}&&{R}_{{\rm{TOA}}}=\displaystyle \sum _{i=1}^{48}60\times 30\times {R}_{{{\rm{TOA}}}_{i}\;},\end{eqnarray} \tag{ 1 }$$

where i is the half hour of the day. ${R}_{{{\rm{TOA}}}_{i}\;},$ was calculated as follows (Liu and Jordan 1960); (Ryu et al 2012):

$$\begin{eqnarray}&&{R}_{{{\rm{TOA}}}_{i}\;}={S}_{{\rm{sc}}}\times \left[1+0.033\;\mathrm{cos}\left(\displaystyle \frac{2\pi {t}_{d}}{365}\right)\right]\times \;\mathrm{cos}\;\beta ,\end{eqnarray} \tag{ 2 }$$

where S_sc is the solar constant (1360 Wm⁻²; (Kopp and Lean 2011), t_d is the day of the year, and β is the solar zenith angle calculated from the latitude, longitude, and local time (Michalsky 1988). We calculated the daily sky clearness index by dividing R_sur recorded at the weather station by R_TOA (equation (3)):

$$\begin{eqnarray}&&{\rm{Sky}}\;{\rm{clearness}}\;{\rm{index}}=\displaystyle \frac{{R}_{{\rm{sur}}}}{{R}_{{\rm{TOA}}}}.\end{eqnarray} \tag{ 3 }$$

The distribution of the sky clearness indices for 43 years of summers (1973–2015) is shown in figure 1.

Zoom In Zoom Out Reset image size

Statistical analyses were performed using SigmaPlot 12.0 software (Systat Software Inc., Chicago, IL, USA). We utilized one-way analysis of variance followed by Tukey's post-hoc test to compare the mean daytime surface temperature of each of the four roof surface materials for the three periods. All results are presented as mean ± 95% CI unless otherwise specified. A p-value of <0.05 was considered to be statistically significant.

The albedos of the roof surface materials varied significantly (table 1). The white roof had the highest mean albedo, followed by the reference, grass, and sedum roofs. The albedo of the white roof decreased significantly between DOY 173 (0.819) and 257 (0.748) (p < 0.05). The grass and sedum roofs had similar albedos (∼0.18) on DOY 173, but the albedo of the sedum roof decreased significantly on DOY 257 (p < 0.05). The reference surface maintained an albedo (∼0.3) higher than that of the grass and sedum roofs from DOY 173 to 257.

Clouds strongly influenced the diurnal patterns of surface temperature among the roof surface materials (figure 2). The diurnal patterns of surface temperature of the white, grass, and sedum roofs differed between clear and cloudy days. The peak surface temperature of the white roof was significantly lower than that of the grass (by up to 5 °C) and sedum (by up to 13 °C) roofs on clear days (figure 2(a)). The surface temperature of the white, grass, and sedum roof materials had similar diurnal patterns and values on cloudy days (figure 2(b)).

Zoom In Zoom Out Reset image size

The relationships between the sky clearness index and the temperature difference between the roof surface materials and the reference surface during daytime throughout summer differed between the white roof and grass/sedum roofs (figure 3). The sky clearness index and surface temperature difference of the white roof showed a marked negative correlation (R = −0.71). Compared with the white roof, the surface temperature difference between the green roofs and the reference showed weaker correlations with the sky clearness index (R = −0.30 for grass, and R = −0.28 for sedum).

Zoom In Zoom Out Reset image size

The variations in mean daytime surface temperature and NDVI of the roof surface materials differed for the three summer periods. The mean daytime surface temperature of the white roof was lowest for all three periods regardless of watering (figure 4(a)) and heatwave occurrence (before heat wave: figure 4(b), after heat wave: figure 4(c)) (p < 0.05). There was no significant difference in the mean daytime surface temperature between the white and irrigated grass roofs (p > 0.05; figure 4(a)). While the NDVI of the sedum roof was similar in the three periods, the NDVI of the grass roof decreased significantly from 0.643 to 0.554 after the end of watering, and from 0.554 to 0.371 after the heatwave (table 2). However, there were no significant differences in mean daytime surface temperature between the grass and the sedum roofs in the three periods (p > 0.05).

Zoom In Zoom Out Reset image size

The mean of the white roof surface temperature was significantly lower than that of the other roofs (1.1 °C) over the entire summer period (table 3). The white, grass, and sedum roofs had lower surface temperatures than the reference during both the daytime and nighttime. The white roof had the lowest peak daytime surface temperature (figure 5; p < 0.05). The mean differences in the surface temperature between the white and green roofs at nighttime (0.1 and 0.8 °C for the grass and sedum roofs, respectively) were significantly smaller than the differences during daytime (1.9 °C and 1.7 °C for the grass and sedum roofs, respectively; table 3). The grass and sedum roofs had similar diurnal patterns in surface temperature.

Zoom In Zoom Out Reset image size

The diurnal pattern and mean of the temperatures beneath the surface differed from those at the surface among the white, grass, and sedum roofs. The mean daytime temperature beneath the surface of the sedum roof was 0.26 °C and 0.8 °C lower than that of the white and grass roofs, respectively (table 3). The temperature beneath the surface of the white roof had a larger peak than that of the green roofs during daytime, but decreased rapidly at night (figure 5(b)). The mean temperature beneath the surface of the white roof was lowest at night, whereas that of the sedum roof was lowest during the day (p < 0.05). The timings of the peaks of the diurnal patterns of the temperature beneath the surface for the grass and sedum roofs were delayed compared with the white roof. The grass and sedum roofs had higher temperatures beneath the surface from 00:00 to 06:00.

Our results indicate that the degree of cloudiness can significantly affect the surface temperature of white roofs during the East Asian summer monsoon. Daytime surface temperatures among the three roof surface materials differed significantly between clear and cloudy days. Under clear sky conditions, surface temperature in white roof was lower than that in grass and sedum up to 10 °C (e.g. figure 2(a)), which is comparable to previous studies (Takebayashi and Moriyama 2007, Zinzi and Agnoli 2012). The lowest surface temperature on clear days appeared in the white roof mainly resulted from its high albedo, which reflected the greatest amount of solar radiation (figure 2(a)). It has been widely reported that white roofs of high albedo maintain a lower daytime surface temperature on clear-sky days (Akbari et al 1997, Takebayashi and Moriyama 2007, Synnefa et al 2012).

Under cloudy days, we observed that the differences in daytime surface temperature among the three roof surface materials were lower (figure 2(b)). This tendency is in agreement with a previous modeling study for London, which showed that the difference in the daytime air temperature between a white and green roof decreased on cloudy days (Virk et al 2015). The modeling study of Georgescu et al (2014) showed that the difference in the air temperature between white and green roofs in California will be greater than the difference in the air temperature among these roof materials in Florida and the Mid-Atlantic in 2100. Yang et al (2015) reported that cooling performance of white roof depends on meteorological and geographical conditions. Although the surface cooling performance by white roof was largely controlled by cloudiness, it showed the lowest mean surface temperature among the four surface materials over the whole summer period (table 3).

The difference in the surface temperature between the white roof and the reference surface showed a stronger negative correlation (R = −0.71) than that between the grass (R = −0.30) and sedum (R = −0.28) roofs and the reference surface. This result indicates that the surface temperature of green roofs was controlled by not only cloudiness but also physiological mechanisms such as stomata closure and transpiration (figure 3). Thus, the cooling performance of grass roofs for diverse sky conditions is highly variable given the extreme weather events such as dry period, heat spell and rainfall.

The white roofs showed consistently lower surface temperatures than the reference for the range of sky clearness index of the 40-year mean ± 1 S.D. (0.172–0.508; figure 3(a)). Thus, we expect that the installed white roof could be cooler than the reference during summer in Seoul for most sky conditions. The annual mean surface solar radiation in Korea increased over the 15 year period from 1990 to 2005 (Wild et al 2009), and the sky clearness index in Seoul has also significantly increased over the 25 years from 1990 to 2014 (0.002 yr⁻¹, p < 0.05 from Seoul weather station). Thus, the cooling effects by white roof in this region are likely to increase in future.

The white roof required minimal maintenance to sustain cooling performance. The daytime surface temperature among the three roof surface materials differed for the three periods (watering, and before and after the heatwave). There were larger variations in surface temperature for the same sky clearness index for the green roofs (figure 3). This indicates that the green roofs were influenced by biological mechanisms, such as stomata conductance and transpiration. While there was no significant difference in the surface temperature between the transpiring grass roof and the white roof during the irrigated period (figure 4(a)), during the non-irrigated periods, the surface temperature of the grass roof was significantly higher than that of the white roof. The effect of the green roofs in mitigating heat is small without a thick soil layer and watering (Thuring et al 2010, Coutts et al 2013, Li et al 2014, Sun et al 2014), because grass has a shallow rooting depth; thus, its metabolic activity is highly sensitive to the level of soil moisture (Ryu et al 2008, 2010). Although there was little change in the NDVI of the sedum roof over the three periods, and a significant decrease in the NDVI of the grass roof between DOY 181 and 223, there was no significant difference in the surface temperature between the sedum and grass roofs on these dates (figure 4 and table 2). This indicates that the sedum tolerated the hot summer conditions well, which was reflected in the stable NDVI, but it did not decrease the daytime surface temperature effectively (Nagase and Dunnett 2010). A previous study showed that a sedum roof that received sufficient water could significantly reduce the sensible heat flux (Coutts et al 2013) thus surface temperature, which differed from the findings in our study. The 2015 summer monsoon brought 395.5 mm of rainfall between June and August (Seoul weather station, Korea Meteorological Administration), which was the lowest recorded summer rainfall since 1954. We found that this amount of rainfall was insufficient to support a covering of healthy grass, which rapidly reached senescence in early August, as shown by the decrease in the NDVI (table 2). Furthermore, the vapor pressure deficit, an indicator for plant water stress, in Seoul has increased over the last 61 years (1954–2014; 56.7 Pa decade⁻¹, p < 0.05 from Seoul weather station). Thus, we expect the management of green roofs in this region to become more difficult in the future.

The accumulated dust on the white roof surface slightly decreased the albedo. A previous study reported that accumulated dust on a white roof surface reduced the surface albedo, which resulted in the decrease in the surface temperature (Gaffin et al 2012). We measured the albedo of the white roof on DOY 173 and 257 and found a significant decrease (from 0.819 to 0.748) during this period (table 1). To maintain consistent cooling performance of white roofs in the megacity Seoul, which experiences significant air pollution, periodic cleaning of the white roof surface would be required, which is likely easier and more cost-effective than the watering required by green roofs.

The white roof showed the most effective surface cooling for both daytime and nighttime throughout the summer. Our results suggest that the performance of the roof surface materials depends on the characteristics of the weather. Although the difference in the surface temperature between the green roofs and the white roof was small on cloudy days, the surface temperature of the white roof was clearly lowest on clear days (figure 2). The white roof had the lowest daytime surface temperature, which is consistent with previous studies (Takebayashi and Moriyama 2007, Coutts et al 2013, Georgescu et al 2014). The white roof exhibited the lowest mean surface temperature among the roof surfaces throughout the entire summer period (table 3).

The peak daytime temperature beneath the surface among the three roof surface materials differed from the peak daytime surface temperature (figure 5). The white roof had the highest daytime temperature beneath the surface (figure 5 and table 3). The heat capacity of the surface materials plays an important role in the variations in the temperature beneath the surface. Because the white roof consisted of a single aluminum panel that had been painted, the temperature difference between the surface and beneath the surface was smaller (0.7 °C ± 0.5 °C) than that of the green roofs (2.7 °C ± 0.8 °C and 3.5 °C ± 1 °C for the grass and sedum roofs, respectively), which were composed of soil layers. The soil layers created an insulating effect during the daytime and released their stored heat at night (Jim 2015). This caused the temperature beneath the surface of the green roofs to be higher than that of the white and reference roofs at night. These diurnal patterns in the temperature beneath the surface were also observed for white and sedum roofs in a previous study (Coutts et al 2013).

Overall, white roof is the most effective strategy for urban heat mitigation during the East Asian monsoon summer. Although the daytime temperature beneath the surface of the white roof was higher than that of the sedum roof (table 3), the mean difference in the surface temperature between the white and sedum roofs (1.3 °C ± 0.4 °C) was greater than the mean difference in the temperature beneath the surface between the white and sedum roofs (0.3 °C ± 0.3 °C) on all days. Previous studies in southern Europe where low-rise buildings dominate tended to focus on reducing the temperature beneath the roof surface to decrease the energy use of the top floor of the building (Synnefa et al 2012, Romeo and Zinzi 2013). In the case of Seoul, a large proportion of the city is composed of high-rise buildings and apartments. Thus, reducing the roof surface temperature is likely more effective for urban heat mitigation than reducing the heat transfer from beneath the roof surface to the top floor of the building.