Effect of cooling roofs on gust wind speed over an urban agglomeration in Southern China

Urbanization has promoted economic growth but it can increase gust wind speed, which may lead to serious damage to infrastructures. This study uses the Weather Research and Forecasting model and a gust parametrization scheme to evaluate the mitigation impact of white roofs and green roofs on wind gust over the Pearl River Delta, an urban agglomeration in Southern China in June, July, and August of 2014. The results show that both white and green roofs decrease the gust wind speed by decreasing the mean wind speed, suppressing the turbulent motion and weakening the convection. The impacts of white roofs are stronger than those of green roofs. The daily mean reductions of gust wind speed are approximately 1.2–1.3 m s−1 (12%–16%) and 0.4–0.6 m s−1 (6%–10%) by white and green roofs, respectively. In general, the contribution of turbulence (60%–85%) to the gust wind speed is the largest, and the contribution of mean wind speed is approximately 10%–30%, however, the effect of deep convection is not obvious (0%–15%) on the decrease of gust wind speed. The effect of cooling roofs on reducing the gust wind speed is stronger during daytime than during nighttime, and the effect is more significant in city areas that have higher building densities. Based on the findings, this study is potentially beneficial for policy-makings in developing urban disaster mitigation methods.


Introduction
Wind gust with large momentary wind speed is a disastrous weather that may induce great socioeconomic impacts as it could cause serious damage to infrastructures such as bridges and power lines, traffic and transportation, industrial and agricultural production, and human safety (Fink et al 2009, Suzuki 2009, Suomi et al 2013, Hawbecker et al 2017, Montenegro et al 2020).Therefore, it is need to develop strategies to mitigate wind gust, especially over urban areas where the above elements are concentrated (Liu et al 2023).Gust wind speed, by which wind gust does harm, is defined as the maximum value in a time series of three-second moving average wind speed during a ten-minute period (World Meteorological Organization 2014).Some researchers also take the hourly or daily maximum wind speed as the gust wind speed (Patlakas et al 2017, ECMWF 2021).
Urbanization has been rapid in China over the past 30 years, bringing about environmental problems such as the urban heat island (UHI) and urban air pollution (Oke 1982, Ryu et al 2013).Many researchers have looked into the influence of urbanization on meteorology, air quality, human thermal comfort, and so forth (e.g., Zhao et al 2014, Georgescu et al 2014, Huszar et al 2018, Li et al 2019, Wang et al 2020).However, there have been few researches investigating the impact of urbanization on wind gust.Liu et al (2023) evaluated the impact of urbanization on gust wind speed in Nanjing megacity, China, finding the gust wind speed was increased by urbanization due to strengthened turbulent flow in the planetary boundary layer (PBL), although in the meantime the mean wind speed was decreased.Therefore, urbanization deteriorates the urban environment not only by raising air temperature and worsening air quality but also by increasing the gust wind speed.Mitigating these problems requires researchers and policy-makers to develop corresponding strategies and methods.
At present some methods are proposed to mitigate the UHI intensity and improve human thermal comfort in city areas, such as cooling roofs (e.g., Li et al 2014, Sun et al 2016, Zhang et al 2017, Imran et al 2018, Jacobs et al 2018, Ma et al 2018).The cooling roofs generally include white roofs (WRs) and green roofs (GRs).WRs refer to the rooftops brushed by light colors and/or coated by other materials reflecting the shortwave radiation.GRs refer to the rooftops coated by vegetations releasing water vapor through evapotranspiration (Li et al 2014, Yang et al 2015).The above-mentioned studies indicate both WRs and GRs can reduce the UHI intensity and improve human thermal comfort in city areas.
However, whether and how WRs and GRs could affect wind gust in urban areas has not been studied yet.Previously, the study on wind gust has primarily been in the field of wind engineering and focused on the local impacts of wind gust on infrastructures such as airports and bridges, etc, while it has received less attention in the research community of atmospheric sciences largely due to lack of direct observations of wind gust.On the other hand, there were few schemes of wind gust parametrization available until recent years when more parametrizations have been developed (e.g., Suomi 2017, Sheridan 2018, ECMWF 2021), as more researchers in the community of atmospheric sciences have become aware of the need to simulate the wind gust because it is severe weather.Therefore, less attention has been paid to the mitigation of wind gust before, let alone the effect of cooling roofs on wind gust in urban areas.
The wind gust results from the turbulent flow in the PBL as well as from the deep convection-induced downward momentum transport into the PBL.Consequently, present parametrizations of the gust wind speed are computed as a function of mean wind speed, turbulence, and deep convection (e.g., Suomi 2017, Kurbatova et al 2018, Sheridan 2018, ECMWF 2021).Therefore, it may be expected that cooling roofs can affect the gust wind speed by reducing the turbulent flow and convection in the PBL, which may be a likely mitigation strategy for wind gust in urban areas.
The Pearl River Delta (PRD) region is a typical urban agglomeration of megacities and one of the regions with the largest population density in China (Guangdong Provincial Bureau of Statistics 2020).The global warming along with urbanization synergistically aggravates the UHI effect that has shown an increasing trend in this area (You et al 2017, Luo and Lau 2017, Xie et al 2020).Methods to improve the urban thermal environment are urgently needed for the PRD areas, which include the application of cooling roofs (Cao et al 2015;Chen et al 2023).Thus, it is need to analyze the impact of cooling roofs on mitigating the gust wind speed over the urban agglomeration in the PRD for more advised city planning and mitigation methods.

Model and Settings
The WRF (Weather Research and Forecasting) model version 3.9.1.1 (Skamarock and Klemp 2008) is used to evaluate the effect of WRs and GRs on gust wind speed.For this work, the land use data for the PRD region of 2014 was the latest availability, while data for more recent years is still being processed.Since the shortwave radiation is strongest and cooling roofs have the largest effects in summer in PRD, the summer of 2014 is selected as the study period.The WRF simulation domain is shown in figure S1 in supporting information.The urban grid pixels in the inner domain are replaced by the Google Earth satellite data (Dai et al 2019), in which the urban areas are then classified into three categories: COI (commercial or industrial), HIR (high intensity residential) and LIR (low intensity residential) areas.There are three tested scenarios in this work: the first one is a baseline hereafter called 'CTL', in which the cooling roofs are not considered at all, the second one is hereafter called 'WRs', in which all the roofs are white roofs, whose albedo is set to be 0.9, the third one is hereafter called 'GRs', in which all the roofs are green roofs.The selection of physical parametrizations is identical with Wang et al (2023).The more detailed model settings are shown in supporting information.

Data and validation
The model performance is validated by comparing the CTL results with observations of the 2 m air temperature (T 2 m ), 2 m relative humidity (RH 2 m ) and 10 m mean wind speed (WS 10 m ) from 45 weather stations.The detailed descriptions of the observational data and validation are shown in supporting information.Overall, the comparisons indicate the model performance is good, and the simulations well reproduce the observations by capturing their trends and being close to their values with small deviations.Therefore, the model is reasonable to assess the impact of cooling roofs on gust wind speed in the PRD urban agglomeration areas.

Wind gust parametrization
The wind gust parametrization used in our work is based on the scheme of the ECMWF (2021).The difference is that we use the mean wind speed at the first model level instead of the WS 10 m to compute the gust wind speed in order to reflect the impact of urban surfaces (Loridan et al 2013, Li and Bou-Zeid 2014, Sun et al 2016).The detailed parametrization is shown in supporting information.

Results and analysis
Many previous researches (Li et al 2014, Cao et al 2015, Imran et al 2018, Jacobs et al 2018) have found that cooling roofs can effectively reduce the UHI effect in urban areas.Our recent study (Wang et al 2023) has also found that cooling roofs can reduce the canopy and surface UHI intensity in the PRD region.In this study, we further analyze the air temperature differences at the first model level between WRs/GRs and CTL experiment (figure S4 in supporting information) and find that the air temperature significantly decreases in the PRD region.In the following sections, we focus on the impact of cooling roofs on wind gust in the PRD region.

Mean and gust wind speed
Figure 1 illustrates the spatial changes in mean wind speed at the first model level (approximately 20 m above ground level) between WRs/GRs and CTL case averaged over the daytime, nighttime and the whole time.The relative differences in the mean wind speed can be found in figure S5 in supporting information.In this study, the daytime spans from 0700 to 1800 (LST), while the nighttime spans from 1900 to 0600 (LST), and the difference of a variable is calculated as the variable in WRs or GRs case minus the counterpart in CTL case.
It is shown that the mean wind speed is decreased in urban areas after the cooling roofs are introduced, the daily mean decreases are about 0.2-0.3m s −1 (4%-12%) and 0.1-0.2m s −1 (4%-8%) by WRs and GRs, respectively.The decrease is due to a cooler thermal environment induced by cooling roofs in urban areas (figure S4), which increases near-surface stability that in turn weakens the transport of momentums, as revealed by previous studies (Sharma et al 2016, Sun et al 2016, Zhang et al 2017, Imran et al 2018).In addition, it is also found that the reductions of the mean wind speed induced both by WRs and GRs are larger during the day than during the night, suggesting the overall effect of cooling roofs on mitigating urban heat is stronger during the day, although the diurnal differences in the reductions induced by GRs are much smaller than those by WRs.WRs and CRs induce 0.3-0.5 m s −1 (12%-16%) and 0.1-0.2m s −1 (4%-8%) reductions of the mean wind speed in the daytime, respectively, indicating WRs are more effective in the mitigation than GRs, as being consistent with previous studies (Sharma et al 2016, Zhang et al 2017, Imran et al 2018).
The spatial changes of the gust wind speed at the first model level are shown in figure 2, and the relative differences are illustrated in figure S6 in supporting information.The spatial and temporal trends of the gust wind speed variation caused by WRs and GRs are basically consistent with those of the mean wind speed difference, but the values are significantly greater.This is because the decrease of the mean wind speed does not necessarily lead to the decrease of the gust wind speed (e.g., Liu et al 2023), and the cooling roofs also affect the gust wind speed by influencing the turbulence and deep convection.The results show that both WRs and GRs are effective in reducing the gust wind speed, and the most significant decreases are induced by WRs and happen in the daytime.During the day, WRs results in a decrease of 1.6-1.8m s −1 (16%-18%) in gust wind speed in urban areas, with a maximum of nearly 2 m s −1 (20%) in COI areas, while GRs leads to a decrease of 0.8-1.0m s −1 (8%-10%) in gust wind speed in urban areas.The daily mean reductions of gust wind speed are 1.2-1.3m s −1 (12%-16%) and 0.4-0.6 m s −1 (6%-10%) by WRs and GRs, respectively.
To further reveal the impact of cooling roofs on the variation of gust wind speed, the probability distributions of the difference in gust wind speed between WRs/GRs and CTL experiment in urban areas in the daytime, nighttime and whole time are shown in figure 3. The results show that both WRs and GRs lead to obvious negative changes for the whole day, indicating that cooling roofs decrease gust wind speed in urban areas.Moreover, WRs induces a significantly greater decrease of the gust wind speed during daytime than during nighttime, as the reduction of the gust wind speed with the largest occurring probability is about 2.0 m s −1 , 0.4 m s −1 and 0.6 m s −1 for the daytime, nighttime and whole time, respectively.GRs induces a slightly greater reduction of the gust wind speed in the daytime than in the nighttime, and the reduction of the gust wind speed with the largest occurring probability is about 1.0 m s −1 , 0.3 m s −1 , and 0.5 m s −1 for the daytime, nighttime and whole time.Overall, the occurrence probability of the reduction of the gust wind speed also indicates that WRs has a larger impact on mitigating the gust wind speed over urban areas than that of GRs.

Turbulence and wind shear
At the right-hand side of equation (S1) in supporting information, the first, second and third terms represent the contribution of the mean wind speed, turbulence and deep convection (regulated by the wind shear between 850-hPa and 950-hPa pressure level) to the gust wind speed, respectively.Since the friction velocity and stability are two important variables characterizing the turbulence in the surface layer, it is need to examine the variations of friction velocity, stability and wind shear to understand variations of the gust wind speed.
Figures 4 and 5 show the spatial changes of the friction velocity, stability (expressed as z L, / where z is approximately 20 m high (the first model level) and L is the Monin-Obukhov length, which is defined by equation (S3)), and mean wind speed variation (i.e., wind shear) between 850-hPa and 950-hPa pressure level between WRs/GRs and CTL experiment averaged over the day, night and whole time.For more information, the spatial differences of the three variables in the percentage as well as the spatial differences of the mean wind speed at 950-hPa and 850-hPa pressure level are also shown in figures S7-S10 in supporting information.
The friction velocity decreases in urban areas when adopting cooling roofs.This is because the friction velocity is related to the standard deviations of the wind speed fluctuation and turbulent intensity.As the latter two decrease, the friction velocity also decreases in urban areas.Overall, the reductions of friction velocity are greater by WRs than those by GRs, as the daily mean reductions are 0.05-0.08m s −1 (12%-16%) and 0.01-0.04m s −1 (8%-12%) by WRs and GRs, respectively.The WRs-induced reductions are significantly larger during   daytime than during nighttime, which exceeds 0.10 m s −1 (20%) over the COI areas.The reductions caused by GRs are slightly larger during daytime than during nighttime, which are between 0.02-0.05m s −1 (8%-12%) in urban areas.
The stability increases (i.e., becoming more stable) when adopting cooling roofs in urban areas.Since the reduction of temperature is more obvious in the daytime, the increases in stability by WRs are more significant in the daytime compared to those in the nighttime.The WRs-induced stability increases approximately 0.10-0.40(5%-25%) at daytime, while it increases 0.10-0.20 (5%-15%) at night over the PRD region.The increases in stability caused by GRs are weaker compared to those by WRs, which are approximately 0.10-0.30(5%-20%) and 0.10-0.20 (5%-15%) during daytime and nighttime, respectively.
The wind shear increases over the PRD region when the cooling roofs are introduced.This is because the temperature reduction leads to a decrease in the mean wind speed within the urban PBL but an increase above the PBL (Sharma et al 2016, Zhang et al 2017, Imran et al 2018, Wang et al 2019, 2020).Moreover, the enhanced stability could also nonlinearly increase the wind shear in urban areas.The increases in wind shear caused by WRs are larger compared to those caused by GRs, and the increases are more obvious during daytime than during nighttime.The daily mean wind shear raises 0.25-0.40m s −1 (25%-40%) and 0.10-0.25 m s −1 (15%-30%) caused by WRs and GRs in the PRD region, respectively.The larger effect of WRs on wind shear is because the variations of the mean wind speed at 850-hPa pressure level caused by WRs are not significantly different from those by GRs, while the mean wind speed at 950-hPa pressure level is more reduced by WRs than by GRs (as shown in figures S9 and S10), as the mean wind speed at 850-hPa pressure level is less impacted by the urban boundary layer than the mean wind speed at 950-hPa pressure level is (Liu et al 2023).
The enhanced wind shear tends to increase the vertical momentum transport.However, the increased stability and enhanced wind shear by the cooling effect have opposite effects on the vertical momentum transport, since the former weakens it while the latter increases it.Consequently, the vertical momentum transport is the result of the combined impact of increased stability and enhanced wind shear by the cooling effect.Although the vertical momentum transport increases the gust wind speed, the gust wind speed is eventually decreased, which is explained in the following chapter.

Explanation for variation of gust wind speed
To further study the relationships of the gust wind speed to the mean wind speed, turbulence and deep convection, the diurnal changes of the mean wind speed, gust wind speed, friction velocity, stability and wind shear in each urban type (i.e., COI, HIR, LIR) and averaged over all the urban grids (i.e., Mean) are shown in figure 6.In general, the effects of WRs on these items are obvious, particularly during daytime and over urban areas with high building densities where the rooftop coverage fractions are greater.The effects of GRs are weaker than those of WRs, and the GRs-induced changes have weaker diurnal variations and are smaller in each of the three urban types.
The results show that the maximum decreases in mean wind speed caused by WRs are close to 0.45 m s −1 and 0.35 m s −1 in COI and LIR areas, respectively.The decrease of mean wind speed caused by GRs is less than 0.2 m s −1 for the whole day.The behavior of the changes in gust wind speed by WRs/GRs is similar to that of the mean wind speed but has greater amplitude.The maximum reduction of WRs-induced gust wind speed is about 2.3, 2.0 and 1.6 m s −1 in COI, HIR and LIR areas, respectively, while the GRs-induced decreases are between 0.2-0.9m s −1 in the three urban types for the whole day.The trends of the friction velocity reductions are close to those of the mean wind speed, which is because the friction velocity is calculated by a parametrization controlled by the mean wind speed and stability (Liu et al 2013).The stability increases in all the urban types, and the largest increment is about 0.55 and 0.30 by WRs and GRs in COI areas, respectively.A stabler atmospheric stratification could reduce turbulence and convection (Wang et al 2019), leading to reductions of the gust wind speed.The cooling roofs can also enhance the wind shear, which results in increases of the gust wind speed.The increases are the largest in the afternoon, which are up to 0.95 m s −1 and 0.45 m s −1 by WRs and GRs in COI areas, respectively.
In order to examine the contributions of the mean wind speed, turbulence and deep convection to the gust wind speed, figure 7 illustrates the diurnal variations of the gust wind speed as well as its components averaged over all the urban grids.The relative contributions by each component can be found in figure S11.Although the wind shear increases the vertical momentum transport that increases the gust wind speed, according to the parametrization equation (S1), the mean wind speed and atmospheric turbulence (i.e., the first and second term on the right-hand side of the equation) have much larger weight than the wind shear (i.e., the third term on the right-hand side of the equation).Therefore, the gust wind speed is eventually decreased by the cooling effect through the reduced mean wind speed and weakened atmospheric turbulence, although the enhanced wind shear tends to increase it.
Overall, both the mean wind speed and turbulence have negative impacts on gust wind speed (i.e., causing the gust wind speed to decrease), whereas the deep convection has a positive impact.The decreases of the gust wind speed caused by WRs are greater than those caused by GRs during daytime, which is because the turbulence generated by WRs is significantly weaker than that by GRs due to a greater reduction of friction velocity and a stabler atmospheric stratification.The contributions of mean wind speed caused by WRs and GRs to the reduction of gust wind speed are 0.1-0.4m s −1 (10%-30%) and 0.0-0.2m s −1 (10%-25%) for the whole day, respectively.The contributions of turbulence caused by WRs and GRs to the reduction of gust wind speed can be up to 1.6 m s −1 (85%) and 0.7 m s −1 (85%) during daytime, respectively.On the other hand, the increase in deep convection is relatively small for most of the day, implying a weaker effect on the reduction of gust wind speed.For the differences of contributions among the three urban types, which are not shown here, the contributions of turbulence both caused by WRs and GRs are greater in COI areas than those in HIR and LIR areas, leading to larger decreases in gust wind speed in COI areas, while the differences in the contributions of mean wind speed and deep convection among the urban types are small.

Conclusions and discussion
In this work, the WRF-v3.9.1.1 model with a gust parametrization scheme is used to assess the impact of cooling roofs on wind gust over the Pearl River Delta, an urban agglomeration in Southern China.Both WRs and GRs decrease the gust wind speed, and the impact of WRs is greater than that of GRs.This is because WRs reflects the shortwave radiation, thus decreasing the total energy in the surface energy balance during daytime, while GRs decreases the sensible heat flux through reassigning part of the total energy from the sensible heat flux to the latent heat flux by evapotranspiration, which essentially does not change the total energy in the surface energy balance.Therefore, by decreasing the total incoming energy, WRs is more effective in cooling the urban thermal environment than GRs, which, in turn, further increases the stability and suppresses the turbulent motion, thus contributing to a larger reduction of the gust wind speed.
The daily mean reductions of gust wind speed are 1.2-1.3m s −1 (12%-16%) and 0.4-0.6 m s −1 (6%-10%) caused by WRs and GRs, respectively.Overall, the turbulence contributes 60%-85% to the decrement of gust wind speed, which is due to significant reduction of friction velocity and stabler atmospheric stratification.The reduction of mean wind speed contributes 10%-30% to the gust wind speed.Although the wind shear obviously increases in the afternoon, due to its relatively small weight in computing the gust wind speed, the deep convection has a small impact (0%-15%) on the decrement of gust wind speed.The effects of WRs and GRs on reducing the gust wind speed are stronger during daytime than during nighttime, and the impacts are most significant in COI areas while least significant in LIR areas, which is because the rooftop coverage fraction is highest in COI areas while lowest in LIR areas.
Urbanization may cause many environmental problems such as UHI and increased wind gust.With the further development of urbanization, these problems will become more serious, thus mitigation strategies are critically needed for these problems.While there have been some measures to mitigate the UHI effect, so far there have been few approaches to mitigate the wind gust in urban areas yet.On the other hand, because of the restricted natural underlying surface and vast area of rooftops over urban areas, WRs and GRs have already become the most cost-effective method to reduce the UHI intensity.However, before this study, cooling roofs are not known to be capable to mitigate the wind gust in urban areas.This study shows that cooling roofs can significantly reduce the gust wind speed in urban areas, which is potentially beneficial for policy-makings in developing urban disaster mitigation measures.

Figure 1 .
Figure 1.Spatial differences of the mean wind speed (WS) at the first model level between WRs/GRs and CTL case (top panel: WRs output minus CTL output, bottom panel: GRs output minus CTL output) averaged over the daytime (left panel), nighttime (middle panel) and whole time (right panel) of June, July and August of 2014.The black hatching indicates the 95% confidence level from the Student's t-test.

Figure 2 .
Figure 2. Spatial differences of the gust wind speed (GWS) at the first model level between WRs/GRs and CTL case (top panel: WRs output minus CTL output, bottom panel: GRs output minus CTL output) averaged over the daytime (left panel), nighttime (middle panel) and whole time (right panel) of June, July and August of 2014.The black hatching indicates the 95% confidence level from the Student's t-test.

Figure 3 .
Figure 3. Probability distributions of the gust wind speed (GWS) difference between WRs/GRs and CTL case (top: WRs output minus CTL output, bottom: GRs output minus CTL output) in the urban areas averaged over the daytime (Day), nighttime (Night) and whole time (Mean) of June, July and August of 2014.

Figure 4 .
Figure 4. Spatial differences of the friction velocity (a, b, c), stability (d, e, f) and wind shear (g, h, i) between WRs and CTL case (WRs output minus CTL output) averaged over the daytime (left panel), nighttime (middle panel) and whole time (right panel) of June, July and August of 2014.The black hatching indicates the 95% confidence level from the Student's t-test.

Figure 5 .
Figure 5. Same as figure 4, but between GRs and CTL case (GRs output minus CTL output).

Figure 6 .
Figure 6.Diurnal variations of the differences of the mean wind speed (a, b), gust wind speed (c, d), friction velocity (e, f), stability (g, h) and wind shear (i, j) between WRs/GRs and CTL case (left panel: WRs output minus CTL output, right panel: GRs output minus CTL output) averaged over three urban types (i.e., COI, HIR and LIR) as well as over all the urban grids (i.e., Mean).

Figure 7 .
Figure 7.The hourly-averaged differences of the gust wind speed and its contributing components between WRs/GRs and CTL case (a: WRs output minus CTL output, b: GRs output minus CTL output) in the urban areas.