Measuring the impacts of a real-world neighborhood-scale cool pavement deployment on albedo and temperatures in Los Angeles

Between 7 October 2019 and 15 November 2019 ∼59 020 m² of cool pavements were deployed in a residential neighborhood in Covina, California, which is located ∼30 km east of downtown LA (see figure 1). The installation area is shown in figure 1 and is hereafter referred to as the 'impact area.' The impact area measures approximately 0.8 km × 0.8 km and is centered at approximately (34.096, −117.921). Full details regarding the installation schedule can be found in figure 2, since measurements were taken before, during, and after all installations were completed. Directly west of the impact area is the 'control area,' which is comparable in size to the impact area, has similar land cover features, and did not have cool pavement installations. (see figure 1). Measurements in an adjacent and upwind control area were necessary in order to properly control for factors other than the cool pavement installation that may affect the temperature comparisons. Wind data from the nearest Automated Surface Observing Systems weather station confirm that the control area was upwind of the impact area (see figure S5 available online at stacks.iop.org/ERL/17/044027/mmedia).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

This pilot installation was coordinated and supervised by the LA County Department of Public Works (LADPW) in order to test the real-world performance and impacts of cool pavement. The companies that produced each cool pavement product were responsible for the installation of the cool pavement. The site was chosen by LADPW primarily for logistical reasons, because the area was under their jurisdiction; but the site also exhibits other ideal traits. For example, the site is located in the eastern side of LA County, which generally has higher temperatures relative to downtown LA. The average maximum summer temperature between 1981 and 2010 at the closest long-term weather station to the site (Pomona Fairplex station) was 31.3 °C, which is 3.6 °C higher than the same metric for downtown LA (www.ncdc.noaa.gov/cdo-web/datatools/normals, last access 3 February 2022). Furthermore, the site is located in census tract 4057.01, which has a tree canopy cover of only ∼3.3% (www.cal-heat.org, last access 3 February 2022). These local characteristics make this site a good candidate for this pilot study.

Albedo, T_surface, and T_air were measured at varying sampling frequencies, time intervals, and spatial coverage. Pavement albedo was measured using two Kipp & Zonen SMP6 Smart Pyranometers. Stationary measurements of albedo were taken at seven distinct locations, both before and after installation (see figure S1). These locations were chosen to represent variability due to the different cool pavement products and road wear from traffic. Albedo spot measurements were sampled at 1 Hz and average albedo was computed using 1 min means. Mobile albedo measurements were also taken at 1 Hz and the data were paired with GPS measurements. For stationary albedo, the height of the bottom pyranometer dome was positioned at ∼50 cm above the ground, according to ASTM E1918-21 [71]. For mobile albedo, the height was positioned at ∼60 cm due to the cart configuration, however diagnostic testing was conducted prior to the campaign to ensure that the minor deviation from ASTM E1918-21 would not significantly impact the albedo measurements (see supplement for details).

Mobile transect measurements of 1.6 m T_air and T_surface were taken at a sampling rate of 1 Hz with an Apogee ST-110 thermistor and Apogee SI-111SS infrared radiometer, respectively. The thermistor was shielded with a white PVC pipe segment (3.81 cm nominal diameter, 15 cm length) in order to minimize the effects of incident and reflected radiation. These instruments were mounted on a car and measurements were taken first in the impact area and then the control area for each unique transect (see figure 3). For each day of measurements, transects were performed every three hours at the top of the hour, with the first transect starting at 09:00 local standard time (LST) and the last transect starting at 21:00 LST. Transect measurements were taken on three days pre-installation, and three days post-installation.

Zoom In Zoom Out Reset image size

Stationary T_air was continuously sampled every five minutes, between 27 August 2019 and 6 November 2019 at four locations in the impact area and five locations in the control area (see figure S2). HOBO U23-004 External Temperature Data Loggers were used with HOBO RS3-B solar radiation shields for stationary T_air. These sensors were placed ∼3 m above ground level on street sign poles (see figure 3). They were installed at 3 m height in order to minimize risks of vandalism or theft. Note that only data from non-daylight hours were used for the stationary T_air analysis due to an increase in reflected shortwave radiation at the surface potentially causing a confounding warming signal.

All GPS measurements were taken with an iPhone 6s at a frequency of 1 Hz, using an application called GPS Tracker. Representative photos of the various instrumentation set-ups are shown in figure 3.

In order to quantify the impact of cool pavements on T_surface and T_air, we used a technique called the DID method. The DID method originates in econometrics but has since been used in many quantitative fields [72–82]. In short, the DID method calculates the direct impact of a treatment/event on some variable by using measurements of the variable of interest in both a control group and an impact group, both before and after the event of interest. This results in four distinct subsets of measurements: (a) pre-event impact, (b) pre-event control, (c), post-event impact, and (d) post-event control. These measurements allow us to calculate how the difference between the impact and the control changes over time. The post-event versus pre-event difference in the impact versus control difference (i.e. the DID) can then be attributed as the direct impact of the event on the variable of interest. A simplified illustration of the DID method is shown in figure 4. Further details regarding the DID method can be found in the supplement.

Zoom In Zoom Out Reset image size

Based on mobile albedo measurements taken on 17–18 September 2019, the mean (± standard deviation) pre-installation pavement albedo of the impact area was 0.08 ± 0.02 (see figure 5). Seven stationary measurement locations (S1 through S7) were distributed throughout the impact area. As shown in figure 5, the albedo at the seven locations ranged from 0.22 to 0.38 on 21 October 2019 five days after cool pavement installation. This corresponds to an albedo increase of 175% to 363% relative to the original asphalt pavement. Subsequently, we observed a decline in albedo at all locations, S1 through S7. On 18 November 2019 the stationary albedo measurements ranged from 0.20 to 0.29. The albedo degradation from the first set of post-installation measurements (21 October 2019) to the third set (18 November 2019) ranged from 8% to 30%. A heavy rain event occurred on 20 November 2019 and the albedo increased at all stationary measurement locations. Albedo measurements on 22 November 2019 two days after the rain event, ranged from 0.23 to 0.31. The albedo increases due to the rainstorm ranged from 7% to 17%, corresponding to an absolute albedo increase of 0.02–0.04. This rain event shows that natural precipitation can wash away dust from deposition and dirt from tire wear to a limited but noticeable extent.

Zoom In Zoom Out Reset image size

Albedo measurements were also taken ∼10 months following the first phase of cool pavement installation (figure 5). Albedo at all spot measurement locations decreased markedly, with values ranging from 0.16 to 0.24. Pavement albedo degradation over the ten month timeframe ranged from 12% to 49%. Spot measurement results are summarized in table 1. Of note, we observed a wide variability in albedo degradation, which we attribute to the performance variability between different cool pavement products and the spatially heterogenous traffic road wear.

The spatial heterogeneity of albedo can be further observed in figure 6. Panels (a)–(c) show pavement albedo approximately one week, one month, and one year after installation. One week after installation, we already observed substantial spatial variability in albedo, which was mainly driven by different initial albedos of the products. The southern half of the impact area was coated with product 3, while the northern half was coated with products 1 and 2. The corresponding albedo is generally higher in the northern area relative to the southern area. This spatial pattern is consistent with the stationary albedo measurements (see figure 5).

Zoom In Zoom Out Reset image size

An additional source of spatial variation was caused by different patterns of tire wear. Corners, intersection, cul-de-sacs, and bends were more likely to exhibit greater tire wear. This effect was observed in the mobile albedo measurements (see figures 6(a)–(c)), and from visual inspection of satellite imagery (see figures 6(d)–(f)). Over the timespan of a year, the average albedo of the three cool pavement products decreased from 0.26 to 0.18. This corresponds to a spatially averaged albedo degradation of 30% over the year.

While others have measured the effects of aging on cool pavement albedo in either laboratory settings or small-scale field measurements [83–85], we present here the most extensive results to date from long-term albedo monitoring of neighborhood-scale cool pavements subjected to both vehicular traffic and natural weathering. A study by Lontorfos et al [86] reported a one-year albedo reduction from 0.26 to 0.15 for a 4600 m² cool pavement installation on a single street in Athens, Greece. This is consistent with the range of albedo degradation we observed in this study. One important distinction between Lontorfos et al [86] and our study is that they reported a majority of the albedo degradation occurring within a month of installation, while we continued to observe albedo degradation following the month timeframe (see figure 5). This implies that cool pavement albedo degradation is not universally consistent and can be dependent on the context of the physical environment and the cool pavement product.

Akbari and Matthews [30] reported that aged asphalt pavements have albedos that range from ∼0.10 to 0.18. In our study, the mean cool pavement albedo decreased to 0.18 within one year. This implies that aged cool pavements subjected to real-world exposure may have albedo values similar to that of aged asphalt pavement. However, it is interesting to note that the rates of albedo change for traditional asphalt and cool pavements may not be equal is magnitude. For example, the control area in our study was freshly paved with traditional black asphalt shortly before the campaign, and we observed that albedo stayed consistent at ∼0.05 over the duration of a whole year. This implies that traditional asphalt pavement may take much longer to reach a steady-state increased albedo than it does for cool pavement to reach a steady-state decreased albedo. An oft-repeated concern with cool pavements is that vehicular traffic could result in rapid decline in its heat mitigation efficacy. This study highlights the critical need to plan for albedo degradation if cool pavements are deployed in the real world, either through improved maintenance, higher performance materials, and/or regular reapplication.

Pavement T_surface was measured as part of the mobile transects described in section 2.2. Approximately one week after installation, we observed lower T_surface throughout the day in the impact area relative to the control area, with mean T_surface reaching up to 40.0 °C and 44.9 °C in the impact and control areas, respectively (see figure S6). Figure 7 shows a matrix of barplots, where each barplot presents the DID at different hours throughout the day. These DID values were calculated using the mobile T_surface measurements from a pair of measurement days. For example, the barplot in the top-left corner shows T_surface DID values calculated using measurements from 23 September 2019 and 05 October 2019. In figure 7, the values shown in the red rectangle refer to the impact of cool pavement installation on T_surface. Our measurements show that T_surface reductions due to cool pavement installation follow a consistent diurnal cycle, with a peak reduction of 5 °C observed at 15:00 LST. The smallest reduction of 0.9 °C was observed at 09:00 LST.

Zoom In Zoom Out Reset image size

The barplots outside of the red outline in figure 7 show the DID values calculated for either pairs of pre-installation dates or pairs of post-installation dates. The pre-installation and post-installation DID values in the blue boxes in figure 7 are nearly all centered about zero, with the 95% confidence intervals implying a null effect. The null results for pre-installation and post-installation date pairs confirm that the DID method is correctly controlling for external variabilities and can robustly quantify the effect of cool pavement installation.

In the two post-installation pair DID plots in the last row of figure 7, the slightly positive DID values (with the exception of 09:00 LST) imply that there was a 'treatment' that occurred between 25 October 2019 and 14 August 2020 that increased the T_surface of the cool pavement. The treatment in this context was the degradation of the cool pavement albedo. In other words, the albedo degradation within a ten month time period reduced the ability of the cool pavement to reduce T_surface by ∼1 °C during the day.

Overall, an average pavement albedo increase of 0.18 led to a maximum T_surface reduction of 5 °C. This corresponds to a T_surface reduction rate of 2.7 °C per 0.1 increase in pavement albedo. These T_surface reductions are consistent with the range of what previous studies have reported. Most recently, Middel et al [22] reported that cool pavements were up to 6 °C cooler than asphalt pavements. Furthermore, we found that an albedo degradation of 0.08 after approximately one year of real-world exposure reduced the cooling efficacy for T_surface by ∼1 °C. By linear extrapolation, we estimate that an albedo degradation of 0.1 will lower the cooling efficacy for T_surface by ∼1.25 °C.

As described in section 2.2, we measured both stationary and mobile T_air. Figure 8 shows a scatterplot of Δ_post versus Δ_pre using data from the stationary T_air measurements, where Δ signifies the difference in T_air between the impact area and the control area, and 'post' or 'pre' signifies whether the measurements were taken post-installation or pre-installation, respectively. Note, the post period for air temperature measurements described here refers to the period between the completion of the first installation phase and the start of the second installation phase. This means that the impact on air temperature described here is attributable to the cool pavements installed in the first phase (see figure 2). Our measurements indicate that cool pavements induced a 3 m T_air reduction between 19:00 and 23:59 LST, though only reductions for the hours of 20:00, 21:00, and 22:00 were statistically distinguishable from zero at a 95% confidence level. A maximum evening 3 m T_air reduction of 0.19 °C occurred at 21:00 LST.

Zoom In Zoom Out Reset image size

From our mobile transect measurements, we found that cool pavements reduced 1.6 m T_air by a mean ± 95% confidence interval of 0.20 ± 0.06 °C at 12:00 LST. A statistically distinguishable DID signal was only obtainable for 12:00 LST due to high micrometeorological variability and the small T_air reduction signal expected from the scale of the cool pavement installation in this study. Note that the absence of statistically significant results for other hours of the day does not preclude cool pavements from having caused changes in 1.6 m T_air at those other hours. For these null hours, the noise due to micrometeorological variability within the timeframe of the mobile transect made it difficult to distinguish the magnitude of the T_air signal. Further details regarding these complications and the statistical procedures can be found in the supplement.

Our stationary and mobile T_air measurements imply that cool pavements can reduce near-surface T_air during the daytime, as well as during the evening. We were unable to identify a definitive diurnal pattern of near-surface T_air reduction. However, we can conclude that the neighborhood-scale cool pavement installation investigated here can reduce near-surface T_air by 0.20 °C around noon, and at least 0.19 °C in the late evening, within the context of the local meteorology and land surface features specific to our field site.

This is the first study to report statistically significant T_air reductions caused by cool pavement installations, using a controlled experimental technique to account for confounding variables. Middel et al [22] reported maximum T_air reductions of ∼0.4 °C–0.5 °C, but pre-installation measurements were not taken as part of that study, which means that confounding variables could not be fully accounted for. We find that our results are consistent (i.e. on the same order of magnitude) with past modeling studies. Taleghani et al [33] used CFD modeling to investigate impacts of cool pavements at neighborhood scale and found that increasing pavement albedo by 0.3 could lead to a neighborhood-averaged reduction in T_air of ∼0.26 °C. Taha [31, 87] used mesoscale modeling to estimate T_air reductions of approximately 0.25 °C–0.5 °C in LA for moderate (0.15) to high (0.25) increases in pavement albedo for a city-scale installation. Mohegh et al [24] also used mesoscale modeling to investigate cool pavement impacts and found that increasing pavement albedo by 0.4 in California cities could reduce T_air by ∼0.18 °C–0.86 °C. Millstein and Levinson [88] used an idealized heat transfer model of air flowing over a 200 m hot plate and estimated a ∼0.15 °C reduction in 1.5 m T_air when albedo was increased by 0.3. This is similar to the maximum reduction of ∼0.20 °C we observed, considering that the maximum east-west street length in our impact area was ∼650 m. Table 2 places our results in the context of these studies by summarizing the abovementioned T_air reductions per 0.1 increase in pavement albedo. Although our results may be specific to neighborhood characteristics and local meteorology of our field site, this study can serve as an important point of reference for hot and dry regions like LA.

Although we do not report pedestrian thermal comfort measurements here in this study, we find it important to discuss the implications of cool pavements on thermal comfort. The cool pavement penalty of increased upwelling shortwave radiation has been noted in previous studies [22, 32, 70, 89, 90], yet the nuances of cool pavement tradeoffs are often overlooked by the public, potentially leading to the misconception that cool pavements can serve as a 'policy panacea' for urban heat mitigation [22]. Here we briefly review the main tradeoffs related to cool pavements' impact on thermal comfort in the context of our results.

The main concern with cool pavements is that the increased reflection of shortwave radiation at the ground level can decrease pedestrian thermal comfort significantly during daylight hours. Most notably, Middel et al [22] reported that cool pavements increased the shortwave radiation load for pedestrians standing directly over the pavement by up to 168 W m⁻². This corresponded to an observed mean radiant temperature (T_MRT) increase of 4 °C.

Although this observed increase in daytime T_MRT is certainly a cause for concern, we suggest that an increased radiant load directly over the pavement does not fully account for the complexities that impact outdoor thermal comfort. Controlling for variability of biometeorological factors, such as evaporative cooling rates, effects of clothing, and metabolic rates; there are only two ways that cool pavements can directly impact thermal comfort: (a) radiative heat transfer, or (b) convective heat transfer. For radiative heat transfer, cool pavements will increase shortwave radiative load on nearby pedestrians, but longwave radiative loads will be decreased due to lower pavement T_surface. For convective heat transfer, cool pavements can lower near-surface T_air, which would tend to make the pedestrian more comfortable in hot weather, all else being equal.

Based on previous measurements [91, 92], the radiative heat transfer coefficient (h_r) and the convective heat transfer coefficients (h_c) for the human body are approximately the same magnitude at low wind speeds (<1 m s⁻¹). This implies that an increase in radiative heat transfer due to a unit increase in T_MRT is similar in magnitude to an increase in convective heat transfer due to a unit increase in T_air. As a hypothetical example, if T_MRT increased by 4 °C (as observed by Middel et al [22] directly over cool pavement), but T_air was also reduced by 4 °C, the thermal comfort change should be negligible under low wind speeds.

The influence of wind speed further complicates the radiation-convection tradeoff. For illustration, the average wind speed at 12:00 LST recorded at LA International Airport during the summer is ∼3 m s⁻¹ [93]. According to de Dear et al [92], h_c/h_r is ∼4 when the wind speed is 3 m s⁻¹. In this case, if T_MRT increased by 4 °C, a simultaneous reduction in T_air of ∼1 °C would compensate for the increased radiative load. The reduction in T_air of 0.20 °C at 12:00 LST observed in our study is not sufficiently large to compensate for a T_MRT increase of 4 °C. However, this balance would change if T_air reductions exceed 1 °C. Based on a number of modeling studies [24, 32, 54], it is generally expected that larger spatial extents of cool pavement deployment would lead to larger reductions in T_air. While T_air reductions are expected to be strongly dependent on the spatial extent, the radiant load effect from increased shortwave radiation is not.

There is also considerable uncertainty regarding the effect of reflected shortwave radiation on pedestrians on the sidewalks, where most foot traffic is expected. It is well-known that the albedo of concrete sidewalks is already higher than that of asphalt pavement, with concrete albedo values generally exceeding 0.2 [30, 61, 94, 95]. Furthermore, the geometry and composition of the 'buffer zones' between sidewalks and roads vary considerably. Some sidewalks may have no buffer, while others may have vegetation distancing the sidewalks substantially from the road, or even parked vehicles or shrubs that block reflected shortwave radiation from the pavement. Although it is clear that the daytime radiative load would increase substantially for a pedestrian standing directly over the cool pavement, it is less clear how thermal comfort would be impacted for pedestrians on sidewalks. Middel et al [22] reported T_MRT on sidewalks versus on cool pavement, and also reported an additional 20–30 W m⁻² of reflected shortwave radiation on the sidewalks during the early evening. However, it is still uncertain how cool pavements impact thermal comfort for pedestrians on sidewalks in a range of real-world configurations.

Lastly, the potential tradeoffs between nighttime decreases in T_air and daytime increases in T_MRT need to be critically examined. There is growing evidence that nighttime temperatures are increasing faster than daytime temperatures, and there is increasing concern about resulting public health impacts [96–99]. Although cool pavements may decrease nighttime temperatures, it remains uncertain whether the potential benefits of nighttime temperature reductions outweigh the consequences of potentially decreasing daytime outdoor thermal comfort.

We suggest that the various tradeoffs (i.e. benefits and penalties) of cool pavements with respect to thermal comfort need to be quantified more explicitly in future research. A large portion of the uncertainty can be attributed to the wide variety of hyper-local features (e.g. sidewalk buffer characteristics, meteorology, vegetation) and the general lack of controlled real-world field measurements. This study confirms that cool pavement can generally reduce T_air, but it remains an open question whether cool pavements can play a viable role in increasing net human health and wellbeing, and just as importantly, in what context.