Micrometeorological effects and thermal-environmental benefits of cool pavements: findings from a detailed observational field study in Pacoima, California

Prior to the start of this study, a community in Pacoima, California was selected by Climate Resolve and GAF for installation of reflective pavements. When the research component was subsequently initiated, two adjacent and similar sections of the community were selected as test and reference areas to compare, side-by-side, the surface and near-surface micrometeorology before and after coating with reflective pavements (figures 1 and 2).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

The total area of the test community (blue polygon in figure 2) is 259,560 m² of which a total area of 59,570 m² (yellow highlight in figure 1) was coated with GAF's reflective materials, which represents 23% of the test community's area. While this appears to be smaller than reported in other studies, e.g., Schenieder et al (2023), the modifications are concentrated in a smaller area (so the modified surfaces exert a larger effect than if more spread out). The area of the adjacent reference community (red polygon in figure 2) is 250,530 m².

The test and reference areas have similar weather and flow patterns; they also have similar tree cover, continuity in and orientation of streets, sky-view factors, building heights and setbacks, urban morphology, and pavement types. Additional information on buildings, streets, the percentages of impervious surfaces and vegetation cover in the project area are provided in table A-1 of the Supplement. Figure 3 shows a typical sky-view factor/aspect ratio of streets in the study area and 1-m resolution tree cover. Buildings do not cast shadows on the streets and, conversely, increased road albedo does not reflect much radiation onto walls or other building surfaces. For the purpose of mobile observations, the segments of each street were defined as E (test) or W (reference) per their position (east or west) relative to Borden Street running at the junction between the test and reference communities (figure 2).

Zoom In Zoom Out Reset image size

Detailed morphological characterizations were carried out to demonstrate similarity between the test and reference areas (pre-coating) and ensure that observed side-by-side micrometeorological differences following reflective-coatings installation were caused by albedo modifications and not by physical or morphological differences. Furthermore, a difference-of-difference approach (discussed below) was also designed to cancel out any possible remnant effects, even if small or negligible, caused by such potential variations.

A multi-scale, multi-platform approach was designed to observe and characterize the effects of cool pavements on the surface and near-surface thermal environment and physical properties. From top of atmosphere down to surface, the platforms were:

• Satellites: MODIS for 500-m albedo, Landsat-9 for 30-m albedo, ECOSTRESS for 70-m land-surface temperature, and Sentinel-2 for 10-m albedo (citations in References section). Broad-band albedo was computed from narrow-band reflectance using suitable weighting functions (e.g., Liang et al 1999).
• Aircraft: NASA/JPL HyTES (Hulley et al 2016): data were obtained for several flyovers in the project area—they provide snapshot thermal imagery and temperature data (figure 4).
• Drone: GAF operated a drone over Pacoima, before, during, and after coating of various surfaces in the project area. The drones provided both visible and thermal imagery.
• Automated weather stations: Two weather stations (type ATMOS 41, METER Group) were installed in test and reference areas on City of LA street lighting poles (figure 5). The purpose was to capture above-canopy background conditions over the test and reference areas and ensure that any thermal environmental changes within the urban canopy were not caused by some changes in background conditions.
• Mobile platform: an electric golf cart was instrumented and used as the main observational platform in extensive mobile transects through the test and reference communities. An electric cart was selected to avoid the effects of tailpipe exhaust heat on temperature readings. A major advantage of the golf cart is that it is a fully-instrumented platform that can continuously cover the entire project area while at speeds lower than those of cars thus allowing a better capture of cool-pavements' effects.
• Reflectometer: Research-grade Solar-I (Surface Optics Corporation, SOC 2021) reflectometer was used for spot, in situ measurements prior to and after reflective coatings were installed (figure 5). Broad-band albedo was derived from various reflectance bands using weighting coefficients per Levinson et al (2010, 2017).

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Aside from the remote sensing and in situ spot measurements listed above, all observations in this study were made by (1) sensors on a mobile platform (electric golf cart) for surface, near surface, and in-canyon street-level observations, and (2) weather stations installed on light poles (one in test, the other in reference communities) for above canopy, background micrometeorological characterizations. The two weather stations operated continuously during the entirety of the project (and continue to do so) whereas the mobile transects were carried out during several days a month and several sorties a day, for a total of 73 full transects over a 1-year period (dates and associated weather conditions can be found in the Supplement figure A-1). Each transect was 6,755-m long (figure 2) and the cart was driven at between 2 and 4 m s⁻¹ with continuous measurements to capture community-wide effects, not just at predetermined locations or street intersections.

The cart was instrumented with research/science-grade sensors (Campbell Scientific and Apogee Instruments), configured and mounted on a frame specifically designed in this study (figure 6). Two METER Group weather stations were installed on City of LA light poles to monitor above-canopy background conditions (figure 5). Variables from cart and stations were logged at 5–10 s intervals including multi-level air temperature, relative humidity, wind speed, wind direction, wind gusts, solar radiation, precipitation, barometric pressure, lightning count, lightning distance, multi-directional short- and long-wave radiation, incoming and outgoing short- and long-wave radiation, surface temperature, surface albedo (net radiometer), and position (GPS). The net radiometer (sensor #11 in figure 6) was mounted on an extendable telescopic arm to avoid the shadow and view-factor effects from the cart. A list of the sensors and their relevant specifications are provided in table A-2 of the Supplement.

Zoom In Zoom Out Reset image size

Comparisons of micrometeorological variables between the test and reference segments of each street (figure 2) were made in two manners. One is in real time (RT), i.e., direct side-by-side differences at the same time, and the other is a difference-of-difference (DofD) applied at each street segment, time of day, and similar weather conditions, so that, for a variable V:

$$\begin{eqnarray*}&&{DofD}=\unicode{x02206}{{RT}}_{{post}}-\unicode{x02206}{{RT}}_{{pre}}\,\end{eqnarray*}$$

$$\begin{eqnarray}&&=\,\left({V}_{{test}.{post}}-{V}_{{ref}.{post}}\right)-\left({V}_{{test}.{pre}}-{V}_{{ref}.{pre}}\right)\end{eqnarray} \tag{ 1 }$$

For recurring similar environmental conditions at the reference community, i.e., when

$$\begin{eqnarray}&&{V}_{{ref}.{post}}\cong {V}_{{ref}.{pre}}\end{eqnarray} \tag{ 2 }$$

DofD is:

$$\begin{eqnarray}&&{DofD}={V}_{{test}.{post}}-{V}_{{test}.{pre}}=\unicode{x02206}V\end{eqnarray} \tag{ 3 }$$

which is the local change in V at the test area resulting from implementing cool pavements. Because DofD also accounts for baseline conditions, it is a more suitable indicator to the effects of cool pavements than RT alone. For micrometeorological variables, both RT and DofD are presented in this paper. For heat-wave conditions and some thermal-comfort metrics, only RT is presented as there were no pre-coating baseline conditions to compare with.

DofD is applied to the averages of readings at each street segment, not to the 10-s readings as it is not feasible to ensure that observations occured at the same exact point in different mobile transects. Further, based on the analysis of pre-coating micrometeorology, the following conditions are to be met when applying DofD but also evaluated on a case-by-case basis.

At street level:

• Clear-sky conditions throughout the day or mostly clear
• Solar radiation at solar noon > 600 W m⁻²
• Average 15 min wind speed < 4 m s⁻¹
• Wind direction SSW—SSE

Background/above canopy:

In addition to the primary, directly-observed variables, several derivatives were computed depending on scale, i.e., community or street level. At the community scale, derivatives included wet-bulb globe temperature, NWS Heat Index, UHI Index, as well as various cumulatives relative to different thresholds. At the street level, and in addition to these derivatives, several thermal indicators were computed at 10-s intervals along every mobile transect for all sorties that were performed. These include MRT, PET, PMV, and PPD, as discussed next. Of those, only MRT was reported in some other studies of cool pavements.

In prior studies, e.g., Middel et al (2020), Schneider et al (2023), and Engel et al (2023), MRT (or TMRT) was used or referenced in evaluating the effects of increased reflected shortwave radiation on thermal comfort. However, the implicit assumption that one degree of radiation is equivalent to one degree of convection applies only when the wind and pedestrian are at a complete standstill. A more realistic approach includes an accounting for radiative (h_r ) and convective (h_.c. ) heat-transfer coefficients (e.g., at the human body) when evaluating the effects of changes in MRT.

Per de Dear et al (1997), the whole-body convective heat-transfer coefficient (h_.c. , W m⁻² K⁻¹) is:

$$\begin{eqnarray}&&{h}_{c\,}=10.4{u}^{0.56}\end{eqnarray} \tag{ 4 }$$

where u is in m s⁻¹. On the other hand h_r is relatively constant at 4.5 W m⁻² K⁻¹ (for a whole-body standing configuration). Thus at 2–4 m s⁻¹ (usual range of wind speed in the Pacoima area), h_.c. /h_r would range from 3.4 to 5.0; meaning a decrease of 1 °C in air temperature won't be canceled out until MRT increases by 3.4 to 5 °C, respectively. Or, if a subject is walking slowly at, say, 1 m s⁻¹ with no wind, h_.c. /h_r would be ∼2.3 (meaning it will require 2.3 °C of MRT increase to cancel out the effect of 1 °C reduction in air temperature). Restated in yet another way, if MRT increases by, say, 3 °C and a pedestrian is moving or subject to a wind speed of 3 m s⁻¹, then the sensation perceived by a pedestrian would be as if air temperature were increased by 0.7 °C.

In addition, MRT may not be the best-suited indicator for outdoor street level thermal comfort assessments as it does not account for physiological, activity, or clothing effects. A better use of MRT would be as one of several input variables to computing the physiological equivalent temperature (PET), as discussed below. Furthermore, while some studies accounted for the shape factor of the human body in calculating MRT, they did not account for the view factor in radiative exchange between a person and the pavements especially when the subject is at a sidewalk, for example, which is more realistic than assuming a person standing in the middle of a street.

In this study, the calculation of MRT is split into diffuse and direct components, similar to its use in the RayMan model of Matzarakis et al (2010), to allow evaluations under different conditions e.g., indirect or direct sun. Thus, for each 10-s interval of every mobile transect:

$$\begin{eqnarray}&&{{MRT}}_{{indirect}}={\left\{\frac{1}{\sigma }\,\left(L+A\,\frac{{K}_{d}}{\varepsilon }\right){\rm{\varnothing }}\right\}}^{0.25}-273.15\end{eqnarray} \tag{ 5 }$$

which is a slight simplification of equation (6), since the sensors measure the integrated effects of various flux components.

$$\begin{eqnarray}&&{{MRT}}_{{indirect}}={\left\{\frac{A}{\sigma \varepsilon }\,\displaystyle \sum _{i=1}^{n}{F}_{i}\,{K}_{{di}}+\frac{1}{\sigma }\,\displaystyle \sum _{i=1}^{n}{F}_{i}\,{L}_{i}\right\}}^{0.25}\end{eqnarray} \tag{ 6 }$$

$$\begin{eqnarray}&&{{MRT}}_{{direct}}={\left\{{\left({{MRT}}_{{indirect}}\right)}^{4}+f\,A\,\frac{{K}_{s}}{\varepsilon \,\sigma }\right\}}^{0.25}-273.15\end{eqnarray} \tag{ 7 }$$

Here, L is upwelling longwave radiation, A shortwave absorptivity of the receiving body, F is the view factor to individual surface components, ε is emissivity of receiving body, ϕ is overall view factor to radiating surfaces, i.e., streets, buildings, vegetation, f is form factor (accounts for the angle between incident solar radiation and a standing or walking person, thus, a function of body posture and solar altitude angle, θ, at the time of observation), and Ks is incoming solar radiation (from the sky). Solar radiation that is reflected from various surfaces is already included in K_d (in the term MRT_indirect ) along with diffuse solar radiation from surfaces and the sky.

In this study, PET (Hoppe 1999) was quantified in two manners: (1) by directly applying the RayMan model (Matzarakis et al 2010) and (2) by using a statistical model of Koopmans et al (2020) that resulted in a 'PET-urban' variation to the indicator. The latter is based on an empirical regression model from the application of RayMan to urban areas. While some of the parameters might have been location-specific to where PET-urban was validated, the range of geolocations and weather variables input to the model can provide additional means to evaluate changes in PET.

Thus, PET-urban was applied in this study to take advantage of various observed quantities in the field and MRT is an implicit input in these calculations (on the other hand, MRT can be an explicit input to RayMan model). Calculated at 10-s intervals for every mobile transect performed in the project, PET-urban was computed per Koopmans et al (2020) as follows:

$$\begin{eqnarray}\begin{array}{rcl}{{PET}}_{{direct}} & = & -13.26+1.25{T}_{a}+0.011\,K\,-\,3.37\,\mathrm{ln}\left[U\right]\\ & & +\,0.078\,{T}_{w}+\,0.0055K\,\mathrm{ln}\left[U\right]+5.56\,\sin \theta \\ & & -0.0103K\,\mathrm{ln}[U]\sin \theta -0.0546\,\beta +1.94\,\omega \end{array}\end{eqnarray} \tag{ 8 }$$

$$\begin{eqnarray}&&\begin{array}{l}{{PET}}_{{indirect}}=-12.14+1.25{T}_{a}\\ -\,1.47\mathrm{ln}\left[U\right]+0.06{T}_{w}+0.015\,\omega \,{K}_{d}\\ +\,0.0060\left[1-\omega \right]\,\sigma \,{\left({T}_{a}+273.15\right)}^{4}\end{array}\end{eqnarray} \tag{ 9 }$$

where T_a is air temperature, K total solar incoming and reflected radiation (from streets and canyon surfaces), U wind speed, T_w wet-bulb temperature (computed from psychrometrics), θ solar altitude angle, β Bowen ratio, ω sky-view factor, K_d diffuse solar radiation including reflected from surfaces (e.g., streets), σ Stefan–Boltzmann constant, and RH relative humidity. 

In addition to calculating PET with the RayMan model, predicted mean vote (PMV) was also computed. Once PMV was calculated, PPD was quantified as:

$$\begin{eqnarray}&&{PPD}=100-95{e}^{-\left(0.03353{{PMV}}^{4}+0.2179{{PMV}}^{2}\right)}\end{eqnarray} \tag{ 10 }$$

A brief comparison of radiative properties of DuraShield-SR against those of existing asphalt pavements in the study area is presented here. Table 1 lists the averages of measurements from several points in the field carried out with the Solar-I reflectometer presented above.

Following application of reflective coatings, the UV albedo (reflectance) increased by a very small, negligible amount (+0.008) and in the case of parking lots, the UV albedo actually decreased (−0.014). These findings are desirable as discussed in the chemistry section below. Furthermore, the larger increases in albedo in DuraShield-SR occurred in the near-infrared (NIR) range, which is also desirable for glare avoidance (also discussed below). The averaged increase in NIR albedo relative to the existing asphalt pavement reached up to about 0.3. At specific locations, however, the increase in NIR albedo was much higher, up to 0.4 or more, as discussed later.

Broad-band albedo was computed from 7-bands reflectometer measurements in the field, using a spectral weighting function of 1.5 atmospheres. Figures 7 and 8 summarize some of the statistical attributes of albedo in the test and reference communities, respectively, showing that there was practically no difference, overall, in the albedo of streets between these two communities prior to coating. The medians for albedo were 0.0550 and 0.0535 and the means were 0.0544 and 0.0536, respectively. There were some small differences in the min/max ranges and the IQR. Skewness was different between the two communities which might have been one factor (in addition to being downwind) why the test community was slightly warmer than the reference community prior to coating.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Whereas broad-band albedo calculations were carried out at 10-s intervals and measured continuously along all transects with a net radiometer on the cart (figure 2), reflectance was also measured (with the portable reflectometer) at several locations in the project area, shown with red circles in figure 9. Examples from these measurements are presented in figure 10 for pre- and post-coating conditions at random locations. While the post-coating broad-band albedo of streets (computed from reflectance) at these locations reached up to about 0.26 on average, likely because of rough surfaces or dirt particles, the post-coating albedo on smoother surfaces, such as in some parts of the streets, school playground, or parking lots (not shown) reached up to 0.33, similar to the nominal albedo of DuraShield-SR

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Since the changes in albedo impart multi-scale effects that are diagnosed at surface, near surface, street canyon, canopy layer, boundary layer, atmosphere, and top of atmosphere, it is of interest to get an idea as to these changes based on the surface modifications in Pacoima. Table 2 lists the increase in albedo as observed from different heights—the values listed here are to be understood as representative but not necessarily applicable to every situation encountered during the study. Changes can be larger than summarized here.

As a radiometer's height above surface increases, the observed increase in albedo (from cool pavements) becomes smaller because areas of unchanged albedo, such as sidewalks, driveways, soil, vegetation, roofs, water, and grass, occupy an increasingly larger percentage of the radiometer's field of view. In addition, atmospheric effects (such as humidity or particulates loading) can come into play as well.

Figure 11 shows pre-coating albedo computed at 10-m resolution from Sentinel-2 satellite, indicating similar albedo in the streets of the project area (also see figures 7 and 8), except for the freeways that have higher albedo. Figure 12 compares the pre- and post-coating albedo computed at 30-m resolution from Landsat 9 satellite where it can be seen that the modified streets, school playground, parking lots, and basketball courts are more reflective as detected from top-of-atmosphere (TOA) following application of cool pavements.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

In figure 12 (top), the school playground and the parking lots/courts have a lower albedo that's similar to most of their surrounds seen in this frame. The test and reference communities are similar in street albedo (no large differences are detected when comparing either sides of Borden Street). After installing the reflective pavements, it is noticed in figure 12 (bottom) that the playground now has a large area of higher albedo and so do the parking lots and basketball courts. The test community (east of Borden) also has a generally higher albedo (following coating) than the reference community and that can be seen clearly demarcated in the figure. Furthermore, streets like Daventry and Sylmar (north-west of the school playground) appear now more clearly with higher albedo values than in pre-coating conditions. In both figures, the darker shade is lower albedo and whiter is higher albedo and each contour level is a stepwise change of 0.04.

A final note related to changes in albedo: the project had a plan in place to track the degradation of reflectance (as measured with reflectometer) over time and also carry out power washing of streets to evaluate the restoration to higher values, but was scrapped because of the heavy rainfalls in California in winter 2022–2023. Continuous readings from net radiometer on mobile platform indicated little degradation in albedo.

The field measurements demonstrated that another aspect of concern with cool pavements, potential glare, is a non-issue in this case. This assessment was based on comparing the spectral reflectance signatures of the reflective coating to those of the existing streets and parking-lot surfaces.

Figure 13 is a comparison between the reflectances in the visible spectrum (VIS) of the existing parking-lot material and the reflective coating that was applied, as measured in the field with the reflectometer. It can be seen that the VIS reflectance of existing parking-lot material (p1, p3, p4) was 0.08–0.10 (except for one instance). The reflectances of the coating were modestly higher in VIS, generally between 0.12 and 0.15, which is not large enough to cause visual/glare discomfort issues (also see figure 1, bottom).

Zoom In Zoom Out Reset image size

However, upon examination of changes in the NIR ranges, as seen in figure 14, it is clear that there was a much larger increase in reflectance of the coating relative to existing material. In other words, the increase in albedo from installing DuraShield-SR resulted mostly from increased reflectance in the NIR, not in the VIS range. As such, the visual/glare environment is not affected negatively but the benefits in terms of cooling are significant. In this case, the NIR reflectance of existing parking-lot material was about 0.1 to 0.15 whereas the NIR reflectance of the reflective coating was between 0.4 and 0.5, a very significant increase.

Zoom In Zoom Out Reset image size

For cool-community measures that include reflective materials, such as cool pavements, the albedo of interest is usually in the range of 0.28 to 2.8 μm, i.e., solar albedo. This includes tropospheric radiation in the UV, VIS, and NIR spectra. In addition to the main effect of changing VIS and NIR albedo, there is interest in the potential effects of possible changes in UV albedo as a result of implementing cool pavements (Taha 2008b) and some studies, e.g., Epstein et al (2017) have raised potential air-quality concerns in this regard. This is so because the energies in the UV wavelength range drive most of the important photodissociation reactions, e.g., those of NO₂, O₃, and PAN, that can have potential implications on ozone formation, i.e., possible negative impacts. However, in reality, the proposed changes in urban surface albedo may have small or no effect at all on UV and, in some cases, the UV albedo actually decreases, as discussed above in relation to measurements with the reflectometer.

In the UV spectrum (UV-A: 0.315–0.400 μm, UV-B: 0.280–0.315 μm, and UV-C: 0.100–0.280 μm), stratospheric oxygen absorbs radiation in the range 0.17–0.24 μm and photodissociates to produce ozone via: O₂ + hν → 2 O (³P) and O (³P) + O₂ + M → O₃ + M

where M = air = N₂, O₂. This leaves mainly UV-B and UV-A radiation to reach the troposphere because the stratospheric ozone produced in the above process absorbs UV at and below 0.29 μm. Thus in the troposphere, wavelengths of relevance to photochemical reactions are those longer than 0.30 μm (Seinfeld and Pandis 1998).

NO₂ absorbs at wavelengths of 0.45 μm or shorter, but because there is little UV radiation reaching the troposphere at or shorter than 0.29 μm, the theoretical critical UV range of interest for NO₂ is thus between 0.30 and 0.45 μm (Cooper and Alley 1994). Further narrowing this range is the fact that 90% of the NO₂ molecules absorb UV energy below 0.4 μm (Stern et al 1984) and as a result, the practical range of importance for NO₂ photodissociation is 0.3 to 0.37 μm (Seinfeld 1975), thus mostly within the UV-A range. For O₃, the critical UV wavelength range is 0.315 μm or shorter (Harrison 1990) and for PAN, the cutoff is 0.35 μm or shorter (Seinfeld and Pandis 1998). Thus the inclusive range of 0.3 to 0.37 μm is the overall 'envelope' that needs to be considered when modifying surface albedo. This range corresponds to the first reflectance band of the reflectometer shown in table 1.

Since UV albedo in this range does not change significantly (or even decreases) with the application of DuraShield-SR coatings (table 1), the UV albedo-chemistry issue is not a concern in this case. As Berdahl et al (2002) and Taha (2005) showed, there are other coatings and materials that increase VIS and NIR albedo significantly without affecting UV albedo.

A range of effects on air temperature was observed when directly comparing the test and reference segments of streets (RT) -- predominantly lower but also slightly higher at times (due to transport and mixing). Surface temperature, on the other hand, was always lower. However, the effective temperature change, i.e., taking into account pre-coating conditions (DofD), was always a reduction in both air and surface temperatures.

While this section presents averages over street segments for typical summer conditions, some of the observed maximum air-temperature reductions occurred during a heat wave event (table 3). These were RT differences, not DofD, since there was no pre-coating heat-wave event to compare to. The maximum differences during that event reached up to −1.9 °C.

Several mobile transects were also carried out during winter months (November/December 2022), to characterize the effects of cool pavements in overcast, wintertime conditions. It was found that the differences were small or negligible in surface and air temperatures between the test and reference communities. This is a desirable outcome as it suggests that the effects of street albedo increase are minimal or non-existent during wintertime which can help avert any potential negative effects on heating-energy needs.

Figure 15(a)–(e) is a random transect, in this case showing a loop from 1040 to 1120 PDT on 10/07/2022 as an example from more typical summer conditions. Figure 15(a) depicts air temperature measured at three levels (0.6 m, 1.3 m, and 2 m AGL) starting at Montford Street in the northeast and ending at Mercer Street in the southern part of the study area (see figures 2 and 9). The general tendency in air temperature, in this case, was a small increase from north to south by ∼1 °C while relative humidity decreased by a small amount, from 56% to 52% (figure 15(b)). On a street-by street basis, however, the temperature tendency was smaller than 0.05 °C and thus no detrending was necessary on this basis. Furthermore, the alternate looping from E→W to W→E was designed to cancel out such tendencies even if small.

Zoom In Zoom Out Reset image size

Figure 15(c) depicts the variations in side-view incoming solar radiation as the sensor alternated between facing the northern and southern hemispheres during each loop. This information helped identify direct and diffuse radiation for the calculation of thermal indicators such as MRT (equations (5)–(7)). For example, in this particular transect, solar radiation below 200 W m⁻² was deemed diffuse (when sensor was facing north) and above that was direct+diffuse (sensor facing south). Furthermore, this type of time series also helped identify outliers, e.g., shaded street spots (such as at the right end of the series) that were then removed from the analysis when developing correlations. In this study, outliers were identified as the more stringent of either V > (Q3 + 1.5 IQR) or V < (Q1–1.5 IQR); or $\left|Z\right|\gt 3.0,$ where Z is the Z-score and IQR is the interquartile range for a variable V. In some subsequent figures, these are labeled 'IQR 1.5' or 'Z3'.

In figure 15(d), the incoming (down) and reflected (up) solar radiation is shown along with downwelled and upwelled longwave radiation (from an Apogee net radiometer on the golf cart) and figure 15(e) is the corresponding surface temperature. The horizontal gaps in the time series in these figures correspond to transitioning from E to W or W to E segments of each street across the divider (Borden Street) between the test and reference areas.

These readings, e.g., for air and surface temperatures, were then averaged over each E and W segment of each street and transect. Table 4 is a random example from the reporting of such averaging results. Lastly, table 5 lists the net radiation and differences at the E and W segments of each street corresponding to the time series of figure 15.

Figure 16 is an overall summary of temperature differences (°C) at street segments (based on a large number of tabulations similar to the example in table 4). These exclude the winter month of December 2022 and heat wave period, as discussed earlier. Table A-3 in the Supplement provides descriptive statistics with respect to this figure. As noted, these are based on street-segment averages, not 10-s data readings that could show much larger but also highly variable effects. This is done to facilitate interpretation of results at street or community scales rather than at particular spot-measurement locations.

Zoom In Zoom Out Reset image size

The observed meteorological conditions, including air temperature, wind speed, wind direction, and solar radiation, corresponding to the differences and changes in air and surface temperatures reported in figure 16 are provided in figure A-1 of the Supplement.

A clear pattern is seen in figure 16 -- that there was a larger decrease in temperatures during daytime and smaller, but still significant, reductions at night. From midnight to pre-sunrise and immediately after sunrise, mean of air-temperature differences ranged from −0.12 to −0.14 °C RT and for surface temperature from −0.72 to −0.79 °C RT. From morning to evening, the mean of differences in air temperature ranged from −0.13 to −0.26 °C RT and −0.60 to −1.21 °C DofD. For surface temperature, the mean of differences ranged from −2.5 to −2.97 °C RT and −4.7 to −5.73 °C DofD. Figure 16 also shows quartiles and maxima indicating that there were large reductions in temperature both RT and DofD, especially in the afternoon (note that the small dots are outliers). There were also increases but generally smaller.

In table 6, examples of the maximum temperature reductions (similar to table 3 above but for non-heat-wave conditions) are provided based on street-segment averages. Note that, statistically, a small number of these could be considered as outliers in some cases (as shown in figure 16). However, these were actual and valid measurements in the field, not errors. Furthermore, a test was carried out to evaluate the statistical significance of changes in surface and air temperatures as a function albedo. The p-values obtained were consistently under 0.0001 indicating high significance of the reported cooling results.

Some other attributes of changes in the air-temperature field that were observed included the local rates of heating or cooling in the test and reference areas during cloud-free days of the monitored periods. The observed air-temperature tendencies on average were:

• Pre solar noon: the test community warmed up by 0.07 °C–0.20 °C h⁻¹ slower than the reference community (i.e., delayed warming)
• Around solar noon: the test community warmed up by between 0.15 °C h⁻¹ faster and 0.31 °C h⁻¹ slower than reference community (that is, a mixed signal but mainly a delayed warming effect)
• Post solar noon: the test community cooled down by 0.50 °C h⁻¹ faster than the reference community (enhanced cooling effect).

It is also of interest to view the cooling achieved by reflective pavements in the context of typical magnitudes of urban heat or location-specific heat islands. Taha (2017) developed the Urban Heat Island Index (UHII) for the State of California and showed that in Pacoima (figure 17), the summer UHII is between 22 and 27 °C·h day⁻¹ (no thresholds). This is equivalent to an average urban heat island (°C·h h⁻¹) of between 0.90 and 1.13 °C (all day/night hours inclusive) and depending on wind direction. Thus, the cooling achieved by the reflective pavement (figure 16) is significant and can offset 25%-50% of the local UHII, depending on time of day.

Zoom In Zoom Out Reset image size

Metrics that were used to evaluate the effects of cool pavements on thermal comfort at street level included MRT, PET-urban, PET-RayMan, PMV, and PPD as introduced above. With respect to MRT, field measurements in this study indicated that relative to existing/typical asphalt, areas with DuraShield-SR had increased reflected solar radiation but also significant decreases in longwave upwelling, which contributed to lowering radiant temperatures. The real-time (RT) observations in Pacoima indicated both increases and decreases in MRT but that the median difference was −0.91 °C and the mean −0.97 °C, community-wide, calculated across a range of weather conditions, times of day, and sun angles.

Figures 18 and 19 depict results from calculations of PET, PMV, and PPD and table A-4 (in Supplement) is a summary of related statistics. As seen in figure 18, the mean of the differences in PET-urban (differences between test and reference communities), where MRT is implicitly accounted for, reached up to −0.36 °C. With the RayMan model, MRT was specified explicitly as one of the input parameters (based on mobile-transect observations) and so had a larger impact on the calculations of PET. The mean of differences in PET-Rayman reached up to −0.66 °C, as indicated by 'MRT explicit' in the figure. Of note, the reductions in PET, percentagewise, were slightly smaller (implying less cooling effects) when in direct sun versus indirect. These changes pertained to real-time (RT), side-by-side differences between the test and reference communities (indicated by 'Real time' in figure 18). When DofD was applied (along with the explicit MRT input), the mean of changes in PET-RayMan reached up to −1.77 °C as shown at the right end of the figure.

Zoom In Zoom Out Reset image size

Zoom In Zoom Out Reset image size

Calculations with the RayMan model showed that reflective coatings reduced PMV by an average of 0.09 RT and an average of 0.32 DofD (figure 19). Since PMV ranges from −3 to +3, these average reductions correspond to 2% and 5%, respectively. Finally, the mean of changes in predicted percent dissatisfied (PPD) was −2.86 percentage points RT and −7.87 percentage points DofD (the range of PPD is from 5% to 100%).

As seen in figures 18 and 19, there are both increases and decreases in the indicators but that the dominant effect is a reduction (benefit). This suggests, overall, no net penalties from cool pavements in terms of thermal comfort at pedestrian level.