Introduction

Impervious surfaces (or macroscopic surfaces in the following of this manuscript), such as paved roads and parking lots, rooftops, building surfaces1,2, cover on average 60% of the total urban area, eventually increasing to 80–100% in central business districts3,4. Interestingly, the high coverage of macroscopic surfaces renders urban environments particularly vulnerable to climate change due to the urban heat island (UHI) effect1. Although macroscopic urban surfaces are highly abundant, their role as sites for heterogeneous reactions with atmospheric gases relevant to air quality has been largely overlooked. While studies have documented the effective uptake and transformation of atmospheric pollutants on simulated or real-world urban grime (i.e., deposits on macroscopic surfaces)5,6,7,8,9, the reactivity of a diverse range of pristine impervious surfaces remains largely unexplored, with only a limited number of studies available in the literature10. Overall, macroscopic urban surfaces may not only serve as a source of atmospheric pollutants but also as a sink, providing extensive surfaces for pollutant scavenging and transformation. This “sponge effect” could play a significant role in the diurnal distribution of atmospheric species and their efficiency to produce secondary pollutants such as ozone (O3) and new particle formation.

Among the various types of macroscopic urban surfaces, asphalt pavements are of particular interest due to their high abundance corresponding to around 40% of the surface of an urban city11, a fraction that keeps increasing with urbanization11,12. Asphalt is composed of a large number or organic compounds reflecting its source from petroleum. They are capable of emitting a wide variety of volatile or semi-volatile organic compounds (VOCs, or s-VOCs) and generally non-methane VOCs (NMVOCs), not only during the application phase, i.e., short-term emissions under high temperatures > 60 °C13,14,15,16, but also on the long-term, as a constant source of NMVOCs under typical ambient temperature conditions16,17.

However, the uptake or transformation of atmospheric pollutants on asphalt pavements is almost undocumented, although it may be an important atmospheric process given the large surface area that asphalt pavements represent. In air quality models, the heterogeneous loss of atmospheric species on asphalt pavements or urban macroscopic surfaces is not always considered so far18,19,20,21,22,23. Even in the case of the recently developed high-resolution city scale models19,20, the heterogeneous loss of urban air pollutants on asphalt pavements or impervious surface areas is not included. Neglecting the contribution of heterogeneous processes in urban environments (areas with high surface to volume ratio), may limit our understanding of the fate of key atmospheric pollutants, such as NOx, O3, and VOCs. This may also be one of the many reasons why atmospheric models fail to reproduce the concentrations of key atmospheric pollutants such as NOx in urban environments18,19,24.

The objective of this article is to investigate the uptake of key atmospheric pollutants, such as nitrogen dioxide (NO2), ozone (O3), and two model organic molecules, methanol (CH3OH) and formaldehyde (HCHO), common urban pollutants, on real world asphalt pavements, i.e., surfaces exposed to weathering and traffic (aged), as well as on non-weathered asphalt substrates. A series of laboratory experiments were performed to determine the kinetic uptake parameters that reflect the steady state heterogeneous loss of atmospheric pollutants on the asphalt pavements. The environmental implications of our laboratory results are discussed through a series of box model simulations in which NO2, O3, and HCHO loss on macroscopic surfaces are incorporated. To the best of our knowledge, this is among the first experimental studies providing quantitative results on the uptake of key urban pollutants comparing the uptake efficiency between fresh and real-world aged asphalt pavements.

Results

Uptake coefficients determined for the different gases studied are summarized in Table S2. In particular, the table displays the masses and the geometrical surface areas of the irregularly shaped samples, the first order uptake rate coefficients, and the corresponding steady state uptake coefficients. The uptake coefficient values presented in this study were determined using both the geometric \({\gamma }_{\mathrm{ss}}\left(\mathrm{geom}\right)\) and the BET surface area of the samples \({\gamma }_{\mathrm{ss}}\left(\mathrm{BET}\right)\). It should be noted that the expression of \({\gamma }_{\mathrm{ss}}\) using the geometric or the BET has no impact on the model simulations described below (atmospheric implication section), since the surface area used to express the uptake coefficients is canceled out in simulation. Overall, using the geometric uptake coefficient is a more straightforward approach in modeling heterogeneous processes particularly for impervious or macroscopic urban surfaces. Furthermore, the number of molecules irreversibly scavenged normalized by the specific surface area are also presented in Table S2. Ns highlights a longer-lasting heterogeneous interaction between the gas and the surface. In the following, the results are discussed separately for each probe gas.

Uptake of NO2 on asphalt pavement samples

A typical uptake experiment of NO2 on an aged asphalt surface (Acity–5–35/50) is presented in Fig. S3, corresponding to the theoretical profile shown in the lower panel of Fig. S2. The initial fast consumption of NO2 due to the very first exposure leads then to constant values of \({k}_{\mathrm{upt}(\mathrm{ss})}\) and thus \({\gamma }_{\mathrm{ss}}\) after a long surface exposure to the gas flow. Spontaneous desorption experiments of adsorbed NO2 performed after an extended period of heterogeneous interaction do not lead to any observable desorption of NO2 at ambient temperature.

Our experiments reveal a significant uptake of NO2 on all weathered road pavements, while within experimental accuracy, no uptake of NO2 was observed on non-weathered asphalt pavement. Uptake coefficient measurements performed in different sample compartments showed no systematic variation of the determined kinetic parameters (see experiments 8 and 11 in Table S2). To evaluate whether edges of the asphalt samples may impact the uptake measurements, a series of experiments were performed with hydrocarbon wax coated samples, where the whole sample except the weathered front surface was dip-coated in hydrocarbon wax in order to eliminate interaction with the probe gas molecules. The absolute (time-independent) uptake of NO2 and the uptake coefficients measured on the different road pavement samples are indistinguishable within experimental uncertainty (compare measurements 3 with 5–7, 8 with 10, and 11 with 12). The latter could be expected because the aged asphalt edges, are non-weathered, and thus no uptake is anticipated based on our measurements. As shown in experiments 4, 8, 11, and 12, we observed that after long exposure, the surface is slowly deactivated, which may indicate the depletion of surface sites, or a rate limiting chemical reactions or diffusion on/in the substrate. However, when the surface is left for several hours isolated and then re-exposed to NO2 it restores its activity (comparison of experiments 3 and 4). Therefore, it could be expected that the reactivity of the asphalt pavements may be restored overtime based on the traffic conditions (low/dense traffic alternation that could be related with the rush hours or day/night traffic). Therefore, these surfaces are proposed as a constant sink for NO2 and will be treated as it in the environmental implications section.

Considering all measurements performed on aged asphalt pavements, the average uptake coefficients of NO2 was found to be:

\({\gamma }_{\mathrm{ss}}\left(\mathrm{geom}\right),\,{\mathrm{NO}}_{2}=(3.5\pm 2.3)\times {10}^{-5}\) while the BET normalized uptake coefficient was

$${\gamma }_{\mathrm{ss}}\left(\mathrm{BET}\right),\,{\mathrm{NO}}_{2}=(2.0\pm 2.0)\times {10}^{-8}$$

where the uncertainty expresses the range of the measurements and includes the variability of the samples (Fig. 1). The uptake coefficients determined in this study are in good agreement with those reported by Trick et al.10, who measured NO2 uptake on asphalt pavements in a smog chamber at room temperature and varying RH levels. Their reported values ranged from 0.9 to 1.3 × 10−5. Despite differences in experimental conditions and sample variability, the results between the two studies are in good agreement.

Fig. 1: Geometric uptake coefficients of key atmospheric species on non-weathered, and aged asphalt pavements.
figure 1

Error bars denote the range of the measurements and include the variabilities due to the origin of the samples, the aging process, and other possible treatments of the samples (see Table S2).

Full size image

Ns for NO2 was found to be in the range of 1011–1012 molecules cm−2, and similarly to the uptake coefficient values, no systematic variation was noticed when using the standard or large SC, or by applying a hydrocarbon wax coating on sample edges. The average value of Ns was calculated using the initial coverage measurements (bold numbers in column 5 of Table S2) and was found to be 1.78 × 1012 molecules cm−2, corresponding to less than 1% of the theoretical monolayer coverage of NO2 (see section calculation of theoretical monolayer coverage in text S2 of the SI). The low estimated surface coverage also suggests that the NO₂ uptake kinetics discussed above follow first-order behavior, consistent with a typical Langmuir–Hinshelwood mechanism, despite the relatively high concentrations used (on the order of 1013 molecules cm−3).

No gas phase products were observed during surface exposition to NO2, or at least their formation is below the detection limits of the instrument. In particular, we focus on the formation of nitric oxide (NO) and nitrous acid (HONO); however, their formation was not observed, and an upper limit of 1% (defined as ΔNO2 consumed versus Δproducts formed) may be given based on the MS sensitivity. Moreover, desorption of NO2 from the asphalt surfaces was not observed, which indicates either a chemical transformation of NO2 leading to surface nonvolatile products (e.g., surface nitrates) or its strong uptake on the surface of asphalt pavement. However, it should be noted that the desorption of NO2 from asphalt surfaces could be a slow process25.

The uptake behavior of NO2 on non-weathered and aged asphalt pavements agrees with the study of Lasne et al.25. These authors studied the emission of NOx from real world asphalt pavements in an atmospheric simulation chamber. Non-weathered samples that had been stored for several years in a clean environment did not emit NOx, even after being exposed to ambient air for several days. On the contrary, Lasne et al., observed that aged asphalt pavements were found to emit NOx once exposed to clean air in the chamber. However, this was a several hours long-lasting process eventually leading to emission exhaustion. Interestingly, when the samples were exposed again to ambient air or to car exhaust, they again took up NOx, and promoting again NOx emission in a clean air environment. The latter observation was also confirmed in the present measurements, as discussed above (experiments 3 and 4). Therefore, the absence of observed desorption in our experiments is likely due to two main factors, (i) the low yield of non-reactive NO₂ uptake, which falls below the detection limit of our instruments, and (ii) the low desorption rate coefficients, making the desorption process negligible under our experimental conditions.

The reaction mechanism of NO2 on asphalt pavements has been poorly investigated in the literature. Trick was among the first to observe the formation of gaseous HONO on asphalt pavements under humid conditions10; however, no specific insights into the underlying mechanism were provided. According to our results, the weathered and traffic aged asphalt surfaces are more reactive compared to non-weathered samples. The latter may possibly indicate that NO2 only reacts with the oxidized asphalt interface, leading to adsorbed phase products. NO2 may also react with polycyclic aromatic hydrocarbons PAHs existing on the asphalt pavement interface26,27. These coefficients are comparable to the rate coefficients measured in the present study (Table S2). However, it should be noted that the rate coefficient is highly dependent on the substrate matrix26,27,28. For example, significantly lower values have been reported for PAHs adsorbed onto silica particles, differing by several orders of magnitude26,27,28. Naphthalene, and its methyl derivatives, and acenaphthene have been identified as PAHs in these particular asphalt samples17, however, the corresponding rate coefficients have yet to be measured. Although PAHs are expected to be present in both aged and non-weathered substrates, the higher NO2 reactivity observed for aged samples indicates that PAHs are not the main contributors to the uptake process. Asphalt exposed to atmospheric photooxidation can form oxygenated aromatics with redox properties such as semiquinones and hydroquinone29. NO2 can be effectively reduced on such compounds with a one-electron transfer mechanism to form surface nitrite8. Another plausible pathway of interaction/reaction of NO2 may be with the catalysts naturally existing in bitumen30, or deposited from traffic31 (car brakes, exhaust pipelines) etc. In addition, the catalytic reaction of NO2 would lead to gas phase product such as NO, HONO, which were not detected32. Nevertheless, it should be mentioned, that HONO, if formed, can rapidly react with asphalt interface to form gaseous NO2 or nitric acid33,34.

Uptake of O3 on asphalt pavement samples

The uptake of O3 has been observed on both non-weathered and aged asphalt pavements, with values on non-weathered samples being markedly smaller than on aged surfaces (Table S2). The differences range from a factor 5 to 10. (Fig. 1). Figure S4 shows the typical O3 uptake experiment totaling three SC open/close cycles. It should be noted that the reduction of O3 concentration is only due to its heterogeneous loss, and any competing gas-phase loss processes, such as reaction of O3 with potentially emitted VOCs is negligible. The following observations support the latter:

  1. a.

    The gas-phase concentration of organics or aromatics in the present reactor was always below the detection limits of the instrument, which was in the 107–1010 molecules cm−3 range. Only levels in the 1013 molecules cm−3 range could explain the observed 10−3 s−1 first-order rate coefficient35 (Table S2);

  2. b.

    alkenes, which are expected to be the most reactive components towards O3, constitute the lowest fraction of asphalt pavement emissions, as noted by Lasne et al.17.

Figure S4 also shows alternating masses of 48 and 32 amu during the open and closed periods of the SC. The former monitors exclusively O3 whereas the latter MS signal is a superposition of both O3 and O2 contributions. The ratio of the O3 to the product O2 flow after correction of mass 32 for the O3 contribution during the SC “open” period changes from 0.62 to 0.88 to 1.14 from the end of the first “open” period to the end of the third “open” period. A ratio of 1.5 between the O2 vs. O3 flow at steady-state conditions would indicate complete catalytic conversion such as 2O3 leads to 3O2. We take the increasing trend of this ratio with exposure time as clear evidence of a chemical reaction of O3 with some components of the organic substrate of asphalt.

As expected, the desorption of O3 was not observed in any experiments since O3 either chemically reacts or decomposes on solid substrates. Averaged the measurements obtained with the non-weathered and aged asphalt pavements the uptake coefficients values were:

\({\gamma }_{\mathrm{ss}}\left(\mathrm{geom}\right),\,{{\rm{O}}}_{3}=(2.4\pm 1.0)\times {10}^{-5}\) and \({\gamma }_{\mathrm{ss}}\left(\mathrm{BET}\right),\,{{\rm{O}}}_{3}=(1.5\pm 0.3)\times {10}^{-8}\) for non-weathered asphalt surfaces

\({\gamma }_{\mathrm{ss}}\left(\mathrm{geom}\right),\,{{\rm{O}}}_{3}=(20\pm 7)\times {10}^{-5}\) \({\gamma }_{\mathrm{ss}}\left(\mathrm{BET}\right),\,{{\rm{O}}}_{3}=(1.4\pm 0.4)\times {10}^{-7}\) for aged asphalt surfaces

where the uncertainty expresses the range of the measurements, and includes the variability of the samples (Fig. 1). Although no measurements were performed with the hydrocarbon wax on aged asphalt pavements, the low reactivity of ozone with the non-weathered surfaces suggests a minor contribution of the non-weathered edges to the total uptake coefficients. The uptake coefficient values were found to be elevated and larger by an order of magnitude compared to those of NO2.

The number of O3 molecules scavenged until reaching steady state conditions is in the range of 1012 molecules cm−2 and by a factor of 10 higher for non-weathered and aged weathered bituminous surfaces respectively (Table S2). In this case the average time-dependent uptake leading to initial O3 coverage is similar to NO2 for the non-weathered pavement and approximately ten times larger for the weathered surface. As with NO2, the low surface coverage observed for O3 is an indirect indication that its uptake follows first-order kinetics under the conditions used, despite the high concentrations.

It is well established in the literature that O3 reacts with unsaturated organic species, such as alkenes36,37, or some selected PAHs when adsorbed to a matrix38, both present in asphalt pavements. However, ozone reacts significantly faster with the aged oxidized samples compared to the non-weathered ones. The latter could be due to differences in the chemical composition, and the slow oxidation of organic species existing in the bitumen, or the deposition, of other unsaturated organics (such as phenols)39 or PAHs from traffic and non-traffic sources. Reaction of ozone with the catalysts naturally existing in bitumen (i.e., copper, nickel, vanadium, zinc, and cadmium)40, or catalysts deposited from other sources (traffic or non-traffic) could be an explanation about the ozone reactivity. The abundance, and availability (bulk or bitumen interface) could explain the higher reactivity observed for aged samples. Overall, the cause of the enhanced reactivity on aged surfaces is not fully understood yet. Several explanations can be proposed and further investigation is required.

Uptake of methanol and formaldehyde on asphalt pavement samples

This section briefly discusses the results on other atmospherically relevant probe gases based on results presented in Table S2. Methanol did not adsorb on non-weathered asphalt surfaces whereas a significant uptake was measured on aged asphalt pavement to be equal to 2.7 × 1014 molecules cm−2 corresponding to 44.7% of a monolayer. Noteworthy is that 10.3% of the methanol taken-up desorbs spontaneously right after adsorption. However, existing data do not unambiguously support a quasi-steady-state loss rate.

There is significant interaction between formaldehyde (HCHO) and weathered/aged asphalt pavement with typical coverages between 12 and 28.8% coupled to variable desorption yields varying from zero to 24%. Figure S5 displays a typical uptake experiment that appears to be similar to NO2 uptake regarding the average steady-state uptake coefficient but leads to a larger time-dependent coverage of approximately 25% of a monolayer. The average uptake coefficient of formaldehyde on aged asphalt pavements was found to be:

\({\gamma }_{\mathrm{ss}}\left(\mathrm{geom}\right),\mathrm{HCHO}=(3.8\pm 1.7)\times {10}^{-5}\) and \({\gamma }_{\mathrm{ss}}\left(\mathrm{BET}\right),\mathrm{HCHO}=(2.7\pm 1.2)\times {10}^{-8}\) for aged asphalt surfaces

where the uncertainty expresses the range of the measurements and includes the variability of the samples.

Overall, regarding the uptake of the VOCs studied, no measurable uptake was observed on fresh asphalt, while aged samples showed detectable uptake of methanol (although without reaching a steady-state uptake coefficient) and formaldehyde, for which a steady-state consumption was observed. This behavior may be attributed to surface aging and the formation of oxygenated functional groups—not only on the organic matrix of asphalt pavements, but also on the mineral-based filler materials. The establishment of a steady-state uptake for formaldehyde implies the occurrence of reactive surface processes rather than simple physisorption, consistent with the presence of chemically active oxygenated sites formed upon aging. These functional groups could facilitate the uptake of polar VOCs via hydrogen bonding or promote the formation of surface-bound products (e.g., formates, which are commonly observed surface species), although no gas-phase products were detected. Attempts were made to investigate the underlying mechanisms of VOC, NO2, and O3 uptake using surface analysis techniques available in our laboratory, such as diffuse reflectance infrared spectroscopy. However, these efforts were unsuccessful due to the high optical absorption (low albedo) of asphalt, which significantly limited surface reflectance and the sensitivity of the measurement.

Model simulations

Here, we evaluate the atmospheric impact of the uptake of the selected trace gases by asphalt. Among others, photolysis of O3 and HCHO are major primary radical sources during the daytime, while NO2 chemistry fuels the ozone production41,42. These compounds play a significant role in controlling atmospheric oxidative capacity; a detailed understanding of their fate is crucial in predicting air quality. As asphalt surfaces cover up to 40% of urban environment surfaces11, they may represent an effective reactive sink for gaseous compounds. To gain more information on the potential impact of asphalt reactivity, we performed a series of zero-dimensional numerical simulations in an idealized urban environment in which heterogeneous loss on macroscopic surfaces is added.

The simulations were made using the Regional Atmospheric Chemical Mechanism RACM43 in the box modeling extensions to the Kinetic Pre Processor (KPP)44, BOXMOX environment45. BOXMOX allows the control of background concentrations, environmental conditions, emissions, depositions, homogeneous reactivity, and photolysis. The chosen conditions are representative of a typical urban environment during summertime in the northern hemisphere. More details about the model conditions are presented in SI (Text S3) and particularly Tables S4 to S9. The model runs cover 7 days from the initial conditions, and species reached equilibrium after 3 days, as indicated by the negligible differences in mixing ratio between days 3 and 4 (Figs. S7 to S12).

Heterogeneous loss is treated as a first-order deposition process:

$$\frac{\mathrm{dX}}{\mathrm{dt}}=-k\times [X]$$
(1)

Where [X] is the concentration of species X (molecule cm−3) k is the first order (s−1) rate coefficient expressed as

$$k=\frac{{\nu }_{d}}{\mathrm{PBLH}}$$
(2)

In (2), \({\nu }_{d}\) is the deposition velocity (cm s−1), PBLH is the planetary boundary layer height (cm). For urban conditions, the aerodynamic resistance Ra and the quasi-laminar layer resistance Rb are minor compared to Rc (see SI for further detail). Therefore, for simplicity, they are not included in our model, and thus \({\nu }_{d}\) depends only on surface resistance, Rc, which itself depends on the uptake coefficient γ according to Eq. (3)46:

$$\frac{1}{{\nu }_{d}}={R}_{c}=\frac{4}{\gamma \times c}$$
(3)

Where c is the mean thermal velocity (cm s−1). According to Eq. (3), for aged asphalt pavement (Table S2), at 298 K, \({\nu }_{d}\) is 0.33, 1.79 and 0.44 cm s−1 for O3, NO2, and HCHO, respectively. The use of geometric uptake coefficients in this calculation is preferred as it allows the comparison of \({\nu }_{d}\) in dimension (geometric surface area) found in the literature.

Table 1 presents the five scenarios designed to assess the relative importance of asphalt reactive uptake on the O3, NO2, HCHO, NO, OH radical, and NO3 radical levels. Emission, photolysis, temperature, background, and initial concentrations are kept constant between each scenario, and only the deposition velocities were varied.

Table 1 A: uptake coefficients and associated deposition velocities (in brackets) of O3, NO2, and HCHO used in the various model scenarios. B: literature data
Full size table

In the base case scenario, the surface is considered 100% urban grime, and the O3 and NO2 surface reactivity are parameterized based only on the uptake measurements on urban grime from Mothes et al. (i.e., uptake coefficients = 10−6)47. Because no data is available on the HCHO surface reactivity on urban grime nor IUPAC recommendation on dust as grime proxy, HCHO surface reactivity is not considered in the base case (Fig. S9). In scenarios 2 to 4, the surface reactivity of NO2, O3, and HCHO has been changed individually to account for asphalt reactivity, and in scenario 5, NO2, O3, and HCHO uptakes are changed altogether. When asphalt reactivity is considered, we assume that asphalt surface accounts for 30% of the total area (seen as a 2D geometric area), and the corresponding uptake from this study is used. The rest of the surface is considered urban grime. This led to the parametrization as indicated in Table 1. As an example, in scenario 2, the NO2 deposition velocity is parametrized as: 0.3 × 3 × 10−5 + 0.7 × 1 × 10−6 = 9.7 × 10−6, which reflects a deposition velocity of 0.09 cm s−1 according to Eq. (3).

In the model, loss via the heterogeneous reaction on aerosol is not considered because it is thought to be a minor effect. Assuming a high aerosol loading of 2 × 104 particles cm−3 with a radius of 100 nm48 and an ozone uptake coefficient of 5 × 10−6 as recommended by the IUPAC evaluation for dust (at O3 mixing ratio = 50 ppb)49, the first order rate coefficient kaerosol (in s−1) calculated according to Eq. (4) is orders of magnitude lower than what is calculated for dry deposition (Eq. (2)). This represents a few percent loss term at maximum and is therefore neglected here.

$${k}_{\mathrm{aerosol}}=\left(\frac{\gamma \times c}{4}\right)\times \frac{S}{V}$$
(4)

where S/V is the aerosol surface to volume ratio (cm−1).

In their model, Ervens and coworkers have used deposition velocities in urban environments at 0.4, 0.4, and 1 cm s−1 for O3, NO2, and HCHO, respectively21. Their O3 and NO2 values compared well with what we calculated for aged asphalt pavement (assuming a 30% asphalt fraction), but their HCHO is one order of magnitude higher than ours (Table 1, scenario 5). In this study, \({\nu }_{d}\) is only based on the surface resistance (Eq. 3), while they parametrized it using a canopy resistance approach50. As the HCHO dry deposition lacks a strong constraint on vegetation51, we believe it can explain the observed difference. Secondly, the HCHO asphalt pavement reactivity would also benefit from better constraint. Thus, overall, our calculated dry deposition data set is consistent with Ervens and coworkers.

In the 1D model developed for urban environments by the Stutz group22,52,53, NO2 and O3 uptakes on soil only are considered, while HCHO is not. O3 uptake is taken from dust proxy studies and set to 5 × 10−5, which allows them to reconcile ozone measurement and modeled data in various urban environments and seasons53,54. This value matches very well the 6.1 × 10−5 value calculated for scenarios 3 and 5. To get NO2 concentration in agreement with observations, the Stutz group has selected NO2 uptakes in the 10−4–10−5 range, which again fall close to our calculated 9.7 × 10−6 value. Finally, Trick10, also used NO2 ground uptake ~10−5 to close the NO2 budget in western European cities. The fact that such close uptakes are obtained using two very different approaches (model and calculation) reinforces our assumption of taking an asphalt fraction of 30%.

Discussion

This section includes the discussion of model simulations and the related atmospheric implications. The modeled mixing ratios for the different scenarios on day 6 between 12:00 and 16:00 are presented in Fig. 2. The time series and the 24 h average on day 6 are available in Figs. S7–S9.

Fig. 2: Statistical summary of the model outputs.
figure 2

Except for NO3 radicals, data from 12:00 to 16:00 on day 6 are used. For NO3, data from 23:00 to 03:00 on days 6 and 7 are used. Scenario 2 = base case + NO2, Scenario 3 = base case + O3, Scenario 4 = base case + HCHO, and Scenario 5 = base case + NO2 + O3 + HCHO. The boxplot represents quartiles and median, the red dot is the average, and the gray dots are the individual data points (6 per hour for 4 h, yielding 24 points). Scenarios are presented in Table 1. Statistical difference between the base case and other scenarios is presented as p-values below 0.05 (*), 0.01 (**), 0.001 (***), and 0.0001 (****). Ns means non-statistical difference. Time series plots are available in Figs. S7–S12.

Full size image

For O3, only implementing an ozone dry deposition parametrization leads to a statistical change in the mean mixing ratio with a 9.5 and 9.1 ppb reduction for scenarios 3 and 5, respectively, compared to the base case. The dry deposition becomes a significant sink of ozone. An increase in NO2 deposition (leading to a reduction of 0.1 ppb in the NO2 level) induces a change of +0.4 ppb (statistically non-significant) in the O3 mixing ratio, which is typical of a NOx saturated regime in an urban environment55. HCHO deposition has almost no effect on the O3 level change of −0.3 ppb (statistically non-significant). For the NO2 case, compared to the base case, a direct increase in NO2 uptake induces a NO2 level reduction of 0.1 ppb (scenario 2), while the HCHO dry deposition does not affect the NO2 burden (scenario 4). In comparison, the reduction of O3 levels in scenarios 3 and 5 leads to an increase in the NO2 mixing ratio mainly because of the reduction of the NO2 + OH (O3 being a primary source of OH) and, to a lesser extent, the NO2 + O3 loss pathways. Regarding HCHO, only scenario 4 shows changes compared to the base case. In other scenarios, trends in the HCHO time series differ from the base case (Fig. S7). However, as HCHO has both primary and secondary sources, changes in chemistry compensate for an increase in the dry deposition.

NO is also affected, especially in scenarios 3 and 5, where O3 levels are lower. Ozone being a major NO sink, the average NO mixing ratios increase by ~41 and ~34% compared to the base case. Figure S11 shows that NO levels are mainly affected during daytime when radical species are at maximum as well. Thus, the RO2 chemistry may shift from a NO to RO2/HO2 dominated loss pathway, affecting the radical cycle and the production of secondary product56.

$${\mathrm{RO}}_{2}+\mathrm{NO}\mathop{\to }\limits^{{k}_{{\mathrm{RO}}_{2}-\mathrm{NO}}}{\rm{P}}+{\mathrm{NO}}_{2}$$
(R1)
$${\mathrm{RO}}_{2}+{\mathrm{RO}}_{2}\mathop{\to }\limits^{{k}_{{\mathrm{RO}}_{2}-{\mathrm{RO}}_{2}}}{\rm{P}}$$
(R2)
$${\mathrm{RO}}_{2}+{\mathrm{HO}}_{2}\mathop{\to }\limits^{{k}_{{\mathrm{RO}}_{2}-{\mathrm{HO}}_{2}}}{\rm{P}}$$
(R3)

In reactions R1 to R3, P is the reaction products and depends on the RO2 molecular composition. Within RACM, P can be HO2, carbonyls or peroxides compounds. We recommend the reader to refer to the original publication for more details43.

Here, we evaluate whether the change in NO concentration for the different scenarios impacts RO2 chemistry. To compete with NO, RO2 and HO2 levels need to be higher than the product of NO concentration with the ratio of their reaction rate coefficient (\({k}_{{\mathrm{RO}}_{2}-\mathrm{NO}}/{k}_{{\mathrm{RO}}_{2}-{\mathrm{RO}}_{2}}\) and \({k}_{{\mathrm{RO}}_{2}-\mathrm{NO}}/{k}_{{\mathrm{RO}}_{2}-{\mathrm{HO}}_{2}}\), respectively). Taking the reaction rate coefficient as in Table 2 and a NO mixing ratio of 500 ppt (about the minimum observed during daytime across all the scenarios), RO2 and HO2 mixing ratios at 2000 and 200 ppt are necessary, respectively. These values are unreasonable and orders of magnitude higher than our model prediction (100 s and 10 s of ppt, respectively). Under the simulated conditions, ozone dry deposition induces changes in NO levels but will not affect the fate of RO2 radical.

Table 2 Reactions and rate coefficients used to estimate the RO2 fate in a changing NO environment
Full size table

The main driver in all the changes observed (Fig. 2, Fig. S7–S12, Tables 3 and 4) is the stronger ground uptake of ozone. In comparison, the NO2 and HCHO increase in deposition velocity is minor. The lower O3 levels will lead to lower OH radicals during daytime and higher NOx, fewer secondary products, as shown for HCHO, and fewer NO3 radicals during nighttime. The overall oxidation capacity is reduced, and the primary compound lifetime is longer.

Table 3 Parameters used for the sensitivity analysis and model maximum concentrations observed on day 6
Full size table
Table 4 Oxidative capacity from the base case scenario and scenario 5 as a function of the time periods and oxidants
Full size table

Implementing the dry deposition on asphalt in the model significantly changes the levels of O3 and NO2, which play an important role in atmospheric chemistry and on radical levels and thus on the atmospheric oxidative capacity. The daytime OH radical levels for scenarios 3 and 5 drop by ~10 and 8%, respectively, compared to the base case (Fig. 2). A reduction in ozone levels would decrease the radical production rates through the photolysis of O3 and the reaction of O3 with unsaturated hydrocarbons. Also, in scenarios 3 and 5, the rise of NOx levels adds to the OH sink and would lead to lower OH concentrations (Figs. 2, and S8–S11). As well, in scenarios 3 and 5, nighttime NO3 radical levels drop by ~58% compared to the base case, which is mainly explained by a strong reduction of its source; the NO2 + O3 reaction (~ −45% O3 nighttime level is observed (Fig. S9)).

As the asphalt fraction was an arbitrary choice which governs uptakes and depositions velocities, and infine the levels of key atmospheric compounds (Fig. 2), sensitivity analysis for scenario 3 has been carried out. The asphalt fraction varied from 0 to 30% in 10% steps, and the associated changes in O3 deposition velocities are presented in Table 3.

Increasing the asphalt fraction reduces O3 and OH daytime (12:00–16:00) and NO3 nighttime (23:00–03:00) maximums. The concentration decrease is linear with the asphalt fraction (R2 > 0.97, Fig. S13). Per unit of O3 \({\nu }_{d}\), the O3, OH, and NO3 max oncentrations are reduced by 33, 20, and 100%, respectively. The results indicate a reduction in maximum oxidant levels compared to the base case, even for a 10% asphalt fraction urban cover. Clearly, this model comparison shows the important effect of dry deposition on urban radical levels.

However, the absolute oxidant concentration does not consider the reaction rate of products removed from the atmosphere after oxidation and thus lacks in picturing the real effect of oxidant change. For that reason, the oxidative capacity metric (OC) has been introduced57 and we follow the definition after Geyer et al., 200158:

$$\mathrm{OC}=\sum ({k}_{{Yi}}\times [{Y}_{i}]\times [X]$$

where kYi is the bimolecular rate coefficient for the reaction of primary emitted molecules Yi (VOC, CO, CH4) with an oxidant X (OH, NO3, and O3).

Table 4 summarizes the absolute OC for each oxidant and their sum in the base case and scenario 5, along with the relative change in scenario 5 compared to the base case. Because the OC depends on the number of primary emitted molecules considered (we have considered only a subset of compounds), absolute OC values are not directly comparable with those in the literature; hence, we focus solely on relative changes between scenarios and on oxidant weighting. On a 24-h basis, and during daytime, total OC is reduced by approximately 5%, mainly driven by a reduction in OH OC (~5%), which accounts for the large majority of total OC (>96%) over these two periods. In comparison, at night, OH OC contributes ~50%, while O3 and NO3 OC account for ~15 and ~40% of the total, respectively. Because the NO3 OC remains constant during that period, the change in total OC reflects the reduction in OH OC, balanced by the unchanged NO3 OC. Total OC is thus reduced by less than 10% between the two scenarios at night. The near-identical absolute NO3 OC at night for both scenarios results from a balance between higher NO3 and lower VOC levels in the base scenario, compared to lower NO3 and higher VOC concentrations in scenario 5. This trend is explained by less O3 in scenario 5, implying a lower NO3 production but less VOC loss via oxidation. Our model results highlight the importance of accurately defining the O3 loss routes, including ground deposition.

To obtain further information on the mechanisms related to ozone loss, Fig. 3 presents the relative contributions of the different ozone loss pathways for the afternoon of day 6 (12:00–16:00) under scenario 5. The homogeneous processes account for approximately 60% of the total, dominated by photolysis (~48% of the total) and the reaction with radicals (~10% of the total). Heterogeneous processes contribute the remaining 40%, dominated by reaction on the asphalt surface. By contrast, in the base case (no asphalt) surface-mediated loss represents only ≈1% of ozone removal. Inclusion of the asphalt reaction increases the absolute first-order rate coefficient of ozone from 5.8 × 10−6 to 9.24 × 10−6 s−1. The comparison between the base case and scenario 5 shows that the O3 dry deposition is mainly controlled by the asphalt reactivity above 98%. In the case of NO2, more than 92% of the dry deposition is also controlled by asphalt reactivity. These values likely represent upper-bound estimates of asphalt surface reactivity effects.

Fig. 3: Modeled relative contribution to ozone loss pathways in scenario 5 between 12:00 and 16:00.
figure 3

The O3 + Alkenes term embeds all the ozone reactions with unsaturated hydrocarbons (SI for more details). The absolute loss rate for the base case and scenario 5 are also presented in the SI.

Full size image

We recognize limitations in using a 0-D box model for simulating certain atmospheric dynamics, such as dry deposition and pollutant dispersion, given the model’s assumption of instantaneous mixing within the PBLH and its idealized treatment of urban structures. Additionally, the simplified emission scenario used here, while sufficient for initiating ozone production, does not capture the full complexity of atmospheric chemistry59,60. Our simulations are not intended to replicate precise atmospheric compound levels but serve instead to assess the relative contributions of various pathways. Future work may benefit from employing a 1-D chemistry model to refine estimates of asphalt reactivity’s influence on atmospheric composition and oxidation capacity53.

Our results demonstrate that the heterogeneous loss of atmospheric reactive pollutants, and particularly O3, on asphalt pavements is an efficient yet not enough considered atmospheric process in urban environments. Our results showing an ozone level reduction and its impact on the chemistry would explain why atmospheric models fail to reproduce the concentrations of NOx in urban environments18,19,20,24. While the results are therefore representative of typical European road surfaces, future studies using a wider range of asphalt types and formulations would be beneficial to assess possible variability in ozone uptake efficiency and to further validate the generality of the observed trends. In addition to this work, pioneer studies showed the existence of positive NO2 flux from the surface to the atmosphere25. The author measured NO2 asphalt emission factors ranging from 0.35 to 7.3 µg m−2 h−1 for temperatures between 23 and 60 °C. These fluxes are in the lower range of the NO2 deposition flux modeled in this study (from ~5 to ~40 µg m−2 h−1). Thus, having a NO2 asphalt source in the atmosphere may reduce or cancel out the effect observed and modeled in this study. Altogether, it is evident that asphalt plays a role in the urban atmosphere, with important implications for the atmospheric oxidative capacity during both day and nighttime. Still, the quantification of these processes remains scarcely studied. This urban setting, where most of the global population lives, presents a major challenge to maintaining air quality and public health and deserves more attention.

Methods

Origin of asphalt pavements

The uptake of trace gases of atmospheric relevance on asphalt pavement was performed on freshly collected and aged real-world asphalt samples that have been exposed to traffic under atmospheric weathering conditions17. This study used two different types of real-world asphalt pavement samples. The first sample is an asphalt pavement with 5% bitumen and a penetration grade of 35/5030, named (Acity–5–35/50) in the following. Acity–5–35/50 corresponds to the most widely applied formulation of asphalt sample in French cities. Our samples originate from Douai city, Northern France, (50°22′17″N, 3°04′48″E) in the metropolitan area of Lille, a densely urbanized area of 7200 km2, with a population of approximately 4 million. It has been collected during road resurfacing works in a street in the center of Douai (rue de Bellain) in March 2021 after 30 years of service (Fig. S1). According to local authorities, approximately 5000 vehicles per day, with 5% of heavy vehicles, were driving on the street on a daily basis. Further details about this sample are presented in the studies of Lasne et al.17. The second real world sample was extracted on August 2021 from a parking lot in IMT Nord Europe Institute (50°22'29.6“N 3°04'05.4“E) after building renovation works. Hereafter, this sample is named (Aparking–5–35/50).

After collection, the asphalt samples were flushed with dry air and washed with distilled water to remove dust and leftover materials and were kept in a clean environment for several weeks before their use (Fig. S1). In addition, samples were cut into rectangular fragments with a thickness of approximately of 1 or 2 cm, ensuring that the uptake of the trace gas of interest interacts mainly with the surface exposed to the atmosphere. To validate that, uptake measurements were also performed by applying a hydrocarbon wax layer to the unexposed, presumably inert sample surfaces, that is, the back and the edges (see results section). To achieve this, the inert sample surfaces were dip-coated in bulk molten hydrocarbon wax at T > 70 °C, following standard surface passivation practices in heterogeneous kinetics studies61,62.

Non-weathered asphalt mixture samples were collected on-site during deposition in Douai in 2018; thus, they have not experienced weathering, but only short-term aging during asphalt production. This asphalt sample also had characteristics of 5% bitumen, and a 35/50 penetration grade. After their collection, the samples were stored in the laboratory at room temperature in a fume hood under controlled ambient air conditions for around two years before being used in experiments (Fig. S1). The objective was to evaluate the uptake behavior of samples not exposed to traffic and weather conditions. In addition, storing the asphalt substrates under clean air eliminated the strong VOCs emission profile, which is anticipated to be elevated a few hours after their application but decrease abruptly with time. In the following of the manuscript, the term (L–5–35/50) is used to identify the non-weathered asphalt pavements stored in the laboratory and used for the uptake measurements.

To perform the uptake measurements, the original samples were cut into several pieces in order to fit into the sample compartment (SC) of the reactor. The latter allowed us to perform repeated measurements using the same sample and evaluate the impact of the surface to volume ratio of the samples to the uptake process. The mass of asphalt pavement samples varied between 11 and 142 g. The geometric surface area of the irregularly shaped asphalt samples was determined by reproducing the front side on a printer/copier and subsequently cutting out the area exposed to the probe gas. The quantitative evaluation after weighing was performed using weighed samples of paper of a defined area. The supporting information also provides information on the measured specific surface area of \({S}_{\mathrm{BET}}=\,\)0.04 ± 0.01 m2 g−1 (see also Text S1, in SI).

Description of the experimental set-up and used protocol

The uptake experiments were performed at room temperature (293 K), and dry conditions under molecular flow conditions inside a Knudsen flow reactor (KFR) coupled with a modulated molecular beam mass spectrometer (MS)63,64. A recent review of this kinetic technique has been published64, and here we limit ourselves to a summary account of the experimental method. Complementary information to those provided in the main information about the KFR and the protocol applied, the calculation of the uptake coefficients, etc., are provided in the SI (Text S2) A scheme of it is displayed in Iannarelli et al. and Zogka et al.63,64. The instrument is configured as a two-chamber flow reactor. The lower chamber is the SC, which contains a piece of the corresponding asphalt pavement sample of known mass (Fig. S1). Two SC, a standard and a large one with corresponding volumes of 98.2 cm3 and 281.4 cm3 (see Table S1 for dimensions), respectively, were used depending on the mass of the examined asphalt sample. This allowed us to change the surface to volume ratio of asphalt pavements in the reactor and elucidate possible contributions to the uptake of pollutants. The upper part of the KFR is the main body of the reactor (plenum), hosting the inlets for the admission of the probe gas of interest. A leak-tight cover plate (plunger) was used to expose or isolate the SC with the sample of interest to the gas flow admitted to the KFR.

The uptake coefficient (γ) of the pollutants on the asphalt surface is determined through the uptake experiments. According to the IUPAC definition; the uptake coefficient is a dimensionless parameter that expresses the probability of a molecule to be taken up by the sample and is used to describe the heterogeneous loss of pollutants on a surface49. In the current study, the steady state uptake coefficient, \({\gamma }_{\mathrm{ss}}\), is determined and denotes the loss probability of a pollutant on the asphalt pavement after a long surface exposition, which is considered to be adequate for atmospheric relevant applications. The uptake experiments also provide information about the total number of molecules scavenged on asphalt samples, \({Ns}\) (molecules cm−2). Further information about the expressions related to the measurement of the uptake coefficients and the associated experimental uncertainties, is given in the SI (Text S2) material.

The protocol followed for the blank experiments (in the absence of asphalt sample) and measurements in the presence of the sample was similar for all species investigated and included several steps. Typical theoretical uptake profiles facilitating the understanding of the protocol are described in Fig. S2. The first step includes measuring the background signal of the species of interest. Thereafter, the SC is isolated (plunger pushed down), and a constant gas flow is introduced into the reactor to establish a steady state flow or concentration. Once a steady state is achieved, the plunger is moved up, and the SC containing the sample (or empty in case of blanks experiments) is exposed to the gas environment. During blank experiments in which the standard SC of the Knudsen cell was used, no observable difference to the temporally recorded signal was noticed upon moving the plunger up, denoting the wall effect is within experimental uncertainty. However, when the large SC was used, a subsequent decrease of the MS signal was noticed (pink line in Fig. S2) due to the extra reactor volume (16% volume increase). This results in a filling transient that instantly reduces the concentration of the gas in the reactor. While negligible, this transient effect, had no impact on the \({\gamma }_{\mathrm{ss}}\) determination (see also Fig. S2), but contributes to the calculation of \({Ns}\) and thus was subtracted from the measurements. The results from other blank experiments performed to validate the inert nature of the hydrocarbon wax are displayed in the SI section (Tables S2 and S3). When the asphalt surface is present in the SC, a sharp decrease in the gas concentration is noticed, owing to the uptake of the gas on the asphalt substrate. This process may either lead to saturation of the surface (i.e., recovery of the gas phase concentration to the pre-exposure levels), or in the case of a constant gas consumption, the flow is stabilized at a lower level, as seen in Fig. S2.

After several minutes or hours of exposure, the SC was sealed against the plenum of the KFR to evaluate the nominal concentration of the gas of interest in the reactor. Thereafter, the gas flow was subsequently halted. The SC held the saturated substrate as a static, closed reactor vessel. Once the gas flow completely vanished and the MS signal returned to the baseline, the SC was opened again for the desorption phase of the experiment (see green marked section in Fig. S2). It is important to realize that all uptake experiments reported here used a KFR with a nominal 1 mm diameter exit orifice without exception, of which the effusion rate constant is given in Table S1. However, for the desorption phase, we started the experiment using the 1 mm diameter exit orifice similar to the uptakes, but was later switched to larger exit orifices (in situ in the order 4, 8, and 14 mm dimeter orifices), when the MS signal decreased, accelerating the desorption process owing to the increased pumping speed. The total desorption yield resulted from the integral of the signal obtained at all four exit orifices and was expressed as a fraction of the original uptake yield.

It should be noted that all experiments were conducted under dry conditions, and the potential influence of RH on the uptake process was not assessed in this study. Asphalt pavements are compact, non-porous materials coated with bitumen, a hydrophobic binder, which limits water adsorption and may reduce competition between water and gaseous pollutants for surface sites. Nonetheless, this assumption remains to be experimentally verified, and the possible influence of RH on pollutant uptake by asphalt surfaces represents an important aspect for future investigation.