High urban NOx triggers a substantial chemical downward flux of ozone

Nitrogen oxides (NOx) play a central role in catalyzing tropospheric ozone formation. Nitrogen dioxide (NO2) has recently reemerged as a key target for air pollution control measures, and observational evidence points toward a limited understanding of ozone in high-NOx environments. A complete understanding of the mechanisms controlling the rapid atmospheric cycling between ozone (O3)–nitric oxide (NO)–NO2 in high-NOx regimes at the surface is therefore paramount but remains challenging because of competing dynamical and chemical effects. Here, we present long-term eddy covariance measurements of O3, NO, and NO2, over an urban area, that allow disentangling important physical and chemical processes. When generalized, our findings suggest that the depositional O3 flux near the surface in urban environments is negligible compared to the flux caused by chemical conversion of O3. This leads to an underestimation of the Leighton ratio and is a key process for modulating urban NO2 mixing ratios. As a consequence, primary NO2 emissions have been significantly overestimated.


INTRODUCTION
Nitrogen oxides (NO and NO 2 ) play a crucial role for controlling the oxidizing power of the atmosphere (1,2). More recently, space-based remote-sensing observations of NO 2 have become a tractable tool to monitor the catalytic efficiency of the atmosphere (3,4). The complex chemical cycling, however, still presents a challenge for accurate ozone predictions in polluted environments (5,6). Nitrogen oxides have also emerged as a primary public health concern in many regions (7,8), and significant efforts have been underway to understand the fate and chemical cycling between O 3 -NO-NO 2 in urban areas (9,(10)(11)(12). Because NO 2 is typically identified as an irritant of the respiratory tract (13), atmospheric levels are regulated and subject to a significant policy debate (14,15). World Health Organization (WHO) air quality (AQ) standards have been implemented in Europe as regulatory action under the EU Thematic Strategy on Air Pollution, which aims to limit urban street canyon NO 2 concentrations to 40 μg/m 3 per year (or 200 μg/m 3 per hour on less than 18 days/year) (16). Compared to the United States (3,17), where NO 2 and NO x have generally been declining steadily, current trends across urban European AQ networks show that NO 2 has not decreased as projected (14,18). Because of slower than expected decrease in NO 2 levels and stricter AQ standards, regulatory thresholds of NO 2 are now violated at many stations (19). More recently, a slowdown of NO 2 concentration reductions has also been reported over the contiguous continental United States (15), and speculation over the causes points toward several possibilities, including smaller than expected NO x reductions from selective catalytic reduction car exhaust systems. Significant primary NO 2 emissions from internal combustion engines (ICEs), particularly Diesel engines (20), are also thought to play a major role. These uncertainties have far-reaching ramifications for air pollution control measures and have played a central role in the Diesel emission scandal (21). Modeling studies (9) have often relied on the assumption that ICEs primarily emit NO x in the form of nitrogen oxide (NO), but key questions in the context of urban NO 2 pollution control remain: What fraction of NO x is directly emitted as NO 2 (9), and to what extent does secondary partitioning within the O 3 -NO-NO 2 triad ultimately control street canyon NO 2 levels in high-NO x environments? Concerning primary NO 2 emissions, fleet averages are particularly difficult to constrain. The reported range from individual exhaust plume and tailpipe measurements, for example, suggests that the NO 2 /NO x emission ratio can vary between 5 and almost 40% (22), depending on car make, driving conditions, emission standard, and vehicle age. The rapid interconversion within the O 3 -NO-NO 2 triad has made it particularly hard to provide a quantitative top-down assessment of urban NO 2 sources based on ambient concentration measurements.
Under sunlight conditions and high-NO x pollution, the cycling between the O 3 -NO-NO 2 triad can be described by the following reaction sequence The chemical relaxation time scale to equilibrium of the NO x triad (Eqs. 1 to 3) (23) can be derived as t ¼ 2 ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi where j represents the photolysis rate (Eq. 1) and k 3 is the reaction rate constant defined by Eq. 3. For typical midday summer conditions, this equates to time scales of about 100 s, comparable to the vertical turbulent exchange time in the urban boundary layer. Because of the rapid interconversion, the partitioning between NO and NO 2 is typically dominated by chemistry, and inferences on primary NO 2 emissions remain very uncertain. Furthermore, the rapid cycling between NO, NO 2 , and O 3 is used to argue that steady-state conditions are achieved in mid-to high-NO x environments. This photostationary state (PSS) can be regarded as a null cycle and, according to Leighton (24), is defined by the following ratio (Φ) A number of studies evaluated this ratio (25)(26)(27) and reported Φ varying around one depending on the abundance of NO x . In more remote or rural regions, peroxy radicals (RO x = RO 2 + HO 2 ) can compete with reaction in Eq. 3, and deviations from PSS lead to Φ being systematically greater than one. This fact has been used extensively to constrain peroxy radicals in the atmosphere and test our current understanding of tropospheric ozone chemistry (26,(28)(29)(30)(31). More recently (32), it was argued that dynamical constraints were not fully considered in past assessments aiming to infer peroxy radicals from PSS. This fosters speculation whether the PSS approach generally under-or overestimates peroxy radicals. Incidentally, recent observations seem to point toward higher ozone production rates in high-NO environments (6). Could PSS have generally been underestimated? Since dynamical effects can lead to either negative or positive deviations from PSS, a conclusive answer whether past assessments were positively or negatively biased is currently not possible because of a lack of fundamental observational constraints.
The most direct approach for measuring turbulent surface fluxes is based on the eddy covariance method (33). Theoretical considerations (34)(35)(36)(37) show that direct flux and concentrations measurements of the O 3 -NO-NO 2 triad allow us to experimentally constrain the set of integrated equations (Eqs. 1 to 3). Here, we present long-term seasonal eddy covariance flux measurements to quantitatively test our current understanding of the urban O 3 -NO-NO 2 triad in a high-NO x environment and combine these with remote sensing data and a theoretical model framework. By experimentally separating dynamical and chemical terms, assumptions inherent to the analysis of PSS can be tested in the context of chemical reactions and net import or export of NO, NO 2 , and O 3 . Furthermore, the fluxes of NO x (<w′NO′> + <w′NO 2 ′>) and O x (<w′O 3 ′> + <w′NO 2 ′>) can be considered conserved and are proportional to the surface NO 2 flux because the measured ozone flux directly represents the fraction of NO reacted to NO 2 on transport time scales (e.g., 100 to 300 s) in the lowest part of the urban boundary layer. The presented observations are therefore uniquely positioned to shed light on the issue of primary city-scale NO 2 emissions from combustion sources and the validity of PSS analysis in urban environments.  Figure 1 gives an overview of the underlying dataset used for this study. Observations reported here span nearly 4 years of continuously measured mixing ratios and fluxes of O 3 , NO, and NO 2 . Data shown in Fig. 1 Fig. 2 and shows the horizontal scale of surface emissions that are captured by measurements at the flux tower. NO x emissions are typically a factor of 2 to 3 lower on weekends than on weekdays. The weekend-weekday effect for the summer of 2018 shows that peak O 3 was about 10 ppbv higher on weekends, reflecting lower NO x mixing ratios. This is comparable to what we observed in 2015, when the difference between weekend and weekday was on the order of 7 to 8 ppbv (18). Surface in situ observations of NO 2 toward the east agree to within the uncertainty of remote sensing observations obtained from a ground-based Pandora system, which has been used extensively to validate Tropospheric Monitoring Instrument (TROPOMI) observations (fig. S1, A and B). The TROPOMI monthly average for August 2018 shows a horizontal gradient of NO 2 from east to west. This gradient is consistent with decreasing surface emissions toward the west. We can take advantage of these data to constrain daytime horizontal advection at the site. The average advection flux of NO 2 was calculated from the horizontal NO 2 concentration gradient inferred from TROPOMI, scaled to surface observations (dC/dx < 10 −4 ppbv/m). The gradient was integrated from the displacement height (d = 18 m) to the measurement height (z m = 42 m) and multiplied by the horizontal wind speed along the valley axis. Relative to the observed vertical flux of NO 2 , we find that the advection flux plays a minor role (i.e., <6% during daytime). Since, typically, 80 to 90% of daytime NO x exists in the form of NO 2 , the NO x advection flux is expected to be quite comparable to NO 2 . It corroborates conventional micrometeorological assumptions that daytime advection often plays a minor role, and we expect a similar behavior at other urban locations where NO x emission fluxes are comparable to observations shown here.

Interpreting the urban Leighton ratio
During daytime, we observed substantial downward fluxes of ozone ( , v e = −flux/concentration) is 1 to 2 cm/s (maximum, 3.6 cm/s). The flux footprint (Fig. 2) shows that 90% of the surface is covered by buildings, roads, or concrete and only 10% is covered by vegetation for the data considered here. Under these conditions, a typical upper limit for a dry deposition velocity for ozone at the surface is on the order of 0.3 cm/s (see the Supplementary Materials), which is <20% of the observed O 3 deposition velocity. A length scale analysis ( fig. S4) demonstrates that the surface influence at 42-m measurement height is 0.17. The ozone flux at measurement height is therefore primarily (i.e., >96%,~0.20 × 0.17) driven by chemistry, in particular, the reaction of NO + O 3 . Dahmköhler numbers are dimensionless quantities that describe the time scale of a reaction to that of transport and can be defined for mixing ratios and fluxes (see the Supplementary Materials). The Damköhler numbers and flux Dahmköhler numbers for ozone for the entire dataset are found to lie in the range of 0.1 to 0.8 and 0.2 to 1.2, respectively. They support the conclusion that both the flux and concentration profiles are significantly influenced by chemical conversion of NO to NO 2 . From an ecological perspective, these findings are quite relevant because such an efficient chemical conversion of ozone implies that urban vegetation is taking up O x in the form of NO 2 rather than O 3 , which contrasts the situation of O x uptake in natural ecosystems, where plant exposure is dominated by O 3 uptake. It has long been recognized that excessive ozone uptake is detrimental to plant health. On the other hand, it has also been shown that plants can quite effectively absorb NO (38) and NO 2 (39) and use NO x as an alternative source of nitrogen, in addition to N fixation from soils. With projected future decreases in NO x , urban plants will gradually be exposed to more ozone. How much could that affect ozone exposure for urban plants? As a sensitivity experiment, we can look at the weekend/weekday ratio for the entire dataset. The yearly average NO x emission flux for weekdays is 22 and 10 nmol/m 2 per second for weekends. The associated increase in ozone mixing ratios on weekends is~10%. An established metric for the exposure of plants to ozone is the AOT40 (accumulated ozone exposure over a threshold of 40 ppbv). If we assume a future decrease in NO x emissions by a factor of 2 and a corresponding increase in ozone by 10%, we find that present-day AOT40 (5840 ppbv h) could increase to 9800 ppbv h, which would be quite significant.
Analogous to an enhanced spurious ozone deposition velocity, the Leighton ratio (Eq. 1) exhibited values close to 1 around 10 UTC but declined with increasing negative O 3 flux. The ratio continued to drop over the course of the day, as the ozone flux became more negative. During conditions of low light (<6 and >18 UTC), the Leighton ratio becomes poorly constrained as j NO2 approaches zero. These periods were therefore not used for the correlation plot in Fig. 3B. We find a decrease in the Leighton ratio with decreasing O 3 fluxes. During midday and high-radiation conditions, we observe a Leighton ratio as low as 0.5 when the negative ozone flux peaked and is in the range of −25 to −30 nmol/m 2 per hour. Within the uncertainty, the observed trends of vertical ozone deposition flux and Leighton ratio are generally reproduced by a simplified second-order closure model, which explicitly calculates second-order turbulence moments, with a predicted minimum of about 0.7. In urban areas, the establishment of the Leighton ratio is thought to be dominated by the O 3 -NO-NO 2 triad due to high-NO emissions. To estimate the influence of RO x chemistry in the present study, we use a chemical box model sensitivity analysis that previously calculated a range of RO x between 2 and 15 parts per trillion by volume ( pptv) (18). This is comparable to typically observed total peroxy radical densities (40). For a RO x mixing ratio of 8.5 pptv, the reaction rate of <NO + RO x > would be approximately 8% compared to the <NO + O 3 > null cycle reaction. Deviations of Φ < 1 are a direct consequence of NO diffusing upward reacting with O 3 . This will lead to a flux divergence of NO, NO 2 , and O 3 (e.g., here dF NO /dz < 0), which, in turn, is the reason for a disbalance of the Leighton ratio. It will approach one when the flux divergence (dF/dz) approaches zero (e.g., top of the surface layer). The divergence of turbulent fluxes is caused by a disequilibrium of chemical destruction and production terms within the triad and potentially other chemical reactions (e.g., RO x ). Mathematically, we can recast the Leighton ratio as following where the overbar represents a temporal average and the last part on the right side is the fluctuating part due to turbulent mixing. For simplification, let us assume that NO is in steady state, and then it becomes clear that the deviation of Φ from 1 is driven by the vertical flux divergence term of NO: Surface NO emissions will drive this flux gradient. A significant downward flux of ozone (and upward flux of NO) then corresponds to a perturbation of Φ toward values below one. The produced fraction of NO 2 is a direct consequence of the NO + O 3 reaction and diffuses upward. In this context, we also constrained the influence of horizontal advection of NO 2 using remote sensing data. We find that advection fluxes are minor compared to vertical turbulent fluxes of the NO-NO 2 -O 3 triad. We find that a key process is how much NO is emitted and how much of ozone is transported downward as it reacts with NO driving a vertical flux gradient (i.e., divergence). While one might argue that measurements in a high-NO environment are dominated by the <NO + O 3 > reaction, it is clear that downward ozone fluxes also occur in areas where <NO + RO x > reactions can compete more substantially with the <NO + O 3 > reaction. Peroxy radicals inferred from the PSS approach therefore generally have to be considered to be biased low under most scenarios where there are sufficient NO surface emissions. We also investigated other seasons and find that ozone fluxes are smallest (i.e., highest) during winter ( fig. S9), when peak daytime fluxes were about three to five times higher than during summer (e.g., −6 nmol/m 2 per second in winter versus −30 nmol/m 2 per second in summer). The reason is that winter ozone mixing ratios are typically more depleted because of excessive NO mixing ratios resulting in significantly smaller absolute ozone fluxes. In addition, convection is quite a bit smaller in winter than in summer. As expected, spring and autumn lie between the two other seasons. As a consequence, the Leighton ratio was, on average, close to 1 (1.05 ± 0.1) during winter (low-ozone flux season), while the seasonal average for summer lies around 0.7 ± 0.1 (high-ozone flux season), with the other seasons in between. Data presented here give a realistic range of the bias of the Leighton ratio as a function of ozone flux. The rapid conversion of NO to NO 2 via reaction with O 3 has a large impact on the formation of NO 2 in the surface layer, making it hard to assess the direct amount of NO 2 emitted from combustion engines, which is a relevant quantity for policy making.

Primary emissions versus secondary production of NO 2
From a policy perspective, NO 2 is considered a particularly relevant species due to its toxicity and ozone-forming potential. AQ thresholds are triggered for NO 2 as a primary toxic air pollutant. This has sparked a debate on the effectiveness of reducing primary NO 2 emissions to meet street-level NO 2 pollution. For most combustion processes, the bulk of NO x is typically thought to be emitted as NO. Discrepancies between declining NO x and NO 2 mixing ratios, however, have indicated that primary emissions of NO 2 might play a significant role due to the emergence of Diesel vehicles (22,41). However, the magnitude of fleet average primary NO 2 emissions has been particularly hard to constrain. Since primary emissions of NO will only convert O 3 into NO 2 due to the NO + O 3 reaction, these emission fluxes would not have an impact on O x surface fluxes. However, direct emissions of NO 2 should be detected in the O x surface emission flux. The overall lifetime of NO x is large enough (several hours) compared to the turbulent time scale (100 s) so that NO x can be regarded as a conserved tracer. As an example, Fig. 4 shows the relationships between O x (:=NO 2 + O 3 ) and NO x (:=NO 2 + NO) fluxes (Fig. 4A) and mixing ratios (Fig. 4B) for the 2018 campaign. All data were preselected for ozone mixing ratios of <5 ppbv. The slope of the O x to NO x flux is directly related to the fraction of primary emitted NO 2 at the street level. We find that the slope of O x versus NO x flux is 10.3 ± 0.1%. For concentration ratios, it is 80% higher (i.e., 18.4 ± 1.5% at the street level and 18.3 ± 2.5% at the flux tower). Looking more closely into the issue of secondary conversion, we find that, even at low ozone mixing ratios (<5 ppbv), the concentration ratio exhibits a dependence on the downward ozone flux. Figure 5 shows the NO 2 versus NO x mixing ratios color-coded by vertical ozone fluxes measured at a close-by street canyon station. Data are filtered for O 3 < 5 ppbv. Air masses that are systematically enriched in NO 2 tend to be associated with higher downward O 3 fluxes. We find a statistically significant difference between higher NO 2 /NO x ratios for larger negative ozone fluxes and lower NO 2 /NO x ratios for less negative (i.e., higher) ozone fluxes (even for O 3 < 5 ppbv). On the basis of the two sample t tests, the hypothesis is rejected that the NO 2 /NO x ratio is independent of the ozone flux. We find that the mean bias at the 5% confidence level is 35% for NO 2 /NO x ratios below an ozone flux threshold of −5 nmol/m 2 per second. The analysis is fairly independent of the chosen threshold but increases with lower thresholds as expected. It is, for example, 40% for an ozone threshold of −10 nmol/m 2 per second. This suggests that low O 3 mixing ratios (e.g., <5 ppbv) alone cannot guarantee that the conversion of NO to NO 2 is negligible because of the distance between emission and detection, because higher ozone aloft can effectively mix down into the street canyon. The downward flux of ozone can therefore strongly modulate street canyon NO 2 production. Incidentally, a recent study (42) has reported rapid interconversion of NO to NO 2 at the tailpipe to curbside stage. Our measurements demonstrate that downward mixing and reaction of ozone must be the most important process driving the chemical conversion of NO to NO 2 in urban areas. A primary NO 2 /NO x ratio of 10.3 ± 0.01% is significantly lower than inferred from more indirect approaches, for example, concentration enhancement ratios (EnR), which are affected by emission, chemistry, and transport of NO, NO 2 , and O 3 . It is also significantly lower than model projections (19) that were constrained to earlier indirect top-down approaches. Figures 4 and 5 illustrate that the interpretation of EnR of O x and NO x mixing ratios is quite a bit more complex than for direct fluxes, because various biases depending on data exclusion procedures can alter regression slopes for street canyon stations but tend to lead to an overestimation. Figure 6 extends the analysis to other years and locations based on commonly accepted data analysis approaches. We find that the bias is consistent across years. We also tested the sensitivity of the EnR method with respect to proximity to the emission source for Innsbruck. Data collected at the Innsbruck Atmospheric Observatory (IAO), about 42 m above street level, yield EnR ratios that are only about 16% higher than EnR data at a busy roadside AQ station near to the flux tower (e.g., orange dashed line versus orange circles in Fig. 6). While EnR data from Innsbruck yield comparable results to other European cities (Fig. 6 and see the Supplementary Materials) and previous policy assessments (14,19), we find that this approach overestimates the primary NO 2 emission ratio by approximately a factor of 2 when compared to the more direct flux ratio method (10.3%), which explicitly accounts for the NO + O 3 reaction.

Implications for AQ policy assessments
The fraction of primary versus secondary emitted NO 2 has consequences for predicting future improvements and trends of AQ. AQ trends in Europe show that NO x and NO 2 mixing ratios have not declined as expected from past assessments (16) and that NO 2 mixing ratios declined less rapidly than NO x . This has brought many cities out of compliance due to stricter AQ limits for NO 2 (40 μg/m 3 ), which were implemented according to WHO guidelines (13) and were based on optimistic projections of declining NO x and NO 2 emission trends. We investigated the magnitude of this effect for the present high-NO x location. On the basis of Eqs. 1 to 3, it follows (see the Supplementary Materials) that a change in urban NO 2 trends due to changing NO x , O 3 , temperature, and photolysis rates can be related according to where j := j NO2 and m := k 3 * [O 3 ]. Similar to many European cities, NO x mixing ratios in Innsbruck have been declining since about 2004. An annual trend for the years between 2004 and 2016 was calculated using in situ data obtained from nearby AQ stations (43). The predicted NO 2 trend [i.e., δ(NO 2 )/(NO 2 )] based on Eq. 7, taking into account changes in NO x , O 3, temperature, and photolysis rates, is −1.7% per year. The individual contributions (Eq. 7, right-hand side) are −2.2% (term 1), +0.5% (term 2), and −0.004% (term 3), respectively. The actually measured annual NO 2 trend was higher (i.e., −1.9% per year) than that of NO x (−2.2%). While this difference could potentially be attributed to an increasing primary NO 2 emission fraction, we find that the effect of increasing ozone mixing ratios (term 2, 0.5%) over the same time period can also explain the difference on an annual basis. As shown here, the downward mixing of ozone plays a major role for urban NO 2 mixing ratios. We find that the individual budget terms (Eq. 7) are internally consistent within the uncertainty. While our analysis supports the commonly accepted view that there is a contribution of primary NO 2 emissions from Diesel-fueled exhaust, we find that the absolute fraction is likely overestimated. In Austria, the fraction of Diesel passenger vehicles has increased from less than 10% before 1985 to 55% in 2018 (Fig. 6). On the basis of NO 2 /NO x EnR trends calculated for representative road-side and inner city locations across Austria and compared to Europe wide assessments, the fraction of primary NO 2 emissions has increased along with the fraction of Diesel vehicles. Trends have recently leveled out, and we argue that this is primarily due to a stabilization of the fraction of Diesel vehicles in conjunction with readjusting ozone mixing ratios. Diesel engines tend to emit a higher fraction of NO x in the form of NO 2 compared to gasoline vehicles (44). While NO x  emissions generally increase over the vehicle's lifetime (45), the fraction of primary NO 2 is thought to decrease (14). The combined effect for an entire vehicle fleet is therefore particularly hard to asses. Direct flux measurements, reported here, show that the amount of primary fleet average NO 2 emissions is substantially lower than estimates from past policy assessments (19) and even lower than more recent indirect analysis methods (i.e., by 80%; Fig. 6). Adjusting EnR trends according to direct flux data yields a fraction of NO 2 /NO x that has stabilized at a level of 10%. By taking this ratio and a NO 2 /NO x emission ratio of 2 to 5% for gasoline vehicles (46,47), we obtain a primary NO 2 /NO x emission ratio of about 15 to 18% (by volume) for the current average Diesel fleet in Europe.
Direct flux observations of the O 3 -NO-NO 2 triad show that the downward flux of ozone into the urban roughness layer is dominantly driven by the rapid conversion of NO to NO 2 . We find that 90% of urban NO 2 production is due to the chemical conversion within the triad. On the scale of air chemistry models used for policy making, the downward ozone flux above urban areas is therefore largely controlled by NO emissions, which become an important factor for subgrid-scale deposition parameterizations of O x . As a consequence, urban plants absorb most O x in the form of NO 2 rather than O 3 , which is considered detrimental to plants. This contrasts deposition in natural habitats. We find that a future reduction of NO x emissions by a factor of 2 can increase AOT40 levels for urban vegetation by 67%. The rapid chemical conversion also has a profound impact on the Leighton ratio, which is significantly modified and can be biased low in high-NO environments. Depending on the ozone flux, peroxy radicals estimated from the photochemical steady state will therefore always be underestimated at the surface and might require a reinterpretation based on findings presented here. These observations support recent studies (6) that seem to point toward higher ozone production rates when NO emissions increase. We also find that a substantially smaller than expected fraction (i.e., 10%) of NO 2 is directly emitted into the urban atmosphere and attribute a fleet average primary NO 2 /NO x emission ratio for Diesel vehicles to be on the order of 15 to 18% in Central Europe. NO x emissions, however, are known to have been significantly underestimated by recent emission inventories across European metropolitan areas and elsewhere because of unreported emissions (18,21,48). Our analysis demonstrates that street canyon NO 2 mixing ratios are ultimately controlled by NO emissions and the vertical ozone flux, which also perturbs the photochemical state. We show that NO 2 mixing ratios have been declining less rapidly than NO x . The discrepancy can be explained by increasing O 3 mixing ratios over the same time period, suggesting that increasing ozone mixing ratios in the future can lead to more efficient secondary conversion of NO to NO 2 in street canyons, slowing the decline of urban NO 2 . Attainment of NO 2 in urban areas will therefore largely remain a function of NO emissions and ozone entrainment fluxes from the free troposphere into the urban boundary layer. For measurements presented here, we constrain the analysis to the east sector that dominates during daytime. It also reflects the direction toward the urban AQ stations operated by the state of Tirol (47°15′45.5″N, 11°23′32.5″E). The mean building height in the east sector was calculated from high-resolution laser scan data provided by the state of Tirol (Land Tirol). The distribution of mean building height within the footprint of the east sector was calculated as z h = 18.2 ± 2.4 m.

Instrumentation
A dual channel chemiluminescence instrument (CLD 899 Y, Ecophysics) was used for high-frequency NO and NO x measurements. The instrument was operated in a fast mode acquiring data at about 5 Hz. A NO standard was periodically introduced for calibration. Zeroing was performed once a day close to midnight. Damping time scales for eddy flux measurements were determined as 0.8 s.
Direct NO 2 measurements were also conducted using a cavity ring-down spectrometer (CARDS) (Los Gatos Inc., USA). The instrument was operated in a fast mode with a cell pressure of about 300 torr. The sample flow was predried with a Nafion dryer. Periodic zeros through a NO 2 scrubber were performed every 180 min. The direct comparison of NO 2 mixing ratios between CARDS and nCLD 899 Y shows a good correlation [slope = 0.94 ± 0.01 and coefficient of determination (R 2 ) = 0.98], suggesting minimal impact of NO y -interfering species for the chemiluminenscence method (49).
High-frequency ozone measurements were obtained from a closed-path fast-response chemiluminescent instrument (Sextant Technology, NZ), which was calibrated using parallel measurements from a closed-path ultraviolet (UV) photometric analyzer (APOA-360, Horiba, Japan). A second closed-path UV photometric analyzer was operated at street level. Chemical measurements at the tower were drawing sample ambient air from a pressure-controlled turbulently purged 3/8″ Teflon line, with an overall lag time of about 1 to 2 s. Calibration procedures for dry chemiluminescence methods based on coumarin have been devised previously (50). A closed path eddy covariance system (CPEC 200, short inlet, enclosed IRGA design, Campbell Scientific) measured three-dimensional winds along with CO 2 and H 2 O. Calibration for CO 2 was performed once a day. A four-channel net radiometer was housed on the flux tower. A second CPEC eddy-covariance system was deployed at street level. AQ data of NO, NO 2 , and O 3 were obtained from two nearby AQ stations.
Application of eddy covariance in urban areas have been assessed extensively (51) and were calculated as the covariance between the rotated vertical wind speed and the tracer mole fraction using routines described previously (33). Quality control was performed according to procedures described by Foken and Wichura (52). Systematic errors due to high-frequency losses were obtained from cospectral analysis according to Massman and colleagues (53).

Models
SOMCRUS (Second-Order Model for Conserved and Reactive Unsteady Scalars) is a one-dimensional second-order closure numerical model to study the vertical turbulent transport of trace reactive species in the (daytime) planetary CBL. The model includes basic reactions of the O 3 -NO-NO 2 triad. The Master Chemical Mechanism (MCM) developed by the National Centre for Atmospheric Sciences at the University of Leeds summarizes the state-of-theart knowledge on tropospheric chemistry. The chemical mechanistic information was taken from the MCM v3.3 via the website: http://mcm.york.ac.uk/.