Comparison of tropospheric NO 2 vertical columns in an urban environment using satellite , multi-axis differential optical absorption spectroscopy , and in situ measurements

Tropospheric NO2 vertical column densities have been retrieved and compared for the first time in Toronto, Canada, using three methods of differing spatial scales. Remotely sensed NO 2 vertical column densities, retrieved from multi-axis differential optical absorption spectroscopy and satellite remote sensing, were evaluated by comparison with in situ vertical column densities estimated using a pair of chemiluminescence monitors situated 0.01 and 0.5 km a.g.l. (above ground level). The chemiluminescence measurements were corrected for the influence of NO z, which reduced the NO 2 concentrations at 0.01 and 0.5 km by an average of 8 ± 1 % and 12± 1 %, respectively. The average absolute decrease in the chemiluminescence NO 2 measurement as a result of this correction was less than 1 ppb. The monthly averaged ratio of the NO 2 concentration at 0.5 to 0.01 km varied seasonally, and exhibited a negative linear dependence on the monthly average temperature, with Pearson’sR = 0.83. During the coldest month, February, this ratio was 0.52± 0.04, while during the warmest month, July, this ratio was 0.34± 0.04, illustrating that NO2 is not well mixed within 0.5 km above ground level. Good correlation was observed between the remotely sensed and in situ NO 2 vertical column densities (Pearson’s R value ranging from 0.72 to 0.81), but the in situ vertical column densities were 52 to 58 % greater than the remotely sensed columns. These results indicate that NO2 horizontal heterogeneity strongly impacted the magnitude of the remotely sensed columns. The in situ columns reflected an urban environment with major traffic sources, while the remotely sensed NO 2 vertical column densities were representative of the region, which included spatial heterogeneity introduced by residential neighbourhoods and Lake Ontario. Despite the difference in absolute values, the reasonable correlation between the vertical column densities determined by three distinct methods increased confidence in the validity of the values provided by each measurement technique.


Introduction
In Canada, nitrogen oxides (NO x = NO + NO 2 ) are classified as criteria air contaminants due to their adverse effects on human health and the environment (Environment Canada, 2013).Anthropogenic production of NO x is largely attributed to fossil fuel combustion, with the transportation sector accounting for 68 % of NO x emissions in Ontario during 2008 (MOE, 2011).Previous studies investigating the spatial distribution of NO x have focused on near-road environments as a means to assess human exposure to trafficrelated air pollution (TRAP), characterize TRAP dilution and chemical evolution, inform urban infrastructure planning and policy, and validate dispersion models (Villena et al., 2011;McAdam et al., 2011; Y. J. Wang et al., 2011;Clements et al., 2009;Beckerman et al., 2008).These studies suggest that variability in traffic volume and fleet characteristics, as well as local topography and meteorological conditions, impact the NO x concentration (distance-decay) gradients observed amongst different near-road environments.Negative Published by Copernicus Publications on behalf of the European Geosciences Union.

D. Mendolia et al.: Comparison of tropospheric NO 2 vertical columns in an urban environment
health effects associated with exposure to NO 2 when used as a marker for TRAP include the exacerbation of asthma symptoms, the increased risk of developing cardiovascular and lung diseases, and increased mortality rates (Andersen et al., 2011;Valari et al., 2011;Pereira et al., 2010;Jerrett et al., 2009;Salam et al., 2008, and references therein;Nafstad et al., 2003).Therefore, an effective means of monitoring the spatio-temporal behaviour of NO 2 could do much to support the development of appropriate pollution mitigation strategies.
Satellite measurements of tropospheric NO 2 vertical column densities (VCDs) can provide long-term spatiotemporal trends on a global-to-regional scale.The Ozone Monitoring Instrument (OMI) was launched on board NASA's EOS-Aura satellite on 15 July 2004, and provides daily tropospheric NO 2 column measurements with a spatial resolution up to 13 × 24 km 2 at nadir.OMI has a local ascending equatorial crossing time of 13:45 EST (Levelt et al., 2006).Previous studies have extended the utility of OMI tropospheric NO 2 column measurements to validate and improve chemical transport models (e.g. S. Wang et al., 2011;X. Wang et al., 2011;Hains et al., 2010;Huijnen et al., 2010), provide top-down emission estimates of NO x and validate bottom-up emission inventories (Shaiganfar et al., 2011;Lamsal et al., 2010), quantify long-term trends in NO x emissions over continental source regions (Boersma et al., 2008b), separate anthropogenic NO 2 emissions from biomass burning events (Mei et al., 2011), estimate the lifetime of tropospheric NO x and account for observed seasonal patterns (Lamsal et al., 2010), infer NO 2 surface concentrations and associated long-term trends (Lee et al., 2011;Lamsal et al., 2008), and analyse the impact of precursor species on surface O 3 formation (Duncan et al., 2010).Although OMI tropospheric NO 2 columns have been applied to inform a variety of objectives, there is still a strong need for independent validation measurements (Vlemmix et al., 2010;Celarier et al., 2008;Irie et al., 2008), especially since previous validation studies have demonstrated that the relationship between OMI and field measurements varies from region to region (Hains et al., 2010).
Previous studies have assessed the relationship between ground-based and satellite tropospheric NO 2 VCDs, and have established that VCDs derived from localized-point measurements can directly capture near-source emissions of NO 2 , while the large spatial footprint (≥ 312 km 2 ) of satellite measurements makes it challenging to retrieve this information (Boersma et al., 2009;Ordóñez et al., 2006;Petritoli et al., 2004).These studies have relied on extrapolating groundbased NO 2 measurements to derive VCDs, either by assuming a well-mixed planetary boundary layer (PBL), or using a computationally intensive global chemical and transport model (CTM) with coarse resolution (> 250 km 2 ) to derive a vertical NO 2 profile.On a regional and short-term scale, Ontario's Air Quality Index informs the public about their potential to experience adverse health effects from outdoor air, relying on hourly average in situ measurements of pollutants such as NO 2 , PM 2.5 and O 3 (MOE, 2010b).In downtown Toronto, NO x monitoring stations situated at 0.01 and 0.5 km a.g.l.(43.663 • N, 79.388 • W and 43.642 • N, 79.387 • W) provided the opportunity to examine tropospheric vertical profiles of NO 2 in a near-road environment, and to demonstrate that NO 2 is not well mixed within 500 m a.g.l.These monitoring stations enabled the retrieval of a NO 2 vertical profile with a horizontal resolution of 2.3 km, which is advantageous in comparison to using a CTM, since few models can derive a NO 2 vertical profile with a horizontal resolution less than 5 km on a local scale.Furthermore, the use of a local-scale CTM requires a well-characterized emissions inventory (Bechle et al., 2013;Simpson et al., 2012), which currently does not exist for Toronto.Studies relying on a local-scale CTM have still used simplifying assumptions regarding the freetropospheric quantity of NO 2 , which has previously been assumed to be constant at 0.02 ppb above 2 km (Bechle et al., 2013).
Multi-axis differential optical absorption spectroscopy (MAX-DOAS) is a relatively new technique used to retrieve tropospheric NO 2 columns that is well suited for the validation of coincident OMI measurements (Halla et al., 2011;Vlemmix et al., 2010).MAX-DOAS employs intensity measurements of scattered sunlight at a series of elevation angles, relying on the narrowband absorbance structures of NO 2 within the UV-visible wavelength range, to obtain tropospheric vertical column densities (VCDs) of NO 2 .Knowledge of average photon trajectories is essential for resolving NO 2 VCDs from MAX-DOAS spectra, but is an involved task, having previously been accomplished with radiative transfer modelling or trigonometric approximations.MAX-DOAS draws upon the advantages of both localized-point and satellite-based measurements, offering average pollutant concentrations covering a horizontal scale which has been reported to vary from 3 to 11 km (λ = 357 nm, aerosol extinction coefficient within 1 km a.g.l.varied from 1.02 to 0.05 km −1 ) with a time resolution on the scale of a few minutes (Irie et al., 2011).
Previous studies have investigated the relationship between OMI and MAX-DOAS tropospheric NO 2 VCDs (Shaiganfar et al., 2011;Halla et al., 2011;Wagner et al., 2010;Vlemmix et al., 2010;Kramer et al., 2008;Brinksma et al., 2008;Celarier et al., 2008;Irie et al., 2008), and have demonstrated that a fair agreement was generally observed.Differences amongst the retrieved tropospheric NO 2 VCDs have been attributed to differences in the spatial resolution of ground and satellite remote sensing techniques, in addition to the vertical sensitivity of each instrument, and the NO 2 VCD retrieval algorithm employed.
The majority of studies suggest that OMI exhibits a positive bias over rural (unpolluted) MAX-DOAS measurement sites, and a negative bias over urban (polluted) MAX-DOAS measurement sites (Shaiganfar et al., 2011;Halla et al., 2011;  (Toronto Transportation Services, 2010).DT refers to the Ontario Ministry of the Environment (MOE) Air Quality Network's downtown Toronto Site.CN refers to the MOE's monitoring site situated at the CN Tower.Kramer et al., 2008;Brinksma et al., 2008;Celarier et al., 2008), despite the use of different OMI products (standard versus DOMINO product and versions of these products), MAX-DOAS instrumentation, and NO 2 VCD retrieval algorithms.While the OMI pixel size is ≥ 312 km 2 , the horizontal resolution of the MAX-DOAS instrument decreases rapidly with increasing aerosol load.Therefore, when MAX-DOAS measurements are conducted at a rural site, coincident OMI pixels may include areas with high NO 2 concentrations, and alternatively, when MAX-DOAS measurements are conducted at an urban site, coincident OMI pixels may include areas with lower NO 2 concentrations.
Previous studies have also investigated the use of stricter coincidence criteria when comparing satellite-and groundbased remote sensing techniques as a means to overcome the differences in spatial resolution of the measurements, and improve their agreement.These criteria have included averaging MAX-DOAS VCDs over several azimuths (Brinksma et al., 2008;Celarier et al., 2008), only including OMI measurements with a pixel centre within 0.1 • × 0.1 • of the MAX-DOAS measurement site (Irie et al., 2008), and excluding MAX-DOAS measurements with a relative uncertainty ≥ 10 % (Vlemmix et al., 2010).These studies demonstrated that using stricter coincidence criteria improves the agreement between MAX-DOAS and OMI measurements.
This study examines the influence of NO 2 horizontal heterogeneity in an urban environment on VCDs determined by three methods of differing spatial scales.The goal of this study was to evaluate the response of remotely sensed NO 2 vertical column measurements in regions with spatially heterogeneous emissions of NO 2 examined using in situ-derived column data.Specifically, remotely sensed tropospheric NO 2 VCD measurements from OMI (DOMINO product, version 2.0) and MAX-DOAS were evaluated through comparison with in situ tropospheric NO 2 VCDs determined based on data collected in the downtown core of Toronto, Ontario during select periods in 2006-2010.Tropospheric NO 2 VCDs were derived using data from in situ (chemiluminescence) monitors situated near roadside at 0.01 and 0.5 km a.g.l.This is the first study that uses stationary in situ measurements of NO 2 collected at two elevations to derive tropospheric NO 2 columns in an urban environment for comparison with remotely sensed data.
MAX-DOAS NO 2 differential slant column densities ( SCDs) were converted to tropospheric vertical column densities (VCDs) using the geometric air mass factor (AMF) approximation (Hönninger et al., 2004) in conjunction with the single-scattering validation criteria discussed by Halla et al. (2011) and Brinksma et al. (2008).The impact of NO 2 horizontal heterogeneity on the remotely sensed VCDs was assessed by comparing MAX-DOAS and OMI tropospheric NO 2 VCDs to those derived in situ (and near road).
This paper is structured as follows: Sect. 2 describes the methodology employed, providing an overview of the downtown Toronto measurement site, along with a description of the in situ, MAX-DOAS, and OMI instruments, and the respective NO 2 VCD retrieval algorithms.Section 3 presents the in situ-derived NO 2 VCD's temporal trends, and the comparison between point-source and remotely sensed tropospheric NO 2 columns.Conclusions are derived based on the findings from Sect. 3, and presented in Sect. 4.

Description of downtown Toronto measurement sites
In  NO z , using the procedure reported by Boersma et al. (2009), which is discussed in Sect.2.2.The in situ measurements of NO 2 provided the opportunity to observe the impact of diurnal traffic patterns and associated meteorology on the spatiotemporal distribution of NO 2 in an urban environment.MAX-DOAS measurements were conducted at the University of Toronto's Wallberg and McLennan Physics buildings, as well as at Toronto's Centre Island, as summarized in Table 2. Azimuth angles were chosen based on the constraints of each monitoring site's topography to ensure buildings and trees did not obstruct the MAX-DOAS instrument's field of view.A map of the MAX-DOAS measurement sites is provided in Fig. 1.
An example OMI pixel from 27 May 2008 with its centre located at DT is also shown in Fig. 1.The in situ and MAX-DOAS measurement sites are well contained within this pixel.Approximately 60 % of the pixel area overlaps mainland Toronto, while the remaining 40 % is over Lake Ontario.

In situ NO 2 VCD retrieval
In situ-derived tropospheric NO 2 VCDs (NO 2 VCD in situ ) were calculated assuming that NO 2 exhibits an exponentially decaying vertical profile, as shown in Eq. ( 1) with z = 12 km.The value selected to describe the height of the troposphere (z) made no difference to the magnitude of the VCD as z H NO 2 .The characteristic height of NO 2 (H NO 2 ) refers to the height at which the in situ NO 2 concentration at DT ([NO 2 ] 10m ) decreases by a factor of e, assuming an exponential vertical profile.Mathematically, the same numeric value for the height could describe a fully mixed layer with a concentration equal to that at DT, or a combination of a fully mixed layer under an exponentially decaying layer.While the assumption of an exponentially decaying vertical profile was reasonable and was the best possible representation based on the available data, these data did not demonstrate that the profile was in fact exponential.
The characteristic height was derived using the in situ NO 2 concentration at DT ([NO 2 ] 10m ) and at CN ([NO 2 ] 445m ), as shown in Eq. ( 2), where z = 435 m.An increase in characteristic height indicates that NO 2 occupies a greater vertical fraction of the atmosphere.The NO 2 VCD in Eq. ( 1) is linearly dependent on H NO 2 in Eq. ( 2).Since z H NO 2 , the term [1 − exp(−z/H NO 2 )] in Eq. ( 1) essentially equals 1 at the tropopause, and as such Eq.( 1) can be simplified to The assumption of a vertically decaying NO 2 profile was founded on the following criteria: NO 2 originates at ground level, and undergoes photochemical conversion and dilution primarily through a first-order elimination process.A similar vertically decaying NO 2 profile shape was observed by Isaac et al. (1998) during summertime aircraft measurements conducted over Egbert, Ontario, between the elevations of 600 and 2900 m (approximately 80 km N of Toronto).These measurements demonstrated that the NO 2 concentration decreases with height by following a profile similar to the water vapour concentration, and has been predominantly associated with the dilution of ground-level emissions.Halla et al. (2011) derived composite tropospheric NO 2 VCDs by combining aircraft with ground-based in situ measurements in Ridgetown, Ontario (approximately 270 km W of Toronto).An exponentially decaying vertical NO 2 profile was assumed to determine the NO 2 vertical column present between aircraft and ground-level measurements.A significant difference was not observed when the composite NO 2 VCDs were compared to MAX-DOAS VCDs at this rural site.There are limited studies available for evaluation that derived NO 2 VCDs from stationary in situ measurements without assuming a well-mixed NO 2 vertical profile in the PBL.Previous research conducted by Schaub et al. (2006) demonstrated that in situ measurements from ground level up to 3.6 km generally obeyed a hyperbolic profile shape.This shape was applied to extract a NO 2 VCD assuming the NO 2 mixing ratio was 0.02 ppb at 8 km.Overall, the application of Eq. ( 2) in this paper aligns with the available research that does not rely on a CTM.Rearranging Eq. ( 2) to calculate [NO 2 ] z shown in Eq. ( 3) (where z is a specified altitude) demonstrates that the concentration of [NO 2 ] at altitude z is directly dependent on the characteristic height.
NO 2 VCD in situ were calculated on a daily basis using data averaged from 12:00 to 14:00 EST to coincide with the OMI overpass time, and hourly averaged NO 2 VCD in situ were compared to coincident MAX-DOAS measurements.Monthly averaged and diurnal profiles of H NO 2 and NO 2 VCD in situ were also derived.
Previous studies have investigated the positive bias associated with the CL detection of NO 2 due to the reduction of other oxidized nitrogen species (NO z = nitric acid: HNO 3 + peroxy acetyl nitrate: C 2 H 3 NO 5 + peroxyacyl nitrates: C x H y O 3 NO 2 + alkyl nitrates: RONO 2 + nitrous acid: HONO. . . ) by the CL monitor's molybdenum (Mo) catalyst, which is unspecific to the reduction of NO 2 to NO.The detection of NO z is dependent on the concentration of reactive nitrogen species at the monitoring site, the relative location of emission sources, meteorology, the conversion efficiency of NO z species by the heated Mo surface, and their respective line losses (Lamsal et al., 2008, and references therein).The positive bias associated with the detection of NO z exhibits seasonal and diurnal trends, reaching a maximum in the summer and during the afternoon when NO z constitutes a larger fraction of NO y (NO z + NO x ).These trends have been associated with the photochemical production of reactive NO z species (such as HNO 3 and PAN) alongside O 3 (Lee et al., 2011;Lamsal et al., 2008;Dunlea et al., 2007;Steinbacher et al., 2007).It was hypothesized that the bias would also vary with elevation, and thus that this positive bias should be accounted for.
The procedure reported by Boersma et al. (2009), which was based on measurements performed by Dunlea et al. (2007), was implemented to remove the influence of NO z from CL NO 2 measurements ([NO 2 ] CL ) after 10:00 EST, as shown in Eq. ( 4), where [NO 2 ] refers to the corrected measurement.
where Boersma et al. (2009) applied Eq. ( 4) to correct CL NO 2 measurements at eight cities in Israel (2006) during the OMI overpass time (13:45 LT).This reduced the CL NO 2 measurement by an average of 8 % (0.7 ppb) during 2006.This approach was validated by using the CHIMERE CTM to determine the NO 2 -to-NO z ratio over Europe, which was typically greater than 90 % for cities exhibiting NO 2 concentrations similar to those in Israel (0-25 ppb).Boersma et al. (2009) demonstrated that the NO 2 -to-NO z ratio using Eq. ( 4) for January, April, July, and October was 0.99, 0.98, 0.95, and 0.98, respectively, which was in excellent Table 3 summarizes the inclusion criteria applied to the in situ NO 2 data during the March 2006-March 2010 period, and corresponding sample size for measurements coinciding with the OMI overpass time.Since the in situ monitors were located approximately 2.3 km apart from one another, differences in the NO x emission characteristics at each measurement site may have influenced the observed NO 2 concentration at 10 m and 0.5 km.As shown in Table 1, the CN site experienced 24 h average weekday vehicle counts that were over 5 times greater than the DT site.Additionally, stack emissions from nearby sources may have impacted the CN measurements, resulting in a higher [NO]/[NO 2 ] than at ground level on some days.Therefore, an effort was made to filter data by removing hours that were strongly influenced by horizontal NO 2 gradients or did not follow a vertically decaying profile.The filtering process recognized that NO x emitted at ground level should consist predominantly of NO and that [NO]/[NO 2 ] should decrease with altitude due to the oxidation of NO.Thus the 17 % of days with a greater [NO]/[NO 2 ] at CN than at DT were excluded, since this suggested behaviour that was inconsistent with a vertically decaying profile, making calculation of the characteristic height impossible based on measurements at the two heights.Overall, 654 of 1426 days (46 %) were maintained for comparison with OMI data.The application of all criteria in Table 3 did not selectively bias the data by season, as during the warmer months of May-August, 50 % of data was retained, while during colder months of December to March, 56 % of data was retained.
Figure 2a and b provide the frequency distribution of the NO 2 volume mixing ratio (corrected for NO z interferencesee Sect.3.1 for results) at the DT and CN sites before and after applying selection criteria 6 and 7 in Table 3.At both measurement sites, the NO 2 volume mixing ratio retains a similar frequency distribution before and after applying these criteria.The NO 2 volume mixing ratio at DT is below 12 ppb for 54 % of the 1181 days considered (Table 3: criteria 1-4), and for 51 % of the 654 days meeting criteria 1-7 in Table 3.
Figure 2c shows the frequency distribution of the [NO 2 ] CN /[NO 2 ] DT volume mixing ratio (corrected for NO z interference) before and after applying selection criteria 6 and 7 in Table 3. Overall, [NO 2 ] CN /[NO 2 ] DT is between 0.15 and 0.75 for the majority (83 %) of the 1181 days considered (Table 3: criteria 1-4), and for all data meeting criteria 1-7 in Table 3 3, criteria 1-4 is 0.47, and the median for Table 3, criteria 1-7 is 0.43.
The results presented in Fig. 2 suggest that the selection criteria in Table 3 did not systematically bias the in situ data.However, both the assumption of an exponentially decaying NO 2 vertical profile and the selection criteria are not without inherent limitations.Deviations from the assumed exponentially decaying profile shape could introduce random errors in the retrieved characteristic height, and systematic errors could also occur.While an exponentially decaying NO 2 profile shape may be a reasonable assumption during the late morning and early afternoon, it cannot account for the presence of NO 2 inversion layers at elevations between the DT and CN sites.The selection criteria in Table 3 did eliminate over half of the in situ measurements, and therefore helped decrease the likelihood of misrepresenting the NO 2 vertical profile.
The use of a consistent NO 2 vertical profile shape within the boundary layer and free troposphere was evaluated by considering the fraction of the NO 2 VCD that exceeded 2 km, which for this study was recognized as an average maximum PBL height under turbulent mixing conditions based on previous measurements conducted in southwestern Ontario (Halla et al., 2011).The rationale for this test is that the NO 2 concentration may not have decayed at a consistent rate within and above the PBL.Overall, the fraction of the NO 2 VCD above 2 km ranged from 0 to 22 %, with a median of 3 %.

MAX-DOAS instrument and NO 2 VCD retrieval algorithm
Measurements were conducted using a commercially available Mini-MAX-DOAS instrument developed by the Institute of Environmental Physics at the University of Heidelberg in collaboration with Hoffmann Messtechnik GmbH (Bobrowski and Filsinger, 2005).The MAX-DOAS instrument is equipped with a commercial crossed Czerny-Turner spectrograph (OceanOptics USB2000) having a linear resolution of 2048 pixels that cover a wavelength range of 290-433 nm.The exposure time for each spectrum was calculated such that the CCD detector reached 80 % of its saturation level.An integrated average of 1000 spectra was used to create the final spectrum.During the hours of 09:00-16:00 EST, the temporal resolution of measurements varied from approximately 1 to 1.5 min, mainly as a result of solar zenith angle.Around solar noon, a sequence of six elevation angles took approximately 8 min to survey."Dark current" and "offset" spectra were taken each time the instrument was set up for monitoring and subtracted from recorded spectra during analysis.The spectrometer was stabilized at 5 • C for all measurements except those conducted during the winter, when the spectrometer was maintained at −5 • C to minimize the spectrometer's exposure to temperature instability.
MAX-DOAS spectra were processed with DOASIS (Kraus, 2003) and QDOAS (Fayt et al., 2011) software to retrieve NO 2 SCDs.Each spectrum was corrected for dark current and offset using DOASIS.The pixel-to-wavelength calibration was initially performed using a spectrum from a mercury lamp with known emission peaks.A calibration polynomial was generated in DOASIS based on the wavelength-to-pixel mapping of these peaks.QDOAS was then used to refine the wavelength calibration of each spectrum while simultaneously characterizing the resolution (expressed as the full width at half maximum -FWHM) of the CCD.Each spectrum was calibrated in QDOAS by performing a non-linear least-squares fit to align the structures of the measured spectrum to those of a high-resolution solar reference spectrum (0.01 nm) (Kurucz et al., 1984a) degraded to the resolution of the instrument.The resolution of the CCD (FWHM) was 0.6 nm at wavelengths greater than 380 nm.
NO 2 tropospheric SCDs were retrieved from calibrated spectra by applying the DOAS technique (Platt and Stutz, 2008;Hönninger et al., 2004;Platt, 1994) using QDOAS (Fayt et al., 2011).The SCD refers to the difference between the average concentration of a trace gas of interest (C) integrated along the average path length (L) traversed by photons prior to entering the spectrometer at elevation angle θ, (CL) θ or SCD θ , and the corresponding observation at an elevation angle of 90 • within a measurement cycle, (CL) 90 • or SCD 90 • , defined in Eqs. ( 5 Fraunhofer reference spectrum (FRS), the influence of stratospheric absorption was removed.
Broadband absorption was accounted for by including a DOAS polynomial of order 3.A first-order offset polynomial was also used in each fit to account for stray light in the spectrometer.The Ring effect (Grainger and Ring, 1962) was accounted for by fitting measured spectra to a synthetic Ring spectrum generated in QDOAS by convolving a highresolution solar spectrum (0.05 nm) (Kurucz et al., 1984b) with rotational Raman scattering cross sections (Fayt and van Roozendael, 2001).
Fitting errors associated with the MAX-DOAS SCD are indicative of the systematic and random errors associated with the QDOAS retrieval.Systematic errors include incorrect spectral calibration, dark current and offset correction, and slit function characterization, as well as errors in the reference spectra, while random errors may be due to the presence of unknown spectral absorbance structures, and a low signal-to-noise ratio (Vlemmix et al., 2010;Fraser et al., 2009).The QDOAS fitting error generally increased as the magnitude of the SCD decreased.A retrieved NO 2 SCD of 1 × 10 17 molec cm −2 had a relative fitting error of ≤ ±2 %, while a NO 2 SCD of 2 × 10 15 molec cm −2 had a relative fitting error of ≤ ±20 %.
Systematic errors associated with the uncertainty and temperature dependence of the NO 2 absorbance cross section have been addressed in previous studies.The measurement uncertainty associated with the NO 2 (294 K) absorption cross section is ≤ 10 % (Vandaele et al., 1998).Vlemmix et al. (2011) demonstrated that NO 2 SCDs measured using a NO 2 absorbance cross section at 295 K should be corrected by a factor of 0.92 for an effective atmospheric temperature of 283 K (determined by considering the vertical temperature profile and the vertical distribution of NO 2 in the lower troposphere).Irie et al. (2012) reported an 11 % decrease in the NO 2 VCD by scaling it with a vertical temperature profile that decreased to 260 K at 2 km.The majority of the MAX-DOAS measurements in the current study were collected in the summer, where the characteristic height of NO 2 was typically only 500m.Any vertical variation in the NO 2 absorbance cross section within this region would have been small, as the relevant vertical variation of temperature was presumably < 5 • C. Thus this temperature sensitivity was not accounted for since, based on previous sensitivity studies, it would only have caused, on average, an error of a few percent.Certainly, some measurements were collected in March, when the average temperature was closer to 273K, and these MAX-DOAS VCDs may be overestimated by up to 10 %.
The total uncertainty associated with the NO 2 SCD was calculated as the root sum of squares of the QDOAS fit error, absorbance cross-section accuracy, and uncertainty associated with the temperature dependence of the differential cross section.The total relative uncertainty of the NO 2 SCD varied from 15 to 25 %.Hourly averaged NO 2 SCDs were calculated to convert MAX-DOAS measurements to the same timescale as the available in situ measurements.

AMF
Previous studies have compared NO 2 VCDs retrieved using radiative transfer modelling to those calculated using the geometric AMF (Wagner et al., 2010(Wagner et al., , 2011;;Halla et al., 2011;Shaiganfar et al., 2011;Vlemmix et al., 2010).Wagner et al. (2011) converted NO 2 SCDs measured at an elevation angle of 18 • in south, north, and west viewing directions to NO 2 VCDs using the geometric approximation (VCD geo ) and the radiative transfer model McArtim (VCD RTM ; λ = 360 nm).The VCD geo /VCD RTM slopes were 0.88 (R 2 = 0.88), 0.96 (R 2 = 0.96), and 0.92 (R 2 = 0.86), respectively.Although no relationship between the geometric NO 2 VCD's negative bias and aerosol optical depth (AOD) was found, this bias exhibited a systematic dependency on the vertical distribution of NO 2 .This systematic dependency is anticipated since the geometric AMF assumes that the tropospheric NO 2 column is contained below the last scattering altitude of photons.Wagner et al. (2011) predicted that for a NO 2 column below 1000 m, the error of the geometric NO 2 VCD is typically within 20 %.Shaiganfar et al. (2011) demonstrated that geometric NO 2 VCDs at elevation angles of 22 and 30 • differed from those determined using McArtim by ±20 % when assuming a 500 m vertical NO 2 box profile (AOD < 1).The geometric NO 2 VCD also exhibited a systematic dependency on relative azimuth angle (difference between the solar azimuth angle and MAX-DOAS viewing direction): the geometric VCD underestimated the true VCD at low relative azimuth angles (∼ 0 • ) and overestimated the VCD at higher relative azimuth angles (≥ 90 • ).Vlemmix et al. (2010) demonstrated that, for relative azimuth angles greater than 30 • , the difference between the geometric VCD at an elevation angle of 30 • and McArtim VCD (λ = 428 nm) reaches a maximum of 25 %, and this offset is strongly a function of solar position relative to the MAX-DOAS instrument as well as AOD.
The above studies did not consider the use of validation criteria in conjunction with the geometric AMF to ensure the NO 2 vertical column was contained below the last scattering altitude of photons into the MAX-DOAS detector.Halla et al. (2011) compared geometrically approximated NO 2 VCDs that met the criteria VCD 10 • = VCD 30 • (±15 %) with those determined using the radiative transfer model McArtim (λ = 413 nm, assuming a vertical NO 2 box profile), and demonstrated geometric VCDs underestimated modelled VCDs by 8-12 %.Wagner et al. (2010) suggest that the strong agreement observed between geometrically derived NO 2 VCDs retrieved at multiple elevation angles (22 and 40 • ) renders the uncertainty involved in using geometrically approximated VCDs associated with aerosol loading < 15 %.Since a similar criterion for employing the geometric AMF was used in this study (NO 2 VCD 10 • = VCD 20 • (or 30 • ) ± 15 %), it is estimated that the uncertainty associated with the geometric NO 2 AMF is ≤ 15 %.The total relative uncertainty associated with the NO 2 VCD was calculated as the root sum of squares of the random and systematic uncertainties associated with the NO 2 SCD and the geometric NO 2 AMF.The total relative uncertainty of the MAX-DOAS NO 2 VCD retrieval ranged from 20 to 29 %.

OMI tropospheric NO 2 columns
Dutch OMI NO 2 (DOMINO version 2.0, collection 3 level 1b) tropospheric columns were obtained from the Tropospheric Emissions Monitoring Internet Service (TEMIS) for the period of March 2006-March 2010.The DOMINO retrieval algorithm has been described in detail by Boersma et al. (2011), and is briefly reviewed here.The DOMINO product (DP) employs the DOAS algorithm (Platt, 1994) to retrieve a NO 2 slant column density (SCD) from a measured spectrum of backscattered solar radiation for the 405-465 nm wavelength range (Bucsela et al., 2006).The stratospheric NO 2 SCD is determined by assimilating the total NO 2 SCD into the TM4 global chemistry and transport model (CTM) (Dirksen et al., 2011).The stratospheric NO 2 SCD is then subtracted from the total NO 2 SCD to determine the tropospheric SCD.The AMF is obtained from the DAK radiative transfer model version 3.0 (Stammes, 2001), and relies on the a priori tropospheric NO 2 profile simulated in TM4, in addition to parameters including cloud radiance fraction, cloud pressure, surface albedo, and satellite viewing geometry.The tropospheric AMF permits the conversion of the tropospheric NO 2 SCD to the VCD (Boersma et al., 2007).The estimated uncertainty of the OMI (version 2.0) tropospheric NO 2 SCD associated with spectral fitting is 0.7 × 10 15 molec cm 2 , and when the uncertainty associated with the AMF is considered, this value increases to 25 % for a NO 2 tropospheric VCD measurement of 1.0 × 10 15 molec cm −2 (Boersma et al., 2011).
The coincidence criteria used to pair OMI tropospheric NO 2 columns with independent measurements was informed by previous studies, which have constrained the sample size according to OMI cloud radiance fraction, pixel size, proximity of pixel centre to independent monitoring site, and overlap of the independent data's averaging interval with OMI overpass time (Lee et al., 2011;Hains et al., 2010;Kramer et al., 2008;Irie et al., 2008).Table 4 summarizes the criteria and coincident hours of in situ data that were used in this study.Results are presented for the 56 days of data that obeyed all criteria.The minimum, median, and maximum OMI NO 2 tropospheric VCD during these days was 3.00 × 10 14 molec cm −2 , 6.95 × 10 15 molec cm −2 , and 2.21 × 10 16 molec cm −2 , respectively.Seventy-two percent of data exhibited a tropospheric NO 2 VCD below 1.00 × 10 16 molec cm −2 .
Overall, of the 56 OMI overpasses with a cloud radiance fraction ≤ 0.3 and pixel centre within 0.1 • × 0.1 • of DT, 10 overpasses coincided with MAX-DOAS NO 2 VCD measurements within ±1 h, and 1 overpass coincided within

Impact of NO z interference on chemiluminescence measurement of NO 2
Table 5 illustrates the impact of removing the NO z interference from the average CL measurement of NO 2 during the hours of 12:00-14:00 EST.This correction reduced the CL NO 2 concentration at DT by 8 ± 1 % and at CN by 12 ± 1 %, suggesting, but not proving, that the NO z interference associated with the CL detection of NO 2 did vary with elevation.Further, the larger NO z interference correction at the higher elevation suggested that the correction is greater for aged air masses, consistent with observations by Boersma et al. (2009) and Steinbacher et al. (2007).
The NO z interference varied seasonally, and had a greater impact on CL NO 2 measurements in the summer versus in the winter.During the months of December-March (n = 222; T = 0.2 ± 0.9 • C), the NO z correction reduced the CL NO 2 concentration at DT by 4 ± 1 % and at CN by 5 ± 1 %, while during the months of May-August (n = 197; T = 22.4 ± 0.7 • C), the NO z correction reduced the CL NO 2 concentration at DT by 11 ± 1 % and at CN by 21 ± 1 %.The larger NO z interference correction at the higher elevation during the summer suggests that a greater concentration of reactive nitrogen species is present, as Dunlea et al. (2007) demonstrated that the positive and linear relationship with [NO z ] and [O 3 ] is due to the photochemical production of reactive nitrogen species alongside O 3 .
The average absolute decrease in the NO 2 concentration after accounting for the NO z interference was less than 1 ppb at both DT and CN, which is similar to observations in rural southwestern Ontario by Lee et al. (2011), who demonstrated that the median difference between the CL and true NO 2 was only 0.9 ppb.Overall, the NO 2 characteristic height decreased by an average of 30 m, and the NO 2 VCD in situ decreased by less than 1 × 10 15 molec cm −2 .

Seasonal and diurnal variation of in situ NO 2 concentration, VCD, and characteristic height
The seasonal variation of the in situ NO 2 concentration at DT and CN is shown in Fig. 3a for data meeting criteria 1-4 in Table 3, and in Fig. 3b for data meeting all criteria in Table 3.The difference between the monthly averaged NO 2 concentrations shown in Fig. 3a versus b is not statistically significant (p value = 0.05).These results demonstrate that the monthly averaged NO 2 concentration at CN and DT determined by applying selection criteria 6 and 7 in Table 3 were representative of the concentrations determined using the original sample size of 1181 days.The monthly averaged NO 2 concentration at both monitoring sites shown in Fig. 3b, and the monthly averaged tropospheric NO 2 VCD shown in Fig. 3c, reached a maximum during autumn and winter (November-March), and a minimum during summer (June-August).This seasonal trend is attributed to increased photolysis during the summer.The photolysis of NO 2 decreases the concentration of NO 2 relative to NO (Seinfeld and Pandis, 2006).The increased photolysis of hydroxyl radical (OH) precursor species (such as O 3 ) also decreases the summertime NO 2 concentration via its reaction with OH to yield nitric acid (HNO 3 ) (Seinfeld and Pandis, 2006;Jacob, 1999).Furthermore, an increase in anthropogenic NO x emissions during winter months due to residential heating may have also increased the wintertime NO 2 VCD.
The seasonal variation of the in situ-derived NO 2 characteristic height is shown in Fig. 3d.These characteristic heights for NO 2 are below the expected midday PBL height, which may reach 2 km in the summer.Overall, the NO 2 characteristic height varies insignificantly within a given season.However statistically significant differences are seen when comparing winter months to spring and summer months.January and February show the greatest NO 2 characteristic heights of 0.71 ± 0.08 km and 0.78 ± 0.08 km (temperature = −2.3± 1.8 • C and −3.9 ± 1.4 • C, solar zenith angle < 55 • ), while minimum values are observed during the spring and summer (i.e. the characteristic height during July is 0.51 ± 0.08 km, temperature = 24.7 ± 0.9 • C, and solar zenith angle < 33 • ).This seasonal pattern for characteristic height is the opposite of that for PBL height, where maximum values occur in the summer.The increased NO 2 characteristic height during winter months indicates that Figure 4a provides the diurnal variation of the NO 2 concentration measured at DT and CN during the autumn and winter months of December-March (average temperature = −0.6 ± 0.2 • C), while Fig. 4b shows this diurnal variation during the spring and summer months of May-August (average temperature = 21.1 ± 0.2 • C).In both figures, the NO 2 concentration at DT exhibits a peak during the morning hours that corresponds to the evolution of the rush hour traffic period from 07:00 to 09:00 EST, and decreases during the afternoon.The loss of NO 2 to HNO 3 reaches a maximum during the afternoon due to the increased concentration of OH (Lamsal et al., 2010;Boersma et al., 2008aBoersma et al., , 2009)).The decrease in NO 2 concentration also occurs as a result of increased wind speed during afternoon hours, resulting in increased mixing, dilution, and advection of vehicular emissions.The average O 3 concentration (not shown), reflective of the OH concentration, reached a maximum during the afternoon hours.The average O 3 concentration at DT during the hours of 13:00-15:00 EST in December-March was 27.9 ± 1.2 ppb, while in May-August it was 41.7 ± 1.8 ppb.These hours generally corresponded to when the NO 2 concentration at DT reached a minimum.
During the months of December-March at 17:00 EST, the NO 2 concentration at DT began to increase in conjunction with the evening rush hour period.During the months of May-August, the NO 2 concentration remained relatively constant from 13:00 to 19:00 EST, due to higher wind speed, and in turn, dilution of emissions during the evening rush hour period.The NO 2 concentration at DT was an average of 25 ± 1 % (5.6 ± 1 ppb) lower in the morning, and 52 ± 2 % (4.0 ± 0.2 ppb) lower in the afternoon/evening during the warmer months than during the colder months.
The NO 2 concentration at CN in Fig. 4a and b follows a similar profile as the NO 2 concentration at DT.In both figures, the morning rush hour peak is not as pronounced as at DT, likely due to the dilution of vehicle emissions with increasing altitude.The NO 2 concentration at CN also reaches a minimum during the afternoon when the O 3 , and presumably OH, concentration at CN reaches a maximum.The average O 3 concentration at CN during the hours of 13:00-15:00 EST in December-March was 33.2 ± 1.1 ppb (19 % greater than at DT), while in May-August it was 47.8 ± 4.7 ppb (15 % greater than at DT).The NO 2 concentration at CN was an average of 27 ± 8 % (2.4 ± 0.2 ppb) lower in the morning, and 52 ± 2 % (4.0 ± 0.1 ppb) lower in the afternoon/evening during the warmer months than during the colder months.
Figure 4c  The NO 2 characteristic height during the months of December-March exhibited a diurnal trend that is similar to the evolution of the PBL, which increases during the morning hours and peaks in the afternoon (Jacob, 1999).The NO 2 characteristic height (and [NO 2 ] CN /[NO 2 ] DT ) is an average of 0.5 ± 0.3 km (0.39 ± 0.02) during the hours of 07:00-10:00 EST, and reaches an average of 0.65 ± 0.01 km (0.47 ± 0.01) during the hours of 11:00-17:00 EST.
During the months of May-August, a statistically significant difference for both the hourly averaged H NO 2 and [NO 2 ] CN /[NO 2 ] DT was not witnessed between the hours of 08:00 and 19:00 EST.The average characteristic height during these hours was 0.56 ± 0.01 km -14 % lower than the average maximum characteristic height in December-March, while the average [NO 2 ] CN /[NO 2 ] DT during these hours was 0.41 ± 0.01 -13 % lower than the average maximum ratio in December-March.These results demonstrate that the vertical distribution of NO 2 within 0.5 km during the period of spring-summer does not follow the typical diurnal profile of the convective PBL, which is typically higher during warmer months than colder months.
Previous research using radiosonde measurements has demonstrated that the height of the PBL in southern Ontario (43.682 • N, 79.612 • W) is an average of 1.5 km during the summer.An increased PBL height during the summer is directly related to the increase in solar radiation and surface heat flux (Ning and Yap, 1986).The NO 2 characteristic heights shown in Fig. 4d are over 60 % lower than these radiosonde measurements of the PBL height, which suggest they cannot be considered representative of the diurnal evolution of the PBL.Overall, Fig. 4d shows that assuming a well-mixed NO 2 concentration within the PBL will lead to an overestimate of the tropospheric NO 2 VCD.
Figure 4e provides the diurnal variation of the in situ NO 2 VCD during the months of December-March, while Fig. 4f shows this diurnal variation during the months of May-August.The diurnal profile shown in both figures closely replicate the diurnal profile of the NO 2 concentration at DT (R = 0.87 and R = 0.93, respectively).During the afternoon and evening hours, the NO 2 VCD is at least 2 times greater during the months of December-March versus May-August.This is due to both the increased concentration of NO 2 DT and the increased [NO 2 ] CN /[NO 2 ] DT during the colder months.Since the diurnal trends presented in Fig. 4e and f rely on the assumption of an exponentially decaying vertical NO 2 profile, the presence of NO 2 inversion layers at elevations between the DT and CN sites could result in an underestimation of the VCD at times, especially during the winter morning hours.

Comparison between in situ and remotely sensed NO 2 VCDs
Figure 5 shows the linear regression of the OMI versus in situ tropospheric NO 2 VCD.Linear regression was performed using data meeting all OMI overpass criteria in Table 4.
The OMI tropospheric NO 2 VCD exhibits a negative bias of 58 % when compared to the in situ NO 2 VCD and good correlation with a Pearson's R value of 0.81.These results are consistent with the spatial footprint represented by these two types of measurements.The in situ monitors measure a point-location "near-roadside" NO 2 column, while the OMI NO 2 VCD reflects an average over the spatially heterogeneous Greater Toronto Area, consisting of roads, residential neighbourhoods, and Lake Ontario.
Figure 6 shows the linear regression results of the MAX-DOAS versus in situ tropospheric NO 2 VCD, and although good agreement is seen (Pearson's R = 0.72), the MAX-DOAS NO 2 VCD is only 42 % of the corresponding value determined using the in situ monitors.The difference between the MAX-DOAS versus in situ-derived NO 2 VCD suggests that differences in the geographic footprint surveyed by each instrument impacted the results.As summarized in Table 2, much of the MAX-DOAS path length was over Lake Ontario during campaigns A, D, and E; was facing east for campaign C; and was located 65 m a.g.l. in campaigns C and D. During all campaigns, the MAX-DOAS instrument's path length was also influenced by the NO 2 concentration in residential areas in the downtown core.
The spatial heterogeneity of the NO 2 concentration across Toronto was evaluated by Jerrett et al. (2007Jerrett et al. ( , 2009)).Passive sampling measurements of NO 2 concentration at 143 sites in the early autumn of 2002 and spring of 2004 were used in conjunction with land-use regression modelling to derive a NO 2 surface concentration map.Despite changes in absolute NO 2 concentration between seasons, overall spatial patterns remained similar: NO 2 exhibited higher concentrations in the west (> 20 ppb), and lower concentrations in the east (10-15 ppb).The downtown core and areas near major highways also exhibited high NO 2 concentrations (> 30 ppb).The MAX-DOAS path length over mainland Toronto consists of a combination of urban background and roadside NO 2 concentrations, and the MAX-DOAS NO 2 VCD is dependent on these gradients.
Figure 7 shows the linear regression results of OMI versus MAX-DOAS tropospheric NO 2 VCDs.A good agreement is seen between the measurement techniques, which have a Pearson's R value of 0.80.The slope presented in Fig. 7 derived using linear regression is 0.93 ± 0.11, or if weighted orthogonal distance regression is applied, then this slope is 0.74 ± 0.16.These slopes are in agreement when their respective errors are considered, and in both cases the slope is not significantly different from 1.Of the 11 available comparison days, 8 OMI and MAX-DOAS measurements agree when the respective uncertainties of each measurement are considered.The results presented in Fig. 7 suggest that MAX-DOAS and OMI measurements both "look at" similar mixes of the spatial heterogeneity in NO 2 concentrations created by the unpolluted regions over Lake Ontario and polluted regions over the city.Thus this finding is indirectly consistent with the majority of previous studies which suggest that OMI exhibits a positive bias over rural (unpolluted) MAX-DOAS measurement sites, and a negative bias over urban (polluted) MAX-DOAS measurement sites (Shaiganfar et al., 2011;Halla et al., 2011;Wagner et al., 2010;Kramer et al., 2008;Brinksma et al., 2008;Celarier et al., 2008).In  3 during the hours of 07:00-19:00 EST for hourly averaged data having H NO 2 within 10th to 90th percentile.the heterogeneous polluted and unpolluted environment of the current study, the two remote sensing methods yield similar vertical columns.More importantly, when these groundand satellite-based measurement methods "look at" the same mixture of environments, then they agree, providing reassurance in the validity of the measurements produced by each method.

Conclusions
In situ measurements of NO 2 in an urban environment were compared with remotely sensed satellite, and multi-axis differential optical absorption spectroscopy (MAX-DOAS), tropospheric NO 2 vertical column densities (VCDs).The chemiluminescence measurements were first corrected for the influence of NO z , which reduced the NO 2 concentrations near ground level and at 445 m by an average of 8 ± 1 % and 12 ± 1 %, respectively.The absolute decrease in the  chemiluminescence NO 2 measurements as a result of this correction was less than 1 ppb.
The monthly averaged ratio of the NO 2 concentration at 0.5 to 0.01 km varied seasonally, and exhibited a negative linear dependence on the monthly average temperature, with Pearson's R = 0.83.During the coldest month, February, this ratio was 0.52 ± 0.04, while during the warmest month, July, this ratio was 0.34 ± 0.04, thereby demonstrating that NO 2 is not well mixed in the PBL.
The NO 2 characteristic height during the months of December-March exhibited a diurnal trend that was similar but much weaker than the evolution of the PBL, while during the months of May-August, a statistically significant difference for both the hourly averaged H NO 2 and [NO 2 ] CN /[NO 2 ] DT was not witnessed between the hours of 08:00 and 19:00 EST.The average characteristic height during these hours was 0.56 ± 0.01 km -14 % lower than the average maximum characteristic height in December-March.These results demonstrate that the vertical distribution of NO 2 within 0.5 km during the period of spring-summer does not follow the typical diurnal profile of the convective PBL, which is generally higher during warmer months than colder months.Although NO 2 profile inversions at elevations between the DT and CN sensors could not be detected using the method employed, overall, the presented results suggest that assuming a well-mixed NO 2 concentration within the PBL will lead to an overestimate of the NO 2 VCD.
Strong correlation was observed between the remotely sensed and in situ NO 2 VCDs (Pearson's R value ranging from 0.72 to 0.81).However, the in situ VCDs were 52 to 58 % greater than the remotely sensed columns due to horizontal spatial heterogeneity, and possibly random and systematic errors associated with each VCD measurement technique.Notably, the in situ VCDs were based on an assumed exponential vertical profile.The in situ NO 2 VCD were representative of a local NO 2 column in a polluted near-road environment, while the remotely sensed (MAX-DOAS and OMI) VCDs were representative of a spatially heterogeneous region, which included the downtown city core, residential neighbourhoods, and Lake Ontario.Overall, the reasonable agreement between the VCD values determined by the three distinct methods increased confidence in the validity of the values provided by each of the measurement techniques.

Fig. 2 .
Fig. 2. (a) Frequency distribution of NO 2 DT volume mixing ratio during OMI overpass time.Light-grey bars refer to data meeting criteria 1-4 in Table 3 (Total Days = 1181) and dark-grey bars refer to data meeting all criteria in Table 3 (Total Days = 654).(b) Same as (a) but for NO 2 CN.(c) Same as (a) but for ratio of [NO 2 ] CN to [NO 2 ] DT .
) and (6).Since the zenith spectrum within each measurement sequence was utilized as the www.atmos-meas-tech.net/6/2907/2013/D. Mendolia et al.: Comparison of tropospheric NO 2 vertical columns in an urban environment

Fig. 3 .
Fig. 3. (a) Monthly averaged in situ NO 2 concentration at DT and CN during the hours of 12:00-14:00 EST for data meeting criteria 1-4 in Table 3 (n = 1181) and (b) for data meeting all criteria in Table 3 (n = 654).(c) Same as (b) but for in situ NO 2 VCD.(d) Same as (b) but for in situ NO 2 characteristic height.
shows the diurnal variation of the NO 2 characteristic height, and [NO 2 ] CN /[NO 2 ] DT during the months of December-March, while Fig. 4d shows this diurnal variation during the months of May-August.Both H NO 2 and [NO 2 ] CN /[NO 2 ] DT follow a similar profile since [NO 2 ] CN /[NO 2 ] DT was used to derive H NO 2 .As shown in Eq. (2), an increase in [NO 2 ] CN /[NO 2 ] DT results in an increased H NO 2 .

Fig. 4 .
Fig. 4. (a) Diurnal profile of [NO 2 ] at DT and CN during the months of December-March and (b) May-August.(c) Diurnal profile of H NO 2 and [NO 2 ] CN /[NO 2 ] DT during the months of December-March and (d) May-August.(e) Diurnal profile of NO 2 VCD during the months of December-March and (f) May-August.Plots were derived by applying all criteria in Table3during the hours of 07:00-19:00 EST for hourly averaged data having H NO 2 within 10th to 90th percentile.

Fig. 5 .
Fig. 5. Linear regression of OMI versus in situ tropospheric NO 2 VCD for data meeting all OMI overpass criteria in Table4.

Table 1 .
Location of in situ monitoring sites and pollutants of interest.N, 79.387 • W) * 24 h average weekday vehicle counts include east and west-bound traffic

Table 2 .
Monitoring period and associated observation geometry for MAX-DOAS measurements.The horizontal distance from the MAX-DOAS instrument (at the azimuth angle surveyed) to Lake Ontario is provided in the "Mainland Toronto" column.
* Campaign B surveyed 1.70 km over Lake Ontario prior to reaching mainland Toronto.

Table 3 .
In situ data selection criteria for the calculation of NO 2 VCD in situ .Day and % of valid data refer to in situ measurements averaged during the hours of 12:00-14:00 EST to correspond with the OMI overpass time.NO 2 data below 10th percentile 654 46 and above 90th percentile); maintain 0.23 km ≤ H NO 2 ≤ 1.33 km agreement with CHIMERE NO 2 -to-NO z ratios of 0.98, 0.96, 0.92, and 0.97.

Table 4 .
Inclusion criteria for OMI measurements.

Table 5 .
Average change in NO 2 concentration at DT and CN, characteristic height (H NO 2 ), and vertical column density (VCD) during the hours of 12:00-14:00 EST as a result of NO z interference correction.