Combining Active and Passive Airborne Remote Sensing to Quantify NO 2 and O x Production near Bakersfield, CA

Authors RV, CS and RMH designed research; all authors performed research. Authors SB, HO, AB and CS analyzed data; authors SB and RV interpreted data and wrote the paper. All authors provided comments

is highest over the city (1.35 kg hr -1 km -2 NO 2 ), and about 30 times lower at background sites (0.04 kg hr -1 km -2 NO 2 ).NO x emissions as represented in the CARB 2010 emission inventory agree well with our measurements over Bakersfield city (within 30%).However, emissions upwind of the city are significantly underestimated.The O x ' production is less variable, found ubiquitous, and accounts for 7.4 kg hr -1 km -2 O x ' at background sites.Interestingly, the maximum of 17.1 kg hr -1 km -2 O x ' production was observed upwind of the city.A plausible explanation for the efficient O x ' production upwind of Bakersfield, CA are favorable volatile organic compound (VOC) to NO x ratios for O x ' production, that are

INTRODUCTION
Ozone (O 3 ) and nitrogen oxides (NO x = NO + NO 2 ) are trace gases that are important components of air pollution.Health concerns of O 3 and NO 2 are related to respiratory illnesses such as chest pain, reduced lung function, asthma, emphysema whereas environmental concerns include reduced vegetation growth and acid rain.Due to these concerns both trace gases are recognized as air pollutants by air quality regulating agencies around the world, and are regulated by air quality standards and guidelines.The World Health Organization (WHO) air quality guideline recommends the standard to be set at 100 μg m -3 (~51 ppb) for O 3 (8 hour mean) and 40 μg m -3 (~21 ppb) for NO 2 (annual mean) [1].The National Ambient Air Quality Standard (NAAQS) set by U.S. Environmental Protection Agency are 75 ppb for O 3 (8 hour maximum) and 53 ppb for NO 2 (annual mean) [2].Similarly, the current air quality standard for the European Union are 120 μg m -3 (~61 ppb) for O 3 (8 hour maximum) and 40 μg m -3 (~21 ppb) for NO 2 (annual mean) [3].Further, O 3 is a greenhouse gas that is relevant to climate discussions [4].The lifecycles of O 3 and NO x are intimately coupled, because NO 2 photolysis by sunlight drives photochemical O 3 production, while emissions of NO destroy O 3 to form NO 2 .The sum of O 3 and NO 2 is called O x , and is a conserved quantity as it implicitly accounts for the destruction of O 3 by NO (O 3 titration).Excess O x is formed from the oxidation of volatile organic compounds (VOCs) in the presence of NO x [5][6][7].Fig. 1 shows a schematic diagram of photochemical O 3 production and evolution of NO, NO 2 , O 3 and O x concentrations upwind (I), within city limits (II) and downwind (III and IV) of an urban area.Different chemistry in these regions results in the characteristic spatial patterns in NO x -O 3 distributions depicted in Fig. 1, which are: (1) Background O 3 present in the upwind region (I).(2) Emission of NO x in the city limits (II), which leads to (3) O 3 removal via reaction with NO to produce NO 2 (titration reaction).( 4) Photochemical production of O 3 from VOC/NO x chemical cycles, which dominates downwind of the city center (III) and results in O 3 concentrations to accumulate.Further (5) the O 3 concentration does no longer accumulate in some distance downwind (IV), when NO x has been oxidized to NO y .NO y is efficiently deposited or lost to aerosols resulting in insufficient NO x to drive VOC-NO x chemical cycles (NO x limited region).Ultimately O 3 removal by photolysis and dry deposition leads to a slowly decreasing O 3 concentration here [6][7][8].
Due to their importance for air quality and human health, NO 2 and O 3 plumes from point sources and urban areas have been extensively studied.Previous studies have estimated NO 2 emission rates from point sources like power plants [9], urban areas [10][11][12][13], O 3 production rates in urban plumes [14,15], the amount of O 3 transported from urban areas and its impact on regional background O 3 [15], and the relationship between O 3 , NO and NO 2 as function of NO x in urban areas [16].However, despite decades of research, models that predict O 3 formations have not been constrained by observations at the scale of cities and immediately downwind of cities.The comparison at the local scale is important, because of uncertain and changing emission of VOCs [17], NO x [18], complicated transport [19][20][21] over cities and downwind of cities, and also uncertainties in non-linear chemistry that couples VOCs, NO x and O 3 .Such chemistry is heavily parameterized in current atmospheric models used to predict O 3 .The net O 3 production by VOC oxidation is related to the conversion of NO to NO 2 by organic peroxy-and hydro peroxy radicals that are formed during the airborne oxidation of VOCs by atmospheric oxidants like OH, NO 3 , O 3 , and Cl radicals [22].Under high NO x conditions, the rate of O 3 production is limited by the availability of VOCs, while availability of NO x controls the rate of O 3 production under low NO x conditions [22][23][24][25].For example, the testing of detailed chemical mechanisms of VOC oxidation using simulation chamber data [26,27], and field observations [28][29][30][31] often predict lower O 3 formation rates than that are actually being observed.The uncertainty in the chemistry of O 3 formation can be of similar relevance as uncertainties in emissions, and transport [30].Further, transport of O x across city, state and international borders causes possible non-attainment of O 3 levels at sites downwind [15,20,32] and the changing boundary conditions complicates enforcement of regulations.
Over the course of the last decade, emission control policies aimed at reducing ambient O 3 levels have resulted into NO x reductions in North America and Europe [18,[33][34][35][36][37][38].NO x sources in the troposphere are primarily related to anthropogenic emissions from on-road motor vehicles and power plants.With more than half of the world population now living in urban areas, cities have developed into hotspots for NO x sources [18,33,34] and provide opportunities for NO x reductions that are relevant on the global scale.This trend towards urbanization on global scales is unique in the history of mankind, and has the potential to change the planet.There is an increasing need for the development of analytical approaches that are effective at quantifying emissions of NO x , provide experimental constraints to O x production rates, and transport in order to refine atmospheric models that are used to manage air resources.
The primary objective of this study is to demonstrate the potential and feasibility of integrated use of passive and active remote sensing instruments and column observations to estimate the rate of NO x emissions, and investigate the O x production from an urban area.We use the mass conservation approach to estimate source strength for NO 2 , and O x from an urban area.Recently, ground based mobile differential optical absorption spectroscopy (DOAS) measurements have been used to estimate NO x emissions from urban areas using this approach [10,11,13].A similar approach has also been used to probe NO x emission from megacities using satellites [12].We have made simultaneous measurements of NO 2 vertical columns, O 3 and wind profiles for the first time from a research aircraft.The data set provides an opportunity to estimate production of individual species and investigate the conserved quantity, O x , which could be significantly impacted by O 3 titration in NO x source areas such as city centers.As a case study, data from a research flight on June 15, 2010 over Bakersfield, California is presented.

METHODOLOGY
We use a mass conservation approach to estimate the emission and production source strength of NO 2 and O x .Neglecting the molecular diffusivity term in the mass conservation equation, the NO 2 and O x source strengths within a given volume can be estimated from their time rate of change within the volume and the horizontal flux divergence across the boundaries enclosing the volume.We have conducted measurements of vertical columns of NO 2 , O 3 , and wind profiles aboard a research aircraft that flew box patterns over and near an urban area.Fig. 1 shows a conceptual schematic illustrating our approach for measuring NO 2 and O x production rates.

METHODOLOGY
We use a mass conservation approach to estimate the emission and production source strength of NO 2 and O x .Neglecting the molecular diffusivity term in the mass conservation equation, the NO 2 and O x source strengths within a given volume can be estimated from their time rate of change within the volume and the horizontal flux divergence across the boundaries enclosing the volume.We have conducted measurements of vertical columns of NO 2 , O 3 , and wind profiles aboard a research aircraft that flew box patterns over and near an urban area.Fig. 1 shows a conceptual schematic illustrating our approach for measuring NO 2 and O x production rates.

METHODOLOGY
We use a mass conservation approach to estimate the emission and production source strength of NO 2 and O x .Neglecting the molecular diffusivity term in the mass conservation equation, the NO 2 and O x source strengths within a given volume can be estimated from their time rate of change within the volume and the horizontal flux divergence across the boundaries enclosing the volume.We have conducted measurements of vertical columns of NO 2 , O 3 , and wind profiles aboard a research aircraft that flew box patterns over and near an urban area.Fig. 1 shows a conceptual schematic illustrating our approach for measuring NO 2 and O x production rates.[39].One of the foci of this deployment was to constrain the emission and production of NO 2 and O 3 upwind, within and downwind of urban areas.

AMAX-DOAS
The CU AMAX-DOAS instrument [41,42] uses scattered sunlight as the light source (passive remote sensing).The scattered sunlight spectra are analyzed for the presence of absorbers like NO 2 , glyoxal (CHOCHO), formaldehyde (HCHO) and oxygen dimer (O 4 ) among others using the DOAS method [43].The instrument and its performance during CalNex and CARES field campaigns are described in detail in Baidar et al. [42].Briefly, a telescope pylon is mounted on the outside of the window plate of the aircraft and includes a rotatable prism to collect scattered photons from different elevation angle (EA) i.e. angle relative to the horizon.Spectra collected from different EA contain information from different layers in the atmosphere and hence can be used to obtain information about vertical distribution of trace gases.The collected photons are transferred to a spectrometer / charge coupled device (CCD) detector system via optical glass fiber bundle.Here we will only present data from nadir viewing geometry from the flight over Bakersfield.
The measured spectra were analyzed, for NO 2 in a wavelength range from 433 to 460 nm, against a fixed zenith reference spectrum recorded during the same flight in a clean environment and flying at relatively high altitudes (3 -5 km; 3.5 km for this flight).Zenith spectra were recorded frequently, and are used to correct for stratospheric NO 2 contributions and NO 2 above the aircraft.The nadir NO 2 differential slant column densities (dSCDs) are observed below the plane and correspond to the average integrated difference in concentration of the absorber along the light path with respect to the reference.Since most of the NO 2 sources in an urban environment are located close to the surface, the retrieved nadir dSCD was considered to be the boundary layer slant column (dSCD bl ).NO 2 nadir measurements were performed every 20-25 s and hence NO 2 data points are available every ~1.5 km horizontally.For the conversion of nadir NO 2 dSCD bl into boundary layer vertical column densities, VCD bl , the geometric Air Mass Factor (AMF geo ) approximation was applied.
Here, SZA refers to solar zenith angle at the time of the measurement.This approach is in good agreement with explicit radiative transfer calculations for California while flying between 2 and 4 km.Radiative transfer calculations for the conditions of the Bakersfield case study (flight altitude: 2 km, SZA: <25º), and comparisons with ground based vertical columns consistently reveal the uncertainty in AMF geo to be less than 7% [42].The overall uncertainty in NO 2 VCD for the Bakersfield case study is estimated to be around 9% (AMF geo : <7%, NO 2 cross-section: ~5%, DOAS fit: ~3%) [42,44].

TOPAZ
NOAA's nadir-looking TOPAZ differential absorption lidar is a compact, solid-state-laserbased O 3 lidar that emits pulsed laser beams at three tunable wavelengths in the UV spectral region between about 285 and 300 nm [45].The differential attenuation of the three wavelengths due to O 3 permits the retrieval of O 3 concentration profiles along the laser beam path [46].TOPAZ O 3 profiles were computed every 10 s (or about 600 m horizontally) with a vertical resolution of 90 m.The ozone profiles extend from about 400 m beneath the plane to near the ground.O 3 values in the lowest two measurement bins (lowest 180m) above ground level (AGL) are typically not used because of poor signal to noise ratio.The TOPAZ lidar also provided aerosol backscatter profiles for the longest (and least absorbed by O 3 ) of the three emitted wavelengths near 300 nm.The time resolution of the aerosol backscatter profile measurements is the same as for O 3 , but the vertical resolution is much finer at 6 m.We used these highly resolved lidar backscatter profile data to retrieve boundary layer height (BLH) by employing a Haar wavelet technique [47].This approach is based on the (often valid) assumption that the aerosol concentration is higher in the boundary layer (BL) than in the lower free troposphere (FT).The altitude at which the strongest aerosol gradient is found by the wavelet technique is used as an estimate of the The TOPAZ lidar also provided aerosol backscatter profiles for the longest (and least absorbed by O 3 ) of the three emitted wavelengths near 300 nm.The time resolution of the aerosol backscatter profile measurements is the same as for O 3 , but the vertical resolution is much finer at 6 m.We used these highly resolved lidar backscatter profile data to retrieve boundary layer height (BLH) by employing a Haar wavelet technique [47].This approach is based on the (often valid) assumption that the aerosol concentration is higher in the boundary layer (BL) than in the lower free troposphere (FT).The altitude at which the strongest aerosol gradient is found by the wavelet technique is used as an estimate of the The TOPAZ lidar also provided aerosol backscatter profiles for the longest (and least absorbed by O 3 ) of the three emitted wavelengths near 300 nm.The time resolution of the aerosol backscatter profile measurements is the same as for O 3 , but the vertical resolution is much finer at 6 m.We used these highly resolved lidar backscatter profile data to retrieve boundary layer height (BLH) by employing a Haar wavelet technique [47].This approach is based on the (often valid) assumption that the aerosol concentration is higher in the boundary layer (BL) than in the lower free troposphere (FT).The altitude at which the strongest aerosol gradient is found by the wavelet technique is used as an estimate of the BLH.At times, the contrast in aerosol backscatter between the BL and the overlying FT is not sufficient to yield reliable results, and the BLH is not reported for such scenarios.
We used the BLH estimates and O 3 profiles measured with the TOPAZ lidar to compute O 3 column data integrated over the depth of the BL.To fill in data gaps in the O 3 profiles close to the ground, we averaged the ozone measurements in the lowest two gates with usable data (typically 200 -300 m AGL) and extrapolated this value to the ground.We then integrated these extrapolated ozone profiles from the surface to the top of the BL to yield BL O 3 column density along the flight track at 10-s resolution.When BLH estimates were not available from the backscatter profile for a given O 3 profile, BLH was interpolated from adjacent measurements to compute O 3 vertical column over the BLH.TOPAZ O 3 measurements have been extensively compared to and agree well (±2-9%) with in situ airborne O 3 observations [48].

Doppler Wind Lidar
Information on the wind structure below the aircraft was provided by the NCAS Doppler lidar [49] mounted in the Twin Otter cabin.The lidar measures the Doppler shift of radiation scattered from atmospheric aerosol particles to estimate the component of wind along the lidar line of sight.Typical precision of the lidar radial wind measurements under acceptable aerosol loading is better than a few tens of cm s -1 .The lidar was mounted in the cabin with the beam transmitted vertically through a small camera port located on the underside of the aircraft (Fig. 2).In order to measure the horizontal component of the winds a rotatable refractive wedge mounted in the port directed the beam to 12.5º off nadir.The original scanner design included two wedges, which provides greater beam deflection and enables vertical pointing; however poor optical quality of the wedges forced us to eliminate the second wedge to reduce total attenuation through the scanner.
During flight operations the wedge was rotated to four different azimuth angles (45, 135, 225, 315º) relative to the flight track.Dwell time at each azimuth angle was 1 s for most of the Doppler measurements during CalNex.A complete rotation among the four azimuths required 8 s, including the time required to rotate the wedge to a new position.At the nominal Twin Otter speed of 60 m s -1 a complete 4-beam scan was completed about every 500 m.Vertical resolution of the lidar wind measurements was roughly 50 m.
Information on aircraft speed and orientation was obtained from the CU AMAX-DOAS motion compensation system [42].Additionally, we used the surface return at the four look angles, for which the only Doppler shift results from motion of the aircraft, to provide additional information on aircraft orientation and velocity.For the case described here, a 19-beam running average of the radial wind estimates at each of the four azimuth angles was computed to improve precision of measurements.After removal of the Doppler shift induced by aircraft motion, the velocities from the four azimuth angles were combined in a leastsquares type algorithm to estimate the mean wind speed and direction in each of the 50 m range gates where backscatter was high enough to provide a strong signal.The wind speed and direction were averaged up to the BLH before further calculations of horizontal flux.The uncertainty in the wind measurement is estimated to be around 6% based on the difference between wind retrievals from a longer (19-beam) and a shorter (3-beam) running average wind fits.

Bakersfield Case Study
Bakersfield is a city located in the southern part of the Central Valley, CA, surrounded mostly by agricultural area and oil and natural gas operations.In the summer months, wind blows predominantly from the northwest down the valley providing steady wind conditions necessary for the method presented here.The Bakersfield area also makes for an interesting case study to probe NO 2 and O x production from a large city influenced by intense agriculture and petrochemical production.In particular, we have probed (i) background air unaffected by urban anthropogenic emissions, (ii) air upwind, influenced by agricultural and petrochemical production, (iii) urban emissions from the city, and, (iv) the chemical evolution downwind, after it is perturbed by urban emission inputs.
The flight plan of the Twin Otter on June 15, 2010 (see Fig. 3) was designed to interrogate NO x emissions and constrain the O 3 production from different source regions enclosed by "boxes" by applying the horizontal flux divergence approach.The flight plan included an enclosed box, over areas with no major emission sources, in the northwest of Bakersfield to characterize the background conditions (Box A; see Fig. 4).Two boxes (Box C and D; D is twice the size of C) were flown over the city of Bakersfield to constrain emissions/productions from the city.In order to contrast NO 2 and O 3 production upwind and downwind of Bakersfield, two additional boxes (Box B and E) were created by interpolating the measured NO 2 , O 3 and wind data for the western legs (shown as diamonds in Fig. 4).A larger trapezoid (Box F) was flown, connecting the three boxes, and enclosing the entire greater Bakersfield region.It took approximately 15, 13 and 18 minutes to complete boxes A, C and D respectively while the larger Box F took ~75 minutes.The entire box patterns were flown at a constant altitude (~2000 m above sea level), well above the BL.Details related to the times and meteorological conditions encountered for each boxes are summarized in Table 1.

Horizontal Flux and Source Strength
For each transect, the gas flux at a point, x along the flight path is obtained by multiplying the column measurement at that location, column(x) by the corresponding wind speed averaged over the BLH, u* avg (x) [10,11].The flux calculation through a surface area, A is shown in equation 2: Where J corresponds to flux at any location, x to the flight direction, z to the altitude, u* to the wind speed orthogonal to the flight direction (x) and is assumed to be constant over the BLH, u * avg to wind speed averaged up to the BLH, and = ∫ ( ).
(3) * = .sin ( ) Here u is wind speed and β is angle between wind direction and flight heading.
Flux measurement in a closed loop can be used to estimate source strength within the enclosed volume [10,11].The general continuity equation in the integral form is given in equation ( 5) and is the basis for the source strength calculation.It involves three terms: source, flux divergence and rate of change of concentration.
i.e. the net source strength of an enclosed volume, Q net is the sum of fluxes through all areas along the closed loop (incoming and outgoing) and change in concentration inside the volume.
We assumed that the time dependence of concentration in the enclosed volume is zero over the time scale of our measurement.Our measurements were performed during the midday when rate of change of NO 2 and O 3 concentration in the Bakersfield area is very small (see Fig. 5).Hence, we neglected the second term on the right hand side in equation ( 5).We also assumed that the net vertical exchange and deposition are negligible over the timescale of our measurement.Hence, the net flux i.e. the difference in fluxes entering and leaving the enclosed volume through the walls gives the source strength for the species of the particular enclosed area at the time of the measurement.

Daily NO x Emission
Daily NO x emission was estimated based on the computed NO 2 source strength, diurnal profile of NO 2 and the NO x to NO 2 ratio measured at the California Air Resource Board (CARB) monitoring station at Bakersfield.It is computed as: Where E is the computed NO 2 production rate from (5), t is hour of the day, t 0 is the hour of our measurement.Therefore, the daily NO x emission is the sum of product of measured NO 2 emission rate at time t 0 , ratio of NO 2 at time t and t 0 and ratio of NO x to NO 2 at time t over the course of the day.In order to minimize potential bias created due NO x accumulation overnight, we only calculated daily NO x emission for the period when NO 2 and NO x measurements at the CARB station were stable (10:00-20:59 PST).Fig. 5 shows NO 2 , NO x and O 3 mixing ratios measured hourly at the Bakersfield CARB monitoring station on that day.The time period of our measurement and time frame for the daily NO x emission calculation are also shown in Fig. 5.

RESULTS AND DISCUSSION
The measurement on June 15, 2010 over Bakersfield, CA was performed at mid-day when the change in NO 2 and O 3 concentration is very small, providing chemically stable conditions most suitable for source strength calculations.This is supported by NO 2 and O 3 measurements at the CARB monitoring station at Bakersfield (see Fig. 5).Column O 3 , BLH from the NOAA TOPAZ lidar, column NO 2 from the CU AMAX-DOAS, and wind speed and direction from NCAS Doppler lidar are shown in Fig. 3. Fig. 3a shows the BLH at the time of the measurement retrieved from TOPAZ backscatter profiles.In general, the highest O 3 and NO 2 VCDs were measured in the southeastern corners of the boxes especially for boxes with significant emission sources (Fig. 3b and 3c), consistent with the prevailing wind conditions.We calculated NO 2 and O x production rates for the six boxes.

RESULTS AND DISCUSSION
The measurement on June 15, 2010 over Bakersfield, CA was performed at mid-day when the change in NO 2 and O 3 concentration is very small, providing chemically stable conditions most suitable for source strength calculations.This is supported by NO 2 and O 3 measurements at the CARB monitoring station at Bakersfield (see Fig. 5).Column O 3 , BLH from the NOAA TOPAZ lidar, column NO 2 from the CU AMAX-DOAS, and wind speed and direction from NCAS Doppler lidar are shown in Fig. 3. Fig. 3a shows the BLH at the time of the measurement retrieved from TOPAZ backscatter profiles.In general, the highest O 3 and NO 2 VCDs were measured in the southeastern corners of the boxes especially for boxes with significant emission sources (Fig. 3b and 3c), consistent with the prevailing wind conditions.We calculated NO 2 and O x production rates for the six boxes.

RESULTS AND DISCUSSION
The measurement on June 15, 2010 over Bakersfield, CA was performed at mid-day when the change in NO 2 and O 3 concentration is very small, providing chemically stable conditions most suitable for source strength calculations.This is supported by NO 2 and O 3 measurements at the CARB monitoring station at Bakersfield (see Fig. 5).Column O 3 , BLH from the NOAA TOPAZ lidar, column NO 2 from the CU AMAX-DOAS, and wind speed and direction from NCAS Doppler lidar are shown in Fig. 3. Fig. 3a shows the BLH at the time of the measurement retrieved from TOPAZ backscatter profiles.In general, the highest O 3 and NO 2 VCDs were measured in the southeastern corners of the boxes especially for boxes with significant emission sources (Fig. 3b and 3c), consistent with the prevailing wind conditions.We calculated NO 2 and O x production rates for the six boxes.The wind speed and direction and BLH during the time of measurements for different boxes are given in Table 1.The wind was blowing predominantly from the northwest and provided ideal conditions as horizontal flux divergence measurements require steady wind fields (also see Fig. 3d).The variability in wind speed and direction was larger for upwind boxes (A and B) compared to downwind boxes (D and E).The BLH range for boxes D and E were also smaller compared to other boxes.The observed BLH variability is most likely a combination of land use changes (irrigated fields vs. dry land vs. urban heat island) and the fact that upslope flow over the foothills east of Bakersfield favor BLH growth, while strong subsidence over the middle of the valley acts to suppress BL growth.We use the BLH variability to estimate the amount of BL air column susceptible to exchange with the FT (see Section 3.1.).Since all the parameters needed to quantify flux are constrained by measurements here, the horizontal variability in BLH does not pose a limitation to our approach.Notably, the variability of BLH remains difficult to predict by atmospheric models, and warrants further investigation.
Background corrected O x (O x ' from here on) column up to the BLH, computed as the sum of NO 2 and background corrected O 3 columns, is shown in Fig. 4. Background correction for O 3 is needed as we are interested in the source strength of the area at the time of measurement i.e. the amount that is being produced locally.The background correction also any potential biases to BL-FT exchange in case of strongly varying BLH.If BLH is constant over a box, then background correction is not necessary, because incoming and outgoing background fluxes are the same and cancel each other.Note that we measure column amount of O 3 and NO 2 .Background O 3 levels were calculated based on the mean O 3 concentration over the BLH in the northwestern corner of the Box A (see Fig. 4).We note that 'background' air in the Central Valley is affected by transport of pollution emitted upwind; indeed we find significant production of O x ' in Box A. However, low and similar amounts of NO 2 are transported into and out of Box A, and net production of measured species is the lowest observed anywhere.We find no evidence for major emission sources within Box A.
The mean and the standard deviation of the background O 3 concentration was 1.20 ± 0.03 x 10 12 molecules cm -3 (48.8 ± 1.2 ppbv, 1ppbv = 2.46 x 10 10 molecules cm -3 ).It was assumed that this background O 3 concentration is representative of the entire area.A background O 3 column was calculated for each measurement point along the flight track by multiplying this O 3 concentration with BLH at that location.This background column was then subtracted from the measured O 3 column to determine O 3 ' , which corresponds to the excess O 3 column at each point along the flight track.NO 2 VCDs were used without further corrections as they were below the detection level over that area (4.2 x 10 14 molecules cm -2 , ~130 pptv).The average column NO 2 to O x ' partition ratio increased from 2% over the background site to 7% over the city of Bakersfield.Thus, NO 2 gas forms a significant portion of O x ' over the urban area and would result in a bias if O 3 production rates were calculated instead of O x .By investigating O x ' we eliminate the uncertainty due to titration of O 3 by NO to form NO 2 in the NO x source regions.The minimum, maximum and mean mixing ratios of NO 2 and O 3 for each box are also given in Table 2.The average NO 2 and O 3 concentrations were lowest for the background box.The NO 2 concentrations show higher variation within a box as well as between the boxes, indicating highly local NO 2 sources in the area.The mean O 3 does not vary much between the boxes (see Table 2).Notably, the maximum O 3 concentrations were observed to be generally related to the O x ' production rates in the box upwind of a box, which expected.Both O 3 and NO 2 showed the highest average concentration over the downwind box (E).Considering that the production rates are lower compared to the boxes upwind (B and D), there could be some accumulation of NO 2 and O 3 taking place in this box.
The enclosed areas are sources for both NO 2 and O x ' for all the boxes investigated.The NO 2 and O x ' production rates calculated for different boxes are given in Table 2.The production rates were calculated using equation ( 5) and have been normalized for the area of the boxes so that they can be directly compared to each other.As expected, the background, Box A, has the lowest production rate for both NO 2 and O x '.The NO 2 production rate in the background box was 0.04 kg hr -1 km -2 .The NO 2 production rate was highest for Box D and amounts to 1.35 kg hr -1 km -2 above Bakersfield.This is consistent with the present knowledge that urban city limits are the dominant source for NO x emissions in California [18].The minimum, maximum and mean mixing ratios of NO 2 and O 3 for each box are also given in Table 2.The average NO 2 and O 3 concentrations were lowest for the background box.The NO 2 concentrations show higher variation within a box as well as between the boxes, indicating highly local NO 2 sources in the area.The mean O 3 does not vary much between the boxes (see Table 2).Notably, the maximum O 3 concentrations were observed to be generally related to the O x ' production rates in the box upwind of a given box, which is expected.Both O 3 and NO 2 showed the highest average concentration over the downwind box (E).Considering that the production rates are lower compared to the boxes upwind (B and D), there could be some accumulation of NO 2 and O 3 taking place in this box.
The enclosed areas are sources for both NO 2 and O x ' for all the boxes investigated.The NO 2 and O x ' production rates calculated for different boxes are given in Table 2.The production rates were calculated using equation ( 5) and have been normalized for the area of the boxes so that they can be directly compared to each other.As expected, the background, Box A, has the lowest production rate for both NO 2 and O x '.The NO 2 production rate in the background box was 0.04 kg hr -1 km -2 .The NO 2 production rate was highest for Box D and amounts to 1.35 kg hr -1 km -2 above Bakersfield.This is consistent with the present knowledge that urban city limits are the dominant source for NO x emissions in California [18].Black areas in the map show active oil and gas wells in the region.
The minimum, maximum and mean mixing ratios of NO 2 and O 3 for each box are also given in Table 2.The average NO 2 and O 3 concentrations were lowest for the background box.The NO 2 concentrations show higher variation within a box as well as between the boxes, indicating highly local NO 2 sources in the area.The mean O 3 does not vary much between the boxes (see Table 2).Notably, the maximum O 3 concentrations were observed to be generally related to the O x ' production rates in the box upwind of a given box, which is expected.Both O 3 and NO 2 showed the highest average concentration over the downwind box (E).Considering that the production rates are lower compared to the boxes upwind (B and D), there could be some accumulation of NO 2 and O 3 taking place in this box.
The enclosed areas are sources for both NO 2 and O x ' for all the boxes investigated.The NO 2 and O x ' production rates calculated for different boxes are given in Table 2.The production rates were calculated using equation ( 5) and have been normalized for the area of the boxes so that they can be directly compared to each other.As expected, the background, Box A, has the lowest production rate for both NO 2 and O x '.The NO 2 production rate in the background box was 0.04 kg hr -1 km -2 .The NO 2 production rate was highest for Box D and amounts to 1.35 kg hr -1 km -2 above Bakersfield.This is consistent with the present knowledge that urban city limits are the dominant source for NO x emissions in California [18].Boxes C and E, located in the northern half of Box D, and immediately downwind of Box D respectively, show about 13 times lower NO 2 production rate, indicating that the NO 2 sources from the urban area are highly localized.Interestingly, the site upwind of Bakersfield (Box B) was also found to be a significant source for NO 2 compared to the downwind site (Box E).Molecular weight Ox (MWOx') = 48 g/mole 4 Error in the production rates represents total propagated measurement uncertainty.Details are provided in Section 3.1.
The O x ' production rate for the background box was 7.4 kg hr -1 km -2 , the lowest of all boxes.Box A likely represents the O x ' production rates for regions in the Bakersfield area that are not affected directly by the urban emissions.Notably, the NO 2 production from within box A was the lowest we have observed in this case study.However, our approach does not attempt to make a correction for NO 2 losses due to photochemistry and deposition, and as such the reported NO 2 production has to be considered a lower limit.While the measured NO 2 flux was essentially zero within error of the measurements, this indicates that comparable amounts of NO 2 enter and exit the box, and that enough NO x was present to produce O 3 .This was confirmed by our observations of elevated O x production in box A. Interestingly, the O x ' production rates over the Bakersfield city limit (Box C and D) and downwind site (Box E) only showed small enhancement (< factor of 1.75) over the background O x ' production rate.This indicates that even though the NO 2 levels in the wider area surrounding Bakersfield are relatively small (~330 pptv), there is enough NO x to sustain photochemical O x ' production in the entire region.Surprisingly, the upwind box (Box B) was found to have the largest O x ' production rate (17.1 kg hr -1 km -2 , 2.3 times that of Box A).The O x ' production rate in Box B was 133% that of urban Box D while the NO 2 production was only about 40% of the urban box.A plausible explanation for our observation of efficient and high O x ' production from Box B could be from enhanced VOC levels due to large oil and natural gas operations in the area, creating favorable conditions for enhanced O 3 production (high VOC/NO x ratio).Oil and natural gas production is a source for atmospheric methane, a greenhouse gas, and other more reactive hydrocarbons as well as NO x .The observed elevated emissions of NO x in box B indicate emissions are active in this area.While higher NO 2 is likely to contribute to the higher O 3 production rate, additional VOC emissions are needed to explain such a high increase in the O 3 production rate.We are unable to conclude about additional VOC sources from our data at this point, but note that some emissions of reactive hydrocarbons are expected from the oil and natural gas production in the area that could help accelerate O 3 formation.Notably, the O 3 produced within Box B was only partially transported into the downwind boxes, where lower O 3 production rates were observed.This decrease in O 3 production rates downwind of Box B are probably related to higher NO 2 concentrations, and a different VOC/NO x ratio as air mixes with urban sources.The net effect of the added emissions from urban sources was a lowering of the O 3 production rates.The black areas in Fig. 4 represent active oil and gas wells in the region [50].The measured NO 2 production rates were used along with the diurnal profile of NO 2 and NO x measured at the CARB monitoring network station of Bakersfield (see Fig. 5) to provide an estimate of daily NO x emissions in Table 3.We only considered the daylight hours (10:00-20:59 PST) when the measured NO 2 and NO x at the CARB station were stable in order to minimize potential bias due to NO x accumulation overnight.The uncertainty in Table 3 only considers error in measured production rates and does not include spatial and temporal variability in NO x and NO 2 across the region.The daily NO x emission from Bakersfield was estimated to be around 10. were comparable to those in the CARB 2010 emission inventory in the same area (29.2 tons).However, there are differences in the locations of the NO x emissions here as well.The background NO x emissions are higher in the emission inventory whereas emissions over the oil and natural gas operations are significantly underestimated.Table 3 also compares hourly emission rates for the hour of our measurement.The measured emissions were lower for all the boxes except for the upwind box, B. Considering that the daily emission estimates are in better agreement compared to the hour of measurement, there is a discrepancy in the timing of NO x emission in the emission inventory.The diurnal profile of NO 2 and the NO x to NO 2 ratio varies between days as well as seasons and hence we do not attempt to scale up to the yearly NO x emission.However with regular flights over different times of the day and course of different seasons, the combination of active and passive remote sensing has the potential to constrain and improve NO x emissions in emission inventories. 2Daily = 10:00-20:59 PST. 3 Hourly = hour of our measurement

Error Estimates
Error in calculated fluxes and source strengths are a function of uncertainties in the measurements of individual species, winds, uncertainties about sinks (dry deposition and oxidation), and the variability of atmospheric state.Previous source strength calculations have found the uncertainty in the wind to be the largest source of error as it was not measured concurrently [10,11].The uncertainty in the wind measurements is estimated to be around 6%. Thus, the uncertainty in the wind measurement itself has a relatively small effect on the production rates in our study.This uncertainty is very likely not representative of wind variability within the boxes but the variability in wind are captured as part of individual wind measurements.
The overall uncertainty in NO 2 VCD is ~9%.The contributions of different error sources in NO 2 VCD uncertainty is given in Section 2.1.It is assumed that all of NO 2 were located inside the BL.Based on the amount of NO 2 above the BL in the vertical profile through the city center (see Fig. 6) we estimate this leads to a systematic error of around a few percent.The lidar O 3 profile measurements at 90 m vertical resolution have an error of typically 6-10% and can be improved by integration and averaging [45,48].O 3 data were integrated vertically (~10 points) and horizontally (2-3 points) for flux calculations at each location.As a result, the statistical uncertainties in the integrated O 3 are reduced to ~2%.O 3 data to the ground were extrapolated from measurement at lowest two gates assuming a well-mixed BL (see section 2.3).Based on the difference between measurement at the CARB stations and extrapolated values, we estimate this systematic error to be around 5%.The uncertainty in BLH retrieval for TOPAZ backscatter profile is ~7%.Considering the standard deviation of background O 3 concentration (~4%), BLH uncertainty, O 3 column uncertainty and NO 2 column uncertainty, the overall error in the O x ' column is ~8%.Thus the total measurement uncertainty in the individual flux measurement is ~10-11%.The error for production rates of NO 2 and O x ' tabulated in Table 2 represents the overall uncertainty due to propagated error in individual column of the species and wind speed and direction for each box.Vertical transport, atmospheric sinks and chemical transformations over the transport time between source and sampling regions are other potential sources of errors in the measured production rates for NO 2 and O x '.Our approach assumes that all transport through the enclosed box occurs horizontally.Entrainment flux is a product of entrainment velocity (w e ) and the difference in trace gas concentrations between the FT and the BL (ΔC) (E flux = w e x ΔC).To constrain the magnitude of potential vertical flux, we have used the rate of BLH growth as the entrainment velocity.This neglects the contribution of large-scale, mean vertical velocity and BLH advection.The northern legs for Box C and D overlap in location, but were flown ~90 minutes apart and provide an opportunity to calculate the BLH growth rate.It was estimated to be 1.2 cm s -1 .NO 2 and O 3 concentrations were determined from the vertical profiles obtained from the low approach over the Bakersfield airport (see Fig. 6).We estimate the vertical flux to be 2.30 x 10 -3 kg km -2 hr -1 for NO 2 and 0.21 kg km -2 hr -1 for O x ' for Box D. Hence, the potential error due to vertical transport is likely to be smaller than 2%.
Typical dry deposition velocities, w d , for O 3 and NO 2 in the continental environment are 0.4 and 0.1 cm s -1 respectively [6,7,52].We calculated the depositional flux as D flux = w d x C, where C is the trace gas concentrations at the lowest layer.This could result in error of ~ 10% in the O x ' and ~1% in the NO 2 production rates.For an air mass transport time of 1 hour (between production and measurement), Ibrahim et al. [10] estimated the error in the production rate due to chemical transformation of NO x , based on average atmospheric NO x lifetime, to be around 10%.Considering that the transport time for our conditions is around 0.5 hour, we estimate the error due to chemical transformation of NO 2 to be less than 5%.The atmospheric lifetime of O 3 is more than 3 times that of NO x.Hence, we estimate the error in O x ' production rate due to chemical transformation to be smaller than 2%.Thus, the potential error due to entrainment, dry deposition and chemical transformation is in the same order as the total measurement uncertainty in the production rates.
We observed very high local variation in BLH and this could also potentially result in error in the calculated production rates.The BLH variability makes the air column susceptible to exchange with the FT via horizontal transport.We have tried to bind the magnitude of this exchange in two ways: Method A calculates it as a product of the relative amount of air column (with respect to the average) susceptible to this exchange based on the difference in the average BLH measured for the upwind and downwind legs and our measured production rates; Method B calculates the same number based on the maximum and minimum BLH, assuming they occur equally frequent within each box.This is likely an upper limit estimate of such transport, since BLH is something in between most of the time.Note that we only know the BLH along the edge of the boxes and not within the boxes.We find that the uncertainty due to this potential FT exchange accounts for 1-8% (method A) of the overall horizontal flux.Method B yields 15-30% as an upper limit for FT exchange.To our knowledge the horizontally variable BLH as a mechanism for BL-FT exchange has not previously been studied, and deserves further investigation.We consider the error in horizontal fluxes from method A to be most likely representative of uncertainty in the production rates listed in Table 2 due to such an exchange.

CONCLUSION AND OUTLOOK
We demonstrated the feasibility of co-deployed active and passive remote sensing instruments aboard a research aircraft to study NO x emissions and O x production rates out of an urban area.NO 2 vertical column, O 3 vertical profile and wind profile measurements aboard the aircraft were used to calculate NO 2 and O x ' fluxes from source areas along the flight track.
The advantages to co-deployment of these three remote sensing instruments on a mobile platform for this kind of study are as follows: 1.The flux calculations are fully experimentally constrained.In particular, measurements of wind and BLH along the flight track decouple horizontal and vertical transport; column measurements integrate pollutant concentrations over the BLH, and are inherently insensitive to vertical inhomogeneity.2. Measurements constrain NO 2 , O 3 , O x (O 3 +NO 2 ), enabling studies of NO x emissions and O x production rates also over NO x source areas, i.e., under conditions when O 3 concentrations are reduced due to NO emissions (O 3 titration to form NO 2 ). 3. Colocation of all three sensors on a single platform minimizes error, and makes the flux calculation straightforward by assuring sampling on similar temporal and spatial scales.
The horizontal flux divergence approach presented here for a case study in the Bakersfield area has comparatively small error for the largest box (Box F) and larger error for the smallest box (Box C).The overall measurement uncertainty in the individual flux measurement is in the order of 10-11%.The potential error due to entrainment, dry deposition and chemical transformation is of a similar magnitude, and can in the future be further reduced through coupling with atmospheric models.Our measured NO 2 and O x ' production rates reveal higher O 3 production upwind of Bakersfield in an area with active oil and natural gas production.This finding is highly significant within experimental error, and spatially well separated from urban source areas.Comparison of NO x emissions with the CARB 2010 emission inventory suggest that the NO x emissions from the urban area are well represented in the inventory.However, the location and timing of the NO x emissions within the urban area could be improved.In contrast, NO x emissions over areas with active oil and natural gas production were found to be significantly underestimated; higher background emissions compensate for these local effects over the entire study area.The impacts of emissions from oil and natural gas production deserve further investigation.
Models that predict O 3 formation have not previously been constrained by observations at the scale of cities and immediately downwind of cities.The synergistic benefit of combining active and passive remote sensing instruments demonstrated here holds great potential as an innovative tool to improve NO x emission inventories (emitted amounts and location) as well as constrain O x production rates experimentally, and over extended areas.The local variations in BLH deserve further investigation as to their role in the exchange of air between the BL and the FT.Further, other trace gases (e.g., formaldehyde and glyoxal) can be measured by AMAX-DOAS and hold largely unexplored potential to extend this approach to the study of VOC oxidation rates.The co-deployment of AMAX-DOAS, TOPAZ lidar and Doppler wind lidar during 51 remaining flights provide a valuable dataset to locate and constrain NO x emissions over much of California especially the South Coast Air Basin, the Bay area, as well to assess the transport of NO 2 and O x across the US-Mexican border.

Fig. 1 .
Fig. 1.Schematic diagram showing cross-section of ozone formation in an urban area under steady wind conditions and horizontal flux divergence measurements in a closed loop for source strength calculations.Evolution of NO (orange), NO 2 (blue), O 3 (maroon) and O x (green) over different urban regions: (I) upwind, (II) urban center, (III) downwind and (IV) further downwind are also illustrated.

Fig. 1 .
Fig. 1.Schematic diagram showing cross-section of ozone formation in an urban area under steady wind conditions and horizontal flux divergence measurements in a closed loop for source strength calculations.Evolution of NO (orange), NO 2 (blue), O 3 (maroon) and O x (green) over different urban regions: (I) upwind, (II) urban center, (III) downwind and (IV) further downwind are also illustrated.

Fig. 1 .
Fig. 1.Schematic diagram showing cross-section of ozone formation in an urban area under steady wind conditions and horizontal flux divergence measurements in a closed loop for source strength calculations.Evolution of NO (orange), NO 2 (blue), O 3 (maroon) and O x (green) over different urban regions: (I) upwind, (II) urban center, (III) downwind and (IV) further downwind are also illustrated.

Fig. 2 .
Fig. 2. Instrumental setup of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar aboard the NOAA Twin Otter research aircraft during CalNex and CARES field campaigns.The yellow, purple and maroon lines represent viewing geometry of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar respectively.The three instruments are also shown in the insets.

BritishFig. 2 .
Fig. 2. Instrumental setup of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar aboard the NOAA Twin Otter research aircraft during CalNex and CARES field campaigns.The yellow, purple and maroon lines represent viewing geometry of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar respectively.The three instruments are also shown in the insets.

BritishFig. 2 .
Fig. 2. Instrumental setup of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar aboard the NOAA Twin Otter research aircraft during CalNex and CARES field campaigns.The yellow, purple and maroon lines represent viewing geometry of CU AMAX-DOAS, NOAA TOPAZ lidar and NCAS Doppler wind lidar respectively.The three instruments are also shown in the insets.

Fig. 3 .
Fig. 3. Maps of (A) Boundary layer height (BLH) above ground level, (B) O 3 vertical columns up to the BLH, (C) NO 2 vertical columns and (D) wind speed and direction from the flight over Bakersfield, CA on June 15, 2010.BLH and O 3 columns were measured by NOAA TOPAZ lidar, NO 2 vertical columns by CU AMAX-DOAS and wind speed and direction by NCAS Doppler wind lidar.Black diamond on A shows the location of CalNex Bakersfield supersite.

Fig. 3 .
Fig. 3. Maps of (A) Boundary layer height (BLH) above ground level, (B) O 3 vertical columns up to the BLH, (C) NO 2 vertical columns and (D) wind speed and direction from the flight over Bakersfield, CA on June 15, 2010.BLH and O 3 columns were measured by NOAA TOPAZ lidar, NO 2 vertical columns by CU AMAX-DOAS and wind speed and direction by NCAS Doppler wind lidar.Black diamond on A shows the location of CalNex Bakersfield supersite.

Fig. 3 .
Fig. 3. Maps of (A) Boundary layer height (BLH) above ground level, (B) O 3 vertical columns up to the BLH, (C) NO 2 vertical columns and (D) wind speed and direction from the flight over Bakersfield, CA on June 15, 2010.BLH and O 3 columns were measured by NOAA TOPAZ lidar, NO 2 vertical columns by CU AMAX-DOAS and wind speed and direction by NCAS Doppler wind lidar.Black diamond on A shows the location of CalNex Bakersfield supersite.

Fig. 4 .
Fig. 4. Map of O x ' vertical columns up to the BLH.Rectangles represent different boxes flown over Bakersfield: (A) upwind background area, (C) over the city, and (F) over the entire area.Colored rectangles in the inset highlight three boxes: (B) upwind, (D) over the city, and (E) downwind.Box B and E were created for comparison purposes by interpolating the western legs.Interpolated data are shown as diamonds.Black areas in the map show active oil and gas wells in the region.

BritishFig. 4 .
Fig. 4. Map of O x ' vertical columns up to the BLH.Rectangles represent different boxes flown over Bakersfield: (A) upwind background area, (C) over the city, and (F) over the entire area.Colored rectangles in the inset highlight three boxes: (B) upwind, (D) over the city, and (E) downwind.Box B and E were created for comparison purposes by interpolating the western legs.Interpolated data are shown as diamonds.Black areas in the map show active oil and gas wells in the region.

BritishFig. 4 .
Fig. 4. Map of O x ' vertical columns up to the BLH.Rectangles represent different boxes flown over Bakersfield: (A) upwind background area, (C) over the city, and (F) over the entire area.Colored rectangles in the inset highlight three boxes: (B) upwind, (D) over the city, and (E) downwind.Box B and E were created for comparison purposes by interpolating the western legs.Interpolated data are shown as diamonds.Black areas in the map show active oil and gas wells in the region.

Fig. 5 .
Fig. 5.Diurnal variation of NO 2 , NO x , O 3 , and O x measured at the Bakersfield CARB monitoring station on June 15, 2010.The red and grey rectangles represent the timeframe of our measurement and time period used for daily NO x emission calculation respectively.
7 metric tons for June 15, 2010 from 10:00-20:59 PST, compared with 13.5 tons NO x for the same time frame in CARB 2010 emission inventory (CalNex-2010 modeling inventory) [51].There is a mismatch in the location of NO x emissions within Bakersfield.Our measurements suggest the portion of emission occurs in the southern half of Box D. Note that NO x emissions of Box C were also part of Box D (i.e., form the northern half of Box D).Emissions in Box C were comparatively very small.In contrast, emissions from Box C form a significant portion of overall emission of Box D in the CARB 2010 emission inventory.The NO x emissions for the entire study area (Box F: 32.1 tons)

Fig. 6 .
Fig. 6.Mean O 3 profile measured by NOAA TOPAZ lidar for Box D and NO 2 profile measured by CU AMAX-DOAS from a low approach at Bakersfield airport.Error bars for O 3 profile is the standard deviation of the mean and NO 2 shows measurement uncertainty.

Table 3 . Daily and hourly NO x emissions calculated for June 15, 2010 from (i) using derived NO 2 production rate and NO x and NO 2 diurnal profiles measured at CARB monitoring station at Bakersfield, CA and (ii) NO X emissions from CARB 2010 emission inventory for that day. Errors represent error due to uncertainty in NO 2 production rates Box NO x Emissions 1 (metric tons) This work CARB 2010 emission inventory Daily 2 Hourly 3 Daily 2 Hourly
1Molecular weight of NOx (MWNOx) = 46 g/mole.