Diurnal variation of midlatitudinal NO 3 column abundance over table mountain facility , California

The column abundance of NO 3 was measured over Table Mountain Facility, CA (34.4 N, 117.7W) from May 2003 through September 2004, using lunar occultation near full moon with a grating spectrometer. The NO 3 column retrieval was performed with the differential optical absorption spectroscopy (DOAS) technique using both the 623 and 662 nm NO 3 absorption bands. Other spectral features such as Fraunhofer lines and absorption from water vapor and oxygen were removed using solar spectra obtained at different airmass factors. We observed a seasonal variation, with nocturnally averaged NO 3 columns between 5− 7× 1013 molec cm−2 during October through March, and 5−22×1013 molec cm−2 during April through September. A subset of the data, with diurnal variability vastly different from the temporal profile obtained from onedimensional stratospheric model calculations, clearly has boundary layer contributions; this was confirmed by simultaneous long-path DOAS measurements. However, even the NO3 columns that did follow the modeled time evolution were often much larger than modeled stratospheric partial columns constrained by realistic temperatures and ozone concentrations. This discrepancy is attributed to substantial tropospheric NO3 in the free troposphere, which may have the same time dependence as stratospheric NO 3. Correspondence to: C. M. Chen (claudine.chen+acp@googlemail.com)


Introduction
NO 3 plays a significant role in the chemistry of the stratosphere and troposphere.In the stratosphere, it influences the partitioning of active nitrogen species NO and NO 2 (NO x ), where NO x is an important component in catalytic ozone loss cycles.The primary source of NO 3 is a reaction between NO 2 and O 3 (R1), and it is consumed by an additional reaction with NO 2 to form the reservoir species N 2 O 5 (R2).The subsequent removal of N 2 O 5 via a heterogeneous reaction with water to form nitric acid, also a reservoir species for NO x but with longer lifetime, contributes to NO x removal.The thermal decomposition of N 2 O 5 is an additional significant source of NO 3 in the upper stratosphere.Since NO 3 photodissociates extremely rapidly at wavelengths less than about 640 nm, we observe significant concentrations only at night.
In the troposphere the same creation and destruction reactions occur, and there is also negligible NO 3 during sunlit hours except for extremely polluted urban settings (Geyer et al., 2003).In the boundary layer, NO 3 additionally is an important nighttime oxidant because it reacts rapidly with many biogenic hydrocarbons such as alkenes, aldehydes and terpenes (Atkinson, 1991;Wayne et al., 1991).
Interest in the role played by NO 3 in atmospheric chemistry increased significantly following the first reports of its detection in the stratosphere and troposphere by Noxon et al. (1978Noxon et al. ( , 1980) ) and Platt et al. (1980).Since then, other measurements of atmospheric NO 3 column at low and midlatitudes at urban-influenced and remote groundbased sites have been made by using the Moon as a light source and employing differential optical absorption spectroscopy (DOAS) (Aliwell andJones, 1996a,b, 1998;Lal et al., 1993;Renard et al., 2001;Solomon et al., 1989).Also, vertical concentration profiles of NO 3 have been inferred from ground-based measurements by observing NO 3 in the slant column during sunrise with direct lunar, zenith sky, and off-axis methods (Allan et al., 2002;Coe et al., 2002;Smith and Solomon, 1990;Smith et al., 1993;von Friedeburg et al., 2002;Weaver et al., 1996).As the solar terminator sweeps from the upper atmosphere down to the surface, photolysis progressively decreases the column of NO 3 , leaving only the column that lies below the terminator altitude.Additionally, stratospheric profiles of NO 3 have been obtained from the SAGE III (Stratospheric Aerosol and Gas Experiment) and SCIAMACHY (SCanning Imaging Absorption spec-troMeter for Atmospheric CartograpHY) (Amekudzi et al., 2005) satellite instruments using lunar occultation, and the GOMOS (Global Ozone Monitoring by Occultation of Stars) (Hauchecorne et al., 2005;Marchand et al., 2004)  A number of measurements have confirmed the role of NO 3 -N 2 O 5 chemistry in the nocturnal boundary layer (Aldener et al., 2006;Allan et al., 2000;Ambrose et al., 2007;Ayers and Simpson, 2006;Brown et al., 2003Brown et al., , 2004;;Carslaw et al., 1997a;Geyer et al., 2001;Geyer and Platt, 2002;Li et al., 2008;Matsumoto et al., 2006;Mihelcic et al., 1993;Nakayama et al., 2008;Smith et al., 1995;Stutz et al., 2004;Vrekoussis et al., 2007;Wang et al., 2006).Fewer have probed above the boundary layer, essentially those using LP-DOAS (Carslaw et al., 1997b), aircraft measurements (Brown et al., 2007a,b) and zenith sky measurements at sunrise (Allan et al., 2002;Coe et al., 2002;von Friedeburg et al., 2002).Due to this relative lack of measurements above the boundary layer, our quantitative understanding of the role of NO 3 -N 2 O 5 chemistry in the free and upper troposphere is incomplete.
Our focus is on the quantification of NO 3 in the free troposphere.We deduce timeresolved estimates of free tropospheric NO 3 using measurements of total column NO 3 , observations of the boundary layer concentration of NO 3 , with stratospheric columns provided by a model.Specifically, we present results of simultaneous measurements of NO 3 column by lunar occultation with the DOAS technique, and surface concentration of NO 3 using LP-DOAS, taken on evenings near full moon in August and September 2004 over Table Mountain Facility (TMF), California.Profiles of NO 3 found using a stratospheric model, shown to be consistent with SAGE III satellite lunar occultation measurements of NO 3 , provided stratospheric partial columns with inputs from a climatology constructed from over 10 yr of lidar measurements at our measurement site.We also used global chemistry and transport model GEOS-Chem to characterize the time evolution and vertical distribution of NO 3 in the troposphere for various locations.The full dataset of NO 3 column measurements was taken from May 2003 through September 2004, and we characterized the magnitude and variability of column NO 3 at TMF, a location near Los Angeles influenced by clean and polluted air masses.Introduction

Conclusions References
Tables Figures

Back Close
Full

Location and measurement frequency
We have acquired direct lunar occultation measurements of the NO 3 column at Table Mountain Facility (TMF), California (34.4 • N, 117.7 • W) at an altitude of 2280 m.TMF is in the San Gabriel Mountains north of the Los Angeles Basin and south of the Mojave Desert.Optimal measurements were acquired within the period two days before and two days after full moon.Full moon conditions offer the longest period of moonlit evening hours and the highest signal to noise due to the intensity of reflected sunlight with lunar phase angle opposition effect (Hapke et al., 1993).Observations from May 2003 through September 2004 consisted of three evenings for each full moon event, weather permitting, and resulted in 40 d of data.

Instrument configuration
The experimental apparatus (Cageao et al., 2001) is shown in Fig. 1.Light was collected by a heliostat and directed into an off-axis telescope with 3× magnification.The 7-cm diameter collimated beam from the telescope was transmitted through a condensing lens, a shutter and order sorting filter (Schott GG-400 glass) to a 0. by the entrance slit projected on the focal plane.For the initial observations, the CCD detector operated in imaging mode, with an integration time of three seconds at peak lunar intensity.The intensity of the moon decreases rapidly off full moon (50% decrease two days from full moon) and with increasing airmass, so the integration time was adjusted to maintain a constant CCD exposure level.Spectra, obtained between sunset and sunrise, were recorded every 10-20 min, yielding at least 30 column abundance values for each evening.In the more recent datasets, June, August, and September of 2004, the collection frequency was increased to a spectrum every minute to improve the time resolution.This increased the number of column measurements to over 400 per night.
To improve the signal-to-noise ratio, five scans were averaged to obtain a single archived spectrum.There were similar integration times among the five scans averaged, but total integration time could differ when comparing averaged data on and off full moon.For measurements in June, August, and September of 2004, the detector was run in a mode that acquired 29 spectra which were summed for each archived spectrum, with a total integration time of 7.25 s at peak lunar intensity.All spectra were dark-corrected, but not flat-fielded, since pixel-to-pixel variability canceled with comparison to the reference spectrum, as described below.

Spectral analysis
Remote and in situ sensing of atmospheric NO 3 make use of the strong vibrational bands at 662 and 623 nm which are assigned to the (0,0) and (1,0) bands, respectively, of the ν 1 symmetric stretch in the A 2 E ← X 2 A 2 electronic transition (Ramsay, 1962).
High resolution laboratory spectroscopy studies have shown these bands are diffuse with cross sections that are weakly dependent on temperature (Cantrell et al., 1987;Ravishankara and Mauldin, 1986;Sander, 1986;Yokelson et al., 1994)  spectral region.There are also absorption bands in this spectral range for O 4 and NO 2 , but the contribution from these bands was negligible at Table Mountain.O 4 scales with the square of O 2 concentration, and therefore exists primarily at the surface.NO 2 has very weak lines in this spectral range that are lost in the noise for this measurement.
The spectrometer dispersion and grating position were selected to give a spectral bandpass of 617-674 nm.In this spectral interval we recorded and analyzed both the 623 and 662 nm absorption bands of NO 3 .The spectra were analyzed using the differential optical absorption spectroscopy (DOAS) approach (Platt and Stutz, 2008) with the spectral analysis and deconvolution program, MFC (Stutz and Platt, 1996).
The principle of DOAS is to use only the high-frequency components of the spectrum to determine the quantity of an optically absorbing atmospheric component.The effect of the DOAS processing steps on the data is shown in Fig. 2. First, the slowly varying low frequency component of the background, from sources such as Rayleigh and Mie scattering effects and solar flux spectral variations, was removed numerically from the lunar occultation spectra.This was done by dividing the raw data with a smoothed version of the same data.Shown in Fig. 2b is the logarithm of this ratio, making the spectrum proportional to column abundance and molecular absorption cross section.
The reflected solar Fraunhofer lines that dominate the raw spectra were then removed using solar reference spectra.The high-pass filtered lunar spectrum was divided by a high-pass filtered solar reference spectrum after aligning the spectra in wavelength with a nonlinear fit with stretch.The resulting spectrum was fit by least squares to high-pass filtered reference spectra of NO 3 , H 2 O, and O 2 (Fig. 2c).
The slant NO 3 column abundances calculated from the fit were converted to vertical column abundances by dividing by the airmass, a factor that describes the amount of air seen through a slant path in the atmosphere compared to a view directly overhead.
The airmass was determined from the reciprocal of the cosine of the lunar zenith angle (a valid approximation for angles up to 80 • ) with an additional small correction for refraction.Figures

Back Close
Full To obtain the reference spectra for the solar features, we recorded direct solar spectra with two ground glass diffuser plates in the light path between the heliostat and telescope.The primary purpose of the diffuser plates was to average the observed radiance over the entire solar disk.Without the plates, only a small fraction of the solar disk was imaged onto the spectrometer slit.The solar spectral features in the nondiffuse spectra differed from those in the lunar spectra.The diffuser plates also helped to attenuate the solar beam, although additional neutral density filtering was used to avoid detector saturation.
Solar reference spectra were acquired over a day for the airmass range 1-7 (SZA 34-81 • ).New solar reference spectra were taken once a month during the time of the full moon datasets to account for small changes in instrument alignment.In addition to solar lines, these spectra contained terrestrial water vapor and O 2 features with optical depths that were proportional to the airmass.The solar reference spectrum used in the processing for a particular lunar spectrum had an airmass within ±0.5 of the airmass of the lunar data, thereby removing much of the water and O 2 column prior to additional processing.To account for the remaining and variable water vapor and O 2 signal, ratios of solar spectra at different airmasses provided empirical water and O 2 reference spectra.Since there is little overlap of these two spectral features, the O 2 and water features were individually isolated and used as empirical spectral references.The low spectral resolution of measurement does not resolve individual lines for water and O 2 , and therefore has little sensitivity to pressure broadening.
The NO 3 reference spectra used were obtained from laboratory absorption cross section studies of NO 3 and have been measured over the temperature range relevant to the troposphere and stratosphere (Cantrell et al., 1987;Ravishankara and Mauldin, 1986;Ravishankara and Wine, 1983;Sander, 1986;Yokelson et al., 1994).The cross sections of Sander (1986) and Yokelson et al. (1994) are in excellent agreement over the range of overlap of temperature.Both studies observed a significant decrease in NO 3 cross sections at the peaks of the 662 and 623 nm bands with decreasing temperature.In contrast, the results of Cantrell et al. (1987)  section with temperature and are assumed to be incorrect.The results of Ravishankara and Mauldin (1986) disagree significantly with those of Yokelson et al. (1994) from the same group, and are assumed to be superseded by the latter.Although the temperature at the peak of the stratospheric NO 3 concentration profile at 40 km is roughly 260 K, the average temperature weighted by the model-predicted NO 3 concentration profile in the stratosphere is closer to 240 K.We have used the spectrum of Yokelson et al. (1994) at 240 K for the column retrievals presented here.Solomon et al. (1989) also used a reference temperature of 240 K, while Aliwell and Jones (1998) used 260 K.

Measurement uncertainty
The overall uncertainty for our measurement of total column NO 3 is approximately 17% RMS.The most important contributions to this uncertainty are systematic errors in cross section, and photon noise.The stated uncertainty in the NO 3 cross section is ±10% (Yokelson et al., 1994), excluding the errors associated with the temperature dependence of the cross sections.Our estimate of total column NO 3 assumes that the absorption is dominated by stratospheric contributions.There is a ±6% error associated with the use of a single cross section at a temperature of 240 K, if the temperature of the column actually varies between 220 and 260 K.If there are contributions from tropospheric NO 3 , the retrieved columns are a lower limit since the cross section of the band peak at 662 nm for 298 K is 17% less than for 240 K. Photon noise in the system contributes an estimated 13% to the uncertainty.The detection limit for the NO 3 slant column abundance is 2×10 12 molec cm −2 .The signal to noise ratio was >5 for most of the evening (during the steady state growth period); this ratio was larger for measurements through larger airmasses or larger NO 3 column amounts.

Long-path DOAS instrument
Horizontal column average measurements were made at TMF for August and September 2004 using a long-path differential optical absorption spectrometer (LP-DOAS).Figures

Back Close
Full LP-DOAS is an active remote sensing technique that gives exceptionally low detection limits by averaging over a long (several kilometers) pathlength.Light from a broadspectrum 500W Xe arc lamp was collimated through a Newtonian telescope and broadcast to an array of corner-cube retroreflectors mounted on a radio tower located on the Blue Ridge in the Angeles National Forest.The distance between the instrument and the retroreflector array was approximately 3.4 km.Light incident on the retroreflector array traveled back through the atmosphere, was collected by the same telescope and transmitted through a fiber-optic cable to the spectrometer and detector.The difference in altitude between the LP-DOAS instrument and the tower-mounted retroreflector array was 298 m.A detailed description of the LP-DOAS instrument and the NO 3 analysis employed here is given in (Geyer et al., 1999).The measurement uncertainty of the LP-DOAS is dominated by the error in the absorption cross-section of NO 3 , which is ±10% (Yokelson et al., 1994) as previously noted.
3 Model descriptions

1-D stratospheric model
A one-dimensional, photochemical steady state model of the stratosphere (Osterman et al., 1997) was run using TMF climatological profiles of temperature and O 3 as inputs, and the modeled NO 3 column abundances were compared to the TMF column measurements.The model calculates diurnally varying species concentrations, assuming each species reaches a balance between production and loss over 24 h for a given temperature and pressure profile and latitude.JPL 2006 cross sections and quantum yields were used to determine photolysis J values, and JPL 2006 kinetic rate constants were used for reaction rates (Sander et al., 2006(Sander et al., , 2003)).Chemical inputs are profiles of O 3 , H 2 O, CH 4 , NO y , Cl y , CO, H 2 , C 2 H 6 , Br y , and aerosol parameters based on a climatology derived from NASA satellite and balloon observations (e.g., Yang et al., 2006), as detailed in Table 1.Figures

Back Close
Full Additionally, SAGE III satellite measurements of O 3 in the stratosphere were used to verify the consistency between the 1-D stratospheric model and SAGE III measurements of NO 3 , testing the current understanding of stratospheric NO 3 chemistry (described in Appendix A).Analysis of the sensitivity of modeled NO 3 column to input parameters and to uncertainties of reaction rates were also conducted and are described in Appendix B.

Table Mountain Facility lidar climatology
Temperature and ozone profiles have been measured at TMF by lidar since 1988 and offer a unique opportunity to compare our measurements with a model with realistic constraints.Temperature profiles were measured between 30-80 km and ozone profiles between 15-50 km, both with 300 m vertical resolution since September 1994 and with 600 m vertical resolution beforehand.Three cases were run using these data: climatological monthly mean values and variability over the 10 yr period 1988-1997 (data extracted from the published contour plots) (Leblanc and McDermid, 2000;Leblanc et al., 1998), and monthly mean profiles for 2003 and for 2004 provided by Leblanc (2005).Temperature and ozone profiles are sufficient for estimating the NO 3 column since NO 3 is primarily determined by these two quantities, as verified from a sensitivity study described in Appendix B1.
The change in NO 3 column at the extremes of variability was probed by running the model with both temperature and ozone variability added or subtracted from the climatological temperature and ozone profiles.The uncertainty in the climatological temperature measurements are 0.6 K at the middle of the altitude range, 8 K at 30 km, and <4 K at 80 km.The uncertainty of the climatological ozone measurements are a few percent at the peak of the ozone, 10-15% at 15 km, and more than 40% above 45 km.The uncertainty in ozone for the 2003 and 2004 monthly mean profiles was a minimum at 6% at the ozone peak, increasing in error above and below this altitude to 10% at 18 km and 42 km.The uncertainty in temperature for the 2003 and 2004 monthly mean profiles varied between 0.5 and 1.3 K over 13-60 km.

Conclusions References
Tables Figures

Back Close
Full The gaps in the ozone and temperature profiles were filled with the climatology from (1) the Upper Atmosphere Research Satellite (UARS) Reference Atmosphere Project (URAP) (Remedios et al., 2007;Wang et al., 1999;Randel et al., 1999), and then (2) a climatology dataset based on ozone data from D ütsch (1974) and ozone and temperature from the Middle Atmosphere Program (Barnett and Corney, 1985;Keating and Young, 1985), with adjustments to the ozone climatology based on in situ ozone measurements in the upper troposphere and lower stratosphere from many field programs.URAP is a compilation of global data from the CLAES, HALOE, HRDI, MLS and ISAMS instruments on the UARS satellite taken from April 1992 through seven years, processed into zonal monthly means with standard deviation.Two types of data were provided: "baseline" data obtained from April 1992 to March 1993, and "extended" datasets averaged over 7 yr.We used extended data where available.The ozone data used were from the extended time range, and the temperature data from the baseline time frame.We also used profiles of H 2 O (extended), CH 4 (extended), and N 2 O (baseline) from URAP.N 2 O was used as a tracer to estimate model inputs for NO y , Cl y , and Br y using well established tracer relations (e.g., Yang et al. (2006) and references therein).
Static profiles for CO and C 2 H 6 from MkIV measurements (Toon, 1991;Sen et al., 1998), and a H 2 profile based on measurements in the stratosphere (Abbas et al., 1996;Dessler et al., 1994;Rockmann et al., 2003) were used for all months.Vertical profiles of sulfate aerosol surface area were based on zonal monthly mean measurements by SAGE II (Thomason et al., 1997) updated to include data acquired during the time of our NO 3 measurements.

GEOS-Chem tropospheric model
We use the GEOS-Chem global 3-D tropospheric chemistry and transport model (Bey et al., 2001;Park et al., 2004;Wu et al., 2007)  7.02.04,http://acmg.seas.harvard.edu/geos/) is driven by the assimilated meteorological GEOS-4 data from NASA Global Modeling and Assimilation Office (GMAO) with 6-h temporal resolution (3-h for surface variables and mixing depths) and a horizontal resolution of 1×1.25 • with 55 layers in the vertical.The horizontal resolution of the GEOS-4 wind fields has been degraded to 2×2.5 • for input into GEOS-Chem.

Experimental results
As seen in model calculations in Fig. 3, the diurnal variation of stratospheric NO 3 can be characterized by four phases: daytime photolysis (negligible NO 3 ), sunset buildup (a rapid rise in NO 3 column), nocturnal steady state (a nearly linear, slow rise in NO 3 column), and sunrise destruction (rapid decrease in NO 3 column).These four stages were also observed in our measurements, except for some variations during the nocturnal steady state stage.Time series that monotonically increase, which occurs almost linearly during the steady state phase, are labeled as "model-like behavior" as seen in our measurements from September 2004 (Fig. 3b); this label does not necessarily mean that these data are purely stratospheric in origin.The remaining data are described as "non-modeled behavior", and displayed a wide variety of different temporal behavior with variability ranging from one to several hours, as seen in our measurements from August 2004 (Fig. 3a).
For purposes of comparison, each night of data was reduced to a time-averaged mean column and a standard deviation over the steady state phase, which was taken to be two hours after sunset up to roughly 30 min before sunrise.An annual plot of all the mean columns, with 2-σ standard deviations as error bars, is shown in months (April through September) was 7.5×10 13 molec cm −2 .The column averaged over winter months (October through March) was 5.5×10 13 molec cm −2 , as summarized in Table 2; cloudy conditions limited the amount of data that could be acquired during winter and none was possible in 2003.We attribute the larger summertime values to a warmer atmosphere (both in the troposphere and stratosphere), which drives the thermal decomposition of N 2 O 5 to form NO 2 and NO 3 .In the 14 remaining cases with non-modeled behavior (closed symbols in Fig. 4), which occurred from May to early October, the range of mean columns was 6-22×10 13 molec cm −2 and the range of standard deviations was 1-5×10 13 molec cm −2 .The low end of the range of standard deviations occurred for cases with relatively flat but decreasing temporal profiles, while the high end of the range was characterized by large oscillations in the NO 3 column.
The large oscillations in total column NO 3 that occurred over an evening did not originate from the stratosphere, since the primary source of variability in the stratosphere is from planetary waves, which have time scales longer than one day (Salby, 1984;Wu and Waters, 1996).Instead these variations are likely to arise from the troposphere.However, the complexity of mountain topography complicates the determination of the origin of the tropospheric air at TMF using traditional back trajectory methods.Depending on the movement of the air masses, we observed from a range of sources as diverse as desert air to polluted urban air and air from aloft due to mountain subsidence and drainage flow.
Model results by Lu and Turco (Lu andTurco, 1995, 1996;Lu et al., 2003) of the Los Angeles basin air flow give an idea of seasonal behavior during quiescent conditions.Their Surface Meteorology and Ozone Generation (SMOG) model calculates winds and tracer transport in the Los Angeles basin and surrounding mountain areas.
Land warming by solar radiation propels onshore winds and upslope mountain flows during the day, with stronger winds in summer than winter.In the summer evenings, there are generally downslope flows from the mountains, and disorganized winds in the basin.Winter evenings have a stronger offshore component from radiative cooling of Introduction

Conclusions References
Tables Figures

Back Close
Full mountaintops and subsidence, driving the air towards the lower pressure off the coast.This suggests that daytime urban air is transported toward the mountains and over the passes year-round, but winter evenings are more efficient at flushing urban air from the mountains to the coast.The two observed regimes for total column NO 3 , model-like and non-modeled behavior, were consistent with the Lu and Turco analysis of local air flow.Between mid-October and April, when only model-like behavior was observed, there was little evidence for tropospheric contribution to the total column NO 3 .In the summer, the daytime onshore component coupled with weakly organized evening flow would often lead to urban air being advected to the TMF site, resulting in large variations in detected NO 3 .A signature of urban influence on column NO 3 was observed for 14 out of 31 d, for data collected between mid-April and mid-October in 2003 and 2004.In Fig. 5, our model-like behavior data are shown compared to other measured column measurements using lunar occultation with grating spectrometers, from Solomon et al. (1989) and Aliwell and Jones (1996b).Data from Solomon et al. (1989) were taken from Fig. 10 of their paper and reduced by 18% to account for the updated NO 3 cross section of Yokelson et al. (1994) at 240 K, which was not available when the paper was published.The result of Aliwell and Jones (1996b) and our data are in good agreement.While some of our data and that of Solomon et al. (1989) have overlapping error bars, the majority of their data is roughly 1-2×10 13 molec cm −2 below the TMF columns.Solomon et al. (1989) confirmed most of their data was primarily stratospheric NO 3 by analyzing the dependence of the NO 3 slant column on the lunar zenith angle (LZA) near the horizon (LZA>80 • ) (Solomon et al., 1989).A tropospheric NO 3 signal would grow much faster than the stratospheric NO 3 at high lunar zenith angles from slant path increases.We were not able to use this method to determine the tropospheric contribution because of pointing system view angles limited to less than 80 As shown in the figure, the short time-scale features in the 29-31 August LP-DOAS data are reproduced in the column data, implying a large boundary layer contribution to the column on those days.Some features seen in the column measurements were not present in the surface concentrations, which could be due to changing thickness of the polluted layer.
In contrast, data from 27-29 September 2004 had a much smaller contribution from the boundary layer.The LP-DOAS instrument confirmed that there were low NO 3 concentrations at the surface, as show in Fig. 6.This period coincided with a Santa Ana wind event, characterized by a northerly downslope flow that advected dry desert air mixed with air from aloft over the measurement site.This circulation is driven by a high pressure system centered north of Southern California.This flow of air from the north over the mountains and through the passes to the LA basin drives wind speeds of 46 km/h and gusts in excess of 90 km/h, carrying urban pollution offshore and away from Table Mountain Facility.Air quality measurements of surface NO 2 , CO, and O 3 from Air Quality Management District (AQMD) stations (California Air Resources Board, 2007) in Victorville (14306 Park Avenue), Azusa, downtown Los Angeles (North Main Street), and West Los Angeles (Westchester Parkway), positioned progressively from the desert in Victorville towards the ocean, verifies that low concentrations of surface urban pollutants were found in the Mojave Desert and into Los Angeles County, and that the diurnal cycle for these chemicals was disrupted for this time period (see Fig. 7).values in April and May.Results from 2004 deviated from the other runs with larger modeled columns for January that decreased to climatological values from March onwards.The TMF seasonal averages for total column NO 3 are listed in Table 2.

Tropospheric model
From the results of the GEOS-Chem 3-D chemical transport model, we investigated the expected range of tropospheric NO 3 column abundances for specific geographic regions.GEOS-Chem could not be used for quantitative comparisons with TMF observations since the grid size is too large to resolve the local meteorology and the detailed transport of pollution from the L.A. Basin.In order to understand the range of the expected NO 3 variability from the model, column abundances are compared for three In addition, we calculate the NO 3 column for the northern midlatitude zonal mean (29-45 • N).The regions were compared for six evenings that coincide with data collection (the evenings of 28-30 August 2004 and 27-29 September 2004).Total columns as well as the partial columns from the boundary layer and free troposphere were calculated.The boundary layer defined by the model for each time step was not used since the boundary layer is shallower during night time and does not reflect the pollution that was distributed throughout the boundary layer in the daytime.Instead, a column was constructed by setting the threshold to the maximum altitude of the top of the boundary layer for that day.This column is labeled as the "maximum boundary layer", with the difference of the total with this quantity labeled as the "minimum free troposphere".
The time evolution of tropospheric NO 3 , shown in Figs. 8 and 9, varied over the different regions but in most cases there was a sawtooth pattern not unlike that for the stratosphere: daytime photolysis with negligible NO 3 , a nearly linear rise in NO 3 column over the evening followed by a rapid decrease in NO 3 column at sunrise.The 20209 Figures

Back Close
Full maximum boundary layer column tended to mimic the total column shape, but for all cases the minimum free tropospheric column consistently had the sawtooth pattern.The diurnal averages, calculated with the same method as with the measurements, are summarized in Table 3.The model grid cell over Los Angeles has large tropospheric NO 3 columns from anthropogenic NO x with roughly 20-70×10 13 molec cm −2 , while the data at a midlatitude mountainous region (Western Colorado) and without nearby urban sources (Northern Midlatitude Pacific) had significantly smaller tropospheric columns (4-5×10 13 molec cm −2 ).The minimum free tropospheric column was at its lowest over the mountain region, (1×10 13 molec cm −2 ).The midlatitude zonal mean value is 6×10 13 molec cm −2 .These values are consistent with the difference between our TMF measurements of total column NO 3 and model amounts of stratospheric partial column NO 3 .As discussed below, this suggests that there is NO 3 in the free troposphere that can reside for days in substantial concentrations, indistinguishable from stratospheric NO 3 based solely on the time evolution (diurnal variation) of the measured signal.

Model and measurement comparison
The measured NO 3 columns along with results from 1-D stratospheric model constrained by measured temperatures and ozone concentrations from TMF showed a seasonal trend with higher NO 3 in the summer months.Measurements and model results from January-March 2004 were consistent within error bars as seen in Fig. 4, even duplicating the decrease in mean NO 3 column over these months not seen in the other model results.However, as seen in Table 2, the measured data are consistently larger than the modeled data by over 2×10 13 molec cm −2 for both summer and winter averages.This suggests that there is significant NO 3 in the troposphere; the stratospheric model correlated well with measured stratospheric NO 3 columns from SAGE III, as discussed in Appendix A, therefore we believe the model is reliable.
While it is clear that our NO 3 columns exhibiting non-modeled behavior has contributions from the boundary layer, even days with established low surface NO 3 20210 Introduction

Conclusions References
Tables Figures

Back Close
Full concentrations, such as 27-30 September 2004 (Fig. 3b), had mean columns of NO 3 that were on average 2×10 13 molec cm −2 more than the model amount.Low levels of NO 3 were detected by LP-DOAS in September, but the measured columns were still on average 2×10 13 molec cm −2 greater than the modeled column, which is more NO 3 than a uniform troposphere with 3 ppt of NO 3 , the detection limit of the instrument.
Other measurements have determined there can be significantly larger concentrations of NO 3 above the surface in the upper boundary layer and lower free troposphere, using zenith sky measurements at sunrise compared to surface DOAS measurements (Allan et al., 2002).GEOS-Chem results for the northern midlatitude band (29-45 • N) for the six days in August and September 2004 highlighted in this study found that the average minimum free tropospheric column was 6×10 13 molec cm −2 while the average maximum boundary layer column was 3×10 13 molec cm −2 .For this case, significant NO 3 existed in the free troposphere with smoothly varying diurnal variation that is indistinguishable from modeled stratospheric NO 3 diurnal variation.This result indicates there is a sizable contribution to the column of NO 3 from the troposphere above the boundary layer.Brown et al. (2007a) reached similar conclusions based on aircraft measurements over the east coast of the United States.

Conclusions
We have measured the diurnal variation of the NO 3 column over Table Mountain Facility, California (34.4 • N, 117.7 • W), using ground-based visible absorption spectroscopy of moonlight.We observed two sets of behavior during the steady state phase of the evening: one described as "model-like behavior" followed the expected slow linear increase (mean columns of 5-9×10 13 molec cm due to variability in the stratosphere and are attributed to boundary layer sources.
Comparison to results from a 1-D photochemical model with temperature and ozone profiles taken from onsite lidar instruments showed that for the most part we measured more NO 3 than found using the model.The model compares well with stratospheric column NO 3 reported by SAGE III.These comparisons suggest significant contributions to total column NO 3 from the free troposphere at all times, with the tropospheric contribution exhibiting a diurnal pattern similar to the stratospheric column.This is supported by simultaneous surface measurements with LP-DOAS in September 2004, and results from the global tropospheric chemical and transport model, GEOS-Chem.

Appendix A
The SAGE III (Meteor-3M) instrument (SAGE III ATBD, 2002) retrieved NO 3 concentration profiles from 20-60 km and O 3 concentration profiles from 15-50 km by satellite lunar occultation at moonset or moonrise.The retrieval process used temperature and pressure profiles from meteorological data from the National Centers for Environmental Prediction (NCEP) (Kalnay et al., 1996).These NCEP temperature and is covered by the SAGE III measurements.The rest of the chemical inputs were the same as described for the TMF model runs.
The NO 3 profiles were integrated over the 18-60 km altitude range to determine a stratospheric partial NO 3 column comparable to the one calculated from SAGE III data.These columns were further reduced to a mean column and standard deviation calculated over the nocturnal steady state period, as done with our measurements.
The modeled stratospheric NO 3 columns are plotted against the values derived from integration of the SAGE III NO 3 vertical profiles in Fig. 10.The bulk of the data points cluster along the one-to-one line.Since both sets of data have significant uncertainties, we used a linear fit that considered both x and y errors (the details are described in Wang et al., 2008), rather than a standard linear fit that considers only errors in y.The uncertainty in the modeled NO 3 column due to uncertainties in input temperature and O 3 profiles was estimated to be 0.7×10 13 molec cm −2 , and uncertainty in the measured NO 3 column derived from the quoted uncertainties in the SAGE III NO 3 retrieved profiles was 0.2×10 13 molec cm −2 .This resulted in a linear fit with a slope of 0.92±0.02and an intercept of 0.42±0.5×10 13molec cm −2 , with a reduced Chi squared, χ 2 red , of 4.3.The reduced Chi squared is the χ 2 statistic normalized by the degrees of freedom, with a value of one indicative of a good fit (residual of fit and data is same order as errors), much less than one an indication of overestimated errors, and much greater than one of underestimated errors.From these fit results we assert that the modeled stratospheric NO 3 columns are consistent with the SAGE III measured columns.

Appendix B B1 Sensitivity Study of modeled NO 3 on input parameters in 1-D stratospheric model
We changes were applied to atmospheric profiles using the URAP climatology for all twelve months.Other chemical profiles not provided by URAP used the same sources as described for the TMF model runs.
We found that out of the various input parameters to the 1-D model, the NO 3 column is most sensitive to temperature and O 3 .An increase or decrease of temperature by 5 K in the model resulted in change in the mean NO 3 column by 36 or −25%, respectively.A linear response was observed below 30 km; from 30-45 km a strongly nonlinear response was observed.A change of ±5% in the ozone concentration resulted in a change in the mean NO 3 column by ±5%, with no altitude dependence in the sensitivity from 18-50 km.This directly proportional, linear relationship between NO 3 and ozone concentration occurs when NO 3 concentration is in steady state.NO y also had a small effect, with the ±5% change in NO y concentration resulting in a ±1% change in the NO 3 column.The effect on NO 3 from changes in the input concentration of other chemicals was negligible.These sensitivity coefficients are summarized in Table 4.

B2 Error propagation of reaction rates to NO 3 columns
Sensitivity of the NO 3 column to the errors in the rates of reactions relevant to NO 3 concentration was probed.The reaction rates of NO 3 creation from NO 2 +O 3 (R1), thermal decomposition of N 2 O 5 , and N 2 O 5 formation (R2), were individually changed by their quoted error (Sander et al., 2003) for the SAGE III runs described in the model description, Sect.3.1.The sensitivity coefficients are summarized in Table 4.The greatest sensitivity of the NO 3 column to reaction rate errors was found for the NO 3 formation reaction from NO 2 +O 3 .The root-mean-square variation for all rate changes that increase NO 3 was 27%, and 32% for changes that decrease NO 3 column.These percent changes were applied to the TMF climatology and are plotted as the pair of dotted lines in Fig. 5.The plotted range of NO 3 columns due to reaction rate errors was of similar magnitude as the range of NO 3 values calculated from the variability observed by the TMF lidar.

Conclusions References
Tables Figures

Conclusions References
Tables Figures

Back Close
Full  Full  and the rate constants were varied by their quoted error limits (Sander et al., 2003).
NO 2 + O 3 → NO 3 + O 2 .(R1) NO 3 + NO 2 + M N 2 O 5 + M. Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | 3 m focal length, f/4 imaging spectrometer (Acton 300i) with a 1200 g/mm blazed grating.A slit width of 150 µm was used, resulting in 0.4 nm (FWHM) spectral resolution.Wavelength calibration for the spectrometer was obtained by observing a neon Penray lamp mounted on the inside of the observatory dome.The shape and line width of these emission lines also provided the instrument lineshape function.The spectrometer was equipped with a 1024×255, back-illuminated CCD detector temperature stabilized with circulating coolant.The pixel spacing of the CCD was Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | showed no dependence of cross Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | to explore the spatial and temporal variability of NO 3 in the troposphere for a few days in August and September 2004, coinciding with our acquisition dates.The GEOS-Chem model (version 20204 Discussion Paper | Discussion Paper | Discussion Paper | Fig. 4. For 26 of the 40 d of analyzed data, the NO 3 columns followed model-like behavior (open symbols in Fig. 4).Within this subset of data a seasonal variation was observed, more clearly shown in Fig. 5.The NO 3 mean column averaged over summer Introduction Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | • .Surface concentration measurements of NO 3 were made with the UCLA LP-DOAS instrument during the August and September 2004 measurement periods.The results are shown along with the NO 3 column amounts measured by lunar occultation in Fig. 6Discussion Paper | Discussion Paper | Discussion Paper | Model results using averaged TMF temperatures over three different time periods are shown in Fig. 4: monthly averages from 2003, from 2004, and over a ten year period, 1988-1997.All three TMF model results exhibited a seasonal variability with higher values during the summer.Results from 2003 followed those of the ten-year average, with November through January having the lowest values of the year, and the highest Discussion Paper | Discussion Paper | Discussion Paper | different locations: TMF (mountainous region with nearby urban pollution sources, 33-35• N, 241.25-243.75 • E), Western Colorado (northern midlatitude mountain area, 37-41 • N, 251.25-253.75 • E), and the Northern Midlatitude Pacific Ocean (no urban sources or orographic influences, 29-45 • N, 178.75-228.75 • E).
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | −2 for 26 out of 40 d), and the other, called "non-modeled behavior", showed large departures from model behavior, often correlated with large mean NO 3 columns (6-22×10 13 molec cm −2 over 14 out of 40 d) and large standard deviations (up to 5×10 13 molec cm −2 ), mostly during May through early October.The changes in NO 3 column seen in the non-modeled data are not likely Introduction Discussion Paper | Discussion Paper | Discussion Paper | pressure profiles along with the SAGE III retrieved O 3 profiles were used as inputs for the stratospheric model and the resulting modeled NO 3 columns were compared with SAGE III NO 3 measurements.Available data spans from May 2002 to October 2005 (the mission was terminated March 2006), and 1184 data points from the latitude band between −70 and 70 • were used; local times of the measurements were between 22:00-02:00 LT.The 1-D stratospheric model described in the paper was run with inputs from SAGE III lunar O 3 measurements as well as the temperature and pressure data from NCEP reanalysis used in the SAGE III retrievals.The O 3 profile below 15 km and above 50 km was filled with a climatology based on D ütsch and the Middle Atmosphere Project, described in the model description in the body of the paper; this has little impact on the scientific interpretation of our results, since the altitude range of interest for NO 3 Discussion Paper | Discussion Paper | Discussion Paper | conducted a sensitivity study of the 1-D stratospheric model to determine to which input parameters the NO 3 column was most sensitive.Changes of ±5% concentration or ±5 K were applied to the entire vertical profile of an individual input parameter.These 20213 Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Table2.Summary of NO 3 seasonal mean columns (in molec cm −2 ) calculated for the subset of observations with "model-like behavior", and for results from the 1-D stratospheric model using profiles from the lidar at TMF with monthly mean fields from 2003 and 2004 and ten year climatological inputs.There were no measurements during the October through March months in 2003.The standard deviation of the mean columns is shown in parentheses.
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper |

Fig. 1 .Fig. 2 .
Fig. 1.Schematic of the instrument light path.(a) Light is collected by the primary of the heliostat (tracker), reflected down to the telescope on the first floor which conditions it to a 7 cm diameter collimated beam.(b) The light is then reflected to a condensing lens, past a shutter, order-sorting filter, and then into the spectrometer, recorded by a CCD.

Fig. 3 .Fig. 4 .Fig. 5 .Fig. 6 .
Fig. 3. Diurnal variation of NO 3 vertical column measured at Table Mountain, California for August and September 2004, along with calculated vertical columns from the 1-D stratospheric model (line).For August 2004 (a), three consecutive evenings of measurements are shown, the evenings of 28 August 2004 (open circle), 29 August 2004 (+), and 30 August 2004 (filled diamond).Also for September 2004 (b), three consecutive evenings of measurements are shown the evenings of 27 September 2004 (open circle), 28 September 2004 (+), and 29 September 2004 (filled diamond).The stratospheric model used monthly mean profiles from TMF lidar measurements from 2004 as input.Data from September, during the steady state nocturnal period, shows only every tenth point to avoid crowding the graph. 39

Fig. 6 .Fig. 7 .Fig. 8 .Fig. 10 .
Fig. 6.Coincident measurements of NO 3 vertical column (in red) using lunar occultation and surface measurements of NO 3 concentration (in black) using long-path DOAS.Many of the large features in the data taken in August 2004 occur in both datasets.Measurements in September 2004 verify that there were very low levels of NO 3 concentration at the surface the whole evening.

Table 1 .
Sources for the input parameters for each of the cases run on the 1-D stratospheric model.

Table 3 .
Mean, median, and standard deviation of column NO 3 from the GEOS-Chem 3-D global transport and chemistry model for the evenings of 28-30 August 2004 and 27-29 September 2004, which coincide with data collection days.Four regions were investigated, Los Angeles (contains TMF), Northern Midlatitude Pacific, Western Colorado, and the northern midlatitude band (30-45• N). the columns are calculated as total, the column below the maximum extent of the boundary layer for the previous day, and the column above.

Table 4 .
Summary of sensitivity coefficients due to variations in temperature, ozone, and relevant reaction rate constants.Temperature was varied by ±5 K, ozone by ±5%, NO y by ±5%,