Vertical profiles of nitrous acid in the nocturnal urban atmosphere of Houston , TX

Introduction Conclusions References


Introduction
Nitrous acid (HONO) is one of the most important hydroxyl radical (OH) precursors in the troposphere, in particular in environments with larger NO x levels (Finlayson- Pitts and Pitts, 2000).Its photolysis (Reaction R1) HONO + hν (λ < 400 nm) −→ OH + NO (R1) produces OH much earlier in the day than photolysis of O 3 and formaldehyde (HCHO) (Finlayson-Pitts and Pitts, 2000).HONO photolysis is thus the dominant primary OH source in the morning.Low levels of HONO can also considerably contribute to the OH budget later during the day.The contribution of HONO photolysis to the total OH budget during the day can reach up to 34% in urban and rural environments Published by Copernicus Publications on behalf of the European Geosciences Union.( Acker et al., 2006;Kleffmann et al., 2005;Aumont et al., 2003;Alicke et al., 2002).Despite the importance of HONO, its formation mechanism is yet not well understood.At night, when photolysis ceases, HONO mixing ratios increase gradually, reaching up to hundreds of ppt in rural areas and up to several ppb in polluted urban boundary layers (Stutz et al., 2010a;Kleffmann et al., 2006;Lammel and Cape, 1996;Platt, 1985).Different HONO formation pathways in urban areas have been identified as being important (Fig. 1).HONO is emitted directly into the atmosphere through combustion processes.Emissions depend on the type of engines and fuel, ranging from <0.01% to 1.8% of NO x (Kurtenbach et al., 2001;Calvert et al., 1994;Pitts et al., 1984;Kessler and Platt, 1984).HONO/NO x emission ratios from traffic tunnels in California and Europe have been measured from 0.29% to 0.8% depending on the traffic density and the type of vehicles (Kurtenbach et al., 2001;Kirchstetter et al., 1996).
HONO can also be formed in gas-phase and heterogeneous chemical reactions.The most significant gas-phase reaction that leads to HONO formation is the reaction of NO with OH.
This reaction is not significant during the night due to the low OH concentrations.Various other gas-phase HONO formation processes have been discussed and proposed in the past decades (Finlayson-Pitts et al., 2003;Li et al., 2008).However, none of these reactions is considered to be important at night.It is currently believed that heterogeneous processes involving conversion of NO 2 on humid surfaces are the major formation pathways of nocturnal HONO as described by the following stoichiometry.
2NO 2 (g) + H 2 O (surface) −→ HONO (g) + HNO 3 (surface) (R3) Laboratory studies show that HONO formation has a first order dependence on NO 2 and depends on relative humidity (Finlayson-Pitts et al., 2003;Kleffmann et al., 2004;Lammel and Cape, 1996).Other heterogeneous reactions such as NO + NO 2 + H 2 O−→2HONO, reaction of NO and HNO 3 on silica surfaces in the presence of water, heterogeneous conversion of NO 2 and other reactions such as on soot surfaces appear to play only a minor role in HONO formation (Saliba et al., 2001;Calvert et al., 1994;Ammann et al., 2005;Gutzwiller et al., 2002;Arens et al., 2001;Kalberer et al., 1999;Kleffmann and Wiesen, 2005).A recent study by Ziemba et al. (2010) proposed that heterogeneous reaction of HNO 3 on fresh organic aerosol can be a potentially significant source of HONO.Photoenhanced sources of HONO from reaction of NO 2 on organic surfaces such as aromatic hydrocarbons and humic acids have been proposed (Stemmler et al., 2006;George et al., 2005) but will not be further discussed in this paper.
While direct emissions of HONO predominantly occur near the surface, i.e. emissions by traffic, heterogeneous conversion of NO 2 to HONO can occur either on the ground and other surfaces such as buildings and plants, or on particles.In the former case one would expect clear vertical gradients of HONO, in particular during the night, when mixing is inefficient.In the case of NO 2 to HONO conversion on particles the vertical profile of HONO would follow the profile of the particles surface area, and would most likely lead to much weaker profiles.It is thus clear that the observation and interpretation of vertical concentration profiles at night can give crucial information on the nature of the surface responsible for the formation of HONO.Few attempts to use this information have been published thus far.Vertical gradients of HONO in the lowest 4 m altitude and fluxes of HONO were measured using differential optical absorption spectroscopy (DOAS) by Stutz et al. (2002) in a polluted rural area in Italy in 1998.Measurements showed that the compensation point for formation and deposition of HONO on grass surface occurred at HONO/NO 2 of ∼0.03, and that emission was a significant source of HONO in polluted areas (Stutz et al., 2002).Kleffmann et al. (2003) performed HONO gradient measurements between 10-190 m altitudes in a semi-rural area in Germany in 2001.Their study showed that HONO formation was dominated by processes on ground surfaces and HONO formation on aerosol surfaces was insignificant (Kleffmann et al., 2003).
Knowledge of the vertical profiles of HONO is also crucial to quantify its nocturnal budget and its contribution to the HO x budget in the morning.Here we present continuous measurements of vertical concentration profiles of HONO, NO 2 and O 3 in Houston, TX in 2006, with a focus on nighttime data.By using a one-dimensional (1-D) chemistry and transport model, the observations will be interpreted to determine the surfaces responsible for HONO formation and to test whether simple model parameterizations of NO 2 to HONO conversion can explain the observations.HONO absorption cross sections are ±5%, ±3%, and ±5%, respectively (Voigt et al., 2001;Stutz et al., 2002;Voigt et al., 2002).The systematic error of the DOAS spectrometer was determined to be <3% (Stutz and Platt, 1996).The observed mixing ratio ranges and detection limits are shown in Table 1.Information on supplemental measurements during TRAMP can be found in Lefer and Rappenglück (2010).

Observations
During the measurement period from 15 August to 20 September in 2006 during TRAMP, the best data on all three light paths were obtained during the first two weeks of September (1-14 September).Therefore, the discussion below is focused on this period (Fig. 3).All three trace gases established vertical gradients at night which disappeared in the morning when the boundary layer became well-mixed.

Meteorology
In-situ measurements of meteorological data, such as wind, temperature and relative humidity were recorded at the top of Moody Tower during the experiment.Wind speed and direction, temperature and relative humidity data are plotted in Fig. 3.During the measurement period from 1 to 14 September 2006, the prevailing winds were mostly northeasterly, with an average wind speed of ∼3 m/s.At night, winds tended to be weaker than during the day, with an average of 2.5 m/s.However, wind directions were mostly northeasterly or southeasterly.Relative humidity during these two weeks ranged from 20% to 90%, with, in general, larger relative humidity at night.Details of the meteorology conditions during TRAMP are discussed in Lefer et al. (2010).

NO 2 and O 3
NO 2 developed negative gradients during the night, with larger mixing ratios near the ground and smaller aloft, while positive gradients of O 3 were typically observed.This was due to the suppression of vertical mixing at night and NO emissions from traffic at the surface that converts O 3 to NO 2 as shown in Reaction (R4).These gradient measurements are consistent with previous field observations (Wang et al., 2006;Stutz et al., 2004b).
During stable and polluted nights (1-2, 2-3, 6-7, 7-8 and 11-12 September), NO 2 mixing ratios at the lower height interval reached above 40 ppb while NO 2 mixing ratios in the upper height interval were ∼1 ppb or less.Differences of NO 2 mixing ratios between lower and upper height intervals reached up to 50 ppb, as observed on 7 September at 04:30 CST.

HONO
During the night, HONO generally followed the trend of NO 2 in all three height intervals.Negative vertical gradients were observed during the night, with larger concentrations near the ground and smaller aloft, similar to previous observations (Wang et al., 2006;Stutz et al., 2002;Kleffmann et al., 2003).Maximum nighttime HONO mixing ratios of 2 ppb and above were observed in the lower height interval on 1-2, 7-8 and 11-12 September.Differences of HONO mixing ratios between the lower and upper height intervals reached up to 1.5 ppb during the nights of 1-2, 6-7, 7-8, 10-11 and 11-12 September.

Selected polluted nights (1-2, 7-8 and 11-12 September)
Three periods (1-2, 7-8 and 11-12 September) (Figs. 4a, 5a, 6a) were selected for a more detailed analysis.These nights were selected based on the distinct NO 2 and O 3 vertical   gradients, which indicate a stable atmosphere.HONO profiles were not considered in the selection of the nights.However, our observations during TRAMP show that nights with stable boundary layers result in strong HONO gradients.1-2 and 7-8 September had similar nocturnal variations of NO 2 , HONO and O 3 with other nights.Vertical gradients of all three trace gases started to develop after sunset and became strongest towards the end of the night.Strong positive gradients of O 3 and negative gradients of NO 2 were observed.O 3 at lower levels became depleted at the end of both nights, while the upper light path showed almost constant mixing ratios of about 60 ppb and 70 ppb throughout the nights.Based on the vertical trace gas profiles, temperature profiles (Day et al., 2010) and the results from our previous NO 3 lifetime analysis (Stutz et al., 2010b) we believe that the upper height interval was partially in the residual layer during both nights.
The residual layer began at 200 m and 160 m respectively for 1-2 and 7-8 September, assuming the residual layer was located right above the nocturnal ground inversion (Day et al., 2010).HONO mixing ratios at all three levels, lower, middle and upper, were positively correlated with NO 2 .The correlation coefficients R 2 were 0.80 and 0.90, respectively, for 1-2 and 7-8 September.Nighttime HONO concentrations were largest about an hour before sunrise and reached 2.15 ppb on 2 September and 2 ppb on 8 September in the lower height interval.During both nights, HONO mixing ratios in the upper height interval remained below 0.1 ppb.Differences in HONO mixing ratios between the lower and upper light paths reached 2.0 ppb and 1.7 ppb during the night of 1-2 and 7-8 September, respectively.The main difference between these two nights was that HONO and NO 2 mixing ratios in the lower height interval reached their maximum at the same time during the night of 1-2 September, but NO 2 peaked an hour earlier than HONO during the night of 7-8 September.
The night of 11-12 September showed a very different picture.HONO and NO 2 mixing ratios at the lower light path did not follow each other closely as was typically observed during other nights.Instead, NO 2 in the lower height interval decreased throughout the night, while HONO accumulated.However, HONO in the middle and upper interval showed a slightly positive correlation with NO 2 , with an R 2 of 0.59.Differences in HONO mixing ratios between the lower and upper light paths reached 1.9 ppb.In addition, vertical gradients of O 3 , HONO and NO 2 were observed already at the beginning of this night, while they were well-mixed at the beginning of the nights of 1-2 and 7-8 September and.This indicated the atmosphere was already stable before sunset.
Following the idea that the HONO to NO 2 ratios allow further analysis of NO 2 to HONO conversion (Stutz et al., 2004a) we also calculated this ratio (Fig. 4a, 5a, 6a).HONO/NO 2 was about 1-2% at the beginning of all three nights.HONO/NO 2 tended to be slightly larger at lower levels than aloft.There were no significant vertical gradients of the ratio except on 11-12 September, during the last few hours of the night.At the lower level, the ratios increased throughout the night and reached a maximum of 4% and 6%, respectively, on 2 September and 12 September at the end of the night when HONO reached a maximum.On 7-8 September, HONO/NO 2 increased to 4% before midnight then remained above 3% throughout the night.Compared with results from Stutz et al. (2004a), the HONO/NO 2 during TRAMP seems to be lower.While HONO/NO 2 of up to 9% were observed in various field experiments as reported in Stutz et al. (2004a), a maximum ratio of 7% was measured during TRAMP.It should be noted here that the values reported by Stutz et al. (2004a) were corrected for direct emis-sion of HONO and are thus expected to be larger.For a better comparison with the model we will use directly observed and modeled HONO/NO 2 , without any further correction.

Model description
To further understand the HONO field measurements taken during TRAMP 2006, we used a 1-D Radical Chemistry and Transport model (RCAT8.2).RCAT8.2 includes the gas-phase Regional Atmospheric Chemistry Mechanism (RACM) and 84 reactive species and 244 gas phase chemical reactions (Stockwell et al., 1997).RCAT8.2 considers temporally varying anthropogenic emissions, predominately from traffic between 0.1 m and 1 m altitude, as well as biogenic emissions at the ground and between 1 m and 10 m altitude.Emission ratios used in the model are based on (Kurtenbach et al., 2001(Kurtenbach et al., , 2002)).Surface uptake resulting in dry deposition and chemistry on aerosol are included as well.Chemistry in the liquid phase is currently not considered in the model.Details of the model are described in Geyer and Stutz (2004a) and Wong and Stutz (2010).

Parameterization of HONO chemistry
In RCAT8.2HONO is directly emitted from cars with an emission ratio HONO/NO x = 0.008 between 0.1-1 m altitude based on the tunnel measurement from Kurtenbach et al. (2001).Chemical HONO reactions currently considered in the model include HONO photolysis (Reaction R1), the HONO reaction with OH and the formation of HONO from NO and OH (Reaction R2).
HONO is formed from conversion from NO 2 on aerosol and ground surfaces, with a yield of 50% following the stoichiometry of Reaction (R3) and an NO 2 reactive uptake coefficient of 10 −5 (Trick, 2004).Deposition rates of HONO and NO 2 are calculated by the number of molecules in the lowest box colliding with the ground and the uptake coefficient.The uptake rate of HONO and NO 2 on aerosols is determined from the aerosol surface area density and the uptake coefficients according to the mass-transfer equation given by (Fuchs and Sutugin, 1971) assuming an average particle diameter of 150 nm.
HONO is taken up on aerosol and ground surfaces with a reactive uptake coefficient of 10 −4 (Trick, 2004).The ground is assumed to be flat in the model and no buildings or trees are considered.The aerosol vertical profile is assumed to be uniform.Vertical transport of HONO, i.e. vertical HONO flux j HONO is calculated based on a flux-gradient relationship and eddy diffusivity (Eq. 1) (Geyer and Stutz, 2004a).Eddy diffusivity is calculated by the Monin-Obukhov theory with the Von Kármán constant κ, the friction velocity u * and the stability correction factor which is a function of altitude and stability through the Obukhov length L (Businger et al., 1971). where

Model initialization
The model was initialized with a well-mixed boundary layer, using observed concentrations after sunset at 18:30 CST on 1 September and 7 September (Table 2).For the period of 11-12 September, the model was initialized with the observed vertical profiles of NO 2 , HONO and O 3 and a stable boundary layer (  As previously described in Geyer and Stutz (2004a) nocturnal vertical profiles of most trace gases, and in particular O 3 and NO x , depend strongly on two factors: vertical stability and NO x emission rates.Both parameters are not known well enough for the TRAMP data to allow an accurate modeling of nocturnal chemistry.To ensure that we describe these two factors correctly before we attempt to simulate HONO profiles we adjusted NO x emission rates and vertical mixing at each hour to make the modeled temporal and vertical profiles of NO 2 and O 3 fit the measurement.Vertical mixing was adjusted indirectly by varying the Monin-Obukhov length, which was used to calculate the eddy diffusivity.It should be noted here that the model was not constrained by any observed trace gas concentrations after initialization.HONO profiles were then calculated based on the adjusted emissions and stability parameters using the chemistry described above.No attempt to adjust the model to the HONO observations was made.

Applicability of 1-D model
One of the obvious challenges in our approach is the use of a 1-D model that may have difficulties in situations where advection plays a role.Emissions from point-sources pose another obstacle for the 1-D model.However, unlike many previous comparisons between observations and model results our 1-D model was used to understand path-averaged LP-DOAS data.These observations were horizontally integrated over 4-5 km and vertically integrated over altitudes of 20-70 m, 70-130 m and 130-300 m.Therefore, any local effects such as freeways and small scale trace gas transport will be averaged and play a much smaller role than in comparisons with in-situ data.
The DOAS measurements are also representative for much larger areas.In Houston the light paths spanned approximately from downtown to the southern edge of the inner ring of Houston, bounded by Freeway 610 (Fig. 2).Since the Houston inner ring consists of urban and suburban areas with similar characteristics, the LP-DOAS observations should also be representative for the rest of the Houston inner ring.

Lateral transport
To investigate how trace gas advection could impact our modeling studies, we compared the vertical mixing timescale in the model with the observed horizontal transport timescales.The vertical mixing timescale in the model is calculated as τ z (z,t) = z κ•u * (t) , where z is the altitude, κ is the Von Kármán constant and u * is the friction velocity.The horizontal timescale was determined using the measured wind speed and assuming that our LP-DOAS measurements are representative for the Houston inner ring with ∼20 km diameter.The 1-D model can be used if the vertical mixing timescale is much less than the horizontal timescale, i.e. τ z τ x .In this case concentration changes due to advection are less than those caused by vertical transport and trace gas distribution in the lower atmosphere is governed by vertical transport rather than lateral transport.Typical vertical transport timescale from the ground to 100 m in the NBL is about 10 3 s and the horizontal transport timescale is about 10 4 s for the area inside the Houston inner ring, for which our LP-DOAS observations are representative.The comparison between the vertical and horizontal timescales showed that our 1-D model can be used to simulate the lower and the middle height intervals.However, care has to be taken in the interpretation of the model calculations of the upper interval (130-300 m) as advection may potentially play a role.This is in agreement with a previous study (Stutz et al., 2010b) that concluded that the upper height interval was frequently in the residual layer in Houston.

Point emission
Another challenge in the use of the 1-D model is upwind point emissions, for example from the petrochemical complex in the Houston ship channel area east of our measurement site.During easterly winds these emissions, which would be inhomogeneously distributed in the atmosphere, could reach our light paths.However, in-situ wind speed and direction measured at 70 m altitude at Moody Tower and radiosonde profiles (Day et al., 2010) showed that wind directions were not easterly during the three modeling periods and thus there was no significant influence of the ship channel emissions on our data.The only exception is 7-8 September between 03:30 to 04:00 CST in the lower and middle height intervals when an increase of about 10 ppb of NO 2 and 0.5 ppb of HONO in the lower interval, and a slight increase of NO 2 and HONO in the middle interval occurred.For our study we will thus treat the Houston inner ring as a metropolitan area where traffic is the dominant source of NO x and HONO emission (Fig. 2).

Sensitivity to emission and vertical mixing
Sensitivity studies based on the model run of 1-2 September were performed to investigate the dependence of NO 2 and O 3 mixing ratios on NO x emission and vertical mixing (Fig. 7).NO x emission, adjusted to yield agreement with the observations, showed the evening rush hours in the first few hours of the model runs, followed by low emissions throughout the night until the onset of the morning rush hours.Similar emission patterns were found in the modeling periods of 7-8 and 11-12 September.Variation of eddy diffusivity, also adjusted for best agreement between modeled NO 2 and O 3 , showed that the boundary layer was weakly stable during the first two hours and stable throughout the rest of the night.Similar eddy diffusivities were used during the night of 7-8 September.11-12 September had a less stable boundary layer since observations showed some mixing between the lower and middle height intervals.
NO 2 mixing ratios in the lower height intervals were very sensitive to changes in NO x emissions.A change of 20% in NO x emissions in the model run will lead to a significant change in modeled NO 2 mixing ratios.Sensitivity of NO 2 to NO x emissions decreases with altitude.As shown in Fig. 7, an uncertainty of +/− 20% in the NO x emission rates seems to capture the disagreement between the best modeled NO 2 mixing ratio and the observations on both light paths throughout the night.NO 2 concentrations did not seem to be very sensitive to variations in eddy diffusivity.However, eddy diffusivity values were small during the night.Thus changes of eddy diffusivity adopted in the sensitivity runs did not make a significant change in the stability or vertical mixing of the boundary layer.Sensitivity of NO 2 to eddy diffusivity also decreases with altitude.
The sensitivity runs also show that our method of adjusting NO x emissions and vertical mixing seems to yield reasonable and fairly accurate results.We thus conclude that the adjusted NO x emissions and vertical mixing parameters provide accurate input parameters for our model runs to study HONO chemistry.

NO 2 and O 3
The modeled NO 2 and O 3 mixing ratios for all three nights (1-2, 7-8 and 11-12 September) agree well with the observations, except for NO 2 in the upper height interval which tends to be larger than the measurements on 1-2 and 7-8 September.Because the upper height interval was most likely in the residual layer during these nights, the weak mixing into the residual layer and/or the possible air mass exchange above 130 m was not simulated accurately in the model.On 11-12 September, mixing of NO 2 at the lower and middle height intervals occurred in the second half of the night.This was simulated until the two were well-mixed, but the modeled NO 2 again developed gradients while the observed NO 2 between the lower and middle light path intervals remained well-mixed.

HONO
The modeled HONO mixing ratios and profiles showed good agreement with the observations for 1-2 and 7-8 September, except for the last few hours of the night when HONO was under-predicted by the model.HONO mixing ratios in the upper height interval were greater in the model than the observations, as the upper height interval was often located in the residual layer.The elevated HONO at this altitude was probably due to an overestimate of vertical mixing into the residual layer by the model and/or transport processes in the residual layer which are not represented well in the model.
Comparison of HONO/NO 2 from the model with the observations allows further analysis of the HONO formation processes and their altitude dependence.Our model calculation showed a surprising increase of HONO/NO 2 with altitude on 1-2 and 7-8 September, which was not reflected in the observations.This suggested that there was either too little NO 2 and/or too much HONO in the middle and upper altitude intervals in the model.Sensitivity studies revealed that increasing NO 2 in the model leads to NO 3 and N 2 O 5 mixing ratios which were considerably larger than the observations (Stutz et al., 2010b).We thus concluded that the increase of HONO/NO 2 aloft was due to excess HONO instead of the lack of NO 2 .Attempts to decrease the HONO/NO 2 at middle and upper altitudes by adjusting the uptake coefficients of NO 2 and/or the conversion factor of NO 2 were not successful.Further analysis of HONO chemistry showed that the elevated HONO/NO 2 was most likely due to an underestimate of the loss of HONO on the aerosol.Modeled HONO/NO 2 in the lower height interval for the three periods was slightly lower than the observations, indicating that HONO was formed at the ground more efficiently than implemented in our original model.Sensitivity studies showed that the underestimated HONO/NO 2 at lower altitude interval was due to the excess loss of HONO by deposition on the ground.
To consider the excess loss of HONO at the ground and the inadequate loss of HONO on aerosol in the model, refined model runs were carried out in which the HONO uptake coefficient at the ground was decreased by a factor of five to 2×10 −5 , while the HONO uptake on the aerosol was increased ten times to 10 −3 .In our original model runs HONO uptake coefficients on aerosol and ground were 10 −4 , based on measurements on asphalt by Trick (2004).In reality, the uptake coefficients are likely to be different on aerosol and ground.HONO uptake coefficients on the order of 10 −4 to 10 −2 were reported by several studies (Yu et al., 2009).HONO uptake coefficients on sulfate particles on the order of 10 −3 were reported by Harrison and Collins (1998).The HONO uptake coefficient used in our refined model runs thus seems to be within the range of published values.
The refined model results (Figs. 4c,5c,6c) showed an increase of HONO mixing ratios at lower altitudes in all three runs compared to the original modeled results, leading to a better agreement between model and observations.There was no change in the modeled NO 2 and O 3 mixing ratios www.atmos-chem-phys.net/11/3595/2011/Atmos.Chem.Phys., 11, 3595-3609, 2011 in the refined model runs.On 1-2 September, the modeled HONO/NO 2 in the lower and middle height intervals agreed very well with the observations until the last few hours of the night when the observed HONO mixing ratios increased and peaked.At 05:30 CST, the observed HONO/NO 2 reached 5% while the modeled ratio was only 2.5%.This was most likely due to a process that leads to large HONO/NO 2 such as HNO 3 conversion on fresh soot aerosol observed by Ziemba et al. (2010) during rush hour.On 7-8 September, the modeled HONO in the lower height interval was slightly smaller in the middle of the night, but it showed a good agreement with the observations during the last few hours of the night.HONO reached a maximum of 1.7 ppb in the model while the observed mixing ratio was 2 ppb.HONO/NO 2 in the lower height interval was better predicted on 7-8 Septenber in the refined model run.In both modeling periods, 1-2 and 7-8 September, the HONO mixing ratios at upper altitude intervals were larger than the observations as the upper light path was located in the residual layer which was not captured accurately by the model.
In general, the refined model improved the agreement between model and observations on 1-2 and 7-8 September.However, the model was not able to reproduce the opposite temporal behavior of HONO and NO 2 in the lower height interval during the night of 11-12 September.The modeled HONO mixing ratios in the lower height interval remained constant at about 0.7 ppb and decreased slightly towards the end of the night while the observed HONO increased steadily throughout the night to 2 ppb at the end of the night.In addition, the modeled HONO/NO 2 remained almost constant at around 2% throughout the night and did not predict the increase of HONO/NO 2 to 6%, as observed.This night differed in its behavior considerably from the other two.11-12 September was the only night throughout the two weeks of measurements (Fig. 3) with opposite temporal behavior of HONO and NO 2 .The decrease of NO 2 at the lower light path throughout the night was due to the low O 3 at the beginning of the night.As surface NO 2 was predominately formed from the reaction of ozone with NO (R4), the NO 2 surface source was small in this night.This interpretation is consistent with the very low NO 3 mixing ratio observations (not shown).It is not clear what caused the increase of HONO with decreasing NO 2 at lower altitude intervals.One possible explanation was the more efficient HONO formation at larger relatively humidity as reported by Stutz et al. (2004a), as this night was more humid than the other two.The average relatively humidity at 70 m was about 85% during this night and 70-75% during the other two periods.In addition, there was precipitation at the beginning of the night of 11-12 September according to the National Climate Data Center (http://www.ncdc.noaa.gov/oa/climateresearch.html).We therefore conclude that precipitation processes wetting the ground and other surfaces during the night of 11-12 September lead to substantially different HONO chemistry, which is currently not implemented in the model.Because our model did not describe the HONO correctly, this night will not be further quantitatively analyzed below.
In the following we will focus on the refined model runs of 1-2 and 7-8 September, as the refined model runs describe the HONO chemistry better.To visualize the vertical mixing ratio profiles and their development throughout the night, the vertical profiles of HONO and NO 2 mixing ratios at 19:30, 00:30 and 05:30 CST were extracted from the model simulations for the two modeling periods, 1-2 and 7-8 September (Fig. 8).HONO slowly accumulated in the lowest 200 m of the atmosphere, with pronounced vertical mixing ratio profiles.The actual vertical profiles should be slightly stronger because the modeled NO 2 and HONO mixing ratios were overpredicted in the upper height interval.Vertical HONO gradients, i.e. comparing 20 m to 300 m altitude, increased throughout the night with changes occurring mainly below 100 m, especially on 7-8 September.Most of the HONO was located in the lowest 200 m of the atmosphere.The modeled HONO mixing ratios below 20 m, which were not observed by the LP-DOAS, showed that HONO near the ground (at 0.55 m) increased from 1.8 ppb at 19:30 CST to 2.4 ppb at 00:30 CST and to 4.6 ppb at 05:30 CST during the night of 7-8 September.This is significantly larger than the modeled and observed values of the lower light path of 0.45 ppb at 19:30, 0.45 ppb at 00:30 CST and 1.7 ppb at 05:30 CST.The vertical profiles of HONO indicated that HONO near the ground (at 0.55 m) can be considerably larger than values from only 20-70 m above the surface.Consequently care has to be taken when extrapolating surface measurements of HONO to the entire boundary layer.

Discussion
Vertical profiles of HONO mixing ratios can give information on the nature of the surface responsible for the formation of HONO.Since HONO developed negative gradients throughout the night and its concentration increased near the ground throughout the night, one can interpret that the source of HONO is near or at the ground.It is however difficult to quantify the contributions of different HONO formation and loss processes and their altitude dependence based on vertical concentration profiles alone.

HONO formation and loss rates profiles
To further understand HONO chemistry and to investigate which processes are important for HONO in the nocturnal boundary layer, modeled HONO formation and loss rates were computed and the vertical profiles of these rates were extracted for the model period of 1-2 September at 19:30, 00:30 and 05:30 CST (Fig. 9).The rate of change of HONO depends on several factors as shown in Eq. (2).HONO is formed at the ground and potentially also on aerosol surfaces from NO 2 conversion (P ground +P aerosol ), emitted from traffic (E) and released through the reaction of NO and OH (P NO+OH→ HONO ).Vertical transport of HONO ( ) can be both a source and a sink.HONO in the nocturnal boundary layer can be lost through reaction with OH and deposition at the ground and uptake on aerosol surfaces (L ground + L aerosol ).dHONO dt = P ground + P aerosol + E + P NO+OH → HONO (2) The rate profiles (Fig. 9) show that HONO formation and loss processes were largest at or near the ground.The dominant formation pathway of HONO was through NO 2 conversion at the ground, followed by direct HONO emission.HONO formation and uptake on aerosol surfaces occurred over the lowest 300 m with rates slowly increasing towards the ground, where NO 2 and HONO mixing ratios are largest, assuming that the aerosol profile was uniform over the NBL.The formation of HONO on aerosol played a less important role.However, please note that aerosol surface profile in our model is oversimplified.The model predicted about ∼10 5 cm −3 nocturnal OH from NO 3 reactions with VOCs and ozonolysis of alkenes near the surface (see also Geyer and Stutz, 2004b).Production of HONO through reaction of NO and OH thus mainly occurred in the lower nocturnal boundary layer where NO mixing ratios were largest and decreased strongly with altitude.HONO formation by reaction of NO and OH was a minor production pathway of HONO in the NBL.Among all the loss pathways, ground deposition of HONO was the largest.Chemical removal of HONO by OH, which decreased exponentially with altitude, was negligible.HONO uptake on aerosol occurred over the lowest 300 m with rates increasing toward the ground where HONO concentrations were largest but, averaged over the NBL, this process was not as large as removal by ground deposition.Finally, HONO rate profiles showed that HONO vertical transport redistributed HONO over the boundary layer and in general transports HONO upward, leading to a loss of HONO near the ground and a gain of HONO at higher altitude.In all model runs, vertical transport was the dominant source of nocturnal HONO above 20 m altitude in average.Since the upper height interval was partially located in the residual layer where advection could play an important role, advection could potentially be a significant removal pathway of HONO in the residual layer.

Integrated HONO production and loss
Nocturnal HONO formation and removal rates change significantly throughout the nights of the two modeling periods due to the changes of meteorological condition and traffic patterns.In order to quantify the overall contribution of each HONO formation and loss process, the integrated HONO production and loss process in the lowest 300 m throughout the night were calculated for 1-2 and 7-8 September (Table 3).
Vertically and nocturnally integrated HONO production ranged from 1.7×10 15 cm −2 to 1.8×10 15 cm −2 in the two modeling periods.Contribution to total HONO production was 65-74% by NO 2 conversion at the ground, 3% by NO 2 conversion on aerosol, 17-21% from traffic, and 6-21% from reaction of NO and OH.Integrated HONO loss ranged from 1.3×10 15 cm −2 to 1.4×10 15 cm −2 .Contribution of total HONO removal was 71-76% by ground deposition, 18% by aerosol uptake and 1% due to reaction of HONO with OH.Vertical transport redistributed HONO over the NBL and even the residual layer.Vertical transport to altitudes above 300 m contributed to 5-10% of total HONO loss in the lowest 300 m.Combining all the factors, the overall HONO net production was 2.7×10 14 cm −2 throughout the night of 1-2 September and 5.3×10 14 cm −2 for 7-8 September.Assuming that all the HONO formed at night photolyzed into NO and OH, these numbers represent the amount of additional OH being contributed by HONO photolysis to the OH budget in the following day.However, because the HONO mixing ratios in the model simulations were smaller than the observations during the last few hours of the night, the actual HONO formation during these two nights should have been even greater.The inclusion of the upper height interval in our integration calculations leads to up to 20% uncertainty in HONO formation and loss on aerosol surfaces,  8% uncertainty in HONO loss due to reaction with OH and <v1% uncertainty in other formation and loss process.Because these processes are relatively small, these lead to only 1% and 4% uncertainty in total integrated HONO formation and removal, respectively.

Dependence of net HONO formation at the ground on vertical stability and emission
As proposed in Stutz et al. (2002Stutz et al. ( , 2004a) ) the HONO/NO 2 system tends towards reaching a pseudo steady state (PSS) between HONO formation and loss.One would expect that in our case the air close to the ground was in this PSS.Our model analysis, however, identified the ground as a net source of HONO during the majority of time.Consequently, the system, and in particular the lowest box of our model was shifted away from the pseudo steady state by vertical mixing processes.One can thus expect that vertical mixing and emission both modulate the HONO to NO 2 ratio at the ground, HONO chemistry at the ground, and as a consequence also the integrated net HONO formation in the NBL.
Vertical mixing plays two roles for HONO chemistry at the ground.It controls the amount of NO 2 available for making HONO at the ground and transports newly formed HONO upwards, away from the ground thus affecting HONO ground deposition.For example, weaker vertical mixing in the NBL can lead to larger NO 2 and HONO at the ground and consequently HONO formation and deposition at the ground (Fig. 10).Sensitivity model runs with constant emission show that the impact of vertical mixing on net HONO tivity of integrated HONO formation and loss rates throughout the night at the ground with diffusivity at 10 m is plotted as a measure of vertical stability.34 Fig. 10.Sensitivity of integrated HONO formation and loss rates throughout the night at the ground with stability.Eddy diffusivity at 10 m is plotted as a measure of vertical stability.
formation can be categorized into two regimes, "transport sensitive" and "NO 2 sensitive".The "transport sensitive" regime occurs when vertical mixing is weak (in this case when K z at 10 m is less than 0.6 m 2 /s) (Fig. 10).During this regime, when vertical mixing becomes stronger, the net formation at the ground increases because vertical mixing transports HONO upward more efficiently leading to a greater decrease in deposition at the ground.This effect outweighs the simultaneous decrease of both NO 2 and thus HONO formation at the ground due to mixing of NO 2 into larger volumes.On the other hand, the "NO 2 sensitive" regime occurs when vertical mixing is efficient (in this case when K z at 10 m is greater than 0.6 m 2 /s) and the reduction of HONO production due to decreased levels of NO 2 is more substantial than the decrease of HONO deposition due to larger transport.Therefore net HONO production decreases in this regime, i.e. when vertical mixing is stronger.
Sensitivity runs were also carried out to test the effect of the emission strength of NO x and HONO on HONO chemistry at the ground.Under the condition of constant nocturnal vertical mixing, when the emission is increased, the model shows an increase of both HONO formation and removal at the ground due to the increase of both HONO and NO 2 , which leads to an increase of net HONO production at the ground.The opposite is true when emission decreases.Therefore, addition of HONO by direct emission leads to larger net HONO formation at the ground when vertical mixing is constant.

Conclusions
During two weeks of the TRAMP 2006 experiment in Houston, TX, (1-14 September, 2006) nocturnal vertical concentration gradients of HONO, NO 2 and O 3 were frequently ob-served by the LP-DOAS.HONO developed negative gradients with larger concentrations near the ground and smaller aloft.Maximum HONO mixing ratios of above 2 ppb were observed in the lower height interval (average 20-70 m) in the late night, while HONO was often below 100 ppt in the upper height interval (130-300 m).These observations thus indicate that care has to be taken in interpreting surface observations of HONO, as they are not representative for the entire nocturnal boundary layer.
Three nights (1-2, 7-8 and 11-12 September) were selected to perform modeling studies and compare to the observations.Our vertically highly resolved 1-D chemistry and transport model (RCAT 8.2) together with the parameterization of NO 2 to HONO conversion and HONO loss rate at the ground (Trick, 2004) reproduced the observed vertical concentration profiles of HONO, NO 2 and O 3 .However, the modeled HONO/NO 2 in the middle (70-130 m) and upper (130-300 m) height intervals was substantially larger than the observations, resulting in an increase of HONO/NO 2 with altitude in the model, that was not observed in the atmosphere.Analysis of HONO chemistry in the model indicated that the uptake coefficients of HONO were too small on aerosol and too large on the ground.Refined model runs with 10 times larger HONO aerosol uptake (10 −3 ) and 5 times smaller ground uptake (2×10 −5 ) showed a much better agreement of HONO/NO 2 with observations for 1-2 and 7-8 September.Better laboratory HONO uptake data is needed to confirm the validity of these uptake coefficients on urban surfaces and aerosol.
The HONO mixing ratios at lower and middle altitude intervals from the refined model showed very good agreement with the observations for 1-2 and 7-8 September, except during the morning rush hour.The greater HONO/NO 2 observed during morning rush hour was most likely due to additional conversion from HNO 3 on organic aerosol, as proposed by Ziemba et al. (2010), which is currently not included in our model.It is also possible that the HONO to NO x emission ratio varied throughout the night due to changes in vehicle fleet composition.Since there was insufficient information on fleet composition and vehicle type specific HONO/NO x emission ratios, sensitivity runs were performed to study the effect of a possible change in emission ratios.Varying the HONO/NO x emission ratios by ±0.5% from the original value of 0.8%, i.e. to 0.3% and 1.3%, led to a 20% change in HONO mixing ratios and HONO/NO 2 ratios in all three height intervals during the rush hour period.During the rest of the night, HONO mixing ratios in the lower height interval varied by 13-19%, and by 13% and 4% in the middle and upper height intervals, respectively.The higher emission ratios could explain the rush hour HONO on 7-8 September but were still too low to explain the rush hour observations on 1-2 and 11-12 September.As a HONO/NO x emission rate of 1.3% is higher than any published value (see for example (Kurtenbach et al., 2001;Kirchstetter et al., 1996)) we conclude that direct HONO traffic emissions were www.atmos-chem-phys.net/11/3595/2011/Atmos.Chem.Phys., 11, 3595-3609, 2011

Fig. 1 .Fig. 1 .
Fig. 1.Schematic figure of HONO chemistry in urban boundary layer, showing HONO formation from direct emission (purple arrow), gas-phase reaction R2 (cyan arrow) and heterogeneous reactions R3 (blue and red arrows).HONO loss processes such as photolysis (R1) and uptake on surfaces are indicated by black arrows.
Fig. 2. Light path arrangement of the LP-DOAS systems during TRAMP 2006.
telescope aimed consecutively at these retroreflectors in order to measure the atmospheric absorptions of HONO, NO 2 , and O 3 , between the University of Houston and downtown Houston.Details of the LP-DOAS instrument can be found in Alicke et al. (2002), Stutz et al. (2010a,b) and Platt and Stutz (2008).The absorption spectra were used to retrieve the path-average concentrations of the trace gases.The retrieved concentrations are horizontally averaged between the Moody Tower and downtown Houston and vertically averaged between the LP-DOAS height intervals (lower: 20-70 m, middle: 70-130 m and upper: 130-300 m) and are representative concentrations for the boxes shown in Fig. 2. The concentration in the upper box is determined by subtracting the concentration of the middle light path scaled by the geometric overlap from the upper light path.Therefore, the concentration in the upper box is weighted more towards downtown.The uncertainties of the NO 2 , O 3 , and

Fig. 3 .Fig. 3 .
Fig. 3. Overview of LP-DOAS observations of NO2, HONO and O3 mixing ratios (lower: 20-70 m, middle: 70-130 m and upper: 130-300 m) from 1-14 September 2006 during TRAMP in Houston, TX.Measurements of temperature, relative humidity and wind speed and direction at 70 m are also shown.Nighttime data are shaded in grey to differentiate from daytime observations.

Fig. 4 .Fig. 4 .
Fig. 4. Comparison of HONO, NO 2 , O 3 mixing ratios and HONO to NO 2 ratios between observations (left) and model results (middle and right) on 1-2 September 2006.(Mixing ratios at lower, middle and upper altitudes are plotted in black, blue and red respectively.)See El.Suppl.for expanded version of figure.

Fig. 5 .
Fig. 5. Comparison of HONO, NO 2 , O 3 mixing ratios and HONO to NO 2 ratios between observations (left) and model results (middle and right) on 7-8 September 2006.(Mixing ratios at lower, middle and upper altitudes are plotted in black, blue and red respectively.)See El.Suppl.for expanded version of figure.

Fig. 5 .
Fig. 5. Comparison of HONO, NO 2 , O 3 mixing ratios and HONO to NO 2 ratios between observations (left) and model results (middle and right) on 7-8 September 2006.(Mixing ratios at lower, middle and upper altitudes are plotted in black, blue and red respectively.)See the Supplement for expanded version of figure.

Fig. 6 .Fig. 6 .
Fig. 6.Comparison of HONO, NO 2 , O 3 mixing ratios and HONO/NO 2 between observations (left) and model results (middle and right) on 11-12 September 2006.(Mixing ratios at lower, middle and upper altitudes are plotted in black, blue and red respectively.)See El.Suppl.for expanded version of figure.

Fig. 7 .Fig. 7 .
Fig. 7. Sensivity of NO2 mixing ratios in the lower and middle light paths intervals to change in NOx emission and eddy diffusivity.NOx emission and eddy diffusivity at 10 m are plotted as well.

Fig. 9 .Fig. 9 .
Fig. 9. Modeled vertical profiles of HONO rates at 19:30 CST (a, d), 00:30 CST (b, e) and 05:30 CST (c, f) during the night of 1-2 September 2006.Panels a, b, and c show the lower 300 m of the atmosphere.Panels d, e, and f provide a closer view of the lowest 30 m of the atmosphere.Numbers in these panels apply to the maxima of the respective processes in the lowest meter above the ground.

Table 1 .
Observed mixing ratio ranges and detection limits during TRAMP 2006.
Table2).Since no observations were available below 20 m, the lowest 9 boxes, which are located below 20 m in the model, were initialized by extrapolating the LP-DOAS observations in in lowest height interval.Because other trace gases such as SO 2 and HCHO did not show any vertical gradients at this time, all trace gases except NO 2 , HONO and O 3 were initialized with a constant vertical profile for this night.

Table 2 .
Initial concentrations in ppb for model simulations.

Table 3 .
Integrated HONO production and loss over the lowest 300 m throughout the nights of the modeling periods, 1-2 and 7-8 September 2006.Percentage contributions of each process are also shown.