Journal of Geophysical Research: Atmospheres

On alkyl nitrates, O3, and the “missing NOy



[1] We describe measurements of NO2, total peroxy nitrates (ΣPNs), total alkyl nitrates (ΣANs), and HNO3 using thermal dissociation followed by laser-induced fluorescence detection of NO2 at three continental locations. The ΣAN observations are unique and provide novel constraints on atmospheric photochemistry. At a rural site in California, measurements over a full annual cycle show that ΣANs are routinely 10–20% of NOy. At this rural site, at a suburban site in California and an urban site in Houston, Texas, both the absolute concentration of ΣANs and the fraction of the higher oxides of nitrogen (NOz) represented by ΣANs are greater than or equal to values reported in any prior observations. Although the contrast with prior observations is striking, we show that large abundances of ΣANs are consistent with simple chemical models of tropospheric ozone production and with the few prior comprehensive model studies. We also show that ΣANs are a large part, if not all, of the “missing NOy” reported in many prior experiments and emphasize that the ratios of ΣANs/NOz and of O3/ΣANs are especially useful for evaluating chemical models and comparing observations at different sites.

1. Introduction

[2] Photochemical production of tropospheric ozone on urban, regional, and global scales is regulated by coupled catalytic cycles involving NOx (NOx ≡ NO + NO2) and HOx (HOx ≡ HO2 + OH + RO2) radicals. Figure 1 shows the elements of the HOx and NOx catalytic cycles (reactions (1)(8)). The HOx cycle involves reaction of OH with a hydrocarbon to produce an organic radical. O2 immediately adds to this radical producing an organic peroxy radical (RO2). RO2 then reacts with NO in a two-channel reaction that either propagates the catalytic cycle producing NO2 and an alkoxy radical (RO) or terminates the cycles by producing an alkyl nitrate (RONO2). In most cases the alkoxy radical reacts with O2 to produce a carbonyl compound (which can undergo further oxidation) and a HO2 radical. HO2 then reacts with NO producing NO2 and regenerating OH, completing the catalytic cycle. The NOx catalytic cycle is coupled to the HOx cycle. The two NO oxidation steps in the HOx cycle produce two NO2 molecules which are photolyzed to regenerate the NO and produce oxygen atoms. The oxygen atoms combine with O2 to form ozone. The net effect (reaction (8)) of the cycles is to produce two ozone molecules.

equation image
equation image
equation image
equation image
equation image
equation image
equation image
equation image

While the basic form of these catalytic cycles is well known, demonstrating that we have an accurate and complete understanding of the chemistry, especially the dependence on the particular hydrocarbons and the role of various chain termination steps, remains elusive and challenging because direct measurements of key radical species and the molecules that are formed during chain termination reactions are not routine. Observational efforts to understand the chemical processes that govern the abundances of these radicals are at the center of efforts to improve air quality, to reduce the contribution of tropospheric ozone to greenhouse forcing, and to understand the past, present, and future global oxidative capacity. Chemistry also affects the spatial scale on which emissions of hydrocarbons and NOx have an effect. For example, oxidation of NO2 to produce water-soluble species such as HNO3 leads to relatively rapid, and consequently small spatial scale removal of NOx from the atmosphere. In contrast, production of insoluble nitrogen oxides such as the most common peroxy nitrate, peroxyacetyl nitrate (PAN), results in long range transport that affects NOx abundances on a global scale.

Figure 1.

Tropospheric ozone production mechanism, composed of coupled catalytic HOx and NOx cycles. For a typical hydrocarbon, RH, one loop around the HOx cycle drives the NOx cycle twice. Consequently, two new ozone molecules are created. Formation of RONO2 (alkyl nitrate) removes one NOx and one HOx molecule from each cycle. Reactions (1)(8) contain balanced reactions for the cycles depicted here.

[3] There are a number of mature technical approaches to observing individual peroxyacyl nitrate compounds and HNO3 [Parrish and Fehsenfeld, 2000]. In contrast, there are few methods for observing alkyl nitrates and none in widespread use [Day et al., 2002; Flocke et al., 1998b; O'Brien et al., 1995; Roberts et al., 1998; Schneider et al., 1998]. Production of alkyl nitrates occurs primarily through the reaction of RO2 with NO,

equation image

Although the relative rate of reaction (3b) versus reaction (3a) may be small for RO2 derived from many hydrocarbons, it is predicted that alkyl nitrate formation is one of the major pathways for converting NOx from radical form to inactive reservoirs near the Earth's surface. For example, Calvert and Madronich [1987] calculated that oxidation of an urban hydrocarbon mixture in NOx-rich conditions produces more than 100 different alkyl and 74 different hydroxyalkyl nitrates. Models not only predict a large variety of alkyl nitrate compounds in the atmosphere, they also predict that alkyl nitrates are a larger fraction of oxidized nitrogen than has been observed. For example, Atherton and Penner [1988] calculated that 5% of NOx could be converted to alkyl nitrates and Trainer et al. [1991] predict that organic nitrates, primarily hydroxy alkyl nitrates derived from the oxidation of isoprene (a biogenic hydrocarbon emitted by trees), can constitute as much as 12–26% of atmospheric nitrogen oxides. Although, it has been demonstrated that nearly 100 alkyl nitrates are present in the atmosphere [Schneider and Ballschmiter, 1999], no prior experiment in the midlatitude, continental boundary layer has shown that alkyl nitrates are a large fraction of NOy. If the formation of alkyl nitrates does indeed represent a substantial contribution to NOy, the large-scale impact of these nitrates is predicted to be significant. Chen et al. [1998] and Liang et al. [1998] estimated that during the summer 7% of NOx emitted in the eastern U.S. and 30% of the NOx in the entire U.S., respectively, are converted to hydroxyalkyl nitrate derivatives of isoprene. These hydroxyalkyl nitrates are thought to rapidly deposit because of their high water solubility [Shepson et al., 1996], thus permanently removing nitrogen oxides from the atmosphere before they are spread to global scales where they would increase tropospheric ozone.

[4] In this paper, we re-evaluate our understanding of the chemical processes that convert NOx to alkyl nitrates (RONO2). Using observations from a novel technique, thermal dissociation combined with laser induced fluorescence detection of NO2 (TD-LIF) [Day et al., 2002], we observe the total abundance of atmospheric alkyl nitrates (ΣANs). We describe observations at three different continental sites, discuss the relationship between ΣANs and O3, and suggest that ΣANs are an explanation for the “missing NOy” in the midlatitude, continental boundary layer. The “missing NOy” is an observation that the sum of measurements of individual nitrogen oxide compounds (ΣNOyi) are less than the total amount of nitrogen oxides (NOy ≡ HNO3 + ΣRO2NO2i + ΣRONO2i + HONO + NO2 + NO + 2xN2O5 + NO3(aerosol) + NO2(aerosol) + …) where the total is measured using a catalyst that converts all of the compounds to NO and is followed by detection of NO.

2. Experiment

[5] We make use of measurements of nitrogen oxides and various other species obtained near the University of California - Blodgett Forest Research Station (UC-BFRS) (1315 m asl, 38.9°N, 120.6°W) from October 2000–December 2001, in the city of La Porte, TX (8 m asl, 29.9°N, 95.0°W) during August 15 and September 15, 2000 and in the town of Granite Bay, CA (110 m asl, 38.7°N, 121.2°W) during July 18 through September 5, 2001. The three sites are expected to exhibit many typical characteristics of rural (UC-BFRS), suburban (Granite Bay) and urban (La Porte) photochemistry.

[6] The UC-BFRS site is 75 km downwind from Sacramento, CA in a sparsely populated region that is largely managed forestland. During the day, upslope winds bring the Sacramento urban plume to the site. At night, downslope flow brings clean, continental background air from the high Sierra. Detailed description of the site and analysis of transport and atmospheric chemistry in the region have been presented previously [Bauer et al., 2000; Dillon et al., 2002; Dreyfus et al., 2002; Goldstein et al., 2000; Lamanna and Goldstein, 1999]. At La Porte, TX, located 40 km southeast of Houston, TX and less than 20 km from the heart of the U.S. petrochemical industry along the Houston shipping channel, a large field study, TEXAQS-2000, included a comprehensive suite of chemical measurements and meteorological variables At Granite Bay, CA, a suburban town at the eastern edge of the Sacramento, CA urban sprawl, measurements of nitrogen oxides, hydrocarbons, ozone, and various meteorological parameters were obtained during a small collaborative study designed to test Lagrangian models of chemistry along the flow from Sacramento to the UC-BFRS.

[7] We use TD-LIF to measure the atmospheric abundances of NO2, the sum total of peroxy nitrates (≡ΣPNs), the sum total of alkyl nitrates (≡ΣANs), and HNO3. The TD-LIF technique and the two different instruments used in these experiments are described in detail in Day et al. [2002]. Briefly, TD-LIF takes advantage of the different characteristic O-N bond energies of NOy compounds. An ambient sample flows rapidly through an inlet and is split into two or four ovens which are either unheated for observing NO2, or heated to 180°C to cause thermal dissociation of ΣPNs (peroxy nitrates: RO2NO2, ∼85–105 kJ/mole), 350°C for ΣANs (alkyl nitrates: RONO2, ∼170 kJ/mole), or 550°C for HNO3 (205 kJ/mole). Upon dissociation, ΣPNs, ΣANs, and HNO3 produce NO2 with unit efficiency (1 NO2 per nitro-oxy group). The NO2 signal, observed by LIF, is the sum of the NO2 contained in all of the compounds that dissociate at or below the oven temperature. Differences between the NO2 signals observed simultaneously from channels heated to adjacent set points are used to derive absolute abundances of each of these four classes. There is less than 1% dissociation of ΣANs at the ΣPNs temperature and less than 2% dissociation of HNO3 at the ΣANs temperature. The dissociation of both compound classes is more than 99% complete before the next compound begins. The NO2 measurements are accurate to 5%. A detailed intercomparison between our LIF measurements and photolysis-chemiluminescence measurements is described in Thornton et al. [2003]. The instrument is capable of routine, continuous, autonomous, in situ measurements for all of the compounds we report. The detection limit at S/N = 2 for these experiments typically ranged from 30–90 ppt for a 10-s average. The precision, of course, varies with the mixing ratio of total NO2 in the two channels used for the subtraction. We conservatively estimate the accuracy of our measurements of ΣPNs, ΣANs, and HNO3 individually to be 15% with an average uncertainty in the zero of ±10 ppt. Most of this systematic error is associated with uncertainty in the NO2 standards and flow system we use to calibrate. This estimate leaves room for interferences we do not yet know about, although in Day et al. [2002] we describe potential interferences at the UC-BFRS and show they are less than 5%.

[8] We use the terms total alkyl nitrates, ΣANs, and total peroxy nitrates, ΣPNs, more broadly than in some previous literature. We use ΣANs to refer to all molecules with the characteristic alkyl nitrate bonding (RONO2), including di-nitrates and multifunctional nitrates (e.g., hydroxyalkyl nitrates - R(OH)ONO2). Similarly, since our detection method does not distinguish between simple peroxy nitrates such as HO2NO2 and CH3O2NO2 which are thought to be unimportant at the surface but we have been observed, using TD-LIF, in the upper troposphere [Murphy et al., 2002] and peroxyacyl nitrates such as PAN and PPN, which are observed throughout the troposphere, we us ΣPNs as a broadly inclusive term representing both categories.

[9] NO and NOy measurements were made at the UC-BFRS with NO-O3 chemiluminescence (Thermo Environmental Co. model 42CTL). A converter of molybdenum oxide (MoO) mesh, maintained at 325 °C to convert NOy species to NO, was located 40 cm downstream from the sample intake. The 30-s measurement cycle was split equally between NOy, NO (MoO converter bypassed), and background measurements. The accuracy is estimated to be better than ±10%. The uncertainty in the zeros was dependant on the time of day and season. During the fall and winter and during the night in all seasons, the uncertainty in the zeros was ±20 ppt. During the spring and summer daylight hours the uncertainty was highest around 10 AM and was ±100 ppt and otherwise was usually ±25 ppt at other times of day, except for the hours of 14–16 during summer when it was ±80 ppt.

[10] The inlets for the chemiluminescence and TD-LIF instruments were very similar and were designed to have a minimum residence time (35–70 ms) and to maximize the transmission of HNO3. HNO3 is known to be extremely difficult to sample under atmospheric conditions due to its high affinity for surfaces, especially under humid conditions. In addition to a short residence time, the inlets consist of entirely molded PFA (TD-LIF) or FEP (Chemiluminescence NO, NOy) Teflon tubing and fittings (∼40 cm long, 3 mm ID) upstream of the converter regions. Neuman et al. [1999] measured the HNO3 loss rate on 30 cm long, 5 mm ID PFA and FEP tubing at trace concentrations (∼5 ppb) and determined that under warm, dry conditions (relative humidity < 60% and T > 10°C) HNO3 was transmitted at nearly 100% efficiency. Under higher humidity or lower temperatures, they show that some of the HNO3 is reversibly adsorbed to the tubing walls. During May–September at UC-BFRS, temperatures usually did not drop below 10 °C and the relative humidity rarely exceeded 70% so we expect that HNO3 did not typically experience significant loss on the inlet. During the rest of the year, the weather is cooler and wetter and the HNO3 transmission efficiency may have been less than 100%. However, since both the NOy and TD-LIF instruments used nearly identical inlets, a comparison of the measurements by the two instruments should not be strongly biased by the inlet effects.

3. Large Abundances of ΣANs

[11] Observations of NO, NO2 ΣPNs, ΣANs, and HNO3 from October 2000 through December 2001 at the UC-BFRS show that ΣANs are a large contribution to reactive nitrogen in all seasons. Figure 2 offers three perspectives on the ΣANs observations. The solid black symbols are 7-day running medians which emphasize seasonal cycles, and the open symbols are half-hour averages whose range reflects the day/night variations. Figure 3 provides an indication of the daily and synoptic variability. Figure 2a shows the ΣAN mixing ratios of observations obtained between 2 PM and midnight, when the site is typically impacted by the Sacramento plume. In Figure 2b we show the ratio of the ΣANs to the ΣNOyi, where ΣNOyi ≡ NO + NO2 + ΣPNs + ΣANs + HNO3. NO was never a large contribution to the total, however, when NO measurements were not available we estimated it using NO2, O3 and a simple photochemical model. Typically NO is one-fifth of NO2 or less. Figure 2c shows ΣANs/NOz, where NOz ≡ ΣPNs + ΣANs + HNO3.

Figure 2.

ΣANs reported as 7-day running median overlaid on half-hour medians at the UC-BFRS from September 2000 through December 2001. Data are shown as (a) absolute concentrations, (b) the fractions of ΣNOyi, and (c) the fractions of NOz. The 7-day median helps to make clear the seasonal cycle which is otherwise obscured by daily and synoptic variability as shown in Figure 3.

Figure 3.

Mixing ratios (ppb) of individual NOy species and ΣNOy for August 6–15, 2001 at the UC-BFRS reported as half-hour medians. The dates indicate noontime of each day when NO and HNO3 peak. Instrument precision is small compared to the points.

[12] ΣAN mixing ratios at the UC-BFRS exhibit a strong seasonal cycle with a minimum in winter and a maximum in summer. The seasonal transition is sharp. Concentrations of ΣANs rise rapidly from 150 ppt at the end of April to 500 ppt by early June. During this transition, ΣANs/ΣNOyi increased from 10 to 20%. The fraction of NOy represented by ΣANs is also high (20%) in many winter months when the total NOy is very low. These observations of ΣANs/NOyi, ranging from 10–20%, are striking since prior reports of the contribution of the sum of individually measured alkyl nitrates to NOy are an order of magnitude lower, ranging from <1 to 3%, for sites in the midlatitude, continental boundary layer [Buhr et al., 1990; Flocke et al., 1998b; O'Brien et al., 1995, 1997; Ridley et al., 1990; Shepson et al., 1993; Thornberry et al., 2001]. Comparison of ΣANs/NOy between different sites is, however, difficult because the contribution of fresh NOx emissions to NOy causes wide variations in the fraction of NOy that is NOz. By contrast, ΣANs/NOz, as shown in Figure 2c, should be a quantity that is more readily comparable among different sites. At the UC-BFRS we observe ΣANs/NOz to be in the range 20–30% during the summer and to be more variable in winter. We observed ΣANs/NOz at 10–20% in Fall 2000. The few measurements we obtained in January 2001 were in the range 40–50%. Most of the springtime measurements were at 20% with some variation as high as 40%. A very high fraction of ΣANs/NOz of 30–50% was observed November–December 2001.

[13] The annual cycle of late afternoon measurements shown in Figure 2, as 7-day running medians, provides a long-term perspective of ΣANs. In Figure 3 we show mixing ratios of NO, NO2, ΣPNs, ΣANs, HNO3, and ΣNOyi at the UC-BFRS for ten days in August 2001, at half-hour time resolution. On the daily time scale, the effects of photochemistry and transport of the Sacramento plume affect the timing of peak concentrations of the various species. On average, during August 2001, NO and HNO3 have daytime maxima at local noon and a diurnal pattern that closely follows the solar illumination. NO2 and ΣPNs peak later in the afternoon at 20 and 18 hours, respectively, indicating that transport or photochemistry that is not directly proportional to sunlight are controlling factors. The diurnal variation in ΣAN mixing ratios was less regular and of smaller amplitude than that of other species. A relatively large background of ΣANs varied on time scales of several days. Variation on daily timescales was less than 50% of the peak mixing ratio.

[14] The daily patterns of ΣANs at Granite Bay, CA and at La Porte, TX were much more strongly influenced by local photochemistry. Figure 4 shows the average diurnal patterns of ΣANs at UC-BFRS, Granite Bay, and La Porte, reported as one-hour median bins. The data for UC-BFRS are from August 2001, a comparable photochemical period to the other two sites. Typically, at Granite Bay and La Porte, ΣANs were characterized by a noontime peak of 900 ppt with evening lows of 400 ppt at Granite Bay and 200 ppt at La Porte.

Figure 4.

Mixing ratios (ppb) of ΣANs at (a) UC-BFRS, (b) Granite Bay, and (c) La Porte reported as medians binned for each hour of the day. The UC-BFRS data used are from August 2001 only, whereas for Granite Bay and La Porte, data from the entire campaigns were used.

[15] ΣANs at all three sites comprised a large fraction of NOz. Results from the entire year at UC-BFRS were presented above (Figure 2) showing a range 10–50%. At Granite Bay, the 900 ppt peak of ΣANs typically observed during the day comprised 10–20% of NOz, and the 400 ppt typically observed during the night was 35–45% of NOz. At La Porte, 900 ppt of ΣANs comprised 8–20% of NOz during the day. Nighttime measurements of HNO3 were unreliable at La Porte because of high humidity making it impossible to quantify ΣANs/NOz at night.

[16] Prior experiments have focused on ΣANs/NOy as an indicator of the extent to which alkyl nitrate formation contributes to the processing of NOx [Buhr et al., 1990; O'Brien et al., 1995, 1997; Ridley et al., 1990; Shepson et al., 1993; Thornberry et al., 2001]. We discussed above the difficulty that proximity to NOx sources places on interpretation of this indicator. For comparison to this prior work we note that, ΣANs/ΣNOyi was highly variable at both Granite Bay and La Porte. Typical values were in the range of 2–12% and <1–10%, respectively, with median values of 5.6% and 3.5%. Even at these urban sites, the high end of the ΣANs/ΣNOyi ratios is without precedent. The lower end of the ΣANs/ΣNOyi measurements is in the range of prior reports of the sum of individually measured ΣANs at comparable sites.

4. ΣANs and the “Missing NOy

[17] The large abundances of ΣANs that we observe suggest that these compounds might represent a solution to a 20-year-old problem in atmospheric chemistry known as the “missing NOy.” In 1984, Fahey et al. [1986] measured NOy, NO, NO2, HNO3, NO3(aerosol) and PAN at Niwot Ridge, CO. They observed that NOy was greater than the sum of the individually measured species (ΣNOyi) by 45% in the summer and 12% in the fall. Based on the seasonal dependence of this difference and its correlation to the abundance of PAN, they suggested that other organic nitrates (including peroxy nitrates other than PAN and alkyl nitrates) might be the chemicals that were not observed.

[18] Many other experiments have produced similar results and no experiment within the nonurban midlatitude continental boundary layer has demonstrated NOy equal to ΣNOyi. For example, Ridley et al. [1990] describe July 1987 measurements of NOx, HNO3, NO3(aerosol), PAN, PPN, NOy, methyl nitrate, n-propyl nitrate, and 2-butyl nitrate. at Niwot Ridge, CO. They reported ΣNOyi was less than NOy by 200–600 ppt or 21–27%. The absolute concentrations of the shortfall correlated well with PAN. Buhr et al. [1990] attempted to improve on this study by adding measurements of the alkyl nitrates, ethyl nitrate, iso-propyl nitrate, n-butyl nitrate, 2-pentyl nitrate, and 3-pentyl nitrate, in July 1988 measurements at Scotia, Pa. They report “missing NOy” of 15% and correlations between “missing NOy” and organic nitrates. Williams et al. [1997] analyze a suite of measurements from Idaho Hill, CO during fall 1993 providing additional indicators that the “missing NOy” is chemically produced and not an artifact of poor instrumentation. They show that the “missing NOy” was correlated with O3 and aerosols and anti-correlated with temperature and the ratio NOx/NOy. The slope of O3 versus “missing NOy” was 10–30 O3 per “missing NOy.” They suggest that the shortfall is likely due to one or more unmeasured NOy compound with a photochemical source. Thornberry et al. [2001] describe the most chemically complete effort to compare ΣNOyi with NOy. They measured NO, NO2, HNO3, NO3(aerosol), NO2(aerosol), PAN, PPN, MPAN, HONO, C2–C5 monofunctional alkyl nitrates, some isoprene-derived nitrates, and NOy at a rural site in northern Michigan during summer 1998. The measurements suggest a larger role for HONO in the NOy budget than had been expected. They report the slope of ΣNOyi versus NOy was 0.81 ± 0.31 (2σ) with a 380 ppt positive offset (ΣNOyi > NOy). They suggest that this result may reflect a systematic interference in one of the measurements and approximately 20% “missing NOy.” Two recent long-time series measurements also provide evidence for “missing NOy.” Hayden et al. [2003] report a seasonal cycle of “missing NOy” at Sutton, Quebec, a rural site, 845 m asl, during 1997–1999. The peak of the cycle corresponds to 30% “missing NOy” or 240–500 ppt. Horii [2002] describes measurements at Harvard Forest during summer and fall 2000. She reports “missing NOy” up to 50% and infers from flux measurements that the “missing NOy” has a deposition rate that is comparable to HNO3. Other reports of “missing NOy” can be found in the following references [Aneja et al., 1996; Atlas et al., 1992; Harrison et al., 1999; Nielsen et al., 1995; Parrish et al., 1993; Patz et al., 2000; Ridley, 1991; Singh et al., 1996; Zellweger et al., 2000].

[19] Because only a subset of the instruments used in many of these studies were state of the art, it is often argued that the “missing NOy” might be an experimental artifact. The paper most often cited in this regard is Parrish et al. [1993], where “missing NOy” of approximately 10% was observed at six rural sites in eastern North America during early fall, 1988. The paper argues that the “missing NOy” can be attributed to the combined systematic errors of the different techniques used to observe NOy and ΣNOyi, or at least such errors cannot be ruled out. However, a chemical model for one of the sites discussed by Parrish et al. predicts that most of the 10–15% “missing NOy” can be attributed to the formation of alkyl and hydroxyalkyl nitrates derived from isoprene [Trainer et al., 1991]. The model accounts for both the average value of “missing NOy” and its variability. It is unlikely that a model of the atmosphere would represent the variability of an experimental artifact, which suggests that alkyl nitrate formation may indeed account for the “missing NOy” at this site.

[20] In summary, observations of NOy and ΣNOyi by more than a dozen research groups at a wide range of locations have all found “missing NOy,” that is, NOy > ΣNOyi, in the midlatitude, continental boundary layer. Our measurements of ΣANs at the UC-BFRS, Granite Bay, and La Porte have all of the characteristics that are associated with the mystery molecules that comprise the “missing NOy.” First, the absolute concentrations we measured at the UC-BFRS during the summer, 200–800 ppt, are comparable to the “missing NOy” reported for other rural sites during summer [Aneja et al., 1996; Buhr et al., 1990; Ridley et al., 1990; Williams et al., 1997]. Second, the fraction of NOy represented by ΣANs at the UC-BFRS (10–20%) is similar to the missing fraction reported at other locations [Aneja et al., 1996; Atlas et al., 1992; Buhr et al., 1990; Ridley et al., 1990; Williams et al., 1997; Zellweger et al., 2000] and predicted by models [Trainer et al., 1991]. As a fraction of NOz, ΣANs observed at the UC-BFRS, Granite Bay, and La Porte are similar. Third, ΣANs are an important fraction of NOy year round at the UC-BFRS, and the seasonal cycle of ΣANs (and ΣANs/ΣNOyi) corresponds closely to observations of seasonal variation in the “missing NOy” [Fahey et al., 1986; Hayden et al., 2003]. Finally, at all three sites we observe a strong positive correlation between ΣANs and a variety of photochemical tracers, namely, O3, ΣPNs, and HNO3 (Figure 5), as have been reported for “missing NOy” [Buhr et al., 1990; Fahey et al., 1986; Nielsen et al., 1995; Ridley et al., 1990; Williams et al., 1997].

Figure 5.

Half-hour medians (ppb) of daytime (10 AM–6 PM) observations of (a) ozone, (b) HNO3, and (c) ΣPNs versus ΣANs at the UC-BFRS. All observations between October 2000 and December 2001 are shown. The lines shown, O3 = 30 + 80(ΣANs), HNO3 = 2.5(ΣANs − 0.05), ΣPNs = 2(ΣANs), were not fit to the data.

[21] Other evidence that ΣANs are the “missing NOy” can be found in a comparison of our own NOy and ΣNOyi measurements at the UC-BFRS. In contrast to virtually every other experimental assessment of the NOy budget, we do not routinely observe NOy greater than ΣNOyi. We compare NOy and ΣNOyi measurements obtained over a full annual cycle, between October 2, 2000 and September 1, 2001 at the UC-BFRS. Figure 6 shows the 3069 30-min points (medians) of NOy versus ΣNOyi from this comparison (156 are off-scale at NOy greater than 5 ppb). We note that this is a more temporally extended comparison of NOy and ΣNOyi than any previously reported. For example, the recent paper by Thornberry et al. [2001], which discusses comparison of NOy and ΣNOyi measurements, is based on 10 days of observations represented by 24-hour averages. Our data span complete diurnal cycles for all four seasons and thus a large range of photochemical regimes.

Figure 6.

Half-hour medians (ppb) of ΣNOyi plotted versus NOy (as measured by catalysis chemiluminescence) for observations made at the UC-BFRS between October 2, 2000 and September 1, 2001. The line shown is a best fit line with slope 1.06 and R2 = 0.66. If the intercept is allowed to vary, no improvement is observed in the fit, but the parameters are a slope of 0.98 and an intercept of 240 ppt.

[22] The linear fit of ΣNOyi to NOy, shown in Figure 6, has a slope of 1.06 ± 0.007 (R2 = 0.66). If the intercept is allowed to vary, no improvement is observed in the fit, but the parameters are a slope of 0.98 and an intercept of 240 ppt. The mean ΣNOyi/NOy ratio is 1.12, and the median difference is 180 ppt (ΣNOyi > NOy). Based on the accuracies of the two instruments (±15% ±10 ppt for TD-LIF and ±10% ±100 ppt for chemiluminescence) this represents agreement. The accuracy of the instruments is a combination of the uncertainty in the concentrations of the reference standards, conversion efficiencies, and any interferences. Comparison of standards used by the two instruments show that the relative uncertainty is smaller than a square root of the addition of the sum of squares of 15% and 10%. Dillon [2002] shows that some combination of a small positive interference in TD-LIF HNO3 observations and/or a small difference in HNO3 transmission efficiencies explains a large part of the deviation of the points shown in the comparison (Figure 6) from a single line. That analysis also shows there is little correlation of the difference between ΣNOyi and NOy with ΣANs. Furthermore, since ΣANs are typically 10–20% of NOy, it is obvious there would be a substantial shortfall of ΣNOyi relative to NOy if ΣANs were 10 times smaller as has been typically reported by methods that measure specific alkyl nitrate compounds.

5. ΣANs and O3 Photochemistry

[23] ΣANs are produced as a minor branch of the reaction of RO2 with NO (reaction (3b)). The main branch (reaction (3a)), results in oxidation of NO to NO2, which is followed by production of ozone (see Figure 1). Because both O3 and ΣANs have their source in the same process, we expect the two species to be correlated with a slope of the correlation related to the average branching ratio for nitrate formation, α = k3b/(k3a + k3b), of the hydrocarbons that contribute to ozone production. Most of the time, the coupled HOx and NOx catalytic cycles (Figure 1 and reactions (1)(8)) produce either 2 O3 molecules or 1 alkyl nitrate per round of the HOx cycle. These numbers for O3 and ΣANs production are typical but hydrocarbons with O3 yields of 1 or 3 are not uncommon and chain termination and deposition can also affect ΔO3/ΔΣANs in the atmosphere. On average, when the HOx cycle chain length is long there will be production of approximately 2(1 − α)O3 and α ΣANs. In the remote atmosphere where CO (α = 0) and CH4 (α ≈ 3 × 10−4) [Flocke et al., 1998a] are the feedstock for ozone production, the slope of ozone versus ΣANs should be in excess of 7000. In contrast, on the continents where isoprene oxidation (α = 0.044 [Chen et al., 1998], 0.12 [Tuazon and Atkinson, 1990]) is a major source of ozone, slopes of the order 15–50 for ozone versus ΣANs should be expected. Branching ratios for hydrocarbons are known to increase with carbon chain length along homologous series. For example, α increases from <1.4% for ethyl peroxy plus NO to 18% for the hexyl peroxy radical reaction [Arey et al., 2001]. Oxidation of alkenes proceeds by addition of OH to the double bond followed by addition of O2 at the adjacent radical site to form hydroxy peroxy radicals. Nitrate yields of these radicals increase from 0.86% for hydroxyethyl peroxy to 5.5% for hydroxyhexyl peroxy [O'Brien et al., 1998]. As a result, the observed correlation of O3 with ΣANs will vary depending on the hydrocarbons that contribute to RO2.

[24] Figure 5a shows the full annual cycle of observations of O3 versus ΣANs during daytime (10 AM–6 PM) at UC-BFRS. The correlation slope is approximately 80 ΔO3/ΔΣANs which implies an average branching ratio for ΣAN production of 2.5%. A 2.5% branching ratio is in the range expected for RO2 radicals [O'Brien et al., 1998] derived from propane or butene. This ratio reflects photochemistry both at the urban source and during transport to the UC-BFRS. If most of the ΣANs are hydroxyalkyl nitrates, then ΣANs will be lost to deposition faster than O3 during transport to the UC-BFRS from the source region and the actual branching ratio of the original hydrocarbon mixture would be higher. We expect that hydroxyalkyl nitrates, derived from isoprene, will be an important contribution to ΣANs since isoprene is expected to be a major contributor to the production of ozone at the UC-BFRS [Dreyfus et al., 2002]. At Granite Bay and La Porte, we also observe strong positive correlations between O3 and ΣANs with slopes in the range of 25–120 which are dependant on the time of day at both sites. Model calculations based on measurements of both anthropogenic and biogenic hydrocarbons at both of these sites predict O3/ΣANs slopes in this range [Cleary et al., 2002; Rosen et al., 2002]. Correlations between ΣANs and other chemicals, notably ΣPNs and HNO3, are also evident in the data for these other two sites.

[25] In addition to serving as an indicator for the peroxy radical source of ozone production, the AN formation reaction (reaction (3b)) also terminates the NOx catalytic cycle. The other major chain termination steps are PN formation and HNO3 formation. Figures 5b and 5c show HNO3 and ΣPNs versus ΣANs. Approximate slopes are 2.5 HNO3/ΣANs and 2 ΣPNs/ΣANs. Assuming losses to deposition are negligible, this implies 36% termination of the NOx cycle by formation of PNs, 46% by HNO3 and 18% by ΣANs. However, it would be reasonable to estimate that approximately 2/3 of the NOy that is converted to HNO3 is deposited during transport [Dillon, 2002]. Accounting for this loss (and assuming no losses of PNs or ANs) implies about 20% termination by formation of PNs, 75% by HNO3 and 10% by ΣANs. Most likely, the AN contribution to termination of the NOx cycle lies within the range of these two estimates (10–18%) because deposition is also likely to remove the hydroxy component of ΣANs efficiently. From a larger regional or global perspective, the relative contribution of ΣANs to overall termination of the NOx cycle will be larger than estimated here since the PNs do not represent a permanent loss for NOx, as they primarily decompose thermally, ultimately regenerating NOx.

[26] There are few theoretical calculations to which we can compare our observations of range of NOx cycle termination steps. Using an observationally constrained model, Thornton et al. [2002] estimate that the ΣAN production rate would be 15% that of HNO3 production under the high NOx conditions present at Nashville, TN and that this ratio would increase at lower NOx conditions. Luecken et al. [1999] describe simulations using three different chemical mechanisms commonly used in air pollution models. For the same location and conditions, the different models predict [ΣANs + peroxyacyl nitrates other than PAN] abundances ranging from 3–35% of NOy. Our observations are consistent with the predictions in both of these references and could serve to substantially refine the models used to make them.

6. Discussion

[27] Our observations of ΣANs suggest that nearly all previous experiments at midlatitude continental boundary layer sites underestimate the role of alkyl nitrates. The highest alkyl nitrate concentrations we are aware of were reported by Grosjean [1983], who observed up to 5 ppb methyl nitrate in Claremont, CA (L.A. basin) during 1980. However, the contribution of ΣANs to NOy or NOz was small since NO2 was in excess of 100 ppb and PAN often exceeded 30 ppb. O'Brien et al. [1995] describe observations of 12 C3–C6 simple alkyl nitrates, 4 C2–C4 hydroxyalkyl nitrates, and one alkyl di-nitrate at a rural site in Ontario during August, 1992. The sum of the alkyl nitrate concentrations ranged from 12 to 140 ppt and contributed 0.5–3% to NOy. Flocke et al. [1998b] measured C1–C8 simple alkyl nitrates at the Schauinsland Station in the Black Forest (Germany) from June 1990 to May 1991. They observed ΣANs ranging from 30 ppt to 630 ppt, with an average value of 130 ppt. The average contribution to NOy was 3%, with little seasonal variation. The average contribution to NOz was 3.4% in summer and 8.0% in winter. In all of the reports describing the midlatitude continental boundary layer, the sum of measured alkyl nitrate concentrations range between <1 and 630 ppt, comprising <1 to 3% of NOy, on average, at both urban and rural sites [Buhr et al., 1990; Flocke et al., 1998b; O'Brien et al., 1995, 1997; Ridley et al., 1990; Roberts et al., 1998; Shepson et al., 1993; Thornberry et al., 2001]. Indeed, despite recent instrumental advances which have allowed the measurement of hydroxyalkyl nitrates and alkyl dinitrates [O'Brien et al., 1995], C1–C15 alkyl nitrates, benzyl nitrate, and various bi-functional nitrates [Schneider et al., 1998], including isoprene nitrates [Grossenbacher et al., 2001], prior to our own experiments ΣANs have never been observed to constitute a large fraction of NOy in the midlatitude, continental boundary layer.

[28] In contrast, high ΣAN/NOy ratios have been observed in remote locations. Bottenheim et al. [1993] and Muthuramu et al. [1994] measured C3–C7 alkyl nitrates and several other nitrates that were not identified at Alert, Canada during polar sunrise in 1988 and 1992, respectively. Both studies observed that the sum was typically between 40 and 200 ppt constituting 10–20% of NOy. Jones et al. [1999] report that the sum of methyl and ethyl nitrate concentrations were more than 50% of the 70 ppt of NOy over Antarctica during the austral summer. Talbot et al. [2000] report that the sum of C1–C4 alkyl nitrates was 20–80% of the 50 ppt of NOy in the marine boundary layer over the equatorial Pacific Ocean in austral springtime. The authors' explanation for these observations of high contributions of alkyl nitrates to NOy in these remote regions include the longer lifetime expected for simple alkyl nitrates relative to other NOy species and possible marine sources. Comparison of these remote measurements to our observations raises additional questions about where ΣANs are removed from the atmosphere.

[29] The experiments described here use observations of a new quantity, ΣANs, to show that at rural, urban, and suburban locations ΣANs are a large fraction of higher oxides of nitrogen (NOz), represent the explanation for the dilemma of “missing NOy,” and act to terminate the catalytic ozone production cycles at comparable rate to HNO3 formation with the consequence that removal of nitrogen oxides from the atmosphere likely often occurs by deposition of ΣANs. We suggest that analysis of the relationship between ΣANs and O3 will provide a more direct test of the hydrocarbon component of models describing ozone production. Along with those analyses, a comprehensive study involving complete hydrocarbon precursor measurements coupled with both TD-LIF ΣANs measurements and individually speciated alkyl nitrate measurements would allow evaluation of our understanding of instrumentation designed to observe the specific compounds that comprise ΣANs and of models of alkyl nitrate production and losses.

[30] Other new opportunities are now possible with ΣAN observations. For example, comparing mixing ratios of HNO3 and ΣANs at different locations is a potential indicator for whether O3 production is NOx-limited or NOx-saturated (VOC-limited), with HNO3 production favored by NOx-saturated conditions and ΣAN production favored under NOx-limited conditions. Much of the current debate in air quality management centers on understanding whether O3 will decrease if NOx is reduced, as it will if the atmosphere is NOx-limited or will increase if NOx is reduced, as it will if the atmosphere is NOx-saturated. Another example of the potential for ΣAN measurements to open new avenues of research is in the area of biosphere-atmosphere exchange. Nitrogen is a limiting nutrient in many ecosystems and it is possible that atmospheric inputs are altering the amount of nitrogen available with consequences for carbon uptake and perhaps climate. Little is known about whether complex organic nitrogen species can be utilized directly by plants as they breathe them in through their stomata. Measurements of ecosystem scale fluxes of ΣANs would be a valuable step toward understanding these phenomena.


[31] Measurements of NOx, NOy, and ΣNOyi described in this paper were supported by the NASA Instrument Incubator Program under contract NAS1-99053, the NOAA Office of Global Programs, the Department of Energy under contract DE-AC03-76SF0009, and the Hellman Family Faculty Fund. Analysis of the measurements at the UC-BFRS was supported by the NSF Atmospheric Chemistry Program under grant ATM-0138669. Joel Thornton gratefully acknowledges support from a NASA Earth Systems Science Fellowship NASA grant NGT5-30219. We gratefully acknowledge Allen Goldstein and Megan McKay for their meteorological and ozone measurements from the UC-BFRS and Granite Bay and Eric Williams for ozone observations at La Porte (TEXAQS). Special thanks to Sheryl Rambeau, Dave Rambeau, and Bob Heald for exceptional logistical support at the UC-BFRS.