Summertime partitioning and budget of NOy compounds in the troposphere over Alaska and Canada: ABLE 3B

As part of NASA's Arctic Boundary Layer Expedition 3A and 3B field measurement programs, measurements of NO x, HNO 3, PAN, PPN, and NOy were made in the middle to lower troposphere over Alaska and Canada during the summers of 1988 and 1990. These measurements are used to assess the degree of closure within the reactive odd nitrogen (NxOy) budget through the comparison of the values of NO measured with a catalytic convertor to the sum of individually measured NOy(i) compounds (i.e., (cid:127)N(cid:127)y(i) = NO n + HNO 3 + PAN + PPN). Significant differences were observed between the various study regions. In the lower 6 km of the troposphere over Alaska and the Hudson Bay lowlands of Canada a significant fraction of the NOy budget (30 to 60%) could not be accounted for by the measured (cid:127)NOy(i). This deficit in the NOy budget is about 100 to 200 parts per trillion by volume (pptv) in the lower troposphere (0.15 to 3 km) and about 200 to 400 pptv in the middle free troposphere (3 to 6.2 km). Conversely, the in the northern Labrador and Quebec of is almost totally accounted for within the combined measurement uncertainties of and the various compounds. A and/or dynamical parameters influencing the tropospheric oxidative potential over these regions. A combination of factors are suggested as the causes for the variability observed in the NOy budget. In addition, the apparent stability of compounds represented by the NOy budget deficit in the lower-altitude range questions the ability of these compounds to participate as reversible reservoirs for active odd nitrogen and suggest that some portion of the NOy budget may consist of relatively unreactive nitrogen-containing compounds.


INTRODUCTION
[e.g., Liu et al., 1983;Chatneides et al., 1987;Carroll et al.; In many regions of the remote troposphere the 1990]. The NO mixing ratios observed in the summertime availability of reactive odd nitrogen in the form of nitric troposphere (0.15 -6.2 kin) over Alaska were, on average, oxide (NO) is a critical factor controlling the slightly larger than those observed over the remote lowerphotochemical production of ozone (03) ß This control of latitude marine environments [Sandhobn et al., 1992]. The 03 production occurs through NO reaction with summertime NO mixing ratios over Alaska were still small hydroperoxy and organoperoxy radicals (HO2 and RO2). enough to result in predictions of a net photochemical loss The photochemical destruction rate for 03 is believed to be of 03 in the lower 6-km column. In these cases the 03 controlled by reactions of 03 with the hydroperoxy radical photochemical lifetime was predicted to be nearly equal to (HO2) and by direct photolysis of 03 leading to hydroxyl the lifetime based on surface deposition. These results radical (OH)production. The absence of NO would result also indicated that the 0 3 photochemical lifetime was in a net photochemical loss of 03 throughout the nearly 2.5 times longer than that predicted in the absence troposphere. Over remote tropical andmidlatitudemarine of NO [Jacob et al., 1992]. The middle to lower environments, observed NO mixing ratios have been small tropospheric summertime measurements of NO made over

NOn and NOy Measure•nents
The spectroscopically selective two-photon/laser-induced fluorescence (TP/LIF) NO technique was applied to the simultaneous measurement of NO, NO:, and NO r Details of this instrument have been previously reported [Bradshaw et al., 1985;Sandhobn et al., 1990;Sandhobn et al., 1992].
The 226-nm and 1.1-/•m laser beams used in the twophoton fluorescence excitation process were passed through three separate ambient sampling cells. One cell was designated for detecting ambient NO. A second cell was designated for detecting NO produced from a photolytic conversion of ambient NO2. The third cell was designated for detecting NO produced from the 300øC gold-catalyzed conversion of ambient NOy using 0.3% CO as a reducing agent. The NO, NO,_, and NOy channels used separate signal acquisition electronics, flow measurement and control systems, and signal normalization/internal reference standards.
Ambient air was continuously drawn in through a common porcelain-glass coated inlet (2.5-cm ID) at a nominal flow rate of 200 liters per minute (lpm). The airstream was sampled perpendicular to the aircraft motion. The sample residence time through the inlet manifold was always less than half a second. Total sample residence time through the NO ambient sampling portion of the instrument was less than i s. Sample residence time in the NO,• portion of the instrument varied between i and 5 s, depending on the photolytic converter system used.
NOy samples were drawn from the center of the ambient sampling manifold at a location near the NOy converter assembly with a residence time • 0.2 s from the manifold to the NOy converter. The sampling manifold and flow line fittings were tested inflight for leaks using a 10 parts per million by volume (ppmv) NO standard as a leak tracer.
During ABLE 3A an excimer laser was used as the photolytic convertor system's photolysis source [Sandholm et al., 1990;Sandhobn et al., 1992]. To reduce the instrument's size, weight, and power consumption, a highpressure xenon arc-lamp-based photolytic converter was used during the ABLE 3B program. This latter converter utilized a 1-kW Cermax short-arc lamp that was spectrally filtered into a photolysis passband of 350 nm < ,X < 420 nm. This converter was designed to operate at 30% to 60% photolytic efficiency with sample residence times ranging from 2 to 4.5 s, respectively. The low wavelength cutoff (10% peak spectral fluence) of 350 nm was chosen to minimize possible interferences from concomitant R-NO• compounds. Deep UV (,X < 330 rim) and Visible/IR (,X > 480 nm) emissions from the lamp were attenuated by > 103-fold relative to the intensity at the center of the passband. The photolysis beam was also spatially filtered to avoid illumination of the photolytic sample cell's walls. In both the excimer laser and the arc-lamp-based converters, the photolytic sample cells were thermally controlled using a high flow rate of ambient air that was passed through an outer jacket of each cell.
The NOy catalytic converter system used with the TP/LIF sensor was adapted from the NOy converter system design developed by NOAA [Bollinger et al., 1983;Fahey et al., 1985;Murphy and Fahey, 1987]. In the ABLE 3A program a short length (--0.3 m) of PFA Teflon tubing was used to connect the NOy converter to the main porcelain-glass-coated sampling manifold. This tubing was replaced by a similar length of heated (-400C) goldcoated nickel tubing in the ABLE 3B program. This change was made to minimize the possible effects of sample line memory during the more rapid descents/ascents planned for the ABLE 3B field program. During the 1989 NASA/GTE Chemical Instrumentation Test and Evaluation (CITE) 3 field program a similar PFA Teflon tubing was used to couple the NOy convertor to the inlet/manifold. In this latter case, a positive artifact was believed to have been induced by sample line memory effects following the rapid descent (500 m/min) through haze layers that occurred near the trade wind inversion over the equatorial western Atlantic Ocean near Natal, Brazil. Inflight and preflight/postflight calibrations were performed by standard additions to either ambient air (inflight) or bottled air (preflight/postflight). Two-stage serial dilution systems were used to dilute the parts per minute by volume mixing ratios of standards contained by high-pressure aluminum cylinders. The diluted standards were injected directly into the ambient sampling inlet and subsequently diluted to final concentration (0.1 to 2 parts per billion by volume (ppbv)). All flow measurements were made with linear mass flow controllers or meters.
These devices were intercompared with positive volume displacement flow measurement instrumentation prior to and after both the ABLE 3A and ABLE 3B field programs.
Primary and secondary gas calibration standards were used for both NO and NO 2. The 50 ppmv (in nitrogen) NO and NO 2 primary standards used for intercalibration have remained stable since their acquisition in 1981 (i.e., < 10% deviation from initial conditions from 1981 to 1991). Standards were intercompared with National Institute of Standards and Technology (NIST) standards as part of the NASA/GTE field measurement program protocol in 1983, 1984, 1986, 1990, and 1991. High-pressure nitroethane (C2HsNO2) in helium standards were developed to provide a more rigorous test of the NOy conversion efficiency. These C2HsNO• standards were made from reagent grade C2HsNO 2 that was further purified by several lowtemperature vacuum distillations in which only the middle third of the distillate was retained. Aluminum cylinders were vacuum baked (< 10 '3 mbar, 50øC) and flushed several times, then filled with a known pressure of the purified C2HsNO 2 and diluted with research purity helium to a final cylinder pressure of about 1600 psig. Final C2HsNO 2 mixing ratios determined from the cylinder contents by UV absorption agreed to within 10% of the value calculated from partial pressure dilution. The C2HsNO • standards have proven to be stable at mixing ratios in the range of 10 ppmv. Inflight calibrations were made using NO and NO2 during ABLE 3A and using NO, NO2, and C2HsNO 2 during ABLE 3B.
The preflight/postflight mixing ratios measured in bottled air to which no standard was added were not subtracted as "blank" values for the NO• or NOy measurements reported here. We believe these "blank" values, which ranged from approximately 40 to 120 pptv, predominantly reflected outgassing, under slow flow conditions (< 40 slpm), of the long length of extra tubing (-15 m) used to reach the aircraft inlet. This belief is based on decay of signal versus time that typically indicated approximately 1 to 2 hours for this system to "clean" up and approach the lower end of the range of mixing ratios stated above. The instrument's background was continuously monitored by blocking of the 1. NO2, and NOy mixing ratios had, at the 95% confidence limit, standard deviations about the mean of +16%, +18%, and +18%, respectively. The limit of detection (LOD) for a signal-to-noise ratio of 2/1 averaged 2 pptv for NO and 6 pptv for NOa using a 3-min signal integration time. The photon statistics based measurement precision at the 95% confidence limit (3 minutes integration) was typically +25% for NO = 15 pptv, +35% for NOa = 35 pptv, and +8.5% for NOy = 700 pptv. Measurement precision was proportional to the square root of mixing ratio and/or integration time. The TP/LIF NO technique was used in both the NASA/GTE CITE 1 (1983) and the CITE 2 (1984) airborne intercomparison programs [Hoell et al., 1987;Gregory et al., 1990a]. These intercomparisons concluded that at low mixing ratios (i.e., < 60 pptv), NO measurements agreed with the stated instruments' precision and accuracy. The level of agreement among instruments was of the order of 35%. On average, individual measurements agreed within 5 to 7 pptv for NO mixing ratios of < 20 pptv. Concurrently, the photofragmentation TP/LIF NOa technique was evaluated during the CITE 2 airborne intercomparison program [Gregory et al., 1990b]. The NOa intercomparison study also concluded that, on average, the NO2 measurements agreed within the stated instrument precision and accuracy at low NOz mixing ratios (i.e., NOa < 200 pptv). Individual measurements showed a 30 to 40% level of agreement (i.e., 15 to 20% from the average). Intercomparison at the lowest NO,_ mixing ratios (i.e., NOa < 50 pptv) revealed individual measurements were uncorrelated due to the randomness associated with approaching the instruments' limits of detection. On average, NOa values still agreed within the instruments' stated accuracy [Gregory et al., 1990a, b]. The instruments used in the ABLE 3A and ABLE 3B field programs were nearly identical in their evaluated precision and accuracy for NO and NO2.

Peroxyacetyl Nitrate (PAN), Peroxypropionyl Nitrate (PPN) Measurements
PAN and PPN were measured using a cryotrap (CT) preconcentration sample loop in conjunction with gas chromatographic (GC) separation, and a tandem columetric electron capture detection (ECD system). In the CT/GC/ECD technique, typically 0.10 to 0.20 standard liters of ambient air, drawn through an aft-facing Teflon inlet, were enriched in a cryotrap held at a constant -150 ø C temperature. The preconcentrated samples were then analyzed using a GC/ECD that was operated at constant pressure (1050 mbar).
Calibration of the PAN instrument was accomplished using diffusion tubes containing PAN standards that were prepared by the CH3CHO/NO2/C12 photolysis method [Singh and Salas, 1983] and dissolved/stored in liquid ntridecane [Gaffhey et al., 1984]. Air was passed over the diffusion tube at a constant rate of flow to provide a calibration gas stream having PAN mixing ratios in the ppbv range. This calibration gas stream was periodically added at a downstream point in the ambient sampling line. PAN mixing ratios in the calibration gas stream were measured onboard the aircraft using a hot molybdenum oxide converter (375 ø C) coupled to a chemiluminescence NO monitor. Calibration of this chemiluminescence system was accomplished using both NO2 and NO standards that were intercompared with NIST primary standards as part of the CITE 2 and ABLE 3B field program protocols. The calibration gas stream was further diluted using a two-stage serial dilution system to provide final PAN calibration mixing ratios in the parts per trillion by volume range. Linear mass flow controllers/meters used in this dilution system were compared with volume displacement standards. The accuracy of the PAN calibration transfer to ambient air measurements is estimated to be +25% at the 95% confidence level.
The CT/GC/ECD PAN instrument was shown to be linear for PAN mixing ratios ranging from 5 to 1000 pptv. The typical limit of detection of the CT/GC/ECD PAN instrument was < 5 pptv for 100 ml of sampled air. Measurement precision was +10% at the 95% confidence level for PAN mixing ratios that were well above the limit of detection (i.e., PAN > 50 pptv).
This CT/GC/ECD PAN instrument was used in the NASA/GTE CITE 2 airborne intercomparison program [Gregory et al., 1990c]. This intercomparison program pointed out the need for onboard verification of PAN standards. As outlined above, onboard PAN standard verification was carried out during both ABLE 3A and ABLE 3B. The CITE 2 airborne intercomparison concluded that the two PAN instruments agreed, on the average, to about 20 pptv for PAN mixing ratios < 100 pptv. At larger PAN mixing ratios, agreement between individual measurements and the average was possible at the +30% level for a 95% confidence limit even though individual pairs of measurements might sometimes fall outside this range. The PAN measurements that were intercompared fell close to agreeing within the stated accuracy and precision of the instruments [Gregory et al., •990c1.

HN03 Measurements
Gas phase HNO3 was measured using a mist-chamber (MC) aqueous scrubber as a preconcentrator for subsequent determination of NO3-by ion chromatography (IC). Details of this technique have been previously described [Talbot et al., 1990;Talbot et al., 1992]. In this technique, gas phase HNO3 and other soluble gases were stripped from ambient air into a dense mist of ultrapure deionized water. The mist was collected on a Teflon filter that recirculated the mist chamber's water supply to the pneumatic nebulizer that formed the mist. A Teflon prefilter was installed in the sampling line, upstream of the mist chamber, in order to segregate against the collection of water soluble particulate-nitrate (p-NO3'). The mist chamber sampled ambient air at a flow rate of 30 to 40 standard liters per minute (slpm) from a main ambient sampling manifold and inlet system that drew in approximately 300 slpm of ambient air. All flow rates were measured with linear mass flow meters that were calibrated prior to and after each field program. A Tefloncoated inlet/sampling manifold was used during the ABLE 3A field program, whereas a 40-ram ID porcelain-glasscoated inlet/sampling manifold was used during the ABLE 3B program. This latter inlet/sampling manifold was nearly identical to the design used by the NO•/NO• instrument described previously. Laboratory tests of the porcelain-glass-coated inlet/manifold demonstrated 100 (+3)% passing efficiency for HNO3 mixing ratios in the 100-300 pptv range using flow rates of approximately 300 slpm. The change of inlet/manifold coating materials was prompted by concerns over possible sampling memory effects from a Teflon-coated system. HNO3 passing efficiency of the 25-mm ID Teflon-coated inlet used in ABLE 3A averaged 80% to 85% [Talbot et al., 1992a]. HNO3 mixing ratios reported for the ABLE-3A program were corrected for a passing efficiency of 80%.
The MC/IC calibrations were based on solutions prepared from dried KNO3. These nitrate standards have agreed within +3% of NIST standard solutions. The IC limit of detection (LOD) for HNO3 was equivalent to 20 pptv for a 30-min sample collection time, using 30 slpm sample flow rates, where the LOD is inversely proportional to changes in the sample collection time. Blanks were obtained prior to each flight by sampling ambient air that was scrubbed of HNO3 after passing through a series of impregnated and activated charcoal filters. Blank values were consistently at or below the LOD of the IC system using a 30-rain sample collection time. Accuracy of the ICs calibration transfer to ambient HNO3 measurements is estimated to be +20% at the 95% confidence level, based on the uncertainties in the IC analysis of NO3', the results of laboratory tests of the MC's collection and the inlet's passing efficiencies for HNO3, and the uncertainties in the measurement of the sampled air volume. The sampling time used during ABLE 3A ranged from 15 minutes to 3 hours, whereas those used during the ABLE-3B program ranged from 3 to 45 minutes.
The MC/IC technique has yet to participate in an airborne intercomparison study. The MC/IC technique has been used in a recent ground-based HNO3 intercomparison. The HNO3 measurement techniques used in this ground-based intercomparison included the MC/IC technique discussed here, a NOAA nylon filter collection technique, and a continuous flow liquid diffusion scrubber (preconcentrator) based system developed by Lind at NCAR.
In all three techniques, ion chromatography was used to measure the NO3' that was collected by the preconcentrators. The preliminary results of this intercomparison indicated that on average the individual techniques agreed within +25% of the mean value formed between pairs of measurements over the HNO 3 mixing ratio range of 100 pptv to 500 pptv. Even so, individual pairs of HNO 3 measurements fell outside this range (E. L. Atlas et al., An intercomparison of three HNO3 measurement techniques, submitted to Journal of Geophysical Research, 1992).

Particulate NOs' Measurements
Particulate NO3' (p-NO3') was measured using an isokinetic aerosol sampling probe with a curved-leading edge nozzle design [Talbot et al., 1992a, b]. A shrouded version of this nozzle design was used during the ABLE 3B field program. The collection efficiency of supermicron particles was greater than that for equivalent inlets using a sharp-leading edge nozzle design similar to those evaluated by Hubert et al. [1990]. Even so, the collection efficiency for supermicron particles must still be considered as uncertain. For submicron particles the curved leading edge nozzle has been shown to give significantly larger passing efficiency than straight edge designs with minimal loss on the nozzle inlet or curved tube sampling manifold wall (R. W. Talbot et al., Improvements in aerosol inlet performance in airborne applications, submitted to Journal of Atmospheric and Oceanographic Technology, 1992).
The Teflon 90-mm filters were mounted on supports made of Delrin and contained by a Delrin housing. Particles were collected on a stacked set of Nuclepore (8 /•m) and Zefluor (1 /•m) filters during the ABLE 3A program, whereas a single Zefluor (2 /•m) filter arrangement was used in the ABLE 3B program. Sample collection times ranged from 15 minutes to 2 hours using a nominal ambient airflow rate of 475 lpm. These filter samples were analyzed for soluble NO 3' by ionchromatography.
Filter blanks for p-NO3' were equivalent to ambient mixing ratios of 4 pptv and were subtracted from reported values. The accuracy of the ambient air equivalent NO 3' that was measured on the filters was estimated to be 520% at the 95% confidence level, excluding uncertainties involving the sample inlet passing/collection efficiency. Some fraction of fine p-NO3' -containing aerosols was most likely collected and converted in the NOy system. This fraction is believed to be significantly smaller than that collected by the curved leading edge aerosol sampling system. Even though there is some degree of uncertainty on an absolute scale, the p-NO3' measurements that were made by the curved leading edge sampler described here should allow for an examination of relative trends in fine p-NO3', with respect to closure of the measured N•O• family budget.

Air Mass Characteristics
The Alaskan portion of the ABLE 3A program (missions 6-26) primarily overflew study regions near Barrow and Bethel, Alaska. Air masses encountered in the ABLE 3A program originated from regions that were typically free from anthropogenic sources based on 3-to 5day air mass back-trajectory analyses. These air mass origins were determined to be from two primary source regions, namely, the Gulf of Alaska/Bering Sea, and northern Siberia/Arctic packice. Missions 14-21 were characterized by air masses originating from the Gulf of Alaska and Bering Sea, whereas missions 6-13 and 24-26 were characterized by air masses originating over northern The various NOy(i) measurements made during the ABLE 3 field programs covered a wide range of sampling time intervals. The HNO3 (and p-NO3') measurement system used the longest sampling/preconcentration times with sampling times, which ranged from 3 min to 3 hours and covered spatial scales from about 25 km to 500 km. In general, the sampling times of this instrument were about 5 times longer during the ABLE 3A program than those used during the ABLE 3B program. PAN and PPN measurements were taken using 1-to 2-min sampling times with individual samples taken every 6-to 10-min. NO• and NOy signals were recorded for 30-s signal integration times with individual measurements reported using longer integration times ranging from 1.5 to 3 min. Selection of the data set used here for investigating closure within the reactive nitrogen budget (i.e., NOy versus ENOy(i)) took To account for effects that might arise from poor temporal overlap of the data during periods with large ambient variability, our NOy to ZNOy(i) budget comparison data subset was filtered at two levels. First, the data were filtered to include only those HNO• measurement periods when all other NOy(i) compounds (PAN, PPN, and NO•) and NOy were also measured. This filter required a 65% temporal coverage by the NO• and NOy measurements and similar temporal coverage by PAN within the framework of the PAN instrument's lower measurement duty cycle (i.e., PAN measurements 20% maximum temporal overlap could yield 13% temporal coverage of the HNO3 measurement interval). The second filter that was used involved the examination of the temporal behavior of NOx, NOy, and PAN during individual HNO• measurement periods. Cases in which significant changes occurred in either NOy, NO•, or PAN mixing ratios that were not accompanied by measurements of all three compounds were excluded from the filtered data set.

Figures 1-3 display three examples of poor temporal
overlap among the various NxOy measurements during periods of ambient variability. In the first two cases, NOy mixing ratios were observed to increase by twofold to threefold in the middle of an individual HNO, measurement period, with PAN measurements occurring either outside of (see Figure 1), or at the center of (see

points. Figure 3 displays an example in which the HNO 3 measurement period coincided with a large increase in
both NOx and NOy. In this case, PAN was measured only during the trailing edge of the event. All three of these cases and others like them were dropped from consideration due to poor temporal coverage during periods of ambient variability, as defined by a greater than 1.5-fold change in NOy mixing ratios over a HNO 3 measurement period without concurrent measurements of PAN and NO,,. There were 84, 134, and 144 HNO 3 measurements made over the Alaska, Hudson Bay lowland, and Labrador boreal forest study regions, respectively. The filtering, discussed above, produced a budget comparison data subset that retained 34, 69, and 89 HNO 3 measurement periods, or 40%, 51%, and 62% of the original database. The larger retention of data from the ABLE 3B database was primarily due to a combination of shorter HNO 3 sampling times and a 1-to 1.5-hour transit time between the base airports and the intensive study areas. These transits generally consisted of stair-step flight profiles that were optimized for the study of N•Oy budgets in the free troposphere. Table I gives a summary of the NxOy composition observed in the filtered data set for the Alaska, the Hudson Bay lowlands, and the northern Labrador/Quebec study regions. These data have been separated into measurements made in the middle free troposphere (from 3 km to 6.2 km) and those made in the planetary boundary layer (taken here as < 3 km). Inspection of Table 1 reveals that mean NO• mixing ratios over both Canadian study regions were larger than those observed over Alaska by approximately 1.5-fold above 3 km and by 2-fold below 3 km. Over Canada the enhanced NO• mixing ratios below 3 km may reflect the combination of warmer temperatures and the approximate threefold larger PAN mixing ratios. This increase in both the average NO• and the PAN mixing ratios has been attributed to biomass burning [Talbot et al.,

Comparison of the Partitioning Observed in Nonurban Air
The partitioning of compounds within the N,,Oy budget is summarized in Table 2. Also included are the partitionings observed in several other field programs that were recently summarized by Ridley [1991]. The comparison listed in Table 2 omits the contributions from p-NO3' due to the uncertainties associated with its contribution to measured NOy, or for that matter the "reactive" N•Oy budget (see also discussions by Atlas et al. [1992]    particulate-nitrate (p-NO3') mixing ratios are given when available, but they were not used in the calculation of residual values. Inspection of Figures 6-8

35% of the NOy accounted for by the •NOy(i). Above 3
km the data are also predominantly out-of-balance but with a larger fraction (---55%) of the NOy accounted for by the •NOy(i). Above 3 km the Hudson Bay lowland data are also more out-of-balance. However, below 3 km the Hudson Bay lowland data are more evenly distributed with respect to overall estimates of the residuals' uncertainty.

For this case, NOy mixing ratios range from approximately 100 to 750 pptv, with a noticeable decrease in average values between the in-and out-of-balance categories (-400 pptv versus --200 pptv, respectively). In contrast, over the northern Labrador/Quebec study region the data from both altitude ranges (i.e., > 3 km and < 3 km cases) are predominantly in-balance with the •NOy(i)/NOy ratio having a mean value near unity. For comparison, a statistical summary of several of the in-and out-of-balance cases discussed above is given in Table 4. Below 3 km the out-of-balance Hudson Bay lowland data have mean mixing ratios that are approximately fourfold smaller for PAN and HNO 3 and approximately twofold smaller for NOx and NOy compared
to the corresponding in-balance data. In this out-ofbalance case the mean mixing ratios of the NOy(i) compounds and NOy are closer to those observed in the lower-altitude Alaska data, which are predominantly outof-balance (if the smaller mixing ratios of HNO 3 observed in the Hudson Bay lowland case were due to increased wet deposition). The in-balance northern Labrador/Quebec data have nearly identical median mixing ratios of NOn both above and below 3 km even though, as might be expected, PAN mixing ratios increased by approximately threefold in the higher-altitude regime. This increase in PAN mixing ratios more than offsets, by about twofold, the decrease in median HNO 3 mixing ratios allowing PAN to be the key species controlling the NOy budget within this altitude range over northern Labrador/Quebec. Talbot  These data were not included in any of the calculated regressions for the out-of-balance data and, in general, the plume data will be discussed separately. In contrast to the change in covariance implied from the

NOx versus NOy regressions, the regression slopes for PAN versus NOy imply that on average there is little change in
the covariance relationship between the in-and the out-ofbalance data. This might suggest that on average the outof-balance data may represent more aged air parcels that had experienced a larger loss of NOx (e.g., via OH + NO x -+ HNO3) relative to PAN. The lack of an enhanced HNO3 to NOy relationship or for that matter the lack of any strongly perceivable relationship in the out-of-balance data appears to contradict this hypothesis. However, as indicated by the small HNO3 mixing ratios in the cloudprocessed plume data, HNO3 solubility precludes it from being a good surrogate for inspecting covariance relationships and suggests the need to examine correlations with respect to other trace gases.  16 15 14 16 17 19 20 15 20 17 13 17 18 14 10 10 13 14  13 17 19 16 13 17 13 19 13 17 18 17 18 14 13 10  establishing the ensemble average of an air mass's relative age. This precludes the unambiguous identification of sources and sinks based on any one "chemical clock's" estimate of an air mass's age and certainly precludes directly determining causation from any individual correlative trend. Thus, whereas it is tempting to argue that the on average correlation between 03 and NOy is suggestive of a common stratospheric source, it is perhaps more appropriate to suggest that this correlation merely reflects the common behavior of two species that are reasonably long-lived in the upper troposphere (as is CO) and that share similar loss rates due to photochemical and/or deposition processes as air parcels become distributed (and diluted) on a hemispheric scale. The small HNO3 fraction of NOy in these data sets also support this latter argument, unless an efficient mechanism can be identified that is capable of converting stratospherically derived HNO3 back into more active forms of NOy.

Recent laboratory measurements indicate that NOx can be generated from the surface catalyzed photolysis of H2SO 4 and HNO3 mixtures [D. Fahey, private communication].
These observations could suggest that a similar aerosol coupled mechanism might be responsible for reliberating the reactive nitrogen pool that has been thought to become irreversibly tied up as HNO3. The implications of such possible mechanism to the distribution of NOx over remote regions certainly warrant careful laboratory study. Another possible explanation is the residence time in the upper troposphere (> 6 km) is sufficiently long to allow OH oxidation and photolysis to slowly reactivate the stratospheric NOx pool that is tied up in HNO3. Ratios of carbon containing compounds can still, however, provide some useful information about a particular air mass's combined photochemical history relative to other air masses. In particular, the ratios C2H2/CO and C3Hs/C2H6 have been used to reflect the degree of combined mixing and photochemical processing that has occurred within an air mass (see discussion by

McKeen and Liu [ibid]). During this atmospheric processing, the dominant tropospheric chemical loss process for all four compounds is their oxidative reactions
with OH. The faster reaction rate coefficients [e.g.,

DeMore et al., 1992] for C2H 2 and C3Hs versus CO and C2H 6 yield expectations that these two pair of ratios should
have similar values that decrease as an isolated air mass photochemically ages. In our ABLE 3A investigations, NOy was well correlated with ratios of C2H2/CO [Sandholm et al., 1992]. This tendency also generally holds for both the in-and out-of-balance data (see Figures 13a  and 13b). However, the out-of-balance cloud-pumped plume data do not follow the trend expected from either the in-balance or the nonplume out-of-balance data. These plumes did appear to be depleted of HNO3 relative to NOx or PAN, as discussed earlier. Somewhat surprisingly though, these plumes tend to follow the general trend exhibited between the ratios of C2H2/CO versus the gNOy(i) (see Figure 13c). This might suggest that the more reactive forms of NOy (i.e., NOx and PAN) may have been in sink with the average degree of atmospheric processing that had occurred in the middle/lower troposphere over these regions, whereas the substantial budget deficit for NOy-ZNOy(i) may represent compounds that were significantly out of sink with the relative degree of atmospheric processing that is indicated by the ratios of C,.H•/CO. Ridley [1991] presented arguments that the abundance of a missing compounds implied from comparison of NOy-gNOy(i) are not enhanced during air mass aging. Our results agree with this argument. However, our results also indicate that the relative abundance of the implied missing compounds increases with the degree of atmospheric processing (see Figure 13d). This might be expected if the implied missing compounds are less susceptible than NOx, PAN, and HNO3 to atmospheric loss processes (e.g., oxidative attack in the case of NO•, thermal decomposition in the case of PAN, and the final dry/wet deposition processes that remove the HNO3 formed from NO• and its reservoirs).
The in-balance mixing ratios of NO v are also correlated on average with the degree of atmospheric processing established by the ratio C3Hs/C2H 6 (see Figure 14a). However, this trend tends to disappear in the out-ofbalance data (cf. Figure 14a and 14b). This supports the argument that the various relative "chemical clocks" of the different air masses represented by these data were all perhaps on somewhat different cycles that reflect different degrees of mixing and photochemical ageing. We believe this is borne out by the loss in correlation between the ratios CsHs/C2H6 and C2H2/CO in going from the inbalance subset of data to the out-of-balance subset of data (cf. Figure 15a and 15b). In general, our attempts at directly investigating the possible factors that influence 'the residuals have been complicated by (1) the apparent random resetting of selected surrogate "chemical clock" relationships, which appears especially pronounced in the out-of-balance data and (2) the generally random nature of the in-balance residual values that appears to be dominated by (1) the random uncertainties of the measurements and (2) the random uncertainties introduced by ambient variability in conjunction with the less than unity temporal overlap between the various NOy(i) and the NOy measurements, including perhaps sporadically enhanced p-NO 3' events occurring within the long sample time of the aerosol measurements.
In addition to 03 mixing ratios (see earlier discussion) the only other compounds that appear to correlate with the values of the out-of-balance residuals are the mixing ratios of PAN and perhaps NO• and HNO 3 (see Figure 16a- masses. These trends might support Jacob et al.'s [1992] argument that in the summertime middle free troposphere over Alaska an additional reservoir-derived source of NOx, besides that produced from the thermal decomposition of PAN, is necessary to account for the observed mixing ratios of NOx. If the NOy budget deficits represented compounds that are capable of generating NOx, then they could also enhance the photochemical production rate of Os and possibly reinforce the observed Os to NOy trends. However, it is doubtful that these missing compounds are simple alkyl nitrates based on the small fraction observed thus far for these compounds relative to NOy or even PAN Buhr et al., 1990]. On average, the residual values are also negatively correlated with temperature (see Figure 16d). This is consistent with there being labile reactive odd nitrogen compounds other than PAN, such as HObNOb, that are thermally stable in the cold upper troposphere [e.g., Logan et al., 1981]. Even though thermally stable at high altitudes, HO•NO• has not been predicted to be a major NOy ( An alternative explanation for these trends that we must also address is the possibility of one or more of the NOy(i) or NOy measurements is in error. Some types of errors in one or more of these measurements would .still be consistent with the out-of-balance residual values being in some way correlated with factors influencing the oxidative potential of the air masses. In the case of NOn it is unlikely that these measurements are in error by the very large factors (i.e., tenfold) that would be necessary to explain the budget deficit. This is based on the level of agreement, which has been shown in intercomparison programs, between this technique and the others (see discussion in section 2.1). In addition, potential NO: measurement interferences have .,.a tendency to have positive values (e.g., HO2NO2 -* NO2), which would argue that the NO• values might already represent upper limits.
In addition, the NO•/NOy and PAN instruments were intercalibrated using common NIST NO and NO: standards. This should have significantly reduced the potential for differences in primary standards being the cause of the NOy budget deficit. budget deficit might exist. In particular, conversion of compounds such as X-CN or X-NH could also be consistent with the correlative trends in O3 and PAN, if these compounds became eventually oxidized to produce NO x. Biomass burning, which impacted all of the ABLE-3 study regions, has been shown to produce X-CN compounds [Lobert et al., 1990]. Therefore this, speculation, that of X-CN compounds, could also be consistent with the ABLE 3A ground-based NOy measurement results. The concurrent ground-based NOy, HNO3, and NOx measurements in the ABLE 3A program showed good agreement in both the magnitude and the partitioning of NOy with the lowest altitude (-0.15 km) measurements taken on the aircraft [Bakwin et al., 1992]. They attributed nearly half of the ground-based measured NOy to "missing" compounds (-100 pptv) and very little to the lablie compound PAN. Their tower flux measurements also indicated that these missing compounds also exhibited a small deposition velocity (i.e., sixfold smaller than 03).
This would suggest that the missing NOy compounds in the lower-altitude range are relatively unreactive. Subtle changes in the catalytic convertor Au-surface's condition might significantly reduce the conversion efficiency for these more difficult to convert compounds (i.e., X-CN) without producing a noticeable change in the conversion efficiency of reducible compounds such as HNO3, NO2, or PAN. This type of'mechanism could possibly explain the observed shift in NOy budget trends for the predominantly in-balance northern Labrador/Quebec data. This hypothesis would also be consistent with the suggestions that some portion of the NOy budget deficit is comprised of relatively unreactive nitrogen-containing compounds. The above hypothesis suggests that more detailed conversion efficiency tests need to be carried out under a wide range of conditions for a number of compounds.

SUMMARY
The summertime partitioning and budget of tropospheric NxOy compounds have been investigated in the middle to lower troposphere over three high-latitude regions of North America. These investigations are the first to use a spectroscopically selective measurement system to detect the NO produced from a NOy Au-catalytic converter. The NxOy budget was analyzed for the degree of closure based on the balance obtained from comparing observed NOy mixing ratios to the sum of individual NOy(i) species (NOx, HNO3, PAN, and PPN). Within the estimated precision and accuracy of the measurements, statistically significant differences were found between the ABLE 3 study regions.
In the middle free troposphere (3 -6 km) over Alaska and the Hudson Bay lowlands, approximately 73% of the analyzed measurements indicate a deficit in the NOy budget at the 95% confidence limit. The NOy budget within the lower altitudes (0.15 -3 km) also had a deficit. In these cases, approximately 87% of the Alaska and 35% of the Hudson Bay lowland measurements indicate a deficit at the 95% confidence limit.
Over the northern Labrador/Quebec region the NOy The large deficit in the NOy budget (having •NOy(i)/NOy < 0.6) over regions of Alaska and the Hudson Bay lowlands appears to be related to factors that influence the oxidative potential of the troposphere over these regions as reflected by the mixing ratios of 03. Based on our discussion of possible measurement errors and the correlative tendencies exhibited with other compounds, it is unlikely that the bulk of this NOy budget deficit is due to a simple artifact of the measurements. A portion of this NOy budget deficit is also suggested to be in the form of relatively unreactive nitrogen-containing compounds, especially at the lower altitudes. However, based on the correlations with 03 and temperature, a larger fraction of the implied NOy budget's missing compounds exhibit trends that are more indicative of the labile NOx-reservoir PAN. This could suggest that these missing compounds either have chemical characteristics that are similar to PAN or that they have a similar atmospheric source as PAN.
perseverance in making and maintailfing logistical arrangements throughout all of ABLE 3B. The investigators at Georgia Tech would like to thank Gerhard Htibler for his helpful suggestion and comments on the development of their NOy system and for providing the authors with preprints of his manuscripts. We also appreciate helpful discussions with Shaw Liu, Stewart McKeen, and Dave Fahey, and the helpful suggestions from our two anonymous reviewers. We would also like to acknowledge the contributions of Sandra Farher and Ding-Jun Yang toward the preparation of this manuscript. This research was sponsored by the National Aeronautics and Space Administration, Tropospheric Chemistry Program.