Atmospheric NO' Measurements to Determine Its Sources, Sinks, and Variations

Measurements of atmospheric N20 concentrations in and around between August 1976 and September 1977 yielded 329.5 ppb (parts per billion) (mole fraction in whole air samples) as the average with a standard deviation of 3.3 ppb. No seasonal or overall trend is discernible. Small, marginally significant diurnal variations of N20 concentrations in the lowest 0.5 m of atmosphere can be seen in our data. Gas collectors applied to soils near our laboratory provided evidence that soils can consume atmospheric N(cid:127).O under some conditions. Soils were seen to be N(cid:127).O sources more often than sinks. Stronger source activity was evident in compost piles. Exhaust samples taken from several conven- tional vehicles showed less than ambient N(cid:127).O concentrations, while catalytic converter equipped autos produce N(cid:127).O. Samples of steaming volcanic vent air from Hawaii indicate that volcanoes are probably a source of atmospheric N(cid:127).O, although emissions at the Sulphur Bank contain much less than ambient values of N(cid:127).O. All measurements were made by electron capture-gas chromatography; techniques and equipment are described.

which is thought to be the dominant source of stratospheric nitrogen oxides [Crutzen, 1971], even though only a few percent of atmospheric N:O undergoes (Rlb). It is also involved in the cycles of nutrient nitrogen in soils and water bodies as a p.roduct of denitrification and other nitrate-reducing processes (see the review by Payne [1973]); nitrous oxide from these origins can escape to the atmosphere. Thus gaseous N:O can serve as an indicator of microbiological activity. Indeed, there is 10 •8 more N2Oin the atmosphere than would exist in an abiological thermodynamic equilibrium atmosphere [Lovelock and Margulls, 1974]. Also, the fact that N20 is produced in certain combustion processes Weiss and Craig, 1976] makes it a potential tracer of combustion plumes. On a larger scale, the vertical and latitudinal profiles of N:O can serve as estimators of global circulation patterns and rates [Schmeltekopf et al., 1977]. Also, if N20 levels ever undergo large increases, there could be a significant trapping of outgoing planetary radiation [Wang et al., 1976]. Because of (R lb) and subsequent chemical reactions of NO and NO2 that catalytically destroy O8 in the middle and high stratosphere, there is a basis for concern over any biospheric activities that might alter global N20 levels. In particular, there is growing interest in the question of how much effect man's increased nitrogen fixation and combustion rates will have on atmospheric N20 [Crutzen, 1974[Crutzen, , 1976McElroy et The N:O measurements reported here address the questions just mentioned. Specifically, we have performed some exploratory studies to try to identify sources of atmospheric N:O and to begin to estimate the strengths of these sources. We have searched for N:O emissions in several nitrogen-rich environments; automobile exhaust, air from cattle feedlots, compost Sample loops are filled to atmospheric pressure; sample loop volumes are 1 ml in most studies, but 5-ml loops are used occasionally. UHP is ultrahigh purity. *Two-column (Molesieve 5A and Porapak Q) tandem system. piles, fertilized soils, and natural soils and volcanoes. Results of these source identification studies are presented in section 3. In the course of our soil gas emission studies we have also found some evidence of N:O consumption by soils. In less exploratory work we have regularly measured N:O ambient concentrations at several locations around Ann Arbor, Michigan, and we report here the results of 1 year of monitoring. We have also sought evidence of diurnal variations of near ground level. We are paying attention both to absolute values and calibrations and to relative variations of Results are presented in section 3 and discussed in section 4.
Our instruments and methods are described in section 2.

INSTRUMENTATION AND TECHNIQUES
In this section we describe the instrumentation and methods that we have used to gather and analyze air samples for N:O. The basic principles and methods of gas chromatographic separation and electron capture detection of N:O have been discussed by Wentworth and Freeman [ 1973] and by Rasmussen et al. [1976]. Table 1 lists the chromatographic equipment and conditions that we have employed. As is mentioned later, in March 1977 we adopted the Porapak Q column and the conditions listed in the first row of Table 1; this led to significantly better precision than had been achieved earlier with Porasil. One reason for this improvement is that CO: and N:O are better separated by Porapak Q than by Porasil B, at least for the conditions listed. Figure 1 shows a typical chromatogram of rural Ann Arbor air (•325 ppb (parts per billion) •330 ppm (parts per million) CO:) with chromatographic configuration A (see Table 1). Although it is not completely germane to the goal of the present paper, the sign of the CO: response in EC-GC systems appears to be determined capriciously, and we note that further investigations are needed. We find a positive CO: peak with configuration C, and P. D. Goldan (private communication, 1977) finds a negative peak with one GC-EC system and a positive peak with another system with the same carrier gas. Also, Lovelock [1974] reported that system sensitivity to was enhanced by CO: in the mixture. Detector geometry and surface condition, carrier gas impurities, and ion-molecule reactions in the system [see Parkes, 1972] all may be involved, and chromatographic interferences are possible. Absolute calibration of the GC-EC system was performed by diluting pure N:O with pretested zero-grade air; commercial calibration mixtures were avoided. Two distinct and independent dilution methods were devised. One is a multistaged feedback flowing dilution system; the other uses an permeation tube which injects a small calibrated flow of pure N:O into a much larger (measured) diluent flow stream. The permeation tube flow is not measured gravimetrically as is common elsewhere [e.g., Dietz et al., 1974] but is measured instead by scanning pressure in a microvolume differential pressure transducer. Both the feedback flow and the permeation tube dilution systems can be used easily between about 50 ppb and 20 ppm (concentrations by volume) of or 1000 times smaller concentrations of CF:CI:. The permeation tube dilution system easily generated 100-1000 ppb N:O and 100 ppt (parts per trillion) to 2 ppb CF:CI:. These flow dilution systems were used to prepare flows with known concentrations of N:O, and the GC-EC response to each concentration was measured. Figure 2 shows one such set of results. Between the various N:O flows, chromatograms of the standard tank to be calibrated were obtained. In this way a calibration curve was constructed, and a secondary standard (described below) was calibrated. By generating a calibration  curve and mapping out the system behavior over a range of concentrations, e.g., 0-600 ppb, one tests the system for linearity and demonstrates that there are no spurious, nonzero signals at zero N:O concentration. The radii of the data points in Figure 2 represent actual standard deviations of each data set. High-precision work like this obviously requires a stable, sensitive GC-EC, pure carrier gas (and carrier stream purifiers), and great care. Occasionally, calibration runs did not yield data points to fit a line as well as is shown in Figure 2 or with a zero intercept. To date, we have gathered seven high-quality calibration curves like Figure 2, four with the feedback flow and three with the (independent) permeation tube flow dilution system. The agreement between the concentrations assigned to the standard tank under study in these seven calibrations can be measured by the statistics of seven independent trials. Their mean value was 383 ppb, and l standard deviation was 10 ppb. Thus we believe that our absolute N:O values are known to +5.2% with 95% confidence. In addition to these absolute calibration studies we have traded two samples each with P. D. Goldan (National Oceanic and Atmospheric Administration (NOAA)) and J. Krasnec and L. Heidt (National Center for Atmospheric Research (NCAR)). We found our values to be I% higher than NCAR's and almost 2% higher than NOAA's.
More extensive intercalibration studies are underway because of the great desirability of directly intercomparing data from many laboratories, some of which disagree by as much as 15% (see the discussion by Pierotti and Rasmussen [1977]). The N:O reference standard mixtures mentioned above are stored in pressurized cylinders and occasionally checked by absolute calibration. One standard was prepared by introducing N:O into an evacuated commercial (•50 l) gas cylinder that had previously contained ultrahigh-purity He. The tank was then pressurized to 1500 psi (10 ? Pa) with zerograde air. During the process we also arranged to obtain nearambient amounts of CO:, CF:CI:, and CFCIa. A second in-ternal reference standard tank was prepared similarly in a 1-1 electropolished stainless steel cylinder at 1000 psi (6.9 X Pa). These two internal standards were prepared 6 months apart, and since the preparation of the second one, they have been measured in ratio for 5 months with no detectable change. Thus we believe that our reference standards have been stable so far but also that longer storage times will require continued checking and possibly more elaborate preparations. The ability of our rather rapid calibration techniques described above (and more fully by Stedman et al. [1977]) to perform frequent tests is advantageous in such work.
Our sample flasks are made from no. 316 stainless steel and are extruded with a center weld. Nupro bellows-seal stainless valves with stainless seats are used through ]-in. (1-cm) stainless tubing, and the flask volume is 2.7 1. The flasks were first electropolished and then rinsed with distilled water followed by absolute ethanol. The tubing and valves were then welded in place, and the assembly was attached to a high-vacuum system (•10 -9 torr) and baked at 400øC. Actual data analysis consists of measuring the GC-EC response to a sample and then the response to our secondary (or transfer) standards, which are calibrated as described above. The GC sample loop is filled to the prevailing atmospheric pressure after being thoroughly flushed with each sample to be injected. Day-to-day variations in atmospheric pressure are of no concern because of this ratio method. In practice, GC response to our secondary standard is measured before and after each sample is analyzed.
Both peak height and peak area analyses have been performed on the data. As is shown later, our ambient N:O data improved in precision in mid-March 1977, at which time we adopted an electronic peak integrator (HP 3380A) and a Porapak Q column. Clearly higher precision is possible by using electronic digital integration than by measuring peaks from a chart record.   Table 2). From the EPA test facility we obtained Tedlar bags fitted with Nupro valves filled with diluted (X 15) exhaust and control bags of background air. Each bag was analyzed within hours; stability tests of N•.O in the bags assured us that tests after 24 hours were still acceptable. Similarly, we found that conventional exhaust systems,    Figure 1) were observed to be much larger and more regular. In most cases, CO•. would attain a peak in the early morning hours and return to normal levels by late afternoon. Perhaps the most interesting behavior was noted on 3 days in May, 8 days in June, and 3 days in July 1977. On these days, N•.O was noted to decrease in the funnels from its initial value (327 ppb) to about 150 ppb in funnels placed in very wet, mostly sun-shaded grassy areas. This behavior was repeatable; removing and cleaning the funnels and placing them in new spots led to identical results in four separate experiments between mid-May and mid-July 1977. CO•. was seen to peak in early morning and to decrease to ambient values by midafternoon each day. These tests were conducted with two or three funnel collectors side by side. The redundancy in our design and the marked contrast to simultaneous test results from drier, sunnier test spots only 25 m away lead us to suggest that soils can consume atmospheric N•.O as well as produce it. From these exploratory studies it appears that season, soil moisture, temperature, and many other factors can regulate this possible sink activity. Clearly, our experiments altered the local environment in and under our funnel collectors; CO•. concentrations rose and fell, and the glass served as a heat greenhouse. On the positive side, however, we did not perturb the subsurface temperature, and our small (625

cm •') covered area was in contact below the surface with a quasi-infinite earth. Also, because our experiments show the ability of soils to consume N•.O from air containing only normal (329 ppb) N•.O concentrations, they supplement the much more elegant and controlled experiments of Blackmer and Bremner [1976]. These authors, while the first to show that soils can consume N•.O, employed initial N•.O concentrations in the 10 -• range. Similar funnel collection studies in an oak forest with natural ground cover (mostly leaves) have not yet shown any sink activity.
We have also used our funnel collection device at organic compost piles containing vegetation and mixed animal manure. There we found more rapid increases of N•.O, usually 1000 ppb/h, if the collector was over a firm, crusted spot. On new compost or on older piles where we intentionally punctured the surface we measured even more rapid increases: up to 20 ppm in 1 hour with little further increase.
A few more exploratory measurements that have provided

indications of N•.O source activity have been performed. First, as was noted above, new or punctured compost piles emit N•.O more rapidly than those with a well-formed crust. When we gathered an air sample from the air a few centimeters away from a new, loose organic compost pile in an open field, we found an N•.O concentration of 3.6 ppm. Similarly, about 50
cm from a crusted dung heap we found 345 ppb, a significantly higher amount than the 327 ppb 25 m away.
To investigate the possibility that volcanoes can produce N•.O, we gathered air samples at five locations on the island of Hawaii; at each site, active steam venting was underway, but no molten lava was flowing or exposed. Table 3 Table 3   contaminated. In any case, the average N20 concentration over the remaining 50 data points of Table 4 is 329.5 ppb with a standard deviation of 3.35 ppb. This standard deviation is similar to that reported by Goldan et al. [1978]; thus Junge's [1974] method to estimate the atmospheric residence time of N20 from its measurement statistics yields the same value when it is applied to our data as when it is applied to their selected highest-precision data, namely, about 14 years. It is clear that Junge's formula is approximate and that a portion of the observed variance is instrumental and not necessarily variance in actual N20 concentrations. Indeed, it is clear from  the standard deviations of data points since February 1977 are smaller than those before that date. Low-precision measurements taken on several days during August 1976 were grouped together as one point, as was another set in late October and early November 1976 (see Table 4 Table 4). In the course of gathering these data we have improved our system precision; thus the most recent data are of higher quality. We have also seen the necessity of analyzing the samples for tracers of man-made pollution such as CF:CI:, CFCla, and CO. After discarding several samples that showed higher  [1978] show that N20 is produced in municipal sewage treatment and water treatment.

The CO•. concentrations in
performed, for useful discussions and many North Lake air samples.