A comparison of aircraft and ground-based measurements at Mauna Loa Observatory, Hawaii, during GTE PEM-West and

. During October 19-20, 1991, one flight of the NASA Global Tropospheric Experiment (GTE) Pacific Exploratory Mission (PEM-West A) mission was conducted near Hawaii as an intercomparison with ground-based measurements of the Mauna Loa Observatory Photochemistry Experiment (MLOPEX 2) and the NOAA Climate Modeling and Diagnostics Laboratory (CMDL). Ozone, reactive mtrogen species, peroxides, hydrocarbons, and halogenated hydrocarbons were measured by investigators aboard the DC-8 aircraft and at the ground site. Lidar cross sections of ozone revealed a complex air mass structure near the island of Hawaii which was evidenced by large variation in some trace gas mixing ratios. This variation limited the time and spatial scales for direct measurement intercomparisons. Where differences occurred between measurements in the same air masses, the intercomparison suggested that biases for some trace gases was due to different calibration scales or, in some cases, instrumental or sampling biases. Relatively large uncertainties were associated with those trace gases present in the low parts per trillion by volme range. Trace gas correlations were used to expand the scope of the intercomparison to identify consistent trends between the different data sets.

related to the individual measurements must be specified to allow meaningful interpretation and to compare with model calculations which rely on the entire suite of chemical measurements. Errors and uncertainties in any of the measurements may have profound implications on the overall data interpretation [e.g., Liu et al., 1992].
A number of different instrument intercomparison exercises have been carried out in recent years to examine the reliability of current measurement techniques. These include aircraftbased programs such as the Global Tropospheric Experiment (GTE) Chemical Instrumentation Test and Evaluation (CITE) series and a number of ground-based studies which address specific measurement issues [e.g, Fehsenfeld et al., 1987;Hoell et al., 1990Hoell et al., , 1993Kleindienst et al., 1988]. These intensive intercomparisons provide valuable information on biases or offsets between techniques in a given experimental setting, but they do not necessarily reflect the "routine" application of the different measurement techniques; in addition, not all chemical species normally measured in a single comprehensive campaign have been intercompared.
As chemical models become more sophisticated, a combination of chemical measurements obtained from different field programs and from different investigators will be requi, red to test model predictions. However, there have been few instances during which different measurement programs can be directly intercompared. The Pacific Exploratory Mission-West (PEM-West) and Mauna Loa Observatory Photochemistry Experiment 2 (MLOPEX 2), both studying aspects of the photochemistry of the troposphere over the Pacific Ocean, were performing many of the same measurements. Thus, colocation of the two programs for one of the PEM-West flights offered the opportunity for an intercomparison under conditions which were "normal" for each measurement program.

The intercomparison which could be conducted near the Mauna Loa Observatory (MLO) site provides an optimum situation to examine data intercomparibility between different investigations.
This report summarizes and compares chemical measurements from the DC-8 aircraft and the ground based MLO site during flight 20 of the PEM-West mission. Not all measurements were performed at both sites, and a description of these data are not included here. More detailed descriptions and interpretations based on the entire data sets are presented in a series of papers elsewhere [see Hoell et al., 1996 (PEM-West A Special Issue]. This report includes comparisons of ozone and a variety of trace gases used to assess oxidant budgets in the remote troposphere. These species include tracer species (halogenated hydrocarbons, nonmethane hydrocarbons), radical reservoirs (peroxides), odd-nitrogen species (NOy, NOx, PAN, nitric acid), and reactive hydrocarbons (alkenes).

Measurements and Techniques
The different measurements which will be compared are given in Table 1. A brief description of the individual techniques and estimated errors are presented here, but details of the instrumentation and methodology are given in references by the individual investigators.

Ozone
For the ground-based measurements, instrumentation is based on UV absorption measurement. The MLOPEX instrument was a pressure and temperature compensated Thermo Environmental Instruments (TECO) Model 4 9 calibrated with a TECO Model 49-PS calibrator.
Data are reported as 1 min averages. The PEM-West instrument is based on ozone-ethylene chemiluminescence which is calibrated with NO-O3 gas phase titration traceable to National Institute of Standards and Technology (NIST)NO standards [Gregory et al., 1983[Gregory et al., , 1992. Each of the ozone instruments is reported to have a precision of at least +_2 ppbv Gregory et al., 1983]. An ozone instrument was also in operation at the CMDL, but that instrument reports hourly averages and the data are not included in this comparison. We note that throughout the MLOPEX intensives, there was good agreement between the CMDL and MLOPEX ozone data. For example, during the free tropospheric flow periods of the first intensive, the mean difference between the two instruments was only 1.1 (+1.8) ppbv (MLOPEX >CMDL), which indicates agreement within the estimated precision of the techniques.
Total Reactive Oxidized Nitrogen (NOy) N Oy was measured in two airborne systems and from a 10-m tower at the MLO site. All instruments rely on the goldcatalyzed reduction of reactive nitrogen to NO [Bollinger et al., 1983] which is detected using either laser induced fluorescence (LIF) [Bradshaw et al., 1985;Sandholm et al., 1990] or chemiluminescence (CL) detection [Hiibler et al., 1992a,b;Kondo et al., 1996]. The airborne instrumentation included one LIF and one CL system; the ground-based instrument was a CL system. Data are reported as 1-min averages for the chemiluminescence-based systems and 90-s averages for the LIF system. Estimated uncertainty of each is 10-20 %.

Nitric Acid and Aerosol Nitrate
Nitric acid was measured using three different techniques at the MLO site, and a fourth technique was used aboard the DC-8. One ground system was based on collection of aerosol nitrate and nitric acid on a Teflon/nylon filter pair ]. The second system used an aqueous-based

NO and NO2
NO was measured using a NO/O3 chemiluminescence technique at both the ground site Ridley et al., 1994] and aboard the DC-8 [Kondo, 1996]. The second airborne system was based on the LIF technique [Bradshaw, 1995]. NO 2 measurements were obtained only with the ground-based system and the airborne LIF instrument. Both instruments measured NO 2 by photolyric conversion of NO 2 to NO, and the ground-based system cooled the photolysis cell to 0øC. Detection limits for both airborne instruments were better than 2 pptv for NO and 4 pptv for NO 2. The ground-based system quotes the uncertainties for mixing ratios <100 pptv as NO, ___(0.9 + 4% reading) pptv; NO 2 (night), ___(2.5 + 4% reading) pptv. Overall uncertainties estimated in these systems were on the order of 15 -20%.

Peroxyacetyl Nitrate (PAN)
PAN measurements in the ground and aircraft systems were based on gas chromatography with electron capture detection.
Both the MLOPEX instrument and the airborne instrument cryogenically preconcentrated air for analysis Singh et al., 1992Singh et al., , 1994]. Limits of detection were <1 and 4 pptv in the ground-based and airborne instruments, respectively. Uncertainties in the measurement are estimated to be in the range of 8 -15%.

Hydrogen Peroxide/Organic Hydroperoxides
Several different systems were used to measure hydrogen and organic peroxides. At the MLO site, hydrogen peroxides were measured by three systems. A continuous dual-enzyme hydroperoxide analytical system provided the highestfrequency data [Lazrus et al., 1986]. A second system using cryogenic collection and high-performance liquid chromatography (HPLC) supplemented the coil data, but provided only hourly integrated data [Kok et al., 1995]. These systems measured both hydrogen and organic peroxides. A tunable diode laser instrument (TDLAS) was the third technique to measure hydrogen peroxide, and half hourly integrated data were reported [Mackay et al.,

this issue]. A detailed comparison of these measurements is presented by Staffelbach et al. [this issue].
Aboard the aircraft, an aqueous coil system similar to the ground-based instrument was used to measure hydrogen and organic peroxides [Heikes et al., 1996]. A second airborne system used high-pressure liquid chromatography to analyze hydrogen and organic peroxides, and this system was used as a basis of standardization of the continuous coil instrument.
Instrument detection limits were on the order of 30 pptv for hydrogen peroxide in the fluorescence based instruments and near 100 pptv in the TDLAS instrument. Organic hydroperoxide detection limits were near 50 pptv. Uncertainties in the peroxide instruments are estimated to be on the order of 15%.

Non-methane Hydrocarbons (NMHC)
Aboard the aircraft, whole air samples were collected in stainless steel canisters, and the NMHC were analyzed using gas chromatography with flame ionization detection [Blake et al., 1994]. The system was calibrated with a secondary whole air standard referenced to NIST scales. The calibration mixture was analyzed between every four sample canisters. Detection limits reported for the aircraft system were between 3 and 5 pptv, depending on the specific compound. NMHC analyses at the MLO site were performed using an automated in situ GC/FID system ]. In this system,

Sulfur Dioxide (SO2)
Sulfur dioxide was measured on the DC-8 as part of an isotope dilution GC/MS system for sulfur species [Thornton et al., 1996]. Integrated samples (40 s) were analyzed every 5 min in this system. Detection limits are estimated at <4 pptv with an overall uncertainty of better than 10%. The groundbased system was a modified TECO pulsed fluorescence instrument. Detection limits from 1-min averaged data were estimated at _+66 pptv, with an overall uncertainty of 20% (G. Hiibler, personal communication, 1992). The purpose of the ground-based instrument was to detect local volcanic emissions at the site rather than to measure background levels of SO 2.
The airborne measurements for the halocarbon compounds were included as part of the analyses of whole air from stainless steel canisters [Blake et al., 1994]. Precision of the halocarbon analyses range from 1 to 5%, with an estimated accuracy of between 2 and 10%. We note that the canister data used in this comparison are modified fi'om the original archive to correct for a calculation/transcription error in the data file.  . Meteorological conditions at the site have been well described and show the site to be strongly influenced by a diurnal upslope/downslope flow regime. For some period during most nights, the downslope flow allows air representative of the free troposphere (FT)near 3.4 km to be sampled. For this reason, the PEM flight was scheduled to be in the vicinity of the site and near the island during a portion of the "normal" downslope flow period (often fi'om 2200 to 1000 HST).
Atmospheric conditions near the island of Hawaii are routinely monitored with twice daily soundings. The sounding from Hilo at 0200 on the night of the flyby shows that MLO (680 mbar)was located above the trade wind inversion (720-750 mbar); winds from 900 to 550 mbar were fi'om the east to east-southeast at 5 -10 knots; dew point decreased from-10øC at 700 mbar to -24øC at 500 mbar ( Figure  1 Hawaii (leg 1). This leg was followed by a chevron pattern at about 400 m altitude off the eastern edge of the island (0500 -0534 HST); the leg ended off the southern shore of the island (leg 2). The plane then ascended to 3.3 km (near the altitude of the MLO site) and proceeded in a counter-clockwise circle of the island (0548 -0652 HST)(leg 3). The island circle was followed by a second chevron at about 4.8 km (leg 4) stacked above the first chevron leg (0700 -0732 HST). After this leg,' the plane headed on a north-south transect near 3.6 km which passed within 20 km of the MLO site at approximately 7:50 HST (7:34 -8:00 HST) (Leg 5). A low-altitude chevron was repeated after the MLO site overflight (leg 6) and before the plane returned to Honolulu. This flight track is illustrated in Figure 3.
The rationale for the flight track allowed for examination of trace gas composition in the marine boundary layer (MBL) as well as a comparison between airborne and ground-based measurements in free tropospheric air, though we consider

Results
The basis for comparing measurements from the aircraft and from the ground site is not as straightforward as it was planned, even for a site in the remote midtroposphere. On a scale that is required for detailed intercomparisons, there were The problem is to choose the most appropriate data for intercomparison in this reasonably complex meteorological situation. Even given the observed winds, there is not sufficient information to be certain that the same air parcel sampled on the ground was also sampled by the aircraft. The  Table 2) and "B" is the overflight (see Table 3).

Long-lived Halocarbons and N20
The agreement between canister sampling (PEM) and in situ analysis for the fluorocarbons CFC-12 (CC12F2)and CFC-11 (CC13F) (NOAA/CMDL) is excellent (within 2%) for both compounds. Average differences are higher (7 and 5%) for methyl chloroform and carbon tetrachloride. For both of these molecules, the canister results are higher than in situ data.  Tables 2 and 3, the entire data sets were used to provide better statistical significance to the comparison. During the full autumn 1991 intensive there was no measurable trend in N20 mixing ratio. Similarly, there was no measurable difference in N20 at different locations and altitudes of the aircraft. Measurements from the autumn 1991 intensive and the full N20 data set from Flight 20 yield an average (_+2 standard error) N20 mixing ratio of 307.9_+.04 (MLO) and 309.25_+.01 (PEM) ppbv.

Tetrachloroethylene
This compound is more reactive than the halocarbon species just mentioned, and its mixing ratio is nearly 2 orders of magnitude lower. This molecule has been used as a tracer of urban source emissions in the remote atmosphere [Atlas et al., 1992a;$ingh et al., 1994;Blake et al., 1994]. During the comparison periods, the two sets of aircraft measurements agreed to within 1.5 ppw. Evaluation of the time series of data through the flight show no consistent bias between the two sets of measurements, but no significant correlation either ( Figure 5). For the entire flight, both aircraft data sets had a mean mixing ratio of 4 pptv with a standard deviation of 1.3 pptv (in situ) and 0.9 pptv (canister). The ground measurement was significantly lower at 2.4 ppw.
The difference could be the result of a variable residual signal or blank level in the analytical systems or from differences in the standards used. The range of mixing ratios seen here is small, and data are insufficient to separate these effects. On the basis of these data we can approximate an uncertainty of _+1 pptv for the measurement of tetrachloroethylene.

Methane and Non-methane Hydrocarbons (NMHC)
There were only a few measurements of hydrocarbons available for direct comparison, but there seemed to be some consistent relationships between the sets of measurements. During the overflight, these measurements differed by only 4 ppbv, and for the ozone maximum period, the difference was 13 ppbv (<0.8%). Tables 2 and 3). Ethane, propane, and benzene appear to be in very good agreement considering the associated analytical uncertainties. Ethyne and the alkenes do not agree well during the comparison periods. Mixing ratios of butane are too near the detection limit for meaningful intercomparison here.

For the NMHC, agreement between the measurement sets depended on the individual species (See
The data comparison can be extended by examining correlations between different hydrocarbon compounds. A correlation plot of ethane versus propane shows a consistent relationship from the two data sets (Figure 6a). For days 290-296, the MLO data are described by the relationship: ethane=435+2.41 *propane; the PEM-West data the relationship is: ethane=458+2.55*propane. Ethyne appears to be consistently higher in the ground-based measurements, and a correlation plot of ethane versus ethyne confirms the consistent bias between the two data sets (Figure 6b). The data indicate an approximate 20 -30% difference in calibration for ethyne or a consistent bias of about 20 pptv higher mixing ratios in the ground-based measurements. For the C2 and C 3 alkenes, measurements are higher in the canister samples than from the in situ ground-based measurements. During the comparison periods, differences are in the range of a few to 1 5 pptv. Correlation plots from the two data sets indicate the same relative slope between ethene and propene; however, the ground-based data extend to lower mixing ratios of ethene and propene compared to the aircraft data. Difference in the minimum levels are about 10 pptv of ethene and 3 pptv of propene (Figure 6c). Certainly, these differences could be due to the limited sampling of the aircraft during this period and uncertainties at the limit of detection, or possibly to small biases between in situ and canister-based sampling.

Greenberg et al. [this issue] describe a comparison of their
own in situ and canister-based NMHC measurements taken during MLOPEX 2. Interestingly, their data point to the same biases or offsets as described here. The differences they observed for ethyne could be ascribed either to calibration differences between different instruments used to analyze canister and in situ samples or to losses of ethyne in canisters. No cause could be ascribed to the positive bias they observed for alkenes in canisters (ethene, 14 pptv; propene; 8 pptv), but it is interesting to note the same level of difference that is reported here between the in situ and aircraft samples. No ground-based canisters were collected during the flyby for comparison of canister-based collections.
In the remote atmosphere, where the reactive hydrocarbon levels are extremely low, levels of reactive alkenes of even a few parts per trillion by volume may be significant to local photochemical processes. Thus it is important to identify the source of a small bias for ethene and propene mixing ratios. The low levels of alkanes and ethyne do not contribute significantly to local photochemistry, but the ratios of these hydrocarbons are often used to infer photochemical "ages" of air masses or to suggest potential emission sources. Biases and uncertainties associated with these ratios can limit their usefulness, but on-going intercalibration efforts [e.g., Apel et al., 1994] should lead to improved data comparability for NMHC.

Carbon Monoxide
Carbon monoxide measurements at the ground site were 1 1 -18% higher than observed aboard the aircraft. Comparison of the CO/ethane correlations from both data sets suggests that there is a consistent bias between the airborne and ground-

NO + NO2
The comparison during the 0 3 maximum was in the nighttime for both the ground-based and airborne data sets. During the dark, NO is expected to be zero, and all systems were in good agreement to within about 1 pptv for this period. One airborne system (LIF) reported data as limit of detection (S/N=2/1; 1.4 -1.5 pptv), while the other (CL) system showed a 1.2 pptv standard deviation around an average of-0.9 pptv. The ground based system averaged -0.2 pptv (+0.5 pptv). During the early morning flyby period, the sun had risen and NO had increased several parts per trillion by volume. At this time, all systems were comparable at 6.7 + 1 pptv. These data are well within the specified uncertainties of the NO instrumentation.
The agreement between the NO 2 measurements was inconsistent for the two comparison periods. NO2 measurements from both instruments were comparable during the overflight period within the observed variation. A difference of 2.6 pptv during this period was within the standard deviation of the measurements. Measurement of NO2 in the ozone maximum, however, was quite different. A factor of 2 difference (=13 pptv) is found between the airborne and ground-based measurement, and this is near the limit expected from the combined instrumental uncertainties. At this time, we have no suggestions to account for this apparent discrepancy.

HNO3 and NO3'
Nitric acid and particulate nitrate data are difficult to evaluate because of the relatively long sample integration times, and, consequently, few data are available for comparison. During the overflight, three sets of ground-based measurements of nitric acid (using different techniques) averaged =64 pptv with a +10% standard deviation. The single integrated measurement from the aircrat• was 21 pptv. During the ozone maximum period, not all ground-based instruments were operating, and there is a discrepancy between the two available measurements. Because of the long integration time of the filter technique and the variable conditions during the ozone maximum period, the fleer data may not be appropriate for comparison here. The NOy difference technique measured nitric acid at =120 pptv during the ozone maximum comparison period.
The denuder measurement came on line shortly at•er the ozone maximum had past MLO site, and it showed comparable mixing ratios and temporal trend compared to the NOy difference technique. The aircraft nitric acid measurement was 22 pptv, still a factor of 2.5 to 5 times less than that observed on the ground. Only one comparison of particul.ate nitrate was made. Both measurements showed low values of aerosol NOs-, with the aircraft measuring 5 pptv higher than the ground site.
The large difference in nitric acid levels observed in the two data sets can have important implications on the reservoirs and cycling of NOx in the troposphere. Earlier side-by-side tests of mist chamber with the denuder system and with the filter pack (E. Atlas, unpublished data, 1991) showed agreement within about +25% or better. This suggests that there were some instrumental problems during During both comparison periods, the airborne instrument measured a higher mixing ratio of PAN compared to the ground site. The differences were from 3 to 7 pptv, and mean values were only twice these deviations. Visual inspection of the data using a correlation of PAN with ozone suggests a possible constant bias of--5 pptv PAN between the data sets (aircraft data higher than the ground-based data), but this is near the combined uncertainties of the two instruments. This level of agreement is better than that suggested in a previous intercomparison between these same laboratories during the GTE-CITE 2 campaign ]. The earlier comparison, which encountered a larger range of PAN mixing ratios, suggested an average bias of 17 pptv for PAN mixing ratios less than 100 pptv (also with NASA > NCAR).

Total Reactive Oxidized Nitrogen (NOy)
Two instruments were measuring NOy aboard the aircraft.
During the overflight, both sets of aircraft measurements were in reasonable agreement with the ground-based measurement, with the mean value from the airborne CL instrument slightly lower than the others. The three sets of measurements (_+1 s.d.) were PEM: LIF = 132_+18; CL=109_+10: MLO: 135_+9 pptv.
However, there were significant differences between measurements for the earlier ozone maximum period. There was a maximum difference of nearly a factor of 3 between the mean mixing ratios from the different instruments, and both airborne instruments measured a lower NOy mixing ratio compared to the ground-based instrument. The mean and standard deviations were PEM: LIF = 135_+11; CL = 74_+10; MLO: 201+18 pptv.
A useful way to evaluate these data sets is with the correlation between ozone and NOy, since a close correlation of ozone with NOy is often observed in the atmosphere [Hiibler et al., 1992a Hydrogen peroxide/organic peroxides Hydrogen peroxide from the dual enzyme, aqueous coil data is consistently lower in the ground-based measurements by 42-47% compared to the aircraft data. The absolute difference in mixing ratio is near 0.2 ppbv. Organic hydroperoxides, which were found to be exclusively methyl hydroperoxide, are also lower in the ground-based measurement by 11 -37%. This is an absolute difference is 0.02 -0.08 ppbv, which is near the uncertainty of the measurement. During this same time, hydrogen peroxide was also measured at MLO by the TDLAS system. During the two short intercomparison periods, the TDLAS system was in good agreement with the ground-based dual enzyme technique. The remainder of the night the TDLAS data were lower than the coil data by an average of 0.14 ppbv. A ground-based comparison of the same instrumentation was conducted during the MLOPEX winter intensive (1992) and is discussed in detail by Staffelbach  peroxide are outside the range generally observed in the subsequent comparison.
Ozone/peroxide correlation plots for the time period on either side of the flyby indicate that the actual period of the flyby had low hydrogen peroxide mixing ratios measured at MLO (Figure 8a). Organic peroxides, though, appeared to be consistent during the flyby compared to adjacent days ( Figure   8b).
There are several possible reasons for the observed lower hydrogen peroxide during the night of October 19-20, though none are fully consistent with other chemical observations at the time. First, the lowered hydrogen peroxide relative to methyl hydroperoxide at MLO suggests preferential removal of hydrogen peroxide by aqueous processing, since hydrogen peroxide is more soluble than methyl hydroperoxide. However, other soluble species do not seem to be affected. For example, the nitric acid/NOy ratio remains uniform in the FT over the same time period that H20 2 is changing. Another possible influence on H20 2 is heterogeneous oxidation of SO 2 to sulfate. Thus we examined possible influence of SO2 from vocanic emission, but no trends or episodes of SO2 emission from the volcano could be correlated with the lower hydrogen peroxide levels. For example, there was no measurable increase in SO 2 during the ozone maximtlm period of the flyby, and hydrogen peroxide levels at the ground remained consistently low compared to the aircraft measurements. This suggests that lower levels of hydrogen peroxide also observed during the overflight period are not influenced by the small SO2 emissions from nearby vents. It appears that unspecified factors, including meteorology,

SO2
Though sulfur dioxide was measured at the ground site, the instrumentation was not designed for accurate measurement of background levels of SO2. Rather, the main use of the instrument was to identify periods at the MLO site which were influenced by volcanic emission. Episodes •:ould be larger than several parts per billion by volume of SO2, but smaller scale emissions, such as that observed during the flyby, could be detected (G. HQbler, personal communication, 1992). Background levels of SO2 measured aboard the aircraft were near 40 pptv, and indicate little influence of volcanic emission on the atmosphere near Hawaii during this flyby experiment. Whether other measurements are influenced by the low-level SO2 emissions cannot be determined by this data intercomparison.

Summary
Most of the instrumental protocols used in the study described here are designed to cover a very wide range of chemical conditions in the atmosphere. Thus, instruments are designed to encounter air masses with a large range of mixing ratios for individual species. In the very broad sense, chemical measurements from MLOPEX and from the PEM-West Flyby indicate a "clean" maritime atmosphere relative to near-source chemical signatures. Under these challenging conditions, instruments are being pushed to perform near the lower limits of detection, and uncertainties can be large. The limited intercomparison reported here illustrates the uncertainties associated with obtaining and interpreting chemical measurements from the remote atmosphere. The data comparison shown here is not from novice experimenters, but represent the efforts of experimenters who are experienced in the use of their instrumentation in challenging environments.
For the most part, the measurements showed good comparability between the data sets, especially considering the low mixing ratios of most species, but there were enough inconsistencies to limit the use of the combined data to understand the chemistry of the mid-Pacific troposphere. One straightforward aspect to consider with respect to the MLO