An Intercomparison of Carbon Monoxide, Nitric Oxide ... - Douglas Davis

1 downloads 0 Views 588KB Size Report
Dec 20, 1984 - 8 Oregon Graduate Center, Beaverton. 9 Washington State University, Pullman. •o Ford Motor Co., Dearborn, Michigan. This paper is not ...
JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 89, NO. D7, PAGES 11,819-11,825,DECEMBER 20, 1984

An Intercomparison of Carbon Monoxide, Nitric Oxide, and Hydroxyl Measurement Techniques'Overview of Results J. M. HOELL, 1 G. L. GREGORY, 1 M. A. CARROLL, 2 M. MCFARLAND, 3 B. A. RIDLEY, n'D. D. DAVIS, SJ.BRADSHAW, M. O. RODGERS, 5 A. L. TORRES, 6 G. W. SACHSE, • G. F. HILL,• E. P. CONDON, 7 R. A. RASMUSSEN, 8 M. C. CAMPBELL, 9 J. C. FARMER, 9 J. C. SHEPPARD, 9 C. C. WANG,xøANDL. I. DAVISxø Resultsfrom an intercomparisonof methods to measurecarbon monoxide (CO), nitric oxide (NO), and the hydroxyl radical (OH) are discussed.The intercomparisonwas conductedat Wallops Island, Virginia, in July 1983 and included a laser differential absorption and three grab sample/gaschromatograph methodsfor CO, a laser-inducedfluorescence(LIF) and two chemiluminescence methodsfor NO, and two LIF methodsand a radiocarbontracer method for OH. The intercomparisonwas conducted as a field measurementprogram involving ambient measurementsof CO (150-300 ppbv) and NO (10-180 pptv) from a commonmanifold with controlledinjectionof CO in incrementalstepsfrom 20 to 500 ppbv and NO in stepsfrom 10 to 220 pptv. Only ambient measurementsof OH were made. The agreement between the techniqueswas on the order of 14% for CO and 17% for NO. Hardware difficultiesduring the OH testsresultedin a data basewith insufficientdata and uncertaintiestoo large to permit a meaningfulintercomparison. INTRODUCTION

For the past decadethere has been considerableinterest in the relative influenceof natural vs. anthropogenicactivitieson the chemicalcompositionof the atmosphere.This interesthas manifesteditself in three major activities:(1) laboratory studies of basic chemicalreactionsand proof of principlesfor instrument concepts,(2) theoreticalmodeling to predict concentrations

and trends associated

with

the natural

and dis-

turbed atmosphere,and (3) field measurementsof key atmosphericspecies.A measureof the successto which theseactivities enhance our understandingof the chemical and physical processesoccurring in the atmosphereis the degreeto which observational data and theoretical predictions agree. Recent experiencewith suchcomparisonshave met with only limited success, due in part to an uncertaintyin our ability to measure some of the key troposphericspecies.Questions concerning the relative accuracyof model predictionsas well as the quality of field measurementshave beenraiseddue to what appears

measurea given speciesis the intercomparisonof atmospheric measurementsobtained at a common site by instrumentsemploying different techniques.As part of the National Aeronautics and Space Administration's(NASA) Global TroposphericExperiment (GTE) [McNeal et al., 1983], three such field missions,one ground based and two airborne, were designedto characterizethe current capability to measure the concentrationsof three key troposphericspecies:carbon monoxide (CO), nitric oxide (NO), and hydroxyl (OH). The purpose of this note is to provide a brief overview of the first mission,conductedduring July 1983 at NASA's experimental rocket launch facility on Wallops Island, Virginia. Overviews for the airborne testsalong with a more detailed review of the resultsfrom each intercomparisonmissionwill be the subject of future publications. INTERCOMPARISONTECHNIQUES

An integral part of the overall GTE intercomparisonstrategy is the use of fundamentally different measurement conagreement between predictions and measurements.In short, cepts for CO, NO, and OH along with specific ancillary one of the limiting factorsin identifyingthe impact of anthro- measurementsto facilitate interpretation of the intercomparison data and to aid in assessing the performanceof specific pogenicactivities on key atmosphericspeciesis our inability techniques. Table 1 lists the principal investigators(PI) and to distinguishbetweentrue atmosphericvariability and instruorganizations responsiblefor the CO, NO, and OH measurement errors. Within recent years there has been a growing realization ments. The ancillary measurements included continuous meteorologicaldata, UV flux, methane, nonmethane hydro[Parsons et al., 1982; Spicer et al., 1982; Chanin, 1983; carbons,ozone, aerosol size distribution and composition, amMcNeal et al., 1983] that one of the more useful methods for monia, and nitric acid. The instrument techniquesthat were characterizingthe performanceof instrumentsdesignedto intercomparedincluded a laser differential absorption technique [Sachseet al., 1983] and three grab samplemethodsfor • NASALangleyResearch Center,Hampton,Virginia. CO ICondon, 1977; Rasmussenand Khalil, 1982]; two chemi2 NOAA, Environmental Research Laboratories and Cooperative luminescencesystems[Ridley and Howlett, 1974; McFarland Institute for Research and Environmental Sciences,Boulder, Colet al., 1979; Torres, 1982] and one laser-induced fluorescence orado. (LIF) [Bradshaw and Davis, 1982] system for NO; and two 3 E. I. DupontDeNemours & Co.,Wilmington, Delaware. different LIF approaches[Wang et al., 1974, 1981; Davis et al., '• NationalCenterfor Atmospheric Research, Boulder,Colorado. 5 GeorgiaInstituteof Technology, Atlanta. 1976] and a radiocarbon tracer technique [Campbell, 1979] • NASAGoddardSpaceFlightCenter,WallopsIsland,Virginia. for OH. 7 NASA AmesResearchCenter,Moffett Field,California. DESCRIPTION OF FACILITIES 8 OregonGraduateCenter,Beaverton. 9 Washington StateUniversity,Pullman. The intercomparisonsite was located at the north end of •o Ford Motor Co.,Dearborn,Michigan. to be excessive scatter in field measurements

and the lack of

NASA's rocket launch facility on Wallops Island, Virginia,

This paper is not subjectto U.S. copyright.Publishedin 1984 by the AmericanGeophysicalUnion.

approximately 1-km inland from the Atlantic coastline. The facilitiesconsistedof trailers set up in an open field to accom-

Paper number 4Dl107.

modate previouslydefinedinstrument,procedural,and oper11,819

11,820

HOELLET AL..'BRIEFREPORT

TABLE 1. Instrumentationfor CO, NO, and OH Measurements at the WallopsInstrumentationIntercomparison Tests Principal Investigator

Organization

Code

MeasurementTechnique

GRAB 1A GRAB lB DACOM GRAB 2

grab sample/gaschromatography direct injection/gaschromatography laser differential absorption cryogenicgrab sample/gaschromatography

CO Techniques E. Condon

Ames Research Center

Glen Sachse R. Rasmussen

Langley ResearchCenter Oregon Graduate Center

A. Torres M. McFarland M. Carroll

Goddard SpaceFlight Center

B. Ridley

National Center forAtmospheric CHEM 2

NO Techniques CHEM 1

chemiluminescence

D. Davis

National Oceanic and Atmospheric /

M. Campbell C. Wang

Washington State University Wayne State University/Ford

D. Davis

Georgia Institute of Technology

Administration

chemiluminescence

Research

Georgia Institute of Technology

Motor

LIP

OH Techniques RCO L/LIF

two-photon, laser-inducedfluorescence

radiocarbon tracer lidar laser-induced fluorescence

Co.

ational requirements.All testsfor CO and NO were conducted using either one of two samplingmanifold systems.The manifoldusedfor the CO and NO ambientand spikedambient intercomparison testswasconstructedof 75-mm ID pyrex pipes and was 35 m long from inlet to exhaust.Ambient air was drawn through this manifold at approximately35 standard liters/s and controlled via a mass flow controller. The manifoldwasdesignedwith the inlet approximately6 m above the surfaceand configuredso that known levelsof CO or NO

I/LIP

in situ laser-induced

fluorescence

workshop data analysisactivities.The major thrust of the preworkshopactivitieswas the intercomparisonof the calibration standardsusedby the CO and NO groupsin order to identify potential sources of systematic errors between measurement results. The main thrust of the on-site activities

was the intercomparison of simultaneous ambient air

measurements by the CO, NO, and OH instruments,respectively.The post-workshop activitiesconsistedof data analysis sessions and wereconductedfor eachspecies separately. could be added to the ambient flow. These additions were The preworkshopactivitieswereinitiatedin May 1983with obtained from a gas dilution systemlocated 17-m and 25-m round robin measurements on common "primary" CO and upstreamof the NO and CO samplingports, respectively.All NO standardssuppliedby the Global TroposphericExperiNO samplingports werelocatedalonga single20-cmsection ment (GTE) projectoffice.This wasfollowedby measurements of the prime manifold.A similar arrangementwasinstalledfor on a separateblind CO and NO standardsentto eachgroup. the CO PI's. The resident time for ambient air from inlet to

In each case the standards were National

the CO and NO sampleports was approximately4 and 3 s, respectively.Preworkshop studiesindicated the gas dilution systemto be accurateto better than 5% at the injectionpoint on the manifold.The gas dilution systemprovidedup to two stagesof dilution. The final stageof dilution occurredwithin

(NBS) standard referencematerials (SRM) or traceableto SRM's. The primary standardwas givento eachgroup,with instructionsto measure•by the best means available--its mixing ratio relative to their respectivecalibration standard. The NBS value for the primary standardwas made known to each PI. Only nominal mixing ratios were providedwith the blind standardssentto eachinstrumentgroup.Measurements of the blind standardswere performedusingthe instrument, procedures, and calibrationstandardto be usedat Wallops. The on-siteactivitieswereformallyinitiatedon July 5, 1983, and werecompletedon July 30, 1983.Testsfor eachspecies, listedin Table 2(a, b), were scheduledindependentlyand, for CO and NO, consistedof a seriesof daytime and nighttime measurements on (1) ambient air drawn through a common manifold,(2) ambientair spikedwith premixedlevelsof CO or NO, and (3) calibrationlevelsof CO in dry nitrogen(CO tests only) provided at each instrument via a common manifold. For a giventest sequence,only the length of the test,the times

the manifold via the mass flow within the manifold. Prework-

shop studies,conductedat CO mixing ratios from 3 to 100 ppmv and NO mixing ratios from 9 to 200 ppbv, showed manifold lossesto be less than 3%. A secondmanifold, lo-

catedadjacentto and designedsimilarlyto the onejust described,was contructedof 25-mm ID pyrex pipe. This manifold was usedto provideknown mixing ratios of CO in ultraclean nitrogen gas at each instrument location. The exhaust fr6m the two manifolds and the exhausts from all instruments

were located at ground level, well away from the inlet to the test manifold.

The inlet of the OH point samplersfrom the GeorgiaInstitute of Technology(GIT) and WashingtonState University (WSU) weremountedon platformsat approximatelythe same elevationand within 5 m of eachother.The OH lidar system from Wayne/Ford Motor Co. (FMC) was housedin a motor van approximately15 m from the sampleplatformsusedby GIT and WSU. All of the OH instruments were located at the site so that ambient measurements could be conducted within

a radius of approximately20 m and at a height of about 4 m above the ground. INTERCOMPARISON PROCEDURES

The Wallops intercomparisonworkshopwas conductedin three phases,consistingof preworkshop,on-site,and post-

Bureau of Standards

at whichthe CO or NO concentration wouldbe changed, and the approximatemaximumconcentration expectedduringthe test sequencewere provided to the PI's. For OH, only ambi-

ent air measurements werefeasiblebecauseof the diversityof samplinggeometriesand the difficultyof generatingan OH standard.The OH testswere scheduledas daytime measurementsfrom !000 to 1600hoursand nighttimetestsbeginning after 1700 hours. The night measurementsof OH and NO werefelt to be particularlyimportantfor evaluatingpotential artifactsand interferences and determiningminimum detectable levelsfor the varioustechniques,sincethe concentration of both OH and NO shouldbe nearzero at night.

HOELL ETAL.:BRIEF REPORT

11,821

TABLE 2a. IntercomparisonSummaryfor CO and NO Tests

Date

Time, EDT

Range of Ambient CO, ppbv

Type*

Range of Spiked Steps, ppbv

Carbon Monoxide

July 8 July 11 July 11 July 12 July 14 July 14 July 18 July 27

1000-1200 1000-1220 1400-1500 2000-2310 0900-1000 1300-1535

Ap ANp• Ap AAp Ap AN$

1530-1930

AN,

1700-2015

AAô

40-140 18-70

180-230

150-200 160-180 180- 300

23-93 18-69

17-380 23-516

260-320

Range of

Range of Spiked Steps,

Ambient NO,

pptv

pptv

July 26 July 27 July 27 July 28 July 28 July 29 July 29 July 29

2100-2300 1900-2000 2230--0050 1200-1605 2(X)0-0(O5 1200-1400 1400-1730 2030--0010

Aõ Aõ AAõ AA AA A AA AA

10-15 120-180 10-15 70-100 5-15 20-60 30-60 10-15

43-145 35-211 11-42 42-132 21-116

Test designation:AA, spikedambient; AN, spikednitrogen;A, ambient. pARC GRAB 1A data contaminated. $OGC techniqueunreliablefor CO in nitrogen. ôARC GRAB 1A and GRAB lB data unreliable. õGIT NO systemnot operational.

An important aspectof any intercomparisonis the choiceof

comparison referenceis susceptibleto biasing by erratic or

a comparisonreference.Sincenone of the techniquesusedin extremevaluesfrom one technique.A groundrule established the CO, NO, or OH testscould be regardedas a standard, for determining the measurementaverageas the comparison two approacheswere adopted for intercomparisonof the test standard was that a valid measurement be available from each results.The distinguishingfeature between the two was the group. As shown by the footnotes of Table 2(a, b) the opermethod involved in creating a "comparisonreference."In the ationalstatusof the instrumentslimited the applicationof this first approach the comparison referencewas defined as the averageof the observationsreported by the PI's during each sampleinterval.The absolutevalue of the respectiveCO, NO, or OH mixing ratio measuredby each technique was compared to this reference.This approachprovidesa comparison of the relative agreementbetweenthe techniquesin which the effect of the ambient variability of the target speciescan be minimized if the measurementsare temporally coordinated.It should be recognized,however, that for this approach the

method of comparisonto about half of the CO and NO tests that were conducted.

None

of the OH

results satisfied the

groundrule for establishingthe averageas a comparisonreference.In the secondapproachfor intercomparison,which can only be usedwith the CO and NO tests,the magnitudeof the GTE-generated step changesin the CO or NO mixing ratio was definedas the comparisonreferenceand comparedto the changesin mixing ratio measuredby eachtechnique.As much as possiblethe changesin mixing ratios as measuredby each

Table 2b. IntercomparisonSummaryfor OH Tests Ambient

Conditions Particle

Dew Date

July 15 July 16 July 20 July 21 July 21 July 22 July 22

Time, EDT

1000-1645II,**,PP 1200-1530II,**,PP 1100-1600II,PP,** 1400-1530**,pp,:[: :[: 1530-2115**,pp,ôô 1100-1700•ô 1700-1935ôô

T, øC

Point, øC

Hydroxyl Radical 3!-34 15-21 34-36 21-23 28-34 22-27 33-34 22-23 27-33 23-26 26-30 10-13 26-28 13-15

IIWSUsystem not on site. **Wayne/FMC systemoperationalbut SNR < 1. •'pGIT OH systemnot operational. $:I:WSUsystemnot operational. ôôWSU data contaminatedby impurities.

03, ppbv

52-104 75-97 48-76 64-75 61-93 33-65 62-71

Count,

number/cm 3

1.2-2.6 12-35 36-85 8-14 15-51 0.7-1.3 0.7-1.6

11,822 488

HOELL ET AL.' BRIEFREPORT -

Standards

DFICOM

A

GRSB1B





The proceduresfor intercomparingstandardswere initiated about 2 months prior to the start of the on-siteactivitiesand completedwith additional measurementson site in July. All GTE-supplied CO and NO standardswere mixed in N2 with mixing ratios in the 1-10 ppmv range. Resultsfrom the CO

A

instruments

288

i

i

i

1888

1188

LOCSL

1288

TIME

5% of the NBS

values certified for

the common(primary)and blind GTE standards.Two CO in zero air standardssupplied by ARC were also exchanged while at Wallops. The CO mixing ratio for these standards were nominally 285 ppbv and 1.3 ppmv. The LRC and OGC groups reported mixing ratios for these two standardsthat were about 4% higher and 3% lower, respectively,than the valuesacceptedby the ARC group. Resultsfrom the NO instruments agreed to within 2% of the NBS value of the common (primary) GTE NO standard. Good agreementwas also observedbetweenthe NO mixing ratio reported by the PI's and the NBS value for the blind standard. The mixing ratios reported by the LIF, CHEM 1, and CHEM 2 groups

188

8

were within

(HOURS)

Fig. 1. Intercomparison of carbon monoxide measurementsobtained during the Wallops Island workshop spiked ambient test on 'July 8, 1983.

for their blind standards were about 6% low, 1% low, and 3%

technique were determined for common time intervals. This approach provides a referencepoint that is traceableto an NBS standard subject to the inaccuraciesof the dilution system,includingsamplemanifoldlosses,and the background variability during the spiked ambient runs. Use of the GTE steps as a comparisonreferenceresults in utilizing a larger percentageof the CO and NO test results, since the only requirementwas, in fact, a valid measurementfrom a given technique. The data-handlingprotocol establishedfor the CO, NO, and OH PI's included (1) reporting of all results from the formal test periods,(2) blind tests(3) submittal of preliminary data within 1 to 2 days after a test sequence,and (4) submittal of final resultswithin 30 days after the completionof on-site activities.When availablethe ancillary data were providedon site without restrictions.GTE personnelperformedpreliminary data comparisonon site, and in caseswhere the results appearedinconsistent,based on experienceor earlier workshop results,the PI's were so notified and asked to examine their data or instrumentprocedures.Even in thesecases,communication was limited, and actual test results were maintained blind. INTERCOMPARISON RESULTS

The two major intercomparisonactivitiesconsistedof intercomparisonof CO and NO standardsand intercomparisonof field measurementsobtained at the Wallops site. In this section the results from intercomparingthe standardswill be presented,followedby a descriptionof the salientresultsfrom the CO, NO, and OH field measurements.

high, respectively,relative to the NBS value. In summary, resultsfrom this phaseof the intercomparisonworkshopsuggest that differences between the various CO

and NO

standards

are statisticallyinsignificantat the 2-sigmalevel. Accordingly, resultsfrom this phaseof the intercomparisonworkshophave not been usedto adjust any of the on-sitemeasurementsdiscussed below. CO Results

Figure 1 illustrates CO results obtained from a daytime spiked-ambienttest sequenceconductedon July 8, 1983. The ambient CO mixing ratios were in the 200-ppbv range, and the CO stepsfrom the GTE mixing systemrangedfrom 40 to 140 ppbv. The times at which the CO steps occurred are indicated along the bottom of the figure. Note that the DACOM system provided continuous CO measurements, while the GRAB lB and GRAB 2 data are from integrated sampleswith sample times ranging from a few secondsto about 6 min. The horizontal bar on the GRAB 2 samples indicatethe samplecollectionperiod. Bars on the GRAB lB data have been omitted, sincethe sampletime is small comparedto the symbolsize.The data shownin Figure 1 illustrate the maximum ambient variability observedduring any of the CO tests,includinga stepchangeof about 23 ppbv observed by the DACOM technique(noted on Figure 1). In this one casethe DACOM data were usedto adjust the magnitudeof the comparison referencestandard (e.g., GTE step). In all other tests the ambient variability was small relative to the magnitudeof any spikedaddition. To provide an overall summaryof the CO testsresults,the

TABLE 3. StatisticsDescribingResultsFrom CO IntercomparisonTests

Category1' Parameter GRAB

1A GRAB

lB DACOM

19 -8.7

4.9

Category2J' GRAB 2 GRAB

1A GRAB

lB DACOM

Category35 GRAB 2 GRAB

19 5.1

19 3.6

0

10 4.0

22 3.1

21 8.4

4.4

7.6

--

30.3

12.3

17.0

1A GRAB

7 -0.43

26.6

N, numberof data points;D, percentage error with respectto reference; ate,standarddeviationin percent. *Spikedambienttestsusingaverageof all measurements asreference. J'Spikedambient testsusingGTE stepsas reference. $SpikednitrogentestsusingGTE stepsas reference.

lB DACOM

8 3.6

10 - 7.2

24.0

2.5

GRAB 2

HOELLETAL.' BRIEFREPORT 150 o

LIF

O

CHEMI

o

CHEM2

the mixing ratio based upon the mixing system parameters. For each technique,D is a measureof its bias relative to the reference,and within a category,D is a measureof the relative agreementbetweenthe techniques.The aD'Sare indicative of the total variability (e.g.,instrumentaland speciesvariability) associatedwith the measurementresultsfor a given technique. The resultsgiven in Table 3 provide usefulinsight relative to the performanceof the CO techniquesindividually as well as a group. The resultsfrom category 1 provide a measureof the relative agreementthat can be expectedbetween the techniques for measurementsof CO under ambient conditions. The relative agreementas indicated by the range of percentage errors is about 14% (e.g., D values ranging from -8.7% to 5.1%). The uncertainty associated with each D value (i.e.,

•'

x

S0

0

i

i

i

i

i

[

i

i

i

i

1

2230

2130

11,823

233e

E)030

O'D/•) isrelatively small andsuggests a truebiasin the

GRAB lB resultswith respectto DACOM and GRAB 2 techniques.The resultsfrom category2 provide a measureof the ability of each technique to measure changesin CO mixing ratios. Sinceonly changesin mixing ratios are compared(i.e., PI-measured vs. GTE mixing system),the results should be relatively free from constant bias errors presentin either the individual measurementsor mixing system. The percentage data have been grouped into three categoriesthat couple the errors for this category indicate that the relative agreement two approachesfor obtaining a referencestandard (described between the techniquesfor measuring changesis about 5% above) and the three test conditions (Table 2a). The three (e.g.,D valuesranging from about 3% to 8%). The relatively categoriesare (1) ambient (i.e.,spikedand unspiked),usingthe large standarddeviation associatedwith each percentageerror average of the PI's reported values as the comparison refer- is indicative of both instrumental and ambient CO variability, ence;(2) spikedambient, usingthe GTE stepsas the compari- with the ambient variability probably having the larger effect son reference;and (3) spikednitrogen,usingGTE stepsas the for this case.Recallthat the comparisonreferencefor category comparisonreference.Statisticsdescribingthe CO resultsfor 2 is a constant mixing ratio for each CO step as determined each category are given in Table 3. The number of data from the mixing system,while the actual CO in the manifold points,N, usedfor each techniquein each categorywas dicta- is from both the injectedand ambient CO. The resultsgiven in ted by the ground rule for determiningthe comparisonrefer- the third categoryprovide a measureof accuracyof the techence.The statisticaldata given for each categoryincludesthe niquesfor absolutemeasurements of CO mixing ratios subject average percentageerror D with respect to the comparison to the uncertaintiesassociatedwith the backgroundCO in the referenceand the standard deviation aD associatedwith D. nitrogen diluent as well as the accuracyof the mixing system The D values for each technique were calculated,in percent, (e.g., •5%). The resultsfor this categoryshould also be more from indicativeof the precisionassociatedwith a given techniquein that the variability of CO is minimum. A measureof the back1 ground CO in the diluent gas can be obtained from the average "zero" obtained during the spiked nitrogen tests. where Y• is the measurementresult and Xi is the correspond- These valueswere 4.6 + 5.5 ppbv, 7.7 + 8 ppbv, and 1.2 + .7 ing value of the comparisonreference.In category1, Y•is the ppbv for the GRAB 1A, GRAB lB, and DACOM techniques, mixing ratio reported for either ambient or spiked ambient respectively.Subject to the limitations noted, the absoluteacruns, and Xi is the average of the mixing ratios reported by curacywith respectto the GTE mixing systemis indicated by eachtechnique;in category2, Y•is the changein mixingratio the averagepercentageerror and standarddeviationgiven for detectedby a techniquefor the spikedambientruns, and X• is each technique.The apparent high accuracysuggestedby the the change in mixing ratio as determined from the mixing low D values for the GRAB 1A and GRAB lB techniques systemparameters;in category3, Y•is the mixing ratio of CO must be acceptedwith caution in light of the correspondingly detectedby a techniquefor the spikednitrogenruns, and Xi is large aDS. LOCfiL

TIME

(HOURS)

Fig. 2. Intercomparisonof nitric oxide measurementsobtained during the Wallops Island workshopspikedambienttestson July 29, 1983. For clarity the temporal centroid of the averagemixing ratio (i.e., symbolswith error bars) has been slightly displacedfor each technique.

Z (r,- x,)/x,

TABLE 4. StatisticsDescribingResultsFrom NO IntercomparisonTests Category 1' Parameters N D

o'D

CHEM 21 6.9

8.3

1

CHEM 21 3.6

12.9

2

Category2'• LIF 21 -9.6

9.8

CHEM

1

26 - 10.6

12.9

CHEM 26 -0.7

14.2

2

LIF 17 -23.9

14.4

N, numberof data points;D, percentageerror with respectto reference'0'D,standarddeviationin percent.

*Spiked ambienttestsusingaverageof all measurements as reference. '•SpikedambienttestsusingGTE stepsas reference.

11,824

HOELL ET AL.' BRIEFREPORT

TABLE 5.

OH Intercomparison Results

107OH/cm3 Date

July 21

July 22

Time, EDT

WSU*

1932

_