Electrochemical concentration cell ozonesonde accuracy and precision

JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 90, NO. D5, PAGES 7881-7887, AUGUST 20, 1985

Electrochemical Concentration Cell OzonesondeAccuracy and Precision ROBERTA. BARNESx AND ALAN R. BANDY ChemistryDepartment,Drexel University,Philadelphia,Pennsylvania ARNOLD

L. TORRES

NASA Goddard Space Flight Center, Wallops Flight Facility, Wallops Island, Virginia

The accuracyand precisionof electrochemicalconcentrationcell ozonesondesusing 1.5% KI sensing solution were measuredas a function of pressurealtitude from 800 to 6 mbar. Measurementswere made in an environmentalchamber designedto monitor the instruments'responsewhile subjectingthem to ozone, pressure,and temperatureprofiles typical of those encounteredduring balloon-borne soundings. Data from dual-instrumentballoon soundingswere used to provide an independentestimate of precision. The precisionwas found to be 6-10% (one standarddeviation) from 800 to 200 mbar, 5-6% from 200 to 10 mbar, and thereafter increasingto 16% at 6 mbar. The accuracy profile indicated a 3-5% positiveerror from 300 to 50 mbar. Larger positiveerrors were observedfrom 800 to 300 mbar (8-14%) and from 50 to about 15 mbar (10%). The error curve then rapidly rolled off toward negative values, reaching - 17% at 6 mbar. Use of new pumping efficiencymeasurementtechniquesshouldimprove both accuracyand precisionby 2-3% at the lowest pressuresbut would increasethe accuracyerror in the 50to 15-mbarrange by about the sameamount.

INTRODUCTION

Hawaii, and Boulder, Colorado, and by the AtmosphericEn-

The electrochemicalconcentrationcell (ECC) ozonesondeis a small, balloon-borne sensor developed at the National Oceanic and Atmospheric Administration (NOAA) [Kornhyr, 1969; Kornhyr and Harris, 1971] for obtaining vertical profiles of atmosphericozone.The sensoris basedon an iodine/iodide redox concentration cell composedof platinum electrodesimmersedin neutral buffered potassimiodide solutionsof differ-

vironment

Service at several locations in Canada.

The widespreaduse of this instrument has led to efforts to evaluate its accuracy,precision,and sourcesof error. Laboratory calibrationsof ECC sondesrelative to acceptedstandards such as UV absorption have been carried out [Torres and Bandy, 1978; Peterson, 1978]. Laboratory studies of specific error sources such as background current [Thornton and Niazy, 1982, 1983] and sampling pump efficiency [Torres, ent concentrations in the anode and cathode chambers. The electrical current generated when air containing ozone is 1981] have been reported. Comparisons of total ozone overpumped into the cathode is electronically converted to a burdens obtained from ECC sonde profiles with those from a signal compatible with standard meteorologicalradiosondes Dobson spectrophotometer have been made [Geraci and in sucha way that both ozonedata and standardmeteorologi- Luers, 1978]. Intercomparisons between ECC sondes and other ozone instruments flow on the same balloon have been cal data are transmittedto a ground-basedreceivingstation. carried out [Attrnannspacherand Diitsch, 1981]. The latter ECC ozonesondesare usedextensivelyto provide both trohave generally involved instrumentsnot too different in prinposphericand stratosphericozone data. Examples include ciple from the ECC sonde,however, so that none of the instrusouthern hemisphereozone profiles [Kirchhoff et al., 1983], analysis of tropospheric ozone production [Fishman et al., ments could be considereda "standard" by which to judge the 1979], and studies of trends in tropospheric ozone [Logan, 1982]. ECC sondesare also frequently used to provide correlative support for other ozone sensors.Examples include the in situ rocketsondeof Hilsenrath and Kirschnet [1980] as well as severaldifferent satellite-borneremote sensors[McCormick et al., 1984; Remsberget al., 1984; Bhartia et al., 1984]. The ROCOZ

rocketborne

ozone sensor obtains data from about

55 km down to about 20 km and dependson the ECC instrument to provide the lower part of the profile so that the compositeprofile can be integrated for total ozone overburden [Holland et al., 1984]. Weekly ECC ozone soundingsare currently made by NASA from Wallops Island, Virginia, and from Natal, Brazil, with launches timed to coincide with the overpassof satellite-borneremote sensorsfor ozone. Regular ECC ozone soundings are also made by NOAA at Hilo,

x Now at Chemal,Inc.,WallopsIsland,Virginia. Copyright 1985 by the AmericanGeophysicalUnion. Paper number 5D0386. 0148-0227/85/005D-0386505.00

others.

Recently, results of the NASA-sponsored Balloon Ozone IntercomparisonCampaign (BOIC) were reported [Hilsenrath et al., 1984]. In this experiment,ECC sondesand other small, wet-chemical sensorswere flown on the same gondola with larger "research grade" instruments that measured ozone using a UV absorption technique. In a later section of this paper, a portion of the BOIC results will be compared with those of the study describedhere. We wish to report here the results of a study designedto evaluate experimentally the accuracy and precision of ECC ozonesondesas prepared and flown by NASA, as a function of altitude. We have designedand constructedan environmental chamber in which ozonesondesare simultaneouslysubjected to the same ozone-, pressure-, and temperature-versus-time profiles that are encounteredin a typical balloon sounding from near ground level to 6-mbar pressure-altitude.By measuring the responseof a seriesof ozonesondesto known inputs of ozone, we have determinedboth averageaccuracyand precision

under

the most

realistic

conditions

obtainable

in the

laboratory. For comparison with the laboratory results, we have useda seriesof soundings,eachwith two ECC sondeson 7881

7882

BARNES ET AL..' ECC

OZONESONDE

TEMPERRTURE

CONTROL LIQ

NZ

THERMRL

CHRMBER uv

PRESSURE

03

TRRNSDUCERS

MONITOR

VENT 03

GEN

GLRSS CRPILLRRY

ROTRMETER

I

03

MONITOR PRESSURE CONTROL

Fig. 1. Environmentalchamberusedin simulatedozonesonde soundings.

the same balloon, to provide a totally independentmeasurement of precision. The environmentalchamberportion of this work has been describedin greaterdetail by Barnes[1982]. DESCRIPTION OF THE EXPERIMENT

The ECC ozonesondesused in this study were obtained commercially (Science Pump Corporation, Camden, New Jersey)and usedwithout modification,with two exceptionsin the chamber work. In the environmental chamber, the signal current developed by the electrochemicalcell was measured directly by means of the voltage developedacrossthe 1000ohm ECC sensorload resistor,bypassingthe two-transistor coupling circuit normally used for compatibility between the ECC and the radiosondetelemetry circuits.Temperature was measured not with the thermistor normally used but rather with a small thermocouple/amplifiercombination interfaced to the data acquisitioncomputer.

along the prescribedpressure-versus-time profile by controlling the actionsof a setof valvesconnectingthe chamberto a vacuum pump. At the beginning of each experiment,the chamberpressurewas reducedto 800 mbar beforestartingthe ECC sonde'sinternalsamplingpump.The pressuredifferential acrossthe glasscapillaryshownin Figure 1 was then sufficient to set up a flow of ozonatedair exceedingthat requiredby the ozonesonde.

Chamber temperaturewas regulatedby passingliquid nitrogen through cooling coils around the chamber surface. Temperatures,including the temperature of the sample gas near the ECC instrument's sampling inlet, were monitored with miniature thermocoupleprobescalibrated in various ice baths and slurriesusingprecisionthermometers.The thermocoupleswere connectedthrough electronicice points (Omega Engineering,Stanford, California) and voltage amplifiersto the data acquisitionand control system. Ozone was photolyricallygeneratedin zero-gradeair using The ozonesondes usedin the environmentalchamberpor- a low-pressuremercurylamp. All surfacescontactedby ozone tion of this work were all 3A models. The dual-instrument were either Teflon or glass.The prescribedozone profile was soundingswere carried out with a mixture of 3A and the generatedby using a motorized light shieldto block varying newer 4A modelshaving a differentsamplingpump design. amounts of radiation from the quartz reactor in the ozone No significant performance differences between the two generator. modelsare expected,if the correctionsfor reducedpumping A fast-respondingethylene-basedchemiluminescence moniefficiencyat low pressures are properlydetermined.The mean tor (built in house)was used for fast feedbackto the control correctionfactorfor 4A pumpsdiffers,for example,by about systemoperatingthe ozonegenerator.For higheraccuracythe 7% at 6 mbar from that for 3A pumpsat the samepressure ozonewas also monitoredwith a UV absorptionphotometer [Torres, 1981]. The variationamong4A pumpswasfoundto manufactured by the Dasibi Environmental Corporation be virtually identical to that observedamong 3A pumps (Glendale, California). The Dasibi instrument was calibrated (Torres [1981], and newer,unpublisheddata).If the pumping in relation to the Wallops 3-m UV ozonephotometerwhichis efficiencycorrectioncurve is measuredfor each pump (as is similar to systemsdescribedby Demoreand Patapoff [1976] the current practiceat Wallops),however,pump differences and by Torres and Bandy [1978]. The Wallops photometer between models should not cause differences in indicated has demonstratedagreementwith referencephotometersat ozone.When the experimentalportionof this studywasbeing the National Bureau of Standards to within 2% [Barnes, carriedout, it was not standardpracticeto use individually 1982]. measuredcorrections.The implicationsof this are discussed in The Validyne Corporation (Northridge, California) pressure later sections. transducerusedin thesemeasurementswas calibratedduring The sensingsolutionusedin both the chamberwork and in the critical low-pressureend of the simulated soundingsby the dual-instrumentsoundingswas 1.5% KI. The effect of manually reading pressuresfrom a 0- to 50-torr Wallace and solutionconcentrationon the cell'sresponseto ozonewill be Tiernan (Bellevile,New Jersey)gauge.The Wallace and Tierdiscussed in a later sectionof this paper. nan gaugewas calibratedat NASA's Langley ReserachCenter A block diagramof the environmentalchambersystemis againsta Ruska 0- to 100-torr quartz bourdon gaugewhich shown in Figure 1. The ozonesondeswere tested in an alu- hastraceabilityto the National Bureauof Standards. minumvacuumchamber,the pressureof whichwasregulated The data acquisition and control system was a micro-

BARNESET AL.' ECC OZONESONDE

TEMPERRTURE -25

-50

-•5

7883

(c)

El

•5

50

5

75

188

ß

--

..'

--

--

.: .: :. .:

-i

o') ice o') L.,J

-...

D_

-- OZONE

'"-.... 500

--- TEMPERFITURE

-

.i

--

48

88

OZONE

1•8

CONC

IBEI

(NB)

Fig. 2. Ozone and temperatureprofiles'usedin simulatedozonesondesoundings.

computer interfaced to the pump values, liquid nitrogen control valve, and the light shield motor in the ozone generator. Analog data from the ozonesondes,temperaturesensors,pressure transducers,and ozone monitors were recorded by this system.

The test profiles were generated by averaging altitudeversus-timedata from several randomly selectedECC sonde soundingsmade at Wallops. Using this averagedaltitude profile (which corresponded to an "ascent" rate of about 290 m/min), ozone, pressure,and temperature values as a function of time (altitude) were then, for convenience,taken from the U.S. Standard Atmosphere(1976) tables. The particular profiles used are, of course, not critical to the objective of this study (namely, how well to ECC sondesmeasure ozone) as long as they are reasonably representativeof the conditions encounteredduring actual soundings. The test profiles for ozone and temperature are shown in Figure 2, where the independentvariable has been converted from time to pressurefor a more familiar appearance.The total "ascent" time during a simulated sounding was just

The variance of the differenceat any pressurerepresentsthe combined precision of the ozonesonde and measurement system'

SDiFF 2-- SECC 2 + SMS 2 The ozonesondeprecisioncan thus be estimated if the precision of the measurementsystemis known. A comprehensive error analysisfor this systemwas carried out EBarnes,1982] which indicated a precision (one standard deviation) equivalent to 1- to 1.5-nbar ozone throughout most of the profile. For simplicity here, an averagevalue of 1.3 nbar was used for SMSthroughout the profile. Dual-ozonesonde soundings were carried out by placing two independentECC sensor/radiosondecombinationson the same balloon. Differencesbetween indicated ozone at a given pressurealtitude were compiledfor 13 dual-instrumentsoundings and usedto estimatethe precisionat that altitude using

SEC C-- [(dl2 + d22 +... + d,32)/(2'13)]'/2

[Youden, 1951]. Differences between pressure sensorsin the two radiosondeson a given balloon can cause a "mismatch" Accuracy and precision were estimated at any given pres- in the two ozone profiles which could not easily be dissure altitude from the mean difference and associated standard tinguishedfrom ECC errors. To avoid this, ozone readings deviation of ozone indicated by the ECC sensorsand that were taken as a function of time during the soundingand then indicated by the Dasibi UV instrument (corrected to the tabulated versuspressureusing only one of the pressuresenchamber pressure).Data from 11 ozonesondeswere used in sors.The resultingprecisionestimateis then representativeof this portion of the experiment,giving 11 comparisonsat any ozonesondesonly and does not contain contributions from pressurealtitude. The UV data were assumedto have no bias. the radiosonde pressure sensor. Although not used during Although no direct measurementof ozonelossesin the glass thesetests,radiosondesequipped with hypsometerscan make capillary and Teflon tubing was possible,indirect evidence pressurealtitude errorsnegligible[Parsonset al., 1985]. suggeststhat these losseswere small. The accuracyestimates reported here and those obtained from the BOIC balloon RESULTS AND DISCUSSION flights (both discussedlater in this paper) did not show the Precision estimates from both the environmental chamber substantial differencesexpected to result from such losses. work and from the dual-instrument soundingsare shown in Also, Barnes [-1982] extrapolated results of Ainsworth et al. •1981] to the present experimental circumstancesand esti- Figure 3. The left portion of this figure showsthe absolute precision(one standard deviation) expressedin ozone partial mated lossesof