Spectral solar UV irradiance data for cycle 21

JOURNAL OF GEOPHYSICAL

RESEARCH, VOL. 106, NO. A10, PAGES 21,569-21,583, OCTOBER 1, 2001

SpectralsolarUV irradiance data for cycle21 Matthew T. DeLand 1 and Richard P. Cebula1 RaytheonITSS, Lanham,Maryland

Abstract. TheNimbus7 SolarBackscatter Ultraviolet(SBUV) instrument, whichbegantaking datain November 1978, was the first instrumentto make solarUV irradiancemeasure-

mentscoveringboththeminimumandmaximumactivitylevelsof a solarcycle. The currentlyarchivedirradiancedatasetwasprocessed with an instrument characterization which failsto completely account for sensor degradation in thelaterpartof thedatarecord,thus limitingtheaccuracy of estimated long-termsolaractivityvariationsandthescientificvalue of thedata. In thispaper,we describe animproved Nimbus7 SBUV spectral irradiance data set,whichutilizesa moreaccurate modelfor instrument sensitivity andtreatsothertimedependent problems in thearchiveddata.Estimated long-termirradiance changes duringsolar cycle21 are 8.3(+2.6)%at 205 nm, and4.9(+1.8)% at 240 nm. The revisedNimbus7 SBUV irradiance dataarein goodagreement withpredictions of solarcyclevariations fromtheMg II indexproxymodel.Thesesolarirradiance changes arealsoconsistent withoverlapping irradiance datafromthedeclining phaseof solarcycle21 measured by theSolarMesosphere Explorer(SME). TheNimbus7 SBUV irradiance datarepresent theearliestcomponent of a 20+ yearcontinuous recordof solarspectralUV activity. 1. Introduction

SpectralIrradianceMonitor(SUSIM) [Floydet al., 1998]and

Knowledgeof the spectraldistributionof solarirradianceis important for atmosphericphotochemistry. At ultraviolet (UV) wavelengths,irradiancevariationsbecome significant on both rotational(-27-day) and solarcycle (-11-year) timescales. The moststrikingand consistentevidencefor correlations betweensolar UV activity and atmosphericresponseis foundin stratospheric ozonedata. Recentwork on this topic includesthe followingpapers: Chandraand McPeters[ 1994] for short and long timescalesat 1-2 mbar; Hood and Zhou [1998] for 27-day variationsin stratospheric ozone and temperature; Bjarnasonand ROgnvaldsson [ 1997] for equatorial total ozone using Total Ozone Mapping Spectrometer (TOMS) data; Jackmanet al. [1996] for 2-D model results; and Chandraet al. [ 1999] for tropospheric ozone. Long-term data setscapableof characterizingsolar UV irradiancevariations on both rotationaland solarcycle timescaleshave been collectedfor more than 20 years,beginningwith Nimbus 7 Solar BackscatterUltraviolet (SBUV) data covering the wavelengthregion 160-400 nm from November 1978 to October 1986 [Heath and Schlesinger,1986; Schlesingerand Heath, 1988]. Additional multiyear irradiancedata setsinclude the Solar MesosphereExplorer (SME), covering 115302 nm from January1982 to June1988 [Rottman,1988]; the NOAA 11 Solar BackscatterUltraviolet, model 2 (SBUV/2) [Cebula et al., 1998; DeLand and Cebula, 1998a], coveting 160-400 nm from December 1988 to October 1994; the Upper AtmosphereResearchSatellite(UARS) SolarUltraviolet

•NowatScience Systems andApplications, Inc.,Lanham, Maryland.

Copyright2001 by theAmericanGeophysical Union. Papernumber2000JA000436. 0148-0227/01/2000JA000436509.00

the UARS Solar Stellar Irradiance ComparisonExperiment (SOLSTICE) [Woods and Rottman, 1998], both covering -•115-410 nm from September1991 to the present; and the Global OzoneMonitoringExperiment(GOME) [Weberet al., 1998], covering240-790 nm from June 1996 to the present. DeLand and Cebula [1998a] providesfurther details about each of these instruments.

A criticalelementin the determinationof solarcyclelength irradiancechangesfrom satelliteinstrumentsis the accuracy of the long-terminstrumentcharacterization.For Nimbus 7 SBUV-the instrumentsensitivitychangedfrom November 1978 throughDecember1986 by-25% at 340 nm and 50% at 255 nm. The time-dependentinstrumentcharacterization derivedfrom the first 3 yearsof datawas foundto give substantialerrorsat shorterwavelengths(3•< 250 nm) when extrapolatedto the end of the continuousscansolardatarecord in late 1986 [Schlesingereta!., 1988]. Furtherdetailsof the long-terminstrumentcharacterizationare discussedin section 2.2. Schlesingerand Cebula[1992] (hereafterSC92) created

a revisedinstrument characterization that greatlyimproved the Nimbus 7 SBUV data quality, reducingtime-dependent uncertainties due to the instrument characterization

to 2-3% at

mostwavelengths.However, SC92 did not producean archival irradiancedatasetusingthe revisedcharacterization.Our goal in this work is to improve the long-termcalibrationof SBUV to producea data set suitablefor studiesof long-term solarvariations,and createa publiclyavailableirradiancedata setthat can be comparedwith othersolarirradiancedatasets. Section2 describesthe archiveddata and the baselinelongterm instrument characterization,the revisions derived by SC92, and the stepswe have takento refine and improvethe SC92 work. Section3 showsour results,includingcomparisons with predictionsfrom the Mg II index proxy model. Section4 presentscomparisonswith overlappingdata from SME during cycle 21. Finally, section5 discussesour conclusions.

21,569

21,570

DELAND AND CEBULA: SPECTRALSOLARUV IRRADIANCE DATA

2. Nimbus-7 SBUV Solar Irradiance Data and Instrument Characterization Corrections

tivity. Changesin the measured solaroutputfrom an initial time to were definedas the productof diffuserreflectivity degradation, instrument throughput (e.g.,optics,electronics) 2.1. InstrumentDescription changes, and true solarirradiancechanges.Cebulaet al. that diffuserreflectivitychanges couldbe TheNimbus 7 SBUVinstrument wasdescribed in detailby [1988] assumed as the productof separable exposureHeathet al. [ 1975]. It is a nadir-viewing Ebert-Fastie double furthercharacterized andtime-dependent terms. Withthisassumption monochromator witha 1.1nmbandpass, designed to measure dependent they defined the following equation for the measuredsolar totalcolumnozoneandstratospheric ozoneprofilesusing output:

observationsof backscattered Earth radiancebetween255 and

340 nm. In normaldiscretemodeoperations, SBUV made measurements at 12 discretewavelengths duringeach32 s scan.Thedetermination of ozonevaluesusingthebackscatteredultraviolet (buv)methodalsorequires measurements of

F(X,t)= F(X,O) P(t)er(x)E(t) es(x)t e-¾(x)G(t) (1) F(3.,t) measured solarirradiance output,mW/m2/nm; F(3.,0) measured irradiance at3.onfirstdayofoperation; P(t) photomultiplier tube(PMT)gainchange; r(3.) exposure-dependent degradation coefficient, hr-•;

solarirradiance to calculate albedovalues[e.g.,Bhartiaetal., 1996]. The solardataareobtained asthe spacecraft crosses thenorthern terminator intodarkness by deploying a diffuser

E(t)

solardiffuserexposure, hours;

plate to direct solar radiation into the instrument. These

s(3.)

time-dependent degradation coefficient, days-J;

measurements are thus full-disk solar observations. The ¾(3.) solaractivityscalefactor; measurements discussed in thispaperweretakenin the conG(t) solarrotational modulation activityindex. tinuousscanmode,whereSBUV scansin wavelength from Thephotomultiplier tubegainchangeP(t) wascharacterized 160to 400 nm in -0.2 nm steps.In normaloperations, three usingthe outputfroma reference photodiode.Solaractivity consecutive scansweretakenon a singleorbiteachdayand was represented only for the purposeof derivinginstrument

averaged to createa dailyirradiance spectrum.Duringse- sensitivity change by theterme-•(;•)G(t), whichwasbased on the Mg II index model developed by Heath and Schlesinger deployment' periods, theobservation frequency wasincreased

lectedintervalsof 2-5 monthsduration,called'accelerated

[1986]. Thistermis not appliedin the creationof corrected to everyorbit(13-14orbitsper day). Onlydailyaverage ir- solarirradiance values. The diffuserexposure rateE(t) was radiance value s arearchived. A mercury lampwasviewedon significantly increased for threeaccelerated deployment peria periodicschedule to trackchanges in thewavelength cali- odsduringthe six yearsof Nimbus7 SBUV operationanabration.No directmonitoring of instrument throughput or lyzedby Cebulaet al. [1988]. Multiplelinearregression fits diffuserreflectivitychanges wasavailable. usingthe logarithmicform of equation(1) thenallowedvalTheNimbus7 satellitewaslaunched in October1978,and uesfor r(3.),s(3.),and7(3.)to be determined.Thisanalysis theirradiance datarecordbeginsonNovember7, 1978. From wasperformedat morethan30 wavelengths throughout the 1978to mid-1983theSBUVinstrument wasusuallyoperated 160-400nm wavelengthregionfor eachperiodof accelerated on a 3 dayson, 1 day off cycleto conserve power. From diffuserdeployment. Regression resultsfor thesolaractivity summer1983untiltheendof thedatarecord,SBUVoperated coefficient7(3.)were foundto be negligibleexceptat short everyday. Thebaselinedatasetfor thispaper,whichis arwavelengths,consistentwith the predictionsof Heath and chivedat theNationalSpaceScience DataCenter(NSSDC), Schlesinger [1986]. Cebulaet al. [1988] foundthatfor data contains8 yearsof dailyaveragecontinuous scansolarirradi-

through1984,changes in r(3.)betweenregression fit intervals

ancedataandrunsthrough October 28, 1986[Schlesinger et were not statisticallysignificant,and adopted a •timeal., 1988]. Additional continuousscanmeasurements were independent parameterization. The time-dependent degradamadethroughFebruary1987,but thesedataarenot archived. tion coefficients(3.) was also assumedto remain constant

The SBUVinstrument beganexperiencing chopper nonsync betweendifferentaccelerated deployment periods. errors in February 1987, and continuousscan irradiance

measurements were terminated soon afterwards due to de-

Figure l a showsthe Nimbus 7 SBUV time seriesat 205

nm fromthe archiveddatafor the period1978-1986. Shortgradeddataquality[Gleason andMcPeters,1995]. Discrete termvariationsat the 2-6% peak-to-peak levelrepresent rota-

modemeasurements continued throughJune1990,although

tionalmodulation. Themagnitude of thechange in F!gurel a (-•30%from 1982 through1986) is considerably largerthan mentnoisecompared with discretedatatakenprior to the predicted solarcyclevariations of 6-10% [e.g.,Lean, 1991]. chopper nonsync period.Afterconducting special operations decrease in 1985-1986,whensolaractivitywas in summer 1990 to extend the lifetime of the TOMS instru- Thecontinued discretesolar data showed a factor of 7 increasein measure-

ment, Nimbus 7 SBUV data qualitydegradedfurther,and SBUV operationswere terminatedin June 1991.

2.2. Time-Dependent Characterization

An accuratecharacterization of long-terminstrumentre-

at a minimumlevel, stronglyimpliesuncorrected instrument changein the data. Althoughsmallerchanges areobserved at longerwavelengths (e.g. 5% decrease at 260 nm in Figure lb), thesechanges arestill a factorof 4 largerthanpredicted solarirradiance variations.The analysisof Schlesinger et al. [ 1988] demonstrated that the observeddrift in irradianceval-

sponse change is criticalto determining solaractivitychanges ues was consistentwith time-dependent variationsin s(3.). on long time scales.The Nimbus7 SBUV instrumentdid not Schlesinger et al. [ 1988]alsofoundthatthereference photocarryan onboard calibration system to directlytrackchanges diodedesigned to monitorlong-termchanges in P(t) wasitin instrumentthroughput. The analysisof Cebulaet al. self experiencing sensitivitychanges.They choseto usea 1

[1988]therefore focused on albedocalibration changes for nm averageof irradiancedatacenteredat 391.3 nm, corrected

ozone processing,and consideredobservedsolar'irradiance measurements as an indicatorof changesin instrumentsensi-

for goniometry (variations in theangularresponse of thesolar diffuserplate)and diffuserdegradation, to represent wave-

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA

9.5

t

21,571

I

9.0

8.5 s.o

7.5

7.0 6.5 1979

1980

œ981

1982

1983

1984

1985

œ956

1987

871-" "1"1"" "1"1"" """" "" ""t • 84 o

•

(b

1979

60 nm

1980

1981

1982

1983

1984

1985

1986

1987

DATE

Figure 1. (a) Time seriesof archivedNimbus7 SolarBackscatter Ultraviolet(SBUV) irradiancedataat 205 nm. (b) Time seriesof archivedNimbus7 SBUV irradiancedataat 260 nm.

length-independent instrumentsensitivitychanges.Normali- animprovement overtheformulation of Cebulaet al. [1988], zationof the Nimbus7 SBUV datato F(391 nm, t) meansthat it does have some limitations. The revised model of SC92 is anyerrorsin s(391nm) will be incorporated into irradiances unableto evaluatechangesin s0•,t) priorto the first accelerat otherwavelengths.Any changesin s00 for otherwave- ateddeployment intervalin 1980. The deriveddegradation lengthswill thereforebe maderelativeto s(391nm). ratesbecomeverysmallfor wavelengths longwardof 300 nm, Schlesingerand Cebula [1992] reviewedthe Nimbus 7 andthe corresponding uncertainties areproportionally larger. SBUV instrumentcharacterization usingthe full 8 yearrecord Sensitivitychangecoefficients for the 1986 accelerated deof continuousscanirradiancedata. They found significant ployment intervalare moreuncertain because of its shorter differencesbetweenvaluesof s00 derivedfrom the first two duration,leadingto increaseduncertaintyin the sensitivity accelerated deployment periodsin 1980 and 1981 andthose changecorrection for the 1985-1986irradiance data. determined fromlaterperiodsin 1984and 1986. Forthe calFigure3a showsthe 205 nm time seriesconstructed by culationof absolutesolarirradiances,a completecharacteriza- SC92, with a 27-dayrunningaveragealso shownto better tion of instrument response is required,but the apportionment illustratelong-termchanges.Solaractivityvariationsat 205 betweenexposure-dependent and time-dependent terms is nm predicted by theMg II indexproxymodel[DeLandand unimportant because corrections areneededfor all instrument Cebula,1993] are shownfor reference. While the Mg II throughputchanges.The coefficients0•,t) thereforeprovides model usesscalingfactorsderivedfrom short-termsolar a representation of all time-dependent responsechanges.The variationsto estimatelong-termirradiancechanges,comparioperationalform adoptedby SC92 for sO•,t)is constantduring sonswith NOAA 11 SBUV/2 irradiance data indicate that the irradiancechangesare accurate to -•1% on solar the earlyand late portionsof the SBUV datarecord,andvar- predicted as well [DeLandand Cebula,1998a]. The ies linearlyin between. Figure2 illustratesthe time depend- cycletimescales enceof s(t) at 205 nm derivedby SC92. The time dependent revised Nimbus 7 data track the Mg II model predictions in lateryearswill be discoefficientssi•_,s3,ands4derivedfrom regressionfits for each fairlywell. The smalldifferences furtherin section2.5. Figure3b showsa time seriesof corresponding intervalare alsoshownwith theirrespectivei 1 cussed c• standarddeviations. While this characterization represents SC92 irradiancedata at 240 nm. Consistentwith the predic-

21,572

DELAND AND CEBULA' SPECTRAL SOLAR UV IRRADIANCE DATA 1.OxlO-4

' ' I ' ' I ' ' I ' '

' ' I ' ' I '; ' I ' ' I ' ' I ' ' I ' ' Ii'

Interval •-•

':

Interval

,

' I:'

'

' I ' ' I ' ' I ' ': I ':'

Interval

3

4_-

-1.0x10 -4

•

-2.0xlO

-3.0xlO

m

-4.0xlO

-5.0xlO

-6.0xlO

-7.0x10

-4

-4

-4

-4

;

-4

-4

-

,,I,,I,,I,,I,,

1979

1980

!t•

t•z

i

I,,

I

1981

ts I,i,1,,

1982

I,,1,,I,,I,,

1983

i 1984

-

t4 i -

', , I , , I , , I , , 11,,• 1985

1986

1987

DATE

Figure2. Timedependence of theinstrument sensitivity change coefficient s(t)at 205nm. Thetimeinter-

vals are defined in SC92.

tionsof the Mg II proxymodel,solaractivityvariationsare August-October 1983. Extrapolating thisrelationship through againobserved,with an amplitudeapproximately one half October1986leadsto predictedwavelengthscalechanges of thatobserved at 205 nm. Figure3c showsthe SC92datatime +0.078 nm at 400 nm and-0.016 nm at 170 nm. No waveseriesat 325 nm, for whichpredictedsolaractivityis much lengthscaledrift correctionwas appliedto the archivedsolar lessthan1% overa solarcycle. Thesedatashowlong-term irradiance data. changesof 2-3%, increasing noticeablyfrom 1984 onward, We havereevaluated the SBUV wavelength scaledrift by which we interpretas uncorrected instrumentsensitivity usingmeasurements of absorption linesin the solarirradiance change. The implicationof 2-3% long-termdrifts in the spectrum.Thesemeasurements providebetterspectraldistriNimbus-7SBUV irradiance dataaddssubstantial uncertaintybutionandtemporalcoveragethanthe Hg lampdataandare to derivedestimates of solarcycleirradiance changes for cy- directlyrelevantto the solarirradiance product.Wavelength cle 21. Our goalin thispaperis to reducetheseuncertainties calibration analysisusingsolarspectrahavebeenpreviously throughreasonablerevisionsof the long-terminstrument presentedfor NOAA 11 SBUV/2 [DeLand and Cebula, characterization. Ourinvestigation of possible adjustments to 1998a],SSBUV [Cebulaet al., 1995/96],andGOME [Casper the SC92 instrument characterization focused on three areas: and Chance,1997]. We used16 absorption features between wavelengthscaledrift, periodicvariations,and long-term 200 and400 nm for thisanalysis.A setof consecutive points sensitivitydrift. wasidentifiedonboththe shortandlongwavelength sidesof eachfeaturewhich changeapproximately linearlyin irradi2.3. Wavelength ScaleDrift anceat the Nimbus7 SBUV resolution.Regression fits were and The Nimbus7 SBUV wavelengthdriveuseda mechanical derivedto eachsetof points,usingdailyaveragespectra, of thosefitscalculated.Theabsolute position cam to selectthe exact wavelengthfor eachmeasurement. theintersection line with an SBUV-type While differentportionsof the camwere usedfor discreteand derivedfor eachsolarabsorption instrument may not equal the high resolution referencevalue continuous scanobservations, thecammadea complete revoby the 1.1nm bandpass.However,because lutionfor eachscanin all operating modes.Thustheaccumu- dueto broadening latedwear on the cam due to continuous use (morethan only relativechangesare of interesthere,the variationsover of long-termchangesin the 3,000,000rotationsover8 years)represents a potentialsource time give a goodrepresentation wavelength scale. Cebula and DeLand [1998] showedthat of long-term driftin thewavelength calibration.Cebulaet al. NOAA 11 SBUV/2 wavelength drift could be characterized to [1988] evaluated measurements fromthe onboard Hg lamp calibrationsystemtakenapproximately twice/week.On the 0.01nm usingthismethod.Figure4 showsanexampleof the basisof datafromMay 1980throughDecember 1983,they Nimbus7 SBUV wavelengthdrift data for the Fe I line at a steadylineartime dederiveda wavelengthscalecorrectionrelationwith a linear 248.82nm. Thesedatademonstrate pendence, except for an unexplained dip of-0.012 nm bespectralandtime dependence: tweenJuly 1979 andJuly 1980. The dip is presentin dataat with smallvariations in durationandampliAk(k,t)= 1.392x10 -7(ko-209) (d-2092) (2) all wavelengths, tude. We omittedthisperiodwhenderivingwavelength drift whered is daynumber(1 -- 1 January1978). Thenominal ratesfor eachline to avoidbiasingthe regression fit. The wavelengthscale was referencedto measurements made in timedependence slopesfor all linesareshownin Figure5. A

DELAND

AND CEBULA:

SPECTRAL

SOLAR UV IRRADIANCE

DATA

21,573

• 39'' I'' I '' I'' I'' i'' I'' I'' I'' I'' I'' I'' I'' I'' I ' ' I ' ' I ' '•

•

•

35

• I • • I • t I • • I • • I • • I • • I , , I , , I , , I , , I , , I , , I , , I , , I , , I , ,•

3a•1979

Ja:1980

J• 1981

$a: 1982

Jan1983

•

1984

3a: 1985

Ja:1986

•• 820•• ,I,, (•)I,, • I, n••I,••I•, • I,,I,,I,•I,,I,,I,,I,,I'' I"'I' ' I' '•

• 810• 800 •

7901

•

•a•

1979

Ja:

1980

J•

1981

Ja•

1982

•an

1983

•a•

1984

Je•

1985

Je•

1986

J•:

1987

DAT•

Figure 3. (a) Time seriesof Nimbus7 SBUV solari•adianccdataat •05 rim,reprocessed usin• the i•s•umcntcharacterization of SC9•. •c heavysolidlinc is a •7-day •unnin• averageof the data. •c dashed ]incis a •7-dayaverage of the solarvariations at •05 nmpredicted by theM• II indexproxymodel,no•alizcdto the sta• of the i•adianccdataandoffsetby 3% fo• clarity. (b) Time seriesof Nimbus7 SBUV solar

i•adianccdataat •40 rim,reprocessed usi• theinstrument characterization of SC9•. •c heavysolidand dashedlineshavethe samemcanin•asin Figure3a. •c predictedsola•variationis offsetby 3% for clarity. (c) Timeseriesof Nimbus7 SBUV solari•adianccdataat 3•5 rim,reprocessed usin•theins•umcntcharacterizationof SC9•. •c heavysolidlinchasthe samemcanin•asin Figure3a. •c predictedsolarvafiatio• is offsetby ] % for clarity.

regression fit to theseslopesasa functionof wavelengthgave

rn= 5.7x10 -6

d < 485,d>_914

a valueof 8.35(+8.54)x10-9nm/day/nm, implying no statisti-

rn= -7.3xl 0-5 rn= 4.9xl 0-5

d = 486-638 d = 639-913.

cally significantspectraldependence to the wavelengthdrift. We thereforecalculatedan averagetime-dependentslope,

withavalueof 5.7(+2.1)xl 0-6nm/day. The total magnitudeof the wavelengthscaledrift fromNovember 1978 to October 1986 calculatedusing the wavelength-independent rate derived above is A)• = +0.017 nm. This is considerablysmaller than the drifts predictedby Schlesingeret al. [1988] at long wavelengths.The explanation lies in the fact thattheir analysisusedonboardcalibration data beginning in mid-1980, where the solar data show a short-termdip. Fitting data from mid-1980 to late 1983 thus gave time-dependentslopesthat were too large. Regression calculationsincludingthe early dip in the A)•(t) data from the solar absorptionlines give an averageincreasein A•,tota I of approximately+0.003 nm, which is less than the estimated accuracyof the linear regressionfit. However, becausewe believethat the dip representsa real aspectof the Nimbus 7

Irradiance changes associatedwith the 8-year wavelength drift correctionwere estimatedby applyinga 0.017 nm nm shiftto the initial SBUV irradiancespectrumandratioingit to the unshiftedspectrum. For the 1 nm binnedspectrum,typical changesare of the order of +0.4%, reaching+1.2% at spectrallocationsassociated with strongabsorptionlines such as Mg II and Ca II. The wavelengthscaledrift correctionis a smallbut importantcomponentof the Nimbus 7 SBUV longterm instrument

2.4.

Periodic

characterization.

Variation

The 205 nm time seriesshownin Figure3a exhibitsregular oscillationsin the later part of the record,where solaractivity is low. Removing solar variationspredictedby the Mg II instrumentbehavior, we have modeled it as a linear decrease index model ("desolarize"the data), as discussedin section from May 1979 to September1979, followed by a linear in- 2.2, leaves a residualtime serieswith constant2% peak-topeak (p-p) fluctuations at approximatelyannual intervals crease from October 1979 to June 1980. The maximum amplitudeof-0.012 nm for the dip represents an averagefor all throughoutthe datarecord(Figure6a). The periodicvariation wavelengths. This approach providesa morespectrally con- is alsoobservedat longerwavelengthswith decreasingamplisistentcorrectionthan usingsmootheddata from one or more tude,reaching---1%p-p at 230 nm (Figure6b) and---0.2%p-p lines. The revisedtime-dependentadjustmentto the reference at 260 nm (Figure 6c). Examinationof desolarizedtime series at 10 nm intervalsshowsan approximatelylinear change wavelengthscalefor eachdateis givenby in amplitude of the periodic variation over the wavelength (3) range 190-260 nm (Figure7). A)•(t) : m(t) (d-2092)

21,574

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA 0.03

Linear regression fit O.OB

0.01

+

+ +

+

•++

+

+-,-

6.00 g'+. *•+%

-0.01 -

+

i

i

; ,,

-0.02

++

+•

_

I • •

Dip region •,

lg?g

I•

I•

lg80

• •1,,

I,

lg81

• I,

1982

,I,,I

1983

•1

•1

•

lg84

I,•

I,•

I,

lg85

lg86

•1,

,

1987

Date

Figure 4. Nimbus 7 SBUV wavelength scale changes derived from measurements oftheFeI solar absorption line at 248.82 nm. The solid line is a linear regression fit, excluding the data between July 1979 and December 1980. Thelength oftheperiod suggests aproblem intheprocessbus7 SBUVdidnotexhibit anyPMTtemperature sensitivity ingalgorithm related to annually varying quantities such as in prelaunch calibration tests [Schlesinger etal.,1988].A1Sun-Earth distance, PMTtemperature, orsolar viewing angle though theSBUV/2instruments doshow a change inradio(i.e.,goniometry). TheSun-Earth distance correction istaken metric response withPMTtemperature change, typical sensifromU.S.Government PrintingOffice[ 1997],andisaccuratetivitiesare0.2%/øC.Evenif weassume theNimbus7 SBUV

tobetter than0.02%.Theextreme values of Sun-Earth dis- PMTgainhadthesame thermal sensitivity asthatobserved tancein earlyJanuary andJulyarenotin phase withthebroad on theSBUV/2instruments, because Nimbus-7SBUVPMT

minima andmaxima shown in theirradiance residuals. Nim- temperatures onlyvariedbetween 20-24øC, uncorrected re11_'

' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' ' I ' ' '_

_ _

_ _

_ _

10--

--

,,-,,q

_

-

m

-

-

•

-

_

•_,

-

_

:

:

• 7•

-

•

•.

i

• '.

•

•.

_

.

ß c

:

•

-

• _ • 4a

_

•

•

•

,,. ..... : •

.

.. •

•..'

•

'"...

'

...

i i

...............

:

•

-

-

:

•

_ -

i

.

.

.

-

,

/

,

,

-: _

.

. :

ß

:

• '.

•

.. " ..

;

_

•

....

:

ß .

...

;

.'

:

-

-

--

; ..

::

$

-

_ _

_ _

1-, 200

, , I , , , I , , , I , • [ } • • • I [ , t { , , t I , , [ { , [ , I , , , { , , ,220 240 260 280 300 320 340 360 380 400 420

Wavelength [n•]

Figure •. Special dependence ofNimbus 7 SBUV wavelen•h scale driR rates derived from solar absoftionlinedata. E•orbars are&2•. •e dashed lineshows theaverage ofallpoints.

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE

DATA

21,575

o -1

-4,

!

I i

i !

Jan19'•9

i !

i i

,;an1980

i i

Jan1981

i i

i i

.Jan 1988

4,--''1''1''1''1'•11'l''1''

i i

i I

I I

Jan19•.•

I''1''

I I

Jan1964

I''l•'l'

I i

i t

Jan1985

' I''

i i

i i

Jan198•

I''1''

i i -,;an1987

I'•-

_

_

_

-2.

-8--

_

-

-4

Jan 1979

.Tan1980

.Tan1981

.T•n 1982

.Tan 1981

Jan 1988

Jan 198.3

,/an 1984

,/an 198.5

•lan 19•

.Ten1987

Jan 198•

•lan 1984

,/an 1985

,/an 198•

Jan 1987

4 ! 0

..............................

-1

-8 -•3

-4, •lan 19't'9

Jan 19•0

DATE

Figure 6. (a) Residual("desolarized") time seriesof SC92irradiancedataat 205 nm, wherethe solarvariationspredictedby the Mg II indexproxymodelhavebeenremoved.(b) DesolarizedSBUV irradiancedata at 230 nm. (c) DesolarizedSBUV irradiancedataat 260 nm.

sponse variations wouldbe lessthanhalf the magnitude re- operational goniometric correction has no spectraldependquiredto producethe variationsshownin Figure6a. Again, ence,laboratoryandinflightdatafrom SBUV/2 instruments thephaseof PMT temperature variations is inconsistent with doshowa wavelength-dependent diffuserbidirectional reflecthe residuals. Other instrument housekeeping parameters, tancedistribution function(BRDF). However,the annual suchasvoltages andmotorcurrents, areextremely stabledur- variationof the solarazimuthangleis alsooutof phasewith ing the lifetimeof the mission.While the Nimbus7 SBUV the irradiancevariationsand has a significantsemiannual

-

- ......

1.2 '-•

Linear regressionfit

--

%

_

-

1.0--

- •• ••

-

--

_

_

•

_

•

-

_

b 0.8-

•

_

•

-

-

x•,,,

_

_

-

-

0.4--

-_

•

-

_

_ _

•

180

1•0

g00

g10

gg0

gg0

g40

gfi0

_

g80

Wavelongth [•m]

Figure 7. Special dependence of pehodicte• identifiedin Nimbus7 SBUV desolahzedresidualtime des between 190 and 260 nm.

21,576

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA

structure thatis not observed in Figure6. Quasi-annual varia- 2.5. Long-Term Sensitivity Drift tions are not observedin other solar data setsduring 19791986, includingirradiancedata from the mechanicallysimilar NOAA 9 SBUV/2 instrument[DeLand and Cebula, 1998b]. Consideringall of the informationpresentedhere, we conclude that the Nimbus 7 SBUV periodic irradiancevariations are dueto an unexplainedinstrumentaleffect. Becausewe have been unable to develop a satisfactory physicalmodel to explain the periodic irradiancevariations shownin Figure6, no correctionhasbeenappliedto the final irradiancespresentedin section3 of this paper. For the bene-

Theirradiance timeseriesresidualat 205nmshownin Figure 8 hasa noticeableupturnin 1984-1986,similarto the 325

nmdatashown in Figure3c. Thisfeature canbeidentified at

manywavelengths,suggesting a systematicerrorin the value of thes34sensitivitychangecoefficient.Schlesinger and Ceand Cebula[ 1992] examinedthe useof an s(t) functionwhich would explicitlyincludethe s4valuescalculatedfor the 1986 accelerated deployment,ratherthan averagingthem with the s3valuesderivedfor 1984. Theycalculatedirradiancespectra for October1986 usingboth s34(•,)and s4(•,). Ratiosof these fit of users who wish to evaluate the Nimbus 7 SBUV data spectrawere in the range0.992-1.014 and exhibitedno syswithout suchartificial behavior,we have developeda charactematictrend. This result implies no significantdifference terizationthat canbe appliedto improvethe appearance of the between these functional forms but does not rule out an error irradianceproduct. On the basisof a linearregression fit to in the s34(•,)values. When we examinednumerousindividual the observedamplitudesshownin Figure 7, it is possibleto irradiancetime series,we foundthat the adjustments to s34(•,) derivea normalizedamplitudefor the periodiccorrectionterm requiredto reduce long-termdrift to +1% rangedbetween givenby

-lx10-5day-• and+2x10 -5day-1,withlargervalues derived at

a(30 = 3.93xl0-2- (1.45xl0-4•)

260 nm. (4)

A completetime-dependentperiodic correctioncan then be written

as

shorterwavelengths. For wavelengthsshortwardof 260 nm the long-termsolarchangeterm duringthe SBUV datarecord calculatedby the Mg II proxy model beginsto exceed 1%. This term must be removedbefore deriving a /•34 value. Achievinga long-termdrift valuebetterthan_+1%from direct

analysis at shortwavelengths by specifying /•34 therefore A()•,t) = A()•) cos{27r[(d+110)/365.25]}

(5)

becomesdependenton the solar model results. To avoid

problems withcircular reasoning, we concentrated onwavewhered is the runningdaynumberas definedpreviously,and a phaseshift of 110 daysis usedto alignthe functionwith the data. Applying this correctionfunctionto the SC92 205 nm irradiance data reduces the fluctuations in the desolarized

residualto---0.5%(Figure8). The dip of-•2% in late 1979 is a consequence of applyingequation(5) to the full data set, since the same time period in Figure 6 is essentiallyflat. Comparableresultsare obtainedat otherwavelengths.

lengths longward of 300 nm, where solar cycle irradiance changeshave been shownto be 0.3% or less, exceptin specific narrowregions[Leanet al., 1997]. When we determinedchangesto s34(•,)that minimizedthe long-termdrift at 10 nm stepsin the 300-400 nm wavelength

region, wefoundanaverage valueof /•34 = 1.2(+0.4) xl 0-5 day 4. Theuncertainty represents theprecision to whichwe can specify/•34 for a givenresidualtime series,ratherthan a

-3

-4

1979

1980

1981

1982

1983

1984

1985

1986

1987

DATE

Figure8. Residual timeseriesof SC92SBUVirradiance dataat 205 nm. Thesolarvariations predicted by theMg II indexproxymodelhavebeenremoved.

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE

DATA

21,577

1 06 1.04 1.02

1,00 0.98 0.96

0.94 0 92

II 1,03k''I'' •' 102F

I' ' I' '1 ' ' I ' ' I' ' I ' ' I '' I ' ' I ' 'l ' ' I ' ' I '' I ' ' I' '1 ' ' •

• 1oo•--................... •

096r- •

• 1.o2_' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I ' ' I "' I" ' I ' ' I ' ' . ' '• ••:g

-

(c) 325 nm

101 --

1.00

--

0.99 _

098

, , I • • I , , I , , I , • I • , I , , I , , I , , I , , I , , I , , I , , I , , I , • I , , I , Jan 1979

Jan 1980

Jan 1981

Jan 1982

Jan 1983 DATE

Jan 1984

Jan 1985

Jan 1986

Jan 1987

Figure9. (a)Timeseries of Nimbus 7 SBUVsolarirradiance dataat205nm,reprocessed usingtheinstru-

mentcharacterization of SC92andtheadjustments described in thispaper.Theheavysolidlineis a 81-day

running average ofthedata.(b)Timeseries ofNimbus 7 SBUVsolarirradiance dataat240nm,reprocessed asin Figure9a. (c)Timeseries ofNimbus 7 SBUVsolarirradiance dataat325nm,reprocessed asin Figure 9a.

statisticalcalculation.Thesechangesare comparable to the 2 • uncertaintiesfor S34listed in Table 4 of SC92 at •. < 300 nm, and approximately a factorof 2 largerat longerwave-

lengths. The uncertainties providedin SC92 were formal

In order to assessthe magnitudeof remaining long-term drifts in the Nimbus 7 SBUV data, we used the method of

DeLand and Cebula [ 1998a] as appliedto NOAA 11 SBUV/2 data.

Desolarized

time series of Nimbus

7 SBUV

data are

statisticalerrorsfor eachwavelength,and did not includea shownin Figure 10 for 200-208 and 240-250 nm. The 200possiblesystematic errorin S34.ThUSthe adjustment recom- 208 nm data (Figure 10a) clearly show the 2% p-p periodic mended here is reasonable within the uncertainties of the

SC92analysis.The sensitivity change•'34 is appliedto all wavelengths. 3. Revised 3.1.

Data

Nimbus

7 SBUV

Irradiance

Data

Set Creation

For the creation of a revised Nimbus 7 SBUV solar irradi-

variationbut no long-termdrift at the 0.5% level. The 240250 nm data decreaseby-2% duringthe first few monthsof operation,remain stableto betterthan 1% through1984, then increaseby-l% in 1985-1986. Theseresultsare very similar to the long-term drift in desolarizedNOAA 11 data for the samewavelengthbands,as shownin Figure4 of DeLand and Cebula [ 1998a]. For examinationof the entirespectralrange, irradiance time series were calculated in 5 nm bands between

170 and 400 nm, desolarizedby removing predicted solar variations,then averagedin 1O-daybins for clarity. We stress that thesecomparisons are not intendedto draw conclusions aboutthe validity of the Mg II model as a representationof solarchange. Rather,they representa validationtool for our estimatesof the long-termaccuracyof the Nimbus 7 irradiance data. Each time seriesis normalizedto the averageirradiancevalue duringthe first week of observations(November 7-14, 1978). Plate 1 showsthat the majority of the data longward of 200 nm exhibit long-term changesof +1% or less, with larger valuesof-2% at 250-270 nm and 340-360 smoothed maximum irradiance value in late 1981 and mininm, and+2% at 215-230 nm. Inspectionof the datawith narmum irradiancevalue in early 1985 both coincidewith ex- rower wavelengthbins indicatesthat theseproblemsare typitremaof the periodicvariation,thusexaggerating the derived cally generatedby spectrallysmallregions. The presenceof solarcycleamplitude.Thesesolarcycleirradiance variations the periodicvariationis clearlyevidentat wavelengthsshort-

ancedataset, we first appliedthe time-dependent sensitivity changecorrections of SC92to thearchiveddata,adjusting the $34(•)valuesasdiscussed in section2.5. We alsoappliedthe wavelengthdrift correctionderivedin section2.3 (equation (3)). Figures9a-9cshowtimeseriesof therevisedirradiance dataat 205, 240, and 325 nm, wherethe time serieshavebeen normalizedto the beginningof the Nimbus7 datasetto more readilyassess solarcyclevariations.The changein solarirradiancefor solarcycle21, asindicatedby an 81-dayrunning average,is -8.3% at 205 nm and 4.9% at 240 nm. When evaluatingthese estimates,it should be noted that the

are consistent with predicted changes of 8.3(50.3)%and 3.7(50.2)%respectively, as derivedfromthe Mg II index

ward of 240 nm in Plate 1.

proxymodelusinga solarcycleamplitude of 7.6% for the81daysmoothed Mg II index.

rapidincreaseduring 1979 andearly 1980. This suggests that the assumptionof time-independentsensitivitychange(So)

Shortward of-185

nm, the Nimbus 7 SBUV data show a

21,578

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA

1.01

(a) 200-208 nm

m•, 1.00 II

'" 0.98 IIIIl[1111111llllllllllllllll[11111111111

ol.02

'1'''''1'''''1'''''1'''',1''•''

I•'''11'''''1'''''_

-

1.01

:8),214 • 1.oo • 0.99 o

0.98

0.97

1g79

1980

1981

1982

1983

1984

1985

1988

1987

DATE

Figure10. (a) Desolarized timeseries of Nimbus7 SBUVirradiance dataat 200-208nm,withpredicted solarvariations removed.(b) Desolarized timeseriesof Nimbus7 SBUVirradiance dataat 240-250nm,with predictedsolarvariationsremoved.

duringthatperiodis incorrectat shortwavelengths.Unfortunately, we have no data on which to basea correctionin this spectralregion Withoutresortingto circular arguments. We thereforeleavethe instrumentcharacterization unchanged and acceptthat an errorlikely existsin the irradiancedata. From 1981 through 1986 the correctedNimbus 7 data in the 170-

i 85 nmregionareconsistent withtheMg II proxyestimate at

the 3% level. Measurementsof cycle 21 from Mg II index, sunspots,and other solar activity indexesshowthat the solar maximum period extendsthroughthe end of 1981. Thus, if the 1979-1980 data are neglected,Nimbus 7 SBUV data at

3.< 185nmcanstil1beused toestimate solar cycle irradiance changes.We alsorecommendcautionin usingthe 1979-1980 datafor studiesof short-termactivity,becausethe anomalous

1.0

0.8

0.8

o.e

o• 0.0 •

-0.4

-0.8

-0.8

-1.0

-,

180

, , I,

180

,,

I , ,,

200

I,

220

, , I,

240

,,

i,

• , I , ,,

280

280

I ,,,

300

I,

320

, , I , ,,

340

I ,,,

380

I,,

380

, I •

400

Wavelength [•%•,]

Fibare ] ]. Co•clafioncoefficients for hncarrecession•s bc•wccnN•mbus7 SB• ancotime sc•cs•aspresented in •hispaper,and•hcNimbus7 ME H index.

• nm avcraEc•ad•-

DELAND

AND CEBULA:

SPECTRAL

SOLAR UV IRRADIANCE

DATA

21,579

o ao o•

[mu] HœON,•q2AV•

21,580

DELAND

AND

CEBULA:

SPECTRAL

SOLAR UV IRRADIANCE

DATA

we adoptasouruncertainty for theinstrument sensitivity adjustment.Leanet al. [1997]suggest a maximum amplitude The influence of residual instrument characterization errors for irradiancechanges longwardof 300 nm of 0.3% overa to a decrease in irradican be evaluatedquantitativelyby derivinglinearregression solarcycle,whichwouldcorrespond fits between irradiancetime seriesand the Nimbus 7 Mg II anceduringtheNimbus7 SBUVtimeperiod,occurring priindex. The correlationcoefficientsfor regressionfits with marilyin 1982-1984.Figure9c shows a change of approxiandtemporallocationin the325 each1 nm irradiancebin areshownin Figure11. High values matelythecorrectmagnitude of-0.90-0.97 are found between 185-215 nm and 230-262 nm data. However,the useof an empirically derivedzXs34 drift at 170 nm becomessignificanton solarrotational(-27day)timescales.

nm, indicatingthat solarvariabilityrepresents80-94% of the

correctionand PMT gain changecorrectionbasedon 391.3

variance in the irradiance data. Reduced correlation values at

nm datamakesit difficultto citethisresultas convincing

•, < 185 nm demonstrate the effect of the 1979-1980

evidence of realsolarchange. 3.2.3. Instrumentsensitivity(interpolation). Schles-

calibra-

tion problems. The dip centeredat 220 nm reflectsthe presenceof uncorrectedlong-termdrifts in this region,as shown by the greenvaluesduring 1985-1986 in Plate 1. Correlation coefficientslongwardof-286 nm are generally+0.6 or less, consistentwith the hypothesisthat solar variations in this region are not well representedby a chromospheric proxy suchas the Mg II index. The high correlationsobservedat 335-365 nm indicatea long-termdrift in the samedirectionas the solarcycleMg II index change. Repeatingthe regression analysisafter applyingthe periodic variation correctionde-

ingeret al. [1988]notethatthesplineinterpolation of s(•,)to get valuesat all wavelengths hasincreased uncertainty at wavelengths between thenodalvaluesusedin thefit. They estimatedthe uncertaintyof s(•,) as 0.1% near a reference

wavelength, increasing to 0.3%between suchwavelengths. 3.2.4. PMT gainchange.In SC92(andSchlesinger et al. [1988]),the functional form of the termP(t) representing

PMT gain changewas the 391 nm irradiancetime series, normalizedto the beginningof the datarecord. The function rived in section 2.4 increases correlation values at •, < 260 nm applied to the first 6 years of SBUV data is shown in by only 0.01-0.05. This demonstratesthat the information Schlesingeret al. [1988], with an overall range of approxicontentof the irradiancedata is essentiallyunaffectedby the mately+2%. Unpublisheddata for 1985-1986 show similar anomalous variation. variations.This correctionassumes no significantsolaractivity at 391 nm over a solar cycle, and also assumesthat the PMT gain changeis wavelength-independent.We assignan 3.2. Long-Term Uncertainty uncertaintyof 0.2% to this term, basedon the scatterin the

The followingtermsrepresentcomponents of the uncer- irradiance data. taintyin the long-termirradiancechangesderivedfromNim3.2.5. Goniometry. The Nimbus 7 orbit was maintained bus7 SBUV irradiance data. Somediscussions wereprevi- at the sameEquator-crossing time through 1984, and drifted ouslypresented in Schlesinger et al. [ 1988]andSC92andare by less than 10 min throughthe end of the SBUV spectral repeatedhere for convenience.

irradiance record in October 1986.

Examination

of individual

3.2.1. Periodicvariation. Althoughwe believethe peri- scandatasuggests a possiblegoniometriceffectduring 1985odicvariationsdiscussed in section2.4 represent an artificial 1986 correlated with orbital drift of no more than +1% in behavior,it is notclearwhetherthevariations areanincrease, daily averageirradiancevalues. However,this erroris in the decrease,or oscillationabout the "true" irradiance. We there-

samesenseas the zXs34 error discussedin section2.5. Thus, if

foremustconsider the full amplitude asan uncertainty term, with a wavelength-dependent amplitudevaryingfrom 1.0%at 200 nm (2% peak-to-peak) to zerolongwardof 260 nm. 3.2.2. Instrument sensitivity(adjustment). In section 2.5, we derivedan averagesensitivitychangecoefficientad-

any goniometryerror associatedwith orbit drift is wavelength-independent, it will be absorbed by the zXs34 correction. The residuallong-termgoniometryerrorin the irradiancedata is thereforesmallerthan the diffuser uncertaintyterm. We have assignedan arbitraryvalue of 0.3% to the goniometry

justment of zXs34 = 1.2(+0.4) xl 0-5day-•. If weconsider this

error term.

termto representour bestestimateof diffuserchanges, then 3.2.6. Interrange ratio. BecauseNimbus 7 SBUV measthe corresponding uncertaintyin irradiancecanbe evaluated ured all gain rangesfrom the last dynodeof the PMT, any by calculatingtime serieswith higherand lower valuesof changesin the ratiosbetweengainrangeswouldhaveto occur zXs34. This analysisyieldsirradiancechangesof 0.6%, which in the instrument electronics. The estimated error of this as-

Table1. Nimbus7 SBUVLong-Term Uncertainty Values a 205 nm

240 nm

300 nm

390 nm

Periodic Variation

Term

2.0%

0.8%

0.0%

0.0%

Instrumentsensitivity(adjustment)

1.2%

1.2%

1.2%

1.2%

Instrument sensitivity (interpolated) b PMTgainchange b Interrange ratio b

0.6% 0.4% 0.6%

0.2% 0.4% 0.6%

0.2% 0.4% 0.6%

0.2% 0.4% 0.6%

Goniometry Wavelengthdrift

0.6% 0.4%

0.6% 0.4%

0.6% 0.4%

0.6% 0.4%

Total (RMS)

2.6%

1.8%

1.6%

1.6%

aAll terms are +2 c• errors.

•From Schlesinger etal.[1988].

DELAND

AND CEBULA:

SPECTRAL

SOLAR UV IRRADIANCE

4 3

DATA

(c)24.0-250 nrn

21,581

' -..f '.: '

2 ß

1

o

-1 "" .':.:i

: ;..

(a) 170-180 nrn .'5'"'" ß

•1,,,,,

o• -10

1982 m

5

•-

4

.

I,,,,, 1983

I,,,,,I,,',,,I,,,,, 1984

1985

1986

1982

1987

(b) 200-208 nm

1983

1984

1985

1986

1987

1985

1986

1987

3

2

•

1

1

O.

0 -1

-1

:"

-2

'""'

-3

1982

1983

1984

1985

1986

1987

Date

1982

1983

1984 Date

Figure12. (a)Difference between Nimbus 7 SBUVandSMEirradiance timeseries at 170-180 nmforthe timeperiod January 1982to October 1986.Datarepresent dailyvalues.Theheavysolidlineis a 27-day running average ofthedifference values.Eachdatasetwasnormalized totheaverage irradiance valuefor January 1-3,1982,priorto calculating differences. (b)Difference between Nimbus 7 SBUVandSMEirradiance timeseries at200-208nm. (c) Difference between Nimbus7 SBUVandSMEirradiance timeseries at240-250nm. (d)Difference between Nimbus7 SBUVandSMEirradiance timeseries at290-300nm.

sumptionis 0.3% during the first 5 years of operation 4. Comparisonswith SME Irradiance Data [Schlesinger et al., 1988]. Measurements of solarUV activitywhichoverlappart of 3.2.7. Wavelength drift. Cebula et al. [1998] showed the Nimbus 7 data record are available from the Solar Mesothat for NOAA 11 SBUV/2, which hasthe samebandpassas sphereExplorer(SME). This instrumentis describedby Nimbus 7 SBUV, errorsin 1 nm binned irradiance data due to Rottman et at. [1982] and mademeasurements in the wavea wavelengthdrift errorof 0.02 nm weregenerally1% or less, length region 115-302 nm. Daily spectral irradiance datain 1 exceeding2% only at strongabsorptionlines. The wavenm bins covering the period January 1982 to June 1988 are lengthdrift correctionderivedin section2.3 givesan uncertainty of-0.007 nm based on the averagetime-dependent available from the National Space Science Data Center and Heath [1988] presented comparislope. This corresponds to a maximumerror of +0.2% for 1 (NSSDC). Schlesinger nm binneddata,increasingto 0.4% near the Mg II and Ca II absorptionfeatures. Table 1 liststhe numericalvaluesof the uncertaintyterms discussed here for 205, 240, 300, and 390 nm. Calculating the RMS 2 o uncertaintyof theseterms,we find that the Nimbus 7 SBUV estimatedvariationfor solarcycle21 at 205 nm is AF205= 8.3(+2.6)%, basedon the 81-day smootheddata shownin Figure9a. For 240 nm the estimatedsolarcycle21 variationis /•7'240----4.9(+_1.8)%. The 20 uncertaintyvalues longwardof 300 nm are considerablylarger than any predictedsolarchanges.Theseresultsare comparableto NOAA 11 SBUV/2 at shortwavelengths,andsomewhatlargerat long wavelengths[Cebuta et at., 1998]. Comparisonsof UARS SUSIM and SOLSTICE irradiancedatasuggestthat the longterm relative accuracyof these data is currently 1-2% and may approach1% when the final instrumentcharacterizations are available[Rottman,1998]. The uncertainties quotedhere for Nimbus 7 SBUV allow comparisonof solar cycle irradiancevariationsto an accuracyof-2%.

sonsbetween archivedNimbus 7 SBUV data, SME data and selectedrocket measurementsof solar UV irradiance. They

focused on absolute irradiance differences between these data

sets,andexaminedthe impactof spectrally dependent noise on determination of solaractivitymagnitude fromirradiance ratios. Theyconcluded that 3-5% noisein the SME spectra availableat that time preventedthe determination of solar

changes at )•> 170nm fromratiosusingthearchived data.

Rottman [1988]reviewed animproved SMEdatasetin light of previous estimates for solarUV andEUV variability. Their estimatedirradiancevariationsfor solar cycle 21 were

approximately 50(+_ 15)%at121.5nm(Lyman alpha), 8(+_5)% at 170-180nm, and7(+_5)%at 205 nm.

In thispaper,we compareNimbus7 SBUV andSME solar irradiancetime seriesfor the overlapperiod from January 1982 to October1986. This coversessentiallythe full range of solaractivityfor solarcycle21. The SBUV datawerefirst reducedto 1 nm binneddaily averagespectrafor consistency with the archivedSME product. Time serieswere then con-

21,582

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA

structedin 10 nm bands,normalizedto the averageirradiance valueson January1-3, 1982, to removedifferencesin absolute irradiances, and smoothed with a 5-day binomialweighted average. Solar cycle irradiancevariation values refer to 27-day averagetime seriesin order to remove the impactof rotationalmodulations. Figure 12a showsthe differencebetweenSBUV and SME time seriesfor the wavelengthband 170-180 nm. The longterm irradiancechangebetween January-February 1982 and September1986 derivedfrom the individualtime seriesdata is ---10%for SBUV and 7% for SME, basedon inspectionof the 27-day averageddata. The Mg II indexproxymodelpredictsa changeof 9.4(+0.6)% overthe sameperiod. A relative drift of 10% between SBUV and SME can be seen during 1982-1984. Inspectionof desolarizeddata suggests that the largerchangesin late 1984 and late 1985 areprimarilycaused by SME data variations. Figure 12b showsthe difference between SBUV

We haveproduced a datasetof Nimbus7 SBUVdailyaveragespectral irradiances in 1 nm binsoverthe wavelength rmlge170-400nm, extendingfrom November7, 1978,to October25, 1986. Thesedataare availableby contacting the authors or on-line through our Web page at

http://ventus.gsfc.nasa.gov/solar.html. Wehavealsoarchived therevisedNimbus7 spectral in'adiance dataat theNational Geophysical Data Center(NGDC). Irradiancedataat the nominalinstrument sampling intervalof 0.2 nm areavailable uponrequest.We arecurrently workingtowardsthecreation of a correctedspectralirradiancedata set for NOAA 9 SBUV/2, which made measurements over a completesolar

cycle(March1985to July1997). Thosedata,combined with the Nimbus 7 SBUV and NOAA 11 SBUV/2 data sets, will allow us to assembleand examine a continuous20+ year re-

cordof solarspectralUV irradiances from nearlyidentical instruments.

and SME irradiance time series for the wave-

lengthband 200-208 nm. The wavelengthrangewasreduced for this figure to avoid complicationsfrom averagingacross the sharpirradianceincreaseat the A1 ionizationedge. Longterm irradiancechangesin this band are essentiallyidentical for both instruments,with the dominantstructurecorresponding to the SBUV periodicvariation. The 240-250 nm band (Figure 12c) and 290-300 nm band (Figure 12d) showvery similarbehavior. No long-termdrift is observedduring19821984, followed by a 2-3% relative drift during 1985-1986. While it canbe difficult to partitionthe trendresultsbetween differentinstruments,examinationof Figure 10b suggests that most of the drift in Figure 12c is due to the SBUV data. The

residualirradiance changes for eachwavelength bandshown in Figure12 arewell withinthe combined long-term uncer-

taintyvaluesquoted in thispaperforNimbus7 SBUVandin

Acknowledgments.We greatlyappreciate thecomments of two anonymous reviewers,whichimprovedthefocusof ourpaper. Barry Schlesinger generously reviewedthe entiremanuscript, andcontributed many technicalinsights. Gary Rottmansuppliedimportant informationabout the SME data set. Tom Woods and Linton Floyd

providedvaluableadvice. This researchwas supported by NASA contract NAS 1-98106.

JanetG. LuhmannthanksScottM. BaileyandClausFr6hlichfor theirassistance in evaluatingthispaper. References Bhartia, P. K., R. D. McPeters,C. L. Mateer, L. E. Flynn, and C. Wellemeyer,Algorithmfor the estimationof verticalozoneprofiles from the backscatteredultraviolet technique,J. Geophys. Res., 101, 18,793-18,806, 1996.

Bjamason, G., andC}.R6gnvaldsson, Coherency between solarUV radiation and equatorialtotal ozone, J. Geophys.Res., 102,

the work of Rottman[1988] for SME.

13,009-13,018, 1997.

Caspar,C., and K. Chance,GOME wavelengthcalibrationusing solar and atmosphericspectra,Proceedingsof 3rd ERS Symposium,Eur. SpaceAgency,Spec.Publ., ESA SP-414, 3, 609-614,

5. Conclusion

May 1997. In thispaper,we havecreated animproved solarspectral Cebula, R. P., H. Park, andD. F. Heath,Characterization of theNimirradiance product fromtheNimbus7 SBUVinstrument for bus-7 SBUV radiometerfor the long-termmonitoringof stratosolarcycle21. Implementation of therevised long-term charsphericozone,J. Atmos.OceanicTechnol.,5, 215-227,1988. acterization derivedbySchlesinger andCebula[ 1992]largely Cebula,R. P., E. Hilsenrath,P. W. DeCamp, K. Laamann,S. Janz, and K. McCullough,The SSBUV experimentwavelengthscale removeddriftsof up to 30% thatwerepresent in thepreviandstability:1988to 1994,Metrologia,32, 633-636,1995/96. ouslyarchived irradiance data.We havederived refinementsCebula, R. P., M. T. DeLand, and E. Hilsenrath,NOAA 11 Solar

to the SC92 instrument characterizationfor residual errors

Backscatter Ultraviolet,model2 (SBUV/2) instrument solarspec-

after1984,andderivedan improved wavelength driftcorrection fromanalysis of solarabsorption lines. An anomalous

tral irradiance measurementsin 1989-1994, 1, Observationsand

periodic variation in thedataat shortwavelengths ()•< 270 nm)wasmodeled empirically, buthasnotbeenremoved from thefinalirradiance product.Theperiodic variation haslittle

long-termcalibration,J. Geophys.Res., 103, 16,235-16,249, 1998.

Chandra,S., andR. D. McPeters,The solarcyclevariationof ozone in the stratosphere inferredfrom theNimbus7 andNOAA 11 satellites,J. Geophys.Res.,99, 20,665-20,671,1994. effect on the informationcontentof the irradiancedata. The Chandra,S., J. R. Ziemke,andR. W. Stewart,An 11-yearsolarcycle in tropospheric ozonefrom TOMS measurements, Geophys.Res. Nimbus7 SBUV dataproduced with theseadjustments show Lett., 26, 185-188, 1999.

a long-term irradiance change of 8.3(+2.6)% at 205nmand DeLand,M. T., and R. P. Cebula,CompositeMg II solaractivity 4.9(+1.8)%at 240nm for solarcycle21, using81-dayaverindexfor solarcycles21 and 22, J. Geophys.Res.,98, 12,809agedtimeseries.Theseirradiance changes arein agreement 12,823, 1993. withpredicted long-term changes of 8.3%at 205 nm and DeLand,M. T., andR. P. Cebula,NOAA 11 SolarBackscatterUltra3.7% at 240 nm, respectively, usingthe Mg II indexproxy model. The Nimbus7 resultsarealsoin agreement with over-

lapping datafromtheSMEsatellite for solarcycle21. The Nimbus7 SBUVdatapresented herecapture thefullrangeof mid-UV solarvariabilityfor solarcycle21 with a long-term

violet, model 2 (SBUV/2) instrumentsolar spectralirradiance

measurements in 1989-1994, 2, Results,validation,and comparisons,J. Geophys. Res.,103, 16,251-16,273,1998a. DeLand,M. T., andR. P. Cebula,SolarUV activityat solarcycle21 and 22 minimum from NOAA-9 SBUV/2 data, Sol. Phys., 177, 105-116, 1998b.

Floyd,L. E., L. C. Herring,D. K. Prinz, andP. C. Crane,Instrument accuracy of---2%.Thesedatacanbeusedtoconstrain models responsivityevolutionof SUSIM UARS, Proc. SPIE Int. Soc. of solarvariability mechanisms andto providerealistic solar Opt.Eng., 3427, 445-456, 1998. Gleason,J. F., andR. D. McPeters,Corrections to theNimbus7 solar inputsfor atmospheric models.

DELAND AND CEBULA: SPECTRAL SOLAR UV IRRADIANCE DATA

21,583

backscatterultravioletdata in the "nonsync"period (February rence,D. W. Rusch,R. W. Sanders, G. E. Thomas,andJ. London, 1987 to June1990),J. Geophys.Res.,100, 16,873-16,877,1995. Solarspectralirradiance,120 to 190 nm, Octobe r 13, 1981, to Heath,D. F., and B. M. Schlesinger, The Mg 280-nmdoubletas a January3, 1982, Geophys.Res.Lett., 9, 587-590, 1982. monitorof changesin solar ultravioletirradiance,J. Geophys. Schlesinger, B. M., and R. P. Cebula,Solar variation1979-1987 Res., 91, 8672-8682, 1986.

Heath,D. F., A. J. Krueger,H. A. Roeder,andB. D. Henderson,The Solar BackscatterUltraviolet and Total Ozone Mapping Spectrometer(SBUV/TOMS) for NimbusG, Opt. Eng., 14, 323-331, 1975.

estimated froman empiricalmodelfor changes with timein the sensitivity of theSolarBackscatter Ultravioletinstrument, J. Geophys.Res.,97, 10,119-10,134, 1992.

Schlesinger, B. M., andD. F. Heath,A comparison of solarirradiancesmeasured by SBUV, SME, androckets, J. Geophys. Res.,93,

7091-7103, 1988. Hood,L. L., and S. Zhou,Stratospheric effectsof 27-daysolarultravioletvariations:An analysisof UARS MLS ozoneandtempera- Schlesinger, B. M., R. P. Cebula,D. F. Heath,andA. J. Fleig,Nimture data,J. Geophys.Res., 103, 3629-3638, 1998. bus-7SolarBackscatter Ultraviolet(SBUV)SpectralScanSolar Jackman,C. H., E. L. Fleming,S. Chandra,D. B. Considine,and J. Irradiance and Earth RadianceProductsUser's Guide,NASA E. Rosenfield,Past,present,andfuturemodeledozonetrendswith Res.Publ.,RP-1199,74 pp., Feb. 1988. comparisons to observedtrends,J. Geophys.Res., 101, 28,753- U.S. Government Printing Office,Astronomical Almanac, Washing28,767, 1996.

Lean,J., Variationsin the Sun'sradiativeoutput,Rev.Geophys.,29, 505-535, 1991.

Lean, J. L., G. J. Rottman,H. L. Kyle, T. N. Woods, J. R. Hickey, and L. C. Puga, Detectionand parameterizationof variationsin solar mid- and near-ultravioletradiation(200-400 nm), J. Geophys.Res.,102, 29,939-29,956, 1997. Rottman,G. J., Observationsof solarUV and EUV variability,Adv. SpaceRes.,8(7), 53-66, 1988. Rottman,G. J., Observationsof the Sunat ultravioletwavelengths:1 to 400 nm, in Solar Analogs,Proceedingsof the SecondLowell Fall Workshop,editedby J. C. Hall, pp. 31-40, Lowell Observatory, Flagstaff, AZ, available at http://www.lowell.edu/users/jch/workshop/sa.html, 1998. Rottman, G. J., C. A. Barth, R. J. Thomas, G. H. Mount, G. M. Law-

ton, D.C., 1997.

Weber, M., J.P. Burrows,and R. P. Cebula,GOME SolarUV/VIS

irradiance measurements in 1995and1996- Firstresults onproxy solaractivitystudies, Sol.Phys.,177, 63-77, 1998. Woods,T. N., andG. J. Rottman, In-flightdegradation resultsforthe

UARS SOLSTICEinstrument, Proc.SPIEInt. Soc.Opt.Eng., 3427, 457-468, 1998.

R. P. CebulaandM. T. DeLand,ScienceSystems andApplications,Inc., 10210 GreenbeltRoad, Suite400, Lanham,MD 20706. (matthew_deland•sesda.com)

(Received November22, 2000; revisedApril 3,2001; accepted April 3,2001.)