Reports

Observations of Solar Irradiance Variability

Abstract. High-precision measurements oftotal solar irradiance, made by the ac-

tive cavity radiometer irradiance monitor on the Solar Maximum Mission satellite,show the irradiance to have been variable throughout the first 153 days of observa-tions. Thas 0.001with higi1368.31,irradian(

developivariabilizone in s

The .diance nthe NA(SMM) s1980, wi

servatioi(1). Theare (i) toon solarcover at(about 2:term pri

shorter-imonths)accurac)

0.05g

cxn

c

0

E

-

CD

0

.0

0.vu

-0.05

-0.10

-0.15

-0.204

Fig. 1. Tovariation;tick markextent ofvertical bkassociatedwith channel A bet.

700

with precision and accuracy not pre-viously achieved in satellite observa-tions. The sensors view the sun througha 5° (full angle) field of view. Wavelengthsensitivity is nearly uniform from the farultraviolet through the far infrared with acavity absorptance of 0.9995 (3). Sepa-rate shutters on each sensor facilitatetheir operation with different frequenciesfor all possible combinations in either au-tomatic or manual modes. The three sen-sors are used in various combinations toprovide periodic cross-references on the

We corrected data resolve orbit-to-orbit variations with uncertainties as small system's performance. This phased usepercent. Irradiancefluctuations are typical ofa band-limited noise spectrum of the three channels is designed to sus-h-frequency cutoffnear 0.15 day-1; their amplitudes about the mean value of tain the precision of ACRIM's observa-watts per square meter approach + 0.05 percent. Two large decreases in tions within 0.1 perceht for at least 1ce of up to 0.2 percent lasting about I week are highly correlated with the year. Comparisons with other sensors onnent of sunspot groups. The magnitude and time scale of the irradiance rocket and Space Shuttle payloadsty suggest that considerable energy storage occurs within the convection should sustain the multiyear precision ofsolar active regions. ACRIM within 0.1 percent for the life of

the SMM. The theory and operation ofactive cavity radiometer irra- variability that are significant for solar ACRIM are described in (2).nonitor (ACRIM) experiment on physics investigations. This report pro- The ACRIM observations over theLSA Solar Maximum Mission vides a brief description of the in- first 153 days of the SMM are presentedspacecraft, launched in February strument's operation and preliminary re- in Fig. 1. Shown are the mean values foras designed to make regular ob- sults of the first 153 days of observa- each orbit as measured by channel A, ad-ns of the total solar irradiance tions. justed to give the total solar irradiance atprinciple goals of the experiment The ACRIM instrument has three ac- 1 A.U. and plotted as percentage varia-begin a climatological data base tive cavity radiometer (ACR) type IV tions about the mean for the 153-day pe-irradiance variability that will sensors, the most recent version of a se- riod.least one solar magnetic cycle ries of flight pyrheliometers developed at Each orbital mean is an integration

2 years) with ± 0.1 percent long- Jet Propulsion Laboratory (2). Each over the solar observing portion of oneecision, and (ii) to provide a ACR sensor is an independent, electri- orbit. It is formed by integrating 32 1-term data base (minutes to cally self-calibrated cavity pyrheliome- second samples from each shutter-openwith maximum precision and ter, capable of defining the radiation period and averaging the results. A maxi-

y for the study of aspects of solar scale at the level of total solar irradiance mum of 28 shutter-open solar observa-tion periods occur per orbit. During thefirst 153 days of the SMM the average

1980 date number of shutter-open periods per orbit3/1 3/15 4/1 4/15 5/1 5/15 6/1 6/15 7/1 7/15 was 24, corresponding to an average so-

| ' ' ' ' lar observing period of about 52 minutesduring each of the 15 orbits per day. Or-bital variations of ACRIM temperatureshave been smaller than predicted, which

I t ; 85 8ii > l ........;\|>^\1enhances the resolution noise limit of the[ii1§eITS i Ev PTM!' l§/ i,2experiment. Although the irradiance

- y r B cj \ # 5 Stl C w 5 h data, sampled every second, have aJ i'.| !! .B single-sample (± I bit of 13) analog-to-

digital uncertainty of ± 0.02 percent, thec B noise is sufficiently oversampled that re-

C ' , sults integrated over longer intervalsshow no influence from the digitization, B i| i limit.

B The following corrections (listed in or-der of significance) are applied to the ir-

Weighted mean 1 A.U. irradiance for period: 1368.31 W/m2 radiance measured by ACRIM: (i) nor-5 60 75 90 105 120 135 150 165 180 195 malization to a distance of 1 A.U. plus

1980 day the projection of the satellite orbit on thetal solar irradiance at 1 A.U. for the ACRIM channel A sensor, shown as a percentage radial direction to the sun, (ii) correctionabout the weighted mean for the first 153 days of the Solar Maximum Mission. Each for the slow decrease in sensitivity ofrepresents the mean irradiance for the sunlit portion of one orbit. The horizontal channel A between days 62 and 163, (iii)

the tick equals the 60-minute maximum duration of the sunlit portion of an orbit. The temperature-dependent corrections forars through the ticks are the + lo- standard errors of the orbital means. The ticks with tepeature-depend e correci nsB and C designations are the channel B and C measurements during comparisons radiation lost through the aperture and

inel A. These results have been corrected for the small decrease in sensitivity of chan- for the temperature coefficient of resist-ween days 62 and 163. ance of the cavity heating elements (1),

0036-8075/81/0213-0700$00.75/0 Copyright C) 1981 AAAS SCIENCE, VOL. 211, 13 FEBRUARY 1981

on

Aug

ust 1

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

on

Aug

ust 1

7, 2

014

ww

w.s

cien

cem

ag.o

rgD

ownl

oade

d fr

om

(iv) correction for relativisitic radiativeeffects due to the relative velocity of thesun and the satellite (4), and (v) correc-tion for the cosine of the angle betweenACRIM's line of sight and the sun's cen-ter. These corrections are small; only thefirst two exceed 0.01 percent. The stan-dard error of the relative measurementsis frequently as small as 0.001 percent forone-orbit averages. There appear to beno residual atmospheric effects or sensi-tivity of the results to particulate fluxes(5).

In-flight intercomparisons of channelsA, B, and C are summarized in Table 1.The only significant change in detectorsensitivity was for channel A, which isused continuously to monitor the irra-diance. The change between days 62 and163 was -0.0153 percent, measured bythe A/B ratio, and -0.0168 percent, mea-sured by the A/C ratio. This drift stabil-ized by day 163 and no further changewas detected by day 199. A slow degra-dation of channel A's cavity absorptancedue to the effects of solar ultraviolet andparticle fluxes was anticipated. Calibra-tion by channels B and C removes its ef-fect on ACRIM's long-term data base towithin the 0.0015 percent change in theC/B ratio over the 101-day period. Theirradiance record of Fig. I was correctedfor drift in channel A, resulting in theconstant relative differences shown forcomparisons on days 62, 163, and 199.The larger differences between channelsA, B, and C during the first few orbits onday 47 are probably due to absorption ofsolar flux by water vapor. Water ad-sorbed by the instrument before launchvaporizes in orbit at a rate dependent onthe temperature of ACRIM and is re-leased from each sensor's interior at arate dependent on the cumulative shut-ter-open time for that channel.The weighted mean value of the total

solar irradiance at 1 A.U. for the first 153days of the SMM is 1368.31 W/m2, withan uncertainty in the International Sys-tem of Units (SI units) of less than + 0.5percent. This value includes the effectsof all variations observed, including thetwo large, temporary decreases in irra-diance centered at days 99 and 146. Theweighting factors used in forming themean are the inverse squares of the stan-dard errors of the weighted daily meanvalues. The latter are similarly formedfrom the orbital means, using their stan-dard errors as weighting factors. Theseweighted means best represent the in-tegrated irradiance for the 153-day peri-od, with higher-precision observationshaving a proportionally larger influenceon the overall result.A more exact value for the "absolute"

13 FEBRUARY 1981

Table 1. In-flight intercomparisons of ACRIM sensors: ratios of irradiances measured inACRIM channel A, B, and C sensors and their change (A) over the 101-day period between days62 and 163. Channel C was not compared with A and B on day 199. The close agreement of theC/B values over 101 days is a measure of ACRIM's potential for long-term precision.

Day AB AA/B A/C AA/C C/B AC/B(1980) ABA/ /

62 1.000731 1.000526 1.000205-0.000153 -0.000168 0.000015

163 1.000578 1.000358 1.0002190

199 1.000578

accuracy ofACRIM measurements in SIunits will be determined by other experi-ments. The ACR type IV sensors aretheoretically capable of defining the radi-ation scale with + 0.1 percent uncer-tainty in the measurement environmentof the SMM. The first principal in-flightintercomparison, conducted on day 62after allowing 2 weeks for the spacecraftand its instruments to reach equilibrium,shows a maximum disagreement be-tween the three independently calibratedsensors of less than + 0.05 percent fromtheir mean, a result consistent with thetheoretical performance. Whatever thefinal uncertainty proves to be, the exper-iment is compiling a total solar irradiancerecord with the precision, accuracy, andtime resolution required to begin a long-term climatological irradiance data baseand provide quantitative new informa-tion on solar physical processes.The most obvious features of the

ACRIM observations are two large tem-porary decreases in irradiance in earlyApril and late May 1980. These events,

00

c

090a

1370136913681367

-1

-2-3

1370136913681367

s 2001000

200N 100

0

1980 day60 80 100 120 140 160 180

15 29 15 30 15 30 15 30 15 30Feb. Mar. Apr. May June

1980 date

Fig. 2. Correlative plots of (a) total solar irra-diance at I A.U., (b) photometric sunspot in-dex (PSI), (c) irradiance adjusted by the PSI,(d) 280-MHz index, and (e) Zurich sunspotnumber.

also observed by the Nimbus 7/ERB ex-periment (6), appear to be related to thebehavior of specific groups of sunspots.The Zurich sunspot numbers-a conven-tional index of solar activity based on butnot directly proportional to the area ofsunspots-show small peak valuesaround the times of these features (Fig.2e), but this index has little significanceon a radiometric scale. As an alternativeto characterizing spot groups, we de-vised a simple method of combiningmeasured spot areas as a predictor of theirradiance deficit due to sunspots:

PSI = aIAs, 2 ) (1)

where PSI is the photometric sunspotindex: a is a normalization factor;Ai = cos (i, where Oi is the angle fromour normal view of the sun; and Si is spotarea. The dependence on Ai approxi-mates the limb-darkening law for the qui-et photosphere. Spot areas and coordi-nates were taken from the Solar-Geo-physical Data issued by the NationalOceanic and Atmospheric Administra-tion. The conversion from spot area to abolometric quantity requires an estimateof the distribution of effective temper-atures within the spot. We took averagevalues (7) for umbral and penumbral areafractions and temperatures, resulting in anormalization factor of a = 0.33. Thusdefined, PSI gives the estimated flux re-duction in parts per million if the sunspotareas are in units of millionths of the so-lar hemisphere. This model, with its ap-proximations and assumptions (8), canonly provide a preliminary test for rela-tions between active regions and total ir-radiance.The solar irradiance and the PSI index

(Fig. 2, a and b) exhibit a high degree ofcorrelation for the 112 days of datashown-a somewhat surprising result inview of the uncertainties of the correla-tive data and the model approximations.The strength of the correlation derivesfrom the strong signals observed around8 April and 25 May (days 99 and 146).TIhe bulk of these signals was producedby large, recently developed or devel-

701

a

A.-/E.f

A. *?

bd*

are large compared with the length of the18.3 11.6 8.5 6.7 5.6 4.7 4.1 data base and the two big dips, which we

assume to be aperiodic, undoubtedly dis-tort the spectrum. The most interestingfeature is the cutoff of spectral intensityfor periods shorter than about 7 days.A second time series analysis was per-

formed on the PSI-adjusted flux to elimi-nate most of the effects of the two bigdips. The result showed essentially thesame spectrum (but with lower relativesignificance for the spectral peaks) withthe 7-day cutoff.The results of these analyses are typi-

cal of a band-limited noise spectrum.The one significant feature at this point isthe cutoff of spectral power density atabout 7 days. This result is consistentwith the emergence and development of

25.6 14.2 9.8 7.5 6.1 5.1 4.4 sunspot groups and associated facularPeriod (days) areas, occurring at apparently unrelated

spectrum of ACRIM daily mean irradiance record for days 52 through 179. intervals and locations, which take onthe order of 7 days or more to completetheir evolution.

oping spot groups crossing the centralsolar meridian; on 8 April Boulder re-

gions 2370 and 2372 on the sun contrib-uted 80 percent of the PSI, and on 25May regions 2469 and 2470 contributed61 percent. These two large variations inirradiance are obviously related to thepresence of such sunspot groups, andour simple PSI model satisfactorily ex-

plains not only the time profile of theseevents but also, in large part, their ampli-tudes.

Outside the times of these strong sig-nals, the correlation between spot indexand solar irradiance is not as good, andsignificant deviations from the patternexist. The first 35 days of data give no

indication of a correlation. Furthermore,there is a difference in detail between thetwo major dips in irradiance describedabove. The difference signal (Fig. 2c)shows no trace of the dip in early April,but shows a large positive excursion atthe time of the May dip, presumably theeffect of the larger facular componentobserved to be associated with the latter.

Finally, a 5-day drop in solar irra-diance centered on 13 May (day 134)seems to have no counterpart in solar ac-

tivity of the type discussed above. Thepresence of such short-term variations ofirradiance, uncorrelated with obvioussolar activity processes, suggests othersources of solar variability. The list ofpossible sources is a long one (5, 9-12)and their identification has just begun.The other major obvious contributor

to the total solar irradiance is the aggre-

gate of white-light faculae. We would ex-

pect the difference between the irra-diance (Fig. 2a) and the sunspot effectestimated by PSI (Fig. 2b) to show a

702

good correlation with some measure ofthese regions of excess solar emission.The 2800-MHz index (Fig. 2d), a reliableindicator of emission activity, does notshow a high correlation with the adjustedirradiance (Fig. 2c) for the entire dataset. Identification of facular contribu-tions to solar irradiance will requiremore precise independent measures offacular brightness than are now available(13).The apparent relation between the

evolution of large sunspot groups and thetwo large irradiance events in April andMay has a bearing on the nature of con-

vective energy flow in solar active re-

gions. We conclude that, at least in thesetwo cases, there is no direct and immedi-ate energy balance between sunspot defi-cit and facular excess; instead, the sun-

spot-deficit energy is stored or delayedwithin the convection zone through theeffects of the magnetic fields associatedwith the sunspot groups involved (14).Subsequent development of related fac-ulae in these regions may provide themechanism for radiating the stored ener-

gy away. Further study of these and sim-ilar events should provide interesting in-formation about the deep structure of ac-

tive regions.The continuous variability of the

ACRIM irradiance record with smalleramplitudes than those of the big dips isanother interesting feature. Preliminarytime series analyses have been per-

formed on the daily mean values to in-vestigate this variability. The power

spectrum of the irradiance at 1 A.U. fordays 52 through 179 is shown in Fig. 3.We assign little significance to the indi-vidual spectral peaks, since their periods

R. C. WILLSONS. GULKIS, M. JANSSEN

Jet Propulsion Laboratory,California Institute of Technology,Pasadena 91103

H. S. HUDSONCenter for Astrophysics and SpaceSciences, University of California,San Diego 92093

G. A. CHAPMANSan Fernando Observatory, CaliforniaState University, Northridge 91330

References and Notes

1. The SMM spacecraft is in a circular orbit at 575km with a 28.5° inclination. During the orbitalperiod of 96 minutes, the spacecraft spendsabout 60 minutes in full view of the sun.

2. R. C. Willson, Appl. Opt. 18, 179 (1979).3. E. Zalewski, J. Geist, R. C. Willson, Proc. Soc.

Phot.-Opt. Instrum. Eng. 196, 152 (1979).4. R. C. Willson and H. S. Hudson, paper present-

ed at the COSPAR (Committee on Space Re-search) meeting, Budapest, June 1980.

5. _, Astrophys. J. Lett., in press.6. J. R. Hickey, personal communication.7. C. W. Allen, Astrophysical Quantities (Athlone,

London, 1973).8. Spot groups differ in properties and can be de-

scribed only approximately by average values.The behavior of small spots may be quite dif-ferent from that of the large ones that dominatethe signal. The PSI values do not include anyestimate of facular excess emission.

9. P. V. Foukal and J. E. Vernazza, Astrophys. J.Lett. 234, 707 (1979).

10. R. K. Ulrich, Science 190, 619 (1975).11. D. 0. Gough, in The Solar Output and Its Varia-

tions, 0. R. White, Ed. (Colorado AssociatedUniv. Press, Boulder, 1977), p. 95.

12. R. H. Dicke, Nature (London) 276, 676 (1978).13. G. A. Chapman, Astrophys. J. Lett., in press.14. For this conclusion to be at variance with the

observations there would have to be either rapidredistribution of the deficit energy to distant re-gions of the sun less visible or invisible toACRIM or conversion to other forms of energynot detectable by ACRIM.

15. We thank R. Noble and M. Woodward for theirvaluable assistance in data processing and inter-pretation. The SMM/ACRIM experiment is sup-ported by NASA grants NAS-7-100 at Jet Pro-pulsion Laboratory, NSG-5322 at the Universityof Califomia at San Diego, and NSG-5330 at theCalifornia State University, Northridge, Foun-dation.

18 September 1980; revised 24 November 1980

SCIENCE, VOL. 211

0.15

0.12

co

C 0.096

0

o.oo

C 0.0a

.Nco

0.06-0z

0.03

0.00 -128.0

Fig. 3. Power