Earthshine Observations of the Earths Reflectance

GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 9, PAGES 1671-1674, MAY 1, 2001

Earthshine Observations of the Earth's Reflectance

P. R. Goode,1 J. Qiu, V. Yurchyshyn,J. Hickey Big Bear Solar Observatory, New Jersey Institute of Technology,Newark, NJ 07102

M-C. Chu Department of Physics, The Chinese University of Hong Kong, Shatin N.T., Hong Kong

E. Kolbe Departement ffir Physik and Astronomie, Universitgt Basel, Basel, Switzerland

C. T. Brown, S. E. Koonin W. K. Kellogg Radiation Laboratory, California Institute of Technology, Pasadena, CA 91125

Abstract. Regular photometric observations of the moon's low-earth orbit view only a small fraction of the earth's sur- "ashenlight" (earthshine)from the Big Bear Solar Obser- face at once so that many observations must be combined vatory (BBSO) since December 1998 have quantified the to obtain global-scale information. It is also a major techni- earth's opticalreflectance. We find large (,,• 5%) daily vari- cal and logistical challenge to keep calibrated photometers ations in the reflectance due to large-scale weather changes continuously in orbit on decadal timescales. on the other side of the globe. Separately, we find compara- The earthshine, or "ashenlight", is the glow of the "dark" ble hourly variations during the courseof many nights as the part of the lunar disk visible to a nighttime observer. It is earth's rotation changes that portion of the earth in view. sunlight reflected from the earth and retroflected from the Our data imply an average terrestrial albedo of 0.297+0.005, lunar surface, and so offers an alternative route to studying which agrees with that from simulations based upon both the earth's reflectance. Earthshine data are complementary changing snow and ice cover and satellite-derived cloud cover to existing satellite data in that the coverage is instanta- (0.2965:0.002).However, we find seasonalvariations roughly neous and hemispheric in scale. Because the earth's phase twice those of the simulation, with the earth being bright- as seen from the moon is supplementary to that of the moon est in the spring. Our results suggest that long-term earth- seen from the earth (i.e., the earth is nearly "full" when shine observations are a useful monitor of the earth's albedo. the moon is a thin crescent),the instantaneousintensity of Comparison with more limited earthshine observations dur- the earthshine near the new moon samples almost half of ing 1994-1995 show a marginally higher albedo then. the earth. For over a quarter century, beginning in 1926, Danjon and his followers[Danjon, 1928, 1954;Dubois, 1947] performed regular earthshine observations from France. We A central parameter of the earth's climate is the planet's have reinvigorated and modernized this nearly forgotten way albedo, A, the fraction (roughly 0.3) of the incident sun- of measuring the earth's albedo. light that it reflects. Simple power balance considerations Our earthshine coronagraph is a 6-inch refractive tele- suggest the importance of this parameter: in the absence of scope with a cooled AP-7 512x512 pixel CCD camera; our feedbacks,the earth's blackbody temperature of some 255 K fiducial regions comprise about 100 pixels each. We mask would increase by roughly 1 K for a 0.01 decreasein A. Pre- the bright portion of the lunar disk (i.e., "moonshine")with cise measurement of the albedo can reveal variation with the a neutraldensity filter of transmission• 10-5, and correct earth's aspect (land is more reflectivethan water), weather the measured earthshine intensity for the effects of moon- (cloud patterns), and seasons(snow and ice coverand veg- shine scattered in the atmosphere and optical train. For de- etation). tails, seehttp://www. bbso.nj it. edu/Research/Earthshine. The albedo is determined most directly by satellite obser- The earthshine intensity is most conveniently determined vations[Gibson et al., 1990;Stephens, Campbell and Vonder by observingthe specificintensities Ia,b of two nearly dia- Haar, 1981;Rossow and Lacis, 1990]. However,satellites in metrically opposite lunar highland regions (each covering nearly 0.1% of the lunar disk) near the edge of the disk - •Also at W. K. KelloggRadiation Laboratory,California In- one (a) in the earthshine,the other (b) in the moonshine.If stitute of Technology, Pasadena, CA 91125 we define the earth's effectivealbedo, A*, as that of a Lam- bert (perfectlydiffusing) sphere having the samereflectance [Flattd, Koonin and MacDonald,1992], then Copyright2001 by the AmericanGeophysical Union.

Papernumber 2000GL012580. 0094-8276/01/2000GL012580505.00 1671 1672 GOODE ET AL.' EARTHSHINE OBSERVATIONS OF THE EARTH'S REFLECTANCE

Here,/• is the lunar phase(/• is nearly -•r at the new moon, snow and ice cover information, we use simulations from increasesto about/• - 0 at mid-month and is nearly/• - •r the Canadian Center for Climate Modeling and Analysis at the end of the lunar month), fb(/•) is the phasefunction (http'//w•-•. cccma.bc . ec . gc . ca), whichgives the monthly of regionb,/• • •r-/• is the earth'sphase angle, and fL (/•) = mean snow/ice cover for each computationalcell (2.5øx . [(•r- I/• [)cos/•+ sin [/• I]/•r is the Lambertphase function. 2.5ø). (All phasefunctions are definedso that f(0)-l.) The lunar Figure i showsthe observed(left panel) and simulated phasefunction, fa("7), arisesfrom the near retrofiectionof (right panel)effective albedo expressed in astronomicalmag- the earthshineby the moon,where -7• < 1ø is the anglebetween nitudes,[A*] - -2.5 logA*, as determinedfrom our earth- the incident and nearly retrofiected earthshine from region shine observations. This standard astronomical definition a. Rem, Re, Roa, Rob, Res and Rms are the earth-moon impliesthat the larger [A*] is, the smallerthe albedo.Note distance, the radius of the earth, the distance of the observer that [A*] -- 1.31 for A* -- 0.3 and that a 1% changein A* from patches a and b, the earth-sun distance and the moon- correspondsvery nearly to a 0.01 changein [A*]. The A*'s sun distance, respectively. from morning and evening observationsand simulations are, Our differential measurement of the earthshine removes on average, nearly indistinguishable. However, the spread many of the complications associated with varying atmo- about the mean of the observed[A*]'s is appreciablylarger spheric absorption and solar intensity. Lunar images around than that of the simulated ones (respectivevariances are a lunar eclipsedetermine the ratio of the geometrical albedos 0.14 and 0.02 about the fitted curves). While the nightly of the two fiducialregions, pb/p•, while photometricobserva- averagein both panels is relatively constant for larger phase tions during many lunations determine the phase functions angles, it decreasessharply toward the full moon, implying that the earth is relatively more reflective at grazing angles To fully account for the effect of atmosphericextinction, (glint). Note that we haveno reliableobservations for phase observations of both the earthshine and the moonshine are angles below 40ø or above 140ø becauseof scattered moon- made as the airmass varies through the night. The true shine and becauseof short observingperiods, respectively. intensitiesto be used in eqn. (1) are greater than the ob- The average albedo can be determined from our obser- servedintensities by a factor of e•'z. The local airmass, z, is vationsand simulations asA - 2 •r A* . the cosecantof the elevation angle for observationsnear the We haveextrapolated [A*] from 40ø •_ _<180 ø to the full zenith, while near the horizon the curvature of the earth re- range of the integration using a constrained least squares quiresthat we usetabulated values [Kondratev, 1973]. The fit in which [A*] is assumedto vary smoothlywith lunar atmospheric extinction coe•cient, c•, ranges between about phase. This method is used widely and successfullyin helio- 0.1 and 0.3, and on a typical night it is determined with a seismology[Goode, 1995]. We testedit by performingnearly relative precision of 1% for the earthshine and about 0.5% 300 additional simulations, for nights other than the 143 be- for the moonshineby fitting that night's observations.Typ- tween January 1999 and August 2000, so as to cover the en- ically, we find c,e- C,m"" 0.05. tire range from 0ø •_/• _• 180ø for the parts of the earth we To determine the earth's reflectancefrom the earthshine, "see"from BBSO. Fitting these and integrating A*, we find we must know the lunar phasefunctions, f•,b - the moonshine intensity of the fiducial regions as a function of the

lunar phase. However, raw moonshinedata near the quarter 2.0 0.16 moon show as much as a 5% variation from month-to-month 1.5 0.25 at the same lunar phase due to varying atmospheric absorption, the lunar declination, and libration. We have addressed 1.0 0.40 this problem by making three successive,linear adjustments 0.5 0.63 to the observed moonshine intensities using the integrated

crescent of the moon as a standard, and then correcting for 0.0 ,. 1.00 the declination and libration of the moon. After applying these three physically required corrections, the aforemen- -0.5 1.58 ß i . . . i . . . , . . . i .

tioned 5% variation in f•,b is reducedto 0.5%. With these 40 80 120 160 40 80 120 160 adjustments, we estimate that we can measure the nightly Lunar Phase (degrees) Lunar Phase (degrees) effective albedo to a relative precisionof about 1% on clear nights. This precision is comparable to those quoted for Figure 1. Observed(left panel) [A*]'s plotted againstthe ab- satellitealbedo data [Gibsonet al., 1990; Stephens,Camp- solute value of the lunar phase (also indicated by the cartoons bell and VonderHaar, 1981;Rossow and Lacis, 1990]. along the x-axis) from 198 clear observing nights from Decem- Our results are best interpreted by comparing them to ber 1998 through the beginning of September 2000. Simulations simulations. In modeling the reflectance of the earth, we (right panel) are shownfor the 143 nightsfor which there were also contemporaneous cloud cover data. Each data point represents a use scene models developed for the Earth Radiation Bud- nightly average of 0.5-3.0 minute earthshine measurements taken get Experiment(ERBE) observations[$uttles et al., 1988]. once every five minutes with 0.1-5.0 second moonshine observa- The cloud cover data, updated every 6 hours, are a whole tions interspersed. Error bars would be within the symbols. Con- earth composite of visible-light satellite images from WSI- strained least square fits to the data and simulations are shown INTELLICAST (http://•-•w. intellicast. com/). For the by the curves. GOODE ET AL.: EARTHSHINE OBSERVATIONS OF THE EARTH'S REFLECTANCE 1673

earthshine at any moment, but there is a relatively longer Mar 24, 1999, Lunar Phase=-92 Oct 31, 1999, Lunar Phase=+90 observing night. Hence, somewhat more than a third of the earth is covered during a given night, roughly independent of phase. The observations are consistent with the simulations for "•:''"" :•'••:•'"':"•':::'::":•':: •';;:":":::''""'•'•" •!..'•.:•:-'•?.... -'•:;: ...:•::.?-:..:-:•&::-;-?:¾ ,•..' ..:?:'::' ...... •.:,•. •,. :'-:•i•-•.:•:i;;•,•,:.--•;•::•::..--....-:½::.?•? •:"...'•i::i"'....•..."' • ' 24 March 1999, which is one of the nights for which the ':};;:;?.' g. -•.*.'• ..•::i-:-:_--.:.-.-.--:.•--':--::•::.:,"-{?::.."?'.•'' ======"' ø¾':':•'•' .-•,•"•2 agreement is quite good. On 31 October 1999, a more typ- -:•:':::::•<•;;:..-...-..... • .--...... -:..:::: .... -...... -.::;;•:.i..;...... ½::-"- '"...... ?.?: i•: :;:•;;• '.!•: ?'•!i:.:::-'-. ======':•:i;;.... .:-..:..¾•!:•.•..:-•*½• '•...½::::---..•*-:•:'..'.%..•.•:.;½•::.:-:•".-•::::-:'---:.:.:...--:--.. ical night, there is a discrepancyof about 5% relative (or :•-:.i.•?.'•:...e. z::....--.::-::•;•.•:.•::.--'-..-'.--'.-'&• • .;'•j.:.-"go.•j.;•.-;-7• '"j:' •;11ii!i•ii•:;: ' :- -:--'::'.':: ::•-::-:'.-:: -:-•:--'-'--':-::!-:½...... •-' "*...... • '?.•.:•:.• ß'•.-:.-'.'•'".:r•'•: 0.015 absolute)in the effectivealbedo. By scalingvariations of the simulated[A*]'s to thoseof the simulatedalbedo, we estimatethat measuring[A*] to an accuracyof 0.01 deter- 0.32 0.32 mines the global albedo, A, to 0.002. As the cloud cover is fairly high at geographical latitudes o.3 '. 0.30 above 45ø, the short timescale variations in reflectanceare 0.30 due to irregularities in the fractional cloud cover at low lat- 0.29 0,28 itudes. In the left panel of Figure 2, one seesthat the observed and calculated maxima in A* at 7:00 UT are due 0.28 0.26 to a relative cloud excessover the Far East, while the calculated local minimum in A* at 11:00 UT arises from the 24 20 16 12 8 4 0 24 20 16 12 8 4 0 cloudless area above India and the Arabian Sea. In the left Universal Time Universal Time panel, there is a nearly 10% change in A* through the pe- riod of the observations. An offset between the observed and Figure 2. In the top panels, the bright areas highlight those calculated effective albedos is evident in the right panel; it parts of the earth that are the sourceof the earthshine (i.e., are simultaneously in the sunshine and are visible from the moon typifies the general result that the observed night-to-night at sometime during the observations);the satellite-derivedWSI variations are greater than those of the simulations. cloud cover data is shown in gray-scale, with brighter areas in- The seasonal variation of the earth's reflectance as deter- dicating greater cloud cover. For 31 October 1999, note that mined from our earthshine observations is shown in Figure the northernmost regions are not sunlit and the southernmost re- 3. As [A*] is a strongfunction of lunar phase,we consider gions do not contribute to earthshine because the moon is fairly the anomaly of each nightly average relative to the best- far north in the sky. Simulations of 400+ nights show that not fit mean trend of all the observations. These anomalies, detecting the pole in this way reduces A by about 0.005 because of the nearly invariant relative brightness of the polar regions. which are shown as fractional changesin A*, carry informa- The boxes indicate the longitudes of maximal contribution to the tion about weather, climate and surface type. There is the earthshineat the UniversalTime (UT) shown. The solid boxesin expected summer minimum, with fall and spring maxima, the lower panels show the observed effective albedo as a function and the earth is brightest in the spring, when it is also the of time (note that the time axis is reversed),while the solid line cloudiest. indicates the simulated A*. The observed seasonal variation is correlated with the corresponding simulations. However, the observations show about twice as much variability, with the differences being A = 0.294. We note that the fit to the full-range of sim- greatest at the extremes. The seasonal variation of albedo ulations is indistinguishable from that shown in the right is the origin of the relatively greater spread among the ob- panel of Figure 1. In all cases,the extrapolationof [A*] served[A*]'s in Figure 1, and mostlikely, the offsetbetween for 101> 140ø contributesabout 0.07 to A, while the region observationsand simulationstypified in the lower right panel for 101< 40ø contributesabout 0.01. From this, we ob- of Figure 2. The muted seasonal amplitude of the simula- tain A = 0.297q- 0.005 for the observations(left panel) and tions may well be associatedwith the coarse binning of the A = 0.296 q- 0.002 for the simulations(right panel) for the scenemodels and/or the useof simulatedsnow and ice cover. two-thirds of the earth we observe. The uncertainties here Some sense of the longer term trend in the earth's re- arise from the large night-to-night and seasonalvariations. flectance can be had by comparing these data with 72 nights Our observations have also determined both the variation of earthshine observationstaken from January 1994 through of the instantaneousA* during each night, and the variation July 1995. (These latter data favor no particular season.) of the nightly average A* throughout the seasons.In Figure The average deviation of the nightly means of these data 2, we show one set of evening and one of morning obser- from the gray curve in the left panel of Fig. 1 translates to vations near the waxing and waning quarter moon, respec- a 2.5%q-1.5% greater A*. Most of the deviation is presentin tively, as well as the corresponding simulations. The top the morning data. The correspondingvalue is 0.0% q- 0.9% panels show that a significant part of the earth contributes for the January 1999 through July 2000 data. We note that to the earthshine during the observations. For larger lunar 1994-1995 was during the most recent solar activity mini- phases,more of the earth contributes instantaneouslyto the mum, and that our result is consistent with the hypothesis, earthshine, but the length of the observations is reduced. based on cloud cover data, that the earth's reflectance de- For smaller lunar phases,less of the earth contributesto the creaseswith increasingsolar activity [$vensrnarkand Friis- 1674 GOODE ET AL.' EARTHSHINE OBSERVATIONS OF THE EARTH'S REFLECTANCE

http-//cloud.ucsd.edu/missions/triana.html) are also of clear interest to intercalibrate the methods and validate < 10 the scene models used.

Acknowledgments. We wish to thank those involved E in earlier earthshine work at BBSO, especially J. Bruls, G. Ey- o 0 chaner, M. Haggerty, J. Maron and H. Zirin. This research was supportedin part by grants from NASA (NAGS-7454) and the Western Regional Center of the National Institute for Global En- vironmental Change. o I References

i i i i i i Danjon, A., Recherches sur la photom•trie de la lumi•re cendr•e et l'albedo de la terre, Ann. Obs. $trasbourg, 2, 165-180, 1928. 1/99 4/99 7/99 10/99 1/00 4/00 7/00 Danjon, A., Albedo, color, and polarization of the earth, in The Earth as a Planet, edited by G.P. Kuiper, pp.726-738, Univer- Figure 3. Seasonalanomalies in the effectivealbedo, A*. From sity of Chicago Press, 1954. December 1998 through the beginning of August 2000, there are Dubois, J., Le photom•tre k ceil-de-chat et ses applicatons, Ciel 170 nights of observationswith 22 nights in the earliest bin, 23 et Terre, 58, 350-361, 1942. nights in the next bin and 15 nights in each of the remaining 9 Dubois, J., Sur l'albedo de la terre, Bull. Astron., 13, 193-196,- bins; the ex.tra nights in the two earliest bins serve to reduce the 1947. uncertainties. From January 1999 onward, there are 143 nights Gibson, G., et al., Characteristics of the earth's radiation budget for which we have both observations and contemporaneous satel- derived from the first year of data from the Earth Radiation lite cloud cover data, which have been averaged in 13 bins with 11 Budget Experiment, $PIE, 1299, 253-256, 1990. nights in each. The x's show the mean of the observations,with Flatt•, S., Koonin, S.E., and MacDonald, G., Global Change and the vertical bars being the standard deviation of the mean. The the Dark of the Moon, Report No. JSR-91-315, The MITRE size of the latter stems from the large night-to-night variations Corporation, McLean, VA, 32 pp., 1992. in the cloud cover, rather than from uncertainties in the obser- Goode, P.R., Internal Rotation and Structure of the Sun, ESA vations. The horizontal bars indicate the temporal span of each SP-376, 121-136, 1995. average. The diamonds indicate the corresponding simulation re- Kondratev, K., in Radiation characteristics of the atmosphere sults. For the 143 nights for which we have observations and and earth's surface, Translated from Russian by Vijay Pan- simulations, the anomalies relative to the separate fitted curves dit, General editor: V.S. Kothekar, Published for NASA by are correlated at the 95% confidence level. Amerind Pub. Co., New Delhi, 1973. Rossow,W.B. and Lacis, A.A., J. Climate 3, 1204-1253, 1990. Stephens, G., Campbell, G. and Vonder Haar, T., Earth radiation budget, J. Geophys. Res., 86, 9739-9760, 1981. Suttles, J.T. et al., Angular radiation models for earth- Christensen,1997]. On the other hand, the differencecould atmosphere systems. Vol I: shortwave radiation, NASA Ref- be associated with the combined effect of the E1 Nifio that erence Publication RP-118•, 26 pp., 1988. These scene models occurred during the earlier observations and the La Nifia give local reflectance as a function of illumination and reflec- (http://www. ogp.noaa. gov/enso) during the latter. tion directions, fractional cloud cover, and surface type. Our demonstration that the earthshine can be observed Svensmark, H. and Friis-Christensen, E., Variation of cosmic ray with sufficient precision to derive meaningful information flux and global cloud coverage- a missing link in solar-climate relationships, J. Arm. and Solar-Terrest. Phys., 59, 1225-1232, about the earth's optical reflectance suggestsseveral fur- 1997. ther investigations. A multi-station network is now being established that will allow continuous global coverage P. R. Goode, J. Qiu, V. Yurchyshyn, and J. Hickey, Big Bear during appropriate lunar phases. With the increased pre- Solar Observatory, New Jersey Institute of Technology, Newark, cision and coverage from the network, we will be able to NJ 07102. (e-mail: [email protected]) monitor long-term variations in cloud cover and aerosols C. T. Brown and S. E. Koonin, W. K. Kellogg Radiation Lab- with sufficient accuracy to bear on climate change. Net- oratory, California Institute of Technology, Pasadena, CA 91125. M-C. Chu, Department of Physics, The Chinese University of work observations of the reflectance in the coming years Hong Kong, Shatin N.T., Hong Kong. will also quantify ENSO events, volcanic eruptions, and E. Kolbe, Departement fiir Physik and Astronomie, Universit/it decadal scale variability, including a more precise explo- Basel, Basel, Switzerland. ration of any solar cycle variation. Comparison of earth- (ReceivedNovember 2, 2000; revisedMarch 8, 2001; shine results with satellite observations(see, for example acceptedMarch 14, 2001.)