REVIEWS
OF GEOPHYSICS
AND
SPACE PHYSICS,
VOL.
18, NO. 3, PAGES 647-658, AUGUST
1980
Evidence in the Auroral Record for Secular Solar Variability
GEORGE
L.
SISCOE
Departmentof Atmospheric
Sciences,Universityof California,LosAngeles,California90024
The historicalrecordof auroraeis continuousand usefullydensefor at leastthe last 2000 years.Revival of interestin the secularvariability in solar activity motivatesa review of the auroral record.The
existenceof secularvariationsin the auroraloccurrencefrequencyhasbeenknown sincethe early 1700's,
includingthe existenceof a significantattenuationof auroralactivityduringthe Maunder Minimum. Investigationof secularvariationsprior to the Maunder Minimum is now possiblebasedon six auroral
catalogsthat have beenpublishedwithin the last 20 years.The catalogscoverthe time period from the
fifth centuryB.C. to the seventeenthcenturyA.D. and combineboth oriental and Europeanobservations. Featurescorrespondingto the previouslyrecognizedMedieval M'mimum, Medieval Maximum,
and the SpCirer
Minimum are clearlyevidentin bothorientaland Europeanrecords.The globalsynchronicity of anomaliesin the auroraloccurrence
frequencyis usedto arguethat they are causedby changes
in the level or stateof solaractivity.The combinedcatalogsprovidea sufficientnumberof eventsin the
Middle Agesto resolvea quasi-80-year
periodicityin the recordedauroraloccurrence
frequency.Alsoin
the unusuallyrich intervalsof the Middle Ages,clear quasi-10-yearperiodicitiesappearin the recorded
occurrence
frequencywaveform. Theseare mostreasonablyinterpretedasmanifestations
of the 1l-year
solar cycle and indicate that the solar cyclewas then operative.
EARLY
RECOGNITION
AURORAL
OF LONG-TERM
VARIATIONS
In 1716, for virtually the first time, the Royal Society and
the Acad6mie
des Sciences carried articles on the aurora bo-
historical behavior of the aurora [von Leibnitz, 1710] led him
to the Saxon Chronicles,in which he found earlier evidence
for long-termvariability in recordsfrom the tenth century.
To gatherdata on the puzzlingtemporalvariationsof aurorae, the causeof which wasitself the subjectof widely diver-
realis in their journals. Both societieswere then a half century
gentspeculation
at that time, many investigators
beganto obold. However, merely 34 yearslater the PhilosophicalTransacserveand record auroral eventssystematically.Others turned
tions had recorded 200 observations of aurorae, and the M•mto historical sourcesto extend the data record into the past.
oires de l'Acad•mie a similar number. The reason for the late
There
soon resulted a number of catalogs,listing contempoand simultaneous debut was the return of the aurora to the
latitudes of London and Paris, in or near which most of the
societies' members
lived.
The auroral eventsof the year 1716most clearly announced
that the prolongedsolar and auroral calm that we now call the
Maunder Minimum [Eddy, 1976a]had ended. But the onsetof
renewedauroral activity was noted already in the previoussolar cycle.In 1707an aurora was seenin Berlin and recordedin
the journal of the Berlin Academy. Curiously, in New England, which is closer to the auroral zone than is London, Paris,
rary and historicalaurorae.By the time that Jean Jacques
Dortous de Mairan publishedhis landmark treatiseon the
aurora in 1733, he had accumulated a sufficient record to
draw two important conclusions:
the auroral occurrencefre-
quencyhadincreased
suddenlyin 1716andhadremainedessentially Constantsince then, and there were a number of
timesin the pastwhen auroral occurrences
had resumedafter
a long absence[de Mairan, 1733].He identified22 suchinstances in the interval 500 A.D. to 1731 and referred to them
as resumptions(reprises).
or Berlin, the aurora returned suddenly in 1719. ContempoThe aurora continuedfrequentlyto visit the populatedlatirary accountsput the first recorded appearanceof an aurora
tudesof Europe,and doubt soonwasexpressed
that suchregin Italy in the 1720's.
ular behavior could ever have been very different. In the secThe return of aurorae to low latitudes ushered in the birth
ond edition of his treatise, published in 1754, de Mairan
of post-Renaissanceauroral studiesand causedearly investivigorouslydefendedthe realityof significanttemporalvariagatorsto look on variability itself over time periods from dections in the •/uroral occurrencefrequencyand, in particular,
ades to centuries as a characteristic of the aurora. The three
men who reportedon the renewal of auroral activity to the societies of London, Paris, and Berlin searchedhistorical records
for exampleswith which to compare it. Contemporary litera-
ture and memory provided no ready examples,although later
searcheswould uncover a number of weaker displays.
In England the reporter was the astronomerEdmund Halley, who referred to several sixteenth-centurydisplays. The
previous comparable display in England that he found occurred more than 140 years earlier, in 1574. In France it was
the astronomer Maraldi, who cited the 1621 exhibition made
the reality of the auroral calm preceding1716.Doubt on the
subjectvanishedwhenthe auroraagainwent into declinefor
a periodof about33 years,between1792and 1826,at a time
when careful, routine observationsof it were being made in
Europe and America.
The subsequentresumptionof the phenomenonin 1827
stimulatednew attemptsto discoverempiricallythe law governingits long-termreturnsand departures.ChristopherHansteenin 1831 inferred a 95-year period in the returns of the
aurora[Hansteen,1831].His resultwasbasedon an identification of 24 suchreturnsin an auroral recordextendingback to
famous by descriptionsof it in the works of Pierre Gassendi.
502 B.C. Denison Olmsted in an 1856 article concluded that
The reporter in Germany was the philosopher-mathematician
Gottfried Wilhelm von Leibnitz, whose researches into the
its greatreturnsoccurat intervalsof from 60 to 65 yearsand
last from 20 to 25 years [Olmsted, 18561.
The articles
by Joseph
Lovering
[Lovering,
1860,1868]and
Copyright ¸ 1980 by the American GeophysicalUnion.
Papernumber80R0212.
0034-6853/80/080R-0212506.00
647
648
SISCOE: HISTORICAL AURORAL EVIDENCE FOR SOLAR VARIABILITY
SUNSPOT
--
NUMBERS
To appreciate the extent to which auroral variations reflect
solarvariations,it is useful to begin with a few post-Maunder
Minimum examples,for which solar and auroral recordscan
be compared directly.
NUMBER OF DAYS WITH AURORAE
• AURORAE
IN
NORWAY
(after
Tromholt)
150
THE
POST-MAUNDER
MINIMUM
RECORD
Two independentattemptsto demonstratesystematically
AURORAE SOUTH OF õ4 • GEOMAGNETIC
LATITUDE
the correspondencebetween aurora and sunspotvariations
140•- /% INU.S.
I• EUROPE
{after
Loomis)
[•
are shown in Figure 1. The top panel displaysauroral data
I00
/•
•
•
from the Scandinaviansectornorth of 55ø geographiclati60
tude. Isoplethsof auroral occurrencefrequency(so-calledisochasms)drawn on a globe of the earth form approximate circlescenteredroughly on the geomagneticpoles.Auroral data
YEAR, A.D.
Fig. 1. (Top) The numberof daysper yearon whch auroraewere are therefore often expressedin terms of latitudes referenced
recordedin Noway. The figurewascomposed
by Harang[1951]from to the geomagneticpoles,so-calledgeomagneticlatitudes.In
data compiledby Tromholt[1902].•e panel relatesto the region the Scandinaviansector,55ø geographiclatitude is also appolewardof approximately
55ø geomagnetic
latitude.(Bottom)S•iproximately 55o geomagneticlatitude.
lar data for eventsrecordedin •efica and Europeequatomardof
The data from which the auroral curve in the lower panel
54ø geomagnetic
latitude.The figurewascompos• by Loomis[1873].
was drawn were compiled by Loomis from eventsrecorded in
both Europe and America. To excludehigh-latitudedata, he
the book by Hermann Fritz [Fritz, 1881] reviewed the subject correctly chose a line of constant auroral occurrence frethoroughly. Lovering demonstratedthe close similarity be- quencyconnectingboth continentsas a northernboundaryto
tween auroral variations recorded in Europe and in America. the observations.He picked the isochasmpassingthrough the
Fritz and Elias Loomis [Loomis, 1873] in the United States northernborderof Massachusetts,
which is approximatelythe
separatelyestablishedthat the variations in the frequency and 54ø geomagneticparallel. The two panelsin the figure thereintensity of the aurora conform closely to variations in solar fore nicely complementeach other, giving separatelythe temspottedness.At this time the I 1-year sunspotcycle recently poral behavior of the occurrencefrequency of high-latitude
had been recognized. Fritz advocated a 55.5-year period for and mid-latitude aurorae.
the secularvariation, a period that Rudolf Wolf had already
One seesin both panelsthe unmistakabletendencyof the
inferred from his collection of sunspotdata, composedof five auroral curvesto conform to the 11-year and secularvariaof the 11-year sunspotcycles. The English astronomer John tionsin the sunspotnumbercurves.(The two sunspotnumber
Herschel, commenting on the importance of the work of Wolf
curvesare slightly different becausethe top panel was conand Fritz, noted that the secular period of 55 years suits the structedabout 70 years after the bottom one and sunspot
auroral observationsbetter than does the period of 65 years numberscontinuedto be refinedin the interim.) The auroral
[Herschel,1864].Loomisfavored a period of around 58 years.
curve in the lower panel revealsthe 11-yearcyclemore disIn summary, prior to the twentieth century, long-term au- tinctly. Also, the minimum in the secular variation centered
roral variations were well established,and their investigation around1810is morepronounced
in it than in eitherthe highwas an important subjectunder the generalheadingof auroral latitude auroral curve or the sunspotnumbercurve.One noresearch.However, to the extent that the aim of the investiga- ticesthat in the low-latitudeauroralcurvethe amplitudeof
tion was to determine the law of the variability it was not par- the secularvariationin generalexceeds
that of the I 1-year
ticularly successful.The 11-year variation was first discovered cycle.
in the sunspotdata, although it was then quickly found to be
The fact that the low-latitude auroral curve is smoother and
presen•in the auroral data also. The attempt to determine the itsvariationsmoredistinctthanthe high-latitude
curvemight
period of the secularvariation, which had yielded sucha vari- reflect an actual change in the behavior of the aurora as a
ety of results, was not long satisfiedwith the answer of 55.5 function of latitude. More probably,it reflectsthe fact that a
years. After further collection and examination of sunspot largergeographicalareacontributedto the low-latitudecurve,
data, Wolf finally favored a secularperiod of 83 years, and a and so it is lesssubjectto aberrations
resulting,for example,
roughly 80-year period was subsequentlyassignedto the au- from year-to-yeardifferencesin regionalcloudiness.Sensitivroral data [Schove,1955; Gleissberg,1965; Link, 1968].
ity of the number of auroraerecordedto regionalinfluences
In addition to increasingthe suggestedperiod for the secu- probablyalsoaccounts
for the factthat the auroralfrequency
lar variation to roughly 80 years, twentieth-centuryresearch curvepresentedby Angot [1897,p. 97], basedon entriesin the
led to suggestionsof variations characterizedby longer peri- auroralcatalogof Fritz [1873]from centralEuropesouthof
ods, namely 200 years [Schove, 1955] and 400 years [Link, 55o latitude, reproducesthe sunspotvariationsover the inter1968]. Attributing to thesevariations the property of periodic- val coveredin Figure 1 lesswell than either curve shownthere
ity is, however, lesswarranted than in the caseof the 80-year and evenmorepoorlyoverthe preceding
portionof the eighcycle. The number of eventsin the pre-eighteenth-centuryau- teenth century. It should be noted, however, that Fritz was
roral record has continued to grow as the result of historical able to identify the 11-yearmaxima and minima in the ausearchesin the twentieth century with additions from eastern roral record over the entire interval on the basis of anecdotal
and westernsources.At the presenttime the longest-termvar- material in the auroral accounts themselves. Such accounts iniations emerge as a clear signal simply by plotting the re- elude information on the brightnessand animation of discordedauroral frequencyas a functionof time. There are now plays,which alsoundergothe 11-yearoscillation.
enough eventsin the aurorally rich intervalsto seedistinct 11Regional influencesare clearly illustratedin Figure 2,
year cycles.
whichcoversthe intervalfromthe endof that in Figure1 up
SISCOE: HISTORICAL AURORAL EVIDENCE FOR SOLAR VARIABILITY
160
649
sistentcycle. All three data recordsare needed to see that the
cycle is in fact continuous.
In summary, both the 11-year cycle and the secularvariation are detectable in time plots of auroral occurrence frequency. The occurrencefrequencyof aurorac seen south of
54ø geomagneticlatitude appearsto be more sensitiveto the
secularvariation than the sunspotnumber itself. The solar-related time variations
can be seen in data records from favor-
ably located sitesand countries,such as Yerkes Observatory
and Norway, but more generally, regional influenceswill degrade the signal. The signal tends to improve with the size of
the region contributingto it. Also, it is apparent from Figure 2
that a signal cannot be perceivedover an interval in a record
if the average auroral occurrencefrequency in that interval
ERE
40
falls below a certain number.
THE MAUNDER
I0
o
,
,
I
GERMANY
,
I
I
IN THE AURORAL
RECORD
The imprint left by the Maunder Minimum in the auroral
record has been describedmany times sincethe original articles appeared announcingthe resumptionof auroral activity
in 1707 and 1716. The primary subjectof the tenth Eclaircissement added by de Mairan to the secondedition of his Trait• is
the documentation of the preceding auroral calm. Fritz repeats and augmentsthis material in the fifth chapter of his
book on the aurora. The two articles by Lovering mentioned
above provide fully corroborativetestimony from the annals
of colonial America. The British and European data again
were re-collectedand supplementedin an article in the October 1886 issueof The EdinburghReview.The auroral evidence
BLUE
H•ILLOBS.
40-
MINIMUM
I
I
for the Maunder
Minimum
had been well reviewed
in these
and other works prior to Maunder's [1894] article. Never30
theless,his article provoked in the following issuea fresh, but
brief, recounting of the concomitant auroral data by Clerke
I
I
20
[1894]. In the present decade the vigorous revival of Maunder's claim led by Eddy has once more surfaced auroral inI
I
I
formation pertaining to that time period [Eddy, 1976a]. Also,
the early colonial auroral record has again been resurveyed
"
I
and discussed[Mendillo and Keady, 1976; Eather, 1980].
1880
90
1900
I0
Except for the articles by Eddy the evidence advanced is
YEAR
anecdotal in form. The argument proceedsby comparing
Fig. 2. Yearly averageWolf sunspotnumber (top panel) and the
number of days per year on which aurorae were recordedat Yerkes statementsabout auroral frequency,brilliance, and animation
Observatory [Meinel et al., 1954], at Blue Hill Observatory [Stetson made by observersbefore, during, and after the interval in
and Brooks, 1942], and in Germany [Schrdder,1972].
question.To complete the argument, the competence,veracity, and interest of the observersare established.To this reviewer the reality of the claimed interval of auroral calm has
to 1965. The two observatorydata panels show results ob- been effectively proven in this way, and the important next
tained from systematic,routine observationsfrom single loca- stepis to quantify its duration and amplitudefor comparison
tions. The observatoriesare very nearly on the samegeomag- with other 'anomalous' intervals, such as the one centered
netic latitude. The much larger number of aurorae recorded at around 1810. Space limitations prohibit recalling here all of
Yerkes Observatoryis probably a consequenceof the fact that the collectedanecdotalmaterial. A few examplesare in order,
the view northward from that site is uncontaminatedby city however, to illustrate the main points that have been made.
lights, whereas Blue Hill Observatory is a few miles south of
The infrequency and feeblenessof the general run of auBoston.An eclecticcatalog of unspecifiedsourcesof aurorae rorae from the mid- 1600'sto 1716 is explicitly documented.
in Germany was used to constructthe bottom data panel. It That the abrupt onset of low-latitude aurorae forced Halley,
probably approximatesmore closelythe usual type of catalog von Leibnitz, and Maraldi to search historical records for examples for comparisonhas already been mentioned. Maraldi
of past aurorae compiled from many sources.
Given only a single auroral data panel, one would see a [1716] beginshis article, 'We have observeda rare and curious
suggestionof an 1l-year cycle but be left in doubt about its phenomenon,which appeared in the month of April in this
permanenceand stability.This is true even for the Yerkes Ob- year 1716.' The German astronomer Gottfried Kirch, who
servatoryrecord, which is certainly the best of the three, but also observedthe 1707 event in Berlin, referred to it as an 'unthe evidencefor the first cycle and the last two cycleswould usual manner of phenomenon' [Kirch, 1710]. The event
not convincea critic who doubtedthe reality of a claimed per- prompted him and his son, Christfried, to begin a catalog of
I
650
SISCOE: HISTORICAL
AURORAL
EVIDENCE
FOR SOLAR VARIABILITY
the aurorae that they observed.Twenty-nine years later it
contained 106 entries,or an averageof nearly four per year
[de Mairan, 1754]. It contains no entries between 1707 and
1716. Christfried Kirch was further inspired by the peculiar
temporal behavior of the phenomenonto compileone of the
earliest catalogsof historical aurorae.
From the observatory at Paris the astronomer Giovanni
DomenicoCassinihad alreadydetectedthe gap in the ringsof
cludedactually more information on auroral phenomenathan
does such an educationtoday. The reasonis that the subject
of meteorologywas then an integral part of a scientificeducation, and in a tradition beginningwith Aristotle, treatmentsof
meteorologyinvariably discussedigneousor fiery meteors,in
which the many speciesof aurorae were enumerated. For example, the second book of Gassendi'sPhysicks[Gassendi,
1658], a standardtextbook during the last half of the sevenSaturn that bears his name and four of its satellites before he
teenth century, was devotedto meteorsgenerally, and the sevdiscoveredthe zodiacal light in 1683 [Cassini, 1666-1699]. enth chapterto the aurora borealisexclusively.
From that year he watchedregularlyfor faint, dispersedlumiThe most popular book on meteorologyin the English lannosityin the heavensto add to the known phenomenologyof guage in the seventeenthcentury was written by William
such occurrences.He recordspossibleauroral glowsonly on Fulke, Doctor of Divinity (who was unfortunately referred to
severalnightsin the summerof 1687.In the languageof mod- by Halley merely as W.F.D.D. and so has since gone unem auroraldescriptionsthey took the form of a quiet, homog- named in auroral literature). The full title of his book [Fulke,
eneousarc, stretchedalong and a little above the northern ho1571] is so long that this excusefor not quoting it here uses
rizon, varying on different nightsfrom feeble to very white in fewer words by more than half than are in the title itself. It is
intensity. He compared his observationswith the account of sometimesreferred to by the first three words in the title, 'A
the 1621 display seenby Gassendiin southernFrance and deGoodlyGallerye.'It was first printed in 1571 with subsequent
scribed by him as being active. Cassini concludedthat active editions in 1634, 1639, 1640, 1654 (the edition referred to by
aurorae must be 'a rare meteor.' (In old usagethe word 'me- Halley), 1655, and 1670.It has thus gonethroughmore editeor' was a generic term whose scopeis roughly captured in tions than probably any other book in English on meteorthe broadestmeaningof the word 'meteorology.')Cassinicon- ology up to the present time. After an introductory section
tinued his work at the observatoryuntil 1712but reportedno discussingcausesthe book beginswith a descriptionof 10 difadditional auroral occurrences.
ferent speciesof aurorae, among them, in modern terminolIt is perhapsjustified to note here for comparisonthe re- ogy, homogeneousbands, rays, corona, pulsating patches,
mark made by Gregory of Tours late in the sixth century con- flaming aurora, and fragmentary arcs.
Studentsat Harvard College in the seventeenthcentury also
cerning the frequency of aurorae observedin that region of
France, south of Paris: 'Portentsappeared. Rays of light were studied Gassendi and, in addition, Charles Morton's Comseenin the northern sky, although, indeed, this happensoften' pendiumPhysicae,which was compiled around 1680 [Morton,
[Gregoryof Tours, 1974]. Still further south and closerin time 1940]. Morton was educatedat Oxford at the time of the forto the commencement of the Maunder Minimum, auroral fre-
mationof theOxfordPhilosophical
Sodety(1648-1649),
the
quencyand brilliancein Italy were suchthat Mario Guiducci,
in a discoursedelivered to the Florentine Academy in 1619,
could refer to them in introducing the subject of the aurora,
'And I know, Academicians,that many of you will have seen
more than once the sky at nighttime illuminated in its northem partsin sucha way that its brightnessyieldsnothing to the
brightest dawn and closely rivals the sun' [Guiducci, 1960].
Guiducci's statement (probably written by Galileo) appears
unremarkable when compared with the post-Maunder Mini-
nucleusfrom whichthe RoyalSocietyerderged
in 1662.The
contentsof the CompendiumPhysicaetherefore reflect a curriculum approvedby the highestlevel of scientificauthority at
that time. When the book turns inevitably to the subject of
meteorology, we see again the prominence of auroral phenomena in the chapter headings:Chapter 12, Of Fiery Meteors; Chapter 13, Of Comets;Chapter 14, Of Aery Meteors
(i.e., winds, etc.); Chapter 15, Of Watery Meteors;and Chapter 16, Of Appearing Meteors (e.g., halos and rainbows). The
mum situation in which at least 52 aurorae were observed in
chapter on fiery meteors includes descriptionsof homogeItaly in the 24 years after 1727 [Lovering,1860]. In describing neousbands and rayed arcs(modern terminology).
an aurora seenin Italy in 1737, Zanotti [ 1738] says:'The AurIt is evident that the paucity of auroral reports during the
ora Borealis,which was formerly a rare phenomenon,and al- Maunder Minimum doesnot reflect a lack of a familiar precemost unknown in our climate, is now becomevery frequent.' dent. All of the early reports after the resumptionof displays
Assertionsof high auroral frequency stand in marked con- referred to Gassendi's account. Nor does it reflect a lack of
trast to the statements of Cassini, Kirch, and Maraldi and to available descriptivecategories,although it is true that by the
an introductorycommentmade by Halley in his article on the time of the resumption the available categoriesseemed oldaurora of 1716 [Halley, 1716]. He referred there to the auroral fashioned and new ones were soon invented. Also, bear in
phenomenonas 'a matter so uncommon,never beforeseenby mind that the Maunder Minimum occurred at a time of rapmy self.' Halley was at this time 60 yearsof age and possessed idly growingscientificcuriosityabout natural phenomenaand
an astronomer'sinterestin night sky happenings,both meteor- at a time when the means had been establishedfor simply
ological and astronomical.He declaredthat of 'all the several publishingsucha thing as an auroral observation.The flood
sorts of meteors I remember to have hitherto heard or read of,
of publishedauroral accountsfollowing 1716 is a measureof
this was the only one I had not as yet seen,and of which I be- the level to which generalscientificinterestand the accessibility of publicationoutletshad risenby the time of the resumpgan to despair.'
Halley's remark raisesthe questionof the preparednessof tion.
The evidence related above refers to latitudes below 55 ø
the observersof that time to recognizeand record in scientific
terms an aurora if one should occur. It is notorious that
and thus to the region that we have seenis particularly sensipeople often fail to see or respondto that for which they are tive to the secularvariation. To obtain a truer soundingof the
unprepared.However, the generalscientificeducationin post- depth of auroral minimum, data from latitudes closerto the
RenaissanceEurope and Britain, and in colonial America in- auroral zone must be examined. Unfortunately, the informa-
SISCOE: HISTORICAL
AURORAL
EVIDENCE
FOR SOLAR VARIABILITY
651
self lists 60 occurrences between 1645 and 1698 in the ex-
tended catalog included in the secondedition of his treatise.
There may even have been a high-latitudeband of restricted
latitudinal extent in which aurorae were fairly common. This
interestingand important possibilitycomesto light in the account of the voyage of Captain Jean Wood to Norway and
NovayaZemlyain 1676[Wood,1705].He statestherethat au,',•,'•,•'-•,•','•'frequently•'•',, i, Greenland •,,,,• ,•,,•,, ,•,.,.•,•i,•,_
ally in Iceland and Norway. In addition,Fritz quotesthe aux v•
,14.v
.. v•
v
•vv...i..i.....
,14......•,,. v...i....j
thor of a book on Greenland fisheries,who during the greatest
depth of the auroralminimumas seenat mid-latitudes(see
discussion
below) wrote that 'The northernlightsshinealone
only for thosewho live in the regionof the arcticcircle.'
Eddyhaslargelyledtheeffortto breakawayfiorareliance
40
30
20
I0
1400
1500
1600
1700
Fig. 3. (a) Data field composedby Eddy [1976a] from the auroral
catalogof Fritz [1873] showingthe number of days on which aurorae
were recordedper decadein Europe southof the polar circle. The interval shownincludesthe Maunder M'mimum,which appearsas a gap
in a curve that risessteeplyfrom a low, pre-Renaissanceto a high,
on purely anecdotalmaterial and to quantify the Maunder
Minimum as a temporal feature in physicalsolar-terrestrial
parameters.
With regardto the auroraldata,whichare our interesthere, he has presenteda plot of the numberof reported
northernhemisphereauroraeper decadesouthof 60ø latitude
listedin the catalogof Fritz [1873].The plot, reproducedhere
as Figure 3a, showsthe Maunder Minimum as a deep recessionin a rapidly (on a scaleof centuries)risingcurve,whose
generalsweeptracesthe progress
of printingand scientificenquiry from the Middle Agesto recenttimes.Otheractualsolar
featureslie half hidden in the curve prior to 1500,and we will
return to these in the next section.
The catalog of Fritz has been criticized for containing a
number of inaccurate entries. Two more recent and more ac-
post-Renaissance
level.(b) Data field showingthe numberof aurorae curatecatalogshave been compiledby Link [1962, 1964].The
per decadecontainedin the more recentcompilationof Link [1964]. first covers the interval 626 B.C. to 1600 A.D., and the second
The attenuated auroral occurrencefrequency during the Maunder
Minimum
is evident in this data set also.
continuesto 1700A.D. Reportsper decadefrom 1400to 1700
basedon the Link catalogsare shownin Figure 3b. Although
details in the two curves are different, the main features are
the same.In particular,the auroral calm coincidentwith the
tion thus far uncoveredon seventeenth-centuryaurorae from
thesehigh latitudes is fairly scarce.However, de Mairan cites
two important sourcesthat point unequivocallyto a diminished auroral presencein the northern regionsprior to 1716.
Anders Celsius, Swedish astronomer and inventer of the centi-
Maunder
Minimum
is evident in both cases. The Link curve
suggeststhat the minimum was deepestjust prior to 1700.
This agreesalsowith Fritz's expressedopinion on the variation of auroral activity throughthe seventeenthcentury [Fritz,
1881,p. 133],althoughit is not soapparentin the curvebased
gradethermometer,was 15 yearsold at the time of the grand on his catalog.
We will touch only briefly on the questionof the contindisplay of 1716. He began a catalog of aurorae observed in
Sweden, which he extended back to 1706 and continued to uation of the l 1-year sunspotcycle through the Maunder
Minimum. Both Maunder and Eddy point to the weak data
1732, at which time it contained 316 entries. He learned from
old men living in Uppsala that the aurorae, which were fre- basesupportingthe identificationof maximaand minima corquent in Swedenafter 1716,were novel to them. On the basis respondingto the 11-yearcycle during this time. Such identiof his attemptsto extend the auroral record backward he ar- fications had been made by Wolf on the basis of available
gued explicitly for the existenceof a prolongedauroral calm sunspotdata, by Fritz using auroral data, and again more recently by Schove[1955] on the basisof auroral data. Neverin Swedenprior to 1716.
De Mairan's secondsourceis a book by Johann Anderson theless,the point was correctlymade that becauseof the small
of maxima and minima presumed
on the natural history of Iceland publishedin 1746, in which samplesizethe assignments
the author states,'It has always appearedextraordinaryto me the existenceof the cycle, whereas the very existenceof the
that the most ancient Icelanders, as they have assured me, cycle itself was an important question.
Subsequentto Eddy's [1976a] article, Schove reexamined
should have been astonishedthemselvesat the frequent appearanceof the aurora in their island [that Andersonhad wit- the data from the seventeenthcentury, using in part Link's
nessed for some time beginning in 1730], declaring that catalog,and published[Schove,1980]a revisedlist of 11-year
formerly they were much less common than today.' Other maxima and minima, which again extendsthrough the Mauneighteenth-centuryauthors (Torfaeus and Pontoppidan) are der Minimum. Link [1978] also has prepared such a list givquotedby de Mairan and Fritz to showthat brilliant displays ing the 1l-year cycle maxima from 1610 to 1712 without a
were sufficientlyrare in the latter half of the seventeenthcen- gap. His assignmentsare basedon suddenchangesin the fretury in Iceland and Norway as to causegeneral alarm and to quency of reports recorded in his catalog (see also Link
[1977]). In addition, Vitinsky[1978] and Gleissberg[1977] arbe taken as prodigies.
Aurorae were not altogether absent during the Maunder gue for the persistence
of the 11-yearcyclethroughthe MaunMinimum either at low or at high latitudes.De Mairan him- der Minimum.
652
SISCOE: HISTORICAL
AURORAL
EVIDENCE FOR SOLAR VARIABILITY
cycle,the seriesof which had first to be identified, a number
correspondingto the maximum annual mean sunspotnumber
for the cycle. The assignednumber was, of course,not absolutely calibrated. On the basisof the indicatorsof solar activity available for a given cycle the cyclewas indexed on a ninepoint scale,from extremelyweak to extremelystrong.To this
nine-point index there was a correspondingfixed scaleof nine
annual mean sunspotnumbers, extending from 45 or less to
160 or greater. The procedure,though subjective,was neverthelessan initial step toward quantifying the historical data
150
,:,
base. It thus allowed one to display graphically the kinds of
I00
..f
5O
variations underlying the conclusionson the secularvariation
1AD
500
I000
1500
2000
that many investigatorshad made since the time of de MaiFig. 4. Two quasi-quantitativeprofilesgiving approximatesimu- ran. In comparison to the results based on pure number
lations of the secularvariation in the level of solar activity. The top countsof aurorae describedbelow, Schove'sprocedurehad at
profilewasderivedon the basisof Inc anomalies
in treerings[Eddy, least in principle the advantage of incorporating information
1976b]. The bottom profile was derived from historical accounts of
on the degree of animation of the displaysand on their geoauroral and naked-eyesunspotobservations[Schove,1955].Apparent
graphical extent for the purpose of determining the general
correspondences
between the main featuresare indicated.
level of solar activity. However, for the advantage to have
been realized fully the algorithm for converting the total
THE PRE-MAUNDER
MINIMUM
RECORD
amount of available information quantitatively into a suitable
The northern lights, along with comets,meteor showers,bo- index would have had to have been specified.
fides, and other heavenly and aerial wonders, are mentioned
The profile of solar activity thus derived by Schove is
in extant histories, annals, and chronologiesdating back to shown in Figure 4, together with the solar activity profile inthe time of imperial Rome. These draw upon even earlier ferredby Eddy [1976a,b] on the basisof Inc anomaliesin tree
sources, now lost, to extend the record of dated auroral dis- rings. We see in the profile by Schove many of the features
plays in the westernhemisphereto the fifth century B.C. The identified by Fritz. Prolongedminima are evident in the sevfirst catalogs devoted specifically to ancient aurorae were enth, fifteenth, and seventeenthcenturies,and there are procompiled shortly after the end of the Maunder Minimum, as longed maxima in the twelfth and sixteenth centuries. The
was noted above. By the time that Fritz pubfishedhis catalog other features noted by Fritz are not as clearly defined. The
in 1873, he could refer to at least eight prior catalogs that profile alsoshowsan isolatedmaximum in the fourteenthcenlisted eventspredatingthe Maunder M'mimum. Fritz's catalog tury, which appearsagain as a distinct and interestingfeature
contains over 400 possible pre-seventeenth-centuryauroral in more recent compilationsdescribedbelow.
occurrences.
The Inc profile has coarserresolutionthan the auroral
In his book on the aurora, published 8 years after the cata- curve. As a 'low-pass'filter it has the advantageof being sensilog, Fritz identified specificfeatures in the secular variation. tive mainly to the more enduring excursionsof solar activity.
The sixth, tenth, twelfth, and sixteenth centuries he described
Five
as being aurorally rich and the seventh,eleventh, thirteenth,
fifteenth, and seventeenth centuries as being aurorally
deficient. As will be seen below, these designationsare preserved with few changes in the most recent compilations.
They are, however, now better resolved as temporal features
correspondencesbetween the two profiles, as noted specifically in the figure. They are most evident in the more recent
features.There is a decided tendency,especiallyin the earlier
events, for a feature in the auroral profile to be older than its
•4Ccounterpart.This might indicatethat the timesof the excursionswere not sufficiently compensatedfor the temporal
sluggishness
of the natural processby which variations in solar activity becomeencodedin the Inc record[Eddy, 1976b].
In any event, the comparisonsuggeststhat the auroral record,
particularly in its most recent, high-resolutionform, could be
of usein determiningthe properdatesto assignto specific14C
and in termsof theirsrelative
amplitudes.
In the present century, Schove has made additions to the
record of historical aurorae. He combined these and the previous listingswith Chinese recordsof aurorae and naked-eye
sunspotsto constructan impressionisticrepresentationof the
variation in the amplitude of the solar sunspot cycle from
classicalGreek and Roman to recent times [Schove,1955]. To
make the representation, he assigned to each 11-year solar
such excursions
are indicated.
It is not difficult
to find
features.
In the last two decades,sevennew catalogsof historical au-
TABLE 1. Recently Compiled Catalogsof Ancient Aurorae
Catalog
Compiler
Date of
Publication
Areal
Coverage
Temporal
Coverage
Link
Link
Stothers
Newton
Dall'Olmo
Keimatsu
1962
1964
1979a
1972
1979
1976
626 B.C. to 1600 A.D.
1600-1700 A.D.
480 B.C. to 333 A.D.
450-1263 A.D.
450-1461 A.D.
687 B.C. to 1600 A.D.
Matsushita
1956
Europe
Europe
Greece and Italy
Europe
Europe
China, Korea, and
Japan
Japan
620-1909 A.D.
*Listed as probable to certain in the assessment
of the compiler.
Number
Entries
385
209
67
65
61
260*
18'
of
SISCOE: HISTORICAL
AURORAL
EVIDENCE
FOR SOLAR VARIABILITY
the entries in Stother's catalog extends the European data
field back to the fifth century B.C.
The histogramexhibits marked variations from century to
century. The most likely candidatesfor the causesof these
variations are changesin the level of solar activity, changesin
social-politicalhistory, and changesin terrestrial magnetism
40
30
CHINA
20
or in climate.
1111111111
BC 4
I
I
2
I
I
2
!
I
I---•.--I
4
I
6
I
8 I0 12 14 16AD
I
I
I
I
I
I
I
I
I
I
I
20
reference
to correlative
data it is
As ß
c•.
a,
as influences on the auroral record are concerned,
in the fifth and thirteenth centuries, which do not coincide
30
EUROPE
40
156
Fig. 5. The number of daysper century on which aurorae were recorded in the orient and in Europe (figure modified from Keimatsu
[1976]). Oriental aurorae are from the catalogof Keimatsu[1976], and
Europeanauroraeare from the catalogsof Link [1962] (solidline) and
Stothers[1979a] (dashedline).
rorae have appeared (see Table 1). The two by Link [1962,
1964], as mentioned earlier, cover the intervals 626 B.C. to
1600 A.D. and 1600-1700 A.D. The first contains 385 entries,
and the second209. Newton [1972] lists 65 displaysfrom the
interval 450-1263 A.D. Dall'Olmo [1979] furnishesan additional 61 occurrencesbetween450 and 1466 A.D. Sixty-seven
aurorae recorded in times of classical Greece and Rome, from
480 B.C. to 333 A.D., are given in a catalog compiled by Stothers [1979a]. These catalogsrecord western or near-eastern
aurorae. Eastern aurorae are listed in a seven-partcatalog by
Keimatsu [1976] covering the period 687 B.C. to 1600 A.D.
and containing a total of 578 auroral entries. Of these, only
258 are assignedby the author a reliability rating of 'probable,' 'very probable,' or 'certain.' The entries in the Matsushita [1956] catalogof aurorae seenin Japan are incorporated
into the Keimatsu catalog up to 1600 A.D. The Matsushita
catalog, however, extends to 1909 A.D. and includes events
the Maunder
Even without
possibleto make a reasonablystrongcasethat at leastthe major isolated fluctuationsin the histogram are the result of
changesin the level of solar activity. The argument proceeds
by the systematicelimination of the other possibilities.
the social-politicalhistoriesof China and Europe throughout
this period are essentiallyindependent.It is true that both regionswere approximately simultaneouslyaffectedby two episodes of invasions of nomads from the Asian steppesand
Mongolia. However, the disruptionsthat they causedpeaked
I_1
within
653
Minimum.
It wasconcluded
in thefirstpartof thisreviewthat,in general, inferencesdrawn from auroral occurrencefrequency
data become more trustworthy as the areal extent of the data
baseis enlarged. The influenceon the record of historical and
climatic accidentsdiminishesas the geographicalregion contributing to the record increases.For this reason the comparison between eastern and western (recorded) auroral occurrencefrequencypresentedby Keimatsu[1976] is especially
interesting.The comparisonis reproducedhere in Figure 5, to
which has been added ClassicalAge aurorae from the catalog
of Stothers.Link's first catalogwas usedto obtain the data on
the other European aurorae. Events rated probable or better
in Keimatsu's catalog composethe China data field.
The figure showsa histogramof the number of aurorae recorded per century in Europe and in China from the second
century B.C. to the sixteenthcentury A.D. The inclusion of
with notable featuresin the figure. We therefore concludethat
thepronounced
fluctuations_occurring
simultaneously
in both
regions,recognizablein Figure 5 as mirror-image anomalies
in the histogrampattern, are not the result of changingconditions in social-politicalhistory.
A change in the strength or the direction of the geomagnetic dipole will alter the frequency of aurorae seen within a
given geographicalregion. The dipole strengthis known from
archeomagneticdata to have performed a quasi-sinusoidaloscillation with a period of roughly 8000 years,which peaked in
the fifth century A.D. at a value approximately 50% greater
than that at present [Cox, 1969]. It is clear that the time scale
for this variation is too long to account for four significant
fluctuations,including both maxima and minima, over a time
span of one millennium.
The changein auroral occurrencefrequencyat a given site
causedby a migration of the geomagneticpole can be profound. To illustrate this, imagine the pole with the presently
existingpattern of auroral isochasmsrigidly attachedto it to
move continually and uniformly along the geographiclatitude
circle that runs through its presentposition. As the pole successivelypassesthrough the meridian and the antimeridian of
the given site, maxima and minima are successivelyrecorded
in the auroral occurrencefrequency. The range of the auroral
occurrencefrequency produced in this way dependsonly on
the geographiclatitude of the site. For example, at 50ø geographiclatitude, roughlythe latitude of London, Paris, Berlin,
and the U.S.-Canadian border, the averagenumber of auroral
nights per year would range from lessthan one to approximately 50. At 40ø, roughly the latitude of Washington,D.C.,
Madrid, Athens, and Peking, the number would range from
lessthan one in 10 yearsto about eight per year. The range is
nearly 2 orders of magnitude in each example. (The occurrence frequencies quoted here are based on the isochasms
given by Fritz [ 1881].)
Polar migration could easily account for the magnitude of
the marked changesin the number of aurorae recorded from
centuryto century.Nevertheless,the in-phasesynchronismof
easternand westernfluctuationseliminatespolar migration as
the causeof these particular variations. Furthermore, studies
of polar migration allow one to conclude that the time required to completea full cycleexceedsthe time betweenconsecutivepositiveor negativeanomaliesin the auroral occurrence frequency histogram [Keirnatsuet al., 1968; Barbetti,
1977].
654
SlSCOE: HISTORICAL AURORAL EVIDENCE FOR SOLAR VARIABILITY
20ORIENTAL
&
EUROPEAN
REPORTS
OF
AURORAE
15-
•j
n
0
20--
I
I
I
'4C-inferred
Medlevol
Moximum
•J
•
•,O--YEAR
SLIDING
AVERAGE
•5--
z
•o
•4½.Inflrrld
--
Medieval
Minimum
5
i
500
I
600
I
700
I
800
i
900
i
I000
I
I100
1200
1300
I
14OO A.O
_
Fig.6. Thenumber
of daysonwhichaurorae
wererecorded
perdecade
in theMiddleAgesascompiled
by Link
[1962],
Newton
[1972],
Da!!'O!mo
[1979],
andKeimatsu
[1976].Thebottom
panelconsists
of three-decade
slidingaverages
of the data given in the top panel.
Climatic variability is the remainingalternativeto solar thereisa corresponding
maximum
in theauroral
data,it took
variabilityto explainthe anomaliesin the 'global'occurrence placein the first or, moreprobably,the secondcenturyB.C.
frequencyof aurorae.Severalkindsof evidenceargueagainst In this case,however,the data from China do not lend supthis possibility.If the Maunder Minimum is consideredto be port to sucha conclusion.Stothersalsois carefulto point out
representativeof anomalous epochs, then the climate ex- that one cannot be certain whether the variations in the replanation is eliminated by the massof data and testimonies cordedauroral occurrencefrequencyin southernEuropefor
thathaveestablished
thisasan epochof greatlyattenuatedso- this period resultfrom plausiblehistoricalcausesor from natlar activity.Furthermore,althoughthe MedievalMaximum in ural causes.
the auroralrecordcoincideswith a climaticoptimumin EuThe time periodscoveredby the sevenrecentauroralcatarope, it coincideswith a period of climatic deteriorationin logsoverlapmaximallyin the 1000-yearintervalbetween450
China [Chu, 1973].Yet a strongauroralmaximumis indicated and 1450A.D., whichis essentially
the perioddesignated
as
from both regions.
the Middle Ages.By combiningthe independent
listingsfrom
Considernextthe specific
featuresin the histogram
in Fig- all catalogsone can increasethe temporalresolutionof the
ure 5. The mirror-imageanomaliesidentify the sixth,twelfth, auroralsecularvariationin thisperiod.The top panelin Figand sixteenthcenturiesas beingrelativelyaurorarich and the ure 6 achieves a factor of 10 increase in resolution over that in
seventhandfifteenthcenturies
asbeingrelativelyaurorapoor. Figure 5. One seesa tendencyfor anomaliesto clusterand for
Theseagreein the mainwith the specialperiodscalledout by changes
from decadeto decadeto be coherent.The general
Fritz and Schove,but now, in effect,they are confirmedwith aspectof the figuresuggests
a spectralcomponent
to the secuan extendeddata set.Comparisonwith Eddy's14Csolaractiv- lar variation lying between10 and 100 years.The bottom
ity profileleadsoneto associate
theseventh-century
depletion panelof the figureshows30-yearresolutionobtainedby a
with the MedievalMinimum,the twelfth-century
surpluswith slidingaverageof the number-per-decade
valuesgivenin the
the MedievalMaximum, and the fifteenth-century
decrease toppanel.Boththeless-than-one-per-century
andthe greaterwith the Sp0rerMinimum. Althoughthereare no featuresin- than-one-per-centurycomponentsof the secularvariation are
dicatedin the 14Cprofileof Figure4 corresponding
to therel- well revealed.This resolutionis perhapsan optimumfor the
ative auroral excesses
in the sixth and sixteenthcenturies, given number of eventsin the interval.
Stuiverand Quay[1980]reporta 14Ctree ring studythat reThe two medievalfeaturesin the inc solaractivityprofile
vealsan increasein solaractivityafter 1530,with a maximum areindicatedin thebottompanel.The temporalshapes
of the
in the interval 1590-1620.
corresponding
featuresin theauroralrecordarequitesharply
With regard to the suggested
Roman Maximum we see in defined.In viewof thefactthatchanges
in solaractivityasreFigure 5 a repeatof the situationpresentedin Figure4. If cordedin theformof incanomalies
in treeringsarelow-pass-
SISCOE: HISTORICAL AURORAL EVIDENCE FOR SOLAR VARIABILITY
20
655
1968].Associatedanomaliesin correlativedata add to its significance. The greatest cluster of pretelescopic,oriental sunspot observationsbearsa remarkable associationwith it, as is
1090 IlOO IO
20
30
40
50
• 2ø
I
60
2ø•B
I'
,o•
70
ll
80
•
90
1200
SUNSPOTS
,olKL
describedbelow.Recentanalysisof 14Canomaliesin tree
rings from this age revealed a pronounced, well-defined decrease, correspondingto a solar activity maximum. The decreasereached a maximum in the decadeof 1370 [Stuiver and
Quay, 1980]. The combinedauroral, sunspot,and 14Caber-
filteredand delayedby the bufferingactionof the atmosphere,
the auroralrecordprobablylocatesthe eventsmore accurately
rations certify the event as a genuine manifestation of a fluctuation in the level of solar activity. If one then acceptsthe assertionthat this event is but an exampleof the generalclassof
maxh-na "• occur approxh-natelyevery ow
tllat
on years, ..•, example perhapsdiffering from the othersin magnitude but not
in essence,it follows that the 80-year auroral cycle evident
here is a manifestation of solar variability.
A higher temporal magnificationis possiblein the autorally
rich intervals, namely, the Medieval Maximum, the 13601370 event, and the interval between the SpOrerand Maunder
minimums. Figure 7 showsthese at a resolutionof 3 years.
The apex of the first 80-year cycle in the Medieval Maximum is resolved into three peaks spaced approximately t0
years apart. These are most reasonablyinterpretedas auroral
expressionsof the I 1-year sunspotcycle. The crestat 1100 appears at the expectedposition if the seriesis continued backward in time, with one cycle missing.Schove[1955] locatesthe
maxima in thesecyclesand the missingor weak cycle exactly
as they appear here.
Although the peak at 1205, correspondingto the second80year cycle in the Medieval Maximum, resemblesthe earlier
ones from the previous 80-year cycle in width and rise, it is
isolated, which precludesidentifying it with the I l-year cycle
on the basis of an evident periodicity in the occurrencefre-
in time. In the auroral record the Medieval
quency.
Z
--
0
1200
Z 30
z
0
,
,
,
60
70
80
90
2O
30
40
50
60
I0
20
30
40
50
60
70
80
90
1600
I0
20
30
40
50
60
70
80
90
1700
I0
1350
,
[-c
• 2o
I
io
<[
o
1500
•-3ø
I
I
•
2o
•o
o
1600
Fig. 7. (a, b) The n_umber
of dayson whichauroraewererecorded
per 3 years,determinedyearly, for the aurorallyrich periodsevident
in Figure 6. (c) The interval from 1500to 1700A.D., compiledfrom
the entriesin the catalogsof Link [1962,1964].The sunspotdata used
to constructFigure 7b were taken from the catalogof Clark and Stephenson[1978].
Minimum
from 590 to 720, and the Medieval Maximum
extends
from 1090 to
1210. Both of the aurorally anomalousintervalsbegin 30-40
years before the correspondingintervals designatedby Eddy
on the basisof the 14Crecord.However,a morerecentanalysisof 14Ctreeringdatafindsa drasticincrease
in solaractivity
from 1070to I tOO,with the maximum extendingfrom t 100 to
1150,decliningthereafterto a minimum around 1280 [Stuiver
and Quay, 1980]. With the exceptionof a secondmaximum
near t 180 the auroral record conformsvery closelyto this determination.
The more rapid semiregularoscillationsevident in the figure have an averagecycle period of 87 years if one replaces
the missingcycle in the Medieval Minimum and counts the
suggestionof a cycle in the thirteenth century. Link [1968]
identified virtually the same featureson the basisof his catalog of western aurorae alone. He associatedthem with the
quasi-80-yearperiodicitythat had been describedpreviously
as operating back through the Middle Ages [Schove, 1955;
Gleissberg,1965].
In addition to the quasi-80-yearoscillation,Link recognizes
a quasi-400-year periodicity. He attributes the seventh-century Medieval Minimum, the deep, unnamed eleventh-century minimum, the fifteenth-centurySp6rer Minimum, and
the twelfth-century Medieval Maximum to its control.
Finally, the largestfeaturein the top panel of the figure deservesspecialmention. Although the 1360 peak conformsto
the expectedmaximum in the 80-year cycle,it differsfrom the
other peaks in its amplitude and its isolation. It is a conspicuousanomaly also in the previous,graphicallydisplayed
summariesof secularauroral variations [Schove,1955; Link,
The change in character from a signal with a clearly defined quasi-10-yearcycleto an irregular signalraisesthe question whether the sunspotcycle was intermittent in this interval. We have simulated signals in which the probability of
occurrenceof an event is proportional to the absolutevalue or
to the squareof a sinewave with an 11-yearperiod. The simulation wasperformedfor.situationsin which the averagenumber of eventsper cycle was 5, 10, 15, and 20. Wave forms composed of 3-year running sums were compiled, conforming to
the procedure used to constructFigure 7.
In the case where the occurrence probability function is
proportional to the absolutevalue of a sine wave, more than
an average of 20 events per cycle is required to perceive a
distinct periodicity in the wave form. However, it is clear from
Figures 1, 2, and 7 that the probability function is usually
more sharplypeakedthan a sinewave. In the caseof the sinesquared probability function a distinct periodic pattern
emergesbetween 15 and 20 eventsper cycleon average.At an
averageof t0 eventsper cycle the simulatedsignal resembles
very closelythat seenin Figure 7 between 1150and 1200 A.D.
Reference to Figure 6 reveals that the recorded occurrence
frequencywas about 15 per cyclein the interval displayingthe
periodicity,but it was about 10 per cycle or lesswhen no distinct peaks were recorded.
On the basisof the simulationsa simple explanation for the
apparentchangefrom periodic to irregular characterof the recorded auroral occurrencefrequencyis that there is a threshold in the average occurrence frequency that must be exceeded before the underlying periodicity in the signal can be
perceived.The secularvariation of solaractivity includingthe
_
_
656
SISCOE: HISTORICAL
AURORAL
EVIDENCE
80-year cycle modulatesthe averagenumber of aurorae that
are recorded during the l 1-year cycle. For the number of
presently known aurorae recorded during the Middle Ages
the secularmodulation carriesthe averageauroral occurrence
frequencyacrossthe thresholdduring the most aurorally rich
FOR SOLAR VARIABILITY
mums. The synchronicity of the anomalies in eastern and
western records,as well as their correspondenceto anomalies
in the 14Crecord, establishesthat they are sympatheticresponsesto secularvariationsin the level or stateof solaractivity.
The solar cause of a briefer
intervals.
That the thresholdof detectionof periodicity has evidently
been crossedfor significantperiods during the Middle Ages
should serve as an incentive and a justification to carry out
further, systematicsearchesfor historical auroral recordings.
As an example, doubling the number of known aurorae recorded in the Middle Ages would possibly expose the complete wave form of the 1l-year, 80-year,and longer-termvariations for this 1000-year interval. It should be noted that
$tothers [1979b] was able to infer the presenceof the I 1-year
cycle of solar activity during the aurorally rich secondcentury
B.C. on the basis of his catalog of aurorae from classicalantiquity.
Returning now to Figure 7b, we encountera very curious
associationbetween the auroral record and the pretelescopic,
oriental sunspotrecord, presentedrecently by Clark and Stephenson[1978]. It was noted earlier that the decadeof 1360
possesses
the largest number of recordedaurorae of all the
decadesin the Middle Ages. The Clark and Stephensoncata-
log contains139 entries.Thesecoverthe interval28 B.C. to
1604 A.D., giving an average of less than one per decade.
While the sunspotcatalog contains two entries in the aurorally rich 1360's,there are 23 entriesin the following decade.
The next largestnumber of entriesper decadein the catalogis
nine. The auroral and sunspotoccurrencefrequency profiles
in Figure 7 each show the distinctive fast rise, slow decay
shapecharacteristicof the 11-year solar activity cycle. However, the peaksin the two profilesare separatedby one cycle.
More important than the asynchronismin the two recordsis
the direct evidencehere for the operation of the I 1-year solar
cyclein yet anotherportion of the Middle Ages.
The bottom panel in Figure 7 coversthe interval from the
last two decadesof the Sp6rer Minimum to nearly the end of
the Maunder Minimum. It therefore includes the aurorally
rich epoch separatingthe two minima. Essentiallythe same
figurehas beenpublishedby Link [1978],who presentedit as
evidence that the 1l-year solar cycle operated prior to the
Maunder Minimum. As in the top panel, one seesa tendency
for the periodicity to become evident at times when the
auroral occurrence frequency averaged over a solar cycle
is high. The signal is indistinctat times when the averageis
low, as during the Maunder Minimum. However, the first
excursion
around
tributed
to a lack of sufficient
information
events in the record to make
available
in the auroral
record
to information
about the level or condition of solar activity. Historical auroral information
includes the secular variation
of the auroral
occurrencefrequency that this review has emphasizedand
also the equatorward extent, the intensity and the degree of
I
-
LATITUDE
EXPLORER
COUNTRY REACHED
G F LYONS
YOUNGER
--
WRANGEL-PF
WE
ANJOU
PARRY
ß
ROSS..
the absolute sine function.
8UCHAN-FRANKLIN.
KDTZESUE
SCORESSY, fhl ELOER
PROSPECTUS
"
EnGLAnO
65
ENGLAND
?6
I•
ENGLANO
?4
.•0
74
20
ENGLANO
LU.TKE
and 1660's,possess
25, 17, and 18 recordedevents.The absence of a clear periodicity in this interval could reflect a
changetoward a lesspeakedwave shape,suchasis typified by
AND
is
manifest such a signal.
The auroral record has proven itself to be a valuable data
sourcefor the investigationof the secularvariation of solar activity. Significantlymore information could be extractedfrom
this source if further systematic historical research were to
double approximately the number of events in the record
prior to 1700. It is possiblethat a good representationof the
last 2000 years of variations in solar activity at a resolution
that includesthe 1l-year cyclecould be achievedin this way.
More effort needsto be expendedon convertingthe kind of
three decades of the Maunder Minimum, the 1640's, 1650's,
SUMMARY
1360 A.D.
similarly established.Its evident membership in a genus of
like anomalies that recur on an average of 87 years in the
Medieval auroral record implicatesa solar causefor the entire
genusand at the sametime exposesa consonantperiodicityin
solar activity itself.
The temporal profile of the auroral occurrencefrequency
exhibits a quasi-10-year periodicity during aurorally rich intervalswithin the Middle Ages and the Renaissance.The pattern and the period of this signalconform to that producedin
modern times by the 11-year solar activity cycle. The absence
of a periodic signal at other times is most conservativelyat-
20
RUSSIA
7'6
RUSSIA
?6:33
ENGLAND
?4
47
ENGLANO
73
41
ENGLAND
?6 54
ENGLAND
80
RUSSIA
65
ENGLAND
81
:•7
12
The secular variation in the level of auroral activity has
been known since the early 1700's. Its main features have
beenprogressively
better resolvedsincethen througha tradition of historical auroral research. The auroral record has now
I
evolvedto the point where the featuresthat have been identio
o
o
,•
o
,•
fied in the 14Crecordstandas distinctive,highly resolved,major departuresfrom an apparentlystable,longer-termaverage. Fig. 8. Principal expeditionsin searchof northeastand northwest
The principal deviationsincludethe Medieval Minimum and passagesto China that sailedduring the auroral calm of the early
Medieval Maximum and the Sp/•rer and Maunder Mini- nineteenth century.
SISCOE: HISTORICAL AURORAL EVIDENCE FOR SOLAR VARIABILITY
animation of displays,the categoriesof auroraeseen,and possibly in some time intervals the location and size of the auroral zones.From these types of data, the permissibleinferences regarding the magnetosphere,the solar wind, and the
sun must be determined.
Historical studiesfocusedon times of specificanomaliesof
solar activity could be particularly valuable. The recent, postMaunder
Minimum
auroral calm centered around
1810 can
be used to illustrate the point. This event occurred at a time
when systematicauroral catalogswere being made by individuals in Europe and America [e.g., Dalton, 1834; Lovering,
i860]. The event itself is well establishedas a solar activity
anomaly by both the sunspotand the auroral record. In this
interval, solar-terrestrial phenomenology departed remarkably from that with which we have direct acquaintance.
Polar exploration aimed at finding northeastand northwest
passagesto China was in full flower in the early nineteenth
century. Figure 8 givesa list of the principal expeditionsthat
sailed during the depth of the auroral calm. Expedition reports in many casescontain auroral accounts.A systematic
study of all of the data available for this event could result in
a descriptionof auroral behavior and of the auroral zone sufficient to infer in fairly specificdetail the correspondingcondition of solar activity.
Gleissberg, W., Betrachtungen zum Maunder-Minimum
nent•itigkeit, Sterne Weltraum, 16, 229-233, 1977.
657
der Son-
Gregory of Tours, The Historyof the Franks,translatedfrom Latin by
L. Thorpe, Penguin,London, 1974.
Guiducci, M., Discourseon the comets,in The Controversyon the
Cometsof 1618,editedby S. Drake and C. D. O'Malley, University
of PennsylvaniaPress,Philadelphia, 1960.
Halley, E., An accountof the late surprisingappearanceof the lights
seen in the air, Phil. Trans. Roy. Soc., London, XXIX, 406-428,
1716.
Hansteen,C., II, Beobachtungentiber das Nordlicht vom 7 Januar
1830, Poggendorffs •4nn., XXII, 536-537, 1831.
Harang, L., The •4urorae,Chapman and Hall, London, 1951.
Herschel, J. F. W., On the solar spots,Quart. J. Sci.. 1.219, 1864.
Keimatsu, M., A chronology of aurorae and sunspotsobserved in
China, Korea and Japan, •4nn. Sci., 13, 1-32, 1976. ,
Keimatsu, M., N. Fukushima, and T. Nagata, Archaeo-aurora and
geomagneticsecularvariation in historic time, J. Geomagn.Geoelec., 20, 45-50, 1968.
Kirch, G. (Kirchii, G.), Brevis descriptio clari qujusdam luminis,
anno 1707, Misc. •4cad. Berolinensia, I, 132, 1710.
Link, F., Observationset cataloguedes autoresbor6alesapparuesen
Occident de 626/t 1600, Geofys.Sb., X, 297-387, 1962.
Link, F., Observationset cataloguedes autoresbor6alesapparuesen
Occident de 1601/t 1700, Geofys.Sb., XII, 501-547, 1964.
Link, F., Auroral and climatic cyclesin the past, J. Brit. •4stron.•4ss.,
78, 195-205, 1968.
Link, F., Sur l'activit6solaire•.u 17•mesi•cle,•4stron.•4strophys.,
54,
857-861, 1977.
Link, F., Solar cyclesbetween 1540 and 1700, Solar Phys.,59, 175178, 1978.
•4cknowledgments.It is a pleasureto acknowledgeinformative discussionsand correspondence
with S. Silverman, R. Eather, J. Feynman, M. Stuiver, and R. Stothers,which were very helpful in writing
this review.This work was supportedin part by the AtmosphericSciencesSectionof the National ScienceFoundation under grant NSF
Loomis, E., Comparison of the mean daily range of the magnetic declination and the number of aurorasobservedeach year, with the extent of the black spoteson the surfaceof the sun, •4mer.J. Sci., Ser.
ATM77-24843.
Lovering, J., On the secularperiodicity of the aurora borealis, Mem.
REFERENCES
Angot, A., The •4urora Borealis,D. Appleton, New York, 1897.
Barbetti, M., Measurementsof recent geomagneticsecularvariation
in southeasternAustralia and the questionof dipole wobble, Earth
Planet. Sci. Lett., 36, 207-218, 1977.
Cassini,J. D., D6couvertede la lumi•re c61estiqui paroist dans le zodiaque, Mem. •4cad. Roy. Sci., VIII, 179-278, 1666-1699.
Chu, K.-C., A preliminary study on the climatic fluctuationsduring
the last 5000 years in China, Sci. Sinica, 16, 226-256, 1973.
Clark, D. H., and F. R. Stephenson,An interpretationof the pre-telescopicsunspotrecordsfrom the orient, Quart. J. Roy. •4stron.Soc.,
3, 5, 245-260, 1873.
•4mer. •4cad. •4rts Sci., 9, 101-111, 1860.
Lovering, J., On the periodicity of the aurora borealis, Mem. •4mer.
•4cad. •4rts Sci., 10, 9-351, 1868.
Maraldi, M., Observationsd'une lumi•re septentrionale,Mem. •4cad.
Roy. Sci., 95-107, 1716.
Matsushita, S., Ancient aurorae seen in Japan, J. Geophys.Res., 61,
297-302, 1956.
Maunder, W., A prolonged sun-spotminimum, Knowledge,17, 173176, 1894.
Meinel, A. F., B. J. Negaard, and J. W. Chamberlain, A statistical
analysis of low-latitude aurorae, J. Geophys.Res., 59, 407-413,
1954.
Mendillo, M., and J. Keady, Watching the aurora from colonial
America, Eos Trans. •4GU, 57, 485-491, 1976.
Clerke, A., A prolonged sun-spot minimum, Knowledge, 17, 206,
1894.
Morton, C., CompendiumPhysicae,vol. XXXIII, reprintedby the Co1onial Societyof Massachusetts,Boston, 1940.
Cox, A., Geomagnetic reversals,Science,163, 237-244, 1969.
Dall'Olmo, U., An additional list of aurorasfrom European sources Newton, R. R., Medieval Chroniclesand the Rotation of the Earth,
Johns Hopkins University Press,Baltimore, Md., 1972.
from 450 to 1466 A.D., J. Geophys.Res., 84, 1525-1535, 1979.
Dalton, J., MeteorologicalObservationsand Essays,2nd ed., Baldwin Olmsted, D., On the recent secular period of the aurora borealis,
19, 387-410, 1978.
and Gradock, London, 1834.
Smithson. Contrib. Knowl., 8, 1-52, 1856.
de Mairan, J. J. D., Trait• Physiqueet Historiquede l'•4uroreBorbale, Schove,D. J., The sunspotcycle, 649 B.C. to A.D. 2000, J. Geophys.
Res., 60, 127-146, 1955.
Imprimerie Royale, Paris, 1733.(Secondedition, 1754.)
Eather, R., The Northern Lights,AGU, Washington,D.C., in press, Schove,D. J., Sunspotmaximum and minimum dates from aurorae,
200 B.C.-A.D. 1712, in Proceedingsof the Solar-Terrestrial Predic1980.
tionsWorkshop,
vol. 3, editedby R. F. Donnelly,U.S. Department
Eddy, J. A., The Maunder Minimum, Science,192, 1189-1202, 1976a.
of Commerce, Washington, D.C., in press, 1980.
Eddy, J. A., The sun sincethe Bronze Age, in Physicsof Solar PlanSchr6der; W., Behavior of aurora in Germany, 1880-1964, J.
etary Environments,edited by D. J. Williams, pp. 958-972, AGU,
Geophys.Res., 77, 6890-6892, 1972.
Washington, D.C., 1976b.
Fritz, H., VerzeichnissBeobachter Polarlichter, C. Gerold, Vienna,
1873.
Fritz, H., Das Polarlicht, F. A. Brockhaus,Leipzig, East Germany,
188!.
Stetson, H. T., and C. F. Brooks, Auroras observed at the Blue Hill
Observatory, 1885-1940, J. Geophys.Res., 47, 21-29, 1942.
Stothers, R., Ancient aurorae, Isis, 70, 85-95, 1979a.
Stothers,R., Solar activity cycleduring classicalantiquity, A stron.As-
Fulke,W.,A Goodly
Gallerye,
Witha MostPleasant
Prospect
Intothe
trophys.,
77,121-127,1979b.
Garden
ofNaturallContemplation,
toBehold
theNaturallCauses
of Stuiver,
M., andP. Quay,Changes
in atmospheric
carbon-14
attribAll KindofMeteors:
etc.,JohnHaviland,
London,
1571.(Laterediutedto a variable
sun,Science,
207,11-19,1980.
tionsin 1634,1639,1640,1654,1655,1670.)
Tromholt,S., Katalogder in Norwegenbis Juni 1878 beobachteten
Gassendi,P., Physics,in OperaOmnia,6 vols.,Montmort,Lyons,
Nordlichter,Kristiania,Oslo,1902.
France,1658.
Gleissberg,
W., Theeighty-year
solarcyclein auroralfrequency
numbers,J. Brit. Astron.Ass.,75, 227-231, 1965.
vonLeibnitz,G. W., (Leibnitii,G. W.), Annotatio
delucequamquidam,auroramborealemvocant,Misc.Acad.Berolinensia,
I, 134135, 1710.
658
SlSCOE: HISTORICAL
AURORAL
EVIDENCE
Vitinsky, Yu. I., Commentson the so-calledMaunder Minimum, Solar Phys., 57, 475-478, 1978.
Woods, J., Woods Voyage, in Collectionof Voyagesand Travels,compiled by J. Harris, p. 617, London, 1705.
Zanotti, E., Descriptionof an auroraborealisobservedat the observa-
FOR SOLAR VARIABILITY
tory of the instituteof Bononia,the night of the 5/16 of December
1737, Phil. Trans. Roy. $oc., London, XLI, 593-600, 1738.
(Received September15, 1979;
acceptedJanuary 8, 1980.)