VOL. 4, NO.

$ WATER RESOURCES RESEARCH OCTOBER 1968

Noah,Ioseph,and Operational
Hydrology

BENOIT B. I•fANDELBROT

JAMES R. WALLIS
International Business Machines Research Center
Yorktown Heights, New York 10598

Dedicated to Harold Edwin Hurst

ß . . were all the fountains of the great deep broken up, and the windows of heaven
were opened. And the rain was upon the earth forty days and forty nights. Genesis, 6,
11-12

ß .. there came seven years of great plenty throughout the land of Egypt. And there
shall arise after them seven years of famine ... Genesis,41, 29-30

Abstract. By 'Noah Effect' we designate the observation that extreme precipitation can be
very extreme indeed, and by 'Joseph Effect' the finding that a long period of unusual (high
or low) precipitation can be extremely long. Current models of statistical hydrology cannot
account for either effect and must be superseded.As a replacement, 'self-similar' models ap-
pear very promising. They account particularly well for the remarkable empirical observa-
tions of Harold Edwin Hurst. The present paper introduces and summarizes a series of
investigations on self-similar operational hydrology. (Key words: Statistics; synthesis; time
series)

INTRODUCTION be of interest to note that they are instances

By 'NoahEffect'we designate the fact that of a broadfamilyof 'self-similar
models.'
The
extremeprecipitation canbe very extremein- conceptof 'self-similarity' originatedin the
deed,and by 'JosephEffect'the fact that a theoryof turbulence, to whichit waslongre-
longperiodof highor lowprecipitation canbe stricted,but it has recentlybecomeof value
extremely long.In a seriesof papersto which in studyinga variety of naturalphenomena
the presentwork servesas Introduction and (see for example,Mandelbrot[1963, 1966,
Summary, we shalldescribein detaila family 1967a,1967b]).
of statisticalmodelsof hydrologywhichwe A word of acknowledgment beforewe pro-
believeadequately accountfor the Noah and ceed.In investigations of currentstatistical
Josepheffects.Differentpapersin the series modelsof hydrology, oneof the mostactive
will be devoted,respectively,to mathematicalgroupshas beenthat foundedby Professor
considerations,to accountsof computersimu- Harold A. Thomas,Jr., at Harvard.In view
lations,and to analysesof empiricalrecords. of the criticaltoneof muchthat follows,the
Later paperswill studyvariousproblems of authorshastento express heretheir personal
water controlengineering as problems of op- indebtednessto Harold Thomas.He directed
erationsresearch, whichthey were long be- B.B.M.'s curiosityto Hurst's work and to
fore the term 'operationsresearch'itself was hydrologyand later initiatedJ.R.W. into the
coined. intricacies of 'synthetic hydrology' and simu-
The models to be described were advanced lation.
and arguedin Mandelbrot [1965] and Mandel-
brot and Van Ness[1968].We havecarried PARTISAN
COMMEI•ITS
Ol•lCURREI•IT
STATISTICAL
outextensive
experiments
of everykindto test tIYDROLOGY
and develop these models,and we have, in Current modelsof hydrologyassumeprecip-
our opinion,confirmedtheir soundness.
It may itation to be randomand Gaussian(i.e., fol-
909
910 I•IANDELBROT AND WALLIS

lowing the normal probability distribution, combininghydrologyand climatology,is far

with its 'Gaiton ogive') with successive
years' from being of Gauss-Markovform.
precipitationseither mutually independentor Other approaches
to hydrologicalmodeling
with a short memory. 'Independence'implies also start with a Gauss-Markovprocessand
in particularthat a large precipitationin one then introducemodifications that tend to be
year hasno 'aftereffect'on the followingyears; more extensivewhen recordsare long than
'short memory'meansthat all aftereffectsdie when they are short. This generalprocedure
out within a few years. The classicalshort can be illustratedusing two examples.The
memorymechanism, the Gauss-Markovprocess, first involvesthe loosebut intuitive idea of
is a 'singlelag linear autoregressive
model.'In the duration of a drought,the secondthe
this ease, aftereffectsdie out in geometric morerigorousbut lessintuitiveconceptof the
progressionand decreaserapidly. More gen- Hurst range.
eral are the 'multiplelag linear autoregressive For droughtsthe point is that, if an inde-
models.'One feature commonto all thesemo- pendent Gauss processor a Gauss-Markov
delsis their belonging to 'the Browniandomain processis chosento fit best the other aspects
of attraction,' a term which we shall define of precipitation,it will greatly underestimate
later. It is our basic belief that modelsin the the durations of the longest drought. There-
Browniandomaincannotaccountfor the Noah fore, suchprocesses must be modifiedby con-
and Josepheffects.These modelsunderesti- sideringmoredurableaftereffects (for example,
mate the complication of hydrologicalfluetua- through'multiplelag'models).Onewhoeonsid-
tions and the difficultyof 'controllingthem erssuchmodifications asnuisance correctionsto
by establishing reservesto make the future a basicGauss-Markov process will naturallytry
less irregular' (to paraphrasethe title of to fit all availabledata with a 'minimal'modi-
Mass• [1946]). fled process, havingas short a span of after-
Disappointmentwith specificmodelsin the effectsas possible.However,when the sample
Brownian domain is today very widespread durationis sufficientlyincreased, 'unexpectedly'
among hydrologists(for example,see Yev- long droughtswill again be observed.This
jevich [1968]). Therefore,our sweepingasset- shows,after the fact, that the 'minimal'model
tion can only be controversialin its blanket had attributed a special significanceto the
condemnationof all modelsin the Brownian longestsample T that was available when it
domain.To try to minimizesuchcontroversy, was constructed. As the sampleincreases, such
we shall now describevarious stop-gapsthat a model must be changed.(For example,the
have beenproposed. We shall point out that, numberof lagsmust be increased.)
in effect, such models end up outside the 'Drought' being,as we said, an elusiveeon-
Browniandomain. eept, let us now proceedto the observedbe-
Some authors eventually conclude that a havior of the 'Hurst range,'which is lessintui-
descriptionof hydrologicalreality requires a tire but easierto study.To defineit, onebegins
Gauss-Markovprocesswith time-varying pa- by evaluatingthe total capacityR(s) a reser-
rameters. Such models must, however, be voit musthave had, in orderto perform'ideally'
changedbefore their consequences have had for s years.'Ideal performance'here means(a)
time to develop fully. For example,Jbefore that the outflow is uniform; (b) that the re-
the sample average of precipitationhas had servoirendsthe periodas full as it began; (c) r
time to 'stabilize'near its expectedvalue, eli- that the dam never overflows; (d) that the
marie changeis assumedto modify that ex- capacityis the smallestcompatiblewith (a),
pectation.We believesuchmodelsto be rather (b) and (c). The conceptof an ideal dam
pointless,becausethe usefulness of a statistical is of coursepurely retrospective,since data
modellies primarily in its large samplepredie- necessaryto design such a dam are only
tions. Since a changingexpectedvalue easily known when it is too late. However, the past
overwhelms Gauss-Markov fluctuations, a dependence of the ideal capacityupon s tells a
Gauss-Markovhydrologicalmodel can be used great deal about the long-run behavior of a
only in conjunctionwith some'mastermodel' river on which an actual dam is to be built.
ruling climatic change.The over-all model, Postponingqualifications
to later papers,let us
 OperationalHydrology
describea striking discoveryH. E. Hurst made
while examiningR(s) for the Nile and other
geophysicalrecords.Hurst divided the capacity
R(s) by the standard deviation $(s) of s suc-
cessivedischarges.The empirical finding, then,
is that save perhaps [or small values o] s,
the rescaledrange R(s)/S(s) is proportionalto
s• with H a constant between 0.5 and 1. Hurst
judged H to be 'typically' near 0.7, but other
estimatesput H much higher, above 0.85. An
independentGaussmodel yields R(s)/S(s) •
0.5. Gauss-Markov models, 'multiple lag' mo-
dels, and all other models in the Brownian do-
main give a more complex prediction' R(s)/
$(s) • sø'5for large s, but R(s)/S(s) grows
faster than sø'5for small or moderate s, which
we shall call the 'initial transient region.' In
this transient region a variety of different be-
haviors may be obtained. Moreover, many
modelsmay lead to the sametransientbehavior,
which makes them indistinguishablefrom the
viewpoint of predictionsconcerningR(s)/$(s).
Then, if one has only the values of R(s)/$(s)
for I _• s _• T (with T a finite duration),
many different modelsof the BrownJandomain
are likely to yield predictions undistinguish-
able from the data. When s exceedsT, how-
ever, the Hurst range of every one of these
processeswill soon merge into the classicalsø'*
pattern. So far such a convergenceto sø'• has
never been observedin hydrology.Thus, again,
those who consider Hurst's effect to be a tran-
sient implicitly attach an undeserved impor-
tance to T, which is typically the currently
available sample sizes.These scholarscondemn
themselvesnever to witnessthe full asymptotic
developmentof the modelsthey postulate.
TOWARDS A CHANGE Or DIRECTION IN
HYDROLOGICAL MODELING
We have now sketched a few reasons,to be.
fully developedlater, why we dislike hydro-
logical models obtained by 'patching up' the
Gauss-Markovprocess.It shouldbe understood
that this criticism does not imply that we ex-
pect to be able, with some other model in-
volving few parameters,to representfully the
tremendouslycomplicatedhydrologicalreality. astronomy.
A model having few parameterscan only be a We realize that a stress on low frequencies
'first approximation,'and we believethat such emphasizesidiosyncrasies.But the purpose of
a first approximationmust endeavorto 'catch' hydrological engineering is to guard against
911
the main features of the problem, namely the
Josephand/or Noah Effects.
To characterizeour proposednon-Brownian
first approximations,the loose distinction be-
tween 'low-frequency'and 'high-frequoncy'phe-
nomena is useful. Using a Gauss-Markov proc-
ess implies fitting high frequency effects first
and worrying about low frequency effectslater.
We propose to invert this order of priorities.
Conveniently, we shall be able to use the term
'low frequency' in either of its main meanings,
to designateeither a rarely occurring phenom-
enon, or an oscillatingphenomenonwith a long
wavelength.
The conceptsof 'low' and 'high' frequency
are, of course, relative. Natural phenomena
cover a continuousspectrum,in which very low
frequenciesof turbulence theory and very high
frequenciesof hydrology overlap around one
cycle per day. This frequency, being funda-
mental in astronomy, may also separate zones
in which intrinsically different mechanismsrule
the fluctuations of precipitation. The same
holds for the wavelengthof one year. The third
important wavelength in hydrology is 50 to
100 years,which we shall refer to as a 'lifetime.'
This is roughly the horizon for which one de-
signs water structures and also, coincidentally,
the length of most hydrological records. The
importance of this wavelength is of human, not
astronomical, origin; it is purely 'anthropo-
centric.' Whereas precipitation fluctuations of
wavelength near one day or one year may
participate in several physical mechanisms,
fluctuationsof precipitation of wavelengthnear
one lifetime are likely to participate in one
mechanism only. Thus, the latter are likely
to be simpler than the former. Now assume
that one wants an approximationvalid over a
wide band of frequencies.It may be convenient
to start by askingfor a goodfit in somenarrow
frequencyband, with the hope that the formula
so obtained will be applicable over the wide
band. Under these circumstances, the band
near one lifetime, although purely anthropo-
centricin its definition,constitutesin our opin-
ion a better basis of extrapolation than the
band near one year, which has meaning in
912 MANDELBROT AND WALLIS

the recurrence of such idiosyncrasies,and one double the variance of each of them. It is un-
cannotafford to neglectany availableinforma- fortunate that this property doesnot yet have
tion. a generally acceptedname. Let it be said im-
We also realize that, modelsin the BrownJan mediatelythat this property fails to hold, either
domain having long been recognizedas ap- for fractional noises or for approximation
plicablein many fieldsof science(beginningof thereto used in Mandelbrot and Wallis [1968a].
course with the BrownJan motion of statistical These processessatisfy, however, both the law
mechanics), their proponents among hydro- of large numbers and the .centrallimit theorem.
logists are often able to identify ready-made If a natural phenomenonobeys the conditions
answers to the standard problems. Our pro- of a validity of all three of these mathematical
posed approach requires more work, but the theorems,it will be called 'smooth' or 'in the
answers appear to be sufficiently better to Brownian domain of attraction.' There exist
make this work worthwhile. Moreover, the phenomenathat fail to obey the conditionsof
conceptof self-similarity,to be discussedlater validity of the third theorem, or of the last two,
in the paper, will bring true simplicity. or even of all three. Such phenomenawill be
called 'erratic.' For example, the average T-•
'SMOOTI-I' AND •ERRATIC' PROCESSES
Z•:•• X(t) may ]ail to tend to any limit. Or it
We finally come to the promised character- may tend to a limit, whereas its distribution
ization of the ']3rownian domain of attraction' does not tend to the Gaussian. Or it may tend
and of related specificmeaningsfor the terms to a Gaussianlimit, whereas'past' and 'future'
'smooth' and 'erratic' time series. We need averages fail to become asymptotically inde-
three resultsof probabilitytheory,two of which pendent. The importanceof this latter circum-
are classical,and all three of which relate to stancefor the hydroIogistlies in the coincidental
averagesof T successive terms of a stationary equality between the order of magnitude of
time seriesdesignatedby X(t). most past records and the horizon of most
One says that X(t) satisfiesa law of large designs(both equal one lifetime), and in the
numberswhen its averagetends to a limit, the fact that, true expectations being unknown,
expectationœX(t), when T -• o•. This law is planning requiresthe determinationof the dif-
the theoretical justification of the common ferencebetweenthe expectedmean flow over a
practiceof taking sampleaveragesas estimates future lifetime and the known past average.
of populationexpectations. It is readily verified that the Gaussianmod-
One says that X(t) satisfiesthe more de- els with a limited memory all assumehydro-
manding Central limit theorem in its original logical phenomenato be 'smooth.' The Noah
form when, for large T, the distributionof the and Josepheffects,on the contrary, not only
averagebecomesapproximatelyGaussian,with suggestthat hydrologicaldata are 'erratic' but
a variance tending to zero as T -• •. This is also express the major two forms of erratic
the justificationof the commonbelief that the behavior. We shall speak of 'Joseph.-erratic'
sampleaverageis likely, if T is not small, to behavior when the wettest decade within a cen-
be a goodestimateof the expectation.A corol- tury includes an extraordinary 'term' of wet
lary of this is that, for large T, eventhe largest years.We shall speakof 'Noah-erratic' behavior
of the T quantities T -• X(t) contributesneg- when a few of the years within the century wit-
ligiblyin relativevalueto the averageT-• Zt:•• ness 'floods' so major as to affect the average
X(t). precipitation for periodsof many years within
The final result is less well known but very which the flood years occurred.Needlessto say,
important in applications.Let us call 'past a processcan be both Joseph-and Noah-erratic
average' the expressionT-• Y•t:-•2 X(t) and simultaneously, a complication that we shall
'future average'the expressionT-• Z•:•• X(t) face much later. 'Pure Joseph-erratic'behavior
and consider the difference between these two. will be said to apply when none of the yearly
The third basic result on the averages of precipitationsduring a 'wet term,' had it stood
random sequences assertsthat, as T -• •, the alone, would have been interpreted as a flood.
two averagescan be consideredindependent, Clearly, the word 'erratic' should not be
so that the variance of their difference is construedto suggesta 'black-and-white' con-
 Operational Hydrology 913
trast. The three theoremsin questionare, in- distant 'outliers.' Also, high water levels, which
deed, asymptotic,but scientificapplicationsof would be considered 'millenium floods' if one
mathematicsalwaysdeal with T's in somefinite extrapolatedthe tails of the histogram from its
horizon. Consider, for example, an infinite body, occur much more frequently than they
(nonrandom)seriesa(t). For the mathemati- shouldunder the Gaussianassumption.
cian, the basic distinction is whether the sum Despite the importance of deviations from
Y.t=•
© a(t) is, respectively,finite or infinite. For the Gaussian,we believe it worthwhile to begin
the scientist,on the other hand, the ultimate our investigationof the JosephEffect by Gaus-
convergence of Y.t=• a(t) is of little import, un- sian processesX(t), which are by definition
lessZ,• • a(t) is alreadycloseto its limit. There- such that the joint distribution of their values
fore, the conceptof 'erratic' must be considered at any finite number of instants is a multivari-
as comingin variousdegreesof intensity,rather ate Gaussianvariable. Such processeswill be
like 'grey.' examined in the next several sections. In the
sectionnear the end of the paper, highly non-
THE ISSUE OF THE MARGINAL DISTRIBUTION OF
Gaussianprocesseswith a Noah Effect will be
TI-IE YEARLY FLOW
mentioned. (Processesthat are only 'locally'
We shall now characterize more accurately Gaussian are studied in Mandelbrot [1968].)
the idea of a 'pure Joseph-erratic'process,
GAUSSIAN PROCESSES AND TI--II• COVARIANCE
using the concept of 'marginal distribution,'
which is defined as the distribution of the values Gaussianprocesses are well known to be fully
of a processirrespectiveof their chronological specifiedby their covariancefunction: If X (t)
order.We believeit reasonableto demandthat, is of zero mean and unit variance,the covari-
when the order of values of a pure Joseph- ance C(s) is the correlation between X(t) and
erratic sampleis scrambled,one shouldbe left X(t -[- s). (Of course,in the caseof Gauss;.an
with a smoothprocess.Thus, the marginal dis- variables zero correlation is identical to in-
tribution of these values will draw a line be- dependence.)Our problem is, then, to use the
tween, on the one hand, Noah-erraticprocesses behaviorof C(s) to classifya Gaussianprocess
and, on the other hand, processesthat are as smoothor Joseph-erratic.What we need here
either smoothor pure Joseph-erratic. is a distinction between a form of high-fre-
The paragonof the pure Joseph-erraticis a quency effect--namely 'short lag' or 'short run'
processwith a Gaussianmarginal distribution. effects--and a form of low-frequency effect--
We shouldtherefore,in every case,begin by namely 'long lag' or 'long run' effects.Short lag
checkingwhether the Gaussianmarginal dis- effectsdependupon the valuesof C(s) for a few
tribution applies.For this, in a first approxima- small valuesof s. Long lag effectsdependupon
tion, 'probabilitypaper' plots are goodenough. the other values of C(s). We shah now examine
They evidently showthat it is not exceptional this dichotomy on four examples.The first is
that the marginal distributionbe either nearly the processof independentincrements,whose
Gaussianor highly non-Gaussian. To stay near covarianceC•(s) satisfiesC•(s) -- 0 for all s
the land of Joseph,an exampleof nearly Gaus- 0. The secondis the Gauss-Markov processof
sian marginal distribution is provided by the covariance Ca(s) -- exp(--]sl/s). The last two
level of the Nile at the Rhoda Gauge near are the processesof covariancesrespectively
Cairo, an example of highly non-Gaussianby equalto Cs(s)-- (1 -]- Isl2/s•) -• and C•(s) --
the annual dischargefrom Lake Albert. To find (1 -[- Isl/s•)-ø'5, wheres•.,s•ands•areconstants.
other examplesof either behavior,it sufficesto The abovefour covariancesdiffer considerably
thumb through Boulos [1951]: Straight line from each other in the long run, but C2(s),
interpolations are quite acceptable in certain C3(s), and C•(s) are all three smoothand mono-
cases,poor but perhapsbearablein someother tone in the short run. Therefore, if the sample
cases, and dreadful in still others. Much less duration is short, and the sample covariance
completedata are availablein other places,but correspondingly'noisy,' the graphs of C•(s),
the existenceof huge deviations from the Gaus- C•(s), Cs(s), and C•(s) may be undistinguish-
sian is very familiar' For example,runoffsdue able, not only to the eye but even from the
to major storms may appear on histogramsas viewpoint of many tests of statistical signifi-
914 MAAIDELBROT AND WALLIS

cancethat examine each value of s singly. That var lAX •] ~ s•'5,where• means'asymptoti-
is, such statistical tests are liable to indicate callyproportionalto.' More generally,let C(u)
that the differencesbetweenthe samplecovari- • u2u-•for large u, with 0.5 < H < 1. Then,
ance and either of the functions C•(s), C.•(s), for large s, var [AX •] --• s2s.Incidentally,as-
C•(s), and C•(s), are not statisticallysignificant sumingimplicitlythat Z•:o•C(u) < oo, G.I.
for most s. The statistician could then conclude Taylor suggested
this infinite sum as a measure
that all the data will be acceptablyfitted as of the spanof temporaldependence in a time
soon as short run data have been fitted. There-series.The Taylor span is thus infinite for
fore, he will advise the hydrologistthat there C•(s), finite and easyto estimatein caseslike
is no evidencethat his data were not generated C•(s) or C•(s), wherethe seriesZ•C(u) con-
by an independentGaussprocess(C•), or a vergesrapidly, and, finally, finite but hard to
Markov-Gauss (C2), or perhaps some more estimate in caseslike C•(s), where the series
involved short-memory process. Z•=o•C(u)
converges
slowly.
This would, however, be a very rash con-
TI-IE RANGE, TI-IE JOSEPH EFFECT, AND
clusion.For example,under the hypothesisthat
I-IURST'S LAWS
the true covarianceis C•(s), one would expect
the relative proportionof positiveto negative Curiously,empiricaldata aboutthe behavior
samplecovariancesto be about one. This pro- of vat [,AX•] in hydrologyhave been exam-
portion would be larger under the hypothesis ined only after thoserelative to anothermeas-
C.o(s), still larger under C•(s), and larger yeture of over-all behavior of a process,namely
under C4(s). Thus, if statisticalcriteria gearedR(s)/S(s), where the sequentialrange R(s)
towardslow-frequencyeffectscan be developed, was defined earlier to be the capacity of a
it is reasonableto expect them to show the reservoircapableof performing'ideally' for s
same data to be significantlycloserto C3(s), or yearsand S(s) to be the standarddeviationof
to C•(s), than to C2(s). yearly flow. Among Gaussianprocesses,the
Our need, then, is to enhancelong-run prop- dependence of R (s)/S(s) upon s sharply dis-
erties of a processwhile eliminatingshort-run tinguishes smooth from Joseph-erratic pro-
wiggles.To do so, the best procedure.is to in- cesses.This is alreadyobviousfor Joseph'sown
tegrate or to use moving averages (fancier exampleof the sevenyearsof highprecipitation
averages will not be consideredhere). Three followedby sevenyears of drought,for which
methods of dealing with long-run effects de- the ideal reservoir would have had to be enor-
serveto be singledout. mous. If wet and dry years alternate at any
pointof a record,thenidealreservoir
sizeis
TI-IE VARIAI•CE OF CUMULATED FLOWS
decreased. It is the task of mathematics to
Let the 'accumulated flow' since time 1 be expressthis reductionin numbers.
definedas Z•q• X(u) and designatedas X•(t). First considerthe casewhen X(t) is an in-
Then, G.I. Taylor's formula (see Friedlander dependentGaussprocess. Then,whens is large,
and Topper [1961]) can be used to evaluate bothR(s) andR(s)/S(s) equalN/s,multiplied
the variance of AX • X •(t + s) -- X •(t) -- by some 'universal' random variable inde-
Z•:,,• • X(u) -- X(t + 1) ... • X(t +s). This pendentof s. The little that is knownabout
variance is given by sC(O) • 2 Z• • (s -- u) those random variables is due to Feller [1951].
C(u), which immediately introduces a basic For the .Gauss-Markovprocessand for other
long-run dichotomy. models for which the memory Z•_-o'C(u) is
When Z• © C(u) • •, var lAX •] -- var finite,the 'N/slaw' remainstrue, but the multi-
[X•(t + s) -- X•(t)] is found to be asymp- plyingrandomvariablesare changed.
toticallyproportional
to sZ•_•C(u),and X(t) The case of series exhibiting the Joseph
is found to be in the BrownJan domain of at- Effect is entirely different. For such series,
traction. the N/s law fails, as was first notedby Harold
When, on the contrary, Z•:o•C(u) diverges Edwin Hurst [1951, 1956, 1965]. For hydro-
sufficiently rapidly, var [.AX•] grows faster logicalseries,as well as for many othernatural
and X is not in the Brownian domain. When, time series,R(s)/S(s) increases
like Cs•. Here,
for example, C(u) -- C,(u), one finds that C and H are positiveconstants;H may range
 Operational Hydrology 915
between 0 and 1 and is seldom near 0.5. We RELATIOAl' BETWEEI• THE JOSEPIE EFFECT,
shall call this empirical finding 'Hurst's law.' HURST'SLAW, AAI'DTHE LOl•G RUN
Moreover, •/var[.AX•], consideredearlier, is BEHAVIOR OF THE COVARIAAI'CES

also proportionalto ss rather than to sø", as

To accountfor the abovelisted findingscon-
suggested by the usualsimplemodels.This will
be called 'Hurst's law for the standard devia- cerningvar(AX*), R(s)/S(s), or the spectrum
tion,' or 'Langbein'scorollaryof I-Iurst'slaw,' has proved to be very hard. For example,
becauseit was first noted in Langbein's com-
perusal of the discussionof Hurst [1951] and
Hurst [1951] demonstrates the kind of des-
ments on Hurst [1956].
perate expedient necessaryto fit his finding
Strictly speaking,Hurst claimeda more de-
within the universe of the simple statistical
manding'one parameter ss law,' accordingto
models. Claiming (incorrectly, as we shall
whichR•s)/S(s) is asymptotically
equalto
demonstrate presently) that there exists no
(s/2)s. His reasonsfor claimingthat C • 2-s
stationary random processwith a range follow-
are not too clear or convincing. Moreover,
ing the ss law, severalamongthe discussants
separateselectionof H and of C can obviously
ensure a better fit. It also yields a different
have suggestedeither giving up statistical sta-
tionarity or invoking nonrandom 'climatic'
estimate of H. For example, Ven Te Chow
changes.
in his discussionof Hurst [1951] found a case
where H goesup from 0.72 to 0.87 when C A more hopeful reaction, already mentioned
is separatelyestimated.Also, we found cases in the partisan commentat the beginningof
where the best estimate of H is below 0.5, which this paper, is exemplified in such works as
contradicts Hurst's assertion that 0.5 ( H ( 1. Ants and Lloyd [1953] and Fiering [1967].
See Mandelbrot and Wallis [1968c] for revised These authors, and others, have constructed
values of H. stationary stochasticprocessesof the usual
Note also that Hurst's original'(s/2) • law' kind (i.e., in the Browntan domain of attrac-
has provendangerous. In somecasesit tempted tion, satisfyingZ•C(u) • •) for which
him, as well as other authors,to estimateH both rangeand standarddeviationare propor-
from a single sample of natural or simulated tional to ss over a finite span of valuesof s.
values of X(t). Such estimatesshouldbe dis- But the usual•/s behaviorstill appliesbeyond
carded. The revised statement we use means
this span.Thus, the ss law is here a property
that estimationof H requiresmany values of s of what we have called a transient span. This
and, for every value of s, a large number of transient may be made arbitrarily long. But
starting points t spread over the total sample long transientscan only be achievedwith com-
of length T. plicated processeswith a long memory. For ex-
On the other hand, every specificmodel of
ample, Fiering [1967] (p. 85) had to use an
the JosephEffect, suchas the fractionalnoise
autoregressivemodel with 20 lags to ensure
to be describedin the sequel,will yield a rela-
that Hurst's law holds over the span I • s •
tion betweenC and H, whoseconformitywith 60.
experiencewill test the value of the model.
An alternative to this approach is based
SPECTRAL AI•ALYSIS: PRIAI'CIPLE AAI'D APPLICATI01• upon the existenceof the self similar random
I1• IEYDROLOGY processes,pointed out by Mandelbrot [1965]
In addition to the behavior of var (•iX •) and examined below. For such processes,
Hurst's law holds for all values of s. Even more
and of Hurst'srange,a third way of lookingat
low frequencyphenomena is throughspectral important from our viewpoint, which empha-
(or Fourier or harmonic) analysis.We only sizes low frequency phenomena, is the exist-
mentionit here to say that the spectraldensity ence of processes for which Hurst's law holds
of hydrologicalrecordspeakssharplyfor very for the short as well as for the long run. For
low frequencies,as is alsothe casefor the so- the standarddeviation,this was already proved
called !:• noises[Mandelbrot, 1967a]. A full whenwe notedin passing that
discussion of this topic is postponedto the
next paper. %/var[AX*]•, s•
•1• •/fANDELBROT AND •ALLIS

whenever it easierto generalize'We claim that the in-

dependentGaussprocessX(t) is so convenient
C(s),'• s2•-•' because
of the possibility
of interpolating
The sameasymptoticbehaviorcan be shown •:• X(u) to continuous timeswith the help
to hold/or the rangeR(s) and the rescaled of a 'self-similar'
randomprocess
B(t), called
rangeR (s)/S(s). Browntanmotion (also calledBachelierpro-
cess,or Wiener process).To define self-simi-
Thisobservationis central
to ourstudyof larity,onemustconsider a portion of B(t),
theJoseph Effect.Before
weexamine it more witht varyingfrom0 to T, andrewriteit as
closely,
let us showhowit .canexplain the B(h T), withh varying from0 to 1. 'Self-
existence
ofmodels in which
Iturst's
lawholds similarity'
thenexpresses the fact that the
in an initialtransient.
The key is that the rescaled
function T-ø'SB(hT) hasthe same
values
ofvar[X•(tq-s) -- X•(t)] fors • T distributionforeveryvalueofT.
areaffected onlyby thevaluesof C(s) for Fromthis,oneimmediately deduces that
s • T. Hence, changesin the covariancefor
s • T willpass
unnoticed
whenonlythespan %/var[B(t+ s) -- B(t)]= sø'5 and
s • T is observable.Now supposethat, start-
ing from (say) the covariance
C(s) - (1 + max B(t •- u) -- min B(t •- u) -- Csø'•,
10s)•-•, long lag covariances(s > T) are
decreasedsufiqcientlyto make Z•_-o•C(u) con- with C a constant.These statementsare forms
vergent. The modified processX(t) is thus of the sø'5law, but they are valid uniformly
'brought back' into the Browntandomain of (that is, for all s) rather than asymptotically
attraction. It could even be made into a (that is, for highs).
'multiple lag' autoregressive
model, which is By analogy, whenstudyingthe lawsof Hurst,
the usualgeneralizationof the Markov model. it is good to know that more generalself-
For suchmodifiedprocesses, Iturst's law con- similar processes exist. A Gaussianprocess
tinuesto hold for s < T, and for sometime X(t), of integralBa(t) = j'o'X(u)du,is Joseph
beyonds = T. On the long run, however,it self-similarif the rescaledfunctionT-•Bs(h T)
will be replacedby the x/s law. For example, is independent
of T in distribution.
That the
the standarddeviationwill equal Cx/s. The ss lawsapplyto Ba(t) canbe seenby simple
value of this constantdepends uponthe tail inspection.
Ba(t)was called'fractionalBrown-
selectedfor the modifiedcovariance.It is ad- tan motion' by Mandelbrot and Van Ness
justable at will and quite arbitrary. [1968].
We considersuch models,in which T plays Unfortunately, the derivative But (t), called
a central role, to be undesirable. 'fractional Gaussian noise,' is too irregular to
be studied directly. As we interpolated the in-
DEFINITION OF SELF-SI•VIILAR ERRATIC
tegral of the independent Gauss processby
GAUSSIAN PROCESSES
Browntan motion, we must now replace X(t)
In criticizingthe usualstatisticalmodels,as by B•(i q- 1) --B•(t). The covariance
of this
appliedto hydrology,we don't underestimate functionB•(t •- 1) --B•(t) is givenfor s •_
their good features. In particular, wherever by
the independentGaussprocessis an acceptable
approximation, it is unbeatable. Where one C,(s) = C[(s- 1)•"-+-(s-+- 1)•"
must amend this process,there are features
with 0 <: H <: 1. If % • H • i and s is
that one will want preserved.For example,all
large, onefinds
the models of the Browntan domain preserve
the assumptionsthat the variance is finite, and
C,s) • [2H(2H -- 1)C]s
Taylor's scale [defined as Z•X(u)] is also
finite. But they destroy another property that This is preciselythe form we have proposedto
makes the independentGaussprocessso con- use to model phenomenaobeyingIturst's law.
venient to manipulate. We shall now express Therefore, our models are approximationsto
this property in an indirect way that will make fractionalGaussiannoise.
 OperationalHydrology 917
APPLICATIONS OF SELF-SIMILARITY sary to explain Hurst's findings, and in the
third paper that it is insufficient.
To apply self-similarity, one proceedsvery
much as in the classical'dimensional analysis'
of fluid and solid mechanics. This is no acci- REFERENCES

dent. Ifydrology can be consideredthe low Anis, A. A., and E. H. Lloyd, On the range of
frequencyapplicationof the theory of turbu- partial sums of a finite number of independent
lence,in which self-similarityand dimensional normal variates, Biometrika, •0, 35, 1953.
analysiswere introducedby yon Karman and Boulos,Naguib, Probability in the Hydrology o[
The Nile, Cairo, Government Press, 1951.
Kolmogoroff (see Friedlander and Topper Feller, W., The asymptotic distribution of the
[1961]. rangeof sumsof independentrandomvariables,
An illustration of dimensionalanalysis is en- Ann. Math. Star., 22, 427, 1951.
counteredwhen dealing with flows of water in Fiering, Myron, StreamflowSynthesis,Cambridge,
vesselsof the same shape but of different di- Harvard University Press, 1967.
mensionsand at proportionatelydifferentveloc- Friedlander, S. K., and L. Topper, Turbulence'
ClassicPapers on Statistical Theory, New York,
ities. Then the calculationsneednot be repeated. Interscience, 1961.
It sufficesto solve all relevant problemsonce. Hurst, H. E., Long-term storage capacity of res-
Solutions to other cases will be obtained by ervoirs, Trans. Am. Soc. Civil Eng., 116, 770,
1951.
mere rescaling.
Hurst, H. E., Methods of using long-term storage
The application of self-similarity to hydrol- in reservoirs, Proc. Inst. Civil Eng., Part 1,
ogy is in the same spirit. Once one has per- 519, 1956.
formed the calculations relative to some 'refer-
Hurst, H. E., R. P. Black, and Y. M. Simaika,
ence'horizon,answersrelative to other horizons Long-Term Storage, an Experimental Study,
can be obtainedby simple rescaling.This may London, Constable, 1965.
make it worthwhile,in the caseof the reference Mandelbrot, B., The variation of certain •ecula-
rive prices, J. BusinessUniv. Chicago, 86, 394,
horizon, to perform some very lengthy and 1963; reprinted in The Random Character ol
involved calculations that would not otherwise Stock Market Prices, Paul /5. Coother (ed.),
be economic. The new hydrology we propose M.I.T. Press,Cambridge, 1964.
may demand readjustmentsof thought. But Mandelbrot, B., Une classede processusstochas-
tiques homothetiques k sol; application k la
there is hopethat the ultimate outcomeof this loi climatologiquede H. E. Hurst. Compt. Rend.
new hydrologywill be a set of new and better Acad. Sci. Paris, 260, 3274, 1965.
'cookbookrecipes.' Mandelbrot, B., Forecasts of future prices, un-
biased markets and 'martingale' models, J.
Business Univ. Chicago, 89, 242, 1966.
FORETASTE OF A DISCUSSION OF THE
Mandelbrot, B., Some noises with 1/f spectrum,
NOAH EFFECT
a bridge between direct current and white
A discussion of the Noah Effect is several noise, Inst. Electrical Electronic Eng., Tra.•s.
Information Theory, 13, 289, 1967a.
papersremovedfrom the presentintroductory Mandelbrot,B., How long is the coas•of Britain?
article. Our approachwill resemblethe methods Statistical self-similarity and fractional dimen-
Mandelbrot [1963] usesto describethe varia- sion, Science,155, 636, 1967b.
tion of commodityprices. The above consid- Mandelbrot, B., Sporadic random functions and
conditional spectral analysis; self*similar ex-
ered function X•(t) -- Z•_•X(u), with X(u) amples and limits. Proceedingso• the Fi•th
the annual flow for the year u, will be the Berkeley Symposiumon Mathematical Statis-
counterpartof the price of a commodityat the tics and Probability, edited by L. M. LeCam
instant t. The very rapid and large changes and J. Neyman, Vol. III, 155, 1967c.
typical of the behaviorof priceswill be com- Mandelbrot, B., Long-run linearity and locally
Gaussianprocesses;roles of J-shaped spectra
paredto floods.Sporadicprocesses
(seeMandel- and of infinite variance,Intern. Econ. Rev. (in
brot [1967c]) will also be needed. press), 1968.
The Noah Effect certainly raises important Mandelbrot,B., and J. W. Van Ness,Fractional
problemsin operationalhydrology.Suchprob- BrownJan motions, fractional noises, and ap-
plications,SIAM Rev. (in press),1968.
lems are, however,separatefrom thoseraised Mandelbrot, B., and J. R. Wallis, Computer
by the JosephEffect. In the next paper we experimentswith fractional noises,Water Re-
shall show that the Noah Effect is not neces- sources Res. (in press), 1968.
918 MANDELBROT AI•D WALLIS

Mandelbrot, B., and J. R. Wallis, Robustnessof
the range statistic, as a measure of long-run
interdependence,Water ResourcesRes., to ap-
pear, 1959a.
Mandlebrot, B., and J. R. Wallis, Some long
run properties of geophysical records, Water
ResourcesRes., to appear,1969b.
Mass•, P., Les r•serves et la r•gularization de
l'avenir dans la vie jconomique, 2 vols., Paris,
Hermann, 194õ.
Yevjevich, V. M., Misconceptionsin hydrology
and their consequences,Water Resources Res.
4, 225-232, 1968.
(Manuscript receivedMay 17, 1968;
revised July 1, 1968.)