You are on page 1of 8

WATER RESOURCES RESEARCH, VOL. 29, NO.

I2, PAGES 3995-4002, DECEMBER 1993

The Hurst Effect: The Scale of Fluctuation Approach


OSCAR J. MESA AND GERMAN POVEDA

Programasde Postgradoen Aprovecharnientode RecursosHidrauticos, UniversidadNacional de Colombia, Medellin

After more than 40 years the so-called Hurst effect remains an open problem in stochastic
hydrology.Historically, its existencehas been explainedeither by preasymptoticbehaviorof the
resca!edadjustedrangeR'n, certainclassesof nonstationarity
in time series,infinitememory,or
erroneous estimation of the Hurst exponent. Various statistical tests to determine whether an
observedtime seriesexhibitsthe Hurst effectare presented.The testsare basedon the fact that for
thefamilyof processes
in theBrownian
domain
of'attraction,
R*n/((On))•/2
converges
in distribution
to a nondegeneraterandom variable with known distribution(functional central limit theorem). The
scaleof fluctuation0, definedas the sumof the correlationfunction,plays a key role. Applicationof
the teststo severalgeophysicaltime seriesseemsto indicatethat they do not exhibit the Hurst effect,
althoughthoseserieshave beenusedas examplesof its existence,and furthermore the traditionalpox
diagrammethodto estimatethe Hurst exponentgivesvalueslargerthan 0.5. It turned out that the
coefficient
in therelationofR*nversus
n, whichis directlyproportional
to thescaleof fluctuation,
was
more important than the exponent.The Hurst effect motivated the popularizationof 1/f noisesand
related ideas of fractals and scaling.This work illustrateshow delicate the proceduresto deal with
infinity must be.

I. INTRODUCTION lems such as turbulence (see, for instance, Mandelbrot


[1974]).
The Hurst effect is one of the most important unsolved
problems
in stochastichydrology.There is ampleevidence 2. ANTECEDENTS
tosuppo_rtthis statement.Ever sinceHurst' s [1951]original
work,there has been a proliferation of papersabout it. Some Let X•, X2, "', Xn be a sequenceof randomvariables
ofthe most important awards of the hydrologiccommunity representing,for instance,the inflows into an infinite reser-
havegone to contribution toward its solution. Hydrologists voir. Denote the partial sum series(cumulativeinflows)by
havebeen divided into two schoolsin the attempt to inter- t

pret the alleged existence of this anomaly in geophysical


records.Some of the most famouspersonalitiesin probabil- So=O St= • Xm t= 1,2,'", n. (1)
m=l
ity theory, like Feller [195!], have devoted time to this
problem.Mandelbrot [1982] has declared that his original The sample mean represents the ideal release from the
investigationinto the Hurst effect was one of the sourcesof reservoir and therefore the adjusted partial sum sequence
inspirationfor his, now in vogue, fractal theory, whose S•, definedby
importancein chaos theory is Without doubt.
Despitethe activity cited above, the problemin hydrology S• = St - (t/n)Sn t = O, 1, 2, ..., n, (2)
is stagnant.There are no clear winners between the short- representsthe fluctuations in the content of this ideal reser-
andlong-memoryschools. Hurst's original motivation was voir. In his studies for the Aswan Dam on the Nile River,
thepracticalproblem of reservoir design,but it was later Hurst [1951] considered the "adjusted range," defined by
discovered that only in someparticular casesdid the so-
calledHurst effect have significantpractical implications Rn = max S'•- min S• for 0 -< t <_n. (3)
[Klemdset al., 1981]. Nevertheless,economicimplications
He showed that the adjusted range is a measure of the
of the use of differenthydrologicmodelsare not insignifi- reservoir capacity required under idealized conditions and
cant,andthe way the persistenceis modeledis very impor- therefore the study of its properties becomesvery pertinent.
tantfrom an economicpoint of view [Zapata, 1987;Mejfa He was particularly interestedin the dependenceof R n on
andMilldn, 1982;Pereira et al., 1984]. the samplesize n. Obviously, R,• increaseswith n, but how
We presentvarious tests for the existenceof the Hurst fast? The fact that the length of the existing streamflow
effect.Applicationsof theseteststo Hurst's originaldata records is rarely of the order of 100 years motivated Hurst
showsthat either there is no Hurst effect or the series is not
[1951] to look at various series of different geophysical
longenoughto providedefiniteanswers.Evenin the latter phenomena. For that purpose, he defined the "resca!ed
case, the Hurst effect is not a natural interpretation. In adjustedrange"
addition
to the importanceof the studyof Hurst'sgeophys-
icalseries,
theproposed
testmayproveto beimportant
in R*n= R•/Dn (4)
otherapplications
whereinfinitememorymodels(nonsum-
whichis a dimensionlessquantity;D n is the samplestandard
mablecorrelograms)havebeenproposedfor physicalprob-
deviationof the X i.
Copyright
1993by the AmericanGeophysical
Union. Hurst [ 1951]used 690 different time seriesof 75 geophys-
Papernumber93WR01686. ical variablessuchas temperature,rainfall, solar spot num-
0043-1397/93/93WR-01686505.00 bers, mud varves, tree rings,etc. His empiricalfindingswere
3995
3996 MESA AND POVEDA: HURST EFFECT

that, in general,R*ngrowslike n to a powerof the orderof standarddeviationof the Xi, say, or.In that case,R*• is
0.72 with some variations for the different records, but in all scaled
bycrbecause
it isthelimitofR*•which
isscaled
by
the casesthe exponentwaslargerthan0.5. This resultwasin Dn.
contradictionwith his own theoretical analysiswhich indi- A consequenceof the above definition and of the invari-
catedthat the exponentshouldbe asymptotically0.5. Feller ance principle is that for a sequenceto exhibit the Hurst
[ 1951] was quick to providefirm theoreticalcomputationsfor effect,it is necessarythat at leastone of the conditionsofthe
the caseof independentidenticallydistributed(IID) random functionalcentral limit theorem be violated. Becauseof the
variableswith finite secondmoments.He computedthe limit generality of these hypotheses something quite dramatic
mean and variance of the rescaled adjusted range, and must be happeningfrom the physical point of view anda
suggestedthat Hurst's [1951] empirical findings might be searchfor the cause of the violation is very much in order.
explained by some kind of Markovian dependence.The The discoveryof anomaliessuchas a nonstationarity or a
discrepancybetween the empirical observationshowingthe strong dependence in that large a class of geophysical
increaseof the rescaledadjustedrangelike a power of 0.7 or recordswill have profound physical implications.
so, and the theoretical expectation that for a wide class of As may be expected,from a retrospectiveviewpoint,the
processesthe exponent is asymptotically 0.5 has becometo explanationsproposed by different authors were relatedto
be known as the Hurst effect. The exponent H in the the violationof the hypothesesof the invarianceprinciple.
empirical relation There have been some theories essentially related to the
dependencestructureof the process,to its stationarity,to
a*n"- nH (5) the existence of an infinite second moment, and to the
preasymptoticbehavior of the limiting process.
is called the Hurst exponent.
The strongdependenceexplanation was initiated by Hurst
To appreciate the different argumentsthat have been put
[1951] himself, and by Feller's [195!] suggestionaboutthe
forward as possibleexplanationsof the Hurst effect, it is
Markovian character of the series being responsiblefor the
necessaryto use a more precisedefinition.Suchprecisionis
Hurst effect. However, soon Barnard [1956] showed the
important also for the proper understandingof the estimation
existence of a 0.5 convergence for that case. Nevertheless,
problem of the next section. To that end, we will use the
Matalas and Huzzen [ 1967]presentedreports of estimatesof
definition of Bhattacharya et al. [1983, p. 651]: "A sequence
H between 0.58 and 0.87 for simulation of first-order autore-
of random variables is said to exhibit the Hurst effect with
exponent
H > 0 5 ff (1/nU)R* converges
' as gressiveMarkovian models with different correlation coeffi-
in distribution,
t!
cients. At that time the Markovian explanation seemed
n goesto infinity, to a non-zero random variable." This is in
adequate, and the autoregressive models were becoming
contrast with a very general result, known as the invariance
fashionable.However, Mandelbrot and Van Ness [ 1968]put
principle or the functional central limit theorem [Ibragimov,
things back in proper perspective by recalling that for
1962;BilIingsley, 1968], which implies that under conditions
processesin the domain of attraction of the Brownian
of stationarity andweakdependence R*•/nø'5converges in
distribution to a randomvariableR*•/n0.5with mean motion, the rescaled adjusted range grows like n to the 0.5
power (notice that Barnard's result goes back to 1956,and
E(n-ø'SR*•)= (0•r/2)1/2 (6) that the invarianceprincipledatesback to Ibragimov [1962]).
Mandelbrot and Wallis [!968, 1969] and Mandelbrot and
and variance Van Ness [1968]proposedan explanationof the Hursteffect
by the strong dependenceof the geophysical seriesand
Var (n-ø'SR*•)= 0 (z-2/6- z-/2) (7) introduced the so-called fractional Brownian motion and the

where 0 is a positive constant, the so-calledscale of fluctu- fractional Brownian noise (1/f noise), which are processes
ation, or correlation length scale, a parameter first intro- with infinitememory, as modelsthat may be usedto simulate
duced by Taylor [1921]. It can be shown that 0 is the sum of the Hurst effect. These important theoretical contributions
the correlation function were not accompaniedby either a physical explanationnor
an investigationof the structure of dependenceof the
differentgeophysical series.The lackof physicaljustification
0= Z p(m) (8) of thistheory was pointedout by authorssuchas Scheideger
rn-- -• [1970]and Klem•s [1974]. There has not been appreciation
for thiskind of questionas the followingquote shows:"I am
where p(m) denotesthe correlationcoefficientbetweenXn preparedto arguethata lackof seriousmotivationina model
and Xn+ m. that fitsand workswell is muchpreferableto lack of fit in a
Equations (6) and (7) correspondto Feller's [195!] previ- model that seems well motivated" [Mandelbrot, !982, p.
ous result, for in that case 0 is one. Besides the mean and the 253].
variance, the whole asymptotic distribution of the rescaled On the estimation side, the contribution of Mandelbrot
adjustedrange are ready availablefrom known resultsin the and Wallis[1968]wasalsoimportant.They pointedoutthat
BrownJan motion case [Bhattacharya et al., 1983]. The the way Hurst [195!] estimated the exponent H was not
condition of weak dependenceimplies a correlation function adequate. Hurst [1951] used the equation
decreasingfast enoughto ensureconvergenceof the seriesin
(8). This explains the use of terms such as strong depen- H l = log(R*n)/log
(n/2) i9)
dence, infinite memory (infinite 0), and short memory(finite whichpresupposes
that the relation(R*) versusn in loga-
0). Normally, existence of finite second moments for the rithmicpaperpasses
throughthepointn = 2 R* = I This
sequence of Xi is assumed. Then the ergodic theorem wasmotivated by weakempiricalarguments,
becauseit is
applied
toX•2 implies
thatDn converges
inprobability
tothe easy to see that using the biased estimator for O'n, one
MESA AND POVEDA: HURST EFFECT 3997

obtains
R*n= 1 whenn = 2. Mandelbrot andWallis[1968] els were used in hydrology to model Hurst's exponents
the so-called"pox" diagram.This and other largerthan 0.5. The broken line model [Rodrlguez-Iturbeet
introduced
waysofestimating
H will bediscussedin thenextsection. al., 1972; Mejia et al., 1972] and the ARMA(1, 1) model
Attempts of [O'Connell, 1974]are the standardpractice. Indeed, Hipel
to explainthe Hurst effectas a consequence
aninfinitesecondmomentwereproposedby Moran [ 1964] and McLeod [1978]demonstratedthat ARMA(p, q) models
andBoesand Salas [1973]. However, Mandelbrotand statisticallypreservethe rescaledadjustedrangeor equiva-
Taqqu
[1979]demonstrated
thatanasymptotic with lently the Hurst exponentestimatedas (9). One can interpret
relation
//= 0.5 holdsfor a sequence of IID randomvariableswith the philosophyof thesemodelsas an attemptto increasethe
distributionand characteristicexponentstrictly less correlation length scale 0 so as to obtain preasymptotic
stable
than2 (seeFeller [1971, p. 169] for a discussionof stable estimates of the Hurst coefficient similar to the observed in
dist•butions). empirical records. The conclusionsof the chapter on the
Hurst [1951] recognized the nonstationaritiesin his origi- Hurst effect in the textbook by Bras and Rodriguez-Iturbe
nalgeophysical
series.He evendesigned
anexperiment
with [1984,p. 265]are very illustrativeof the standardpracticein
"probability
cards"whichproduced sudden
changesin the engineeringhydrology. This is most noteworthy if one
meanof the process,and obtainedempiricalestimationsof considersthat the physical problem is unsolved.
theexponentH near 0.71. Klemds[1974]andPotter [1975] In fact, there is no physical explanation yet for the
developed simulationswith nonstationary modelsthat pro- occurrenceof infinitememoryin geophysicalseries.There is
ducedempirical series exhibiting the Hurst effect. Some- no systematic(physicalor empirical)studyof the correlation
thingsimilarwasobtainedby Boesand Salas[1978]with the structureof theseprocesses.Even thoughthe nonstationar-
shiftinglevels model. It is worth noting that all of these ity of some geophysicalprocessesmay be argued on some
nonstationarymodels belong to the Brownian domain of physicalgrounds[Leopold et al., 1964, p. 61], it is only in
attraction. very generalterms and it remains to be explained why the
Bhattacharya et al. [1983] provided clear mathematical trendsproducethe sameHurst exponentin variousgeophys-
demonstration
of the existence of the Hurst effect for weakly ical series. Moreover, the issues of estimation raised by the
processesperturbedby smallmonotonictrends. preasymptoticexplanationsare very relevant and have no
dependent
Asan example, let Y• be a sequenceof weakly dependent definiteanswers.The importance of the estimationproblem
randomvariables, say IID normal variables with zero mean is reinforced if one considers the claims about the robustness
andunit variance, then of the range analysis in current literature on fractals [Man-
delbrot, 1982, p. 382; Feder, 1988, p. 194].
Xn -- Yn+ c(m+ n)• (10)
willexhibit the Hurst effect with exponent H dependent on 3. TESTS FOR THE EXISTENCE OF THE HURST EFFECT
•e valueof the parameter/3as follows:for - 1/2 </3 < 0, H
isequalto 1 + /3, for/3 > 0, the exponentH is 1, and for Notice that the whole puzzle rests on the empirical evi-
/3< -1/2, and for /3 = 0, the Hurst exponentis I/2 [see dence of the exponent H being larger than 0.5. Also, recall
Bhattacharyaet al., 1983]; notice the discontinuityat/3 = 0. thatthequestion isnotrelatedtominorthings: no.7isalmost
In all the cases,m and c are arbitrary parameters(c > 0 and 4 timesn0.sfor n = 1000, andthe factorkeepsincreasing
m -> 0). The significanceof this demonstrationis twofold. with n. With all the history behind but the perspective of
First, the class of processeswith infinite memory is no huge unsolved questions present, the least that can be done
longerthe only theoretically proved class of processes is to look further into the empirical evidence.
exhibitingthe Hurst effect. Second, the estimationproblem
3.1. Estimators of H
becomes very important, for it is possiblethat very small
trendsmay be responsiblefor the appearanceof the Hurst As was pointed out before, Hurst [1951] originally esti-
effect. mated H by means of (9), but this practice was shortly
In additionto the infinite memory and the nonstationarity abandonedand substitutedby the least squaresslope in the
explanations,there have been theories that present the linear relationof log (n) versuslog R* alongwith some
Hursteffectonly as a preasymptoticbehavior.This means other variations. How good are those estimators?Very little
theconvergence to the theoretical0.5 exponentis slow,and theoretical work has been done along these lines. As an
thereforethe empiricalobservations for finitesamplesizes alternative for the complex theoretical issues involved,
maygive Hurst exponentslarger than 0.5 [Lloyd, 1967]. extensivecomputer experimentswere performed by Poveda
SaMsand Boes [1974] consideredthe equation for the [1987] using Bhattacharya et al.'s [1983] nonstationary
expectedvalue of R n for finite n and the case of IID model(equation(10)), given its capacityto producevaluesof
variablesshowingthe preasymptoticbehaviorof the ad- H, at will, by fixing fl. This analysis consideredmost of the
justedrange,andproposed a differentway of estimation of estimators reported in the literature [see also Poveda and
theexponent H. Gomide[1975,1978]considered the caseof Mesa, 1993]; Table 1 presents Poveda's [1987] results for
MarkovJan processes and showedhow the preasymptoticone caseand someof the estimators.The performanceof all
regionis expandedwith valuesof p(1) near one, and pro- estimatorsin all caseswas poor. Slightly better resultswere
poseda new way of estimationof the exponentH. Salas et obtainedwith a new estimator proposed by Poveda [1987],
al. [1979a, hi, using autoregressive movingaverage which consistsof the slope Sn of the regressionof sample
(ARMA)(1,1)models,showed region valuesof R*mversusm takingonlyvaluesof m largerthann,
thatthepreasymptotic
isexpanded because of eitherasymmetricmarginal
distribu- but some degree of arbitrariness remains regarding the
tion,largebutfinitememory(large0) andnonstationarity.choice of n.
It is worth notingthat parallelto the preasymptotic A conclusionof these computerexperimentsis that all the
explanation
of the Hurst effect,variousshort-memory
rood- estimatorsof the Hurst's [1951] exponent H performed
3998 MESA AND POVEDA: HURST EFFECT

TABLE 1. EstimatedValues of the Hurst ExponentH for a 3.3. Statistical Tests


SimulatedSequenceof the Bhatthacharya et aI.'s [1983]
Nonstationary Model, Equation (10), With The visualtest in the GEOS diagrammay be improved
/3 = -0.3, c = 1, and m -- 1,000 and
20,000 Record Length
substantially
if an independent
estimateof the limit of the
sequence
{R*n/n
0.5}isknown.
Underthehypotheses
ofthe
Wallis and functionalcentrallimit theoremthis limit is a randomvail.
Hurst Matalas Gomide Poveda
able with known distribution; therefore samplevalues
[ 1951] [ 1970] [ 1975] [ 1987]
shouldbe aroundthe mean,with deviationsof theorderof
5 0.7172 0.1343 0.5408 thestandard
deviation.
In fact,forshortmemory
stationary
10 0.6901 0.6544 0.3773 0.5370 processes
themeanandthevariancearegivenby (6)and(7).
25 0.6571 0.6216 0.4137 0.5357
As a consequence,
standardstatisticaltechniques
maybe
100 0.6195 0.5889 0.5487 0.5399
250 0.6049 0.5760 0.4788 0.5508
employedto test the hypothesisof absenceof the Hurst
500 0.5939 0.5661 0.4660 0.5760 effect. The only extra parameter needed is the scaleof
1,000 0.5753 0.5517 0.5020 0.6257 fluctuationor correlationlength scale 0. Clearly, because
of
2,500 0.5563 0.5346 0.4966 0.7086 the emphasison the exponent previous studies have over
5,000 0.5474 0.5223 0.4710 0.7998 looked the proportionalityconstantin the asymptoticex.
10,000 0.5637 0.5248 0.5037 0.7921
pression(5).
15,000 0.5855 0.5372 0.5198 0.4528
20,000 0.5812 0.5421 0.5177 3.3.1. Test1. Given a sequenceX1, X2, ßß' of random
variables,with known scale of fluctuation0, a sample
sequence x l, x2, "', and a level of confidence a, the
sequencedoes not exhibit the Hurst effect if the sample
poorly. This contradicts the alleged robustnessof the range values
ofR*n/n
0.5remains
intheinterval
(q,,
- q,+) forlarge
analysis [Mandelbrot, 1982, p. 386]. enough
n (whereq• andq•+ arethe 1 - od2andaY2quan-
tilesoftheasymptotic
distribution
ofR*•/n0.5,respectively).
In practicalapplicationsq,- and q,+ canbe approximated
3.2. Visual Tests
by the mean asymptotic value (equation (6)) ---2.3timesthe
Thebehaviorof R*n/n
H (H = 0.5 andH > 0.5) is the standarddeviation(squareroot of (7)). The value2.3 surely
most natural thing to examine to test for the Hurst effect in exceedsthe values correspondingto the confidencelevelof
a geophysicaltime seriesinstead of the logarithmicregres- 0.95 in the asymptotic distribution. More precise valuesof
sion of R*n on n. A useful set of diagramswas designedfor q,- and qa+ maybe computed if desired-
that purpose, the so-called "GEOS" (geophysicalrecord) Test 1 is an immediate consequenceof the definitionof the
diagrams , with n on the abscissaand R*n/n0.5 on the Hurst effect. Obviously, a test with a not known valueof 0is
ordinates.
Recallthatif thereexiststheHursteffectR*•/n0.5 needed. Various alternative ways for estimating 0 from
will eventually diverge to infinity, whereas if there is no stationary random time series have been presentedin the
HursteffectR*n/n
o.5willconverge
toafinitelimit,withsmall literature [Vanmarcke, 1988, p. 327]. A short summaryis
random variation around it. Therefore samplepoints for a presented next.
time series which exhibits the Hurst effect will increase The first procedure is by means of the samplecorrelation
indefinitely in the GEOS diagram. Instead, a time serieswill function using (8). However, this estimator is inconsistent,
not possessthe Hurst effect when its G EOS diagram con- since its variance does not vanish when the record length
verges to a finite limit. becomesvery large; indeed, it exhibits a high coefficient of
In a similar way, it is possibleto check the convergenceof variation [Vanmarcke, 1988, p. 325]. On the other hand,
R*n/n
•, H > 0.5, byscaling
theverticalaxisbynH. In this consistent estimators of 0 can be obtained by usingthe
case a geophysical time series which exhibits the Hurst variance function F( ) and the known fact that under a
effect, with exponent H, will converge to a nonzero limit, condition of weak dependence(finite first moment of
whereasa time serieswithout the Hurst effect will converge [Vanmarcke, 1988, p. 188]) the scale of fluctuationis als0
to zero. given by
GEOS diagrams are visual tests for the existence of the
0= lim TF(T). (11)
Hurst effect in any time series with a long enoughrecord. T---> oc
These diagrams become more powerful tools than the so-
called pox diagrams.This superiorityis due to the fact that Recallthat the variancefunctionF(T) is simplythe variance
the diagramis scaleddownproperly, not only with respectto of the T averageof the original process.
the meanof R* but alsowith respectto the varianceand Nevertheless,the ordinaryvariancefunctionestimator is
other moments as well. Therefore deviations from the ex-
biasedand a correctionis requiredfor the estimation of 0.
pected behavior have the proper significancethroughthe FollowingVanmarcke [1988,p. 336],the expected valueof
whole range of n values. Besides, no slope estimationis the estimator of the variance function for an n long zero
involvedand, as was pointed out, factors of the order of 4 or mean unit variance sample is
more are involved providing a magnifyingview that should
help discriminate the existence of the Hurst effect. In r(r) - r(n)
applications, the main limitation of these visual tests is that r_< n.
1 - F(n)
if the lengthof the recordis not longenoughit may not be
easy to draw definite conclusions. For instance a conver- This motivates a corrected estimate F*c as followsß
gencefrom below may be wrongly interpretedas a continu-
ous increase. r'AT) = r(n) + - r(n)]. •13}
M•.s^ AND POVEDA: HURST EFFECT 3999

8'

o
.................................
;...............
..........
........
.............................................
2 o []

0
' ' "" '-' "i0 ....... ....... 0 20 40 60 80 1O0 120
n

Fig. I•. GEOS dia•m for a time se•s of tree •ngs of a Fig. 1c. GEOS diagram for St. Lawrence river discharges.
•u•ss Hr(Snake•ver). HoHzont• dashedlinescodespondto the
•ymptoticme• •d •2 stand•d deviations.
-
the factorsmultiplyingq,• and q, are affectedby the new
situationof 0* being an estimate. In a processexhibiting the
SinceF(n) is not known, finding an estimate 0* of the
scaleof fluctuation will require some iteration. Using an
Hursteffect,theincrease
of R*•/nø'5withn will eventually
dominate. The hypothesis of the test is also weak depen-
approximate
modelfor the variancefunction(F(T) • O/T,
dence and stationarity.
forlargeT) provides the following expression[Vanmarcke,
Notice that the estimation of 0 is by itself a test of the
1988,p. 337]:
existence of the Hurst effect for stationary processes. In
F*(T)Tn fact, because of the functional central limit theorem, if 0 is
O* = T<n. (14) finite the exponent h is 0.5 and there is no Hurst effect.
- r*(r)]n- T
Otherwise, if 0 is infinite then there is Hurst effect and it is
Anotherpossibleway of estimating 0 is by means of the not necessaryto perform the tests. In fact, for long memory
one-sideunit area spectral density function at zero. Also, if time series the estimation of the scale of fluctuation in (14)
a shortmemory theoretical model is adjusted to the data, 0 does not converge to a finite limit and therefore test 2 is not
couldbe estimated from the model, according to the theo- suitable. For the nonstationary process of Bhattacharya et
reticalexpressionsfor 0 in terms of the parameters of the al. [1983] the estimation of 0 by any of the means presented
model. above also show divergence to infinity, in concordance with
the theoretical result about the existence of the Hurst effect.
3.3.2. Test2 (Outline). Supposethe scaleof fluctuation
0 is not given, estimate it by means of 0* using any of the For those processestest 2 is not applicable either.
methodsdiscussed above, estimate also the size of the In view of the above, the recommendation to deal with an
variance of R*n/(O*n) ø'5 andperforma testin classical observed time series is to proceed first to the estimation of 0
terms. Even though some technical details need to be usingfor instancethe variancefunction approachof (14) (see
workedout, the test is in the same spirit of test 1. For this Figure 3b). Stabilization of the estimator with n indicates
reason,formal substitution of 0* for the scale of fluctuation finite memory and test 2 may be performed. If there is no
intest1 is proposed.The idea is that this will not affectthe stabilizationthere may be three possible causes:the series
powerof the test significantly.Only small modificationsin comesfrom an infinite memory process, or it comes from a
nonstationaryprocess or the length of the record is insuffi-
cient to estimate 0. As an easy cheek, if the value of n (the
length of the record) divided by the estimated 0 is less than,
say, 15, the record is short. In these cases, if possible, the
length of the record should be increased, and the estimation
of 0 repeated. On the other side, there are various ways of
testing and removing nonstationarities.If the problem re-
mains, no conclusion can be inferred from the data alone. In
fact, all extrapolationsof the behavior of either the range or
the estimator of 0 are equally arbitrary from a statistical
point of view, and any decisionshouldbe basedon physical
reasoning.
In a related problem, Burg [1967] observed that the
0.5 problemwith conventionalFourier spectralanalysisof finite
time seriesis that only a finite numberof correlationlagsare
estimableand that the truncation in lag spaceresults in a
0 5• 1•)0 "1• 2(•)0 '2•0 300 smoothingof the true spectralfunctionin frequencyspace.
n
Burg [1967]arguedthat the criterion for extrapolationshould
Fig.lb. GEOSdiagram
fortemperature
incentral
England
series. be to obtainthe spectraldensityestimatethat correspondsto
4000 MESA AND POVEDA:HURSTEFFECT

TABLE 2. Estimationof H Accordingto Siddiqui[1976] andMcLeod[1978]on ARMA(p, q) modelsfittedto some


Hipel and
ofthoseseriesallowedestimation of 0 using(15).In Figures
SeriesCode n 0 Poveda [1987] McLeod [1978] l a, lb, and lc threeof thoseGEOS diagrams areshown.
Forthemajority ofthecases,samplevaluesofR*n/n o..•seem
Mstouis 96 1.6 0.451 0.591 to indicateconvergence to the asymptotictheoreticaldistil.
Neumunas 132 1.4 0.499 0.591
Danubio 120 1.0 0.495 0.495
butionof the fittedARMA(p, q) model.Moreover,estima.
Rhin 150 1.0 0.4984 0.484 tionof the Hurstexponent
H in the way suggested
by
Odgen 97 12.1 0.436 0.929 Siddiqui[ 1976]alsoshowsthe lackof existenceof theHurst
Gota 150 1.6 0.504 0.636 effectinthesetofgeophysical
timeseriesanalyzed
byHipel
Espafiola 350 44.8 0.455 0.927 andMcLeod[1978](seeTable 2). Accordingto theestimated
Temp 255 1.6 0.521 0.640
Precip 100 1.0 0.473 0.473 scaleof fluctuationall the seriesbut OdgenandEspafiola
Minimum 848 24.6 0.462 0.746 haveratiosof n over 0 largerthan30, indicating
adequate
Snake 669 3.9 0.475 0.663 length of the records.
Exshaw 506 3.9 0.420 0.580 Nevertheless, time series correspondingto mud varyes
Naramata 515 2.1 0.435 0.543
exhibit GEOS diagramsthat always increasewith the value
Dell 655 3.8 0.475 0.667
Lakeview 544 5.9 0.499 0.706 of n (seefor instanceFigure 2), althoughthe asymptotic
Ninemile 771 7.4 0.466 0.642 value given by (6) for the fitted short memory modelis
Eaglecol 858 9.3 0.485 0.701 unknown in these cases. Two facts could explain this situa.
Navajo 700 2.9 0.468 0.584 tion:thetimeseriesis notlongenoughto reachtheasymp-
Brice 625 4.0 0.513 0.727
Tioga 661 3.5 0.498 0.691
toticmeanof R*•/nø'5or thosetimeseriesactuallyexhibit
Bigcone 509 3.6 0.404 0.691 the Hurst effecteitherbecauseof nonstationarity
or long
Whitment 1164 2.5 0.53 0.627 dependence.
For further illustration of the ideas presented,a 18,000
long series of vertical wind velocity sampledevery 0.I s
collected with a very precise instrument was analyzed.
the most random or unpredictablestochasticprocesswhose Figure 3a shows the pox diagram with the exponentH
correlationfunction is consistentwith the given information. estimatedusing traditional estimationindicatingthe pres.
Using the maximum entropy method he derived a widely enceof the Hurst effect (least squaresslopeof 0.773). Figure
used procedure for the estimation of the spectral density. 3b shows the estimation of the scale of fluctuation; 0 canbe
The techniqueis basedon obtaininga data modelthat is least estimatedto be of the order of 55 notwithstandingsampling
informative with respectto data that are not available. This fluctuationsand after observing stabilization. Figure 3c
maximum entropy method for estimation of the spectral showsthe GEOS diagram and the proposed test indicates
density is equivalent to the linear prediction method that that there is no Hurst effect that can be inferred from this
assumethat the underlyingdata model is an autoregressive record(asymptoticmean around 9.3). A straightline of slope
(finite memory) model [Roy et al., 1991]. 0.5 and interceptequalto ((0•-/2))1/2was alsodrawnin
Figure 3a for further clarification of the test. The least
4. EMPIRICAL EVIDENCE
squaresline in the pox diagram has both the slopeand
intercept free and fits the preasymptotic domain. The 0.5
At this point an obvious question arises, Do the classical slopeline hasno fittingparameterand is supposedto predict
time series which have been used to illustrate the existence the asymptotic behavior of the rescaled adjusted range.
of the Hurst effect really possessit? To elucidate the Nevertheless, some may extrapolate the 0.773 powerlaw
questionGEOS diagramswere plottedfor severalgeophys- increaseand othersmay considerthat the lengthof the series
ical time series. Basically, the formal version of test 2 was doesnot permitdefiniteconclusionsyet. In any case,thetest
performedusing GEOS diagramswith 0 estimatorsbasedon may be repeated if a longer series is available.
the parameters of short memory models fitted to the obser-
vations.Indeed, for ARMA(p, q) modelsin the senseof Box
7
and Jenkins [1970], the scale of fluctuation is given by
[Siddiqui, 1976]
6-
2

5-
1--Ea j
1 •4-
0* = -- (15)
2 •3-
1- qbi
i=1

where •'0 is the ratio of the variance of the processto the


variance of the noise, and qband a are the ARMA(p, q)
parameters. This procedure to estimate 0 was used because
the originalserieswere not available.Values of R*n versus n øo 6oo goo 2800
for variousgeophysicalserieswere taken from the extensive n

work by Hurst et al. [1965]. Additionally, resultsby Hipel Fig. 2. G EOS diagramfor Lake Saki mud varves.
MESA AND POVEDA: HURST EFFECT 4001

10000• 16

14'

12-

lOOO
10- Im

m c•

6-
100:

lO ' t i i i i
lO •oo ..... i'6'oo.... i 6• i •36ooo 00 3_ • •) • 1'0 1'2' 1'•' 1'6 1'8' 20
TIME(0.1SECONDS) TIME (0.1 SECONDS)
(Thousands)
Fig. 3a. Pox diagramfor a vertical wind velocity time series.
Thedottedline is a least squaresfit with slopeof 0.773,and the solid Fig. 3c. GEOS diagram for a vertical wind velocity time series.
lineisa 0.5slopewith((•r0/2))1/2
intercept.

gence to a finite limit, the tests presented here are not


5. CONCLUSIONS
conclusive.In these casesany extrapolation of the asymp-
To determine if a finite time series exhibits the Hurst effect totic behaviorof the rescaledadjustedrangeis not supported
isa delicatematter. Pox diagramsalone are not sufficientand by observationsand should be based on physical evidence.
furthertestsare proposedin that respect. Various statistical Application of the tests to data used by Hurst do not show
testswere developedto determine the existenceof the Hurst existenceof the Hurst effect. Only in the case of mud varyes
effect.They are formulated in precise mathematicalterms. is there spacefor some speculation. However, even in that
Theirbasisis the so-called GEOSdiagrams of R*n/n ø'5 case, long memory does not appear to be the natural
versusn that for a short memory stationary processwill explanation of the evidence which might be related to
convergeto a known distribution, whereas for a series nonstationarity of the records.
exhibiting the Hurst effect will divergeto infinity. The scale of fluctuation is, with the mean and the variance,
Theestimationof the scaleof fluctuationis itself a way of one of the most important parametersin hydrological sto-
testing for shortmemory stationarityand for computingthe chasticmodeling, since they completely describestatistical
asymptotic value for the GEOS diagram.Additionally, the propertiesof a time series such as central tendency, fluctu-
scaleof fluctuationprovidesa way of determiningthe length ations around the mean, the whole correlation structure, the
of a time series. "Hurst characteristics" of the time series, the low- and
The testsintroducedin this work do not have adjustable high-frequencycomponentsof the signal power spectrum,
parameters.The scale of fluctuation needs to be estimated the so-called Noah and Joseph effects [Mandelbrot and
byindependent means,but evenwithoutit GEOS diagrams Wallis, 1968], etc. Additional advantagesof a better knowl-
mayserveto indicate the presenceof the Hurst effect. If the edge of the scale of fluctuation of a stochasticprocess,
estimationof the scale of fluctuation does not show conver- besidesits role in the definition of the Hurst effect, lies on its
importance to study some characteristics of a time series
suchas runs,run lengths,level crossings,passagetimes and,
in general, probability distribution of extreme values.
The so-called Hurst effect and other related anomalities in
geophysicaltime series are probably the result of a mixture
of scalesmore than infinite memory. Knowledgeof those
scalesis a more fundamentalissue from a physical view-
point. Modelingthat mixture of scalesis more simpleand
down to earth than infinite memory modeling.

Acknowledgment. Wind data used in this work were kindly


provided by Roger Shaw (The University of California at Davis)
20'
from an experiment funded by Environment Canada under the
directionof Hartogand Neumannof the AtmosphericEnvironment
Service of Canada.
10-

,,,

REFERENCES
0 1• 260 360 46O 560 600
TIME (0.1SECONDS) Barnard,G. A., Discussionof Hurst, Proc. Inst. Civ. Eng., 5(5),
552-553, 1956.
Fig. 3b. Scale of fluctuation estimation for a vertical wind
Bhattacharya,R. N., V. K. Gupta, and E. C. Waymire,The Hurst
velocity
timeseries.The solidcurveis TF(T), andthedottedcurve effectundertrends, J. Appl. Probab., 20(3), 649-662, 1983.
istheestimator
givenby(14).Unitsofthescaleoffluctuation
arein Billingsley,
P., Convergence
of ProbabilityMeasures,JohnWiley,
0. Is. New York, 1968.
4002 MESA AND POVEDA: HURST EFFECT

Boes, D.C., and J. D. Salas, On the expectedrange of partial sums cont•nfasis


enel Problema
dela Informaci6n
Escasa,
Univer.
of exchangeablerandom variables,J. Appl. Probab., 10, 671-677, sidad Nacional, Facultad de Minas, Medellfn, 1982.
1973. Mejfa, J. M., I. Rodrfguez-Iturbe,
and D. R. Dawdy, Streamflow
Boes, D.C., and J. D. Salas, Nonstationarity of the mean and the simulation, 2, The brokenline processas a potentialmodelfor
Hurst phenomenon, Water Resour. Res., 14(1), 135-143, 1978. hydrologicsimulation,WaterResour.Res., 8(4), 931-941,1972.
Box, G. E. P., and G. Jenkins, Time Series Analysis, Forecasting Moran,P. A. P., On therangeof cumulative sums,Ann.Inst.Stat.
and Control, Holden-Day, Oakland, Calif., 1970. Math., 16, 109-112, 1964.
Bras, R., and I. Rodriguez-Iturbe, Random Functions and Hydrol- O'Connell,P. E., Stochastic
modellingof longtermpersistencein
ogy, Addison-Wesley, Reading, Mass., 1984. streamflow sequences,Ph.D. thesis,Dep. of Civ. Eng.,Imperial
Burg, J.P., Maximum entropyspectralanalysis,in Proceedingsof Coil., London, 1974.
the 37th Annual International SEG Meeting, Society of Explora- Pereira,M. V. F., G. C. Oliveira,C. G. Costa,and J. Kelman.
tion Geophysicists,Oklahoma City, Okla., 1967. Stochasticstreamflowmodelingfor hydroelectricsystems,Water
Feder, J., Fractals, Plenum, New York, 1988. Resour. Res., 20(3), 379-390, 1984.
Feller, W., The asymptotic distribution of the range of sums of Potter,K. W., Commenton "The Hurst phenomenon:
A puzzle:"
by V. Klem•,s, Water Resour. Res., ••(2), 373-374, 1975.
independent random variables, Ann. Math. Stat., 22, 427-432,
1951. Poveda,G., El Fen6menode Hurst, M.S. thesis,Univ. Nacionalde
Colombia, Medellfn, 1987.
Feller, W., An Introduction to Probability Theory and Its Applica-
Poveda,G., and O. J. Mesa, Estimation of the Hurst exponentand
tions, vol. 2, John Wiley, New York, 1971.
GEOS diagramsfor a non-stationarystochasticprocess,in Pro-
Gomide, F. L. S., Range and deficit analysisusingMarkov chains,
ceedingsof Stochasticand StatisticalMethodsin Hydrologyand
Hydr. Pap. 79, Colo. State Univ., Fort Collins, 1975.
Environmental Engineering: An International Conferencein
Gomide, F. L. S., Markovian inputs and the Hurst phenomenon,J. Honor of Professor T. E. Unny, University of Waterloo, Water-
Hydrol., 37, 23-45, 1978. loo, Ont., Canada, 1993.
Hipel, K. W., and A. I. McLeod, Preservation of rescaledadjusted Rodrfguez-Iturbe, I., J. M. Mejfa, and D. R. Dawdy, Streamflow
range, 2, Simulation studies using Box-Jenkins models, Water simulation, 1, A new look at Markovian models, fractional Gaus.
Resour. Res., 14(3), 509-516, 1978. siannoiseand crossingtheory, Water Resour. Res., 8(4), 921-930,
Hurst, H. E., Long term storagecapacity of reservoirs, Trans. Am. 1972.
Soc. Civ. Eng., 116,770-779, 1951. Roy, R., B. G. Sumpter, G. A. Pfeifer, S. K. Gray, and D. W. Noid,
Hurst, H. E., R. P. Black, and V. M. Simaika, Long-Term Storage: Novel methodsfor spectral analysis, Phys. Rep., 205(3), 109-152,
An Experimental Study, !45 pp., Constable,London, 1965. 1991.
Ibragimov, I. A., Some limit theorems for stationary processes, Salas,J. D., and D.C. Boes, Expected range and adjustedrangeof
Theor. Probl. Appl., 7, 349-382, 1962. hydrologicsequences,Water Resour. Res., I0(3), 457--463,1974.
Klem•.s, V., The Hurst phenomenon--a puzzle?, Water Resour. Salas,J. D., D.C. Boes, V. Yevjevich, and G. G. S. Pegram,Onthe
Res., 10(4), 675-688, 1974. Hurst phenomenon,in Modeling HydrologicProcesses,editedby
Klem•s, V., R. Srikanthan, and T. A. McMahon, Long-memory H. J. Morel-Seytoux, Water ResourcesPublications,Fort Collins,
flow models in reservoir analysis:What is their practicalvalue?, Colo., 1979a.
Water Resour. Res., 17(3), 737-751, 1981. Salas, J. D., D.C. Boes, V. Yevjevich, and G. G. S. Pegram,Hurst
Leopold, L. B., M. G. Wolman, and J.P. Miller, Fluvial Processes phenomenonas a pre-asymptoticbehavior, J. Hydrol., 44, 1-15,
in Geomorphology, W. H. Freeman, New York, 1964. 1979b.
Lloyd, E. H., Stochastic reservoir theory, in Advances in Hydro- Scheideger,A. E., Stochasticmodelsin hydrology, WaterResour.
science, vol. 4, edited by V. T. Chow, pp. 281-339, Academic, Res., 6(3), 750-755, 1970.
San Diego, Calif., !967. Siddiqui, M. M., The asymptotic distributionof the rangeandother
Mandelbrot, B. B., Intermittent turbulence in self-similar cascades: functionsof partial sumsof stationary processes,Water Resour.
divergence of high moments and dimensionof carrier, J. Fluid Res., •2(6), 1271-1276, 1976.
Mech., 62, 331-358, 1974. Taylor, G.I., Diffusion by continuousmovements, Proc. London
Mandelbrot, B. B., The Fractal Geometry of Nature, W. H. Math. Soc., 2(20), !96-211, 1921.
Freeman, New York, 1982. Vanmarcke, E., Random Fields: Analysis and Synthesis,MIT
Mandelbrot, B. B., and M. Taqqu, Robust R/S analysisof long run Press, Cambridge, Mass., 1988.
serialcorrelation, paperpresentedat 42nd international Statistical Wallis, J. R., and N. C. Matalas, Small samplepropertiesof hand
Institute, Manila, 1979. estimators of the Hurst coefficient h, Water Resour. Res., 6(6),
Mandelbrot, B. B., and J. W. Van Ness, Fractional Brownian 1583-1594, 1970.
motions, Fractional Gaussian noises and applications, SIAM Zapata, A.D., El Problema de la Informaci6nEscasaen la
Rev., 10(4), 422--437, 1968. Operaci6nde un SistemaHidrotdrmico, M.S. thesis,Univ. Na-
Mandelbrot, B. B., and J. Wallis, Noah, Joseph, and operational cional de Colombia, Medellfn, 1987.
hydrology, Water Resour. Res., 4(5), 909-918, 1968.
Mandelbrot, B. B., and J. Wallis, Roboustnessof rescaledrangeR/S O. J. Mesa and G. Poveda, Programas de Postgradoen
in measurementof noncyclic long run staticaldependence,Water Aprovechamiento de RecursosHidraulicos,UniversidadNacional
Resour. Res., 5(5), 967-988, 1969. de Colombia,ApartadoA•reo 1027, Medellin, Colombia.
Matalas, N. C., and C. S. Huzzen, A property of the range of partial
sums, Proc. Int. Hydrol. Syrup., 1,252-257, 1967.
Mejfa, J. J., and J. MilMn, Efecto de la memoria larga de los (Received October 6, 1992;
procesoshidro16gicossobre la operaci6n de los sistemashidro- revised June 1, 1993;
t6rmicosde generaci6ne16ctrica,in Seminario Sobre Hidrologia accepted June 21, !993.)

You might also like