XEROX

CLlMATIC DETERMINISl\1
EDWARD N. LORENZ

National Center for Atmospheric Research, Boulder, Colo.1

ABSTRACT ·.

· The often-accepted hypothesis that the physical laws governing the behavior of an atmosphere determine
a unique climate is examined critically. lt is noted that there are sorne physical systems (transitive systems)
whose statistics taken over infinite time intervals are uniquely determined by the governing laws and the
environmental conditions, and other systerns (intransitive systems) where this is not the case. There are
also certain transitive systems (almost intransitive systems) whose statistics taken over very long but
finite intervals differ considerably from one such interval to another. The possibility that long-term climatic .
changes may result from the almost-intransitivity of the atmosphere rather than from environmental
changes is suggested.

l. Introduction these physical laws completely determine the climate,

It is indeed a privilege to be allowed to present the
opening paper at this symposium on Causes of Clima tic
Change. For a long time we have been faced with well-
nigh incontrovertible evidente that the climate during
previous centuries or millenia has differed from the
climate of today, and we are presently gathered to ask
to what extent we can account for these climatic
-changes. During 'the course of this symposium we can
expect many altemative and sometimes conflicting ex-
-. phmations to be offered. Sorne of these will attribute
. . .. t changes of climate to changes in the properties of
the oceans. Others may call upon variations in voÍcanic
they certainly exert their influence upon it, By and
large, the physical processes which affect the climate
are the same ones which we must take into account if
we wish to make the best possible weather forecasts.
To be sure, sorne processes which must have an im-
portant bearing upon climate, and upon long-terrn
· weathervariations, can be virtually ignored in dáy-
to-day forecasting. For example, in
preparing a one-
day forecast we need not consider the possibility of
growth or decay of existing glaciers, and we ca11
probably disregard changes in the present ocean sur-
face ternperatures. �/
� ,-'._,-,�'.,.··/·,;?��:fr;,.�:);é'. j

�:h:.:;:�::::::�:7.on of governing physical . ,}i!{ '.

activity. Stíll others will invoke fluctuations in the out-
put of the sun. 20

�f��l!���1fi��:j�l�:tlffi. · �,�J�itt�f
<,

The types of explanation which I have chosen to

mention have in common the feature that they seek a
change in the environment of the atmosphere as the
most likely cause of a change in the behavior of the
'1
atmosphere itself. Those who advance such explana-
tions might very well assume that if environmental press tbe time derivative of each relevant quantity.· ·-r {:;;:r}':;.�J:t �
· 1 influences long ago had been the same as at present, in terrns of the instantaneous overall state-of the ;: •:e- ,,•. �i;;.' -
4
the ancient climate would have been t� same as . atmosphere and its environrnent. As a result, the;\}<('.(. . . ;\'!
today's; hence the necessity for invoking environ- mathematical problern of weather forecasting is far. ' .':·rtt,r.�.
-:)J
mental changes asan explanation. In short, they might more straightforward than the problem of deducing :'. -. }·: ;;?;ri;f�j
the clima te frorn the physical laws. In fact, one of the ' -, <··: ,, · -�- · l
. . ··.'······ . �"-' -·�
-draw the not unreasonable conclusion that the nature
__

of the atmosphere's environment, together with the more promising methods of deducing the climateseerns . 1
internal physical nature of the atmosphere, should to be· to go through the mechanics of weather fore- J.
-deterrnine the climate in sorne more or less unique casting, grinding out numerical solutions oI the equa- 1
fashion. · tions for extended periods of time, and then to compile :-, ¡
It is this hypothesis, that the physical laws whichstatistics from the solutions (Lorenz, 1964). .·r �
govern the atmosphere are responsible for determining More generally, the problem of deducing the climate .�
a unique climate, that I wish to examine critically. from the physical laws which influence the climate may -_¡
I shall assume without further ado that whether or not · be viewed as a special case of an easily stated mathe- ·:,,·: ·�
1
Permanent affiliation: Massachusetts Institute oí Tech- matical problem : Given a clo�ed system of equations, · ·.�
nology, Cambridge. to deduce the set of long-term statistics oí the solu-
1
/ . ·i

. � ..:� .. . . ' -,:

1
.. >r
_;
·1
,1

". :,,.:.... ,.. . _:· \ r,.::· �: ,

. -·� :.\ ·.';:.'
� -�. . ,.,,._ -�-
. ' "" .. ;._
- ........ .
. ·:-:.:.;i; •..�-_ ,:-:...... ' !f
 ' s;·'

METEOROLOGICAL MONOGRAPHS VoL. 8, No. 30

tions of these equations. This more general problem indefinitely, but a pattern containing five waves, if
has received considerable attention from mathemati­ established instead, will also persist indefinitely. This
cians. Questions concerning the existence and unique­ . is a truly intransitive system. Such external tampering
ness of long­term statistics fall into the realm of ergodic as stirring the water with a pencil may change the
theory. · flow from a four­wave to a five­wave pattern, or vice
Particularly when the equations governing a physi­ versa. Under slightly different fixed external condi­
cal · systern are . linear, a unique set of long­term tions, such as a slightly higher rotation rate, only the
statistics can often be expressed in analytic form. five­wave pattern will occur. In this case the system is
However, the equations governing the atmosphere are transitive. Transitivity is a qualitative feature of the
highly nonlinear. The physical process responsible for experiment, but it depends very definitely upon quan­
the most troublesome nonlinearity is advection­the titative features of the input.
transporting of sorne property of the atmosphere from Another. example is afforded by simple numerical
one location to another by the motion of the atino­ models which simulate the gross features of atmo­
sphere itself, Since the motion of the atmosphere is also spheric motion [see, e.g., Lorenz (i963)]. Both transi­
·. one of the properties of the atmosphere represented by tive and intransitive systems are easily constructed.
the dependent variables, those terms in the equations The difference between them may be simply the
which represent advection will be quadratic, thus numerical value of one preassigned parameter, ·
:,'i­endering the complete system nonlinear. How about the real atmosphere? Is it transitive? .
. · In .the case of nonlinear equations, the uniqueness We do not know. The atmosphere is neither alabora­
oflong­term statistics is not assured. From the way in tory experiment nora set of riumbers in a computer,
which the problem is forrnulated, the system of egua­ and we cannot turn it off and then set it in motion
. tions, expressed in deterministic form, together with a again to see whether a new climate develops. Neither
specified set of'mitial conditions, determines a time­ does current mathematical theory give us the answer ..
dependent solution extending indefinitely into the So far, 1 have been tacitly identifying "climate"
future, and therefore determines a set of long­term with the set of long­terrn statistics, and in addition.
statistics. The question rernains as to whether such I have been assuming that "long­terrn" averages. mean
statistics are independent of the choice of initial condi­ averages over infinitely long time intervals extending
tions. We define a -t,ansitive system of equations as forward from the present. This is not the universally
one where this is the case. rr, however, there are two accepted concept of climate. In fact, if climate were
or more sets of long­term statistics, each of which has defined in this manner, climatic change would by
a greater­than­zero probability of resulting from ran­ definition be impossible. A concept of climate more
dornly chosen initial conditions, the system is called compatible with the purposes of this symposium
intransitive. , would be the set of statistics taken over a long but
So far I have just introduced definitions. Mathernat­ finite interval of time.
ical theory now tells us, however, that both transitive
and intransitive systems exist. Moreover, no simple ­ 4. The almost intransitive system
way has been discovered · for examining an arbitrary This leads us to the concept of a special type of
system of equations · and determining whether it is transitive system which, for want of a standard mathe­
transitive or intransitive. matical term, I shall call almost intransitive. In an
almost intransitive system, statistics taken over infi­
3. Examples of an intransitive system nitely long time intervals are independent of initial
Since an intransitive physical system, where the conditions, but statistics taken over very long but _,
physical laws do not uniquely determine the climate, finite intervals depend very much upon initial condi­
may be a somewhat unfamiliar concept, let me give tions. Alternatively, a particular solution extending
a few examples. One is provided by the laboratory over an infin,ite time interval will possess successive
experiments which have been designed to simulate very long periods with markedly different sets of
certain features · of the atmosphere [see Fultz et al. statistics.
(1959) and Hide (1958)]. Tbe apparatus consists in I am not aware that the mathematical theory of
essence of a. rotating basin, containing water which is almost intransitive systems has been very highly
subjected to differential heating. The resulting motion developed, · but the existence of systems having the
of the water is made visible by a tracer. Under suitable proper qualifications is well established. I do not know
conditions a set of waves develops and progresses whether the experimentalists in the laboratory have
about the axis of rotation. found occasion to seek or study such systems. The
Under certain fixed external conditions, a pattern sirnplest numerical models simulating the atrnosphére
containing four waves, once established, will persist may, however, be made almost intransitive through
 FEBRUARY 1968 METEOROLOGICAL MONOGRAPHS 3

suitable adjustrnents of one or more of the preassigned of the system, and equations for the time derivatives
parameters. J\s the models are made more compli­ of properties of this portion of the environment are
cated, it seems possible that almost­intransitivity may included 'in the complete system of equations. The
·hecome the rule rather than the exception. sun, and presumably those layers of the earth from
A íew years ago I had occasion to work with a model which volcanic activity arises, may still be regarded
in which the instantaneous state oí the atmosphere as part oí the external environment, but the oceans,
was represented by a set of 28 numbers (Lorenz, 1965). . and also the glacial ice, should be part of the system.
The solution of the system of 28 equations was ex­ Perhaps almost­intransitivity of such an augmented
tended overa four year simulated period. Even such system is less difficult to visualize, Nevertheless,
basic quantities as mean overall westerly wind speed almost­intransitivity oí the atmosphere of a hypotheti­
and pole­to­equator temperature contrast assumed cal earth devoid of oceans, ice and dust rérnains a
markedly different averages during different years; at theoretical possibility which may sorne <lay be verifi­
times the instantaneous departures from normal woukl able by an extensive mathematical model.
retain one sign throughout periods as long as four
· months. This was despite the fact that the time steps 5. Conclusions
in the numerical solution were only 3 hr. The model
Despite our meager knowledge of almost­intransi­
did not indude any heat storage in the underlying
tivity, we can draw a few conclusions. For one thing,
ocean or ground surfaces nor any seasonal variations
the mere existence oí long­term climatic changes can­
in the forcing. On a suitable time scale, this system
not by itself be taken as proof of environmental
was almost intransitive. Whether the model could
changes; alternative explanations are now available.
have produced century­to­century differences in aver­
Finally, what about the not unlikely possibility that
age properties is another question.
More recently, together with Dr. Eric Kraus of the · the atmosphere would be alrnost­intransitive if the
environmental influences were constant, while at the
Woods Hole Oceanographic Institution, 1 have studied
· same time externa} environmental changes actually
a model in which seasonal variations of externa} heat­
ing are included (Kraus and Lorenz, 1966). We ex­ are taking place? The effect of these changes will then
tended the solution overa simulated period oí 100 yr, be harder to detect, and causative connections witl be
·.,.
--�
Although the separate summers were very much alike, more difficult to establish. For example, an environ­
.

they differed enough so that the ensuing winters were mental change which ought to bring about a 2C
·· ?le to differ considerably from one another; a single temperature rise might occur just at the time when
· ... ,vinter would not suffice for , the determination of the temperature was in the process oí falling 2C as a
long­terrn climatological statistics. result of almost­intransitivity. The environmental
How about the real atmosphere again? lt was not change rnight then go unnoticed simply because no.
my original intention to put inrá plug for almost­: one would see any reason to Iook for it ..
intransitivity as a majar cause of climatic change. ­In summary, climate mayor may not be determi­
However, alrnost­intransitivity is too important a nistic. We shall probably never know for sure, but as /·
phenomenon todisregard altogether, and, in examining further mathematical theory is developed, and as.
the program for this symposium, 1 gained the decided more realistic models are consfructed, we may become ,
impression that no one else would put in any such more and more ­eonfident of our opinions.
plug. Let me say, then, that I find it conceivable that ·
almost­intransitivity could be a principal cause of
REFERENCES
climatic change, although I would not be ¡>repared to
say that it is the most likely cause. Perhaps more can Fultz, D., R. R. Long, G. V. Owens, \V. Bohan, R.
be said when we have had the opportunity to extend J. Weil, 1959: Studies of thermal convection in a rotating .
cylinder with sorne implications Ior large­scale atmospheric ·
the solutions of much larger numerical models over motions. Meteor. Monog., 4, No. 21, 104 pp.
much longer time intervals, to see whether almost­ Hide, R., 1958: An experimental study oí thermal convection
intransitivity on the scale of centuries rather than in a rotating liquid. Phil, Trans. Roy. ss« London, A250,
years can occur, 441­478.
A word oí warning may be needed here. The mathe­: Kraus, E. B., and E. N. Lorenz, 1966: Numerical experiments
matical theory which I have quoted applies to systems with large­scale seasonal forcing. J. Atmos. Sci., 23, 3­12.
Lorenz, E. N., 1963: The mechanics of vacillation. J. Atmos. Sci.,
where environmental influences, if present at ali, are
20, 448­464.
not affected in turn by the system. The atmosphere ­­, 1964: The problem oí deducing the clirnate from the
by itself is not a system of this sort; the theory is governing equations. Tellus, 16, 1­11. .... ....... ·
more appropriate if all parts of the environment which ­­, 1965: A study of the predictability oí a 28­variable at­,
.e influenced by the atmosphere are included as part mospheric model. Tellus, 17, 321­333.