






































Microsoft Word - 09harper.doc


Proceedings of the International Commission on History of Meteorology 1.1 (2004)

9.  The Scandinavian Tag-Team: Providers of atmospheric reality to

numerical weather prediction efforts in the United States (1948-1955)

Kristine C. Harper

History of Science Program
Oregon State University
Corvallis, Oregon, USA

Introduction

In late 1945, the distinguished mathematician John von Neumann needed a
suitably difficult scientific problem amenable to a numerical solution to showcase the
capabilities of his proposed computer. Although there were numerous candidates from
the physical sciences1, von Neumann settled on the weather prediction problem. In their
brief accounts of the development of numerical weather prediction, William Aspray’s
John von Neumann and the Origins of Modern Computing and Frederik Nebeker’s
Calculating the Weather: Meteorology in the 20th Century give von Neumann primary
credit for starting and leading the Meteorology Project at the Institute for Advanced
Study. Given significantly less credit are Carl-Gustav Rossby, Jule Charney, and a series
of Scandinavian meteorologists who significantly influenced the entire project.2 I will
argue that the Scandinavian “tag-team”, invited by Charney and supported by Rossby,
was not only critical to its ultimate success, but that differences in the cultures of
American and Scandinavian meteorology made the Scandinavians better suited to
accomplish the work at hand than their American counterparts. That the Scandinavians
possessed both practical forecasting skills as well as superb analysis and theoretical
knowledge enabled the answering of this question: Is the computer predicted
representation of the atmosphere a valid one?

The Beginnings of the Meteorology Project



Proc. ICHM 1.1 (2004) 85

John von Neumann’s quest to prove the worth of an electronic computing device
needed a sufficiently complex problem to fully exploit its power and be practical enough
to attract funding and public attention. By the end of 1945, the Navy was interested in
funding von Neumann’s efforts without restriction.3 World War II’s war fighting and
countermeasures efforts had been heavily based on applied physics. The advantages
presented by a computer for problem solving as well as strategic and tactical decision
making could not have been lost on Navy officials.

However, after discussing the issue with Vladimir Zworykin of the Radio
Corporation of America’s Princeton laboratories, the weather prediction problem started
to look appealing. Von Neumann had no meteorology background, but he had worked on
hydrodynamical problems. Weather prediction was a problem whose solution was
potentially useful to everyone. A non-linear mathematical description of a moving ocean
of air surrounding a rotating body was ideal because it could only be attacked numerically,
not analytically.

Having met several times with Weather Bureau personnel, in early February 1946,
Harry Wexler, Director of Research, encouraged von Neumann to present his ideas to
Carl-Gustav Rossby.4 Rossby, of the University of Chicago’s Department of
Meteorology, was probably the most influential academic meteorologist in the United
States. A naturalized U.S. citizen from Sweden, Rossby had spearheaded the training of
several thousand military personnel in meteorology during World War II. A strong
proponent of theory-based meteorology, he had many ideas about moving the art of
weather prediction into a solid scientific endeavor.

Von Neumann proposed to Rossby that the general circulation of the atmosphere
be modeled in its most “simplified and schematic form” – a homogeneous rotating earth
that allowed for variations in incoming solar radiation by latitude, but assumed that all
data were zonal, i.e., longitudinal placement had no effect.5 Rossby subsequently met
with von Neumann and reported the substance of the meeting to Francis W. Reichelderfer,
Chief of the Weather Bureau. Von Neumann’s interest in the pursuit of basic
meteorological problems by reducing the dynamic equations to a computer solvable
version was a potentially great asset to the progress of the science. Rossby wrote, “To
stimulate him further, and to lay the foundation for a computational approach to
theoretical meteorology, it would be necessary to organize a small and versatile group of
competent theoretical meteorologists to work with Professor von Neumann”.6

Von Neumann’s requirement for additional stimulation is debatable. The need for a
theoretical foundation to allow for a computational solution to meteorological problems is
not. The mathematical approach to meteorology had been weak before the war. The
existing hydrodynamical equations had to be simplified to be numerically solvable. The
resulting simplification had to produce a realistic representation of the atmosphere.

Therein lay the primary problem outside of building the new computer: who
would create such a simplified mathematical system? Rossby created a list of those who
might be both available and capable of addressing this problem. He suggested to von
Neumann that he propose a research project to the Navy’s Office of Research and
Inventions (later the Office of Naval Research – ONR) to be based at IAS. Due to



Scandanavian Tag-Team and NWP 86

discussions with the Navy’s LCDR Daniel F. Rex, he knew the Navy would make a
renewable annual funding commitment. The project would “examine the foundation of our
ideas concerning the general circulation of the atmosphere with the intention of
determining the steady state of the general circulation of the atmosphere and its response
to arbitrarily applied external influences”. Having arranged the contract, he provided von
Neumann with a complete budget proposal, personnel suggestions and a proposed
starting date.7 As von Neumann himself later wrote to Rossby in 1949, “[you] more than
anyone else were responsible for getting theoretical work started at IAS under the
auspices of the contract with ONR”.8

With the addition of technical information on the proposed computer, the contract
proposal echoing Rossby’s discussion went to the Navy. The proposed personnel would
be younger meteorologists led by Harry Wexler and prominent consultants including
meteorologists and oceanographers C.-G. Rossby, Harald Sverdrup and Jacob Bjerknes –
all Scandinavians.9

With the contract successfully negotiated, the Meteorology Project became a
reality on July 1, 1946. A meeting of some 20 prominent meteorologists in late August
established the working assignments. However, for a variety of reasons the personnel did
not materialize. After consulting Rossby and Reichelderfer, von Neumann reorganized the
Project into a smaller group. Only two people worked full-time (Paul Queney of France,
and Albert Cahn, University of Chicago) and they left in December. Israeli Chaim Pekeris
and New York University’s Hans Panofsky worked1/3 time and Army Captain Gilbert
Hunt, a mathematician, participated in some discussions. Army-Air Force Lieutenant
Philip D. Thompson took over Cahn’s work on the hydrodynamical equations. As 1947
dawned, only one person worked full-time for the Meteorology “group” – Thompson.10

His meteorology background was due to wartime training at the University of Chicago, he
was largely self-taught beyond the basics, and he lacked familiarity with the existing
literature. However, he did have some very good ideas. Jule Charney, who had attended
the August 1946 meeting in Princeton, visited the group in March 1947 just before
departing for the University of Oslo. Writing to Rossby, he noted that the meteorological
aspect was the “weak sister” of the Meteorology Project which did not exactly bode well
for its success.11

Von Neumann focused on numerical solutions to the hydrodynamical equations,
but of necessity his efforts were divided between the short-handed meteorology group
and the Computer Project itself. Since little could be accomplished by the isolated
Thompson, the Project was extremely fortunate that Charney was developing ways to
numerically treat the dynamic equations of the atmosphere. By the end of 1947, he had
found a “filtering” method which would remove “noise” – energy waves that did not
contribute to the solution, but which complicated the calculations.12 Von Neumann was
pleased when he found out that Charney was interested in the Meteorology Project. He
also inquired as to the availability of Arnt Eliassen, a Norwegian meteorologist who had
been collaborating with Charney.13

Charney, who had been offered a position at UCLA by Jacob Bjerknes,14

accepted von Neumann’s offer in part because he was concerned about the Project’s



Proc. ICHM 1.1 (2004) 87

future if it continued on its existing path. In writing to Bjerknes, Charney made clear his
view, “unless some physical ideas are brought to bear, the project will die out through
mathematical sterility”.15 Charney also recognized that the addition of physics would not
be enough. The Meteorology Project needed people who knew enough about meteorology
to know “when and how to make approximations” that would inevitably be required to
reduce the applicable equations to a computer solvable version. To this end Charney
recommended bringing in Arnt Eliassen from the Norwegian Meteorological Institute.16

The Scandinavian Tag-Team Begins

Arnt Eliassen thus unwittingly became the first in a line of Scandinavians to join
the Project. Eliassen possessed the ideal combination of skills: experienced in synoptic
and
theoretical meteorology, he was also interested in numerical solutions. As Charney saw it,
the Project needed to be combining theoretical work with empirical data. That meant that
some members of the group had to have “intimate experience with actual weather
processes”.17 Eliassen was the only member of the group who met that requirement.

Charney and Eliassen arrived in Princeton in the summer of 1948 joining
Thompson and Hunt.18 Not on scene, but a major player nonetheless, was its primary
meteorological supporter – the always forward looking Rossby.

If this group was focused on the development of meteorological theory, why the
need for personnel with synoptic experience? Synoptic meteorology relied on data
collected worldwide and analyzed locally to make predictions. A very subjective
endeavor, it was considered by theory-based dynamicists to be more an art than a science.
However, Rossby recognized that any theory used as a basis for a computational solution
had to first include those factors which were either consciously or unconsciously used by
the forecaster. After all, they were adding significant skill to turn raw data into a
representation of the atmosphere from which they could make a prediction. Any
additional variables could be added once the approximations and assumptions of the
forecasters had been included.19

Therein lay a potential problem for the fledging group. If the team members were
looking strictly at elegant numerical solutions to the hydrodynamical equations, then they
could develop internally consistent models. Such models could produce forecasts for
conditions at multiple atmospheric levels correctly correlating with each other, but not
necessarily having any relation to reality. As Rossby noted, the equations needed to be
viewed as tools to studying problems suggested by the atmosphere, not as an end in
themselves.20 Without solid synoptic support, Charney’s fear of the group becoming
mathematically sterile would become a reality.

Charney and Eliassen developed a very close collaborative effort producing a
major article entitled “A numerical method for predicting the perturbations of middle
latitude westerlies”.21 It was important because it demonstrated that an objective method
could successfully replace more subjective methods. Further, the heuristic nature of their
approach enabled them to try a model, see how it worked, and modify it as required.22 Of



Scandanavian Tag-Team and NWP 88

course, how would they know whether it worked or not? The only way to do that was to
compare the computer output to analyses produced by synopticians. When the output
appeared meteorological and gave results similar to those anticipated by an outstanding
analyst then it could be considered valid. Eliassen’s value was not just his theoretical and
mathematical knowledge, but the synoptic experience that allowed him to analyze the
data with a significant degree of expertise.

Eliassen, however, had accepted a position at UCLA with Jacob Bjerknes and
would not be in Princeton forever. Charney needed a replacement. He settled on another
Norwegian, Ragnar Fjörtoft, because of his ability to mix into the general nature of the
problems being explored by the Project and of course, because he had an excellent
synoptic background through his work with the Norwegian Weather Service.23 He arrived
in September 1949.24

Fjörtoft slid easily into Eliassen’s spot and was on hand for the first ENIAC (the
Aberdeen Proving Grounds computer) “expedition” in March 1950. With Charney and
von Neumann, he wrote an article describing the results of their barotropic model runs.25

However, he needed to return to Scandinavia in June 1950. Charney once again needed a
replacement – preferably someone like Eliassen or Fjörtoft.26

Rossby had a new prospect for Charney’s group – the Swede Bert Bolin. Bolin
was an excellent analyst and had had enough theoretical training that he would be able to
carry out the needed synoptic studies. The barotropic model had failed in some regions
and they needed to determine the non-computational error sources.27  Charney thought
baroclinicity might be the problem.28 Rossby thought it might have been a “rather hidden
appearance of cyclonic vorticity aloft”. It would take careful analysis of the atmospheric
situation to determine the error source. That would be Bolin’s task.

As it turned out, Fjörtoft returned for another six months stay. During that
period, with Bolin, Fjörtoft and Davies on board in addition to Joseph and Margaret
Smagorinsky, Charney had sufficient personnel to attempt a baroclinic model.29

Unfortunately, this
abundance of personnel riches came to an abrupt halt in July1951 when for a period of
time Smagorinsky was the only one left with the group. In the fall of 1951, only Charney,
Norman Phillips and Joseph Smagorinsky were working full time. In the absence of
experienced analysts, members of Rossby’s International Meteorological Research
Institute in Stockholm performed a large part of the preliminary work needed for
evaluating the new models.30 However, Charney needed people on site in Princeton and
the staffing situation was rather desperate. Once again Rossby arranged a personnel
transfer to the beleaguered Charney. By the middle of 1952, Ernst Hovmöller of the
Swedish Meteorological and Hydrological Institute (SMHI) arrived to undertake the
synoptic duties associated with the switch from the barotropic to the baroclinic models.31

Still short of people, Charney continued to look abroad. One name on his list, Eady, was
from England, but the other two were the Norwegians Eliassen and Fjörtoft.
Unfortunately, they were all unavailable until 1953.32

These meteorologists would provide relief, but it was too distant. Von Neumann
made a rather desperate plea to Hovmöller’s boss, Anders Angström, to let him stay



Proc. ICHM 1.1 (2004) 89

beyond the end of 1952. Von Neumann wrote that “His association with us is perhaps
the first example of the kind of cooperating that will ultimately have to take place
between theoretical and synoptic meteorologists if and when numerical forecasting is
integrated into the governmental weather services”.33

Angström deflected von Neumann’s plea and instead offered Roy Berggren. Von
Neumann was happy to take Berggren, but still wanted Hovmöller. Angström firmly
closed that door, but undeterred von Neumann tried again. He needed an outstanding
synoptic meteorologist and needed him right then. “To obtain such a person we are
willing to bring him over from Europe…” Berggren arrived in March 1953 and stayed
until the end of the year.34

In early 1953, the Princeton computer was working well and the Project members
were kept very busy running models and analyzing output. Phillips in Princeton and
Bolin in Stockholm changed places in the summer. Bolin stayed with the Princeton group
for nine months where he was joined by Eady and Fjörtoft.35 Fjörtoft remained until the
end of June 1955. The last of the tag-team at the Meteorology Project was fittingly the
first - Eliassen – and he stayed for the six months ending April 1956.

By the mid-Fifties, the Princeton group’s emphasis had shifted to general
circulation versus weather prediction due to the opening of the Joint Numerical Weather
Prediction Unit. With von Neumann’s departure for the Atomic Energy Commission and
support from IAS fading, Charney left for MIT and the Meteorology Project ceased to
exist.

Conclusion

So why the Scandinavians? Were there no Americans capable of performing the
detailed analyses needed to validate the models and locate sources of error? Prior to World
War II there were very few university trained meteorologists in the United States.  Most
operational forecasters received their training on the job. When the war required many
more meteorologists, Rossby’s University Meteorology Committee only accepted for
training those who had extensive backgrounds in either physics or mathematics.
Consequently, those who remained in meteorology after the war often had a different
approach to the science than those already in the field. All of the Americans working in
the Meteorology Project came into meteorology because of the war and they did not have
the practical weather experience which develops a solid feel for the atmosphere.

Charney realized that it was absolutely imperative that at least some of the people
working in numerical modeling have a more subjective feel for the atmosphere and the
weather parameters necessary to making a valid prediction if they were going to overcome
the skepticism that greeted them from the meteorological community. The “older”
meteorologists possessed the necessary backgrounds, but many were either teaching the
returning servicemen crowding their campuses or conducting research that had been
postponed by the war. Some, perhaps, were just not convinced of the need for numerical
methods. In any case, those active in the field in the United States were not clamoring to
join the highly respected von Neumann.



Scandanavian Tag-Team and NWP 90

Charney needed meteorologists with weather experience coupled with a solid
theoretical background; people who were excellent analysts and interpreters of data and
who could also communicate physical concepts in terms of complex equations. Charney
already knew people who fit the requirements because of his stay in Norway. He also
knew Rossby could provide the people he needed to keep the project moving.

In the United States, academic meteorologists were not people who were
interested in actually predicting the weather. On the other hand the people making the
predictions were not considered “meteorologists”, but forecasters; a lower caste of non-
professional, non-scientific workers. In Scandinavia these two groups were often one and
the same. Therefore, their meteorologists were multi-faceted scientists ready to be tapped
for the development which would change the face of meteorology in the middle of the
Twentieth Century – numerical weather prediction – and thus provided a dose of
atmospheric reality to a tremendous achievement of applied mathematics.

Endnotes

1 Among them fluid dynamics, elasticity and plasticity theory, electrodynamics and
quantum theory.

2 William Aspray, John von Neumann and the Origins of Modern Computing,
Cambridge, Mass., MIT Press, 1990. Chapter 6 addresses the early period in numerical
weather prediction. Frederik Nebeker, Calculating the Weather: Meteorology in the 20th

Century, San Diego, Academic Press, 1995 addresses the growth of calculations as a tool of
weather prediction from the 19th century to the introduction and acceptance of the
computers in the advance of numerical weather prediction. He mentions the development of
meteorological instrumentation by RCA as a possible reason for Zworykin’s interest in
weather prediction.

3 Memo to John von Neumann from Frank Aydelotte of November 9, 1945 (von
Neumann Papers, Library of Congress, Box 12, Folder 1) (Hereafter: von Neumann Papers)

4 Letter to C.-G. Rossby from John von Neumann of February 6, 1946 (von
Neumann Papers, Box 15, F 7)

5 Ibid.
6 Letter to F. W. Reichelderfer from C.-G. Rossby of April 16, 1946 (von Neumann

Papers, Box 15, F 7)
7 Letter to John von Neumann from C.-G. Rossby of April 23, 1946 (von Neumann

Papers, Box 15, F 7)
8 Letter to C.-G. Rossby from John von Neumann of June 13, 1949 (Charney Papers,

MIT Archive, MC 184, Box 14, F 459) (Hereafter: Charney Papers)
9 Letter to CDR D. F. Rex from Frank Aydelotte of May 8, 1946 (von Neumann

Papers, Box 15, F 6)
10 Progress Report, Meteorology Project, July 1, 1946 to November 15, 1946 (von

Neumann Papers, Box 15, F 6)
11 Letter to C.-G. Rossby from Jule Charney of March 19, 1947 (Charney Papers,

Box 14, F 460)
12 Letter to P. D. Thompson from Jule Charney of November 12, 1947 (Philip D.

Thompson Papers, National Center for Atmospheric Research, Boulder, Colorado, Folder
Correspondence Jule Charney, 1947-1950) (Hereafter: Thompson Papers)

13 Letter to Jule Charney from John von Neumann of November 19, 1947 (von
Neumann Papers, Box 15, F 1)

14 Letter to Jule Charney from Jacob Bjerknes of November 4, 1947 (Charney
Papers,  Box 4, F 120)



Proc. ICHM 1.1 (2004) 91

15 Letter to Jacob Bjerknes from Jule Charney of January 14, 1948 (Charney Papers,
Box 4, F 120)

16 Letter to John von Neumann from Jule Charney of January 2, 1948 (von
Neumann Papers, Box 15, F 1)

17 Letter to John von Neumann from Jule Charney of January 2, 1948 (von
Neumann papers, Box 15, F 1)

18 Letter to Jule Charney from John von Neumann of February 6, 1948 (von
Neumann papers, Box 15, F 1)

19 Letter to J. Charney from C.-G. Rossby of October 28, 1948 (Charney Papers,
Box 14, F 460)                

20 Letter to Jule Charney from C.-G. Rossby of January 9, 1949 (Charney Papers,
Box 14, F 460)

21 J.G. Charney and A. Eliassen, “A numerical method for predicting the
perturbations of middle latitude westerlies”, Tellus, 1 (1949), 2: 38-54.

22 Letter to J. Charney from C.-G. Rossby of April 23, 1949 (Charney Papers, Box
14, F 459)

23 Letter to C.-G. Rossby from Jule Charney of May 24, 1949 (Charney Papers, Box
14, F 459)

24 Progress Report, Meteorology Project, July 1, 1949 to June 30, 1950 (Charney
Papers, Box 14, F 459)

25 J.G. Charney, R. Fjörtoft and J. von Neumann, “Numerical integration of the
barotropic vorticity equation”, Tellus, 2 (1950), 237-254. In this barotropic model, density is
assumed constant on a pressure surface.

26 Letter to C.-G. Rossby from Jule Charney of June 5, 1950 (Charney Papers, Box
14, F 459)

27 Letter to Jule Charney from C.-G. Rossby of June 13, 1950 (Charney Papers, Box
14, F 459)

28 Letter to George Platzman from J. Charney of June 22, 1950 (Charney Papers,
Box 14, F 451)

29 Letter to George Platzman from Jule Charney of June 22, 1950 (Charney Papers,
Box 14 F 451). The baroclinic model assumes the advection of both the circulation and
thermal fields.

30 Progress Report, Meteorology Group, July 1, 1951 to September 30, 1951
(Charney Papers, Box 9, F 304)

31 Letter to Jule Charney from C.-G. Rossby of April 10, 1952 (Charney Papers, Box
14, F 459)

32 Letter to John von Neumann from Jule Charney of November 17, 1952 (Charney
Papers, Box 14, F 517)

33 Letter to Anders Angström from J. von Neumann of November 14, 1952
(Charney Papers, Box 16, F 517)

34 Letter to Anders Angström from J. von Neumann of February 3, 1953 (Charney
Papers, Box 16, F 517)

35 Letter to Bert Bolin from Jule Charney of February 3, 1953 (Charney Papers, Box
4, F 121)


