Interview With Professor K.H. Hinkelmann
Offenbach is located on the River Main a short distance upstream from Frankfurt and near the centre of the Federal Republic of Germany. Its name was first mentioned in a document dated AD 977, although in those days the neighbouring community of Burgel was more important, being where the Frankish Kaiser Heinrich II held the imperial parliament in 1018. For more than three centuries the village belonged to the hereditary Counts of Isenburg, but the Congress of Vienna stripped the last reigning prince of his title to the lands for having supported Napoleon, and thus it was that in 1816 Offenbach became part of the Grand Duchy of Hesse. The nineteenth century saw the start of the town's rise to fame as it developed into an industrial centre, becoming known the world over for its leather goods. In 1956, the Council of Europe awarded the first Europa Prize to Offenbach and its sister-town of Puteaux-sur-Seine (France).
The headquarters of the Deutscher Wetterdienst has been located at Offenbach since 1957. There are hand-written accounts of exceptional weather events dating back to the sixth century, and with the introduction of the printing press, weather records came to be included in calendars and ephemerides. The first regular weather journals were kept in the sixteenth and seventeenth centuries, and a synoptic approach was introduced by Landgrave Hermann of Hesse; indeed, in 1635 he published weather observations made in the provinces of Hesse and Pomerania on a daily basis. In 1679, Professor Reyher of Kiel began systematic observations using the newly-invented instruments to measure the atmospheric pressure and temperature, and continued this work for 35 years. Incidentally, he is said to be the author of the first handbook on meteorological instruments which appeared in 1668.
The Meteorological Society of Mannheim was established in 1780 on the initiative of Elector Karl Theodor of the Palatinate, and the story is continued in earlier articles (WMO Bulletin 27 (3) pp. 179-182; 29 (4) pp. 235-238; 30 (2) pp. 104-106).
On 10 June 1985, Dr H. Taba visited the small town of Obertshausen, a few kilometres ESE of Offenbach, in order to interview Professor Karl Heinz Hinkelmann, a truly outstanding scientist in the field of numerical weather prediction. As a student at university, he already realized that there was a need to introduce more objective and mathematical methods into weather forecasting, and so he and his fellow-student, E. Lingelbach, determined to orient their studies in that direction. They took their inspiration from the ideas of L. F. Richardson and Vilhelm Bjerknes.
However, the Second World War was in progress, and scientists had to work largely in ignorance of research results in other countries. The major limiting factor in putting theory to practical test was the capability of performing the complex calculations that Hinkelmann's methods involved. Whereas scientists in the USA, USSR, United Kingdom and Scandinavia were working more or less exclusively with barotropic models and using geostrophic approximations and a divergence equation to filter undesirable effects from the computations, Hinkelmann was convinced that the best approach lay through the primitive equations. He met C.-G. Rossby and, like many another young meteorologist, fell under the spell of this outstanding man. For his part, Rossby saw in Hinkelmann an upcoming scientist of great promise and many original ideas.
At last, in the late 1950s things began to improve, and Hinkelmann's group was able to gain access to some of the early electronic computers. Hinkelmann brought numerical weather prediction in the Deutscher Wetterdienst from theoretical research, through experimentation, up to a highly satisfactory routine operation. That achieved, he felt he could relinquish his post in the Meteorological Service and devote his efforts to university work. He therefore accepted a professorship in theoretical meteorology at the University of Mainz in 1968.
Dr W. Edelmann was a long-time disciple and colleague of Professor Hinkelmann ' in the Deutscher Wetterdienst. Here are some recollections as conveyed to Dr Taba by Dr Edelmann:
'With Hinkelmann, research in the Federal Republic of Germany took a new direction. In the USA an early programmable computer had permitted the first predictions based on a filtered barotropic model, but even while that was going on, Hinkelmann saw that a good forecast using numerical methods must take into account a three-dimensional baroclinic atmosphere. That was the problem be tackled from the very beginning, and it demanded great faith and not a little optimism in view of the astronomical demands that this concept would make on computer facilities.
Whilst I was studying at the Free University in Berlin in the early 1950s, thanks to the recommendation of Professor Scherhag I was able to join Hinkelmann's small research group at Bad Kissingen for the winter semester 1952/53. I remember my first impressions: the narrow room under the sloping roof, the dense clouds of cigarette smoke and overflowing ashtray, a blackboard completely covered with hieroglyphs and chalk-dust everywhere, and the unmistakable voice of Hinkelmann retaining some of its Saxon overtones. In 1954 I came to join the group full time, and worked alongside Hollmann, Reiser (now President of the Deutscher Wetterdienst) and Wippermann.
For the historical record, I think it is worth recalling how the first trial three-dimensional 24-hour forecast field was made. The initial analyses - geopoten-tials at three levels - had been prepared manually by the synoptic division.
Then the vorticity and even the Jacobians of the quasi-geostropic model had to be evaluated by graphical addition and subtraction and more elaborate methods, producing a whole lot of maps; that took several days and was not very precise. Then a square grid was placed over the Jacobian maps and values interpolated for each grid point. The grid for the three levels was then enlarged and reproduced on a huge piece of white paper. Our young lady assistants transferred the Jacobians to this grid in very small writing, because much space was needed for our main work, namely the solution of a three-dimensional elliptic equation for tendency by relaxationandmany many iterations. One girl read out the figures to the other who sat at a noisy and slow mechanical calculator. So it went on through each grid point and each iteration. After a couple of hours, the girls would change places; after a couple of days they had done an iteration for the whole field, and after several weeks we had figures we assumed to be the solution to the elliptic equation. The tendency was converted back to a map and graphically added to the initial field, giving us a forecast for 12 hours. Then the entire operation had to be repeated to give a 24-hour prediction. The result did not look totally unreasonable. We never dreamt that less than 30 years later all this computation would be performed much better in a matter of minutes.
Hinkelmann is not only an excellent scientist, he is also a first-rate teacher and a splendid organizer. We were so captivated by our scientific problems that it seemed perfectly natural that we discuss them in all sorts of places. Hinkelmann's formulas were very likely to be found scribbled on table covers in restaurants or in blank spaces on the page of a newpaper. Not surprisingly, he frequently left personal items like cigarettes or umbrellas in cafes or tramcars. He has that most valuable talent of being able to reduce a problem to its essential elements, so that the solution (or source of error) becomes manifest. Finally, Hinkelmann has a great sense of humour; apart from his rich collection of jokes, he can also say outrageous things with a perfectly straight face, leaving his audience in a state of doubt or even shock, before relaxing the tension with a broad grin.'
H.T. - Professor Hinkelmann, please will you tell readers first about your background and education.
K.H - I was born in 1915 in a small village called Geringswalde. It is now within the German Democratic Republic about 45 km south-east of Leipzig. It was famous for chair-making, and my father had his own factory. I went to primary school in Geringswalde and secondary school in nearby Rochlitz where I found my special interest was in mathematics, physics and chemistry. When I left school at the age of 19 it was very hard to find a job; there were lots of unemployed people then. So there seemed to be nothing to do except join the army, and it was quite dreadful. I was a private soldier from 1934 to 1936 and then at last I was made a corporal. From 1937 to 1941 I was attached to the administration. Then in 1941 1 was able to go to technical college in Dresden for four semesters where I mainly studied mathematics and theoretical and experimental physics. After that I went to the University of Leipzig and joined the Meteorological Institute which was still under the direction of Professor Ludwig Weickmann who, some ten years earlier, had been working with the Scandinavians Tor Bergeron, Jack Bjerknes and Erik Palmén on the then new concept of three-dimensional analysis.
H.T. - What method of forecasting was favoured at that time?
K.H. - It was really an extrapolation of the analysed fields based on what observations we had, using linear manual operations for the purpose. Now I and some of the other students, who included Ernst Lingelbach, felt convinced that we should apply theoretical physics to this problem as was already done in astronomy and astrophysics. Obviously this would be far more complicated because we should have to deal with non-linear differential equations. At any rate I suggested to Professor Weickmann that some of the lecture time devoted to meteorology should be given to theoretical physics; I was a good student and Weickmann had taken to me, so I did not feel that was too presumptuous.
H.T. - This was a completely new approach, was it?
K.H. - You must remember that the Second World War was going on and that meteorologists in Germany had no way of knowing what was being done in other countries, but certainly as far as we in Germany were concerned it was a new concept. Vilhelm Bjerknes had introduced the use of linear equations at Leipzig when he was director of the meteorology and geophysics department, and this had been carried on ever since. But with this method you miss crucial processes such as cyclogenesis, frontogenesis and occlusion because no account is taken of interaction between the various wave motions. Perhaps we owed our inspiration to Charles Babbage and L. F. Richardson. Richardson had the right idea, and if he had not run into numerical instability he would have had much more encouragement and acknowledgment. Of course, we had only the vaguest notions about the feasibility of computers, so that our constraint was the same as Richardson's, namely doing all the complex calculations in a reasonably short time. A few years after the war we were in contact with the mathematician Professor A. Walther, who had known John Aitken, and with the engineer K. Zuse who built the first programmable computer. Both of them advised us to acquaint ourselves with the use of high-speed computers and helped us to do so.
H.T. - Were you in the military service up to the end of the war?
K.H. - Yes. 1 gained a diploma from the University and then was sent to a few operational meteorological centres to gain experience. But by now it was 1944 or 1945, and things had become terribly difficult in Germany; there was no fuel left for the aircraft and meteorologists were of little military value. By this time I found myself drafted to a parachute unit, but the advance of the Allied Forces was so rapid that I never had to jump.
H.T. - Could you get a job in meteorology when the war was over?
K.H. - To start with I worked in a small shop as electrician, repairing radios and wiring apartments. I maintained a climatological station in my garden, and had my first scientific paper published: it was on how to measure the density of the air from a falling object. This topic had become rather important because of the interest in rockets.
H.T. - When did you become a full-time meteorologist?
K.H. - That was in 1949. Professor Weickmann had been appointed director of the weather service run by Germans in the American-occupied zone at Bad Kissingen, and he asked me to join the research department there under Professor Hermann Flohn. You may remember from your interview with Professor Flohn* that immediately after the war there were separate services in the American-, British- and French-occupied zones; it was only in 1952 that the Deutscher Wetterdienst - the national Meteorological Service of the Federal Republic of Germany — was created.
H.T. - You would be able to get hold of research papers from all over the world now that the war was over. Which works impressed you most?
K.H. - I was extremely interested to read of the work being done in the USA, largely under the leadership of Jule Charney. I was most impressed that, in his papers on the scale of atmospheric motions and numerical prediction of large-scale motions, Charney pointed to the reasons for baroclinic instability. With him in the forefront at that time were Arnt Eliassen and R. FjØrtoft from Norway and Norman Phillips, greatly helped by John von Neumann and Carl-Gustav Rossby. I was also able to read about the first numerical integrations of the barotropic vorticity equation. You know, there was a highly-gifted countryman of mine — Horst Philipps — who had formulated an equation for the ageostrophic wind component, and I found that, by applying the divergence factor to this wind, it gave the vorticity equation as stated by Charney and his colleagues. So I was very pleased that physics was now really being brought into meteorology. My main interest at this time was the problem of 'noise'; that is, the effect of short-period oscillations obscuring a longer-term tendency. I had a paper on this published in about 1951.
H.T. - In this paper you show that a trend in meteorological elements can be computed with sufficient accuracy by a numerical integration of the primitive equations.
K.H. - To integrate all the wave-values, including 'noise' which travels at about the speed of sound, would require time steps of a few seconds which would be quite impractical, even with the super-efficient computers of today. So we had to suppress at least the vertically-propagating noise signals by replacing prognostic equations by the hydrostatic and implicit Richardson equations. I still consider the elimination or dampening of noise to be the crucial problem in numerical weather analysis and prediction. Later on I found there exists an ordered hierarchy of diagnostic equations which, if applied to initial data, can guarantee that noise amplitudes are reduced in unfiltered systems. In filtered systems, they can replace prognostic equations and by going to increasingly higher orders they ensure that the synoptically important Rossby waves are conserved.
H.T. - Were your views shared by your colleagues in the Weather Service?
K.H. - I regarded myself at the time I joined the Deutscher Wetterdienst as being in the rising younger generation of meteorologists. The pre-war generation was mostly immersed in what I might call phenomenology, and they regarded our numerical approach with great suspicion. However, there were two notable exceptions, namely Professor Weickmann and Professor Flohn who both encouraged me wholeheartedly. In 1951 there was a symposium arranged at Bad Kissingen by the Meteorological Society of the Federal Republic of Germany, and one of the main speakers was Professor Rossby. I presented a paper on our work on simple barotropic and baroclinic equations. On Weickmann's recommendation, Professor Rossby invited me to go to his Meteorological Institute in the University of Stockholm for a few months, and, needless to say, I was very happy to accept. You will appreciate that there was very little money to support research in the Federal Republic of Germany in those days.
H.T. — I seem to remember that in Stockholm you were in favour of baroclinic equations, whilst Rossby was more for barotropic equations. How did he react to your ideas?
K.H. - My impression was that Rossby was coming to the conclusion that he had exhausted all the possibilities of barotropic models, but he was right in insisting that before going on to baroclinic models we must have thoroughly mastered the use of barotropic equations. You see, we in Europe had not yet had experience in using computers because there were so few of them. In fact there was only one at Princeton where they had built the ENIAC under von Neumann's supervision. Over there they were also thinking seriously about baroclinic models at this time; I had very valuable discussions with Arnt Eliassen and I read papers by Charney and by Eady on how to operate a baroclinic model. But at the 500-hPa level the barotropic model produced pretty good results, and it was to be a very long time before we developed a baroclinic model which could do as well, let alone better.
H.T. - What was your work when you got back to the Federal Republic of Germany?
K.H. - Professor Weickmann succeeded in obtaining a contract with the Geophysics Research Directorate of the U.S. Air Force* which gave us the means to set up a small numerical weather prediction unit within the research department and the welcome opportunity of working in close co-operation with Phil Thompson and his staff. Amongst others in my group there was W. Edelmann, Giinther Hollmann, Heinz Reiser and Friedrich Wippermann. Our work was basically analytical research, with a baroclinic model as the main objective, integration having to be done by manual operations. The group took courses in programming and software manipulation, and in fact we had access to a small primitive Remington machine using mercury tubes for internal memory. We had to think about what would be the best relaxation method and eventually chose eigenvalues for block relaxation. We realized that the Remington calculator was hopeless for our purpose, and therefore used some of our GRD funds to buy time on an IBM 704 computer installed in Paris — it cost one Deutschmark per second — and did our first baroclinic computations with it. But that was far too expensive for our limited means, so we managed to get access to a computer of the General Electric Company at Lynn, near Boston in the USA. A bit later, George Cressman arranged for us to use the computer at the National Bureau of Standards in Washington, D. C.
H.T. - All this work you were doing was research, not actual forecasting?
K.H. - It was research, in which we used actual data and integrated our baroclinic model in short time steps to predict over periods of up to a few days, but the forecasts were not in real time. I knew that we must have our own computer in the Federal Republic of Germany if we were to advance reasonably quickly, but I had a lot of difficulty in convincing our authorities that we should spend so much on what they regarded as the newfangled unproven concept of NWP.
H.T. - During this period I believe you went a stage further and introduced the primitive equations in place of quasi-geostrophic approximations?
K.H. - That is true. I felt the time had come to attempt numerical integration using the primitive equations for several reasons. For one thing, I was certain from my studies that noise amplitudes could be kept reasonably small by appropriate data initialization. Then the primitive equations promised to simulate atmospheric dynamics and energetics more realistically, and, moreover, they should also work in the tropics where the geostrophic approximation is invalid. Finally, the extra computer time required because of time steps in minutes rather than hours would, to some extent, be offset by much simpler solution operations and avoiding time-consuming relaxations and iterations. On my first attempt, using idealized initial data, I got a most encouraging result which reproduced new developments, occlusions, and even the kinks of isobars along a front.*
H.T. - Was any work along these lines going on in the USA at this time?
K.H. - I remember that soon after we had done that first run with the primitive equations I went to see Professor Smagorinsky in Washington, D. C. After seeing our results, he said that we had done a fine job, but added that his group had also had good results with primitive equations and intended to use them exclusively from that time on. So in fact our independent research efforts had both led to the same conclusion. I consider that the change from quasi-geostrophic models to primitive equations was a very important step in simulating atmospheric processes.
H.T. - You had done all that work without having your own computer which makes the achievement even more meritorious. When was it that the Deutscher Wetterdienst eventually acquired one?
K. H. - The then President of the Deutscher Wetterdienst, Dr Georg Bell, hesitated a long time about this, but thanks largely to the support of Dr E. Sussenberger (then in the Ministry of Transport) and his assistant Ernst Lingelbach, we purchased and installed a CDC 3400/3800 computer unit in 1965/66. We started using baro-tropic models and then went on to baroclinic models with several levels, but we always used primitive equations. I saw the transition of NWP through from a research to an operational activity in the Service. Then we went on to develop a computerized symbol-, chart- and TEMP-plotting facility, as well as objective isoline-drawing techniques from the plotted data. Here again, once it had been proved fit for operational use, this activity passed to the synoptic division.
H.T. - I understand that when Professor Flohn went to take up his appointment at the University of Bonn in 1961, you succeeded him as chief of the research department. How long did you stay in that post?
K.H. - Until I left the national Weather Service to go to the University of Mainz in 1968. My decision to take up academic life was based on several reasons. I felt that numerical weather prediction was now working satisfactorily in the Deutscher Wetterdienst, and that further progress would depend upon looking closely into the physics - work that could better be done in a university environment. Another reason was that I enjoy lecturing, and I think I am quite good at it. Moreover, the upper age limit for a professor to join a university was 55, and I was already 53.
H.T. - Was there already a meteorological department at the University of Mainz?
K.H. - There was a geophysics department, but the authorities realized that more theoretical physics needed to be injected into the meteorology because of the new interest in numerical modelling. That was why they offered me a chair of theoretical meteorology. I gave lectures and I also did research work. Atmospheric physics was my main subject, and for that I tried to separate different sub-disciplines such as convection, cloud physics, turbulence, the general circulation and thermodynamics. On average I had between ten and twelve students in each lecture, and I saw to it that during their first four semesters they concentrated on mathematics and physics, because only that way were they able to understand theoretical meteorology. I was very satisfied at the progress they made. My primary interest was the thermodynamics of irreversible processes and extending the linear Onsager theory to turbulent systems so that parameterized turbulent fluxes might be included in models.
H.T. - Looking back over your career, which European meteorologists have left the greatest impression on you?
K.H. - One of them would have to be Rossby because he was so full of ideas and gave to physical meteorology such a revolutionary impetus. Another is Jacques van Mieghem because of his comprehensive, meticulously exact and logical analysis of all problems in theoretical meteorology.
H.T. - How do you see numerical weather prediction developing in the future?
K.H. - I feel that national Meteorological Services should concentrate their efforts more on their zones of responsibility and leave the larger-scale circulation to centres like the ECMWF and WMCs. They can do this by nesting their zones with higher resolution within the global-scale products they receive, and then zoom down to mesoscale or even small-scale non-hydrostatic noise-filtered models. Here they should take in more and more physics; fields such as radiation, convection, turbulence, the boundary layer and atmospheric chemistry. Being of an irreversible nature, these are difficult to reproduce in larger-scale models. Of course, these smaller-scale models will require the storage and incorporation of various terrestrial parameters (for example topography, albedo, surface roughness or conductivity) in order to compute surface/air mass, energy and momentum exchanges. However, that would enable the Services to make detailed and realistic analyses and forecasts over limited areas, and to predict the evolution of an eventual hazardous emission (pollution or radioactivity, for instance) under any meteorological conditions. Universities should devote their efforts to formulating the physics of atmospheric processes in terms of adequate but simplified equations which could be handled by the operational models of Meteorological Services. Much progress in this domain has been made by our departments. Then again, interest is now shifting to climate modelling. As I expect you know, in my country, Professor Klaus Hasselmann at the Max Planck Institute in Hamburg is active in this field. Even on this scale we still depend on physics as the main tool.
H.T. - How do you pass your time these days, Professor Hinkelmann?
K.H. - I have little contact with meteorology following my serious illness in 1980. At my age one learns more slowly and forgets much more rapidly. I work in the garden and do the shopping, and I enjoy reading a good book. I still read some scientific works; I recently turned to Einstein's theory of relativity, and learnt that the metric form plays the fundamental role there as well as in atmospheric physics.
H.T. - What would be your advice to a young person contemplating meteorology as a career?
K.H. — I should first warn him or her that there are more candidates than vacant posts in meteorology. If this person is convinced that he or she wants to be a meteorologist, my advice would be to study mathematics and physics very very thoroughly. This would not be wasted even if there were no meteorologist posts available at the end of the course. Such a background will be immensely helpful in other fields of natural sciences. However, if this person enters the profession of meteorology, he or she should choose a domain in which to specialize, and try to become a real expert in it.
H.T. — Professor Hinkelmann, it has been very pleasant meeting you again, and this interview has been most instructive. Thank you very much for having given it, and I wish you many more years of congenial retirement.