Akira Kasahara – Session I

Edwards:

Let me just start by saying some things for the tape. So we're here at NCAR interviewing Akira Kasahara about his career. It's the 2nd of November 1998, and also here are Joe Tribbia, Diane Rabson and Warren Washington. You guys, if you want to say anything, ask questions or anything, you're really going to have to speak loudly. This will pick him up very well, and me, but not you where you're sitting. The first thing I'd like to ask you is some particulars about your very early life — where you were born, where you grew up, what your parents did, and how you got interested in meteorology.

Kasahara:

Well, I'll be brief because that's not very interesting to talk about, for me. I was born in Tokyo. My parents worked in the application of photography and tried to make toys out of that. For example, like blueprint — I think blueprints were used more in those days, but that was mostly for technical applications. My father was working to make toys using the “wet plate” photography and blueprint technology.^[1] I don’t have much of special reminiscence concerning early days of my childhood.

Washington:

I have one. I discovered that Akira met his wife in kindergarten (laughter).

Kasahara:

(overlapping dialogue; inaudible) That is irrelevant.

Edwards:

But it's still relevant (laughter).

Kasahara:

My wife and I went to kindergarten together and elementary school. Then of course separated, but met later on and married. Obviously I had no idea whether we even lived together in those days (laughter). I went to middle school and high school, then entered to the University of Tokyo in 1944. I was interested in —

Edwards:

So the war was still on.

Kasahara:

Yes. That's right.

Edwards:

What were your experiences of World War II like? In Tokyo you had to have —

Kasahara:

Well, in fact, when I joined university in 1944, the Tokyo bombing was started. So you could not carry out classroom in Tokyo and we evacuated to countryside.^[2]

Edwards:

And continued to have the classes?

Kasahara:

And continued the classes. When I entered there the war was still going on, but it ended in 1944?

Edwards:

'45.

Kasahara:

That's right, '45. So I then returned to Tokyo and continued. In the fourth year, we had to select a particular area in the Faculty of Science and I joined the meteorology section in the Geophysical Institute.

Edwards:

Why?

Kasahara:

I was interested in astronomy and meteorology in general. Actually I was interested in seismology too, but I found from the professor in meteorology that the seismology's kind of more traditional, while meteorology is a younger science. So I was interested in trying to find out how young it is (laughter). I think my motivation is not as deep as some people. Anyway, I picked the meteorology.

Edwards:

OK. So this was under Professor Syono?

Kasahara:

Yes.

Edwards:

And part of what's interesting about the fact that you studied with him is that all the general circulation modeling efforts in this country, in the United States, except for Leith’s at Livermore, began as collaboration between an American and a Japanese émigré who had studied under Professor Syono at some point.

Kasahara:

Yes.

Edwards:

So you were saying (overlapping dialogue; inaudible) —

Kasahara:

Well, in fact that's just why I wanted to talk about that.

Edwards:

Yes.

Kasahara:

I graduated Geophysical Institute in 1948, and I received a Bachelor's of Science degree and still in the meteorology group. And I got into the graduate school and continued work on meteorology. About that time Jule Charney started the Princeton group, and in fact the first numerical weather prediction was performed in 1950 using ENIAC computer. During the war obviously we didn’t have any information coming from the U.S., but then after the war we received various publications —

Edwards:

Journals?

Kasahara:

Journals. With the help of Allied occupation armies which set up a library in Tokyo, they tried to bring Western countries’ cultures and information to Japan. So in that effort, they brought some of the meteorological publications too. And so —

Washington:

Can I —

Kasahara:

Yes?

Washington:

One interesting thing would be helpful to kind of understand, what you — did Syono have any collaboration with U.S. scientists?

Kasahara:

Not until 1957 or so. He was one of the very eminent Japanese meteorologists, and essentially developed similar theories as Rossby did, but of course his publications were in Japanese.

Washington:

But I mean he didn’t travel to international conferences, say in the '30s?

Kasahara:

I think the first time he went abroad was 1957 or ‘58.

Washington:

Oh. So he worked in isolation.

Kasahara:

'57. Yeah. Pretty much in isolation, right. But he developed pretty high level meteorological theories for mid-latitude disturbances. The Meteorological Society has a pretty long history. Now it's about 100 years.

Edwards:

The Japanese Meteorological Society?

Kasahara:

Yes. The Meteorological Society of Japan. The level of meteorology has been quite high, and there were many well-known meteorologists appeared in history, including Fujiwara and Okada and so on, who traveled, of course, to Western countries because that was before the war. But Syono was much younger and his activities took place pretty much during the war, so he never had any opportunity to travel abroad until after the war. Well, then — so we received much information on meteorological activities in the U.S. through publications available at the library arranged by the occupation army. So we started reading those publications such as Journal of Meteorology, Bulletin of the American Meteorological Society, Quarterly journal of the Royal Meteorological Society and so on. Those are the major publications nowadays. Before the war, the literatures available were mostly German publications, such as Meteorologische Zeitshrift, but none of the American journals were available during the war. And we realized that there was a tremendous development took place in the 1940s, '45 to late '40s, and —

Edwards:

Were these journals the first place that you heard about computers?

Kasahara:

No. The computers — the first numerical weather prediction run on a computer was performed in 1950.

Edwards:

Yeah, I know, but I'm saying during that period there must have been some kind of mention that computers had been developed and were starting to be used. Did you hear about that at all during the war?

Kasahara:

Well, I heard about electric computers had been developed around that time, but not obviously electronic computers. Relay computers were available in the early '50s.

Edwards:

To you?

Kasahara:

Not me, but in Japan. So for example — well, maybe we will come back to that later.

Edwards:

OK.

Kasahara:

So after reading those publications, the Japanese started developing — tried to catch up with those developments. And we formed a group, the Numerical Weather Prediction group. In the very beginning the name wasn't Numerical Weather Prediction, but then particularly in the 1950s, when the publication of the first numerical weather prediction appeared, that obviously became a trigger for more active studies toward numerical weather prediction. Then Kanzaburo Gambo, who was a research associate for the professor, was invited by Jule Charney to visit Institute for Advanced Study, Princeton. That was 1952. And then he started writing to me about numerical weather prediction activities in Princeton. He almost wrote a letter every week, and —

Edwards:

Why to you?

Kasahara:

Well, because I was at the University of Tokyo. In fact, I became a research associate in 1952 when he moved to Princeton, because his position became available —

Edwards:

Oh, I see.

Kasahara:

— because Gambo was invited by Princeton. So that position became available. I was a student when the position was offered, Gambo's position. Because of that connection, Gambo kept writing to me. And I passed that information to Professor Syono. Also, I passed the information to my colleagues — ^[3]

Edwards:

Oh, colleagues, yeah.

Kasahara:

At the Meteorological Research Institute and Japan Meteorological Agency, Weather Forecasting Group. So then we started developing — essentially tried to repeat what they have done. In those days the model was the barotropic model based on the conservation of vorticity expressed by the Laplacian of stream function. So there was a need of solving a lot of Poisson equations which we did by hand. Yes, in the early days we tried to do it by hand.

Washington:

Is that using graphical methods, or actually —

Kasahara:

No, (overlapping dialogue; inaudible) that was later developed by Fjørtoft. That was not very popular. I think more direct relaxation method; Southwell’s relaxation method was very popular. And in fact, in those days in Sweden too people were doing by hand and also they were using graphical methods. So the level of development was about the same in Japan as in other countries. So that happened in 1952 to 1954. But then —

Edwards:

But you're still doing all of this solution by hand at this point?

Kasahara:

Yes, that's right. But also in those days already some of the baroclinic model — a two-level model was developed and we tried to solve two-level models by hand, again by the relaxation method. But mostly it was just a tendency calculation. It’s not really time integrations. Actually, the time tendency gives a tremendous information. Then in 1954 Gambo decided to come back.

Edwards:

And you were out of a job.

Kasahara:

That’s right. I was out of a job. So I started looking around and, you know, obviously that was right after the war and jobs were scarce. I couldn’t find a decent job. Then, Professor Hidaka, who had visited Texas A&M, oceanography department and just came back, mentioned to me that they are interested in having a research assistant at the Oceanography Department in Texas A&M. Professor Hidaka suggested to me that I maybe able to apply. So I wrote to Prof. John Freeman. A part of the reason I went to Texas A&M was that John Freeman was one of the scientists who worked at Princeton in early days. So I thought that, you know, visiting John Freeman I may have a chance to get more information on numerical weather prediction.^[4]

Washington:

Can I just mention one thing to Paul?

Edwards:

Yeah.

Washington:

In that famous photograph that they show of the Princeton Group, John Freeman is one of those people.

Edwards:

OK. Before we go on to the Texas A&M, I want to ask a couple more questions about the Tokyo group. First just some general things about what it was like to work with Syono and how big the group was, and secondly, I'm interested in any memories you might have of Manabe and Arakawa from that time.

Kasahara:

OK. Actually, you know, Manabe is quite younger.

Edwards:

Yeah.

Kasahara:

I think he's about five years younger than I am. Arakawa is about the same as my age, although maybe a year or so younger.

Edwards:

A year or two younger than you, yeah.

Kasahara:

Yeah. Arakawa was really active working in the numerical prediction group. And Manabe, in fact, wasn't active at the time I left Japan in '54. Manabe may have been in the group, but because he was quite a bit younger, I don't remember much. I remember more about Miyakoda, because Miyakoda is about the same as my age, so we had interactions. And in fact for that matter, Matsuno is much younger. In fact, I haven't had any contact with him during the school years.

Edwards:

What was it like to work with Syono?

Kasahara:

He's a typical professor (laughter) who was interested in research but not so much on job opportunity (laughter). Well, in fact that's one of the reasons most of his students emigrated rather than getting job in Japan. While in other sections, other professors tried to create some positions within Japan, you know, creating research groups or something like that. But Syono has never tried to do that. He was really interested in doing research, but he wasn't very strong on personal contact or to influence people and so on. But I think he's a thoroughly scholarly person. He pretty much left to us whatever we like to do. So he's not really dictating, you know, but he's overwhelming. But one thing which was extremely interesting and I can never forget happened before I left in 1954. At the time I was at Meteorological Institute, Syono found a paper by Jule Charney on the baroclinic instability of the westerly current. That was published actually in 1947, but we found it quite a bit later around 1952 or ‘53. So it took a long time for us to see it. But Prof. Syono was extremely excited and he said that all meteorology had to change by this paper.

Edwards:

This is the famous paper on the numerical integration of the barotropic model?

Kasahara:

No, it's about the baroclinic instabilities in the westerlies — ^[5]

Edwards:

OK.

Kasahara:

It was published in Journal of Meteorology in 1947. In fact that was kind of opening to modern meteorology. Clearly Syono was able to recognize its value, because he had been working on the same topic, but he was not able to come to that point. So he wanted to let us know how things should have been done. Anyway between 1952 and '54, before I left, there was a lot of active effort to study the atmospheric dynamics as well as numerical prediction. That sort of trend continued even after I left in 1954. Then, I think during 1958 to 1959, '60 and those days, the numerical prediction group tried to use relay computers —

Edwards:

Oh yeah?

Kasahara:

— developed by Fuji Telecommunications Company. I think that was the predecessor of Fujitsu. That was the first time they used an electronic device to do this numerical prediction. Now, changing the subject a bit, I was at Texas A&M from 1954 to '56. One notable thing was that around that time many hurricanes hit the East Coast. In 1955, I think, Diane and Connie, two hurricanes hit the East Coast particularly. Then active efforts were made by the U.S. Government to promote research in tropical storms, hurricanes. In 1956 when I was still in Texas A&M, I was asked to attend a research planning committee at the Weather Bureau in Washington —

Edwards:

OK.

Kasahara:

I met, for the first time, with Jule Charney, George Platzman, Harry Wexler, Herbert Riehl, Robert Simpson, and so on. My research during graduate school was the structure and development of tropical cyclones and typhoons. So I had knowledge on the tropical cyclones and I was able to participate in that meeting. Then George Platzman said that he's going to have a contract from the U.S. Weather Bureau on tropical cyclones and asked me to join his project. That's why I decided to join the University of Chicago, working with George Platzman in 1956 and stayed for six years — I left Chicago in 1962. So for six years I was working on the numerical prediction of tropical cyclones, the movement of tropical cyclones, and later worked on the formation of tropical cyclones. But, one interesting thing happened during my stay in Chicago was that in 1960 Prof. Syono organized the first international conference on numerical weather prediction in Japan. Of course, I was not in Japan then and the conference was arranged by Syono with a help of numerical prediction group. So the members of the numerical prediction group, for example like Arakawa, Miyakoda, Manabe and so on, helped to organize the conference. But what is unique about the international conference —

Edwards:

Is this held in Japan?

Kasahara:

In Tokyo, at the Meteorological Agency. And most of the research scientists working on the numerical prediction around the world joined, including of course Jule Charney, Norman Phillips, Ed Lorenz, George Platzman, Fred Shuman, and so on. So, practically all people who were working on the numerical prediction in those days joined, as well as from Norway and Sweden, Fjørtoft, Eliassen, Bolin, Bo Döös, and from England, Knighting, and so on. So this conference exposed a great deal of Japanese activity and then that became an occasion to really exchange the information between Japan and various other countries. So like Yale Mintz talked to Arakawa, and that in fact became the occasion later for Yale Mintz to invite Arakawa to join UCLA. And so —

Edwards:

I think that's the following year, like 1961.

Kasahara:

Something like that.

Washington:

Can I add one little thing to that? And maybe I mentioned this in my earlier interview, that he invited Arakawa to come and, you know, later to come as a professor at UCLA.

Kasahara:

I don't think as a professor for the first time, but then in the second visit he became professor —

Washington:

The second time?

Kasahara:

Oh, the second time, right.

Washington:

And he did not have many publications and caused an enormous amount of difficulty in the UCLA system (laughter) though he had very strong letters from Jule Charney and others, but he didn't have the normal set of publications and that was something that was characteristic of Arakawa, very few publications.

Kasahara:

Now during 1954 — no, '56 when I joined Chicago and until '62, when I left, during the six years, I visited GFDL very frequently. Although George Platzman had a contract from the U.S. Weather Bureau and we worked on the numerical prediction of hurricanes, University of Chicago didn’t have any computers available to that project. The only way to use electronic computers was to work with the IBM 701 at JNWP.

Edwards:

I see.

Kasahara:

So I started using IBM 701 around 1954 and '55. Smagorinsky arranged with George Platzman that his project staff can use the computers. And in fact Ferd Baer who was working on the spectral method for weather prediction with George Platzman was also using the machines during that period. When I was visiting GFDL in connection with the use of computers around '56 or ‘57, Joe Smagorinsky asked me that he wants to invite a scientist from Japan to work on the GCM development. He said that he wants to have a person who has a strong background in atmospheric physics; because Smagorinsky himself is a modeler and so he wanted to have a person with a background in physics. He asked me who would be a good person to do so, and I recommended Manabe to him. That's how Manabe was invited to join his group. I think he came 1958 or so.

Edwards:

'58. Yeah. Now how did you know Manabe?

Kasahara:

I knew Manabe because, for example in 1954, he was a student and we worked in the numerical prediction group together, so I knew his work.

Edwards:

OK.

Kasahara:

And I remember in fact, after Manabe was arranged to come to GFDL, Syono asked him to finish PhD thesis. So at the time he joined GFDL, he just finished PhD. Joe Smagorinsky made very systematic efforts to try to help Manabe. Joe realized that Manabe need a good training on atmospheric radiation. So Joe arranged him to work with Fritz M?ller. And in fact they had a paper together.

Edwards:

Yeah, they do.

Kasahara:

Actually, Wiin-Nielsen joined the GFDL about the same time, but he didn't stay for long. And he came in fact to NCAR around '60 or '61. Joe also hired Doug Lilly about the same time that Manabe did. Doug Lilly didn't stay too long either and came to NCAR around ’63 or so. But Manabe stayed on, and became kind of a successful fixture at GFDL.

Edwards:

One question about this period through 1962, you mentioned here on this description of your research that you gave me that you used the IBM 709 at Argonne National Lab.

Kasahara:

Yes.

Edwards:

That was an AEC lab at that time, Atomic Energy Commission.

Kasahara:

I don't remember how the arrangement was made, but I used the 709 and 7090 which was a transistorized version. I did also at the National Bureau of Standards in Washington, too. Of course at that time I didn’t have citizenship so I was not able to go inside. In fact, I could not get in beyond the guards’ office. So I had to work at the guards' office.

Edwards:

At what? I'm sorry.

Kasahara:

You know that —

Edwards:

Oh, at the guards' office?

Kasahara:

Guards’ office! (laughter). There was a small room in the guards' office, and people took my program deck and brought me back the computer outputs.

Washington:

Akira, though, jumping this tiny bit ahead, you must have been able to get around this problem when you went to the Courant Institute because they had some security there because it was an AEC computer.

Kasahara:

Yes, I'm sure that's right, but I didn't have any problems of running a job at the Courant. Well, the computer belonged to Atomic Energy Commission. Yeah, I don't know how my computer time was arranged. It was all arranged by Bob Richtmyer, who was Director of Computing Center at the time I went. That was 1962 to ‘63, just before coming here.

Edwards:

I see, you went to the Courant Institute.

Kasahara:

After I left Chicago.

Edwards:

Yeah, and I gather you first met Warren there.

Washington:

He doesn't remember . . .

Kasahara:

(overlapping dialogue; inaudible) (laughter) OK, let me finish then I'll —

Edwards:

Sure.

Kasahara:

Why I decided to leave Chicago? Toward the end of my stay I started working on the development of tropical cyclones, as I said. Then in 1961, I received an invitation from Phil Thompson to join NCAR.

Edwards:

Had you met Thompson at that point?

Kasahara:

Well, yes. Before then I met him during my stay in Chicago. We had a conference on numerical weather prediction in Chicago. It was a big conference, around 1958 or ‘59. That was one of the U.S. activities complementing international conferences took place in various countries. So the Chicago conference was one of them. There I met many scientists, including Phil Thompson, though I don't think we talked very much. The reason that Phil Thompson gave me an offer, I believe, was by a suggestion of my colleague Ferdinand Baer, who was also working with George Platzman and worked on the spectral method. Ferd left Chicago 1960 or so and moved to Fort Collins to join the new Department of Atmospheric Science, organized by Herbert Riehl, as an assistant professor. I am sure that Ferd Baer and Phil Thompson had been talking and Phil obviously asked to Ferd Baer about me. So I think that was how Phil sent me an invitation and I visited Phil in December of 1961. Actually, Wiin-Nielsen picked me up at the airport and drove me to Boulder. In those days there were only a handful of people there at NCAR. Vin Lally was there. The offices were located at the old armory of the University of Colorado. Of course, the director was Walter Orr Roberts. Phil Thompson was associate director and Aksel Wiin-Nielsen was assistant director, and so on. So only a handful of people were there when I visited. Actually, I started to worry because at that time I was thinking about going back to Japan after a few years. I wanted to come to NCAR and maybe stay a few years and then go back, but then when I visited NCAR for interview it was just a handful of people. So if I were to stay a few years in Boulder, I may not get very much out of it. In those days, George Morikawa was also working on the tropical cyclones at the Courant Institute of Mathematical Sciences in NYU. Apparently, George Morikawa heard about that I may be moving to Boulder and suggested to Eugene Isaacson that it may be a good idea to have me at the Courant. Then Isaacson talked to Richtmyer who was Director of AEC Computing Center and I got an offer from Bob Richtmyer. That was how I decided to go to the Courant before coming directly to Boulder. I wrote to Phil Thompson that I want to take a leave of absence and in fact it was mentioned to that effect in NCAR annual report. If you look at the '61 or '62 NCAR annual report, there is the mention about I joined NCAR but I'm taking a sabbatical leave right now and working at the Courant. (laughter) I stayed at the Courant a year and a half. When I joined the Courant, James Stoker who was Director of the Institute mentioned to me that he is interested in meteorology and he had a theory of the occlusion process of extra tropical cyclones, namely how the cold front moves faster than the warm front so that eventually the cyclones occlude. And, he said that why don't you work on the numerical simulation of the occlusion process of frontal cyclones. So that was the job I took and completed a report within a year.^[6] Now they started saying, well, why don't you stay at the Courant, because they were very much interested in expanding a meteorological group. But obviously I'm not a mathematician, so I said I will go to Boulder. So I joined NCAR in 1963, June. I think you were already there or maybe —

Washington:

I think I came in September.

Kasahara:

September? Is that right? And then Phil Thompson was the head of department —

Washington:

Laboratory of Atmospheric Science.

Kasahara:

Yes, Laboratory of Atmosphere Science. When I saw Phil, obviously I asked him what I should work on, but Phil said you can work on whatever you want. (laughter) And so I started to continuing work on the tropical cyclone problem and then I did some calculations using a CDC 3600 or no — it was the university's — ^[7]

Washington:

IBM 709.

Kasahara:

709, yes. Then I got to know Warren. One day Warren came to me and said well, I want to work to build a general circulation model (GCM). So I said wow, I was really startled because he just graduated from Penn State. But then when he said that I was quite excited, because I had been thinking about the same thing. Actually I heard Norman Phillips’ talk on GCM when he presented in 1956 at an AMS conference [the 134th AMS National meeting on January 24 in New York City], because I attended the same conference to present a paper. I remember that Norman’s talk was very inspiring for me or I wanted to do something like that too. But even before then during Chicago, I learned through George Platzman about the book of L. F. Richardson [Weather prediction by numerical process]. So I ordered a copy of the book and to my surprise that book arrived. I still have that first edition of 1922 publication. So I had been interested in the GCM modeling. When Warren said he was interested in working, I said, OK, let's work together. And we talked about how we should approach. At that time already GFDL was developing a numerical model, and UCLA had a model, and Chuck Leith had a model, and so on. So I said that let us work on L. F. Richardson's formulation because that's something different. If it didn't work out then we can bring some other model, but at least if we do from scratch we can learn more. Besides, I was interested in demonstrating that the L.F. Richardson's approach may not be a failure. Saying failure, if the work is not good, it's a failure, but he actually didn't have any computers and couldn't try that. (laughter) So that was how we started working together and beyond that I'm sure Warren had talked about it.

Edwards:

Let me ask you something about a document you gave me in this set of things you sent earlier. This is the Report of Research and Facilities Programs, 1964 from NCAR and what it says is General Circulation Experiments: “Thompson has worked out a model which describes the long-term fluctuations resembling atmospheric flow patterns.” I won't read the rest of this, but the point is that this is a gross physical characterization of the state of general circulation. And then it says, “…the numerical scheme for solving this model has been worked out by Washington.”

Kasahara:

Warren was working with Phil Thompson as one of his earlier work.

Edwards:

So I don't think we talked about this when I talked to Warren. Say a little bit about that work. Was this really a General Circulation Model or was it just — Washington: No. Well, it was. (laughter) But it was a model that was kind of a theoretical model and that uses simplified equations. He actually simplified them down to some almost zonally averaged, deviation from zonally averaged flow equations, and the equations looked quite different than the ones that we would solve in a general circulation model. They were equations where you had things like Coriolis minus the shear of the zonal flow multiplied by other kinds of terms. And so I never thought that the model worked very satisfactorily because it had lots of assumptions in it which — I think Phil had trouble ever publishing it. Is that your recollection?

Edwards:

Yeah. So this didn't have much influence, or did it, on what you then did… Washington: No, it was pretty much a Phil Thompson theoretical model which —

Washington:

— as far as I could tell. And so he needed somebody to help put it into a computer and I programmed it for him. Maybe some other people were involved but I think I basically did it myself.

Edwards:

Well, why don't you talk some — maybe both of you could talk about the division of labor when you began to work on this GCM.

Kasahara:

Well, when we started, we decided to work together. That means when I do the formulation, Warren checks; when Warren develops, I check. And then we decided to write papers together, depending on who spends more becomes the first author. So obviously first I started to formulating and Warren did check it. And some of the details Warren started to work on like the diffusion term or something like that, and then I did checking. But fortunately in those days we got many programmers to assist us.

Edwards:

Programmers?

Kasahara:

Programmers.

Edwards:

Yeah.

Kasahara:

So for some of the output and input portions, like Bernie O’Lear, was helping us. What we programmed was the main scientific core. Both of us literally worked together in programming. Again, either one of us wrote the code and the other checked. So both of us we pretty much wrote the main code.

Washington:

I think somewhere in the archives there are some of those early notes — you remember — we wrote the equations out. I think you put them into a non-dimensional form and a derivation of those equations is in the archives.

Kasahara:

And also, I may have notes.

Edwards:

Oh, OK. I have a question. Regarding which groups, if any, you consulted with, because there were several ongoing modeling activities in the States at that time. Did you talk to any of them, get any advice of what to do or what not to do in setting things up?

Kasahara:

Well, as far as I know, Warren and Chuck Leith had a very close collaboration and, in fact, Warren visited him a number of times.

Edwards:

This was when Leith was still at Livermore?

Kasahara:

Yeah, Livermore.

Edwards:

He came to NCAR in 1965 maybe? When did Chuck come?

Kasahara:

Probably around that time. Chuck Leith didn't come to work on the GCM, but tried to work on the turbulence problem which Phil Thompson was interested in. In fact, I think Chuck Leith — it's my recollection Chuck Leith’s model is the first workable General Circulation Model.

Washington:

That's right.

Kasahara:

His model included the thermal convection effect. Around 1958 or so, Joe Smagorinsky was having a lot of problems. While I was visiting GFDL, I noticed that when he included the convective heating effect, small-scale convection tends to develop and it overwhelms the large-scale flow. So how to control was a major problem. And when Chuck Leith worked on the GCM, I was not quite sure how he was able to run without problem. But later on I found out what he did was to essentially reduce the heating rate in some fashion. Although he had a separate work to find out how to control the connective instability, what he did was essentially to reduce heating rate depending on stability.

Washington:

A thought occurred that what Chuck did — correct me if I'm wrong — was to have a parameterization which allowed for large vertical diffusion when the lapse rate got close to unstable.

Kasahara:

No, instead I think he reduced the heating rate, because if you have large dissipation or large diffusion terms that affects computational stability, unless you used an implicit method. And I don't think he used implicit method, because he was using essentially a semi-Lagrangian method for advection computations. Of course, he could use implicit-type vertical diffusion but I don't think he used that. Instead he reduced the heating rate. Because after all, the heating rate at one grid point should be adjusted according to the grid area. So if you get a very strong heating at one grid point and use that value to represent the entire grid area, it's obviously too much, you know. Well, in fact, I don't know why I didn't realize that in the first place (laughter). I had a lot of problems doing similar things for simulation of the tropical cyclone development.

Washington:

Well, I keep thinking, and maybe it's my poor memory, that you kind of found this out early with the tropical cyclone.

Kasahara:

That’s right. I found it difficult. I didn't offer how to solve it. (laughter)

Washington:

Yeah, but I thought that you kind of maintained that you had to keep the lapse rate stable.

Kasahara:

Well, I didn't do it. That's what Manabe did.

Washington:

OK.

Kasahara:

In fact, Manabe even suggested to me that I should use his technique for tropical cyclone development. But I didn't do it because I thought that's not right.

Washington:

OK. Because you remember, in our first running of the model with Joe Smagorinsky that we put together, that we came up with this very simple scheme of maintaining the internal temperature —

Kasahara:

Yeah, but...

Washington:

— in between two layers and whenever it was unstable we would conserve the internal energy, I should say, and therefore the lower layer became a little colder and the top layer became a little bit warmer and it made it stable. And the calculation was stable at that point with that simple trick.

Kasahara:

Yeah, but that knowledge was already available at the time we were working on it. So that's not my invention and somehow that convective adjustment was a practice used in general circulation modeling. But somehow I felt that's not good to use for tropical cyclone modeling and that was obviously my short sight. (laughter)

Edwards:

Just a question, just going back to that, what was the contact you had with other groups, you've been talking a bit about Leith. What about (overlapping dialogue; inaudible)?

Kasahara:

OK. Yeah. For example, I had a strong contact with GFDL, before I came to NCAR, but not after I joined NCAR, because we were developing a model based on the L.F. Richardson's and no other group ever tried that. When you look at the formulations, it's very clear and completely different. Different from the model either used at GFDL or UCLA. Now, in the case of UCLA, the model was essentially formulated by Yale Mintz. Obviously it was modified by Arakawa’s effort. Yale Mintz was essentially a synoptician and didn't have much of a modeling experience but he had a strong interest and that was why he wanted to work with Arakawa. And I think my understanding is that Yale Mintz really wanted to work on modeling. But Arakawa felt that the first thing he should do was to have a very good numerical scheme. So Arakawa was spending most of his time on developing a numerical scheme and in fact I heard at some time Yale Mintz was very irritated that Arakawa was spending so much time on developing what we now know as the Arakawa-Jacobian. But obviously once it was done then Arakawa more or less took over Mintz's activities and pushed the general circulation modeling. At GFDL after Manabe succeeded in integrating with the convective adjustment ideas, then the modeling activity was taken over by Manabe. So although Joe did a lot of the earlier development, I think Manabe sort of finished the work.

Tribbia:

I was going to ask you, Akira, you mention that convective adjustment was a step forward in terms of maintaining stability but before that there were some numerical questions in stability, integrating the primitive equations for a long period of time, and Arakawa was one solution to that dilemma. I know there were others. What years was that problem solved directly?

Kasahara:

Well, as far as the NCAR model was concerned, obviously we felt that that's an important topic and David Houghton joined NCAR about same time — when was that?

Washington:

Yeah, '63 or (overlapping dialogue; inaudible).

Kasahara:

And so I asked Dave Houghton to work for me and I think Dave worked on the various schemes. When we started essentially we took the Lax-Wendroff type approach, which I heard about at the Courant, and also Norman Phillips used that in his numerical model.

Washington:

In fact, Akira, what seemed to be an influence on us, too, as I recall, was a technical note, I think it was prepared by Richtmyer? Which was kind of a survey of numerical techniques but didn't include Arakawa's scheme but it essentially dealt with schemes that came out of the Courant Institute, I think.

Kasahara:

In fact, there was an interesting story floating around, namely that the Courant Institute people did not recognize the Arakawa's Jacobian study. And I don't know to what extent that was a serious matter. My impression was Arakawa felt that the Courant Institute people were not recognizing his work. Now the Courant Institute people obviously knew Arakawa's work, but probably I should say the Courant Institute people had their own development and Arakawa's work didn't quite fit what they had been working. And that was why they just didn't pay attention. I don't think it's not correct to say that they even opposed — I don't think they did that, although some people had some different views.

Washington:

OK, well, Akira probably knows this better than I do, but my impression that the people at the Courant Institute were used to dealing with shock problems and front problems so that when something got down to the grid scale and you damped it somehow, whereas Arakawa was more interested in conservation and keeping energy from going into the smallest scale, and so when you're dealing with shocks or fronts or whatever, you want dissipation to be up to the smallest scale. And so a different sort of philosophy, I think.

Kasahara:

Well, it may be true that since we had been — Warren and I — using Lax-Wendroff approach of numerical scheme, people felt that we may be ignoring their development. Well, it's true that we didn't quite follow Arakawa's. We tried to develop our own way even though Arakawa spent so much time on developing numerical scheme perfection. So, they might say well NCAR was ignoring use of Arakawa's scheme, but that's not true. Since I had been at the Courant, I was exposed to that scheme. Bob Richtmyer was visiting NCAR and wrote a technical report. David Williamson joined NCAR about the same time, and continued work on the numerical scheme, and Dave Houghton worked on some of long-term integration programs. So we had a sort of the component for developing GCM. Now incidentally, we had Takashi Sasamori, who was a student of a well-known atmospheric physicist, Yamamoto, who had developed the so-called Yamamoto chart for graphical computing methods of radiation. He was visiting Julius London at the University of Colorado. So after his work with Julius London finished he joined NCAR and he was the person to work on the atmospheric radiation.

Edwards:

Let me back up just one question, I mean one — a little bit, and ask a sort of sociological question. Why is the Courant Institute important? I don't know much about it or what it is, and I hear you guys saying that you have focused on their methods and what they thought, or methods of people there, as opposed to Arakawa. Is there a prestige thing there?

Kasahara:

Well, no, if you go through what Phil Thompson has done, Phil had a lot of contact with the people in mathematics and well, our connection to the Courant was obviously a part of business, partly because I came from the Courant Institute. And there was a kind of almost bible in those days written by Bob Richtmyer. What was it called?

Tribbia:

“Difference Methods for Initial-value Problems”. Something like that.

Kasahara:

Yeah. That was probably the first kind of book appeared in the applied math. Nothing was available and in fact, there's an interesting connection also. A student of C.G. Rossby, Heinz Kreiss in Sweden, was very much interested in numerical methods. Let's see, when was that? But anyway, around '62, Heinz Kreiss had been working with Eugene Isaacson at the Courant. Phil Thompson knew Heinz Kreiss, because Phil was visiting the Institute of Meteorology in Stockholm. In fact, NCAR was started even before Phil came to Boulder. Phil joined NCAR at the time he was at Stockholm. So, Phil had a lot of contacts with people in Sweden and Heinz Kreiss was one of them, because Heinz was a programmer for C.G. Rossby. Anyway, Heinz Kreiss worked closely with Eugene Isaacson and Bob Richtmyer at the Courant. So that was kind of connection we had. I think the Courant people realized that the application of numerical methods to meteorology is one of the important topics. In fact, that's the reason the Courant Institute always wanted to have a meteorological group. They continue to have a small group even now and, in fact, Joe (Tribbia) visits there also from time to time.

Edwards:

This is really a question for you, Warren. You said a minute ago that they were —Courant Institute people — were their main paradigm is shock waves and I'm curious why that was true. Were these people who had been working on the bomb?

Washington:

Yes. They all came — well, I wouldn't say all — but they worked with the Manhattan Project at Los Alamos and helped out with the Manhattan Project. A number of them have.

Tribbia:

Bob Richtmyer in particular.

Washington:

Right.

Tribbia:

And Bob had found a numerical technique for stabilizing shocks, the artificial viscosity and so this goes precisely along the lines of what Warren was saying. I might add, my impression is that the Courant group had been working for the decade of the '50s on the problem of numerical fluid dynamics, computational fluid dynamics, primarily motivated by the problems associated with simulation of —

Edwards:

Simulation of explosions.

Tribbia:

Right. It was also known that their expertise would help in general computational problems and so Eugene Isaacson and Bob Richtmyer and the general group there were an obvious group to collaborate with on a number of problems along those lines.

Edwards:

Interesting.

Kasahara:

Well, clearly, Phil Thompson was very much familiar with those activities. And obviously he approved of these activities, although he didn't try to engineer any kind. But obviously with his understanding we were able to do quite freely to contact any group or seek any connection. Also in the early days of NCAR, one of Phil's ideas was to try to have top notch scientists and then let them to try to mingle with the younger generations. So everyone tried to promote that idea.

Washington:

And actually, partly his question about, did we have much contact, that — I think even in the early days, '63, '64, '65, we had a number of visitors coming to NCAR all the time, from Europe and other places. Mintz came many times. I don't think Smagorinsky visited, but Chuck Leith came many times. Seemed like this was a place where there were a lot of meetings, people stopping in all the time.

Edwards:

OK. Oh, I had one more question about from this little document of 1964; this is a short description of your early efforts on the GCM. The last sentence here says, "Jordan is working on the preparation of initial data to be used in the calculation."

Washington:

Who is that?

Edwards:

Jordan.

Washington:

Jordan?

Washington:

Oh, I know who it is. Yes, yes. We —

Kasahara:

Cuban scientist [Placido Jordan — ed.].

Washington:

Yeah, yeah. Let me — (laughter) — Akira probably knows this better than I do, but apparently when Castro took over, a lot of people fled. I don't know exactly what Jordan's status was, but he was high in the weather service. And there was some pressure on us to employ these Cubans who, as kind of like they were refugees. I don't know how he was hired, but he shared an office with Dave Baumhefner and I remember he smoked continuously cigars. (laughter) And he wasn't much of a —well, I should be gracious. He didn't seem to know very much.

Edwards:

That's being gracious?

Washington:

He didn't like the fact that he'd been some high official in the weather service in Cuba, right? And then he was here drawing maps under the direction of a young scientist like myself. I don't think it agreed with him very well, he was very angry all the time.

Kasahara:

He was working for actually Wiin-Nielsen before he started working with us.

Washington:

Is that right? Oh, I'd forgotten that. OK. I'm glad it happened, personally.

Edwards:

OK.

Edwards:

I brought it up not so much because — to find out who he was as to talk about this issue that I'd spoken about with you of how you tested your model —

Kasahara:

Well, I think that aspect maybe I should mention. When we built the model we realized that someone has to do checking. Since neither of us is very strong on synoptic situations, we decided to hire a person who has a strong background in synoptics. I don't know how we came about but we heard about David Baumhefner, who was a student of Yale Mintz.

Washington:

I thought it was Krishna [T.N. Krishnamurti]?

Kasahara:

Oh that's right, Krishna. Maybe Krishnamurti was the one who introduced us. But anyway Dave applied the job, tried to do the data business, namely to look at our outputs to see if the model is doing all right or not. We worked on the climate model which is global in nature and tried to compare with reality, but we really didn't have any global maps. One way to check the performance of the model is to do global real data forecasting and to see how well the model performs. But we didn't have any input data in those days. So David Baumhefner’s first job was to prepare global maps, I think five days or so with the help of Harry van Loon, particularly for the southern hemisphere. Well, the northern hemisphere wasn't too bad, but the quality of observations and the number of observations in the southern hemisphere was very poor, but there were special observations called IGY data during 1950s.

Washington:

'57, '58.

Kasahara:

Yeah, something like that. And using that data and with the help of Harry van Loon’s climatology, David Baumhefner was able to analyze five days or so for three levels.

Washington:

Akira, wasn't —

Kasahara:

Surface, 500 mb, and 200 mb.

Washington:

I'm just trying to remember but was that the first global data set —

Kasahara:

That was the first daily global data set ever analyzed. Used for global forecasting, testing global forecasting. Well, of course then the time was getting into the GARP.

Edwards:

Let me just ask one other question about this early period of development which is sort of '63, '64, to '68, shall we say '67, which is the first publication of the model itself. I already asked Warren this question but I want to get your answer to it also. Did you have — was the issue of climate change at all in your mind as a reason for developing this model?

Kasahara:

Well, as far as I'm concerned, I was not really deeply thinking of how to use the model after we developed the model.

Edwards:

You just wanted to build it.

Kasahara:

I think my primary motivation was to demonstrate whether the L.F. Richardson formulation was a complete development. Warren maybe had a slightly different view. Because, for example, one of the things that Warren did was to make a movie after we developed the model. The first thing was a two-level simulation. We tried to show how baroclinic waves develop from the beginning and I thought that was a very nice way of demonstrating how baroclinic waves start develop from the initial start of no motion. Also Warren made monsoon movies. So I was not really trying to use that model for climate change or something like that.

Washington:

I didn't hear much talk about how to use the model even among anyone of the modeling group. Actually no one was saying that they were going to use the model for say global warming or climate change. It was more trying to make it simulate the observed atmosphere. That was the goal.

Kasahara:

Yeah. I think my primary motivation was to do an accurate simulation, how accurately, like verifying using actual data and try to do a global forecasting and short-range forecasting. But we haven't heard much talk about how to use the model for global change. The one thing Warren did an interesting experiment in those days. I guess someone probably suggested him to work on, but the problem was to study the impact of metropolitan warming by thermal pollution.

Washington:

Yeah, that's right.

Kasahara:

What is the thermal pollution effect on the weather? Warren made an experiment using our GCM. I remember that you did an experiment by adding a positive heat. Also you did the same by subtracting heat. To my surprise, they both came out with the same kind of changes, wasn't it?

Washington:

Right. Very small differences.

Kasahara:

That really shocked me. In fact, when Warren showed me I wasn't quite sure how to (laughter) reconcile that phenomenon. But if I think about it now that was an early experiment of predictability.

Washington:

Well, in fact, I mentioned to him in my interview that the signal was so small it kind of worried us so we then started, I think, Steve Schneider and Bob Chervin started coming up with a student T-test to see if the signal to noise actually gave you something, and so I don't think we carried out any other experiments before that that had any — that were kind of climate change experiments. I think that was 1969 or '70, '71, somewhere around that time. So I think that was our first. And then also I should add, I mentioned in my interview that we already know from Manabe's experiments with a radiative convective model about global warming, or increased CO₂ concentration. So we kind of knew that was the direction that things were going to go, but actual three dimensional models I don't think were being applied at that point.

Kasahara:

I left GCM modeling work around 1972 when I went to spend a year at Stockholm, University of Stockholm. So Warren pretty much took over after that and worked with David Williamson to continue the work of NCAR approach. But then the spectral technique came along with the participation of Chuck Leith. Chuck Leith really liked the spectral technique so the model started gradually changing, changed to spectral. But the original NCAR formulation was taken over by Bob Dickinson and applied it to the thermospheric model. He mainly tried to push up the model top, higher up in the thermosphere. He and Ray Roble continued the line of approach we started. I don't know what kind of model they're running now, but that model uses the height coordinate. When you use the sigma system, you can have a fine vertical grid resolution in the troposphere. But when you go higher up you will have a rather coarse vertical grid resolution. So you cannot use the P system or sigma system at high levels in the thermosphere. Since our model is based on the height coordinate, that's why I believe they took the height coordinate approach. L.F. Richardson's approach uses the height coordinate which we tried. The only difference is that Richardson's model went to infinity. Obviously that doesn't work in a model of finite top, so we had to change the formulation to apply it to a finite top.

Edwards:

I see.

Kasahara:

So that's the only modification we made, other than of course the numerical integration scheme and various physical processes and so on. So the basic dynamic formulation was based on L.F. Richardson's.

Edwards:

Well, listen, we've been going for almost two hours, and I know Diane has to leave. Maybe this would be a good time to stop for today and take this up again tomorrow? OK. Thanks to all of you for being with us.