24
Francis Bretherton
coefficients that do not depend on individual data but only on their statistics and
array geometry, then the assumption of normal distributions was no longer necessary.
The framework could then best be described as “linear estimation with least mean
square error.” He also later found that the central concepts had been anticipated in a
meteorological context by Gandin in Leningrad in 1963 and indeed by Gauss in the
nineteenth century (Bretherton, Davis, and Fandry, 1976). I have since noted several
other disciplines in which the technique has been independently rediscovered.
Meanwhile, the remainder of the MODE Program was also making progress.
In the Scientific Council, there was consensus that the primary objective was to
map for a 3-month period the density and horizontal velocity fields throughout the
water column below the seasonal thermocline, in an area away from the Gulf Stream
that included both smooth abyssal plain and moderately hilly topography. The site
was tentatively identified. A secondary objective was to assess the capabilities of
promising instrument systems that might contribute to more extensive measurements
in the future. Array 1 of current meters and hydrographic soundings was deployed
to obtain preliminary measurements of what might be expected. A Program Office
staffed by Dennis Moore was opened in Cambridge, Massachusetts. The Executive
Committee began to meet every 2 weeks at MIT, and I had a loud, if naıve, voice in
its deliberations, during which I began to appreciate the complexities of organizing
and coordinating an enterprise on this scale. On my return home to Baltimore in the
evening, my daughter used to sniff the tobacco smoke in my clothing and exclaim,
“Daddy, I know where you’ve been!”
The summer of 1972 was a critical time for finalizing the plans for the Field
Program a year later. The Theoretical Panel met for 3 weeks at NCAR in Colorado
with intense and sometimes heated discussion. On the side Mike Karweit was busy
in the computing center and our modeling effort made great progress. Meanwhile
the Observationalists deliberated less formally on the East Coast, though there was
substantial cross-fertilization by visitors in both directions and individuals with roots
in both groups. As I remember it, the major conclusion of the Theoretical Panel
was that the proposed array of moorings at the center and corners of a hexagon of
side 100 km was too widely spaced to allow mapping of velocities throughout the
containing circle to an accuracy of ±20%. This conclusion derived from estimates
of expected mean square error computed from the measurements made in Array 1
(also known as MODE 0), using the technique of objective mapping described above.
Unsurprisingly, it was also controversial, because most oceangoing scientists were
accustomed to making their own judgments about such matters, and here was a bunch
of theoreticians telling them what to do. After the ensuing debate the actual design
retained the 100-km circle but three moorings were inserted closer to the center. In
the event, Murphy’s Law took its bite and a large number of the current meters in
the main array failed, substantially degrading the outcome. The major achievement
was that, in the MODE-1 Atlas (1977) that became the definitive summary of the
observational results, for every variable there were objective analyses of the observed
Francis Bretherton
coefficients that do not depend on individual data but only on their statistics and
array geometry, then the assumption of normal distributions was no longer necessary.
The framework could then best be described as “linear estimation with least mean
square error.” He also later found that the central concepts had been anticipated in a
meteorological context by Gandin in Leningrad in 1963 and indeed by Gauss in the
nineteenth century (Bretherton, Davis, and Fandry, 1976). I have since noted several
other disciplines in which the technique has been independently rediscovered.
Meanwhile, the remainder of the MODE Program was also making progress.
In the Scientific Council, there was consensus that the primary objective was to
map for a 3-month period the density and horizontal velocity fields throughout the
water column below the seasonal thermocline, in an area away from the Gulf Stream
that included both smooth abyssal plain and moderately hilly topography. The site
was tentatively identified. A secondary objective was to assess the capabilities of
promising instrument systems that might contribute to more extensive measurements
in the future. Array 1 of current meters and hydrographic soundings was deployed
to obtain preliminary measurements of what might be expected. A Program Office
staffed by Dennis Moore was opened in Cambridge, Massachusetts. The Executive
Committee began to meet every 2 weeks at MIT, and I had a loud, if naıve, voice in
its deliberations, during which I began to appreciate the complexities of organizing
and coordinating an enterprise on this scale. On my return home to Baltimore in the
evening, my daughter used to sniff the tobacco smoke in my clothing and exclaim,
“Daddy, I know where you’ve been!”
The summer of 1972 was a critical time for finalizing the plans for the Field
Program a year later. The Theoretical Panel met for 3 weeks at NCAR in Colorado
with intense and sometimes heated discussion. On the side Mike Karweit was busy
in the computing center and our modeling effort made great progress. Meanwhile
the Observationalists deliberated less formally on the East Coast, though there was
substantial cross-fertilization by visitors in both directions and individuals with roots
in both groups. As I remember it, the major conclusion of the Theoretical Panel
was that the proposed array of moorings at the center and corners of a hexagon of
side 100 km was too widely spaced to allow mapping of velocities throughout the
containing circle to an accuracy of ±20%. This conclusion derived from estimates
of expected mean square error computed from the measurements made in Array 1
(also known as MODE 0), using the technique of objective mapping described above.
Unsurprisingly, it was also controversial, because most oceangoing scientists were
accustomed to making their own judgments about such matters, and here was a bunch
of theoreticians telling them what to do. After the ensuing debate the actual design
retained the 100-km circle but three moorings were inserted closer to the center. In
the event, Murphy’s Law took its bite and a large number of the current meters in
the main array failed, substantially degrading the outcome. The major achievement
was that, in the MODE-1 Atlas (1977) that became the definitive summary of the
observational results, for every variable there were objective analyses of the observed