COMPUTING YESTERDAY'S WEATHER

    Review of "Calculating the Weather: Meteorology in the 20th
    Century" by Frederik Nebeker Academic Press, 1995.

    How did meteorology join the ranks of the exact sciences, and how
    is weather prediction done today?  With few general accounts of
    computer forecasting available, the appearance of Nebeker's book
    aroused hopes that here at last was a comprehensive discussion of
    these questions.  However, disappointment awaits the reader seeking
    an understanding of modern weather forecasting. The book treats the
    ever-increasing role of calculation in meteorology but fails to
    provide a satisfactory picture of the state of play today.

    The central theme of the book is that computational methods have
    brought about a unification of meteorology. In 1900 three main
    strands were clearly discernible: climatology, dynamical theory and
    practical forecasting.  During the century, these three strands
    have become more quantitative and computational and a greater unity
    is apparent. To the unifying forces acting in the three areas
    Nebeker gives the self-explanatory if prosaic names "data-push",
    "theory-pull" and "science-not-art".

    Part 1 reviews the three strands in 1900.  The then status of
    climatology is captured in a quote from Joseph Henry:  "There is
    perhaps no branch of science relative to which so many observations
    have been made and so many records accumulated, and yet from which
    so few general principles have been deduced".  The theoretical
    tradition is reviewed but the treatment is patchy. The work of
    Ferrel, Hann and Exner is outlined and there is a mini-essay on
    Shaw, but one is left without any overall picture of the status of
    dynamic meteorology in 1900. Things improve with the chapter on
    practical prediction. The synoptic method is well described, giving
    a good feeling for the plight of the hapless forecaster.

    Part 2 deals with developments up to 1945. Bjerknes' programme to
    make forecasting a science is described, though the key 1904
    reference is missing. The graphical solution method is well
    illustrated by an example of calculating vertical velocity near the
    ground. Richardson's method and forecast are the subject of a long
    and interesting chapter.  Nebeker states that Richardson's work was
    highly regarded, but he does not satisfactorily explain why no-one
    followed it up even though Richardson had pointed to smoothing as
    the solution to his problems.

    The growth of meteorology between the wars is considered next. The
    dramatic influence of frontal theory on practical forecasting is
    described. Later (p129) Nebeker makes the dubious claim that the
    Bergen school methods were "probably more subjective than earlier
    procedures".  There is a brief but interesting biographical sketch
    of Rossby.  The remainder of this part concerns calculating aids
    used before and during the war.  The extensive treatment of
    punched-card equipment is rather dull but the descriptions of the
    D-day forecasts and of the war-time importance of meteorology are
    good.

    Part 3 considers the emergence of computer forecasting. The
    Princeton project of von Neumann and the crucial role played by
    Jule Charney are discussed in an excellent chapter. The subsequent
    introduction of computational methods into general operations and
    the greater cohesiveness which has resulted are then treated, but
    with virtually no material on work since 1970.  The European Centre
    for Medium Range Weather Forecasts is not even mentioned.

    The final chapter looks at Lorenz's discovery of chaos, which has
    brought about a scientific revolution and has profound implications
    for predictability of weather.  A major thrust of recent research
    has been to estimate the probability of different weather scenarios
    using ensembles of forecasts. This effort springs directly from
    Lorenz's work on sensitivity to initial conditions. Yet ensemble
    forecasting, which is inextricably intertwined with
    super-computing, is not mentioned; such an omission in a book of
    this nature is extraordinary.

    The style of writing is good and the book pleasant to read,
    although the end-notes beloved of publishers are a nuisance for
    readers.  About 450 references and a good index are included.
    Several errors were detected in the few equations displayed and
    even the list of equations on p51 is wrong. But the major complaint
    is that the book might have been written thirty years ago. Readers
    interested in the early history of numerical forecasting will find
    it a useful summary. Given the title, most scientists will be
    unhappy with the absence of a description of modern forecasting
    methods.