With chapters written by specialists, The Development of Atmospheric General Circulation Models is aimed at researchers, but at those elsewhere within geophysics and related areas, not just at those working directly on GCMs. So the reader is not expected to have any detailed knowledge of specific models or their implementation, just an understanding of the basics of atmospheric science and some idea how a circulation model works. The result is a work that will be accessible to many outsiders as well. It is well illustrated with diagrams and maps, some of which are repeated in colour in a separate sixteen page section.
In "From Richardson to early numerical weather prediction" Peter Lynch goes back to the earliest circulation models: Richardson's 1910 hand-computation, which "blew up" due to numerical problems, and the ENIAC computation of 1950, which has been reimplemented and even ported to run on a mobile phone. Early applications were to weather forecasting, but Norman Phillips ran the first long-range simulation of the atmospheric circulation in 1956.
Looking at "evolution and future research goals", Warren Washington and Akira Kasahara trace the history of the major early research groups (GFDL, UCLA, LAM, NCAR) and the subsequent proliferation of GCM work. They chart the steady increase in model complexity and look at some of the goals of current climate and Earth System modelling.
A chapter by James Rodger Fleming focuses on the later career of Harry Wexler, in particular his role in early applications of computers to meteorology. This was tied up with ideas about controlling the weather and climate (which once again have salience, with consideration of geoengineering to avert global warming). As something of an aside, Fleming also gives a brief account of Gilbert Plass' early warnings about the dangers of anthropogenic carbon release.
Drawing on experience with the Met Office Unified Model, Senior et al. survey some of the synergies between climate models and numerical weather prediction. One approach is to build a "seamless" model that can be run at different scales for different purposes. They give an example of modelling orographic drag on the Alps, where there is a steady "hand over" from parameterisation to explicitly resolved forces as the resolution moves from 60km down to 1km. Such unified models have applications in data assimilation and reanalyses, the use of initial tendencies to diagnose model errors, ensemble generation, model initialisation, and diagnostic techniques and metrics.
Ngar-Cheung Lau surveys "the contributions of observational studies to the evaluation and diagnosis of atmospheric GCM simulations", focusing on large-scale dynamical processes. In the early years the focus here was on the energy cycle and zonal-mean features of the circulation; there followed a shift to looking at frequency dependency and longitudinal variability, local interactions between transient fluctuations and the time-mean circulation, and interactions with sea surface temperatures. Such traditional comparisons between observations and model output remain important even with long-term projections involving Earth system models, in the verification of hindcasts.
Ocean models have some similarities to atmospheric GCMs, but also have major differences (notably in having much smaller relative density contrasts and being largely opaque to radiation). Kirk Bryan also considers some of the different coordinate systems used in ocean models. Exploring "coupling atmospheric general circulation to oceans" he looks at how coupled models compare with observations — simulation of the thermocline is a challenge — and how they have been used to study paleoclimates, El Niño, and multi-decadal climate variability.
Robert Dickinson looks at "coupling atmospheric circulation models to biophysical, biochemical, and biological processes at the land surface". Land surface models have steadily increased in sophistication, with handling of evapotranspiration and radiative fluxes incorporating models of land use change and dynamic vegetation. And there has been much work on the terrestrial contribution to carbon cycling and its uncertainties — poorly characterized "old carbon", with long residence times, includes "charcoal and carbon locked within soil microaggregates" as well as stores in permafrost and wetlands. Surface modelling is typically done at a much greater spatial resolution than that of the attached GCM.
In "The Evolution of Complexity In General Circulation Models" David Randall glances over a number of topics. He gives a very brief history of core numerics, touching on the introduction of spectral methods and the more recent shift away from them. He looks at the difficulties of cumulus cloud parameterisation, illustrating the problems facing attempts to build modular models. He suggests two-track modelling as a way of combining bleeding-edge research and operational weather and climate services. And he considers the challenges posed in data processing and storage by high resolution cloud-resolving models, where "simulation of one annual cycle will produce several petabytes of model output".
The development of GCMs has spanned much the same time-span as the rise of concerns over anthropogenic global warming. A final chapter by Richard Somerville looks at how the use of GCMs in the IPCC process has evolved since the First Assessment Report, where "assessment of GCM results was largely process-oriented, devoted to exploring the behavior of an important but restricted part of the climate system". This focuses down a little on coupled models and cloud processes, but also steps back to give a broad view of the IPCC process and the surrounding policy debate.
There are inevitably compromises that have to be made in covering such a broad area — a chapter on the handling of uncertainty would have been a nice inclusion — but The Development of Atmospheric General Circulation Models is not a collection of loosely related papers but a nicely presented volume that makes sense as a book. It gives a feel for the challenges, past and present, faced by modellers, and for the growing sophistication of their work.