POSTSCRIPT

In this book, we have presented, in historical order, many of the studies of global warming that have been conducted since the end of the nineteenth century, when Arrhenius conducted his pioneering study described in chapter 2. These studies have used a hierarchy of climate models with increasing complexity, such as an energy balance model, a 1-D radiative-convective model, and a 3-D GCM of the coupled atmosphere-ocean-land system. These models have been very useful not only for predicting, but also for understanding climate change.

As described in chapter 4, a GCM of the atmosphere consists of prognostic equations of state variables such as wind, temperature, specific humidity, and surface pressure. Each prognostic equation usually consists of two parts. The first is based upon the laws of physics, such as the equation of motion, the thermodynamic equation, Kirchhoff’s law of radiative transfer, Planck’s function of blackbody radiation, and the Clausius-Clapeyron equation of saturation vapor pressure. The second part includes parameterizations of various sub-grid-scale processes such as moist and dry convection, the formation and disappearance of cloud in the atmosphere, the budget of snow and soil moisture at the continental surface, and the formation and disappearance of sea ice at the oceanic surface, among many others. In the 1960s and 1970s, when early versions of GCMs were developed at various institutions, electronic computers were in the early stages of their development and their capabilities were limited. This is an important reason why the parameterizations of these sub-grid-scale processes in early models were made as simple as possible. Nevertheless, it was encouraging that these models successfully simulated many salient features of the general circulation of the atmosphere and the distributions of temperature and precipitation, as shown, for example, in chapter 4. It is also encouraging that a GCM of the coupled atmosphere-ocean-land system, constructed about 30 years ago, successfully simulated the geographic pattern of surface temperature change that has been observed during the past several decades, as noted, for example, by Stouffer and Manabe (2017).

Owing partly to the simplicity of parameterizations and low resolution, the computational requirements of these models are much smaller than the climate models that are currently used for predicting climate change. Thus, it has been possible to conduct countless numerical experiments, changing one factor at a time, using the models as a virtual laboratory for exploring the inner workings of the climate system. In fact, the simplicity of parameterizations has facilitated greatly the diagnostic analysis of the results obtained. For these reasons, climate models with relatively simple parameterizations have been and are likely to remain very powerful tools for exploring climatic change of not only the industrial present but also the geologic past.