The basic design of ARM expects that models of the radiation field will be based on observations of the state of the atmosphere rather than on predictions. There are two basic approaches that can be taken to specify which atmospheric observations are required to support models of the radiation field. The first approach would be to ask which set of physical conditions would a GCM either use now or anticipate using to predict the radiation field at some time in the future. The second approach would be to ask which observations in general would be required to predict the radiation field, without restricting our considerations to the necessarily limited set of physical conditions available to a GCM.
In setting the requirements for the atmospheric observations we have chosen the latter approach, making sure that the GCM requirements are a subset of the ARM observations. The reason for this choice is very simple. As described by Gates (1985) the development of a GCM is a matter of choices. The choices reflect the necessary trade-off between capturing the critical physics and the computational limitations of modern supercomputers. Since ARM will guide the treatment of the radiation field in GCMs, it too must contend with this trade-off. In essence, ARM's approach is to fix the resolution and computational characteristics and then to seek the appropriate physics of radiation and cloud processes.
The set of observations required to specify the atmospheric state fall into the categories of basic meteorological measurements, atmospheric composition, geometrical structure and surface properties.