The purpose of ARM is to perform an intensive series of atmospheric radiative transfer measurements to validate cloud models and existing radiative transfer codes and ultimately improve the treatment of radiative processes in GCMs. To be successful, these measurements must encompass all climatologically significant conditions that influence large-scale climate, particularly those that interact to produce strong feedback mechanisms.
The focus of the ARM measurements is the basic physics of GCMs. However, these physics are not immutable, as in the sense of a physical law. GCMs integrate elements from theory, basic physics, and observation. They are computational tools and, as such, only approximate reality. This approximate treatment is very much at issue in the discussion of the parameterizations used in the models. As has been described here several times, the use of ARM data is not only to confirm the details of the basic physical processes, but to understand what physical processes and effects must be preserved as the problem is solved in the coarse-resolution GCM case.
The application of the first two criteria, climatic significance and climate sampling, captures the essence of the multiple site problem. The parameterization of clouds in GCMs is so important that it is absolutely necessary to confirm observationally the correctness of those parameterizations in those regions of the globe that are important to climate modeling. More than one region is important. Further, there is sufficient diversity among the climatically important parameters at a particular site that no single site can be thought to adequately explore the meteorology to ensure proper parameterizations for GCMs.
A single location cannot experience all of the necessary conditions to form a definitive study. A single location cannot be found that will experience both a humid tropical climate as well as mid-latitude and subarctic regime. One of the big differences will be in cloud properties, a major focus of this experiment. Low-level tropical clouds are composed of water and water vapor and have temperatures above freezing; ice and supercooled water are usually found in only the upper portions of thunderstorms. Mid-latitude cloud systems are more variable but are predominantly "cold clouds" in winter and at middle and upper levels all year. Other latitude-dependent parameters such as tropospheric depth, varying from 10,000 to 20,000 km, will be important.
Another large difference in cloud properties is determined by the air-mass characteristics during cloud development. Marine air masses contain relatively few cloud condensation nuclei (CCN), the small hygroscopic particles that form the center of cloud droplets. Continental air masses, on the other hand, have a larger number of CCN. As a result of this difference, continental clouds have a high density of small droplets, while marine clouds have a much larger average drop size, but a lower number density of droplets. The ARM experiment must make measurements in all of these conditions to be successful.