The Atmospheric General Circulation Model (AGCM) is derived from weather forecast models, and differs not only in the breadth of physics that is included, but also in resolution, and in the length of calculations. The climate models must be lower resolution so that we can afford to make a computation of months, years, or decades, instead of the 24 or 48 hours of a weather forecast. The model that we have been using is the Genesis AGCM , which is part of a larger Earth Systems Model developed at the National Center for Atmospheric Research Climate and Global Dynamics Section

There are many ways to use the AGCM in the study of atmospheric dynamics. It can be run as just a part of the entire earth system model, which includes a dynamic ocean and dynamic biosphere, so that the entire system reacts to changes in solar forcing (or some other external variable) in the same manner that our real earth system does. This may be a good method to investigate some oscillation of the system such as El Nino, or climate changes on the scale of ice ages, but it is difficult to use actual historical data or look at processes in this way. A more specific use of the AGCM is to specify a priori the field of some environmental variable (e.g. sea surface temperature) and see what happens to the atmospheric portion of the system. We have done that using sea surface temperatures (SST) observed over the decade from 1979 to 1988 using the data set generated for the Atmospheric Model Intercomparison Project (AMIP). In this manner we can see how much of the variability in the climate model results is due to ocean surface forcing by contrasting a decade of simulation forced by the AMIP observed sst's with a Control run forced by climatalogical sst's that are the same year after year. The results show that the winter precipitation, for example, has considerable variability in the control run (lower panel), but much more variability when the actual sst's are used (center panel).

Precipitation, as we know from everyday experience, is often a hit-or-miss affair, being transient and highly localized; this is well represented in AGCM simulations. This simulation, using climatological SST's at a higher resolution than those above, shows the precipitation every 5 days for 4 months from January through April (24 pentads of data), and you can see that there is little continuity to the rainfall patterns. To get a smooth-looking climatological pattern, an average must be made over many weeks. Over land, the earth itself can act as the smoother, with the soil soaking up moisture from rain and returning it to the atmosphere at a later time, mainly as transpired moisture from plants. We can see the effect of the earth's surface smoothing out the variations of rainfall by looking at the soil moisture variable for the same 24 time periods, and we see that it is much smoother than the rainfall given above. The line of 50% snowcover is also shown, to indicate the area where the soil moisture is locked up as ice. The surface interaction acts as a damper on the atmospheric motion, much in the same way that the ocean acts to store heat.

Most of the soil moisture is available to the atmosphere by being taken up in the root zone of plants, and this process is simulated in the AGCM. We can see the northwards progression of spring in the transpiration for the same 24 time periods as before. The vegetation type also enters into the calculation, and the one shown here is unrealistic in some respects, but the effect is evident.

Future experiments will examine more closely the role of anomalous SST and soil moisture distributions in forcing year-to-year variations in seasonal temperature and rainfall.