adequate in temporal and spatial coverage, in parameters measured, and in precision to permit meaningful validation or rejection of models. It is equally important that models be designed to permit confrontation with the real world through observations and that they be tested sufficiently and explored, including creating ensemble runs under differing conditions.
Over the past decade there has been remarkable progress in modeling, not only in simulating the principal individual subsystems but also in treating key linkages such as those between the ocean and atmosphere. This record of progress within the U.S. Global Change Research Program (USGCRP) makes it reasonable to expect that within the next 10 years of the USGCRP the scientific community will develop fully coupled dynamical (prognostic) models of the full Earth system (see Figure 10.11) that can be used on multidecadal timescales and at spatial scales relevant to important policy formulation and impact assessment. Such models exist in rudimentary form today. Future models will advance in completeness, sophistication, and proven predictive capability. The key will be to demonstrate some degree of prognostic skill in these future coupled models of the Earth system.
This development process will not be isolated from the needs of policy and decision making. Some of these Earth system models will be integrated into more encompassing models that link human and nonhuman processes or will be employed in various analytical or deliberative processes to inform decisions. Providing useful insights to inform decision making on global change will require dynamic representations of complex possible cause-effect-cause patterns linking human and nonhuman components of the Earth system. To develop and validate such models, observations of the Earth system must include data on human impacts from, and contributions and responses to, global change. At present, human influences generally are treated only through emission scenarios that provide external forcings to the Earth system. In future comprehensive models, human activities will interact with the dynamics of physical, chemical, and biological subsystems through a diverse set of contributing activities, impacts, feedbacks, and responses.
The focus of this chapter is on the path for realizing and evaluating a suite of such Earth system models. It should be recognized at the outset that the multi-decadal timescale places important constraints and demands on the character of such models. The most important constraint is that models must confront the ever-expanding (though still inadequate) set of time series data, both in situ and remote. The canonical example of the extraordinary value of time series information is the Keeling Record, the daily measured atmospheric concentration of carbon dioxide from Mauna Loa (see Figure 2.10 in Chapter 2).a The importance
It is worthwhile to note that obtaining this unique record was threatened more than once by budget cuts and shortsighted federal managers.