Providing valid long-term projections of geologic systems for policy decisions; can we succeed? Should we try?
Providing valid long-term projections of geologic systems for policy decisions; can we succeed? Should we try? (in A paradox of power; voices of warning and reason in the geosciences, Charles W. Welby (editor) and Monica E. Gowan (editor))
Reviews in Engineering Geology (1998) 12: 177-184
Many environmental problems require modeling of long-term interactions of man-made and geologic systems. By using the term "model," we acknowledge that we will never know whether our descriptions of geologic features, events, and processes are unique and represent absolute reality. "Validation" of a long-term predictive model means that, on the basis of tests of the assumptions, inputs, outputs, and sensitivities, the model adequately reflects the recognized behavior of the portion of the system it intends to represent. Adequacy is driven by the needs of the application for which the model is developed. Most environmental applications are overprinted with political, scientific, and social requirements that add subjective influences to the definitions and interpretations of measures of adequacy. Therefore no single measure can be developed for the adequacy of a model. The public, on the other hand, may expect "validation" to describe an unobtainable absolute demonstration of "truth." Scientists assessing long-term risk use various mechanisms to establish the adequacy of their models such as: (1) expert judgment to assign appropriate ranges of parameters where data are sparse, controversial, or unobtainable; (2) conservatism in assigning parameter values and process descriptions, including ignoring some potentially mitigating processes; and (3) stochastic simulation to assess the effect of uncertainty in descriptions and the sensitivity of performance predictions to uncertainty, and to examine alternative scenarios and process models. Other measures are under-taken to demonstrate that the effort to ensure validity has been comprehensive, including: (1) documentation of the structure of models, including justification for assumptions and simplifications, as well as the examination of alternative conceptualizations for the system; and (2) review by the scientific community and those who have a stake in the decisions that these models support. Legal precedent recognizes that decisions with large consequences often demand judgments of the validity of conflicting descriptions of reality. The stochastic nature of natural phenomena and the value of models as guides to judgment (not final answers) are both acknowledged. The problem of compounding conservatisms, which leads to an inherently unreal and thus invalid, model, may also be understood. The scientific community can usefully project the range of future behavior of systems and must do so if well-reasoned choices are to be made about how humans should affect those systems. Unfortunately, the consequences (costs and risks) of abandoning or overburdening efforts to make long-term projections are rarely examined as critically as are the efforts to project interactions of man-made systems with the geologic environment.