9602201 Green This study will investigate a method for characterizing uncertainty in coupled mechanistic models. In order to study the possible impacts of global climate change on a living system, it seems intuitively obvious that investigators should pursue models which are well grounded in the science underlying the system. This is because by exploring the possible effects of climate change, one is necessarily extrapolating beyond the range of the data. Hence there is little reason to believe predictions from empirical, data-interpolation models. An unfortunate consequence of relying on models developed in accord with scientific understanding of the system is that it is often problematic to place any sort of confidence bounds on the estimates, or provide measures of the uncertainty of the estimates. Since the estimates are almost surely biased, this is a serious problem. This project will address this problem and demonstrate a solution by means of a case study, in which two mechanistic models will be linked, and estimates of the uncertainty of the outputs will be provided. Most current work on global change involves the use of coupled mechanistic models. Raftery et al. (1995) have presented a method of characterizing uncertainty for mechanistic simulation models. This project will extend their method to coupled models. This will be accomplished via a case study involving MAESTRO, a carbon flux model, and Pipestem, a carbon balance forest growth model. The project team will specifically study the growth of loblolly pine stands under several scenarios for atmospheric carbon dioxide and weather. The result will be estimates of loblolly pine stand development under the assumed scenarios complete with statistically defensible measures of the uncertainty of the estimates. The latter are unavailable with current methods. This approach to characterizing uncertainty in coupled mechanistic models should be readily generalizable to other integrated assessments of potential global change.
