Current Projects Read or Print all Projects

Analysis of Uncertainty, Structural Error, and Reachable Futures. There has always been uncertainty; and always will be. The advent of its formal analysis is a sign of a maturing subject. Two sea-changes of outlook over the past two decades, within the analysis of uncertainty in environmental stewardship, are striking. First, there has been a shift away from the presumption of a search for the ultimate truth of a singular model structure — just one (and only one) approximation of the way things happen. Second, there has been a shift away from the singularity of what the one (and only) decision-maker might want of the future, to the plurality of what the people want (and do not want). We acknowledge that considerations of the quantitative aspects of uncertainty attaching to the former (crudely stated as Uncertainty{Computational Model}) may be dwarfed by the qualitative aspects attaching to the latter (i.e., Uncertainty{Human Behavior}) and befuddled by its sheer unpredictability. Yet it would be unprofessional of us to put aside the analysis of Uncertainty{Computational Model} because of this. We move forward accordingly, thus, now expecting there to be structural change, especially over the increasing spans of our forecasting (and observational) horizons, and structural uncertainty/error, possibly expressed as a plurality of candidate model structures populated by multitudes of candidate parameterizations. Can we, then, detect structural change in the presence of so much confounding uncertainty? Can we characterize structural error/uncertainty and account for its consequences in any forecasts? Should we try to resolve such structural uncertainty, through model structure identification? Or is the path forward to be essentially pragmatic: accept the uncertain empirical record; accept the ambiguities of interpreting the past through a multiplicity of candidate model structures; generate ensembles of forecasts shot through with uncertainty; and act when the preferred course of action still emerges as "robustly preferred" above the fog of all the uncertainty? These questions, whose logic flows from the past and into the future, are worthy of sustained attention. But we know from adaptive community learning and the inverse approaches of Sustainability Science that this logic can be turned around: to start from the plurality of people's hopes and fears for the future (of their cherished piece of the environment), to assess their reachability (their plausibility/implausibility), and to identify the few, key, scientific uncertainties requiring the purchase of more science — because the coming of the more reachable futures, or more apocalyptic futures, hinges crucially upon them, all the gross uncertainty notwithstanding. Would purchase of the same key pieces of new science, in the here and now, emerge from evaluating the democratic plurality of what the people want, across the generations into the distant future? We have achieved proof-of-concept for the associated, core, computational development. It rejoices in the acronym of RIMME (or Random-search Inverse Methodology for Model Evaluation). It seems, remarkably, to be as relevant to the analysis of reachable futures as it is to the analysis of uncertainty and sensitivity in the VHOMs (very high order models) of predictive, multi-media exposure assessments, and to assuring the quality of those VHOMs as tools designed to fulfil such predictive tasks.