GC53B-1211
Articulating and Locating Uncertainty for the Users of Hydrological Models: some considerations from climate informatics
Friday, 18 December 2015
Poster Hall (Moscone South)
Kimberly C Brumble, Indiana University Bloomington, History and Philosophy of Science, Bloomington, IN, United States
Abstract:
Climate models and constructions increasingly involve handling massive data sets which incorporate multiple and compounding sources and kinds of uncertainty. Kinds of uncertainty include measurement uncertainty involved in homogenizing data and handling data-scarce environments, as well as structural uncertainty arising from statistical activities like assimilating data and adopting data-handling conventions. Furthermore model uncertainty arises when doing intercomparisions of diverse models into ensembles. All of the above are examples of sources of typed uncertainty in model-building and interpretation. These different uncertainties also compound, “cascading” into one another. This has led climate modelers and those in climate informatics to explore methods for quantifying uncertainty in order to give the users of model outputs and conclusions explicit expression of the uncertainty involved. However, uncertainty quantification methods each have virtues and limitations in accounting for uncertainty and few explicitly locate the sources and kinds of uncertainty involved in accessible ways. Articulating and distinguishing these uncertainties accessibly is vital for policy users of models because applications of model outputs may depend heavily on particular limited scopes of possible scenarios or applications. The users of integrated climate impact and hydrological models in particular need uncertainty which is described and localized in the modeling process in order to interpret and utilize model projections. Methods for locating and articulating uncertainty in the modeling process are discussed and evaluated, and some suggestions for future projects are explored.