Using a Multi-Model Ensemble Approach to Constrain Uncertainties in Climate Change Impact Studies.

Many previous studies reported that different impact models often provide divergent or sometimes even contradictory results using the same input data. Researches also showed that “all models are not created equal” and they have different emphasis and assumptions, therefore with great spatial and temporal (climate-wise) variability in terms of model reliability. Multi-model ensemble, of which the main goal is to pool the results from multiple models to reduce uncertainty in model simulations, has been successfully applied mostly in atmospheric science, but received little attention in climate impact studies. In this study, we develop a strategy to identify spatial/temporal patterns of model reliability and a framework to synthesize the simulation results from multiple impact models by taking into account of model reliability patterns. The results of a case study on corn and sorghum yield simulations in the U. S. Great Plains will be presented.