Library: The University of Georgia, Athens, USA

Simulation modeling is arguably the most versatile scientific tool for predicting the future environment. However, the reliability of model-based predictions is limited to the behavior domain defined by the historical data employed for conceptualizing and calibrating the model. Future changes in external inputs and internal structure tend to produce system behavior significantly different from prior predictions. To abate this seeming lack of credibility, it is now customary to qualify model predictions with uncertainty estimates. This dissertation explores the complementary approach of back-ca...