Associating a Level of Uncertainty with Agricultural Model Predictions

This paper describes a comprehensive approach for evaluating and communicating the uncertainty of predictions made by agroecosystem models. Recognizing that models capture ecosystem complexity only imperfectly, the authors compare simulations to observed data and propagate quantified uncertainties in inputs and parameters. They tailor the evaluation process to different model uses—prediction, decision support, and diagnosis—by identifying relevant reliability indicators. Ten case studies demonstrate how to adapt methodological tools to specific needs and constraints. The project’s results, including articles, a book, and training sessions, are made available to the agricultural modeling community to enhance model reliability and uptake.