Abstract

There are a multitude of sources of uncertainty in the science of climate change, many of which are related to the extensive use of climate models to answer research questions. This chapter, which complements Morrison and Lawrence (chapter “Understanding Model-Based Uncertainty in Climate Science,” this volume), examines how various sources of uncertainty in climate models – structural, parameter, scenario, and initial condition – contribute to uncertainty in the ability to project climate impacts and changes to extremes, understand equilibrium climate sensitivity and transient climate response, and identify the causal contributions of a changing climate to disastrous weather events (attribution). The second component of this chapter moves beyond descriptions of the consequences of uncertainty to discuss how scientists have sought to decrease and understand model-based uncertainties. Practices examined include the use of model ensembles in various forms (multi-model, large single model, and perturbed parameter), benchmarking, out-of-sample testing, and machine learning. Also explored are the philosophical discussions related to these practices, commenting on interpretations of model pluralisms, the epistemic opacity of machine learning, and the epistemic role of models in informing decisions related to adaptation and resilience.