Predicting weather and climate: Uncertainty, ensembles and probability Wendy S. Parker 1 Ohio University, Department of Philosophy, Ellis Hall 202, Athens, OH 45701, USA article info Keywords: Simulation Climate Weather Uncertainty Ensemble Prediction abstract Simulation-based weather and climate prediction now involves the use of methods that reflect a deep concern with uncertainty. These methods, known as ensemble prediction methods, produce multiple simulations for predictive periods of interest, using different initial conditions, parameter values and/or model structures. This paper provides a non-technical overview of current ensemble methods and considers how the results of studies employing these methods should be interpreted, paying special attention to probabilistic interpretations. A key conclusion is that, while complicated inductive arguments might be given for the trustworthiness of probabilistic weather forecasts obtained from ensemble studies, analogous arguments are out of reach in the case of long-term climate prediction. In light of this, the paper considers how predictive uncertainty should be conveyed to decision makers. & 2010 Elsevier Ltd. All rights reserved. When citing this paper, please use the full journal title Studies in History and Philosophy of Modern Physics 1. Introduction Computer simulation models are ubiquitous in the study of weather and climate. Their value as aids to understanding is widely acknowledged, but for prediction they are considered truly indispensable. Over the last half-century, substantial resources have been devoted to the development of these weather and climate models and to the expansion of observing networks that help to provide initial conditions for them, often with the explicit goal of more accurate prediction. Some of the fruit of this investment can be seen in the considerable increase in the skill of weather forecasts that has occurred in recent decades (e.g. Simmons & Hollingsworth, 2002). It is important to recognize, however, that simulation-based weather and climate prediction today differs from that undertaken 50 years ago in ways that extend far beyond having more comprehensive observing systems and higher-resolution models that represent more physical processes in greater detail. Most notably, weather and climate prediction today involves the use of methods that reflect a deep concern with uncertainty. These methods, known as ensemble prediction methods, produce multiple simulations for predictive periods of interest, using different initial conditions, parameter values and/or model structures. Ensemble methods are judged to be among the best ways forward when it comes to predicting weather and climate in the face of uncertainty. Yet interpreting results from studies employing ensem- ble methods is a complex matter and, in the context of climate prediction, is currently the subject of much discussion and debate (e.g. Collins, 2007; Stainforth, Allen, Tredger, & Smith, 2007). Particularly contentious is whether the results of ensemble prediction studies should be transformed into probabilistic forecasts that are offered as such to decision makers—as they choose policies for reducing greenhouse gas emissions, decide how high to build sea walls to protect coastal cities from strong storms, etc. The discussion that follows has two main goals. First, it aims to provide a non-technical introduction to today’s ensemble prediction methods in a way that gives a sense of the richness and complexity of current practice. As we will see, weather and climate prediction involves much more than simply ‘‘running a model’’. Second, the discussion will consider how results produced using the latest ensemble methods should be interpreted. A key conclusion will be that, while complicated inductive arguments might be given for the trustworthiness of probabilistic weather forecasts obtained from ensemble studies, analogous arguments are out of reach in the case of long-term climate prediction. In light of this, options for conveying predictive uncertainty to decision makers are presented. 2. Representational uncertainty in global weather and climate prediction A computer simulation model is a computer-implemented set of instructions for repeatedly solving a set of equations in order to Contents lists available at ScienceDirect journal homepage: www.elsevier.com/locate/shpsb Studies in History and Philosophy of Modern Physics 1355-2198/$ - see front matter & 2010 Elsevier Ltd. All rights reserved. doi:10.1016/j.shpsb.2010.07.006 E-mail address: parkerw@ohio.edu 1 Thanks to Lenny Smith for helpful discussion, to Gavin Schmidt for providing Fig. 1, and to Tim Palmer and Royal Society Publishing for permission to use Fig. 2. This material is based on work supported by the National Science Foundation under Grant no. 0824287. Studies in History and Philosophy of Modern Physics 41 (2010) 263–272