Philosophical attention on climate simulation has focused primarily on empirical methods aimed at the 20^th century, entirely neglecting the historical study of Earth’s climate by paleoclimatologists. This dissertation integrates historical climatology into the broader philosophical context, articulating its import for understanding future climate.

In Chapter 1, I present a taxonomy of models used by paleoclimatologists, arguing that simulation models are an amalgamation of theoretical models and data models designed to emulate the behavior of some system over time. A product of theoretical and data driven considerations, I argue that climate simulation models should not be identified with theory.

In Chapter 2, I argue that neither should climate simulation models be identified with scientific experiment. After analyzing the epistemically significant features of experiment, I argue that experiment and simulation are distinct as a result of exhibiting different justificatory focuses. I defend this account from an apparent counterexample: numerical experiment.

In Chapter 3, I argue that climate simulations are inferential tools. I consider Otávio Bueno’s (2014) inferential account of computer simulation, arguing that revisions must be accepted to ensure externally valid climate simulation. I argue that climate simulation models allow for two kinds of inference: dynamic inference and variable inference.

Chapter 4 surveys a number of empirical methods required for externally valid paleoclimate simulation. I conclude with a discussion of what credentials a simulation earns through successful performance, and when we should expect credentials to carry from contemporary contexts to the past.

Chapter 5 considers the three predominant views of future-directed climate simulation assessment. I argue that each of the views suffers epistemic shortcomings. I introduce the threshold problem, an important shortcoming for the adequacy-for-purpose view. I conclude with an argument that the adequacy-for-purpose view, however, can apply to paleoclimate simulation modeling.

Chapter 6 tackles the threshold problem discussed in chapter 5. I argue that the threshold problem can be dissolved with the right paleoclimate partial analogues. I articulate three homomorphic conditions required for a past climate system to serve as an analogue for a future climate system. Paleoclimate simulations thus play a crucial role in supporting future-directed climate simulations.