Adrian Currie (University of Calgary)

Title: Paleo-Climate & Future-Climate

Understanding the climate’s deep past is essential for understanding its future, however the nature of the paleoclimatological record is problematic. My aim in this paper is to get a conceptual handle on what paleoclimatology can offer climate projections, and what challenges this faces. Climate simulations are validated if it is shown that they are adequate for some purpose, for instance, predicting global mean temperature to within a certain margin, given different carbon emission scenarios, for the second half of the 21 s t Century. The paleoclimatological record potentially aids in two ways. First, analogues of future climates: if it is reasonable to believe that some features of future climate matches features of past climate, then the record can be drawn upon both to confirm overall simulation performance and that of particular components. Second, successfully predicting aspects of the deep past could support what I call ‘realist’ hypotheses about climate simulations, that is, hypotheses that some structure in a climate simulation captures the dynamics of some aspect of the climate system. Such hypotheses are important because they potentially licence extrapolating climatological simulations into situations which potentially radically depart from those we have data for. However, the paleoclimatological record is problematic. Two worries concern the nature of the data. First, historical data is patchy and requires reconstruction, asking heavy lifting from background theory in order to provide meaningful characterization. Thus, our picture of the past is could be highly ‘unstable’: highly sensitive to background theory. Second, the paleoclimatological record has enormous amounts of data, and large data-sets come with their own problems. Specifically, as data increases, splitting signal from noise becomes increasingly difficult. Large data sets can be so awash in patterns that justifying why one pattern is considered empirically relevant as opposed to another is contentious. Another worry concerns the nature of the climate itself. If Earth’s climate displays historicity, that is, if it is highly path dependent and sensitive to initial conditions, then its past may be no guide to its future. I discuss how these concerns relate to the two uses of the paleoclimatological record identified and argue that (1) reconstruction and signal/noise splitting can be unproblematic if there is ‘vertical independence’ between the theory drawn on to reconstruct and the theories behind climate simulations; (2) in cases where independence fails, we can develop criteria which view loss of independence as trading off other epistemic benefits: this allows us to draw on our current theories to both stabilize reconstruction and identify empirically relevant patterns; (3) the thesis that Earth’s climate exhibits historicity is both a testable empirical hypothesis and one which may apply to some aspects of the climate more than others: it could be that some properties are relatively robust, while others are highly sensitive to initial conditions. This grounds a future research project which aims to identify (and work out how to identify) which aspects of climate systems are more or less fragile.