Unknowability in Climate Science: Chaos, Structure, and Society

Abstract

Unknowability in Climate Science:Chaos, Structure, and Society Gavin A. Schmidt (bio) CLIMATE SCIENCE AS A DISCIPLINE We are the Borg. You will be assimilated. —The Borg (Star Trek: Voyager 1999) The field of climate science is a relatively new discipline that has arisen from the understanding that the previously distinct fields of meteorology, oceanography, hydrology, studies of the cryosphere, atmospheric chemistry, and such are fundamentally linked and that changes in one or another involve interactions and feedbacks across all these domains. This has always been known at a basic level, but due to the complexity of the overall system, the number of scientists working across these domains was relatively small until recent decades. With new understandings of coupled atmosphere/ocean variability of phenomena like the El Niño–Southern Oscillation, and the growing appreciation of the magnitude of human-caused climate change that emerged in the 1980s, the concentration of scientists working on climate, per se, as opposed to the previously disparate domains, has increased enormously. The challenges within the newly emerged discipline are substantial. Events happening at the smallest spatial and fastest temporal scales—for instance, the nucleation of cloud-condensing nuclei at the micrometer/microsecond scales—are as important to climate as [End Page 133] the multi-millennial global shifts in the orbital parameters for Earth. Integrating understanding across such vast scales (some 14 orders of magnitude in time and space) is technically challenging and, because there is so much ground to cover, very difficult to synthesize into a single coherent theory. Additionally, the enormity of the observational data now available from weather stations, ocean buoys, and, importantly, remote sensing by satellites defies simple characterization. Even data from the past—whether old instrumental records or proxies for climate changes in the ice, trees, or ocean sediments—require complex processing before they can be correctly interpreted. Complexity is thus the background upon which all work in climate science is situated, and the increasing tractability of complex problems over time has both made this work necessary and driven it forward. THE NECESSITY FOR CLIMATE MODELS Si Dieu n'existait pas, il faudrait l'inventer. —Voltaire (1768) The advances in computational processing have allowed the interconnections to be quantified and processed. Physics-based, multi-process, multi-domain general circulation models (GCMs) have steadily become more advanced, more complex, and more complete over the last few decades (figure 1). These models are playing an ever-greater role in integrating observations, deepening understanding, and making forecasts (from weather timescales to multiple decades). Although it may sound a little clichéd, the substantial role that large-scale simulation plays in climate science now, compared to 30 or 40 years ago, can be usefully described as a new paradigm (Kuhn 1962). But what has driven this change? Certainly, it is related to the enormous improvements in computational capacity and accessibility over the last 50 years that have made once inconceivable calculations [End Page 134] routine. However, the biggest driver is the success of early efforts that demonstrated that not only were old questions answered in better ways, but also new questions could be posed that would have been impossible to address before. Climate models have enabled better understanding, improved attribution, and skillful prediction. More relevant for this paper, they have also clarified the limits to that knowledge. Click for larger view View full resolution Figure 1. History of climate modeling showing the progressive integration of previously separate efforts over the last 60 years (modified from CarbonBrief 2018). Climate science as a whole is much broader than the development and application of these models, clearly. But these models form the backdrop for all substantive prediction efforts in weather and climate. Their limitations are an excellent starting point for considering where the limits of knowledge in climate science lie and how much of what is beyond those limits is truly unknowable as opposed to simply, as yet, unknown. The construction, calibration, and application of these models are complex (Schmidt and Sherwood 2015; Schmidt et al. 2017), but a number of points are clear. First, no single model is (or will ever be) either complete or perfect. There will always be processes that need to be approximated...