Humans are altering climate at a pace that unprecedented in Earth’s geologic history. We cannot predict unprecedented events using simple statistical models. Instead, we must rely highly complex scientific models. What are the limitations of these models? This becomes a pressing question now that we are using these models to assess future climate risk and make major policy decisions.
In this talk, we address this question by tracing the history and philosophical underpinnings of climate prediction. The evolution of modern climate prediction is intertwined with advances in digital computing. Models have grown complexity over time as computers have become more powerful. However, in the most recent IPCC assessment, some of the most complex models have been downweighted as they were deemed as being too sensitive to carbon dioxide concentrations. This suggest that increased model complexity may be yielding diminishing, or even negative, returns. Just as weather models are subject to a predictability limit, climate models may be facing a “reducibility limit”.
Host: Lorenzo Polvani, Maurice Ewing and J. Lamar Worzel Professor of Geophysics, Ocean and Climate Physics.
The Earth Science Colloquium Series, sponsored by Lamont-Doherty Earth Observatory and Columbia University Department of Earth and Environmental Sciences (DEES), provides a lively forum for discussing a wide variety of topics within the Earth sciences and related fields. Colloquia are attended by the full range of scientific and technical staff at LDEO. Colloquium attendance is required of all pre-orals DEES graduate students. The Colloquium Series supports the Lamont Seminar Diversity Initiative.