ICAD Presentation - Ghosts in the Machine: climate modelling, uncertainty and user expectations.
National Centre for Atmospheric Sciences, ES4 Summer School at Edinburgh University.
18th April 2013.
The traditional approach to climate modelling has been to build the most complicated model possible fitted within the largest computer available, run it, and wait-to-see what happens. Single ‘best-guess’ outputs, of this kind, often fail to fully capture the role played by uncertainty, however. Heeding calls from within, and beyond, the climate modelling community, the drive to quantify uncertainty through multi-model ensembles and probabilistic projections has gathered pace. Better technical understandings and attempts to reduce uncertainty will, it’s assumed, lead to better decision-making. But is this always the case? Taking the example of the UK’s latest climate projections, UKCP09, this talk explores how the push for probabilistic projections emerged from a political preference for defensibility yet, in so doing, encountered numerous modelling challenges in the form of computing resources, methodological novelty and competing user needs. Making the projections work involved a series of explicit, and sometimes implicit, assumptions that come back to affect the way the tool is used – ghosts in the machine. It is important, therefore to ask: does quantifying uncertainty produce better climate science? Does the visualisation of uncertainty help users to make better decisions? And how does the issue of uncertainty change the relationship between the users and producers of climate science?
This presentation will be given by Dr James Porter and a copy of the slides is available.