Statistical methods for understanding and reducing uncertainty in process-based models of the Earth system

This meeting on statistical methods was fully booked (90 odd delegates) and justifiably so. Lots of great speakers who did alternating theoretical and practical talks with plenty of time for questions.

Lots of useful tools/datasets were suggested, including: the uncertweb elicitation tool from Lucy Bastin (project leader Dan pointed to other software they developed, which may or may not be useful, some of which he would use again), the Match elicitation tool (online); the Shelf elicitation tool (based on R); several people advocated MUCM toolkit for emulating, the UKCIP09 UK climate dataset; the lhs R add-on for latin hypercube sampling; lots of people mentioned DiceKriging, also used for emulation via R; Lindsay Lee talked about fast99, a tool for calculating the main effect and the total effect and strongly encouraged reading literature by Saltelli; and Doug McNeall used Paper by 53 to make his cartoons/sketches and they look awesome.

Peter Challinor gave a great talk which finally got me to understand latin hypercubes and why you’d want to use them.

While I was down at the Met Office, Andy Wiltshire told me about the JASMIN / JULES project, which sounds like it’s going to be brilliant.

A very useful two half days.