Tamsin Edwards

Tamsin Edwards  (Geographical Sciences ) and co-applicant Jonty Rougier ( Maths) won funding from the Cabot Institute's annual Open Call 2012 / 2013 to put on a Black Box Workshop.  Here they outline what this is and why such a workshop is important:

What are the best methods for assessing our uncertainty about the environment? The earth system is so complex, and the sources of information (observations and physical modelling) about it so diverse in quality and coverage, that there is often no consensus on the most suitable approaches to statistical inference and uncertainty assessment when it comes to understanding the past and predicting the future of the planet.

We will host a “Black Box Workshop” in which researchers from many diverse disciplines can share their approaches within three research topics:

(a) statistical modelling of complex datasets,

(b) statistical emulation of physical models, and

(c) inference from physical model ensembles.

The name of the meeting reflects the needs and advantages of such wide-ranging expertise: we will ask attendees to draw back from their usual research topic-focused perspective and consider their dataset or physical model in the most abstract terms possible.

Given that each discipline uses their own lexicon, we will distribute a standardised set of terms and notations before the meeting, and ask speakers to describe their work in these terms – transforming each research problem into an opaque “black box” – so as to allow colleagues from very different disciplines to understand their problems and provide useful critiques.

This will be a two day meeting, with generous time allowed for small and large group discussions, including a meeting dinner. To maintain methodological focus, priority will be given to Bayesian statistical approaches. A draft programme is:

1. Statistical Modelling of Complex Datasets

• Spatio-temporal modelling in palaeoclimate, palaeo sea level and present day Antarctica
• Multilevel modelling in biological, ecological and veterinary sciences
• Cross-cutting topic: Priors

2. Statistical Modelling of Complex Simulators

• Emulation of climate models
• Emulation of galaxy formation
• Cross-cutting topic: Approaches to structural uncertainty

3. Inference from Physical Model Ensembles

• Past and future climate
• Numerical Weather Prediction
• Planet formation
• Cross-cutting topic: Decision-making under uncertainty: lessons from climate, flooding and engineering

The workshop will be held during the first half of 2013. To express an interest, please email tamsin.edwards@bristol.ac.uk