Real world risks and extremes: correlation and quantification
Extreme events which occur rarely but have high impact are features of both natural and socio-technical systems. Examples include earthquakes, financial crises, hurricanes and large scale power failures. Governments, industry and scientists need to quantify extreme risks in order to manage the long term operation and cost of human systems. Modelling extreme events is intrinsically difficult because they are rare. This difficulty is exacerbated by the possibility of correlated or cascading extremes. Examples include repeated extreme floods, human-induced seismicity or the economic impact of the Fukushima disaster.
Considerable sophistication has been achieved in risk quantification in areas like finance, catastrophe modelling and climate science. A combination of approaches drawn from the mathematical, statistical and physical sciences are used but the quantification of correlated fluctuations remains thorny. From a physics perspective, correlated fluctuations become most interesting when they are non-Gaussian and long ranged but from a rigorous mathematical or statistical perspective, they are most tractable when they are Gaussian and short ranged. The real world is somewhere in between.
Within this problem space there are thus two axes: different methodological approaches versus the potential for coupling between different human and natural systems. The purpose of this meeting is to promote a high-level dialogue about correlated fluctuations across both these dimensions. It will offer academics, industry practitioners and policy-makers an opportunity to listen to, and engage with, leading experts whose expertise spans these worlds.