Home > CATS > Research Grants > GLIMPSE


Gauging Limitations of Imperfect Model Prediction with Sculpted Ensembles

Novel, potentially deployable approaches to operational adaptation of infrastructure and transport for the future climate would prove of great value, given the very limited insight current climate model simulations are able to provide at the high spatial/temporal resolutions previously considered by design engineers and regulators. GLIMPSE will explore a game-changing approach to extreme event in the unknown climate of the future.

A moderate-risk/high-reward proposal, it will develop mathematical methods allowing fundamental changes to the way we design, regulate and operate weather sensitive infrastructure. It will be the first component of a larger research program (Just Enough Decisive Information, JEDI) between the LSE, the University of Chicago and stakeholders worldwide. 

Effective design of large-scale operational infrastructure and safety-critical systems with a lifetime of many decades is challenged by our inability to produce decision-relevant probability distributions of the weather expected under future climate. Nuclear power plants, for example, must be designed for the one in ten-thousand year event in, say, 2050. It is not generally held that direct simulation using current models or general methods from extreme value statistics could estimate the nature of events at this threshold with high fidelity. In cases where such estimates might become available, the regulatory consideration of “worst plausible case” scenarios can require design based on low mitigation, high impact scenarios; scenarios which, hopefully, will not be realized. This “safety margin” is likely (in fact expected) to result in extraordinarily expensive over-engineering.

An alternative to deploying designs to operate through such extreme extremes is to design much less expensive infrastructure able to survive in a “shut-down or avoid” state given early warning. GLIMPSE will develop algorithms for very early warning of low probability events, to enable a new approach to engineering design and operation. It is the first step towards our wider Just Enough Decisive Information (JEDI) scheme of climate adaptation.  Under GLIMPSE the postdoctoral researcher will:

(a) explore four or five potential applications in the energy, transport, coastal infra-structure, coastal safety (Royal National Lifeboat Institution) and shipping sectors with existing CATS industrial partners. The aim is to determine the lead time and likelihood levels required for each sector, regulatory restrictions, and openness to a JEDI approach.

(b) improve early warning by combining ensemble members in the medium range from multiple launch times (working with ECMWF and forecast archives from the US models already in-house at LSE)

(c) develop the novel  “sculpting” of existing simulations to generate initial conditions indicative of possible high-impact events, noting that “high-impact” events need not require meteorologically severe weather. Proof of concept will be achieved in moderate dimensional nonlinear dynamical systems (Lorenz 95, Lorenz 2004, Quasi-geostrophic models…) including consideration of structural model error. Tests with operational models are beyond the scope of this proposal, but discussion on this will be conducted with ECMWF and NCAR.

This is clearly an ambitious proposal; we note that the reward would be significant and that any progress would prove of great value as at present the traditional approach both requires thresholds we cannot estimate realistically and all but insuring over-design.