Jim Smith : Reifying Probabilistic Digital Twins for Flooding Risk Assessments using Expert Judgement.

In the light of the more frequent extreme weather events recent climate change is inducing different utilities need to develop strategic plans to systematically strengthen the defences of their assets. The challenge here is that such incidents do not only affect the assets of individual companies but also simultaneously the assets of connected utilities. Under CreDo, a team of researchers, co-ordinated through Cambridge,  developed a prototype digital twin. This was designed to help co-ordinate understanding about the mutual risks companies might face in the light of different types of extreme flooding events that threatened a chosen geographical region of the UK.

We were given unique access to experts from three different asset owners in the region as well as flooding specialists. From these we could elicit not only the vulnerabilities of each individual asset within the network but also their connectivity. The interdependence across the network of assets would be represented by a digital twin over the geographical region. Here the nodes of the graph would represent the location of the assets and the edges their interdependencies. A probabilistic model of a given flooding incident would then induce conditions that might lead directly to various assets within the network failing. The graph could display how these failures might cascade through the whole network.

Demonstrators of such a system were presented in COP 26 in November 2021. However, to be part of a suite of decision support tools our probabilistic model would obviously need to be reified so that the cascade of failures was calibrated to expert judgements. This talk will describe how Kevin Wilson, Chris Dent and I are developing technologies that will enable us to build realistic probabilistic digital twins to be used for strategic flood defence planning by individual asset owners.