Partner Information

RMS (Risk Management Solutions) - Risk Consultancy

How to price risk when events occur very rarely?
- Create virtual history of 50,000 years with hope to capture frequency of event of concern
- Solution is measuring ‘Exposure’ -> everything exposed to impacts from hazard
Risk assessment scope
- Started focussed on insurance sector – but now used and applied by cities and governments also (e.g. Sendai 2015 process)
- How to define progress in reducing impact of disasters, Sendai set targets to achieve by 2030 with progress measured by fatalities in the 10 years prior – empirical data for this is inappropriate with such large infrequent events (how to model progress in an area that hasn’t experienced disaster in some time)
- To monitor progress – parameterise vulnerability, get timely exposure data (same so-called vintage) in order to assess evolution rather than exposure data compiled across longer time period
How to capture exposure data with low latency?
- Use remote sensed data with exact dates to infer params of greatest interest - e.g. convert satellite image data to disaster risk (in terms of casualties or infrastructure) - think of indirect observations such as tarpaulins on rooves, junk in backyards etc...
- In future countries may need to demonstrate that they are making progress in protection from disasters

Main contact - Hélène Galy, [email protected]
Others - Thomson, Marie-Kristina [email protected]; Foote, Matthew [email protected]; Simon Solvsten, [email protected]

Matt Foote – Head of Climate risk analysis
- Earth Obs has never been at the centre for management and response & consistency over the whole globe as needs differ is difficult
- There is need to respond and predict how things will develop when events do happen
- Find ways to harness AI for inference from aerial data (ESA Copernicus, NASA SERVIR)
Tina Thomson – Re-insurance broker arm (Willis Re)
- Detail of input data available for models is difficult – capture of information is increasing but there are barriers: GDPR, inequalities in coverage between countries, insurance penetration etc…
- EO is fast moving with lots of potential, how to deploy these datasets and make them useful? Also considering licensing issues, prohibitive costs etc…
Hélène discussion points
- Calibration of models requires detailed ground analysis post-disasters
- Multi-hazard assessment – they do need to be kept separate/siloed but also joined as some can influence others (wind more vulnerable to flooding etc…)