Share

In this webinar organized by UCL EPICentre, London, GEM’s Vitor Silva discussed the current challenges in existing risk models, as well as how big data and machine learning can be incorporated in the risk assessment process to improve the current practice.

With close to a hundred online participants, Vitor presented and discussed the current practice of earthquake hazard assessment, exposure modelling, and data collection using various methods – from mobile apps to satellite imagery. He also touched on validating risk models and scenario modelling.

In his presentation, Vitor emphasized how cities around the world have expanded at a rapid rate within just 30 years or less explaining that there are several factors that can influence future exposure modelling such as proximity to major roads and to other cities, population, land value, and many other factors.

Presenting a simulation of San Jose, Costa Rica urban growth, Vitor pointed out that “Rapid urbanization and increasing population are sources of epistemic and aleatory uncertainty that can be introduced in the model to get a range of trajectories of how the risk is expected to vary in the future.”

He further added that the real importance of being able to incorporate uncertainties in the model “is not so much predicting the risk but introducing risk mitigation measures.”

“What if we start retrofitting 10 000 buildings per year in the country? What if we improve the design regulation? What if we start discouraging certain building classes? What if we start making sure that people do not build on soft soils or close to active faults?,” Vitor asked to emphasize his point on risk mitigation measures against risk prediction.

All of this information, according to Vitor, can be introduced in the model to see different trajectories of risk that can help a modeler pinpoint which one is going to lower the risk to an acceptable value.

The Earthquake Risk Assessment: Current Challenges & Future Trends webinar was organized by the University College London (UCL) EPICentre as part of its Online Seminars series. For more information, visit https://www.ucl.ac.uk/epicentre.