NatCat and Earthquake Modelling Workshop

Date: November 24, 2017, 13-17 pm
Organizer: Mario V. Wüthrich, RiskLab, ETH Zurich
Place: ETH Zurich
Room: There is a limited attendance.



Cat Model Developments at Swiss Re
Swiss Re has been developing its own in-house cat modeling platform over more than 20 years. Depending on the scenario (peril & market), the scenario's accumulation potential and typical business written in the market a more sophisticated or simpler model and kernel are used. We will give some insights to this modelling solution, what kind of models run in that platform and how this helps allocating capital and steering business.

Some Recent Advances in CAT Risk Engineering for Earthquake, Wind, and Flood
Probabilistic catastrophe (CAT) loss models are becoming increasingly popular tools for estimating potential loss due to natural hazards. Such models incorporate detailed databases and scientific understanding of the highly complex physical phenomena of natural hazards and engineering expertise about how infrastructure, buildings, and their contents respond to those hazards. This talk will present some recent advances in CAT modelling for earthquake, wind, and flood hazard, namely: 1) the use of physics-based simulated ground motion for seismic risk modelling; 2) the development of a real-time CAT modelling framework for designing engineering applications of earthquake early warning; 3) the development of tools for modelling risk to offshore wind energy assets. The talk will finally introduce and discuss some perspectives in CAT modelling for cascading and multiple hazards at different spatial scales (from a single asset to portfolio of buildings), with special focus on developing countries.

The Role of Spatial Correlation in Seismic Risk Assessment
Ground Motion Prediction Equations (GMPEs) are empirical models used in Probabilistic Seismic Hazard Assessment (PSHA) to estimate the ground motion intensity measures (IMs) and their corresponding uncertainties. A review of current practice of GMPE development will be presented and the limitations on current estimation methods are discussed. Improved GMPE estimation, that correctly accounts for spatial correlation of earthquake ground motion IMs will be explained. The newly developed GMPEs can be used to calculate the shaking intensities at various locations for several earthquake scenarios including both historical events as well as forecasts in space. This is particularly beneficial when assessing earthquake risk, for distributed system (e.g. lifeline network, portfolio buildings and transport).

Advanced Estimation Methods for Ground Motion Models with Spatial Correlation
Methods which separate calibrations of spatial correlation model and ground motion prediction function often suffer from statistical deficiencies such as loss of efficiency. This talk will explain how to adapt consistent joint estimation based on the method of Scoring and Expectation Maximisation (EM) algorithm.