# Earthquake hazard

An important domain of application of computational seismology is the problem of simulating the **ground motion**
following large earthquakes. Until today the estimation of shaking
hazard and the associated risk is based on relatively crude assumptions
using empirical relationships. Because large earthquakes happen
infrequently it is impossible to build shaking hazard maps for
particular **earthquake scenarios** from observations. Therefore the *deterministic*
approach to earthquake scenario simulations is rapidly gaining
momentum. Here, for the given local crustal seismic velocity model and
appropriate near surface structure the complete ground motion for a
whole region can be calculated.

The synthetic data can be converted into **shaking hazard maps**
by considering (at each point on the surface grid) the peak ground
velocity (acceleration) and other properties of the wavefield such as
shaking duration. While this is a powerful concept, there are still
many open questions some of which we detail briefly: (1) Ground motion
is strongly influenced by near-surface seismic velocity structure.
These velocities may have values as low as 300m/s. This poses an
enormous computational problem. How can we incorporate the **near surface structure**
in numerical modelling? Can we use variable grid sizes? If so, is it
possible to use space-dependent time stepping? (2) For very large
earthquakes (e.g. Mexico City 1985) regions even at great distances
(e.g. 400km) are at risk. For these cases is it useful to combine 2.5 D
approaches with local 3D grids in order to fully capture the possibly
more relevant basin effects? (3) In regions with strong topography
(e.g. Los Angeles basin, the Alps) topography may have a strong effect
on shaking hazard through interference effects. It is at present not
clear what the best technique is to incorporate strong topography even
though several approaches exist.