This thesis examines the ground motion prediction component of earthquake
loss estimation (ELE) frameworks and is based on the assertion that
reducing the uncertainty in ground motion prediction will result in improved
accuracy of loss estimates. The objective is to obtain improved ground motion
predictions by identifying and quantifying the sources of uncertainty
in the predictions, with particular focus on the portion of the uncertainty
that can be reduced.
The work presented in this thesis starts with an examination of ground
motion measures commonly used in ELE and their relative utility. The
ground motion measure Arias Intensity is identified as well-suited to application
in a number of problems in earthquake engineering and this along
with the lack of a robust equation for its prediction, leads to the development
of a new predictive equation for Arias Intensity. Next, the prediction of
Arias Intensity at spatially separated locations is studied in order to develop
a model for the spatial correlation of Arias Intensity so that loss estimates
for spatially distributed portfolios may be obtained. Thirdly, the sources
of uncertainties in the predicted values of Arias Intensity are investigated
and the uncertainties are characterised and quantified in order to establish
whether or not they may be reduced. The impacts of these uncertainties
on the new predictive equation for Arias Intensity are also examined. The
final part of the thesis focusses on the use of GIS to display the information
described in the previous sections on ground motion prediction. Particular
attention is given to enhancing the display of uncertainties in ground motion
predictions.
This thesis demonstrates that the impacts of uncertainty on ground motion
predictions and therefore earthquake loss estimation are significant,
making this research of particular importance in this field.