Most data from Les Saintes (21/11/2004, Mw6.3) earthquake and aftershocks.
Aim to develop model for use in predicting macroseismic intensities shortly after an earthquake. Hence
model should be applicable for wide range of magnitudes and distances and, ideally, independent of tectonic
context and depth. Hence use simple functional form.
Many epicentres (95) offshore.
Data from 14 stations on both rock and soil.
Note that due to lack of quadratic magnitude term PGAs could be underestimated for large rhyp.
Because magnitudes follow power law and there are more records from short distances, apply weights
that are proportion to magnitude and power of rhyp to give more weight to larger earthquakes and longer
Note that high σ may be due to lack of site term, wide range of magnitudes and distances and too simple
a functional form.
Examine residuals w.r.t. magnitude and rhyp. Find no trends for entire magnitude range but significant
underestimation in PGA for R < 15km.
Compare observed and predicted PGAs for 6 earthquakes, including some not used to derive model, and
compute average of the residuals. Find slight evidence for underestimation. Find PGA for one soil station
underestimated by factor 10.