The tenor of Lee's discussion does not really justify a debate, particularly when it is based on 42 references, 37 of which come from his team. We suspect that some of the points raised by Lee are almost exclusively included as pretexts to present their extensive bibliography. However, we provide here some general comments addressed to end users of our equations and separately in the appendix we offer more detailed discussion on the nine points raised by Lee which will be of interest mainly to researchers in engineering seismology.
We must stress at the outset that our paper is concerned only with the attenuation of absolute acceleration spectral ordinates and peak ground acceleration and discussion regarding other strong-motion aspects is not particularly relevant.
The equations that we have presented enable reliable estimates of zero-period and spectral accelerations to be obtained in seismic hazard assessments in Europe on the basis of parameters that can be quantified for future earthquakes with relative ease. The desire to publish our results stems from our concern regarding the widespread use in Europe of standard spectral shapes to construct design spectra and the use of frequencydependent attenuation relations which are based either on unnecessarily limited datasets or else on data from outside the Eurasian region. The important point is that no attenuation equations for this whole region have been derived previously.
We have used all the European strong-motion data currently available to us without applying any exclusion criteria other than an unacceptable quality of digital data. The predictive relations that we have derived are robust and the scatter associated with our equations is not higher than that usually found in attenuation studies.
We did refer in our work to four other attenuation studies for parts of Europe (two for Italy, one for Greece and one for the Eastern Mediterranean) and it was perhaps an oversight on our part not to make reference to the work of Lee on strong-motion attenuation in Yugoslavia nor to the work of Caillot and Bard for Italy. We can find no significant variations among regions and doubt whether the discrepancies between our results and those of Lee are genuinely due to regional differences rather than other factors such as the methodology of analysis and the distributions of the datasets.
On other points we do not feel that in the present case a lengthy reply is particularly useful, but some brief comments are offered for consideration. Regarding the proposal to apply an instrument correction to all records using standard dynamic characteristics, Lee himself points out that the influence is likely to be very small in the period range we consider. On the issue of the period range, we have deliberately limited ourselves to a maximum period of 2•0 s, as have Boore et al. and Lee. We do not have great confidence in the data at longer periods and it seems that nor does Lee who states that the Yugoslavian data 'beyond ¹'2•0 s is not adequate for meaningful regression at this time'.
Lee challenges us to employ more complex characterization of the local geology at the recording sites through the use of two separate terms as he and his co-workers have done in several of their studies. This may well represent a desirable improvement to the model but at the present time sufficient data is simply not available and the choice is essentially to derive relatively straightforward attenuation equations for the engineer as we have done, or to adopt a more complex model whereby the usable dataset would become so small as to be unreliable for regression analysis. In such cases over-parameterization is not an improvement