Geoscience Reference
In-Depth Information
Euroseistesttoevaluatehowmuchdetailwasnecessaryinamodeltopredicttheseismic
ground response of this valley. They evaluated both 1D and 2D models at the centre of
the valley. Their conclusion was that even a simplified 2D model was better than a very
detailed 1D model to predict site response. These results are a warning that tells us that
we must be careful when deciding where to concentrate the efforts. Clearly, for the end
of understanding site response at the centre of the Euroseistest valley, a rough 2D model
is better than detailed 1Danalysis.
Two-dimensional models go one step further in that they allow variation of properties
or geometry along one horizontal dimension. Three decades ago, a great many papers
dealingwiththesimulationofgroundmotioninsuchirregularconfigurationswerebeing
published. Trifunac (1971) presented the analytical solution for the ground motion at the
surface of a semi-circular, 2D alluvial valley subjected to plane SH incident waves. This
solutionwasformanyyearsthebenchmarkagainstwhichtoshowthereliabilityofsolu-
tions computed using numerical methods. In fact, when we revise the papers presenting
resultsofnumericalmodellingofgroundmotionforirregularlocalconditionsduringthe
1980s, all those papers included some sort of validation of the method used, either com-
parison with Trifunac's solution or with that computed using other numerical method.
This is no longer the case. Papers that present results of numerical modelling these days
neednolongerneedtovalidatethemodellingmethod,astheyhavebecomeveryreliable.
Modelling methods suchasfinite elements, finitedifferences, or boundary methods have
been shown to be very reliable to compute the seismic response of complex configura-
tions,anddonotposeparticularproblems(althoughthehugeamountsofcomputerpower
necessary to make the computations are still an issue in some cases). For this reason, the
problemof modelling seismicresponsehasshiftedfromthemethod and the computer to
the building of the model; the properties to use, the geometry, and the reliability of all
those parameters. These days, the problem of defining the properties of the subsoil to be
usedinamodelislargerthanthatposedbythemodellingmethoditself.Asecondobsta-
cle is that there must be a way to evaluate the results of the modelling, when often there
are no independent measures. This can be exemplified with the case of Parkway basin,
modelledbyCh´avez-Garc´ıa(2003).Therewerenotenoughdatatobuildamodelforthis
basin. For this reason, the model used was very crude; only two different media were
considered and the interface between sediments and basement was interpolated starting
from the depth estimates obtained from observed values of dominant period measured at
thesurface.Surprisingly,theresultswereuseful.Inthiscase,validationofthemodelwas
made using previous estimates of site effects obtained from the analyses of earthquake
datafromatemporaryseismographnetwork(Figure3.5).Detailsoftheobservedtransfer
functions were well reproduced by the model, as were the preferred directions of propa-
gationofdiffractedsurfacewaves.Inaddition,anestimateoftheadditionalamplification
brought about by the 3D configuration of the basin (in addition to the amplification due
solely to impedance contrast) could be obtained. One of the conclusions of that paper
is that even a crude model can be useful. The next paragraphs present some comments
regarding the methods that have been used to obtain enough data to build a model from
which tocompute local response.
