Spatial Variation of Earthquake Ground Motion (SVEGM) is clearly indicated in data recorded at dense seismographic arrays

The main purpose of this paper is to study the influence of SVEGM on the seismic response of large embankment dams. To this

end, the Masjed Soleyman embankment dam, constructed in Iran is selected as a numerical example. The spatially varying ground

motion time histories are generated using spectral representation method. According to this methodology, the generated time

histories are compatible with prescribed response spectra and reflect the wave passage and loss of coherence effects. To

investigate the sensitivity of the dam responses to the degree of incoherency, three different coherency models are used to simulate

spatially variable seismic ground motions. Finally, the seismic response of the dam under multi-support excitation is analyzed

and compared to that due to uniform ground motion. Also, the Newmark's method is used to estimate seismic-induced permanent

displacements of the embankment dam. The analysis results reveal that the dam responses can be sensitive to the assumed spatial

variation of ground motion along its base. As a general trend, it is concluded that the use of multi-support excitation, which is

more realistic assumption, results in lower acceleration and displacement responses than those due to uniform excitation.

Among the different ways of in-situ soil investigation, cone penetration test data are selected to evaluate the spatial variability

of geomaterials and the scale of fluctuations is chosen to evaluate the correlation structure of CPT data. In this regard six case

studies in sandy materials from Australia, U.S.A. and Iraq are selected. Various techniques for the calculation of the scale of

fluctuation of geotechnical parameters are suggested in literature e.g. VXP, SAI, AMF, BLM and VRF without any preference or

privilege for any specific procedure. In order to isolate the stochastic portion of cone tip resistance, deterministic trend was first

removed by regression analysis. This study suggests that quadratic trend removal is more suitable for selected CPT data

soundings. The closeness of the estimated scale of fluctuation using different approaches is assessed too. Mean value of the scale

of fluctuation by five established methods ranges between 0.44 to 1.52 meter for six different cases and the coefficient of

variation for the scale of fluctuation calculated by these methods varies between 12 to 27 % showing that available established

methods produce almost compatible and comparable results.