Seismic compression is defined as the accrual of contractive volumetric strains in unsaturated soil during strong shaking from earthquakes. While ground deformations from seismic compression have been reported in the literature , it contains few case histories in which the amount of ground deformation was known accurately from pre– and post–earthquake surveys. In this report, two such case histories are documented in detail and analyzed. Both case studies involve deep canyon fills in Santa Clarita, California, an area strongly shaken by the Northridge earthquake (peak accelerations on rock ≈ 0.3–0.7 g). The performance of the fills was quite different. In one case (denoted S ite A) ground settlements up to ∼18 cm occurred, which damaged a structure, while in the other case (Site B) settlements were < ∼6 cm.
One important thrust of the present work involved cyclic simple shear laboratory testing of four reconstituted soil samples from the two subject sites. These samples all have fines contents near 50% (such that the fines fraction controls the soil behavior), but have varying levels of fines plasticity. Each specimen was compacted to a range of formation dry densities and degrees of saturation. The results significantly extend th e seismic compression literature, which has consisted primarily of laboratory testing of clean uniform sands. The test results show that seismic compression susceptibility increases with decreasing density and increasing shear strain amplitude. Saturation is found to be important for soils with plastic fines but relatively unimportant for soils with nonplastic fines. Comparisons of test results for soils with and without fines suggest that for many cases, fines decrease seismic compression potential relative to clean sands. For soils with fines, it appears that seismic compression is most pronounced when the fines are nonplastic, or when the fines are plastic and the soil has a clod structure. We observe clod structures in plastic soils compacted dry of the line of optimums or at low densities, but not in nonplastic soils.
The objectives of analyses performed for the two sites were (1) to investigate the degree to which seismic compression can explain the observed ground displacements and (2) to evaluate the sensitivity of calculated settlements to variability in input parameters as well as the dispersion of calculated settlements given the overall parametric variability. The analysis procedure that is used de-couples the calculation of shear strain from that of volumetric strain. The shear strain calculations involved one- and two-dimensional ground response analyses employing site-specific dynamic soil properties and a suite of input motions appropriate for the respective sites. Volumetric strains are evaluated from the shear strains using material-specific models derived from the simple shear laboratory test results.
Parametric variability in all significant model parameters is estimated, and the analyses are repeated according to a logic tree approach in which a weight is assigned to each possible realization of the model parameters. The analyses results provide probabilistic distributions of shear strain, volumetric strain, and settlement, the last of which can be compared to observed field settlements. Calculated ground settlements at Site B match observations between the 30th and 70 th percentile levels. At Site A, the analyses successfully predict the shape of the settlement profile along a section, but the weighted averag e predictions are biased slightly low (match occurs at the 50th to 70th percentile level). We speculate th at the underprediction likely results from imperfect knowledge of site stratigraphy and/or underestimation of volumetric strains from the laboratory tests as a result of the non-reproducibility of the field soil’s clod structure.
Sensitivity studies reveal that the mean value of calculated settlements is highly sensitive to shear strain amplitude and compaction condition, while the standard deviation is mostly strongly influenced by variability in the shear strains. The median and standard deviation of shear strains, in turn, are strongly influenced by the site shear wave velocity profile, ground motion characteristics, and the method of site response analysis (i.e., 1-D versus 2-D). The various sources of parametric variability combine to form a coefficient of variation of about 0.5 to 1.0, being closer the low end of the range if 2-D analyses are performed (∼0.5–0.7) and the upper end of the range if 1-D analyses are performed (∼0.8–1.0).
Full List of PEER Reports: click here.
