UCLA

Levee systems along Kushiro and Tokachi Rivers in Hokkaido, Japan, rest on significant deposits of peat and organic soils in downstream regions. Both levee systems were subjected to strong shaking during the 1993 M 7.6 Kushiro-oki and 2003 M 8.2 Tokachi-oki earthquakes. Local levee staff with the Japan Ministry of Land, Infrastructure, and Tourism (MLIT) performed thorough inspections of the full length of these levee systems after both events, which documented the location and severity of damages. This record of field performance presents a valuable dataset for investigating the factors that given rise to different levels of seismic performance. To my knowledge, this is the only such data set world-wide of levee performance when founded on peaty organic soils and subjected to strong earthquake shaking.

A crucial requirement for an investigation of the seismic performance of these levee systems is to understand the levels of seismic shaking they experienced. This is accomplished using a Kriging routine that operates on event-adjusted residuals between observed ground motions from local recording stations and ground motion models. Two ground motion models are considered, with some accommodations made to the path and site components of the ground motion models.

The site response component of the ground motion models is not able to capture the effect of the local geologies in the downstream regions, where the soft peat and organic soils are well outside of the range present in global site databases. Accordingly, a regional site amplification model is developed using recordings from the downstream portion of the Tokachi River system in combination with nonlinear ground response analyses (GRA) with representative profiles. The profiles are based on information from the literature, local field offices, and a subsurface exploration program conducted as part of this research using the spectral analysis of surface waves (SASW) method and ambient noise measurements at 21 sites. The fundamental site period is estimated from the horizontal to vertical spectra ratio (HVSR) of the ambient noise and used as a predictive parameter for the empirical site response model. Dispersion curves are inverted to obtain shear wave velocity profiles for GRA and estimates of VS30 along the levees. The empirical amplification is higher and exhibits less nonlinearity than the amplification model derived from simulations. The regional model is used in place of the ergodic site terms in the ground motion models for predicting PGA at the levee segments with similar foundation conditions.

Seismic levee fragility is expressed as the probability of exceeding a damage level given the peak ground acceleration. The levee system is discretised into 50 m segments, each of which is assigned damage levels based on crack depth, crack width and subsidence from the MLIT reconnaissance. Around a third of the 9,768 levee segments have peat within the foundations. Within the levee systems examined, levees on peat are found to be significantly more fragile than levees on inorganic soils.

Detailed analyses were performed for ten cross sections along the Tokachi River where strong motion recordings on the levees are available for the 2003 earthquake. Typical geotechnical performance assessment methods (liquefaction susceptibility, triggering, and consequence) are applied to examine the degree to which the observed field performance can be predicted. 2-D limit equilibrium models are constructed to evaluate slope displacements from Newmark analysis. A composite prediction framework considering both liquefaction severity indices and slope displacements is proposed to account for damage from multiple failure mechanisms and the consequence of liquefaction in the foundation and/or body of the levee.