Quantification of seismic site effects on slopes in Wellington
Degree GrantorUniversity of Canterbury
Degree NameDoctor of Philosophy
Wellington lies in an area of high seismicity, with multiple known active faults in the region and a major fault intersecting the city itself. Although the Wellington city centre is situated on relatively flat ground, vital infrastructure and many dwellings are located on the surrounding hills and at the toe of steep slopes. The urban development has also led to construction of numerous cut slopes, often unsupported. Although Wellington city has experienced shaking of MM intensity VI–VII many times over the past 150 years, it has not experienced severe shaking since 1855. Most of the current cuts have thus not been tested by strong shaking.
Seismic site effects in hilly terrain are the result of complex interaction between the site topography and geology, but are also affected by the properties of the incoming wave field. The site effects can change the amplitude, frequency content and duration of seismic shaking. In most cases, the site effects have been observed to amplify the shaking at the slope crests and deamplify at the slope toes. Quantitatively, however, there is significant variability in the published data. This variability reflects differences in the methodologies used to analyse site effects, in the quality of the analysed data, and the variability in the topography and geology of the investigated slopes.
In this thesis, we have investigated the weak motion seismic response of slopes using 4 characteristic slopes in the area (Owhiro Bay Quarry, Breaker Bay Cliff, Mitchell Street Ridge and Saint Gerard’s Monastery) using seismic records from the field and numerical models. Seismic arrays were deployed on the sites for periods ranging from 4 to over 11 months. Continuous seismic records of a combined length of over 100 months have been gathered from the arrays. The seismic signals recorded by the arrays were analysed in terms of the peak ground accelerations, Standard Spectral Ratios and Horizontal to Vertical Spectral Ratios of 10 second s-wave windows. Polarisation of the site response and source backazimuth effects were also analysed.
Dynamic models of the sites were created in FLAC 2D to further investigate the site effects on the sites, and to assess the reliability of the numerical models to predict how other similar slopes in the Wellington area may respond to earthquake shaking.
All our stations at the tops of the slopes show relatively similar, low mean amplification ratios (2–3.1 for the horizontal peak ground acceleration). Stations in the central parts of the slopes show larger scatter of the mean amplification ratios (from 0.9 to 7.2), with the amplification ratios correlating with the slope shape. The wavelength of the fundamental resonance, on the other hand, correlates well with the site dimensions (height and length). The central parts of slopes show relatively strong polarisation, controlled mainly by the local shape of the terrain. The slope crests mostly exhibit relatively complex polarisation patterns reflecting the 3D shapes of the investigated sites.
All stations also display strong effects of source backazimuth on horizontal amplification, and mostly limited effects on vertical amplification. In terms of peak ground acceleration, the effects of source backazimuth are stronger than the effects caused by the site polarisation. The effects of source backazimuth affect nearly the full range of investigated frequencies. While in some cases the effects of source backazimuth can be linked to the shape of the investigated topographic features, in other cases the effects seems to be controlled by the topography surrounding the investigates sites.
In accordance with previous observations, the numerical models were found to mostly underpredict the site amplification compared to the field data. The match between the models and the field data is, however, strongly site specific: while the models of some sites provided a good match with the field data, for other sites the models predicted amplification of an order of magnitude lower than that observed in the field. Use of field data for verification of numerical models is, therefore, absolutely critical.