Regional and local factors in attenuation modelling: Hong Kong case study

Abstract

Seismic attenuation behaviour is controlled by a large number of wave modification mechanisms. The characteristics of some of these mechanisms are specific to a local area, whilst the remainder can be generalised to the entire seismic region. Factors representing these mechanisms are often not resolved. A new attenuation modelling approach is demonstrated in this paper (using Hong Kong as a case study), to evaluate individual regional and local wave modification factors. Shear wave velocity (SWV) information for the four prevalent geological formations found in Hong Kong was first obtained: (a) at shallow depths from instrumented boreholes; (b) at depths of up to 100–200 m from measurements using the Microtremor SPatial Auto-Correlation (SPAC) technique; (c) at depths of up to 1.5 km from the monitoring of quarry blasts; and (d) at depths from 1.5 to 8 km in the hard basement rock layers from results of seismological refraction surveys. The upper-crust amplification factor calculated from the four modelled rock SWV profiles was then combined with predicted attenuation parameters to determine the upper-crust modification factor (filter function) incorporating the local wave modification characteristics associated with Hong Kong geological formations. Such functions may then be combined with the regional attenuation characteristics in that part of the South China region. A seismic attenuation model was developed by combining the upper-crust modification factor with the regional source function of intra-plate earthquakes, based on stochastic simulations. The ground shaking model developed from the presented methodology is supported by the comparison with macro-seismic data of seven historical earthquake events affecting Hong Kong.

Introduction

Seismic hazard assessment requires a representative ground shaking attenuation relationship to be developed. Attenuation relationships provide predictions for the intensity of ground shaking for any given earthquake scenario, expressed principally in terms of a combination of earthquake magnitude (M) and source-site distance (R).

The attenuation behaviour of earthquake ground shaking is highly complex, but can be approximated by a series of ‘filters’, each of which represents a seismic wave generation (or modification) mechanism along its entire transmission path between the source (at depth) and the site of interest (on the surface). The properties of these filters can be generalised to a region, an area or a site depending on the considered mechanism and the method of modelling. Thus, attenuation factors can be classified into (i) regional factors, (ii) area factors (local factors) and (iii) site factors. A recent study by Tsang and Chandler (in press) gives an example of how such a division of attenuation factors can be useful in conducting area-specific and event-specific seismic hazard assessment, especially in regions like South China (including Hong Kong) having large spatial variations in seismic activity rates and in the regional and local attenuation characteristics.

Regional factors characterise the seismic wave generation and transmission mechanisms that can be generalised to the whole region, and comprise the following: (i) source factors (representing properties of seismic waves generated at the source of the earthquake); (ii) geometrical attenuation factors (representing the spatial spread of the radiated seismic energy); and (iii) whole path attenuation factors (representing the dissipation of energy along the wave transmission path before the seismic waves reach the ‘upper crust’, comprising approximately the upper 4 km of the earth's crust). Local factors characterise the extent of amplification and attenuation (energy dissipation) mechanisms in the upper crust. Site factors characterise the filtering mechanisms within the soil sedimentary layers overlying bedrock, which operate on much smaller distance scales.

Despite distinctions between the three tiers of mechanism, their effects have seldom been resolved in existing attenuation models. In regions of high seismicity such as California where strong motion records are abundant, attenuation relationships are developed typically by regression of recorded ground shaking parameters (e.g. Sadigh et al., 1997). In this conventional modelling approach, regional and local factors contributing to the attenuation behaviour of the ground motion are not parameterised separately and are incorporated collectively into an attenuation relationship expressed as a function of magnitude, source-site distance and site classification. Consequently, ‘local’ conditions (as distinguished from ‘site’ conditions) within the region have not been parameterised.

In the absence of recorded strong motion data in regions of low to moderate seismicity, historical seismic intensity (MMI) data presented in the form of iso-seismal contour maps has typically been used to develop intensity attenuation relationships. The development of such relationships requires well-documented archives of historical earthquake events spanning a long period of time, in order that regions possessing different conditions are represented in the database. In Australia, for example, the available records only permit attenuation relationships to be developed for broad sub-regions, namely Western Australia, South–eastern Australia and North–eastern Australia (Gaull et al., 1990). Data are limited to a few major historical events clustered in certain locations. As for other attenuation relationships, intra-regional (local) variations have not been parameterised.

Seismic hazard modelling may alternatively be based on ground motions simulated stochastically in accordance with the seismological model, which characterises earthquake properties by their frequency content. Importantly, the various regional and local mechanisms identified above are represented by separate source and path factors within such a model. Refer to Section 2 for an overview of the seismological model and the listing of its factors.

Numerous seismological models have been developed for the stable continental region (SCR) of Central and Eastern North America (CENA), as reviewed by Atkinson and Boore (1998). Crustal factors can be very significant but there exists relatively little crustal amplification, or attenuation, in CENA because hard rock conditions in the ancient (Archean) rock formations of that region are characterised by very shallow SWV gradients. It is noted that crustal conditions in SCR's around the world can be very different to the conditions in CENA. For example, the eastern part of Australia (east of the Flinder's Ranges) is covered by much younger rock formations than most of Western and Central Australia, or CENA, even though the Australian continent is wholly a SCR. Furthermore, there could be significant variations within a region.

The concept of employing separate factors in the seismological model to represent regional and local mechanisms constituting the earthquake process is logical, and gives the model transparency. In the Hong Kong case study presented in this paper, the generic source factor for intra-plate earthquakes has been combined with regional attenuation factors identified for South China (refer Section 3). These regional factors have then been combined with local crustal factors based on the SWV profile developed specifically for each geological formation, to complete a regional average seismological model for Hong Kong (refer 4 SWV profiling for common geological formations in Hong Kong, 5 Determination of local factors, 6 Combined regional and local modification factors). Response spectra for rock outcrops obtained by stochastic simulations of the developed seismological model have then been obtained (refer Section 7).

The objective of presenting the Hong Kong case study in this paper is to stimulate a wider application of the modelling methodology to different areas around the world, and to explore alternative means of evaluating parameters to construct representative seismological models for a diversity of conditions.

The proposed methodology has been based on stochastic simulations and hence is subject to the usual limitations of not modelling azimuth and directivity effects in near-fault conditions. Such limitations are considered not to be important in SCR's, given that specific details of the potential fault source and wave paths are usually very limited and are insufficient for accurate modelling of such effects. Basin edge effects have similarly not been modelled herein. Despite these limitations, stochastic methods have been found to provide a reasonable representation of seismic hazard, according to the review by Atkinson and Somerville (1994). At the end of this paper (Section 7), ground shaking parameters have been compared with the same inferred from MMI data of historical earthquakes, in order to indicate their broad agreement.

Site filtering mechanisms occurring within soil sedimentary layers overlying bedrock are normally taken into account by site factors in codes of practice. Alternatively, one-dimensional non-linear shear wave analyses (by programs such as SHAKE developed by Schnabel et al., 1972) have been used as a popular engineering tool in modelling site effects including that of multiple-wave reflections within the soil layers causing conditions pertaining to resonance behaviour. The modelling of site modification factors (filters) is beyond the scope of this paper, but has been dealt with extensively elsewhere (Kramer, 1996).