VIEWED AS AN INVERSE PROBLEM
6. ARRIVAL TIME DATA IN SOUTH-EAST AUSTRALIA: A LINEARIZED INVERSION
6.1 The dataset: Travel times from natural and controlled seismic sources
6.1.1. Earthquake data
6. ARRIVAL TIME DATA IN SOUTH-EAST AUSTRALIA:
A LINEARIZED INVERSION
The success of any study of seismic velocity structure will ultimately depend upon
the quality and quantity of the data available. As Lanczos (1961) put it, "no mathematical
trickery will ever make up for a fundamental lack of data". This is intuitively obvious but
nevertheless a very important factor. In the next section we give details of the dataset
compiled for the 3-D inversion, together with an account of the selection procedure involved.
We then examine the question of adequate parameterisation of the earth model under the
confines of a linear regime. Next the inversion process itself is examined and a suitable
scheme presented. Finally results of the linear inversion are discussed together with those
arising from various modifications to the initial scheme.
6.1 The dataset: Travel times from natural and controlled seismic sources
The dataset for the current work consists of two major types: arrival times generated
from local earthquakes recorded at a regional network of seismic stations and travel times
from blasts recorded at a series of temporary geophones. The earthquake data has been
selected from recordings of the South-east Australian network over more than twenty five
years of operation. The travel time data has been derived from five separate Crustal
refraction surveys performed by the Bureau of Mineral Resources, Geology & Geophysics
(B.M.R.), during the 1970's.
6.1.1. Earthquake data
The south-east Australian seismic network, (Fig. 2.5) has been discussed briefly in
section 2. At present there exist sixteen permanent stations. The three most northerly are
operated by the Sydney Metropolitan Water Board and the rest by the Australian National
3-D Linear Inversion 6.2
network to the south west operated by the Phillip Institute of Technology and, more recently,
by two extra stations maintained by the B.M.R. The signals from the thirteen stations
operated by the A.N.U. are sent to the Research School of Earth Sciences via radio telemetry
and recorded on a multi-channel photographic recording system.
While monitoring the local seismicity for more than a quarter of a century, over 6000
events have been routinely located. All times are picked by an observer from photographic
traces projected onto a screen. The amount of arrival time data that may be reliably
determined is restricted by both the signal to noise ratio and the clarity of the projected
image. Over regional distance scales first arriving longitudinal P-wave and transverse S-
wave phases are by far the clearest and easiest to pick. Later phases are occasionally
observed and used in the routine earthquake location. Picking errors are estimated at 0.1 s
for the first P phase and 0.3 s for all other phases. All of the arrival times available have
been re-examined for the current study.
6 .1 2 Earthquake selection criteria
The seismicity of the south-east Australian region is generally diffuse (Fig. 6.1).
However there are areas such as the Dalton-Gunning and Young seismic zones, in the central
part of the region, where the level of seismicity is significantly higher than in surrounding
areas (Cleary 1967). Conversely there exist regions such as southern New South Wales
where a lower level is quite noticeable. To achieve the maximum data coverage possible
across the region, the entire earthquake catalogue was scanned for suitable events. Initially
all events with poorly constrained hypocentres were rejected. The remainder were examined
for clarity of recording, as judged by the original observer, and number of stations detecting
the event. All events detected by less than six stations were subsequently rejected. Finally events lying within the network with magnitude mL < 3, and outside with mL < 4 outside
were also rejected.
As expected within the network, many earthquakes were found to satisfy all selection
Robertson Picton Talbingo Dalton - Gunning Eucumbene * Melbourne □ M > 5 0 4 < M < 5 o 3 < M < 3-9 x 2-5 < M < 2-9 . M <2-5 100 km Gippsland
Figure 6.1 Seismicity of S.E. Australia from 1960-1983, with the more active regions indicated by place names.
L a ti tu d e 8 L ongitude
Depths
Figure 6.2 Epicentres of 315 events used in the inversion study and their rclaiion to the 16 station network. Epicentres are those determined by the fully nonlinear location algorithm presented in section 2.
Figure 6.3 A histogram of the depth distribution of all 315 earthquakes used in the inversion study. All events were relocated using the nonlinear location algorithm.
3-D Linear Inversion 6.3
With the aim of producing the highest quality dataset possible, central events were again
filtered and those with less than nine stations rejected. This resulted in a further 30%
reduction in travel times. Since the central region is well sampled, this rejection of possibly
useful data is not thought to be significant.
The station source separation also played an important role in the selection process.
To achieve raypath sampling across as wide a range of depths as possible, special
consideration was given to large magnitude events at large epicentral distances. (In practice
it was found that most of these were netted by the initial procedure anyway).
Of the large number of events recorded by the south-east Australian network only a
relatively small percentage were finally chosen for the inversion. Fig. 6.2 shows the
epicentral distribution of these events in relation to the network geometry. The diagram also
shows the perimeter of the region considered for the inversion. Obviously a large number are concentrated in the central region within the network. However several mL ~ 4 - 5 have
been recorded to the west and provide coverage there. The earthquake data is summarized in
Fig. 6.7. In total over 300 events were found to be suitable. These produced over 4000
raypaths crossing the region from varying azimuths and depths (see Fig. 6.3). Overall the
distribution of earthquake sources is thought to provide reasonable coverage across the entire
network and in the immediate surroundings.