Lateral spreading in the Canterbury earthquakes -

Observations and empirical prediction methods

H.J. Bowen, M.E. Jacka, S. van Ballegooy & T.J.E. Sinclair

Tonkin & Taylor Ltd, New Zealand

H.A. Cowan
New Zealand Earthquake Commission

SUMMARY:
The recent large earthquakes in Canterbury, New Zealand have caused significant damage to tens of thousands
of houses across the Canterbury region. The most severe damage to houses was often associated with
liquefaction-induced lateral spreading. The occurrence, magnitude and distribution of lateral spreading was
different for each major earthquake in the sequence, and was dependent on a number of different factors such as
ground shaking intensity, liquefaction susceptibility, topography and geological conditions.

Lateral spreading observations are summarised for two events, the Mw 7.1 earthquake of 4 September 2010 and
Mw 6.2 earthquake of 22 February 2011. Data was collected from a variety of sources, including extensive post-
earthquake mapping, pre- and post- earthquake LiDAR scanning, and crack width mapping. Information from an
extensive geotechnical investigation and strong motion records was used to compare the observed lateral spread
to predictions of a widely used empirical method (Youd et al., 2002). Lateral spread displacements were found to
be highly dependent on geological and topographical features, and for this case study empirical methods
generally over-predicted lateral spread displacement magnitude at large distances from the free face.

Keywords: liquefaction, lateral spreading, Christchurch earthquakes, image coregistration

1. INTRODUCTION

Lateral spreading was a major cause of damage in the recent Canterbury earthquake sequence; the
eastern suburbs in particular incurred severe damage to houses, bridges, underground services and
roads. Despite the significant economic impact of lateral spreading, the phenomenon is not well
understood, and design methods are necessarily approximate in nature due to large aleatoric
uncertainties inherent in both the nature of the ground conditions and of earthquake shaking. The
objective of this paper is to summarise observations from this earthquake sequence, as part of a
longer-term goal of enhancing design methods.

This series of earthquakes has subjected liquefaction-susceptible soils in the region to four strong
events, each with different characteristics varying significantly across the city. Combined with detailed
mapping of observations, numerous strong motion records and extensive geotechnical investigations,
analysis of these earthquakes provides an opportunity to examine key aspects of lateral spreading.
Three such aspects are discussed in this paper: the triggering of lateral spreading, the displacement
pattern and extent of lateral spreading and the magnitude of lateral spreading displacements.

2. GEOLOGICAL SETTING

Christchurch is located on the east cost of New Zealand’s South Island. Figure 1 shows a broad
overview of the primary geological units in Christchurch; it can be seen that the city is founded on the
boundary of alluvial Springston Formation and the marine Christchurch Formation.
 Christchurch CBD Christchurch
Formation

Sand dunes
Springston Formation
Overbank deposits
Drained
lagoons
Historic flood channels

Study areas in Fig. 2

Port Hills - Volcanic Rock

Figure 1 – Map of Christchurch showing primary geological units – Springston Formation (spy) in yellow and
the Christchurch Formation (ch) in green (from Brown and Weeber 1992)

The Springston Formation comprises post- glacial fluvial channel and overbank sediments, with three
distinct units, (1) alluvial overbank silt which becomes sandy closer to the coast, (2) flood channels
infilled with gravel, and (3) peat deposits of former swamps. The Christchurch Formation comprises
beach dune sands, and silts and sands from estuarine, lagoonal and coastal swamp deposits. These
units are derived from material from the Southern Alps, transported out to sea by the Waimakariri
River located to the north of the city.

3. CANTERBURY EARTHQUAKE SEQUENCE

Since the Mw7.1 earthquake occurred on 4 September 2010, a sequence of significant earthquakes has
struck the greater Christchurch region, as summarised in Table 1. All four earthquakes caused lateral
spreading to some degree; the most damaging earthquake occurred on 22 February 2011.

Table 1 – Summary of major earthquakes and recorded ground motions in the Canterbury earthquake sequence
Earthquake Moment magnitude, Mw Distance from epicentre Typical horizontal PGA in
in eastern suburbs eastern suburbs
4 September 2010 7.1 40 – 42km 0.16 – 0.23g
22 February 2011 6.2 6 – 10km 0.22 – 0.67g
13 June 2011 6.0 7 – 10km 0.22 – 0.46g
23 December 2011 5.9 4 – 6km 0.30 – 0.32g

4. TRIGGERING OF LATERAL SPREADING

Figure 2 compares the extent of lateral spreading that occurred in the 4 September 2010 and 22
February 2011 earthquakes. It can be seen that 22 February earthquake caused lateral spreading to
occur in areas that did not suffer spreading in September. Land that underwent lateral spreading in
September was the worst affected in February. Of particular interest are sites where liquefaction
occurred in both earthquakes, but lateral spreading occurred in February but not September. One such
case study is in the suburb of Avondale; two sites labelled A and B in Figure 2.

Site investigations indicate the soil profile is similar across Avondale, with 8-9m of Springston
Formation silt and sand overlying denser Christchurch Formation sand. The soil profile at sites A and
B comprises low plasticity sandy silt from 0 to 3m below ground level, overlying fine grained loose to
medium-dense sand with trace silt from 3 to 9m, overlying dense sands from 9m depth. Both sites are
a similar distance from the riverbank, the height of the riverbank and the depth of the river is similar at
both sites. The key difference is the density of the intermediate sand layer – at Site A the sand is loose
with CPT tip resistance varying from qc = 2-4MPa. At Site B the sand is medium dense, with qc = 4-
8MPa, as shown in Figure 3.

Analysis indicates that the factor of safety (FOS) against liquefaction, calculated using the method of
Idriss and Boulanger (2008), is lower at Site A than Site B. Table 2 summarises the results, which
indicate that the observed lateral spread correlates well with the FOS against liquefaction. This
suggests that for this case study rapid and complete liquefaction of a soil layer is required for lateral
spread to occur. Note that conventional Newmark sliding block analysis with post- liquefaction soil
strength would suggest both Sites A and B would undergo lateral spread in the September earthquake.

Figure 2 - Map of eastern Christchurch indicating the extent of liquefaction induced lateral spreading:
(a) 4 September 2010 earthquake, with detail locations in Figure 5; (b) 22 February 2011 earthquake
 CPT tip resistance, qc (MPa) CPT tip resistance, qc (MPa)
0 5 10 15 20 0 5 10 15 20
0 0
1 1
CPT @ A CPT @ B
2 2
3 3
4 4
Depth (m)

5 5
6 6
7 7
8 8
9 9
10 10

Figure 3 – Cone Penetration Test results at two locations in the suburb of Avondale. At location A lateral
spreading occurred in both the 4-Sep-2010 and 22-Feb-2011 earthquakes; at location B lateral spreading
occurred in February only, despite liquefaction also occurring in September.

Table 2 – Calculated factor of safety against liquefaction in critical layer at Sites A and B
4 September 2010 22 February 2011
Site
Mw=7.1, PGA=0.19g Mw=6.2, PGA=0.37g
A FOS = 0.5 – 0.6 FOS = 0.4 – 0.5
B FOS = 0.8 – 1.0 FOS = 0.6 – 0.8

5. LATERAL SPREADING DISPLACEMENT PATTERNS & EXTENT

After each of the four main events to date in the Canterbury earthquake sequence, LiDAR survey of
ground levels was undertaken to provide information to assist recovery efforts. By comparing the
digital elevation models from before and after earthquake events, a pattern-matching image
coregistration process was applied to estimate the horizontal ground movements which have occurred.
This work was commissioned by EQC and undertaken by Imagin' Labs Corporation and GNS Science,
refer to Leprince et al. (2007) for details of this analysis technique. This provides estimated horizontal
ground movements from each earthquake, on a 4m grid across much of the Greater Christchurch urban
area. By combining the absolute ground movement from the LiDAR analysis with the regional
tectonic movements from a fault rupture model, an estimate was made of the local component of
ground movement (caused by lateral spreading and other local forms of permanent ground
displacement).

As one method for ground-truthing the horizontal displacements from the LiDAR analysis, the
displacements were compared to the ground-based mapping of post-earthquake observations on each
individual property, as summarised in Figure 4 for the 22-Feb-2011 earthquake. It should be noted that
these properties are generally set back from the river’s edge – the LiDAR analysis and observation
indicates that greater ground displacements occurred in the road and reserve land closer to the free-
face. The following observations can be made from Figure 4:

The LiDAR analysis indicates large local horizontal ground displacements (typically 150 –
700mm) on properties where ground-based mapping identified that severe lateral spreading
had occurred. This magnitude of displacement is in general agreement with the ground
cracking observed in these areas.
 Moderate local horizontal ground displacements (typically 100 – 500mm) are indicated for
properties with severe liquefaction (likely due to ground-oscillation or sloping-ground effects)
and moderate lateral spreading (likely due to free-face effects). This magnitude of
displacement is in general agreement with the ground cracking observed in these areas.

The analysis indicates minor local horizontal displacements (typically 100 – 300mm) for
properties with moderate, minor and no liquefaction observed. This magnitude of
displacement is greater than implied by the limited ground cracking observed in these areas.
Much of the displacement indicated by the analysis is likely to be noise, due to the limited
accuracy of the LiDAR data and tectonic model. However some of this displacement may be
real, associated with minor cyclic seismic displacements within the deep soil profile, slope
movement, or similar effects not associated with lateral spreading.

100%
Cumulative Frequency (Property Count)

90%

80%
Post-Earthquake Mapping Observations
70%
Figures in brackets are number of properties
in each category within lidar coverage area
60%
Not mapped (12714)
50%
No liquefaction observed (8474)
40% Minor liquefaction (2150)

30% Moderate liquefaction (22987)

Severe liquefaction (5000)
20%
Major lateral spreading (4219)
10%
Severe lateral spreading (794)
0%
0 100 200 300 400 500 600 700 800 900 1000 1100 1200
Horizontal Ground Displacement in Feb-2011 Earthquake (mm, with tectonic displacement removed)

Figure 4 – Correlation between ground-based post-earthquake mapping and displacement from LiDAR analysis,
for 22 February 2011earthquake.

Figure 5 presents results from the analysis of LiDAR data for three areas within the eastern suburbs of
Christchurch. Figure 6 summarises the observed trends of reducing ground displacement with distance
away from the free edge. The ground displacement patterns in each of these areas show a number of
different effects as discussed below.

Figure 5(a) shows the suburbs of Richmond, Avonside and Dallington, with the Avon River
meandering through the centre. The ground elevation map on the left clearly shows the position of
river terraces and lower-lying point bar deposits. The map on the right shows the magnitude of local
horizontal displacement that occurred due to the 22 February 2011 earthquake (after correction to
remove the tectonic component of ground movement). Ground movement is predominantly directly
towards the nearest riverbank, driven by both free-edge effects and the moderately-sloping
topography. It can be seen that in this area, the most significant lateral spreading ground displacements
are generally constrained within the boundaries of the lower river terraces. It is inferred that on the
upper terraces, where significant lateral spreading was not observed despite widespread liquefaction
occurring, the denser soils suffered less-severe strength loss following liquefaction triggering.

Figure 5(b) shows that the suburb of Avondale (to the east of the river) is predominantly near-flat low-
lying land, with some higher ground to the south and north. In the near-flat region (0.1 – 0.2% slope)
there appear to be no distinct topographical or geological constraints to the lateral spreading ground
movements (the nearest significant change is the higher and denser dune deposits about 700m to the
east, just off the edge of the map). Lateral ground movements appear to be driven predominantly by
free-face effects, with little effect from inland topography. Without constraints, the magnitude of
ground displacement reduces gradually with distance back from the free-face over a width of about
150 – 200m (beyond this distance any lateral spreading toward the free-face is lost in the background
noise and other forms of ground movement). In the northern and southern areas, minor lateral ground
movements extend further inland, driven by the more steeply-sloping higher ground.

Figure 5(c) shows the suburb of Horseshoe Lake, which is surrounded by an oxbow lake from the
historic meandering river channel. The lowest-lying land in the southeast is approximately 1.2m below
the higher ground in the northwest and southwest, and 0.6m below the ground in the west (ground
slope of 0.1 – 0.2%). Along the western and northern reaches of the lake, lateral spreading has
occurred towards the nearby free-face only in areas which are very close to the edge. The predominant
driver of lateral ground movement across the majority of the Horseshoe Lake suburb appears to be the
overall ground topography sloping towards the low-lying south eastern area.

6. MAGNITUDE OF LATERAL SPREADING DISPLACEMENTS

The magnitude of the lateral spreading displacements obtained from the LiDAR survey as described in
Section 5 was compared with a widely used empirical method for predicting lateral spreading
displacements (Youd et al. 2002). The measured displacements have been baseline-corrected to zero
once the displacement profile reaches a near-constant value – this is intended to isolate the component
of ground movement associated with lateral spreading towards the free edge, filtering out the noise
and other types of ground movement present in the data.

The predicted displacement was determined for the 22 February 2011 earthquake using information
from a comprehensive post- earthquake geotechnical investigation. 28 boreholes were drilled
alongside the Avon River in areas affected by lateral spreading, the location of these boreholes are
shown in Figure 2 and Figure 5. In addition, hundreds of CPTs and kilometres of MASW geophysical
survey lines were also undertaken in these areas. Table 3 summarises the sources of key information
used in this assessment.

(5a) Avonside
 (5b) Avondale

(5c) Horseshoe Lake

Figure 5 – LiDAR analysis for selected suburbs of eastern Christchurch. Left image is LiDAR elevation survey;
changes in elevation correspond to the presence of different geological units such as river terraces or sand dunes.
Right image shows the magnitude (colours) and direction (vectors) of lateral spreading displacements, calculated
from the difference between pre- and post- earthquake LiDAR surveys for the 22 February 2011 earthquake.
 1600 1600
Lateral Ground Displacement (mm)

Lateral Ground Displacement (mm)

1400 Avonside 1400 Dallington
1200 1200

1000 1000

800 800

600 600

400 400

200 200

0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Distance From Free Edge (m) Distance From Free Edge (m)
1600 1600

Lateral Ground Displacement (mm)

Lateral Ground Displacement (mm)

1400 Avondale 1400 Horseshoe Lake

1200 1200

1000 1000

800 800

600 600

400 400

200 200

0 0
0 50 100 150 200 250 300 0 50 100 150 200 250 300
Distance From Free Edge (m) Distance From Free Edge (m)

Figure 6 – Observed displacement vs. distance profiles for selected suburbs, derived from LiDAR analysis.
Displacement profiles have been baseline-corrected to zero displacement at the minimum point – to isolate the
component of ground movement associated with lateral spreading towards the free edge (filtering out the noise
and other types of ground movement present in the data).

Table 3 – Parameters used in Youd et al. (2002) method

Parameter Information source
Magnitude, M M = 6.2
Source distance, R Determined two ways:
(a) R = actual distance from earthquake epicentre
(b) R = Req, equivalent source distance calculated using peak
ground accelerations recorded nearby
Height of free face, H LiDAR survey and pre- earthquake riverbed survey
Thickness of liquefiable deposit, T15 SPT testing in boreholes adjacent to the Avon River
Average fines content, F15 Laboratory tests
Average mean grain size, D5015 Laboratory tests & interpretation from logged soil description

The Youd et al. Method was developed for western U.S. and Japan where attenuation of seismic
waves is relatively high. For other seismic regions, Youd et al. recommend the use of an equivalent
source distance, which is determined using attenuation models that incorporate the earthquake
magnitude and the peak ground acceleration. In general, the peak ground accelerations experienced in
eastern Christchurch were much higher than conventional attenuation models suggest. Therefore, for
the purposes of this paper, both the equivalent source distance (calculated using the observed peak
ground accelerations recorded nearby) and the actual source distance were used to prepare predictions.
Figure 7 compares the predicted and observed displacements. In general, displacements immediately
next to the free edge are affected by instability and are difficult to predict. Distances further back are
 1 1
30m from free face 30m from free face

Measured lateral spreading disp. (m)

0.9 0.9
Measured lateral spreading disp. (m)

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Predicted lateral spreading displacement (m) Predicted lateral spreading displacement (m)

1 1
50m from free face 50m from free face

Measured lateral spreading disp. (m)

0.9 0.9
Measured lateral spreading disp. (m)

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Predicted lateral spreading displacement (m) Predicted lateral spreading displacement (m)

1
1
100m from free face 100m from free face
0.9 0.9
Measured lateral spreading disp. (m)
Measured lateral spreading disp. (m)

0.8 0.8
0.7 0.7
0.6 0.6
0.5 0.5
0.4 0.4
0.3 0.3
0.2 0.2
0.1 0.1
0 0
0.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Predicted lateral spreading displacement (m) Predicted lateral spreading displacement (m)

Figure 7 – Comparison of predicted lateral spreading displacement using Youd et al. (2002) versus lateral
spreading displacement measured using LIDAR survey. For the plots on the left hand side (blue symbols) the
predicted displacements were calculated using the actual distance to the source. On the right hand side (red
symbols) the predicted displacements were calculated using the equivalent source distance. The solid line
represents predicted displacement = measured displacement, and the top and bottom dashed lines represent
predicted displacement = 0.5 times and 2 times the measured displacement respectively. As shown in Figure 6,
measured displacement profiles have been baseline-corrected to zero displacement at the minimum point.
typically of more relevance to engineers, so comparisons at 30m, 50m and 100m distance from the
free edge are presented. For the plots on the left hand side (blue symbols) the predicted displacements
were calculated using the actual distance to the source. On the right hand side (red symbols) the
predicted displacements were calculated using the equivalent source distance.

From inspection of the data it can be seen that:

There is a large amount of scatter in the data, however at 30m and 50m from the riverbank the
measured displacements were broadly accurate within a factor of 2 when using the actual
source distance.
Use of the equivalent source distance results in an over-prediction of displacement for these
case studies. This may be due to the strong ground motions experienced in this particular
earthquake, rather than the prediction model itself. It does demonstrate however that the
prediction method is very sensitive to small magnitude earthquakes close to the source.
For these case studies, the method over-predicts lateral-spreading displacements at larger
distances (greater than 100m) back from the riverbank.

7. CONCLUSIONS

This paper provides a summary of observations and preliminary analysis on lateral spreading during
the Christchurch earthquakes. Key outcomes from this work include:

Rapid and complete liquefaction of susceptible layers is required to trigger lateral spreading
for this case study of moderate-magnitude earthquakes.
Geological and topographical features have a great influence on the displacement pattern and
extent of lateral spreading.
Application of various analysis techniques to the LiDAR data provides valuable case-study
information that can be used to further develop empirical lateral spreading analysis methods.
Displacements predicted by the Youd et al. method were greater than measured displacements
at large distances from the free edge for this case study, possibly related to the high PGA
experienced close to the fault in this moderate-magnitude earthquake.

ACKNOWLEDGEMENT
This work would not have been possible without the data obtained by the New Zealand hazard monitoring
system (GeoNet) and the extensive remote sensing and ground investigations of land damage, sponsored by the
New Zealand Government through its agencies the Earthquake Commission of New Zealand, the Ministry of
Civil Defence and Emergency Management and Land Information New Zealand.

REFERENCES

Brown, L.J. and Weeber, J.H. (1992). Geology of the Christchurch Urban Area. Scale 1:25,000. Institute of
Geological & Nuclear Sciences Geological Map 1.
Idriss, I.M. and Boulanger, R.W. (2008). Soil liquefaction during earthquakes. EERI MNO-12.
Leprince, S., Barbot, S., Ayoub, F. and Avouac, J. P. (2007). Automatic and Precise Ortho-rectification,
Coregistration, and Subpixel Correlation of Satellite Images, Application to Ground Deformation
Measurements", IEEE Transactions on Geoscience and Remote Sensing, Vol.45, No.6, June 2007.
Youd, T.L., Hansen, C.M., and Bartlett, S.F., (2002). Revised multi-linear regression equations for prediction of
lateral spread displacement, J. Geotechnical and Geoenvironmental Eng. 128(12), 1007-017.