Cone Penetration Tests (CPT) derived from the Hamilton section of the Waikato Expressway were analysed within CLiqTM software. The derived Liquefaction Potential Index (LPI) from each CPT was then combined with LIDAR, pedological and geological maps for statistical analysis. A soil model that incorporates the conditions of modern soil development with these derived LPI values was developed as a preliminary assessment tool for liquefaction potential within Hamilton Basin soils. The model shows that liquefaction is more likely to occur on interfluvial areas where there is little topographical relief. Pedological soils with high organic component are also a likely indicator of high liquefaction susceptibility.
Liquefaction is the process by which an increase in pore water pressure occurs from seismic stressors, resulting in loss of shear strength and resultant fluid-like behaviour of the affected soil (Obermeier, 2009). The ideal conditions for such an event include relatively recent (Holocene to Late Pleistocene) sediments that are cohesionless, dominated by coarse silt to fine sand, loosely packed, with a shallow water table (Eslami, Mola-Abasi, & Shourijeh, 2014; Owen & Moretti, 2011; Robertson & Wride, 1998). Within the Hamilton Basin, the Hinuera Formation comprises extensive Late Pleistocene (c. 22 to 17 cal ka) volcanogenic (mainly rhyolitic) alluvium dominated by unconsolidated and highly variable sediments, including pumiceous, ranging from cross-stratified gravelly or slightly gravelly sands, sandy gravels, and silts that show marked changes in lithology both vertically and horizontally over short distances (Hume, Sherwood, & Nelson, 1975). The Hinuera Formation has the potential to liquefy in the event of an earthquake because its physical properties in many places meet the criteria mentioned above. Paleoliquefaction features have occurred within the Hinuera Formation (Clayton & Johnson, 2013; Kleyburg, 2015). Pedological soils (at the land surface, usually to ~1 m depth) tend to reflect underlying characteristics of their parent materials in terms of texture, drainage, and water table level, these features being encompassed by ‘soil family’ for this particular study. A soil family is the fourth category in the New Zealand Soil Classification (NZSC) (Webb, 2011), with families being defined on the basis of physical properties, not genesis. These characteristics are also those which control the susceptibility, or otherwise, of subsurface materials to liquefaction. This paper investigates the thesis (following Kleyburg et al., 2015) that the mapped pedological soils on the surface of the Hinuera Formation provide a means of predicting the likely liquefaction susceptibility of a site.
2.1 Site selection and data acquisition
The Hamilton Basin was chosen as the area of study because of the hazard a liquefaction event would pose to its many residents. The data were selected on the basis of availability from the New Zealand Geotechnical Database (NZGD), which provided a record of the initial CPT testing carried out before development began on the Hamilton section of the Waikato Expressway. LIDAR data were sourced from Waikato Regional Council. The pedological soil map was provided by Waikato Regional Council with the permission of Landcare Research. A total of 216 CPT data points were available in NZGD over a distance of ~18.5 km from Horsham Downs (north of Hamilton) to Tamahere (south of Hamilton).
2.2 CLiqTM analysis
Raw CPT data were entered into CLiqTM software to obtain the predicted Liquefaction Potential Index (LPI) developed by Iwasaki et al. in 1978 for assessing liquefaction potential based on the depth of the liquefiable layer in relation to ground surface as well as its thickness and its computed factor of safety (FS) (Toprak & Holzer, 2003). LPI is classified within CLiqTM as either low (<5), moderate (5-15) or high (>15) potential based on the Robertson and Wride (NCEER 1998, 2009) calculation method for three separate depths of 3 m, 5 m and 10 m. Liquefaction potential (derived from LPI) was considered at a depth of 3 m as shallow liquefaction will be the most damaging in general, yet shallower materials are unlikely to liquefy due to lack of normal stress (Luo, Wang, & Li, 2013). A maximum depth of 10 m was considered because liquefaction deeper than this is likely to extend to the ground surface only for very large proximal earthquakes (Huang & Yu, 2013). The parameters used to compute LPI were derived from multiple sources. Input parameters included: horizontal peak ground acceleration (pga); earthquake moment magnitude (Mw); water table depth; and fines content. Horizontal pga was computed as 0.38 (g) using ah = ZRC where Z (Hazard factor) = 0.16, R (Return period factor) = 1.8, and C (Site response factor) = 1.33 (NZ Transport Agency, 2014 & Technical Committee BD-006-04-11, 2004). A magnitude Mw = 7 event was assumed as worse-case selected from ranges of 5.5-7 Mw suggested for the Kerepehi Fault (Persaud et al., 2016; Wallace, Hamling, Holden, Villamor, & Williams, 2016). There is a relatively wide range within the literature regarding ground water table depth within the Hamilton Basin, with values as low as 0.6 m and as high as 4 m recorded in the field area. To account for water table level variation a depth of 1 m was used following the examples of Opus (2014) and (Ministry of Business Innovation and Employment, 2017; Opus, 2014) Fines content was also included in analysis as the finer component of a soil has a significant influence on liquefaction potential (Thevanayagam, 2000).
2.3 Data analysis
Data were compiled in Excel into a uniform format that could be easily imported into the software that would carry out analysis. The Excel spreadsheet consisted of each CPT having an appropriate label, alongside GPS coordinates, and the specific cumulative LPI values for 3, 5, and 10 m depths derived from the raw results of CLiqTM. Factors of slope, elevation (derived from LIDAR DEM data), and soil families and associated soil siblings (defined by other physical properties and recorded using numbers) from S-map (Lilburne et al., 2011; Landcare Research, 2016) were included in the spreadsheet that was then imported into STATISTICATM. Soil families, identified by a geographical name with suffix f, are defined by the nature of soil profile material to 100 cm depth, parent rock (if present), dominant texture class to 60 cm depth, and permeability of the slowest horizon within 100 cm depth (Webb & Lilburne, 2011).
Parameters of LPI, slope, elevation, soil family and soil sibling were put into STATISTICA software to determine the most influential factors to liquefaction. Regressive Exhaustive CHAID (chi-square automatic interaction detection) was developed to identify and categorise parameters that are of higher influence to LPI where liquefaction potential was input as the predictive value (dependent variable), with slope, elevation and soil texture then input as continuous variables and soil family as a categorical variable (independent variables). The resultant graph then displays each independent variable as separate classes (low, moderate and high liquefaction potential) based on their associated LPI. Due to limited data availability some soil families only had one CPT test undertaken within them and therefore only one recorded LPI value. These families were not included within the analysis as only one LPI value could not faithfully represent the liquefaction potential for that particular family. GIS (Geographical Information Systems) was then used to depict the results from the statistical analysis in terms of liquefaction susceptibility.
3.1 CLiqTM analysis
Liquefaction susceptibility, displayed as data points to the west of Hamilton City, show an increasing cumulative susceptibility with an increase in depth (Figure 1). At 3 m depth susceptibility is low (119 sites) to moderate (97 sites) with the majority of sites being considered as having a low susceptibility to liquefaction (LPI <5). At greater depths susceptibility can be seen to range from low to high with 10 m depths being dominated by high liquefaction susceptibility. When referring to Figure 1A, areas of moderate susceptibility appear more concentrated toward the southern end of the field area; within Figure 1B, high susceptibility appears within the middle/lower region of the field area also.
Figure 2 illustrates eight individual CLiqTM output traces (3 m depth), with colours (see key) representing normalized Soil Behaviour Type (SBTn) and the yellow line indicating what the recorded soil behaviour was at each particular depth. These eight profiles were chosen because they illustrate how variation in soil behaviour can affect the resultant liquefaction potential, with each profile illustrating a different degree of LPI (from no liquefaction to moderate-high potential liquefaction occurrence). Profiles were derived from upper (northern), middle, and lower (southern) sections of the field area to give an accurate representation of entire data set.
Those that are of no to low susceptibility (Figure 2A & B) have a soil behaviour that is nearer to that of a granular soil or in contrast has a soil behaviour that is too-fines rich, typically >2.6 Ic (soil behaviour type index). As soil behaviour moves into more silt/sand textures susceptibility increases. The main difference that can be seen between those that read low moderate compared with those of high moderate is the influence of an increased fines content toward the top of the profile reaching a soil behaviour value of Ic ~2.6-3. Moderate susceptibility profiles show a similar pattern but do not exceed Ic=3, having a higher proportion of clay dominated fines.
Figure 1: Field area (Hamilton city with Waikato Expressway Hamilton Section under construction to the east). Data points represent CPT sites and subsequent LPI based on Robertson and Wride (NCEER 1998, 2009) calculation method. (A) 3 m depth, (B) 5m depth, (C) 10 m depth.
Figure 2: SBTn plots showing changes in proposed soil behaviour from 0.1- 3 m with each profile illustrating a different LPI value (Fig 2A-0, Fig2B-0.5, Fig2C-1.8, Fig2D-3.5, Fig2E-5.1, Fig2F-6.5, Fig2G-8.8 & Fig2H-11.28).
3.2 Statistical analysis
The results of the Exhaustive CHAID analysis can be seen in Figure 3. Of the independent variables, soil family, slope, elevation and sibling number, it can be seen that soil family is the most influential (dependent variable) when it comes to predicting liquefaction potential (shown in Figure 3 at 3 m depth but also shown at 5 m and 10 m depth when analysed). The soil families can be classified into three classes of liquefaction potential, low, medium and high. Pukehinaf, Moeatoaf, Kohuratahif, Kainuif, and Rotokaurif families are classified as low LPI (mean value of 3.72); Otorohangaf, Matakanaf, and Te Puningaf families as medium LPI (mean value of 5.3); and Utuhinaf and Kaipakif families as high LPI (mean value of 7.57). Note, however, there is a relatively high variance within each LPI class (between 5–8), reflecting the variability within the soil families. With the exception of the high susceptibility node (ID 4–7.5 LPI) which terminates at soil family, the second most influential factor is elevation (Figure 3 has been ‘pruned’ for simplification and so does not show further splits based on less influential independent variables). For lower susceptibility soils (ID 2) elevation is, on average, higher than moderate susceptibility (ID 3) at approximately 36–38 m and 24–29 m, respectively. It is apparent from the data that liquefaction potential is at its highest when topography is less complex with little to no relief.
It can be seen within Figure 4, that on average, the susceptibility class predominantly ranges from low (green) to low-moderate (orange) with high-moderate values being confined to the Northeast of Hamilton City. The three main topographical features that can be identified on the surface of the Hinuera Formation are small ridges (mounds) up to a few metres in height, channels (paleo and current), and the interfluve zones (flat areas with little topographical relief). From Figure 4 it is apparent that when the topography becomes more complex the susceptibility is seen to decrease, such as in the drainage channels as well as along the low ridges. The areas that are displaying moderate to high susceptibilities are mainly within the interfluvial zone where ground is near to level.