Abstract
Liquefaction and associated lateral spreading during the 1987 Mw 6.5 Edgecumbe earthquake caused severe damage within parts of the Whakatane township. Liquefaction primarily manifested proximal to the Whakatane River in areas underlain by recent fluvial and marine sediments. The development of ground motion intensity and groundwater models for the Whakatane region for the Edgecumbe earthquake enabled CPT based liquefaction assessments to be applied using an extensive CPT dataset. Liquefaction assessments undertaken using the median PGA model and median groundwater model were found to closely correspond with the observed severity of liquefaction manifestation at sites known to have surface manifestation. However, significant over prediction of manifestation severity was evident in the Central Business District (CBD) under these conditions. Sensitivity analyses of the PGA and groundwater models were not able to account for these issues, with reduction in PGA and lowering of water table providing some improvement in areas without manifestation while at the same time providing underestimates in areas with manifestation. Overall, the findings suggest that standard CPT based methods of liquefaction assessment may be causing conservative predictions in the Whakatane CBD; further research is required to examine potential factors behind the inconsistent predictions.
1 INTRODUCTION
Earthquake-induced liquefaction and associated lateral spreading pose a significant hazard to the built environment. It is critical that liquefaction hazards are able to be adequately assessed so that the hazard can be effectively managed and the associated impacts minimized. The simplified liquefaction triggering procedures for assessing liquefaction hazards have been derived from liquefaction case histories collected following large earthquakes (e.g. Idriss and Boulanger, 2008 and Boulanger and Idriss, 2014). Cases where liquefaction is predicted by the simplified procedures, yet was not observed, provide important insights into the limitations of the current assessment methodologies, including their applicability in various soil types.
The 1987 MW 6.5 Edgecumbe earthquake caused localised liquefaction and lateral spreading in parts of the Rangitaki Plains in the Bay of Plenty, New Zealand. Liquefaction and lateral spreading resulted in severe damage to infrastructure and lifelines, including NZ$10 million worth of damage to flood control and drainage schemes within the region (Pender & Robertson, 1987; Christensen, 1995; Dowrick & Rhoades, 1990; Berrill, et al., 2001). The township of Whakatane experienced localized liquefaction and lateral spreading proximal to the Whakatane River, however no liquefaction was reported within the in the Central Business District (CBD).
In this paper, the extents of liquefaction manifestation within Whakatane for the Edgecumbe earthquake as collated from historical photographs, technical reports, and publications are presented. The collated extents are then compared with the predicted liquefaction severity estimated from the extensive CPT dataset collated for the Whakatane township using modelled peak ground accelerations (PGA) and the depth to groundwater at the time of the earthquake. Three peak ground acceleration models and three groundwater models were developed for the Whakatane region to account for the uncertainty in these variables during the earthquake (a median, lower estimate and upper estimate).
2 LIQUEFACTION OBSERVATIONS
Observations of liquefaction related land damage, including sand ejecta and lateral spreading associated with the Edgecumbe earthquake sequence, were compiled from technical reports, publications, and historical photographs (i.e. Pender & Roberson (1987), Franks et al. (1989) and Jennings et al. (1988)). The records were digitally compiled and presented geospatially for the purpose of this research (Figure 1). The accuracy in location and extent of the observations varies significantly as a function of the method of observations. For example, observations transcribed from aerial photographs onto hand drawn maps are significantly more innacurate than ejecta locations that have been digitised through precise descriptions of where samples have been taken (Such as in Pender & Robertson, (1987)) and near field photographs where landmarks are visible. Because of the above two categories have been chosen to distinguish between the confidence in the location of liquefaction manifestation, these being “Liquefaction Manifestation Confirmed” and “Possible Liquefaction Manifestation”.
The most severe liquefaction manifestation in Whakatane occurred in point-bar deposits within the inside bends of the Whakatane River, particularly the James Street Loop and the Netball Courts, and in paleo-channel deposits along the western side of the Landing Road Bridge (SH2). It is noted that there was no evidence of liquefaction having occurred throughout the Central Business District (CBD). Areas of liquefaction manifestation in Whakatane are summarised in Figure 1.
3 PEAK GROUND ACCELERATION MODEL DEVELOPMENT
The ground motion characteristics of the Edgecumbe earthquake were recorded by four strong motions stations (SMSs) within approximately 100 km of the causative fault plane, however there were no SMSs located within Whakatane. The closest SMS was Matahina Dam, which was approximately 11 km from the fault plane recorded a geometric mean PGA of 0.26g, while the western edge of Whakatane was approximately 9 km from the fault plane.
In order to estimate the Peak Ground Accelerations (PGAs) in Whakatane a number of ground motion models (GMMs) were assessed. From these, the Bradley (2013) GMM was chosen as the recorded PGA was well approximated by the 79th percentile of the Bradley GMM at each SMS location. Using this recorded PGA data, this percentile was then assumed to represent the median PGA for the Edgecumbe earthquake and the inter-event uncertainty was removed. To represent the uncertainty in the PGA only the intra-event standard deviation was used. This approach was taken as in the absence of any recorded data the combined inter- and intra-event uncertainty would result in a wide distribution of potential PGA values in Whakatane. By assigning a modified median representative of this event, the uncertainty reduces and is expected to be more representative of the actual PGA experienced in Whakatane.
The 15th and 85th percentile of the new probability distributions were used to represent upper and lower estimates of the PGA across the region. To provide another assessment of the accuracy of this model, the median PGA models was converted to equivalent MMI using the PGA-MMI relationships of Wald et al. (1999). The resulting MMI values were shown to correlate well with the reported MMI values across Whakatane. From this model a geospatial representation of the PGA across Whakatane was developed, with contours representing the respective PGA estimates presented in Figures 1-3.
4 GROUNDWATER MODEL DEVELOPMENT
Groundwater at the time of the Edgecumbe earthquake was modelled using a kriging interpolation method to develop a groundwater elevation surface between depth to groundwater at monitoring wells and river levels. The groundwater elevation surface was subsequently subtracted from a digital elevation model (DEM) to build a model of groundwater depth. River level data from 1956 to present from monitoring stations was used to derive a river level model at 50 m increments along the river assuming a constant gradient between the recording stations. Using groundwater data available from a range of sources, a correlation was identified indicating that groundwater is governed by the river levels. The groundwater data from more recent investigations were subsequently adjusted to levels expected during the Edgecumbe earthquake based on the river levels at the time of the earthquake. The standard deviation of the correlation between the river levels and groundwater levels was used to create upper and lower depth to groundwater estimates. Two standard deviations were added for the upper estimate of the depth to groundwater, while two standard deviations were subtracted for the lower estimate of the depth to groundwater. A more comprehensive description of the development of the groundwater models is discussed by Mellsop (2017).
5 LIQUEFACTION ASSESSMENT
The factor of safety against liquefaction was evaluated with depth at each sounding location using the Boulanger & Idriss (2014) liquefaction triggering methodology to evaluate the likelihood of liquefaction. The soil’s fines content (FC) was estimated using the default Boulanger and Idriss (2014) FC-Ic correlation with the CFC fitting parameter set to zero. The cyclic resistance ratio (CRR) curves for a probability of liquefaction (PL) of 15% were adopted for the liquefaction triggering analyses. Soil layers with Ic values greater than 2.6 were considered plastic in behaviour to liquefy (Robertson and Wride, 1998).
Predicted land damage was subsequently calculated for each CPT using the Liquefaction Severity Number (LSN) (van Ballegooy, et al., 2014) and Liquefaction Potential Index (LPI) (Iwasaki, et al., 1984) liquefaction manifestation severity parameters. Previous studies have shown a good correlation between the LSN and LPI with observed liquefaction manifestation (Juang, et al., 2005a; Juang, et al., 2005b; Toprak & Holzer, 2003; Maurer, et al., 2014; van Ballegooy, et al., 2014). These studies generally find that LSNs greater than 16 and LPIs greater than 5 coincide with minor to moderate liquefaction manifestation, while LSNs and LPIs greater than 26 and 15, respectively align with moderate to severe liquefaction manifestation. It is important to acknowledge that the LPI and LSN are not intended to consider lateral spreading, and therefore may not account for the liquefaction severity observed proximal to the Whakatane River. However, the collated reports indicate sand boils formed in the flat land adjacent to the lateral spread sites and thus should correspond with LSNs and LPIs higher than 16 and 5 respectively.
