Liquefaction perspective of soil ageing

Sand liquefaction has become an important topic of geotechnical engineering since the 1960s and significant efforts have been made on the assessment of liquefaction resistance of in situ soils. The current methodology for this assessment commonly relies on in situ measurements such as SPT-N (number of blows for 30-cm penetration in standard penetration tests; Seed & Idriss (1982) and Tatsuoka et al. (1980)).

Case histories indicate that most of the damaging liquefaction disasters occurred in geologically very young, sandy deposits, such as those found in manmade islands, former river channels, filled lakes, swamps and so on (Youd & Perkins, 1978; Wakamatsu, 2000), as well as pipeline backfill. For example, during the 1964 Niigata earthquake, significant liquefaction occurred in the Kawagishi-cho area where the sandy ground had been constructed by pouring dune sand into water in the 1920s. Thus, the age of the liquefied ground was around 40 years at the time of the earthquake. Kobe Port Island, which developed significant liquefaction in 1995, was constructed by dumping gravelly sand into the sea in the late 1960s and 1970s. Thus, the age of liquefied soil was about 30 years or less. During the 2011 gigantic Tohoku earthquake in eastern Japan, liquefaction in the Tokyo Bay area occurred predominantly in recent manmade islands (Towhata et al., 2014). As those islands were constructed after the late 1960s, the age of soil was around 40 years. Fig. 1 illustrates the distribution of liquefaction in Urayasu City on the Tokyo Bay. Here, the land in the liquefied areas was constructed after the 1960s, whereas the ground in the liquefaction-free areas such as ‘Nekozane’ and ‘Todaijima’ had been in place 200 years ago according to Tadataka Inoh's map, which was based on modern survey technology at the beginning of the nineteenth century (Fig. 2). As both the liquefied and unliquefied areas possessed very similar SPT-N values (Fig. 3), it is evident that liquefaction vulnerability assessment can be improved by accounting for the age of the deposit. However, age is not generally considered in current practice, and in situ penetration resistance is solely relied on.

Ageing is the change of material properties which occurs over time. Although geological ageing lasts for thousands to millions of years, the present paper addresses engineering ageing of a shorter timescale, such as a few days to 100 years (Schmertmann, 1991), over which such soil properties as SPT-N and liquefaction resistance have been known to increase.

Youd & Hoose (1977) together with Seed (1979) stated that most liquefaction during earthquakes occurs in ‘young’ subsoil. Youd & Perkins (1978) qualitatively stated that cohesionless sediments younger than 500 years old have higher susceptibility to liquefaction than older soils. Arango et al. (2000) conducted liquefaction tests on undisturbed aged sand to show that the liquefaction resistance of Miocene sand is greater than what is suggested by design formulae developed for Holocene sands. However, most studies on ageing are limited to Pleistocene or older sands, and not much information is available on more recent soils, and particularly on those younger than 1000 years old. Lewis et al. (2008) tested undisturbed samples and demonstrated the higher liquefaction resistance of soils with ages of 5 to 23 million years. For a quantitative study on ageing, Leon et al. (2006) discussed the timescale of thousands of years or more on the basis of a small number of available data. Yoshimi et al. (1984), Goto (1995), Goto et al. (1987, 1992), Hatanaka et al. (1985, 1988), Hori et al. (2002) and Nagase et al. (1999) employed undisturbed frozen samples, whereas Seki et al. (1992) conducted block sampling, to compare the liquefaction resistance of aged soil with laboratory-reconstituted (young) specimens. Mulilis et al. (1977) studied several sites to show the increase of liquefaction resistance over thousands of years. Hayati & Andrus (2009) interpreted many reports on laboratory tests and field interpretations to propose an empirical formula for the effect of ageing. The authors note that, although there have been several US studies on aged soils, these have involved ages greater than 10 000 years (e.g. Heidari & Andrus, 2012) and there are very few data available on soils aged within the engineering timescale of interest in this paper of 100 to 1000 years.

Current practice assesses the liquefaction resistance predominantly by SPT-N values. Hence, if both SPT-N value and liquefaction resistance increase with time at the same rate, there would be no further need to consider the effect of ageing on liquefaction vulnerability. Fig. 3 compares SPT-N profiles at a young liquefied site and an unliquefied site of greater age that was shaken by the 2011 Tohoku earthquake (moment magnitude M_w = 9·0) in Urayasu, Japan. Although both sites have similar SPT-N values in the top 2–3 m of the fine sandy layer, liquefaction damage was found only at the site of the most recent deposits. In line with this, Matsuo (2004) stated that aged alluvial sand is systematically more resistant to liquefaction than recently deposited manmade islands, even if normalised N₁ values are identical. Although Skempton (1986) suggested the increase of SPT-N over 100 years, Mesri et al. (1990), Baxter & Mitchell (2004), Mitchell & Solymer (1984) and Massarsch & Fellenius (2002) stated that the increase of SPT-N or CPT (cone penetration test) resistance is completed within 1 year after land construction or significant compaction. This short-term increase of SPT-N value implies that the ageing of SPT-N cannot account for the observed improvement in liquefaction resistance that occurs over a timescale of hundreds or thousands of years. Roy (2008) mentioned that CPT resistance increases with age more slowly than S wave velocity.

In Fig. 4, the horizontal axis designates the elapsed time after liquefaction and fresh deposition of sand grains, and the vertical axis shows the increment of SPT-N value after fresh deposition. The increment was calculated from the pre-liquefaction SPT-N. In the tests by Kiyota and quoted by Towhata et al. (2012), the pre-liquefaction grain structure was destroyed by liquefaction during the 2011 Tohoku earthquake and grains redeposited to form a new structure. In Kusunoki et al. (2013), blast-induced liquefaction destroyed the original grain structure and a new one was created. In both cases, the penetration resistance was measured by Swedish weight sounding at practically the same places for 900 days and was converted to equivalent SPT-N values using a formula by Inada (1960). Penetration resistance increased for a short period of only 1 year after fresh deposition.

Figure 5 presents schematically the authors' concept of ageing that SPT-N and liquefaction resistance of soil (R) increase with time at different rates. After some elapsed time the real liquefaction resistance becomes substantially greater than what increased SPT-N implies and thus a quantitative study is necessary on the additional liquefaction resistance produced by ageing.

Laboratory tests were conducted by Mulilis et al. (1977) and Tatsuoka et al. (1988) on sand with different consolidation times. Although the employed consolidation time was limited to several months owing to the experimental constraints, the liquefaction resistance increased by, at maximum, 40% or so from that of the freshly deposited samples.

Past studies on the causative mechanisms of engineering ageing are classified into two groups. The first group addresses cementation (e.g. Kokusho et al., 2012). As the bonding mechanism, Mitchell & Solymer (1984) proposed sedimentation of silicate material at granular contacts for which the silicate material is supplied from soil grains. The second idea concerns the change of granular packing induced by weak ground vibrations and so on (Wichtmann et al., 2005; Kiyota et al., 2009; Kondo, 2013; Dobry et al., 2015). The authors suppose that the induced dislocation of sand grains fills big voids, improves the mechanical stability of fabric structures, intergranular locking or the number of grain contacts and increases the liquefaction resistance. So far, no conclusion has been drawn on which mechanism is more predominant.

The present paper aims to shed more light on the ageing of geologically young soils. Although Towhata et al. (2015) briefly addressed the framework of this study, the page limitation did not allow any details to be introduced regarding the hypotheses employed and the idea on the causative mechanism of ageing. In this regard, the present paper provides more details to offer readers a better understanding.

This section describes the authors' quantitative study on ageing that evolves over hundreds of years (briefly reported by Towhata et al. (2014)), focusing on reclaimed land in the Tokyo Bay area which suffered significant liquefaction damage in 2011. The Tokyo Bay area benefits from a comprehensive historical record of land reclamation since the end of the sixteenth century (Fig. 6). Information relating to incidents of liquefaction in this area during the 2011 Tohoku earthquake is available (Fig. 1). Additionally, the ground motion records for the region in 2011 are also available from K-Net stations. The wealth of available information allows the liquefaction resistance of sites within the Tokyo Bay area to be assessed using the Highway bridge design code (JRA, 2012) or the Recommendations for design of building foundations (Tokimatsu & Yoshimi, 1983; AIJ, 2001). For example, the highway bridge design code assesses the effects of SPT-N, the effective vertical stress prior to shaking (σ′_v in kPa) and the fines content (F_c) on the liquefaction resistance of loose sand (R) undergoing strong ground motions that occur at the boundary of tectonic plates using equation (1)

In case of F_c < 10%, C₁= 1 and C₂= 0 when the normalised SPT-N, which is defined by N₁= [170N/(σ′_v (kN/m²) + 70)], is less than 14. Note that σ′_v stands for the effective vertical stress prior to the application of earthquake loads and that C₁ and C₂ increase with F_c. For reference, these two codes are compared with US practice in the Appendix.

The seismic load (L) is assessed using the K-Net surface ground motion records, which allow the factor of safety, F_L= R/L, to be determined and examined against the field observations of liquefaction incidents with reference to the soil age group. This would allow the boundary value of F_L between liquefaction and non-liquefaction to be determined taking into account soil age. Note that the boundary value should be equal to unity, but its real value probably changes with soil age.

The present study encountered many inevitable uncertainties, as described below.

• (a)

  At sites where evidence of liquefaction such as sand ejecta or subsidence of buildings was observed, it is not always possible to identify which soil stratum liquefied unless the F_L method is precise. Because the present study attempts to improve the F_L method by taking into account soil age, F_L cannot be considered precise.

• (b)

  At sites without observable surface manifestation of liquefaction, it is possible that soils at greater depths liquefied without affecting the ground surface.

• (c)

  The majority of K-Net stations are situated on unliquefied sites.

• (d)

  Sites of K-Net stations and those of analysed borehole data are not necessarily identical and the F_L values have to be calculated by using the nearest available K-Net data.

• (e)

  Alluvial deposits are generally assumed to be older than reclaimed manmade islands. However, in the extreme case, it is possible that deposition occurred in the past few hundred years.

• (f)

  Liquefaction can occur in more recent backfill of lifelines within much older natural deposits.

Bearing these uncertainties in mind, the present study employed a conservative approach to avoid significant misinterpretation. Cases of liquefaction of lifeline backfill were removed from the dataset.

Figure 7 illustrates the method employed to reduce the impact of uncertainties (a) and (b). At liquefaction sites, F_L was calculated and plotted in Fig. 7(a) for the entire soil profile that consists of the surface artificial fill and the underlying alluvial sand. In this figure, it is not yet clear which sandy layer liquefied and which did not. However, the surface evidence of liquefaction ensures that liquefaction occurred somewhere in the profile (Fig. 7(a)). Therefore, the boundary value of F_L between liquefaction and non-liquefaction is greater than the minimum value, F_Lmin, in the profile. This is because, if the boundary value were less than F_Lmin, F_L would be greater than the boundary value at all depths and no liquefaction could have occurred in the entire soil profile, which contradicts the observation. On the contrary, at sites without manifestation of liquefaction, the boundary value of F_L is less than the maximum value, F_Lmax (Fig. 7(b)). Otherwise, F_L would be less than the boundary value at all depths and liquefaction could have occurred in the entire soil profile, which is against the observation. Consequently, the range of the boundary F_L is determined as

Four hundred and fifty borehole profiles were collected from many databases in eastern Tokyo and adjacent areas and F_L profiles such as that shown in Fig. 7 were calculated using the Highway bridge design code of Japan. Note that all the employed borehole data were obtained prior to the 2011 earthquake. F_Lmax and F_Lmin were then assessed from the profiles and plotted against the era of land reclamation. The year of land reclamation can only be approximately determined owing to the limited accuracy of the available historical information and because the process of reclamation takes many years. Furthermore, although it is assumed that the alluvial soil is older than manmade ground, this may not always be the case (e). Accordingly, in Fig. 8, F_Lmax is plotted with downward arrows because the boundary value is less than F_Lmax (equation (2)). Conversely, F_Lmin is plotted with upward arrows because the boundary is greater than F_Lmin. Hence, the desired boundary values should lie between the upward and downward arrows.

The shaded area in Fig. 8 indicates the possible range of boundary F_L. It is interesting that the suggested boundary decreases as the soil becomes older, implying that the greater the age of the soil, the less likely it is to liquefy despite F_L being less than unity. As the measured SPT-N values were employed in the calculation, the idea of additional increase of liquefaction resistance (Fig. 5) can be examined whether SPT-N becomes constant after a short time (Fig. 6) or possibly keeps increasing. Also, care must be taken of the very wide range of the shaded area that was caused by uncertainties of (c) and (d). The uncertainty of (e) was solved in the Urayasu City by referring to the old map by Tadataka Inoh as stated previously. Moreover, the boundary value of the youngest soil is greater than unity, which is not consistent with the theory of factor of safety.

In order to improve Fig. 8, three measures were taken, as listed below.

• (a)

  The 2011 Tohoku earthquake had a seismic magnitude of 9·0 and generated an unusually large number of shaking cycles (Fig. 9), which was able to promote the development of excess pore water pressure more profoundly than the 15 to 20 cycles considered in conventional design codes. This situation was exacerbated by a strong aftershock of magnitude (M_w) = 7·7 that occurred 30 min after the main shock. Thus, F_L was reduced by 20%, by which both the significant number of cycles in M = 9·0 main event and that of the smaller aftershock are considered. Yasuda & Ishikawa (2012) proposed a reduction of 18%. Note that the Highway bridge design code does not require the earthquake magnitude for input, whereas the building foundation code of Japan does.

• (b)

  Noting that F_L in Fig. 8 was calculated using the biggest value of the east–west and north–south peak accelerations, the effect of the bi-directional shaking, $EW2+NS2$⁠, was absent. In order to account for the reduction in liquefaction resistance due to superposition of the components of strong motion (Seed et al., 1978), the calculated F_L value was reduced by10%; Ishihara & Yamazaki (1980).

• (c)

  The acceleration records employed in Fig. 8 were obtained at K-Net sites where liquefaction did not occur (uncertainty (c)). Because of the softer soil, the acceleration at liquefied sites was most likely greater than the K-Net records (amplification by soft soil). For example, Fig. 10 compares the K-Net acceleration in Urayasu and those obtained in the reclaimed and liquefaction-prone land of the same city, all at similar epicentral distances. Moreover, data at the K-Net Inagi site, which liquefied, are added. Peak accelerations at sites of (significant) liquefaction were greater than that at the unliquefied K-Net site (see 1:1 line). The range of data variation is substantial owing to many underlying uncertainties. With some conservatism, the maximum acceleration at the K-Net site is multiplied by 4/3 to assess the near-upper-bound values of accelerations in liquefaction-prone sites. Thus, the factor of safety, F_L, was reduced to 3/4 at sites of liquefaction. In summary, the previously calculated F_L values were modified by multiplying by 0·8 × 0·9 × (3/4) or 0·8 × 0·9, depending upon liquefaction or no liquefaction at respective sites.

The modified result (Fig. 11) clearly demonstrates a reduction of the range of uncertainty in the boundary F_L values compared to Fig. 8 and younger deposits have the boundary values of unity, which is consistent with design practice. The new data from Amagasaki and Kyo-ga-shima and other sites in Kobe will be discussed in the later sections. It is noteworthy that more aged soils were unlikely to liquefy in spite of the small values of F_L. This implies that the true resistance of older soils is greater than that assumed by the current design codes without considering soil age.

The data from the 2011 Tohoku earthquake were limited to the Tokyo Bay area where the age of liquefied sites was, at maximum, 100 years or so. Therefore, data on older liquefied sites were sought. The requirements for reliable data are

• (a)

  evidence of the onset of or lack of liquefaction

• (b)

  availability of borehole data

• (c)

  availability of earthquake records at or near the studied borehole sites

• (d)

  known age of subsoil.

These requirements were satisfied by the Tsukiji area in Amagasaki City and Kyo-ga-shima Island, Kobe.

The Tsukiji area in Amagasaki City (Fig. 12) was constructed in shallow sea in front of the Amagasaki Castle at the beginning of the seventeenth century and suffered significant liquefaction during the 1995 Kobe earthquake of M_w = 6·9. The land reclamation started in 1617 and was completed by the middle of the seventeenth century. Hence, the age of the reclaimed subsoil at Kobe earthquake was approximately 400 years.

A typical borehole profile in Tsukiji is illustrated in Fig. 13. The earthquake motion record was obtained by the Ministry of Transport in the Amagasaki Harbour and its peak acceleration value was input to the liquefaction analysis. The obtained F_Lmin was 0·147. Only (b) and (c) measures were employed to improve the calculated F_L value, while (a) was not, as the number of cycles generated by the Kobe earthquake was within the range of consideration of the present design codes. Consequently, F_Lmin = 0·099 was plotted by a solid dot as an older soil in Fig. 11.

Kyo-ga-shima Island was constructed near the end of the twelfth century in the present Kobe Harbour by Taira-no-Kiyomori, who intended to construct an island in front of his harbour (Hyogo in Fig. 12) to protect international trading ships from southeast winds and waves. A classic descriptive poem ‘Heike Monogatari’ states that the construction started in 1161 and that the area of the island was 37 ha. Later, the knowledge of the exact location of this island was lost and the island is believed to be buried under more recent land reclamation activities. The most likely location of the ancient Kyo-ga-shima Island is the Hyogo area of Kobe (Fig. 14), which is the most traditional part of the Kobe Harbour and contains a number of historical cultural structures, such as old temples and statues. Strong ground motion of peak ground acceleration, PGA = 616 cm/s² in the east–west direction was recorded at the nearby Takatori railway station, where no liquefaction was observed. Combined extensive borehole records for the area and good records of liquefaction locations satisfy data requirements (a) to (c).

The area in Fig. 14 is situated to the west of Hyogo Wharf, which was the main part of the Kobe Harbour until the nineteenth century. The authors interpreted a geotechnical borehole database as published by Kansai Geoinformatics Agency in order to identify precisely the location of Kyo-ga-shima. The island was constructed by transporting soils from local mountains into the sea. Because the local Rokko mountains are made of granite, the location of the missing island can be identified by a deposit of gravelly–sandy materials lying above former seabed materials such as fine sand and silt.

After detailed investigations, cross-sections (Figs 14 and 15) pointed to the potential location of Kyo-ga-shima. Fig. 15 shows that a gravelly–sandy deposit lies on a layer of finer sand and silt. The separation of the gravelly mass from the underlying natural deposits of sand and silt suggests that it is an artificial deposit in a formerly inundated area. The convex shape of the sand–silt layer suggests consolidation settlement of the original seabed under the weight of the infill. Fig. 16 indicates the distribution of different soils at −4 m below the sea level. A gravelly–sandy deposit at the centre shows the probable size of the ancient island, and the sand–silt deposit to its south suggests the entrance channel of the harbour. These findings mean that the location of the Kyo-ga-shima was most likely detected by the present study.

No liquefaction occurred in the ancient Kyo-ga-shima area (Wakamatsu, 2011). Thus, hollow circles (F_Lmax) were drawn in Fig. 11 together with downward arrows. The studies at other sites in Kobe provided data on liquefied manmade islands (Karumo Island in Fig. 12 and others) that were constructed in the first half of the twentieth century and were nearly 100 years old at the time of the 1995 Kobe earthquake.

Consequently, Fig. 11 presents all data and evidently demonstrates that the boundary F_L value decreases with increasing age of soil. Thus, the older soil is unlikely to liquefy even though the calculated factor of safety, F_L, is less than unity, and ageing produces additional resistance against liquefaction. But it is noteworthy that this ageing effect is not considered in the current SPT-based design codes and that the liquefaction resistance of aged soil is underestimated. Although Fig. 4 might suggest minor increase in SPT-N for hundreds of years, this increase is not significant enough to account for ageing in liquefaction resistance. This is because the SPT-based F_L in aged soil is still less than unity, in spite of lack of liquefaction. Similar ageing was found from the Recommendations for design of building foundations (AIJ, 2001) as well (Fig. 17).

The boundary F_L values are able to quantitatively reveal the increase of liquefaction resistance with soil age. For example, if the boundary value is 0·5, the liquefaction resistance is 2 times greater than the SPT-based design code calculation. By using the upper and lower bounds of the shaded area in Fig. 11, the ageing effect was quantitatively drawn in Fig. 18. The derived age-induced increase is consistent with the results of former studies obtained by undisturbed samples or other case history studies. Because the current design code calculation takes into account only the limited effect of temporal increase of SPT-N, the present study indicates the ratio of

whereas other studies on ageing consider the combined effects of increased SPT-N and ageing, except field tests by Hayati & Andrus (2009) and Lewis et al. (2008) in Fig. 18. Note that studies by other researchers indicate an increase of more than 40% after 50 years or so, although the present study suggests a greater increase after 1000 years. Accordingly, it is possible herein to state with sufficient conservatism that the ageing becomes prominent after 100 years and that at least a 40% increase can be expected after 400 years.

To shed light on the causative mechanism of ageing, laboratory tests were conducted in which dislocation of sand grains under sustained pressure was observed microscopically (Shintaku & Towhata, 2013). Although grain shape may be important and mineralogy may affect the development of cementation, the present paper focuses on a hypothesis that the dislocated sand grains generate more grain-to-grain contact points or fill big voids to create more interlocking and more uniform stress distribution so that the mechanical properties of sand as a mass may be improved (Fig. 19). The dislocation of one particle triggers that of the next grain and so on. This procedure occurs rather slowly and hence the improvement of the mechanical properties of sandy ground takes as long as hundreds of years.

Dry sand with different non-plastic fines content was placed by air pluviation in a soil tank that measured 6 cm × 15cm × 20 cm. Then the sand was compressed under 25 kPa. Photographs were taken of the vertical cross-section of the sand deposit through a transparent side wall of the tank by using a microscopic camera every day and compared with each other to see whether or not there was any movement of grains. The extent of dislocation was expressed by the number of moved particles per square millimetre.

Three kinds of silica sand (G_s = 2·65) were used and grain size distribution was controlled by mixing several kinds of silica sands; see Fig. 20 and Table 1. These sands are non-plastic and have nearly identical D₅₀ size, while the uniformity coefficient (U_c) is different. Because the fines content has no cohesion, the present study addresses the effect of non-plastic fines.

Figure 21 shows two pictures of grains of sand S1 (U_c = 4·0) on the third and fourth days of the experiment. The two photographs were compared to find out the change of grain locations. It can be seen that the grain in the circle changed its position, showing the occurrence of dislocation. The dislocation continued for weeks but gradually calmed down.

Figure 22 indicates the temporal increase of the number of dislocated grains for different relative densities (D_r) and uniformity coefficients (U_c). Because the initial packing immediately after sample preparation was unstable, the first 3 days were excluded from the study. When U_c is constant, looser sand exhibits more particle dislocations than denser sand because the grain structure of loose sand is less stable. On the other hand, the sand with greater U_c developed more significant dislocations. This is consistent with the general perception that sands with more fines achieve more significant ageing than uniformly grained sand (Wakamatsu et al., 1991; Sawada et al., 2005).

Cyclic undrained triaxial tests were conducted on the same sands in order to show the relationship between the fines content, the rate of dislocation and the ageing of liquefaction resistance. All the specimens with diameter of 5 cm and 10 cm high were prepared by air pluviation in a rigid cylinder and then were subjected to one-dimensional consolidation under 25 kPa of overburden. The duration of consolidation varied from 1 to 64 days. After consolidation, specimens were set in a triaxial device to be consolidated isotropically under 25 kPa. Then, cyclic undrained tests were performed.

Figure 23 compares the liquefaction resistance of tested specimens. Tests on S1 sand failed and are not presented here. While the sand with more non-plastic fines (U_c = 10) is weaker against liquefaction (greater uniformity coefficient in Fig. 23), as has been reported elsewhere (Towhata et al. (2013), among others), this sand increased liquefaction resistance with time more rapidly than sand with less fines (U_c = 1·5). This finding is consistent with Fig. 22, where more dislocation occurred in sand with higher uniformity coefficient. Thus, the hypothesis in Fig. 19 is promising. As dislocation occurs slowly with time, it is a microscopic perception of soil creep. Finally, the higher rate of dislocation for greater fines content implies that soils with greater fines content can develop higher liquefaction resistance as current design codes specify.

Earthquake damage histories indicate that subsoil liquefaction has occurred in young sandy deposits. Because more aged sand did not liquefy, it is possible that ageing increases the liquefaction resistance of loose sand over hundreds of years. To quantitatively evaluate the ageing, the behaviour of sandy subsoil during past earthquakes was investigated. Consequently, the following conclusions were drawn.

• (a)

  The 2011 Tohoku earthquake revealed that liquefaction occurs more profoundly in younger subsoil.

• (b)

  Borehole data and the intensity of seismic shaking were analysed to calculate the factor of safety against liquefaction, F_L, for different age groups. The assessed F_L at liquefied and unliquefied sites were compared against the age of soil to show that more aged sand did not liquefy in spite of F_L less than unity.

• (c)

  This suggests that more aged sand has additional liquefaction resistance, which is called the ageing effect.

• (d)

  For practice, it is proposed that soil older than 400 years has 40% greater resistance against liquefaction than what SPT-N implies in the current seismic design codes. Note that this proposal is of reasonable conservatism.

• (e)

  Laboratory tests were conducted to verify a hypothetical mechanism and results indicate that sand grains under static load dislocate with time and fill the larger voids. As a result, big voids, which are mechanical weakness of grain structure, disappear and a more compacted, stable, granular structure is developed.

The present study was supported financially by Tokyo Gas Company. Important subsoil data were provided by the Geotechnical Database for Kanto Region by JGS, Geotechnical Database by the Ministry of Land, Infrastructure, Transport and Tourism, Kansai Geo-informatics Database, Geoenvironmental Information Bank of Chiba and Dr Seiji Suwa. The intensity of ground motion was obtained from K-Net of the National Research Institute for Earth Science and Disaster Prevention. Dr Way Way Sim of Imperial College London kindly assisted with editing the text. This various support and assistance is greatly appreciated by the authors.

Liquefaction risk assessed by the two Japanese codes used in this paper, which are the Highway bridge design code (JRA, 2012) and Recommendations for design of building foundations (AIJ, 2001), are compared with recent US practice (Idriss & Boulanger, 2006). The assumed soil condition is made of clean sand (fines content = 0%) and the depth of interest, z, is either 4 m or 10 m, while the groundwater table is located at 2 m depth; see Fig. 24. The design earthquake has the maximum acceleration of 0·4g and its magnitude is 8·0. The energy efficiency of Japanese SPT practice was set equal to 70% by referring to Seed et al. (1985). By varying SPT-N, the change of F_L was plotted in Fig. 24 where three examined codes are consistent with one another, irrespective of the effective stress level, except for the range of SPT-N > 17 at shallow depth.

C₁, C₂

parameters for fines-content effect

D_r

relative density

D₅₀

mean grain size of sand

F_c

fines content

F_L

factor of safety of soil against liquefaction

F_L,max

maximum value of F_L in soil profile at sites without liquefaction

F_L,min

minimum value of F_L in soil profile at sites of liquefaction

g

gravitational acceleration

L

seismic load

M_w

moment magnitude of earthquake

N

SPT blow counts

R

liquefaction resistance of sand

U_c

uniformity coefficient of grain distribution of sand

σ′_v

effective vertical stress prior to earthquake shaking