Introduction

The 1999 Kocaeli earthquake in Turkey (M w  = 7.4) resulted in significant loss of life and structural damage. It also triggered liquefaction in numerous inland alluvial areas as well as along the coast. Sand boiling, serious lateral spreading and settlement were widely observed in the city of Adapazari and along the southern coasts of Sapanaca Lake and Izmit Bay in the eastern and western parts of the earthquake-affected region respectively (Fig. 1). The large-scale liquefaction and associated ground failures in the eastern Marmara Region due to this earthquake attracted the attention of researchers both from Turkey and other countries and the occurrence and consequences of liquefaction and lateral spreading were documented by a number of reconnaissance teams (e.g. Kasapoglu et al. 1999; Hamada et al. 1999; Aydan et al. 2000; MCEER 2000; EDM 2000; Sucuoglu et al. 2000).

Fig. 1
figure 1

Location of the main ground deformations related to the different type of failures, which occurred during the Kocaeli earthquake and measured magnitudes of displacements (Aydan and Ulusay 2000)

Full size image

Most of the investigations into the liquefaction concentrated on the city of Adapazari (e.g. Bray et al. 2001, 2004; Erken 2001; Yasuda et al. 2001; Sancio et al. 2002) and along the southern coast of Sapanca Lake (Cetin et al. 2002; Inuzuka 2002; Kanibir 2003; Aydan et al. 2004, 2005; Kanibir et al. 2006a, b), while investigations into the associated ground failures and inundations were generally carried out at local sites along the southern shore of Izmit Bay (Fig. 1). Arel and Kiper (2000) indicated that high ground acceleration due to faulting near Gölcük town and the presence of highly liquefiable layers on the coast was responsible for the occurrence of the submarine landslide at Degirmendere Peninsula (Fig. 1). Aydan and Ulusay (2000), who investigated the type, distribution, direction and magnitude of the liquefaction-induced ground failures along the southern coasts of Izmit Bay and Sapanca Lake, reported that the maximum ground displacements reached up to 5 m. According to Yasuda et al. (2001), most of the coastal problems were related to the steep slopes of river fans. They also indicated that the fans on the southern shores of Izmit Bay slid due to loss of support following submarine disturbance. Towhata et al. (2001) suggested that pull-apart action along the fault line is the most likely mechanism for these movements, although the steep angle of the recent deposits was also a factor.

The results from the study by Ratjhe et al. (2004) indicated that both liquefaction and tectonic subsidence contributed to the failures and sea inundation within the pull-apart basins, and major liquefaction sites are situated at the prograding nose of active delta fans. They also emphasized that the identification of regions susceptible to both tectonic subsidence and liquefaction are important when evaluating seismic hazards. Cetin et al. (2004a) studied the ground failure at Degirmendere Nose during the Kocaeli earthquake. Based on their analyses, they concluded that liquefaction and earthquake-induced lateral forces could have triggered the coastal landslide at this location. Cetin et al. (2004b) analyzed the magnitude of expected lateral displacements at some locations in the same region using a semi-empirical and two empirical models and noted they resulted in very different predictions of the lateral ground displacements.

Many of the residential areas and industrial facilities located along the coastal zone between the shore and hilly grounds to the south of Izmit Bay suffered serious damage during the Kocaeli earthquake. This study was undertaken to assess the liquefaction potential of the southern part of Izmit Bay and to establish a liquefaction-based microzonation map, which could be compared with the sites where liquefaction and associated ground deformations were observed after the earthquake. For this purpose, a zone extending approximately E–W and bounded by the shoreline in the north and the junction of Quaternary deposits and exposed bedrock in the south was chosen. The area was selected because there was evidence of liquefaction and associated ground failure, both reported and still visible, and a considerable amount of data from geotechnical boreholes drilled at the site before and after the 1999 Kocaeli earthquake was available.

Data obtained from reconnaissance fieldwork and geotechnical information from the local municipalities were then evaluated. The SPT-based method was used to calculate the factor of safety against liquefaction and the calculation of liquefaction severity indices, from which a liquefaction potential map was prepared. This map was compared with the field observations carried out by the second author and other investigators immediately following the 1999 Kocaeli earthquake.

Liquefaction damage along the southern shores of Izmit Bay

Liquefaction and lateral spreading generally occurred in open areas along the southern shores of the bay, with significant damage to the industrial port facilities, particularly from about Gölcük to the east (Fig. 1). In addition, as a consequence of the coastal failures, some buildings were moved laterally into the bay. Sand boils were observed particularly in the vicinity of the inundated eastern portion of Gölcük town (Figs. 2 and 3a). As reported by several reconnaissance teams (e.g. Kasapoglu et al. 1999; Hamada et al. 1999; Aydan et al. 2000; MCEER 2000; EDM 2000; Sucuoglu et al. 2000), fine sand was ejected due to liquefaction between Degirmendere (west of Gölcük) and the Ford Otosan Factory, which was under construction at the time of the Kocaeli earthquake. The largest coastal failure occurred at Degirmendere Nose, where buildings were carried out in to the bay (Figs. 2 and 3b).

Fig. 2
figure 2

Locations of observed liquefaction, lateral spreading and subsidence sites along the southeastern coast of Izmit Bay (compiled from http://www.ce.utexas.edu/prof/rathje/research/turkey.html and based on the observations by the second author following the 1999 earthquake)

Full size image
Fig. 3
figure 3

a Liquefaction of the ground in the inundated Kavakli district of Gölcük town (Aydan et al. 2000); b area of slope failure at Degirmendere (Cetin et al. 2004b); c view from the Gölcük normal fault passing near Gölcük football stadium (Aydan et al. 2000); d view of the inundation at Yenikoy; e submerged quay at MKE’s Ship dismantling plant, Seymen; f lateral spreading of a park at Seymen (Aydan et al. 2000); g lateral spreading at a soccer field (http://www.peer.berkeley.edu/turkey); h lateral spreading at the eastern end of Izmit Bay, Basiskele (Aydan and Ulusay 2000)

Full size image

The analyses carried out by Ulusay et al. (2002), Rathje et al. (2004) and Cetin et al. (2004b) indicated that in addition to seismically induced inertial forces, the liquefaction of deeper soils is likely to be partly responsible for this landslide. The main cause of the ground subsidence and coastal failures between Gölcük and Ford Otosan Factory was a result of a secondary normal faulting associated with the main lateral strike-slip event during the earthquake (Fig. 3c; Aydan et al. 2000; Rathje et al. 2004). Indeed, it was concluded that regional subsidence at the western part of the investigated area was more significant than the liquefaction/lateral spreading per se.

A zone of liquefaction-induced lateral spreading and resultant cracking was evident along the shoreline between Gölcük and the eastern end of Izmit Bay, with traces of the ground deformations still visible some 6 years after the earthquake (Fig. 3d). A typical example of lateral spreading is shown in Fig. 3e where the land and the quay at the MKE Ship Dismounting Plant were shifted towards the sea; permanent horizontal ground displacements measured along two parallel lines were 4.77 and 4.37 m; and a concrete cabin and electric pylon were tilted due to lateral spreading (Aydan and Ulusay 2000). To the west, at Yeniköy, a park was also subjected to ground deformations due to lateral spreading. Based on the measurements of surface cracks along two scan-lines by Aydan and Ulusay (2000) at this park, the permanent displacements were 3.25 and 1.03 m horizontally and 4.09 and 0.28 m vertically (Fig. 3f). In this locality, the direction of movement towards the bay was NE. Lateral spreading was also observed in the soccer field 9 km east of Gölcük town (Fig. 1) where some 1.2 m of displacement was measured, while the lateral and horizontal displacements at Basiskele in the south east corner of the bay (Fig. 3g; Cetin et al. 2004b) were 3.29 and 0.43 m, respectively. In addition, sand boils were clearly visible in the shallow waters just offshore.

Geological setting

Izmit Bay is located in an E–W trending pull-apart basin bounded by two horsts—the Kocaeli Peninsula to the north and the Armutlu Peninsula to the south. The study area is situated at the northern part of the Armutlu Peninsula where the natural slope angles are between 0.2% (1:500) and 4% (1:25) along the shore. As seen from the geological map in Fig. 4 three different geological units have been identified in this area (MTA 2003). The Pamukova metamorphics (schists, marbles and metabasic rocks) form the basement rocks and are present in the hills to the southwest. They are overlain by the Pliocene clastics resembling molasse sedimentation, locally affected by neotectonic events (Goncuoglu et al. 1986). They are generally composed of sandstone, poorly bedded conglomerate, mudstone and marl alternations. The youngest units are the Quaternary alluvial and deltaic deposits (Seymen 1995), which are present along the shoreline and in the V-shaped delta fans. They are composed of loose and saturated sediments with varying grain sizes. Borehole data indicate sand and silty sand layers highly susceptible to liquefaction are present in these alluvial sequences. The depth to groundwater generally ranges between 0.5 and 6.5 m; the mean groundwater depth near the shoreline is about 2 m.

Fig. 4
figure 4

Geological map of the investigated area (Q: Quaternary alluvial delta fan, flood plain and fluvial deposits; Pl3: Pliocene conglomerates, sandstone, mudstones and marls; TRj: Palaeozoic schists, marbles and metabasics; MTA 2003)

Full size image

The Kocaeli earthquake occurred due to movement on the northern segments of the strike-slip North Anatolian Fault Zone (NAFZ; the Sapanca segment on the east of Izmit Bay and the Gölcük and Yalova segments running parallel to the shore of the bay; Fig. 5). The interaction between the Gölcük and Sapanca segments resulted in development of the pull-apart Izmit Basin. A normal secondary fault, referred to as the Gölcük Fault, is present between the Kavaklı district of Gölcük town and Hisareyn (Figs. 2 and 5). Some 2 m of vertical movement was measured on this fault during the Kocaeli earthquake. The strike of the fault was 45–60 NW (Aydan et al. 2000).

Fig. 5
figure 5

Main segments of the North Anatolian Fault Zone (NAFZ) in the vicinity of Izmit Bay (Lettis et al. 2000)

Full size image

Main characteristics of the Kocaeli earthquake

The Kocaeli earthquake, with a magnitude of 7.4 (M w ), occurred on 17 August 1999. It resulted in a death toll of more than 20,000 as well as heavy damage, particularly affecting the eastern Marmara Region. The epicentre of the earthquake was at Gölcük, with an estimated depth of 15 km. A number of aftershocks with a maximum magnitude of 5.8 also occurred. The earthquake resulted in approximately 120 km of surface rupture and right lateral offsets reaching up to 4.5 m.

During the earthquake, the largest peak horizontal ground acceleration (PHGA) was recorded at Sakarya (0.407 g) in the city of Adapazari, east of Kocaeli. The locations of the strong ground motion stations and the maximum PHGA values recorded are shown in Fig. 6. The closest station to the study site is the Yarimca station (YPT) where the maximum PHGA recorded was 0.3 g. This station, some 4 km from the study site, is founded on soft alluvial soils similar to those in the investigated area.

Fig. 6
figure 6

Strong ground motion stations in the vicinity of Izmit Bay and maximum PHGA values recorded at these stations during the Kocaeli earthquake (PHGA values from GDDA 2000, Safak and Erdik 2000 and Celebi et al. 2000)

Full size image

Assessment of liquefaction potential

Site and data selection

The principal factor governing the selection of the site between Degirmendere and Basiskele along the south eastern shore of Izmit Bay due to the fact that the liquefaction phenomena and associated ground deformations have been commonly observed along this coast. The second factor was the availability of high quality field reconnaissance data obtained by several investigators (including the second author) within a few weeks of the earthquake. In addition, records of some 345 relevant geotechnical boreholes were available, drilled for different purposes by private companies (Tekar 1994; Belirti 2000; Arel and Kiper 2000; Kasktas 2000; Geo 2000; STFA 2000; ABM 2000, 2001; http://www.peer.berkeley.edu/turkey/adapazari/phase4). Laboratory test results and SPTs were also obtained from various local municipalities and from the Iller Bank of Turkey.

Experience has shown that liquefiable layers are generally found at shallows depths and 15 m is generally adequate for a realistic assessment. The boreholes were therefore classified into three groups: <10, 10–15 and >15 m in depth. Those extending into the very stiff Pliocene units (not susceptible to liquefaction) were ignored, as were some drilled in the Quaternary deposits, which had insufficient data. As a consequence, the total number of the boreholes considered decreased from 345 to 135 (Fig. 7). The database included depths of SPTs and groundwater table together with laboratory test results (e.g. particle size analysis, Atterberg limits).

Fig. 7
figure 7

Liquefaction severity map of the investigated area

Full size image

Evaluation of liquefaction potential

In this study, the simplified procedure proposed by Seed and Idriss (1971) and Seed et al. (1985) was employed to evaluate the liquefaction resistance (CRR) based on the SPT data. Modifications suggested by Youd et al. (2001) were also taken into consideration. Normalized SPT-N values were calculated using the correction factor (C N) proposed by Liao and Whitman (1986); borehole correction factors were selected from the list given by Youd et al. (2001). The 7.4 (M W ) magnitude measured for the Kocaeli earthquake was assumed for all the analyses and the PHGA of 0.3 g was recorded at Yarimca, the closest station to the site and also situated on alluvium. Cetin et al. (2004a) using one-dimensional equivalent linear site response studies concluded 0.3–0.4 g would be realistic for the study area. As a consequence, 0.35 g was assumed. The procedure mentioned above was used to calculate the liquefaction resistance and factor of safety against liquefaction (F L) at the sampled depths of each borehole.

The areas where sand boils, settlement and lateral spreading was found were classified as liquefied sites and gave calculated values of F L less than unity. It is of note that the thickness of the cap soil at these sites is between 0 and 3 m. The fact that F L values of less than or only slightly greater than one were obtained for some other sites where the effects of liquefaction were not clearly observed is probably due to the presence of a thicker cap soil.

The deterministic method may not be the best choice to judge whether liquefaction occurred in a post earthquake investigation as an unknown degree of conservatism is generally built into it (Yuan et al. 2003). The liquefaction potential as defined by Seed and Idriss (1971) and Youd et al. (2001) is evaluated for a soil at a particular depth. However, the severity of ground damage cannot be evaluated solely by the factor of safety (F L). In order to quantify the severity of liquefaction Iwasaki et al. (1982) suggested the liquefaction potential index (I L). The I L which employs F L, is an integrated effect of the likely liquefaction over the entire depth of the profile and hence could be a base for mapping the liquefaction potential of an area. Based on I L, four categories of liquefaction potential (very low, low, high and very high) were proposed by its originators. However, the categories of “non-susceptible” and “moderate” are lacking hence Sonmez (2003) modified the classification to include these two categories. This author suggested that when F L > 1.2 throughout the soil column to a depth of 20 m, the column is classified as “non-liquefiable”, but this threshold value is open to discussion. Taking into account that the probability of liquefaction decreases as F L increases, Sonmez and Gokceoglu (2005) concluded that the use of a probability-based impact factor for the F L would be more realistic and suggested a new “liquefaction severity index” (L S) calculated by the following equations.

$$ L_{{\text{s}}} = {\int\limits_0^{20} {P_{{\text{L}}} (z)W(z){\text{d}}z} } $$
$$ P_{{\text{L}}} \, = \,\frac{1} {{1 + {\left( {\raise0.5ex\hbox{$\scriptstyle {F_{{\text{L}}} }$} \kern-0.1em/\kern-0.15em \lower0.25ex\hbox{$\scriptstyle {0.96}$}} \right)}^{{4.5}} }}\quad {\text{for}}\;F_{{\text{L}}} \le \,1.411 $$
$$ P_{{\text{L}}} (z) = 0\quad {\text{for}}\;F_{{\text{L}}} > 1.411 $$
$$ W\,(z) = 10 - 0.5z\quad z < 20\,{\text{m}} $$
$$ W\,(z) = 0\quad z < 20\,{\text{m}} $$

where z is the depth of the mid point of the soil layer in metres and F L is the factor of safety against liquefaction. Boundary values of L S and corresponding liquefaction severity classes are given in Table 1.

Table 1 Liquefaction severity classification (Sonmez and Gokceoglu 2005)
Full size table

In this study, an attempt was made to establish a microzonation map for the study area based on the liquefaction severity index. The F L values calculated for different borehole locations were used to estimate L S and the liquefaction severity classes were determined from Table 1. As in some areas the number of boreholes was limited, if the observations of reconnaissance teams immediately following the Kocaeli earthquake suggested liquefaction, triangulation was undertaken. In addition, areas where the F L values suggest no liquefaction and the number of boreholes was limited were categorized as “non-liquefiable” and their boundaries shown by dashed lines.

As seen in Fig. 7, the study area was divided into five zones from non-liquefied to very high liquefaction. A comparison of Figs. 2 and 7 suggests that there is a general agreement between the predicted zones and the sites observed after the earthquake. Particularly, Degirmendere Nose, the shoreline of the Kavaklı district of Gölcük, the Ford Otosan Factory site and the eastern parts of these settlements where sand boils, lateral spreading, etc. were observed, fall into the very high to moderate liquefaction severity classes. Towards the south, the liquefaction severity of the soil layers becomes very low (Fig. 7), probably due to the presence of finer grained soils and thicker cap soils.

Conclusion

Based on the available geotechnical data and information from the site observations following the 1999 Kocaeli earthquake, the liquefaction potential of the southeastern part of Izmit Bay was evaluated. The liquefaction severity map established was compared to the liquefaction-affected areas. The main conclusions drawn are:

  1. 1.

    Liquefaction along the southeastern part of Izmit Bay appears to have occurred primarily within the Quaternary deposits at shallow depth. The major regions of liquefaction and associated ground deformations are particularly located along the creeks and shorelines and associated with delta fans.

  2. 2.

    The recent rapid deposition of sediments and very shallow groundwater table, particularly near the shoreline, contributed to the creation of conditions favorable to liquefaction. However, regional subsidence was more important than liquefaction-induced lateral spreading.

  3. 3.

    A liquefaction severity map for the southern part of Izmit Bay was prepared. The zones representing different degrees of liquefaction severity show general agreement with the observations carried out after the 1999 event. Towards the south, the liquefaction severity of the soil layers becomes very low, probably due to the presence of finer grained soils and thicker cap soils.

This study from an earthquake-affected area clearly indicates that the role of geo-environmental information is becoming increasingly important for urban planning. When citing settlements in coastal areas, it is important to consider ground failures due to liquefaction-induced lateral spreading and to establish microzonation maps showing the extent of such liquefaction-susceptible areas. A lack of engineering geological and geotechnical information which would allow zoning areas for different purposes appears to be the main problem in establishing an equilibrium between urban development and protection against such geohazards.