1. Introduction
The prevention of utility pipe damage caused by earthquakes is significant for sustaining urban life, and experimental and numerical analysis-based studies on this subject are in progress worldwide and in our country. Considering that the damage that may occur in drinking water, stormwater, and sewer lines during earthquakes can negatively impact life, it is believed that pipe damage can be prevented by taking certain measures in the design process.
Numerous experimental studies and numerical analyses have provided critical data regarding the uplift potential of buried structures during significant earthquakes. Studies conducted by Yang [
1] on the George Massey Tunnel and Sun [
2] on the BART Tunnel have revealed significant uplift events. These findings have also been supported by centrifuge experiments conducted by Adalier [
3] and Chou [
4]. In these experiments, it has been observed that the presence of liquefiable sandy soil beneath buried structures significantly affects the vertical displacement of the structure. It has been determined that the uplift behavior is influenced by the intensity of earthquake shaking and the increase in pore water pressure. Sasaki [
5] observed significant uplift displacements of pipes in centrifuge tests under high input accelerations or with low sand densities. Since the pipe uplift mechanism has not been fully understood, various methods are used to determine the uplift forces acting on pipelines during soil liquefaction. In general, this force is calculated using the formula
FBuoyancy = ρsatgV. Here, saturated soil is considered as a material that has been fluidized whose unit weight is equivalent to its saturated unit weight [
6,
7]. It should be noted that the buoyancy force is also estimated in terms of the excess pore water pressure [
8].
The safety factor against flotation proposed by Koseki [
9] has been developed and verified through centrifuge tests for both pipes [
10] and manholes [
11]. The specified safety factor procedures only indicate the uplift triggering condition and do not accurately predict the final uplift displacements of underground structures [
11].
Numerical analyses have been carried out to investigate the dynamic responses of soil and pipes up to the liquefaction stage [
12]. In the past fifteen years, the behavior of underground structures (tunnels and pipes) under dynamic conditions has been investigated through numerical analyses [
2,
13,
14,
15,
16,
17,
18,
19,
20] and physical model experiments [
3,
4,
21,
22]. Azadi and Hosseini [
13] have conducted a numerical study using 2D finite difference software FLAC (2008) to investigate the effects of tunnel uplift due to soil liquefaction, evaluating pore pressure variations during earthquakes, considering tunnel diameters, burial depths, and soil stiffness and strength parameters.
Based on a series of model experiments, Koseki [
9] has identified three main reasons for the uplift of underground structures exposed to liquefaction: lateral deformation of the soil beneath the structure, followed by the movement of pore water towards the base of the structure, and finally the reconsolidation of the liquefied soil. Numerical and experimental studies have been conducted to understand better the uplift response of underground structures with circular cross-sections buried in homogeneous liquefiable soil [
22]. In these analyses, the deformation patterns occurring in the ground surrounding the structure are similar to those observed by Koseki [
9]. Chian [
22] found in numerical analyses concerning the performance of underground sewer pipes near Tokyo in Urayasu City following the 2011 Great East Japan Earthquake that significant pipe uplift displacements occurred in liquefiable soil, which was consistent with observed pipe damage in the field. Sudevan [
23] have conducted numerical analyses to investigate the uplift of an underground pipe structure due to liquefaction. The liquefiable soil has been modeled using the elastic, perfectly plastic Mohr–Coulomb model, incorporating Finn Byrne’s pore water pressure formula. In the study by Huang [
24], the uplift of shallow buried pipelines subjected to seismic vibrations in liquefied areas has been investigated using dynamic centrifuge tests. Shafari et al. [
25] used the Finn Model with the FLAC2D software to study the uplift behavior for Nevada sand parameters at different relative densities, various diameters, and burial depths. It was shown for the various depths that the amount of uplift increases as the diameter ratio increases, and decreases as the relative density and burial depth increase.
Damage occurring on critical pipelines during an earthquake can potentially lead to delays in critical services, infrastructure disruptions, explosions, structural collapses, environmental pollution, and economic losses. The safe operation of natural gas pipelines is of great importance in meeting the energy demands of modern societies, but geological hazards pose significant challenges to their integrity
Most seismic disaster studies have shown that the likelihood of damage to buried pipes in liquefied soil is higher than in non-liquefied soil [
26,
27]. In liquefied sands, shallowly buried pipes may exhibit bending deformation, creating an upward buoyancy behavior, which increases the degree of damage. During the 1964 Niigata earthquake, 68% of a 470 km-long pipeline was destroyed. Most of the deformations observed were due to pipe uplifting behavior. Similar seismic damage to pipelines, characterized by pipe uplifting behavior, has been observed in major seismic events such as the Loma Prietra earthquake and Nansei-Oki (Okushiri) earthquakes [
28,
29,
30,
31].
In a study conducted by Manshoori [
32], pipeline damage in industrial areas and oil pipelines was evaluated following critical past earthquakes. Teng’s study [
33] implemented a seismic scenario analysis using a spatial grid to evaluate earthquake damage to underground pipelines in an urban area. Types of damage to underground pipelines were classified, pipeline disaster management procedures were discussed, and improvement measures such as establishing a geographic information platform and conducting disaster impact assessments for hazardous material pipelines were proposed. Borfecchia [
34] conducted a first-level spatial assessment of earthquake-induced damage in the region hosting Europe’s largest onshore oilfield and oil/gas extraction and pretreatment plant, with a pipeline that transports the plant’s products to the refinery in Taranto on the Ionian seacoast. In a study by Unal et al. [
35], the effects of geological hazards such as landslides, high peak ground acceleration (PGA), liquefaction, and surface rupture on natural gas pipelines during the Kahramanmaraş earthquake were evaluated.
In the study conducted by Kitaura [
36], the damage to water supply pipelines during the 1995 Hyogoken-Nambu earthquake was investigated. An overview of the damage to water supply pipelines in the Hanshin region was presented, and the relationship between the damage to water supply pipelines in Kobe city, geological characteristics, and building damage was discussed. It was determined that liquefaction caused extensive damage to the pipelines, and earthquake damage in various pipe types and diameters was examined.
In the study by Bouziou [
37], the effects of the Christchurch earthquake on the water distribution system were investigated through geospatial analyses using the LiDAR method. Areas prone to liquefaction and the repaired pipelines in these areas were shown.
The study conducted by O’Callaghan et al. [
38] presents firsthand observations and experiences specific to different pipeline systems in earthquakes in New Zealand, which occurred in a period of 27 years, from 1987 to 2014.
In the realistic and economical design of engineering structures, laboratory model experiments and advanced constitutive models can be utilized alongside numerical analyses. They accurately predict potential large soil deformations after liquefaction, which remains challenging with traditional methods. Soil conditions and seismic characteristics influence the displacements after liquefaction. In this study, the uplift behavior of pipes in liquefiable soils has been investigated. In this context, a centrifuge test has been numerically modeled to examine the behavior of buried pipes in liquefiable soil during earthquakes. In the analysis, the user-defined PM4Sand model in PLAXIS 2D (
https://www.bentley.com/software/plaxis-2d/ (accessed on 21 May 2024)) has been used, and the uplift movement of a concrete pipe has been examined. Experimental measurements at a depth of 1.1D and a seismic acceleration of 0.22 g have been compared to numerical analysis results to validate the model. In some areas of Iskenderun, damage occurred in stormwater and drinking water pipelines and facilities during the earthquakes in Kahramanmaraş on 6 February 2023, negatively affecting the continuity of urban life. Numerical analyses have been conducted using field soil conditions and actual earthquake records to evaluate the damage observed in steel pipelines in two areas in the Iskenderun district. In Study Area 1, the damage at the Iskenderun port has been examined using earthquake station data from the Kahramanmaraş Pazarcık earthquake (records of stations 3115 and 3116). In this area, the behavior of steel pipes with diameters of 1500 mm and 2000 mm has been modeled, and the effects of expanding the model boundaries on the pipes and the surrounding soil have been investigated. In Study Area 2, using records of station 3115, the damage that occurred in 1500 mm diameter steel pipes due to liquefaction on Atatürk Boulevard in Iskenderun have been investigated through numerical analyses and evaluated based on field observations.
This study aims to evaluate the practical applicability of existing liquefaction analysis methods. For this purpose, it was conducted using the PM4Sand model, which has been developed to model liquefaction behavior and is continuously updated in this regard. The methodology used is designed to assess whether existing methods accurately evaluate liquefaction risk in pipelines.
Our study utilizes existing calculation methods to analyze potential liquefaction consequences in pipelines in more detail. These analyses are crucial for determining how suitable existing methodologies are for real-world scenarios.
Our results can provide valuable insights into the applicability and accuracy of existing methods. We emphasize the importance of our study for conducting liquefaction analyses, which directly impact the safety and durability of pipelines, in a more reliable manner.
We believe this study will also serve as a foundation for future research, comparable to that presented by O’Callaghan et al. [
38], on liquefaction-induced pipeline damage in the seismically active Iskenderun region.
Thus, this study aims to shed light on predicting potential damage and implementing necessary improvement measures where critical pipeline networks pass through liquefiable areas.
3. PM4Sand Model and Verification Analyses
The Plaxis PM4Sand (Version 3.3) plasticity model has been used for the numerical analysis of soil liquefaction. The model follows the basic framework of the strain-controlled, critical state-based, bounding surface plasticity model initially developed by Manzari and Dafalias [
46] and extended by Dafalias and Manzari [
47]. In the model, the difference between the current relative density D
R and the critical state relative density D
R, cs is expressed as the relative state parameter index ξ
R. It has been described in the guides (starting from Boulanger [
48], with the latest version being Version 3.3 in 2023) and in associated publications [
49]. The model has been encoded as a dynamic link library (DLL) for use with the dynamic module of PLAXIS 2D.
The PM4Sand model is a constitutive model developed to account for excessive pore water pressure, liquefaction, and post-liquefaction behavior observed during earthquakes. This model is based on the critical state concept and is formulated using the bounding surface plasticity theory. The constitutive model is stress ratio-controlled, critical state compatible, and operates based on the bounding surface plasticity theory. In this model, the behavior of the soil is represented by dilation and critical state surfaces and bounded by the bounding surface plasticity to simulate the response of soil under seismic loading and confinement pressure within a specific density range. The formulation of the model can simulate the volumetric response of soil to repeated loading and directly simulate excessive pore water pressures [
49].
This constitutive model includes the following: revising the fabric formation/destruction to depend on plastic shear rather than plastic volumetric strains; adding fabric history and cumulative fabric formation terms; modifying the plastic modulus relationship and making it dependent on fabric; modifying the dilatancy relationships to include dependence on fabric and fabric history and to provide more distinct control of volumetric contraction versus expansion behavior; providing a constraint on the dilatancy during volumetric expansion so that it is consistent with Bolton’s (1986) dilatancy relationship [
50]; modifying the elastic modulus relationship to include dependence on stress ratio and fabric history; modifying the logic for tracking previous initial back-stress ratios (i.e., loading history effect); recasting the critical state framework to be in terms of a relative state parameter index; simplifying the formulation by restraining it to plane strain without Lode angle dependency for the bounding and dilatancy surfaces; incorporating a methodology for improved modeling of post-liquefaction reconsolidation strains; and providing default values for all but three primary input parameters [
49].
The initial relative density value has been estimated based on the SPT N value using the correlation suggested by Idriss and Boulanger [
51]:
The fundamental variable controlling the small-strain shear modulus is G
max. G
max is selected to match the estimated or measured shear wave velocities (
Gmax = ρVs2). The value is calculated based on the modified correlation between SPT (N
1)
60 according to Equation (2).
The third fundamental input parameter is the constant h
po used during calibration for a specific relative density value. The contraction rate parameter h
po, allows for calibration to specific values of the cyclic resistance ratio (CRR). This parameter should be calibrated last, after assigning the values of the other parameters. In this study, target CRR values have been determined according to the liquefaction-triggering correlation proposed by Idriss and Boulanger [
52], as shown in
Figure 3.
The h
po values are obtained by matching the CRR values found from direct simple shear (DSS) simulations with the CRR values calculated using the SPT-based liquefaction triggering correlation developed by Idriss and Boulanger [
52] with
CRR M = 7.5 values. It is assumed that the SPT-based CRR prediction, for an earthquake of magnitude M = 7.5 and 1 atm, the effective overburden stress is approximately equal to the CRR value corresponding to a 3% peak shear strain deformation induced by 15 uniform loading cycles in direct, simple shear loading [
49].
Numerical Model and Analyses
The numerical analysis of the centrifuge experiment conducted by Chian [
22] to investigate the uplift behavior of buried pipes in liquefiable soils was performed using the Plaxis PM4Sand material model. In the centrifuge experiment, a pipe with a diameter of 5 m was subjected to an acceleration of 66.7 g, corresponding to a burial depth of 1.1D = 5.5 m. Verification analyses for a single pipe at a depth of 1.1D were conducted considering the boundary conditions and material parameters in the centrifuge experiment. The properties of the pipe used in the model included a specific gravity, Gs = 0.3, and an equivalent wall thickness of 0.35 m.
The finite element mesh of the numerical model and the earthquake input motion are shown in
Figure 4. The finite element mesh consists of 3084 triangular elements with 15 nodes each. In the created model, the average element size is 0.71 m, and the earthquake data were applied to the analysis for 30 s at equal intervals of 0.005 s. Since there was no rock at the base of the experimental model, fixed base conditions (none) were selected, and tied degrees of freedom were used as boundary conditions at the lateral boundaries. During the application of earthquake data, drift correction was selected. For verification, numerical analyses were conducted considering parameters determined from Equations (1) and (2) for 0.9 D
R, 0.8 D
R, and 0.85 D
R for sand at 45% relative density. The PM4Sand constitutive model parameters for the case of 0.9 D
R and the parameters determined for the pipe are presented in
Table 1 and
Table 2, respectively.
In the modeling process, the initial step was created using the K0 procedure. In the second step, an HSS (Hardening Small Strain) soil model was assigned using parameters appropriate for sandy soil characteristics to create a new step. The main purpose of this was to accurately determine the initial stress state in the created soil profile. While the PM4Sand constitutive equations can successfully model the behavior of cohesionless soils under dynamic loads, they are inadequate under static conditions [
55]. In the third step, dynamic analyses were performed using the PM4Sand model. At this stage, the drainage type
“Undrained A
” was selected to observe the formation of pore water pressure.
Determining the damping parameters of the soil layer is essential in dynamic analyses conducted with finite elements. The Plaxis 2D program determines Rayleigh damping parameters using the C damping, M mass, and K stiffness matrices through Equation (3).
The Rayleigh damping parameters were determined in Plaxis using two different variables: Target 1 (
f1) and Target 2 (
f2). Here, (
f1) represents the dominant frequency of the entire soil profile and was calculated using Equation (4). In this equation,
VS, the average is the average shear wave velocity of the soil layer, and H is the layer thickness.
f2 is determined by rounding the value obtained from Equation (5), which depends on the parameters
feq and
f1, to the nearest integer [
55].
In the PM4Sand soil model, the target damping ratio for sand soil was set to 2%, and analyses were conducted using the values αR = 0.1761 and βR = 0.001906 corresponding to Target 1 and Target 2, respectively. Although PM4Sand was a successful constitutive model for modeling liquefaction, it was inadequate in obtaining initial stress conditions. Therefore, initial stress conditions were determined using the Hardening Small Strain (HSS) constitutive model.
In measurements of pore water pressure increase around the pipe, surges originating from structure of the pipe occurred, and it was observed that the liquefaction ratio was calculated more accurately as one moves away from the pipe. The excess pore water pressure ratio (Ru) value reached
Ru = 1, indicating the occurrence of liquefaction behavior on and around the pipe. It was determined that the calculated pore water pressures at the edge and base of the pipe were consistent with the centrifuge model experiment. Due to the fluctuations around the pipe, the excess pore water pressure ratios calculated in the far field were consistent with those measured in the experiment [
22] (
Figure 5a–d). Near the ground surface regions, the effect of pore water pressure increments, as shown in
Figure 5a, is calculated to be slightly lower (
Ru = 1) compared to the centrifuge model. This is because pore water pressures during earthquakes are more easily dissipated in near-surface areas. However, the lower values of pore water pressure increments calculated by the PM4Sand model are due to the damping effect of the soil. Nevertheless, the actual behavior occurring in the field can be determined by means of on-site measurements during field experiments.
In numerical analyses conducted for verification, material parameters were calibrated to model the behavior of liquefiable sand under seismic effects with PM4Sand, and it was determined that the displacements calculated with 0.9 D
R were consistent with measurements (
Figure 6).
The centrifuge test was modeled by Chian [
22] using the FLAC 2D finite differences software. In these analyses, the Wang elastic model [
56], capable of modeling liquefaction, was used, and a pipe uplift of 25% less than that determined in the test was computed. In Sudevan [
23], the centrifuge model was simulated using the FLAC3D Finn–Byrne formulation, and a structure uplift exceeding 1.0 m was calculated. The pipe uplift results calculated in those analyses are collectively presented in
Figure 6.
In this article, it was determined that the uplift value calculated from verification analyses is consistent with Chian’s [
22] experimental measurements. Liquefaction behavior could be realistically modeled using the PM4Sand model in Plaxis 2D.
4. Evaluation of Pipeline Damage Caused by Earthquake in the Iskenderun Region
After the earthquakes on 6 February, damage in the infrastructure facilities and pipelines in the region due to liquefaction were detected.
Figure 7 illustrates the layout of the sewage and stormwater pipelines, as well as pumping stations in Iskenderun, Hatay. The sewage and stormwater pipes have diameters ranging from Ø 300 mm to Ø 1200 mm. The stormwater pipelines on Atatürk Boulevard start from Ø 600 mm and reach up to Ø 800 mm, finally discharging at a diameter of Ø 1200 mm. After the earthquakes, the State Hydraulic Works planned to repair these damaged pipelines and pumping stations.
In
Figure 8a, a manhole cover and manhole shaft on Atatürk Boulevard that were damaged due to the earthquake are shown. In
Figure 8b,c, the area affected by the damage in the stormwater and sewage pipes on the same boulevard is seen. As can be seen in
Figure 8c, the pipes in this area were broken due to the occurrence of settlements caused by liquefaction after the earthquake, and as the existing inclination was insufficient, obstructed water flow caused blockages and ponding.
The U-shaped French stormwater channel shown in
Figure 7 and its associated pumping station, Shelter Sewer pumping station no.1, and Maternity House stormwater pumping station 2 were subjected to uplift reaching 30–40 cm in the caisson foundations of the structures (
Figure 9a,b). Damage occurred in the pipes inside these structures (
Figure 9c). In
Figure 10a, a photograph taken after the earthquake in the Iskenderun port area shows settlement reaching 1 m, resulting from liquefaction.
Figure 10b shows the approximately 0.70 m uplift of the shallow buried pipe in Çay District, covering the rear part of Atatürk Boulevard.
After the earthquake, it was stated in the Post-Earthquake Activity Report prepared by the Ministry of Agriculture and Forestry—State Hydraulic Works (DSİ) (dated 6 February–28 February 2023) that water was supplied to the city center of Hatay province from Harbiye Springs and Büyük Karaçay Dam. For this purpose, two steel flanges of Ø 1100 mm, 102 m of DN 800 cast concrete pipe, and three pieces of DN800 sliding sleeves were used to repair the transmission line. In order to rebuild and reroute a 750-m transmission line supplying water to İskenderun city center, Ø 1500 cast iron pipes were used [
59].
The damaged water supply line was observed at three different locations in Atatürk Boulevard (Work Area 2), where ductile steel pipes were located, as indicated in
Figure 11a–c. Pipes Ø 100 mm to Ø 1000 mm in diameter suffered broken joints due to uplift caused by liquefaction.
In the coastal area, in the Sarıseki neighborhood (Work Area 1), broken pipe joints due to liquefaction, a landslide which occurred along the water course near the spring, and rock fragments fallen onto the line caused significant damage, and the repairs were delayed due to difficult access. A total of 27 locations (
Figure 12) along the transmission line which suffered damage were accessible and were fixed by welding [
57,
59]. A 480 m long section of the pipeline on a cliff was repaired with difficulty by DSI and HATSU [
57]. Although not all of the leaks in the landslide zone could be dealt with, water can still be obtained from this pipeline [
57].
In 2018, a microzoning study was conducted by the Hatay Metropolitan Municipality in Iskenderun, and the geological setting of the Iskenderun coastline was revealed (
Figure 13a) [
60]. In the microzonation survey, 400 boreholes were drilled to depths ranging from 6 m to 30 m. Of these, 128 encountered 15 m to 30 m thick alluvium along the coastline. In those boreholes, there is a topsoil layer, 0.50 m thick, underlain by sandy, silty clay, sandy clay gravel, and coarse gravel with large boulders (alluvial unit). In the areas identified as terraces (Trç/yellow areas), conglomerate layers were reached within 0.5 m below the surface. The Okçular Formation (Theo/open pink areas) consists of limestone units. Hatay ophiolites (Kha/green areas) comprise cumulates, diabase, pillow lavas, and volcanic sediments. The area not covered by the microzonation survey is indicated in red (
Figure 13a).
In this article, the behavior of buried pipelines under seismic effects was investigated using numerical analyses, considering the soil conditions in the areas referred to as Study Area 1 and Study Area 2. Seismic records obtained from stations 3115 and 3116 in Iskenderun during the Kahramanmaraş and Pazarcik earthquakes were utilized for this investigation. In Study Area 1, numerical analyses were conducted for buried steel pipes with diameters of Ø 1500 mm and Ø 2000 mm at burial depths of 1.2D, 1D, 1.2D, and 1.5D using data from accelerometer station 3115 (located on soil) and accelerometer station 3116 (located on rock). The analysis model initially considered a width equal to the thickness of the soil (H) (50 m), and later, in repeated analyses, it was expanded to approximately 3H (120 m) in width. In Study Area 2, the behavior of a steel pipe with a diameter of Ø 1500 mm at a burial depth of 1.2D was investigated. The analysis model had a soil thickness (H) of 30 m and a width of approximately 3H (100 m). The analysis was conducted using data from station 3115 (located on soil).
4.1. Discussion of Results with Numerical Analysis of 3115 Station Data for Study Area 1
The drinking water pipelines along the coastline in the Sarıseki zone in Study Area 1 suffered damage caused by liquefaction, while landslide-induced damage occurred in the spring water transmission pipelines on sloping terrain. The transmission lines of the Azganlık water treatment plant located near the border of the Sarıseki neighborhood were put into use in order to provide sufficient water supply to the region. The Ø 1500 mm water treatment plant transmission line was inspected and no significant damage was detected; only a rerouting process [
57,
59] was carried out on this line to restore it to operational condition after the earthquake.
Figure 13c depicts geophysical measurements conducted in the Sarıseki location and borehole ISK-113.
Figure 14 presents the soil profile of borehole ISK-113 and geophysical survey (MASW-131) results. The geophysical survey showed that the shear wave velocity was 140 m/s in the top 4 m below the ground surface. Between depths of 4 m and 10 m, it went up to 210 m/s and then to 290 m/s below a 10 m depth. SPT-N
30 = 6–11 were obtained in this layer. Considering the geophysical survey results and SPT N values, the silty clay (alluvium) unit was classified as liquefiable (
Figure 14). The alluvium is underlain by a conglomerate (Terrace Fr.) layer with a low shear wave velocity, which is not liquefiable (Vs = 290 m/s). In areas where coastal reclamation works are being carried out, an artificial fill layer consisting of gravel and boulders was placed on the natural ground surface. The 7 m thick layer of clayey gravel with boulders having a low shear wave velocity as revealed in borehole ISK-113 and geophysical survey MASW 131 represents the man-made fill. In order to evaluate the liquefaction potential of lithological units we applied the standard penetration test (SPT) and used simplified methods for the calculation of the factor of safety (Fs), the cyclical resistance ratio (CRR), and cyclical stress ratio (CSR). These methods were originally developed by Seed and Idriss [
61], and later updated by Seed et al. [
53,
54,
62], Youd and Idriss [
62], and Youd et al. [
54]. According to these, the variation in the safety factors obtained for the ISK-113 drilling with depth is given in
Figure 14b. For local ground class ZE and a probability of exceedance of 10% in 50 years (with a return period of 475 years) along the Iskenderun coast (AFAD, [
39]), a short-period design spectral acceleration coefficient of S
DS = 0.977 was considered for calculating the safety factor (
Table 3).
At accelerometer station 3115, the peak ground acceleration (PGA) on the ground surface was measured as 287 cm/s
2. The coordinates of the station are (36.54634 and 36.16459), and it is located at an elevation of 180 m above sea level. Its distance to the epicenter of the Pazarcık earthquake is 113.57 km. The input seismic data () were initiated from a 1 m thick highly weathered rock layer, located at a 40 m depth at the base of the numerical model (
Figure 15), referred to as a residual soil layer. At the base of station 3115, the shear wave velocity was measured as Vs30 = 424 m/s.
The model shown in
Figure 15 consists of 6236 15-node triangular elements. In this model, the average element size is 0.858 m, and seismic record data were applied in the analysis at equal intervals of 0.01 s up to 99.9 s. Since there was no bedrock at the model’s base, ‘none’ boundary conditions were selected at the base. The soil and pipe material properties are presented in
Table 1 and
Table 2, respectively. A damping ratio of 2% was selected for liquefiable soil layers.
When the model width was expanded from 50 m (1H) to three times the layer thickness (3H), i.e., 120 m, the model created (
Figure 15) used 15,017 15-noded triangular elements, and the average element size was 0.8551 m. The boundary conditions for the model were selected as ‘none’ at the base.
As can be seen in
Figure 16b, the numerical analysis conducted in Study Area 1 showed that liquefaction occurred in the soil, as the excess pore water pressure ratio at a distance in the far field at the bottom of the pipe, Ru = 1. In
Figure 16c, during the first 50 s of the earthquake in the 50 m wide model, an uplift of 50 mm was determined in the Ø 1500 mm diameter pipe. Between 50 and 60 s, the layers beneath the pipe underwent an approximately 0.17 m downward displacement due to the overall liquefaction effect in the surrounding soil, followed by a rise of 3.6 cm (
Figure 16c). In the analysis conducted for a Ø 2000 mm diameter pipe under the same conditions, an uplift of 2 cm occurred within the first 50 s. Subsequently, between 50 and 60 s, there was downward displacement in the pipe, and after the 60 s mark, an uplift of 0.15 m was calculated (
Figure 16c). After the earthquake, following the settlement and upward displacements in the field, the Ø 2000 mm pipe returned to its initial position in the model (
Figure 16c). Since the effect of the analysis boundaries on the behavior of the pipe was determined, when the analyses were repeated after expanding the model boundaries, it was observed that the pipe was less affected by the boundary conditions (
Figure 16d).
As shown in
Figure 17, in 50 m and 120 m wide models, for the DC Ø 1500 mm diameter pipe, the downward displacements (settlements) at the ground surface 10 m away from the pipe were calculated as 1.25 m and 0.16 m, respectively (
Figure 17a,b). In the 50 m wide model in Study Area 1, downward displacements of up to 2 m were calculated at the boundaries, and these settlements also affected the zone around of the pipe (
Figure 17a). The high downward displacement immediately next to the pipe indicates that the boundary conditions affected the structure.
Figure 16c,d present the results of the analyses conducted with the 50 m and 120 m wide model, demonstrating the effect of model width and pipe diameter on the uplift behavior. For a burial depth of H = 1.2D in the 120 m wide model, uplifts of 0.245 m and 0.30 m were calculated for pipes of Ø 1500 mm and Ø 2000 mm, respectively (
Figure 16d).
With reference to the photograph in
Figure 12, the actual occurrence of liquefaction-induced broken pipe joints observed in the field support the pipe damage due to uplift as obtained in the analyses. The pipe uplift demonstrated by the analysis results presented in
Figure 17b indicates that the uplift behavior observed and measured up to 40 cm in the caisson foundations shown in
Figure 9b was the result of liquefaction. The settlement behavior observed in areas far from the pipe in the analyses represents the actual settlements observed near the port of Iskenderun in alluvial soils and artificial fill layers (
Figure 10a). On the other hand, the pipe uplift occurrence in the analyses was supported by the observed pipe uplift up to 70 cm as shown in
Figure 10b. When the boundaries in the analyses were extended to 120 m, the maximum downward displacement (settlement) dropped from about 2 m to 1.70 m (
Figure 17b). The effects of the boundary conditions on the settlements around the pipe were reduced.
In order to investigate the effect of the width of the analysis model on pipe displacements, various model widths were used. The total uplift initially calculated as 17 cm for the DC Ø 2000 mm steel pipe (
Figure 16c) increased to 30 cm when wider model boundaries were adopted (
Figure 16d). These results indicate that in seismic analyses where significant displacements occur, increasing the model width is crucial to mitigate the misleading effect of boundary conditions on pipe behavior.
In
Figure 18, the change in Ru value after the earthquake is given in the 120 m wide cross section containing a Ø 2000 mm pipe, where the red areas indicate Ru = 1. The PM4Sand model was not used for the conglomerate layer beneath the liquefiable layer. The HSS model was used for this layer, and therefore Ru values were not calculated (
Figure 18).
4.2. Discussion of Results with Numerical Analysis with 3116 Station Data for Study Area 1
In Study Area 1, the seismic performance of the Ø 1500 mm pipeline between the treatment plant and Isdemir steel plant was evaluated, taking into account the repair works carried out as described in the DSI (State Hydraulic Works) assessment report [
59].
After the earthquake, HATSU (Hatay Water Authority) recommissioned the treatment plant, declaring that the existing pipes did not sustain any damage [
57]. Study Area 1 is located near the Sarıseki neighborhood (
Figure 2), and based on results of the microzonation study, the seismic shear wave velocity Vs30 is higher in that area (
Figure 13c), suggesting that the bedrock is closer to the surface. Therefore, the seismic record of station 3116 was used.
The coordinates of station 3116 are (36.61618 and 36.20661), and its elevation is 33 m. The epicentral distance is calculated as 105.38 km. Analyses were repeated using the record from station 3116 in the soil profile of Study Area 1. In the record from station 3116, PGA is 160.8 cm/s
2, and the shear wave velocity is Vs30 = 870 m/s. Therefore, a 1 m thick bedrock layer was defined at the base instead of residual soil. Material parameters (linear elastic, γ
unsat = 24 kN/m
3, E = 250 × 10
−3 kN/m
2, ν = 0.25) for the rock layer used in the numerical analysis for Study Area 1 were utilized. The input data were applied from rock layers located at 40 m at the base of the analysis model, and analyses were conducted employing a model width of 50 m (
Figure 15).
In the finite element mesh of the 50 m long model, 6236 triangular elements with 15 nodes each were used. The average element size in the model was 0.858 m, and seismic data were applied in the analysis at equal intervals of 0.01 s up to 95.7 s. Since there was rock at the model’s base, the boundary conditions were selected as a compliant base, and free-field boundaries were used at the lateral boundaries. Since the input seismic data were applied from rock, the ‘Compliant base’ option was selected, and a prescribed displacement of 0.5 m magnitude was applied along the
x-direction across the base of the model in the numerical analysis. The
y-component of the strong ground motion was selected as fixed (
Figure 15).
Considering that the seismic acceleration is exerted from rock, the peak ground acceleration (PGA) is 160.8 cm/s2, and settlements are small (), the model width was selected as 50 m, assuming that the pipe was not significantly affected by the boundary conditions. Expanding the model boundaries extends the duration of the analysis. Using a seismic record obtained from bedrock at station 3116, the behavior of a Ø 1500 mm diameter pipe was analyzed at different burial depths (H/D = 1, H/D = 1.2, H/D = 1.5).
According to the analysis results, the uplift of the Ø 1500 mm pipe decreases as the burial ratio (H/D), i.e., burial depth, increases. For the burial ratio of H = 1D, the calculated uplift is 0.075 m, while for the H = 1.5D scenario, the uplift is calculated as 0.035 m (
Figure 19c and
Figure 20). The variation in vertical displacements is shown in
Figure 20. The analysis results (
Figure 19b) indicate liquefaction in the soil due to the increase in excess pore water pressure ratio, Ru = 1. However, the excess pore water pressures resulting from the proximity of the bedrock to the surface and the thickness of the liquefiable soil layer have been insufficient for the pipe to undergo displacements significant enough to cause rupture. In the Azganlık region, the shear wave velocity Vs30 values are higher and they fall within the green zone in
Figure 13c, representing a range of values between 400 m/s and 550 m/s. In the Sarıseki Region, the shear wave velocity values fall within the dark blue zone in
Figure 13c, i.e., they are in the range of 250 m/s to 350 m/s, providing more favorable conditions for liquefaction. According to the information provided by HATSU (Hatay Water Authority), the pipeline between the treatment plant and Isdemir steel manufacturing plant did not suffer significant liquefaction damage due to the earthquake. The analysis results also demonstrate low uplift behavior, ranging from 0.035 m to 0.075 m, indicating no significant liquefaction-induced damage in this area. The main reason for this is the proximity of the bedrock to the surface along the pipeline.
The modeling in Study Area 1 demonstrated the impact of seismic records obtained from the rock at station 3116 on underground pipe displacement. Based on the geophysical data of station 3116 (Vs30 = 870 m/s), analyses were conducted under the assumption that a 1 m thick layer of rock underlay the conglomerate layer despite the lack of information on the presence of solid rock below the alluvium.
In Study Area 1, the seismic records obtained from stations 3115 and 3116, as well as the variation in bedrock depth, significantly influenced the behavior of the pipe. The uplift of the pipe decreased by 70% when the bedrock seismic record from station 3116 was used, as shown in
Figure 16d and
Figure 19c. According to the results of the analysis, it can be concluded that in regions where the bedrock layers are close to the surface, the soil exhibits smaller settlement and, consequently, smaller pipe uplift. As shown in
Figure 16c, the displacement around the pipe decreased by 0.30 m in total settlement when the model width was expanded to 120 m. This resulted in greater uplift in the pipe (
Figure 16d). There was no downward deformation during the uplift of the pipe. However, at a distance of approximately 10 m from the pipe, vertical displacements (settlements) on the surface were at about 0.15 m. As shown in
Figure 19c, in the analysis conducted using the seismic record from station 3116 and assuming that the bedrock was close to the surface, lower peak ground accelerations (PGA) resulted in smaller uplift. The significant difference between soil conditions at stations 3115 and 3116, characterized by different shear wave velocities (Vs30 = 424 m/s and Vs30 = 870 m/s, respectively), resulted in different seismic effects on the corresponding units. As a result, areas analyzed using the seismic record from station 3115 were observed to experience displacements exceeding 1.7 m, as shown in
Figure 17b. On the other hand displacements of up to 0.56 m were calculated when the seismic record from station 3116 was used in areas where the bedrock was close to the surface (
Figure 20).
These analyses demonstrate the importance of determining the model width to minimize the impact of model boundaries and boundary conditions on the behavior of the buried pipe. However, increasing the model width also significantly increases analysis durations. Therefore, in numerical analyses, model boundaries should be selected in such a way that they do not affect the behavior of the buried structure.
4.3. Discussion of Results with Numerical Analysis with 3115 Station Data for Study Area 2
Atatürk Boulevard is located along the coast of Iskenderun (
Figure 2). While the seaside of this boulevard is designated for parks and social areas, the opposite side is occupied by buildings. According to the Preliminary Assessment Report for the Kahramanmaraş–Pazarcık Mw = 7.7 and Elbistan Mw = 7.6 earthquakes on 6 February 2023, the majority of the buildings in this location were found to have suffered moderate to severe damage. This report provides images and locations of damaged buildings in the Çay Neighborhood [
58]. The effects of liquefaction have been observed both in buildings and buried structures along the coastline.
According to the microzonation report [
60], most of the Iskenderun coastline consists of alluvial units, as shown in
Figure 13a.
Figure 13b displays the MASW Vs30 map of Study Area 2 along the Iskenderun coastline, as well as the location of borehole ISK-259. The soil profile according to the boreholes drilled in the area are given in
Section 1 in
Figure 13d. The orientation of the section is shown in
Figure 13b.
After the earthquake, liquefaction-induced damage occurred in stormwater and sewage pipelines along Atatürk Boulevard lying parallel to the coastline, and some pipes were subjected to vertical displacement and consequently became blocked, causing local flooding on the streets.
The borehole and geophysical surveys representing Study Area 2 are presented in
Figure 21a. According to the MASW-386 survey, the shear wave velocity is 210 m/s in the top 7 m of the ground, rises to 250 m/s between depths of 7 and 12 m, and becomes 300 m/s below. According to borehole ISK-259 representing this area, the top layer is sandy clay, 3 m thick (58% D
R), followed by a layer of gravelly sand, 17 m thick (
Figure 21a). Borehole ISK 258 in
Section 1 in
Figure 13d indicates that the thickness of the alluvium is not less than 30 m. Both layers, as particularly indicated by the SPT-N
30 (5–12) results, show liquefaction potential; therefore, they can be considered liquefiable also based on seismic velocities (
Figure 13d and
Figure 21a). The variation with depth of the safety factors obtained using relationships derived from Seed and Idriss [
61] and later updated [
53,
54,
62] is presented in
Figure 21b. For local ground class ZE and a probability of exceedance of 10% in 50 years (with a return period of 475 years) along the İskenderun coast, a short-period design spectral acceleration coefficient of SDS = 0.972 was considered for calculating the safety factor (
Table 4). Although no information is available on the layer beneath the alluvium, it has been demonstrated in the study by Ozener [
63] that the bedrock lies at a depth greater than 50 m based on seismic velocities. Therefore, at a depth of 30 m, the presence of an extremely weathered rock (residual soil) layer was assumed (as shown in
Table 2). In the analysis, Station 3115 records were used, and a damping ratio of 2% was considered for the liquefied soil layer.
To mitigate the effects of boundary conditions in Study Area 2, a steel pipe having a diameter of DN 1500 mm was selected in a model width of 100 m, which corresponds to more than three times the size of the 30 m thick layer (
Figure 22). The pipe and soil material parameters are presented in
Table 2 and
Table 5. In the finite element mesh, 9806 triangular elements with 15 nodes each were used, resulting in an average element size of 0.82 m in the model. The boundary conditions shown in
Figure 22 were applied.
According to the analysis results (
Figure 23a) obtained using the seismic input record provided in
Figure 17a, with the excess pore water pressure ratio, Ru = 1, liquefaction occurred at a depth of 3.2 m below the surface. The upper layer within the first 3 m is a fill layer composed of sandy clay with a relative density of 58% D
R. Liquefaction effects are more pronounced in the underlying layer with a relative density of 38% D
R. Displacements due to liquefaction reached 1.0 m at a distance of 10 m from the pipe and vertical displacements in the pipe were calculated as 0.27 m (
Figure 23b). The settlement at the section where the pipe was located was reduced by up to 0.70 m due to the uplift effect in the pipe. Therefore, the excess pore water pressure caused by the presence of the pipe reduced the settlement occurring in the soil. Between 60 and 70 s, the settlement movement of the pipe ceased, and then it continued to settle again (
Figure 23b). The total vertical displacements after the earthquake motion are shown in
Figure 24.
In the analyses, the Ru value reached 1 and liquefaction occurred in the alluvial unit with a relative density of 38% D
R beneath the 3 m thick layer of sandy clay/silty sand (artificial fill), with SPT-N = 16 and D
R = 58% at the surface. A thin denser soil layer at the surface (artificial fill) and the underlying 27 m thick alluvium layer contributed to the occurrence of a relatively large settlement. Therefore, the soil profile constrained the uplift of the pipe, giving way to settlement in the pipe along with the soil. In the photograph of Atatürk Boulevard taken one year after the earthquake, as shown in
Figure 10, it is noted that settlement and rupture occurred instead of uplift in the sewer manhole. All the drainage lines, especially the stormwater and sewer lines located on the coast, as shown in the plan obtained from HATSU (Hatay Water Authority) [
57] (
Figure 7), were damaged due to settlements, and repairs have not been completed yet. The entire sewer and stormwater infrastructure serving the city center collapsed (
Figure 7). Fractures occurred in the manhole, as shown in
Figure 8a, similar to the expected outcome of the settlement behavior suggested by the analysis results.
The effect of the thickness of the liquefiable layers on the pipe uplift behavior was clearly demonstrated. After the earthquake, displacement along the coastline resulted in flooding in the streets and avenues. The analysis results confirm occurrence of significant displacements (
Figure 24). The ground displacements and the collapse of stormwater and sewage pipelines led to groundwater contamination in the damaged areas, resulting in environmental pollution and flooding.
Drinking water pipelines also suffered similar settlement problems, as suggested by the analyses.
Figure 11 shows plan views of three localities where drinking water pipelines were damaged due to liquefaction-induced settlements. In the field, fractures have occurred, particularly at the connections of pipelines to buildings, due to different structural behaviors and boundary conditions.
6. Conclusions
In the article, the behavior of buried pipes during the 6 February 2023, Kahramanmaraş Pazarcık earthquake was investigated through numerical analyses using the records of accelerometer stations 3115 and 3116 in Iskenderun, considering geological profiles in two areas with liquefaction potential. The effects of soil conditions and seismic acceleration on pipe behavior at the recorded station were determined. Liquefaction occurred in the coastal area, and the excess pore water pressure ratio (Ru value) reached 1. In the studied area, a vertical displacement of maximum 0.56 m was calculated when bedrock was shallow, whereas in areas where the bedrock lay at a considerable depth, the calculated displacements range from 0.56 m to 1.7 m. The field observations and damage records confirm the amounts of settlement calculated. Field observations revealed liquefaction-induced settlements ranging from 0.30 m to 1.50 m along the coastline.
In Study Area 1, the effects of shallow and deep rock layers, pipe diameter, burial depths, and boundary conditions were investigated using records from accelerometer stations 3115 and 3116. A small amount of pipe uplift was calculated when there was a shallow rock layer. In shallow rock conditions, different burial depths (1D, 1.2D, and 1.5D) were tried, and increasing the burial depth reduced the amount of uplift. The absence of significant damage in the transmission line of the treatment plant under station 3116 supports the analysis results. When the rock layer was deep (station 3115 records), greater pipe uplift was calculated, indicating that boundary conditions influence the model to a greater extent. When the model boundaries were expanded, settlements around the pipe were reduced, and the obstructive effect of this behavior on pipe uplift diminished. It is concluded that in seismic analyses where significant displacements occur, increasing the model width is essential to reduce the misleading effect of boundary conditions on pipe behavior. Additionally, the increase in pipe diameter resulted in increased pipe uplift in the analyses. The analysis results and the observed damage at pipe joints due to liquefaction in coastal sections of the Sarıseki spring water transmission line are consistent. The main cause of the significant pipe damage in this area was slope failures following the earthquake.
In Study Area 2, the effect of pipe uplift in a thick alluvial layer was investigated using seismic records obtained from the soil layers at station 3115 (where the bedrock was deep). In the analyses, significant settlements occurred at a distance from the pipe. In the section where the pipe was located, however, the settlement amount decreased due to the uplift effect of the pipe. The flooding on Atatürk Boulevard due to damage inflicted on sewer and stormwater pipelines, caused by the occurrence of post-earthquake settlements governed by the thickness of the liquefied layer, validates the analysis results. The uplift observed in the caisson foundations of the stormwater and sewage pumping station on the coast supports the calculated uplift behavior in the region.
The results of this study lead to the conclusion that the dynamic modeling of critical pipes in liquefaction-prone areas can support measures to prevent damage during earthquakes, by predicting seismic effects. This study covers the results obtained from numerical analyses performed utilizing actual specific ground and seismic data. Solution proposals can be generated by conducting numerical analyses for different ground conditions.