COASTAL FAILURES DURING THE 1999 KOCAELI EARTHQUAKE IN TURKEY
Ellen M. Rathje1, Ismail Karatas2, Stephen G. Wright3, and Jeff Bachhuber4
During the 1999 Kocaeli earthquake (Mw=7.4) in Turkey, coastal failures and sea
inundation were observed and were particularly concentrated along the margins of Izmit Bay and
Lake Sapanca, in pull-part basins created by stepovers in the fault rupture. Geotechnical site
characterization, geologic mapping, liquefaction evaluation, and slope stability analysis were
carried out to identify the principal contributing factors of the coastal failures. Results from this
study indicate that both liquefaction and tectonic subsidence contributed to the failures and sea
inundation within the pull-apart basins. Most of the liquefaction sites were situated at the
prograding nose of active delta fans, where the presence of steep slopes coupled with the loose
sediments found within young active delta fan deposits resulted in liquefaction-induced slope
failures and sea inundation. Liquefaction in other coastal deposits outside of the actively
prograding delta fans caused limited lateral spreading and only minor sea inundation. Outside of
the delta fans, where soils were not liquefiable, tectonic subsidence associated with normal
faulting was the cause of the observed sea inundation. Generally, tectonic subsidence caused the
most severe sea inundation. Based on these observations, the identification of regions
susceptible to both tectonic subsidence and liquefaction are important when evaluating seismic
KEYWORDS: Subsidence, Ground Failure, Liquefaction, Faulting, Coastal Failures
Assistant Professor, Department of Civil Engineering, University of Texas, Austin, TX 78712.
E-mail: email@example.com, FAX: 1-512-471-6548
Staff Engineer, GeoSyntec Consultants, Huntington Beach, CA.
Brunswick-Abernathy Professor, Department of Civil Engineering, University of Texas,
Austin, TX 78712
Principal Engineering Geologist, William Lettis and Assoc., Walnut Creek, CA 94570
The Kocaeli earthquake (Mw = 7.4) occurred on 17 August 1999 in the northwestern part
of Turkey along the North Anatolian fault. The bilateral strike-slip fault rupture involved
displacement on four distinct segments of the North Anatolian fault (Figure 1). These strike-slip
fault segments are separated by right-releasing stepovers, which accommodated significant
normal-slip displacement (up to 2.4 m) during the earthquake . The stepovers in the fault
rupture coincide with distinct pull-apart basins that are filled with thick Quaternary deposits.
Severe coastal subsidence, ground failure, and sea inundation were observed within the pull-
apart basins located in coastal areas. This study focuses on identifying the main geotechnical
and geologic factors that contributed to the coastal failures and sea inundation in the coastal pull-
apart basins during the Kocaeli earthquake.
The main coastal and near-shore submarine stepovers involved in the Kocaeli earthquake
fault rupture were the Karamursel, Golcuk, and Sapanca stepovers (Figure 1). Each of these
stepovers is associated with a distinct, sediment filled pull-apart basin and a topographically low-
lying area. Widespread subsidence and sea inundation were observed within each basin, causing
some of the most dramatic damage and destruction of coastal areas and facilities from the
earthquake. Geotechnical site investigation, geologic mapping, liquefaction evaluation, and
slope stability analysis were carried out to investigate the localized failures and subsidence
observed at several sites within the coastal pull-apart basins. These sites are Degirmendere
within the Karamursel pull-apart basin; Golcuk, Yenikoy, and Seymen within the Golcuk pull-
apart basin; and Esme within the Sapanca pull-apart basin (Figure 1). The collected data were
evaluated to study the interaction and relative contribution of different mechanisms to the
observed coastal failures.
The 1999 Kocaeli earthquake occurred on the western portion of the North Anatolian
fault, which is a major strike-slip fault extending over 1,500 km across northern and western
Turkey. This fault zone has experienced numerous past earthquakes (e.g., 1939 Erzincan, 1942
Erbaa, 1944 Bolu-Gerede; ), and the Kocaeli earthquake occurred on a portion of the fault that
was formerly a seismic gap. The earthquake ruptured a 126 km section of the fault which is
comprised of four distinct segments separated by three extensional pull-apart basins: the
Karamursel, Golcuk, and Sapanca basins . Pull-apart basins formed as a result of extension
within the basins that is accommodated by vertical displacement along basin-bounding normal
faults (Figure 2) and causes structural downdropping and warping. In coastal lowland areas and
delta plains, this downdropping allows for sea inundation and forms topographic lows that are
loci for sedimentation. The pull-apart stepovers in Turkey have been infilled with thick
Quaternary deposits over time, creating deep alluvial basins. The large dimension and thick
sediment infilling of these basins indicate that they are long-lived features in the displacement
history of the fault .
Numerous prograding delta fans have formed along the coastal margins of the pull-apart
basins along the North Anatolian fault. These deltas are constructed at the discharge points of
creeks and rivers that transport sediment from the adjacent elevated areas and deposit their
sediment load into the topographic depocenter and water bodies at the basin margin. The
depositional processes within delta fans produce loose, saturated sediments that often contain
fine sand and silty sand layers that are susceptible to liquefaction. Additionally, the submerged
prograding delta noses typically are relatively steep (greater than 10 to 15 degrees) and quasi-
stable. The combination of loose sediments and steep unstable delta nose slopes makes delta fan
deposits prone to liquefaction-induced ground failure (i.e., lateral spreading, slope failures).
This study included areas of ground failure, subsidence, and sea inundation in each of the
three coastal pull-apart basins affected by the Kocaeli earthquake (Figure 1). The Karamursel
basin is a submarine basin within Izmit Bay that was identified by evaluation of bathymetric and
geophysical data. Displacement on the Golcuk fault, entering the east side of the basin, was
rapidly attenuated within the basin, and only negligible to no fault displacement was observed on
the Yalova fault that exits the west side of the basin and crosses the Hersek Peninsula . The
coastal geology along the bay margin here consists mainly of Plio-Pleistocene sedimentary
bedrock overlain in some areas by Pleistocene marine terrace deposits and Holocene sediments.
The Holocene deposits consist of (1) alluvial fans and stream channel deposits that are incised
within older sediments and bedrock, and (2) actively prograding delta fans and modern beach
deposits along the narrow coastline. The active delta fans occur at the mouth of incised creeks
that originate in the hills immediately to the south of the coastline. Because of the close
proximity of the sediment source in the steep hills to the south, the delta fan sediments in this
area are relatively coarse (sand, fine gravel), poorly sorted, and laterally discontinuous.
The Golcuk pull-apart basin is located east of the Karamursel basin and encompasses a
coastal area that extends east from the city of Golcuk (Figure 1). It is approximately 2 km wide
and 6 km long, and is located within a right-lateral step between the Golcuk and Sapanca fault
segments. The shoreline geometry is controlled both by local faulting and alluvial deposition.
Portions of the coastline are coincident with, and subparallel to, active splays of the North
Anatolian fault that ruptured during the Kocaeli earthquake. In other areas, the shoreline is
formed by coalescing alluvial deltas that prograde into Izmit Bay from alluvial plains, fans, and
bedrock hills to the south. Repeated tectonic subsidence within the Golcuk pull-apart basin has
formed a localized depocenter for delta progradation between the towns of Golcuk and Seymen.
The basin includes a broad, late Pleistocene to Holocene alluvial plain that is bordered by the
Golcuk normal fault at the west and south, and older Pleistocene fans that extend out from the
range south of Izmit Bay. Stream channels crossing the plain have shifted repeatedly over time,
forming broad delta fans at the discharge points with laterally interfingering packages of
sediment. In general, delta deposits become finer grained with increasing distance from the
mountain range front. The youngest, most actively prograding parts of the deltas occur at the
mouth of major streams and are defined by triangular-shaped cones of sediment that project as
much as 100 m seaward from the shoreline. These active delta lobes can be differentiated on
aerial photographs and consist of unconsolidated, loose sediments that are considerably less
consolidated than adjoining older parts of the delta. The creeks feeding the Golcuk basin delta
fans cross the wide alluvial plain and originate in the bedrock hills and older fans deposits further
inland than the streams feeding the delta fans that border the Karamursel basin. As a result, the
Holocene delta sediments within the Golcuk basin are significantly more fine grained (i.e., fine
sand, silt, and clay) than those bordering the Karamursel basin.
The Sapanca pull-apart basin is located within a right-lateral step between the Sapanca
and Sakarya fault segments (Figure 1). The 60-m deep topographic depression caused formed by
the Sapanca pull-apart basin is infilled by Lake Sapanca and serves as a local, internally-drained
depocenter for sedimentation for streams that issue from the bedrock hills surrounding the basin.
The margins of much of the lake consist of coalescing late Pleistocene to Holocene alluvial fans
and terraces, and late Holocene beach deposits and prograding delta fans at the mouths of creeks.
Pliocene and Pleistocene bedrock and older alluvial fans occur inland from the current lake
margin and were deposited in a somewhat larger ancient tectonic basin that appears to have
narrowed along the active fault traces. The creeks feeding the delta fans bordering Lake Sapanca
originate in incised stream vallies close to the lake shore, and the fan sediments consist of
laterally-discontinouos, relatively coarse sand and gravel with some interfingering alluvial silt
and clay lenses and fine-grained lake deposits
Bordering the Karamursel basin, the most severe sea inundation took place along a
Holocene delta fan at the mouth of Degirmen creek in the city of Degirmendere. The delta fan at
Degirmendere is located at the interface between a large Plio-Pleistocene alluvial fan complex to
the south, and a series of Pleistocene marine terraces that form a flight of relatively flat
topographic surfaces subparallel to the shoreline. The Holocene delta is incised into these older
deposits, and projects north into Izmit Bay. The shoreline delta deposits are comprised of
reworked Plio-Pleistocene fan and marine terrace sediments, and are relatively coarse and
laterally discontinuous. During the Kocaeli earthquake a large section of coastline, which
coincides with the Holocene delta fan, failed into Izmit Bay, extending approximately 300 m
along the coast and 75 m inland (Figure 3). A curvilinear, well-defined 1 to 2-m high head scarp
defined the landward extent of the failure zone and formed the post-earthquake shoreline. The
headscarp was located approximately 800 m south of the offshore projection of the Golcuk fault,
and did not appear to be associated with primary surface fault rupture. The headscarp shape and
location indicated that the coastal retreat was caused by a distinct slope failure rather than
tectonic subsidence or lateral spreading. Minor cracking extended an additional 100 m inland of
the coastal scarp to the approximate contact between late Holocene delta deposits and older
alluvial deposits. The cumulative amount of extension across the cracks in this zone suggest that
up to about 0.45 m of lateral seaward movement occurred behind the slide headscarp. No ejected
sand boils or other evidence of near-surface liquefaction was observed in the failure area;
however, some buildings inland of the failure experienced slight to moderate levels of
settlement. The pattern of cracking and lack of significant surficial differential movements
suggests that the slope failure was a deep slump or slide-type of failure rather than a classic
shallow liquefaction-induced lateral spread.
The pre-earthquake geometry at Degirmendere was constructed using a bathymetry map
, while the post-earthquake geometry was developed from bathymetric data provided by
Degirmendere city officials. Pre- and post-earthquake cross sections taken perpendicular to the
post-earthquake shoreline in the center of the failure zone are shown in Figure 4. The pre-
earthquake geometry shows that the offshore delta fan had a relatively steep slope, with a
maximum slope angle of about 18 degrees at a location about 100 m offshore. At a distance of
about 175 m from the coastline, the seafloor was almost flat. The post-earthquake geometry
shows a significant change in the shoreline and seafloor, with a steep (30 degree to vertical)
shoreline headscarp, and locally-steepened nearshore slope. The seafloor slope inclination
offshore of the headscarp zone was reduced to about 5 degrees after the failure. Comparisons of
the pre- and post-earthquake profiles shows that the maximum thickness of sediment lost by the
landslide approached 25 m, and that the failure appeared to be relatively deep-seated.
The geotechnical site investigation at Degirmendere consisted of 3 SPT borings and 2
CPT soundings (Figure 3) performed by the local firm ZETAS. These borings and soundings
covered a distance of about 50 m along the shoreline, and extended 10 to 15 m from the slide
headscarp and new shoreline. Rotary wash borings were performed with SPT measurements
taken every 1.5 m. The SPT were performed in accordance with ASTM D1586 , using a
safety hammer, rope and cathead, and AWJ drill rods. SPT energy measurements were not
performed. The CPT soundings were performed with standard CPT equipment manufactured by
A.P. van den Berg. Two SPT borings and two CPT soundings at Degirmendere were located
very close to the slide headscarp/shoreline, in an attempt to sample material that is similar to that
which was involved in the slide. However, because the delta sediments are laterally variable,
and the slide zone occurred mainly offshore, the collected subsurface data may vary somewhat
from that within the body of the slide.
In situ data from the two CPT and two SPT performed along the shoreline at the mouth of
the creek in Degirmendere are shown in Figure 5. The subsurface materials encountered in these
borings consist primarily of medium dense sand and fine gravel in the top 30 m, with some
layers of silty clay and silty sand. The sand layers generally display CPT qc values between 10
and 20 MPa, while the SPT blowcounts are between 10 and 30. Some silty sand and silty clay
layers show qc values as low as 2-3 MPa, but these weaker layers were not found at consistent
depths within the different borings and soundings performed along the shoreline (Figure 5). The
discontinuity of layering is most likely the result of the complex depositional processes at work
along the coastline of this delta fan. The CPT and SPT data show that the sediments exhibit an
increase in resistance at about 10 m. Shear wave velocity measurements made at the site using
the Spectral Analysis of Surface Waves method  indicate a similar trend, with the shear wave
velocity increasing from about 200 m/s in the top 9 m to about 270 m/s at depths from 9 to 14 m.
The CPT and SPT data were used to assess the liquefaction susceptibility of the soils at
Degirmendere . The procedures outlined by Youd et al.  were used to evaluate liquefaction
susceptibility in terms of the cyclic stress ratio induced by the earthquake (CSR) and the cyclic
resistance ratio (CRR). The earthquake-induced CSR was computed as:
CSR = / vo = 0.65 (PGA/g) (vo / vo) rd (1)
where PGA is the peak ground acceleration, is the average shear stress, g is the acceleration of
gravity, vo and vo are total and effective vertical stress, respectively, and rd is the stress
reduction coefficient. The cyclic resistance ratio (CRR) was computed using measured in situ
test parameters and appropriate liquefaction correlations, as described by . For the CPT data,
the CRR was evaluated only for layers where the soil behavior type index (Ic, ) was less than
2.6. This value of Ic represents the boundary between predominantly granular materials (sand to
silty sand) and predominantly fine-grained materials (silts and clays). Although some soils with
Ic greater than 2.6 may liquefy , this possibility was not considered in this study.
For the liquefaction evaluation at Degirmendere, a PGA value of 0.3 g was used.
Because no strong motion station was located in the vicinity of Degirmendere, this PGA value
was inferred from the Yarimca Petkin (YPT) strong motion record that was recorded
approximately 4 km north of Degirmendere on the opposite site of Izmit Bay . The YPT
station is situated on deep alluvium and recorded a PGA of 0.27 g (geometric mean of the two
horizontal components). Based on this recording, the PGA at Degirmendere was taken as 0.3 g,
and the computed CSR ranged from 0.25 to 0.35. The two CPT and two SPT in Figure 5
consistently predict that the soil from about 5 to 10 m (qc ~ 8-12 MPa, Ic = 1.5, N1,60 ~ 10-20)
was liquefiable during the Kocaeli earthquake, with factors of safety (FS) between 0.5 and 0.9.
At depths below 10 m, the various CPT and SPT data do not consistently predict the same
liquefaction potential. DN-CPT2 indicates highly liquefiable soils from 10 to 25 m (FS < 0.6),
while adjacent DN-SPT1 indicates all soils below 15 m have a FS greater than 1.1. DN-CPT1,
located 25 m east of DN-CPT2 and DN-SPT1, shows no soils below 10 m as liquefiable (FS >
1.5). Finally, DN-SPT2, located 25 m east of DN-CPT1, indicates dense sand from 10 to 25 m,
but liquefiable silty sand between 25 and 30 m (N1,60 ~ 10).
The in situ test data indicate liquefiable soils are present within the deposit at
Degirmendere. However, these liquefiable soils are generally restricted to a depth of about 10 m.
The geometry of the failure zone, scarp shape and height, and cross section suggests that the
Degirmendere failure was deeper than 10 m, and that the shallow liquefiable layers encountered
in the borings cannot explain the deep-seated failure at Degirmendere. Therefore, slope stability
analyses were performed to gain further insight into the failure mechanism at Degirmendere.
Slope stability analyses using the program UTEXAS4  were performed for both the
pre-earthquake and post-earthquake geometries at Degirmendere. The subsurface materials
under pre-earthquake conditions were modeled as granular soils with an effective friction angle
of 35 degrees. As expected for a slope consisting of granular materials under saturated
conditions, the critical slip surface indicated an infinite mode of failure along the steepest section
of the slope (offshore nose of the delta fan). The computed factor of safety was 1.75, indicating
that the slope was adequately stable under static conditions before the earthquake. Using drained
strengths, a minimum yield seismic coefficient (ky) of 0.1 was computed for the slope. This ky
corresponds to an infinite slip surface, which is not compatible with the observed deep-seated
Slope stability analyses were performed next considering liquefied strengths for the
susceptible layers encountered in the borings and CPT soundings at depths of between 5 and 10
m, and between 25 and 30 m. A post-liquefaction residual strength (SR) of 40 kPa was assigned
to the shallow liquefiable layer based on an average N1,60 of 15 (FC~5%) and the
recommendations from Seed and Harder  and Baziar and Dobry . The deeper liquefiable
layer was assigned an SR of 20 kPa based on N1,60 ~ 10 and FC ~ 25%. Analyses with residual
strengths assigned only to the shallow liquefiable layer predicted an infinite slope failure mode
because the shallow liquefiable layer coincides with the flatter part of the slope. Analyses with
residual strengths assigned to both the shallow and deep liquefiable layers predicted a deep-
seated failure with a factor of safety equal to 1.05. The yield seismic coefficient for this slip
surface (using drained strengths) is 0.19. Considering a PGA of 0.3 and reducing that value to
kmax = 0.2 to correct for averaging effects over the depth of the sliding mass  results in a
ky/kmax close to 1.0, suggesting that inertial effects alone, without soil strength reduction by
liquefaction, could not have triggered the failure at Degirmendere. The failure appears to have
initiated on a deep liquefied soil layer in the steeper part of the offshore delta nose, and may have
expanded laterally in a progressive mode. The zone of cracking inland of the failure scarp
suggests that this zone was quasi-stable and experienced some minor extensional movements in
response to formation of the slide headscarp and partial liquefaction of underlying soil layers.
GOLCUK PULL-APART BASIN
The Golcuk pull-apart basin is located on the south shore of Izmit Bay and encompasses
the coastal area that extends east from the city of Golcuk (Figure 6). The Golcuk basin is
bounded on the southwest by the Golcuk normal fault that accommodated between 0.5 and 2.4 m
of vertical displacement during the Kocaeli earthquake. The normal faulting produced global
subsidence of the basin, localized coastal subsidence, and sea inundation. The most dramatic sea
inundation from this earthquake occurred within the Golcuk basin, where approximately 0.5 km2
of the basin was inundated by the sea and another 0.75 km2 experienced substantial subsidence
but remained above sea level. The Golcuk normal fault follows a higher (3 to 6 m) paleoscarp in
Holocene and late Pleistocene deposits, suggesting that the fault has experienced at least several
previous episodes of normal displacement. Three sites of subsidence were investigated within
the Golcuk pull-apart basin: central Golcuk, Yenikoy, and Seymen (Figure 6).
In central Golcuk, severe coastal subsidence occurred along the western margin of the
Golcuk pull-apart basin, near the Golcuk normal fault (Figure 6). Here, shoreline retreat
extended inland over 300 m. Minor evidence of liquefaction was reported around the failed area
and local residents reported that settlements continued for weeks after the earthquake .
The zone of intense sea inundation coincides with a Holocene delta fan that formerly
prograded as much as 100 m northward into Izmit Bay. The subsidence is concentrated in a
cone-shaped area at the nose of the delta fan, at the intersection between the normal fault and
shoreline. Approximately 1.5 m of vertical fault displacement occurred on the Golcuk normal
fault adjacent to this inundated area. Much of the subsidence here can be directly related to this
vertical downdropping. However, the magnitude and locally more-severe subsidence within the
Holocene delta fan suggests that other processes in additional to normal faulting contributed to
the inundation. Pre-earthquake aerial photographs of central Golcuk taken in 1974 indicate that
the subsidence and inundation zone coincides with an area that was an active delta in 1974.
Between 1974 and 1999, the creek that was the source of this delta was diverted into a channel,
and the active delta area was developed. Additionally, a bathymetry map of Izmit Bay 
indicates that the pre-earthquake slope offshore of the subsidence zone was steep (about 15 to 20
degrees), and formed by the rapidly accumulating delta sediments. Post-earthquake bathymetry
data suggest that the offshore area experienced some subsidence and that the seafloor slope was
flattened, although the data are not sufficient to differentiate the geometry of the offshore failure.
The concentrated, intense subsidence in this delta area and the flattened offshore delta nose
suggest that liquefaction-induced ground failure also contributed to the localized inland
Three CPT soundings were performed in the vicinity of central Golcuk. Two of these
soundings (GL-CPT1, GL-CPT2) were performed west of the subsidence zone, while the other
(GL-CPT3) was performed adjacent to the subsidence zone, immediately east of the creek. Due
to rocky fill being placed to reclaim the subsided area after the earthquake, soundings could not
be located immediately within the subsidence zone. No SPT borings were performed in this
The CPT sounding adjacent to the subsidence zone (GL-CPT3), along with the
interpreted subsurface profile, is shown in Figure 7. The top 10 m consist of loose, alluvial silty
sand (qc ~ 4-8 MPa, Ic ~ 1.5-2.0) with some interfingering alluvial and marine clay seams, which
is in turn underlain by about 11 m of alluvial and marine clay. Dense sand is found below the
clay layer (qc ~ 30 MPa). The CPT data were used to assess the liquefaction susceptibility of the
soils in central Golcuk using the procedures outlined previously. The PGA at Golcuk was
estimated as 0.3 g, based on the YPT recording on the opposite side of Izmit Bay. The CPT data
indicate that the soil from about 2 to 10 m is highly liquefiable, with a factor of safety of about
0.5. This result further suggests that liquefaction played a role in the subsidence in central
The measured values of CPT tip resistance in the top 10 m from the three CPT soundings
in central Golcuk are shown in Figure 8. GL-CPT2 was performed about 200 m west of the
subsidence zone and GL-CPT1 was performed about 500 m west of the subsidence zone, in an
area comprised of older geologic materials (Figure 6). All of the soundings indicate loose,
liquefiable sand layers in the top 10 m, with some interbedded layers of silty clay. The similarity
in the soundings is surprising, because liquefaction was not observed near GL-CPT1 or GL-
CPT2. However, soundings GL-CPT1 and GL-CPT2 indicate thicker and more frequent silty
clay layers near the surface, which may have masked any evidence of liquefaction. Additionally,
the offshore slope in this area was flatter than at the delta nose in the subsidence area, suggesting
that steep offshore slopes near GL-CPT3 may have exacerbated the effects of liquefaction and
resulted in more intense surface effects.
Yenikoy is located in the Golcuk basin east of the city of Golcuk on the eastern margin of
a large alluvial plain and deltaic peninsula that extends northward into Izmit Bay. Yenikoy is
underlain by interfingering Holocene alluvial deposits (Figure 6) consisting of coalescing
individual delta fans. The streams crossing the alluvial plain and older delta fans are relatively
small and have sources within the distant hills south of the coastline. The sediment carried by
these streams is relatively fine-grained, and the soils underlying the Yenikoy site have significant
percentages of silt and clay.
A large zone of coastal subsidence and sea inundation occurred along the shoreline in
Yenikoy, with shoreline retreat ranging from about 50 to 350 m in this area. Both the shoreline
and inundation zone are subparallel to the trend of the Golcuk normal fault, located about 1.5 km
to the southwest. Bathymetry data  indicate that the offshore slope within 150 m of the pre-
earthquake shoreline is flat, averaging between 2 and 4 degrees. Post-earthquake bathymetry
suggest that the seafloor remained relatively flat, but possibly steepened somewhat near the
shoreline. It is possible that tectonic tilting and minor secondary displacements on other
shoreline-parallel normal faults actually steepened the topography in this area. Watermarks on
subsided structures that formerly were at the shoreline suggest that the shoreline subsided
between 1 and 3 m. One CPT sounding was performed along the post-earthquake shoreline coast
in Yenikoy (Figure 6). The CPT data indicate that the top 17 m consists predominantly of silty
clay and clay (qc ~ 0.5 – 1.0 MPa, Ic ~ 3), with some dense sand layers (qc ~ 30 MPa). Some of
the silty and clayey layers contained shells and appear to be marine deposits that are
interfingered with the fine-grained delta sediments. None of the soils encountered at Yenikoy
are liquefiable, and no surficial evidence of liquefaction or ejected sand/silt was found in this
area. As a result, the subsidence at Yenikoy appears to be solely attributed to global tectonic
downdropping along the Golcuk normal fault, with possible additional vertical movement
associated with fault block tilting and secondary normal faults at the shoreline. The relatively
flat pre- and post-earthquake bathymetry profiles suggest that a large-scale, near shore slope
failure did not occur.
The town of Seymen is situated near the southeastern margin of the Golcuk pull-apart
basin (Figure 6), approximately 5 km east of central Golcuk. Coastal subsidence and sea
inundation, as well as distinct lateral spreading in some areas, occurred along a 1 km segment of
the coast. The distressed area in Seymen coincides with the latest Holocene prograding part of a
delta that has formed a small peninsula into Izmit Bay at the mouth of a creek.
The two main areas of subsidence and ground failure in Seymen are the MKE Scrapyard
and an adjacent tea garden. The MKE Scrapyard is located at the nose of the Holocene delta fan
and shoreline retreat as large as 70 m occurred here, causing submergence of the shoreline
facilities within the MKE Scrapyard. A prominent 0.5 to 1-meter high, curvilinear scarp formed
the post-earthquake shoreline, and a secondary zone of cracking occurred within a 150 m wide
zone inland of the post-earthquake shoreline, across the nose of the delta fan. The lateral
displacements across cracks and joints in a concrete wall that traverses the zone of secondary
cracking were measured and indicated about 1.5 m of cumulative onshore lateral displacement in
the zone behind the offshore slide scarp.
The pre-earthquake cross-section at the nose of the delta fan at the MKE Scrapyard was
developed from bathymetry data provided by ZETAS . The maximum offshore slope angle
at the nose of the delta fan at the scrapyard was 22 degrees and the height of this slope was about
10 m. The maximum slope occurred approximately 25 m offshore, while the slope was flatter
further offshore with an angle of 6 degrees. To develop the post-earthquake offshore geometry,
water depth was measured using a sonar device (Lowrance X-85) that is typically used for
recreational boating applications. These data indicate that the inclination of the post-earthquake
offshore geometry at the delta nose is about 3 degrees.
At the MKE Scrapyard, one CPT (SY-CPT5) was performed at the nose of the delta fan
in the area adjacent to the coastal slide and one CPT (SY-CPT4) and one SPT (SY-SPT1) were
performed approximately 150 m inland (Figure 6), along the wall where cracking was mapped.
Inland, the subsurface soils consist mainly of medium stiff, highly plastic clays and silts (PI 20-
40) including both alluvial and marine sediments. Some thin silty sand alluvial layers were
identified in the top 15 m, and a layer of medium dense silty sand (N1,60~17, 15-30% fines) was
encountered between 15 and 20 m. Liquefaction analyses using a PGA of 0.3 g indicate that the
factor of safety against liquefaction for this deeper silty sand layer is close to 1.0. The results
from the CPT performed at the nose of the delta fan along the shoreline at the scrapyard (SY-
CPT5) are shown in Figure 9. These data indicate that the top 10 m consist of loose, deltaic silty
sand (qc ~ 5 MPa, Ic ~ 2.2) with interbedded clay seams. The underlying soils consist mainly of
clays, which are in part marine in origin.
The liquefaction susceptibility of the soils at the nose of the delta fan at the MKE
Scrapyard (SY-CPT5) was assessed using the procedures outlined previously and a PGA of 0.3
g. These analyses reveal that at the nose of the delta fan most of the soil in the top 10 m of the
profile is liquefiable, with a factor of safety less than 0.5 . The elevation of this liquefiable
layer coincides with the location of the steep offshore slope, indicating that a liquefaction-
induced slope failure that daylighted at the base of the steep delta nose caused the localized sea
inundation at the MKE Scrapyard. It is possible that the failure occurred progressively, initiating
at the steep delta nose and propagating landward. The wide extensional zone behind the main
scarp is similar to the extensional zone observed behind the slide scarp at Degimendere, and
similarly suggests that this zone was quasi-stable and experienced limited lateral movement into
the slide zone possibly along partially liquefied layers.
The Seymen tea garden is located approximately 300 m west of the MKE Scrapyard,
along the edge of the Seymen delta. Deformations characteristic of classical lateral spreading
were observed, but with only minor shoreline retreat and sea inundation. Pre-earthquake
bathymetry in the tea garden area revealed an offshore slope of less than 2 degrees . Post-
earthquake water depth measurements performed as part of this study revealed a similar offshore
geometry, indicating that the ground failure did not extend far offshore or did not involve large-
scale displacements of the seafloor. The ground cracking at the tea garden site was not mapped
during post-earthquake reconnaissance efforts, but a photograph taken from the air after the
earthquake  was used to estimate the cumulative lateral width of ground cracking. Using
landmarks (e.g., sidewalks, curbs) in the photograph that were measured in the field during this
study, the cumulative lateral crack width was roughly estimated to be between 2 and 4 m.
One CPT and one SPT were performed along the shoreline at the Seymen tea garden.
The data from these CPT and SPT (SY-CPT2, SY-SPT2) are shown in Figure 10. The CPT and
SPT data indicate a 5-m thick layer of soft, silty clay at the surface, underlain by over 15 m of
loose silty sand (qc ~ 4-6 MPa, N1,60 ~ 6-12) with interbedded layers of silty clay. Below the
silty sand layer is highly plastic clay (PI 30-40). Liquefaction analyses using a PGA of 0.3 g
reveal that 5 to 10 m of the silty sand layer is highly liquefiable, with a factor of safety less than
0.5 . This liquefiable layer is the most likely cause of the observed lateral spread cracking at
the tea garden. The 5-m thick silty clay layer at the surface presumably prevented any soil ejecta
from surfacing, and thus sand boils were not observed after the earthquake.
The lateral spread displacement estimated from the photograph at the Seymen tea garden
was compared with the predicted lateral displacement from the empirical relationship of Youd et
al.  for sloping ground. The empirical relationship is a function of earthquake magnitude
(M), distance to the fault (R), ground slope (S), and the thickness (T15), fines content (F15), and
median gain size (D5015) of the liquefiable materials with N1,60 less than 15. These parameters
were estimated for the tea garden site using the CPT and SPT data, grain size distributions from
split spoon samples, and field observations (Table 1). Using these parameters, the Youd et al
 relationship for sloping ground predicts displacements between 7 and 14 m, which are much
larger than observed. However, the ground accelerations during the Kocaeli earthquake were
smaller than expected for a Mw = 7.4 event . This effect can be taken into account by using
an attenuation relationship for acceleration (e.g., Abrahamson and Silva ) to compute the
equivalent distance for a specified acceleration level and using this distance in the Youd et al.
 relationship. For Mw = 7.4 and the estimated PGA of 0.3 g, the equivalent distance is about
20 km. Using 20 km in the Youd et al.  relationship results in a displacement prediction of
between 1.0 and 2.0 m, which is much closer to the estimated field displacements. The need to
use equivalent distance in the Youd et al.  relationship arises because these relationships use
magnitude and distance as indicators of earthquake intensity, rather than directly using peak
ground acceleration. It may be useful in the future to use peak ground acceleration directly as an
indicator of intensity in predictive relationships for lateral spread displacement rather than using
magnitude and distance as indirect indicators of intensity.
SAPANCA PULL-APART BASIN
The Sapanca pull-apart basin, which encompasses Sapanca Lake, is located within a
right-lateral step between the Sapanca and Sakarya fault segments (Figure 1). Right-lateral
displacement of 3 m was observed along the Sapanca fault segment where the fault enters the
north margin of the lake. Post-earthquake bathymetry indicates that the Sapanca fault segment
extends into the lake and displayed some normal displacement during the earthquake.
Significant right-lateral displacement (2 to 5 m) and minor vertical displacement (0.25 to 0.5 m)
were measured on the Sakarya fault segment on the southeast shoreline of Lake Sapanca.
Normal fault displacements were also documented in this area during the 1967 Mudurnu Valley
The topographic depression caused by the Sapanca pull-apart basin serves as a
depocenter for sedimentation, with Holocene delta fans prograding into the lake at the mouths of
many creeks. The severe subsidence observed along the margins of Lake Sapanca during the
Kocaeli earthquake was mainly concentrated at the nose of delta fans. The most significant
failure occurred along the south margin of the lake at the Hotel Sapanca, and a smaller failure
occurred along the north margin of the lake within the town of Esme . Both of these sites are
situated on Holocene delta fans. The Hotel Sapanca site experienced as much as 50 m of
shoreline retreat , and geotechnical investigations by other researchers found liquefiable silty
sand and sand in the top 10 m of the site with qc < 10 MPa, N60 < 10 , and a factor of safety
against liquefaction of less than 0.5.
The Esme site is located on a Holocene delta fan along the north shore of Lake Sapanca
(Figure 11). This delta is located near the base of a bedrock hill front, which is approximately
500 m to the north. Bedrock is mapped as tertiary sedimentary rock, and the delta sediments are
derived from erosion of this bedrock. The close distance between the hill front and delta reduces
the amount of reworking and grain size sorting that occurs in the delta sediments. As a result,
the near surface delta deposits are relatively coarse, consisting mainly of sand with silt and
gravel lenses. The active portion of the delta has constructed a narrow nose that extends into
Lake Sapanca, and this is the area that experienced inundation during the Kocaeli earthquake.
The inundated area extended approximately 50 m along the coast and shoreline incursion was
estimated as approximately 35 m back from the pre-earthquake shoreline. Onshore ground
cracking was purely translational and extended approximately 150 m inland, but no soil ejecta
was observed at the site . The normal faulting that was observed in post-earthquake
bathymetry data occurred approximately 750 m offshore from Esme and most likely did not
contribute to the failure.
Pre-earthquake bathymetry data were not available for Lake Sapanca, and therefore the
offshore slope at Esme before the earthquake is unknown. The post-earthquake geometry of the
Esme site was developed using water depth measurements collected as part of this study. These
data indicate that the post-earthquake offshore slope of the delta nose near the shore is relatively
flat, with a slope angle of approximately 7 degrees. Approximately 25 m offshore the slope
angle increases to 15 degrees. The onshore portion of the delta exhibits a low gradient, on the
order of 2 to 4 degrees.
The site investigation at Esme included one SPT boring and three CPT soundings (Figure
11). One SPT boring (ES-SPT1) and one CPT sounding (ES-CPT3) were located at the
shoreline within the ground failure zone, while another CPT (ES-CPT4) was located
approximately 100 m inland along the creek that is actively depositing the delta sediments. The
third CPT sounding was performed 1.5 km west of the failure in a non-failed portion of the delta
fan. The in situ data from ES-SPT1 and ES-CPT3 are shown in Figure 12. The top 10 m
consists of loose sand to silty sand (qc ~ 3-6 MPa, N60 ~ 5-13), with the fines content varying
from less than 10% to about 30%. This sand layer is underlain by a 5-m thick nonplastic silt
layer and a medium-dense, silty sand layer (qc ~ 20 MPa, N60 ~ 20). Stiff, silty clay (PI 10-20) is
found below 20 m. The CPT performed inland (ES-CPT4) indicated similar subsurface
conditions, but the layers were somewhat thinner due to the effect of topography on deposition.
The CPT and SPT data were used to assess the liquefaction susceptibility of the soils at
Esme. No strong motion station was situated at Esme, but the Sakarya station (SKR) is located 5
km east of Esme. The SKR station is situated on shallow soil and recorded a PGA of 0.4 g.
Based on this recording and considering the effect of the soil conditions on ground shaking, the
PGA at Esme was estimated as between 0.3 and 0.5 g. These values were used to estimate the
earthquake-induced CSR and the CRR values were computed from both SPT and CPT data. The
liquefaction susceptibility analysis indicates that most of the top 10 m of silty sand is highly
liquefiable with factors of safety well below 0.5. The medium dense sand between 15 and 20 m
also is liquefiable (qc ~ 20 Mpa, N60 ~ 20) with factors of safety between 0.75 and 1.0. These
liquefiable layers are the most likely the cause of the coastal failure and sea inundation at Esme.
Severe coastal subsidence and sea inundation were observed within the coastal pull-apart
basins during the 1999 Kocaeli earthquake in Turkey. Geotechnical and geologic site
investigations were performed at several coastal sites to assess the failure mechanisms that
caused the observed subsidence. The collected information was used to evaluate the two most
likely causes of the onshore sea inundation: tectonic subsidence associated with normal faulting
and liquefaction-induced ground failure.
A summary of the observations from the coastal sites described in this paper is given in
Table 2 and discussed below. The Degirmendere site, within the Karamursel pull-apart basin,
experienced up to 75 m of shoreline retreat at the nose of a delta fan. Due to the active
progradation of the delta, the offshore slope at Degirmendere was relatively steep (18 degrees).
The near-surface soils (top 10 m) were liquefiable but could not explain the observed deep-
seated slope failure. Slope stability analyses indicated a factor of safety of 1.05 for the slope,
assuming that a deeper soil layer (25-30 m deep) liquefied. These analyses indicate that
liquefaction of deeper soils most likely caused the failure at Degirmendere.
Within the Golcuk pull-apart basin, four sites of coastal subsidence were investigated. In
central Golcuk, sea inundation extended as far as 300 m inland at the nose of an active delta fan.
This area coincides with an area close to the Golcuk normal fault that experienced approximately
1.5 m of vertical fault displacement. The pre-earthquake offshore slope was as steep as 20
degrees and highly liquefiable soils were encountered in the top 10 m of the site. Consequently,
it appears that both tectonic subsidence and liquefaction played a role in the sea inundation in
central Golcuk. In Yenikoy, the sea inundation reached as far as 350 m inland. However, no
liquefiable soils were encountered here and the offshore slope was flat (less than 4 degrees).
Therefore, the coastal subsidence in Yenikoy is solely attributed to normal faulting within the
Golcuk pull-apart basin. In Seymen, the MKE Scrapyard experienced 70 m of sea inundation at
the nose of an active delta fan. The offshore slope was as steep as 22 degrees and liquefiable
soils were found in the top 10 m. Only minor tectonic subsidence was observed in this area, and
therefore, the sea inundation is mostly attributed to a liquefaction-induced slope failure at the
delta nose. At the tea garden in Seymen, lateral spread deformations were observed with only
minor sea inundation. The site is situated away from the nose of the active delta fan in Seymen,
with an offshore slope of less than 3 degrees. Liquefiable soils were identified in the top 15 m of
the deposit at the tea garden and are the most likely cause of the lateral spreading. Lateral spread
deformations estimated from Youd et al.  agreed well with those estimated in the field, after
accounting for the smaller than expected ground motions from the Kocaeli earthquake.
In the Sapanca pull-apart basin, the two main sites of coastal failure (Esme, Hotel
Sapanca) were situated at the nose of active delta fans. The Esme site experienced up to 35 m of
shoreline retreat, while the Hotel Sapanca experienced up to 50 m of inundation. Because no
pre-earthquake bathymetry was available for Lake Sapanca, it was not possible to estimate the
offshore slopes at either of these sites. However, highly liquefiable soils were encountered at
both sites and are the most likely cause of the failures and sea inundation.
The field investigation summary in Table 2 indicates that the largest inland extent of sea
inundation and subsidence occurred at sites that experienced more than a meter of tectonic
subsidence due to normal faulting. In the case of central Golcuk, liquefaction may have
enhanced the extent of subsidence. The sites with moderate sea inundation (25 to 100 m of
shoreline retreat) predominantly were situated at the nose of delta fans with quasi-stable steep
slopes and liquefiable soils. Most of these sites did not experience significant tectonic
subsidence, but did undergo localized deformation. Finally, the site with the least sea inundation
(Seymen-Tea Garden) was not situated at the nose of a delta fan, but classic lateral spread
deformations occurred due to the presence of liquefiable soils and sloping ground.
The collected field observations, field data, and associated analyses indicate that both
tectonic subsidence and liquefaction-induced slope failures can cause significant sea incursion in
coastal pull-apart basins. During the Kocaeli earthquake, tectonic subsidence caused up to 350
m of coastal retreat and subsidence. Additionally, liquefaction-induced coastal slope failures
caused shoreline retreat of between 25 and 75 m, while lateral spreading caused only minor sea
inundation. Whereas tectonic subsidence was broadly distributed within the pull-apart basins,
particularly near to the basin-bounding Golcuk normal fault, liquefaction and slope failures were
largely restricted to the most active prograding parts of Holocene delta fans. It is important to
note that these prograding portions of delta fans are readily identifiable by geologic mapping and
aerial photograph analyses. Conversely, subsurface conditions encountered in SPT borings and
CPT soundings often appeared to be relatively similar in both failed zones in latest Holocene
deposits, and adjacent non-failed zones in older Holocene or even late-Pleistocene deposits. This
comparison suggests that careful geologic mapping and age estimation of units may be very
useful in distinguishing problematic areas.
Based on this study, when siting facilities in coastal areas, it is important to consider
coastal failures that can cause tens to hundreds of meters of shoreline retreat. In these cases, a
study should focus on identifying both: (1) the extent of extensional pull-apart basins and the
locations of basin-bounding normal faults, as well as (2) more localized areas susceptible
liquefaction and liquefaction-induced slope failures, including actively prograding delta fans.
Financial support was provided by the United States Geological Survey under grants
01HQGR0042 and 02HQGR0059 to the University of Texas and grants 01HQGR0043 and
02HQGR0030 to William Lettis and Associates. This support is gratefully acknowledged. The
site investigation was performed by ZETAS Corporation of Istanbul, Turkey with the help of Dr.
Turan Durgunoglu and Mr. Turhan Karadayilar. Jason Holmberg of William Lettis & Associates
prepared many of the figures for this paper.
1. Lettis, W., Bachhuber, J.L., Witter, R., Brankman, C., Randolph, C.E., Barka, A., Page,
W.D., Kaya, A. Influence of the Releasing Stepovers on Surface Fault Rupture and Fault
Segmentation: Examples from the 17 August 1999 Izmit Earthquake on the North Anatolian
Fault, Turkey. Bulletin of the Seismological Society of America 2002; 92(1): 19-42.
2. Barka, A.A. and Kadinsky-Cade, K. Strike-Slip Fault Geometry in Turkey and its Influence
on Earthquake Activity. Tectonics 1988; 7; 663-684.
3. Witter, R.C., Lettis, W.R., Bachhuber, J., Barka, A., Evren, E., Cakir, Z., Page, W.D.,
Hengesh, J., and Seitz, G. Paleoseismic Trenching Study Across the Yalova Segment of the
North Anatolian Fault, Hersek Peninsula, Turkey. In: Barka, A., Kozaci, O., Aykuz, S. and
Altunel, E., editors. The 1999 Izmit and Duzce Earthquakes Preliminary Results. 2000.
Istanbul Technical University, Turkey, pp. 329-339.
4. Turkish Navy Turkiye, Marmara Denizi, Izmit Limani. Department of Hydrography and
Oceanography (in Turkish), 1997.
5. American Society for Testing and Materials. ASTM D1586-99 Standard Test Method for
Penetration Test and Split-Barrel Sampling of Soils. Annual Book of ASTM Standards, West
Conshohocken, PA, 2003.
6. Cox, B. Shear Wave Velocity Profiles at Sites Liquefied by the 1999 Kocaeli, Turkey
Earthquake. M.S. Thesis, Utah State University, 2001: 274 pp.
7. Karatas, I. Evaluation of Ground Failure in Pull-Apart Basins during the 1999 Kocaeli
Earthquake. M.S. Thesis, University of Texas at Austin, 2002: 238 pp.
8. Youd, T.L., R.D. Andrus, I. Aragon, G. Castro, J. T. Christian, R. Dobry, W.D.L. Finn, L.F.
Harder Jr., M.E. Hynes, K. Ishihara, J.P. Koester, S.S.C. Liao, W.F. Marcuson, III, G.R.
Martin, J.K. Mitchell, Y. Moriwaki, M.S. Power, P.K. Robertson, R.B. Seed and K.H.
Stokoe, II. Liquefaction Resistance of Soils: Summary Report from the 1996 NCEER/NSF
Workshops on Evaluation of Liquefaction Resistance of Soils. ASCE Journal of
Geotechnical and Geoenvironmental Engineering 2001;127(10): 297-313.
9. Robertson, P.K. and Wride, C.E. Evaluating Cyclic Liquefaction Potential Using the Cone
Penetration Test. Canadian Geotechnical Journal 1998; 35: 442-459.
10. Rathje, E.M., Idriss, I.M., and Somerville, P. Strong Ground Motions and Site Effects. 1999
Kocaeli, Turkey, Earthquake Reconnaissance Report in Earthquake Spectra 2000; 16(A): 65-
11. Wright, S.G. UTEXAS4 – A Computer Program for Slope Stability Calculations. Shinoak
Software, Austin, Texas, 1999.
12. Seed, R.B., and Harder, L.F. SPT-Based Analysis of Cyclic Pore Pressure Generation and
Undrained Residual Strength. In: Proceedings of the H.B. Seed Memorial Symposium,
BiTech Publishing, Vancouver, B.C., Canada, 1990; 2: 351-376.
13. Baziar, M.H. and Dobry, R. Residual Strength and Large-Deformation Potential of Loose
Silty Sands. ASCE Journal of Geotechnical Engineering 1995; 121(12): 896-906.
14. Makdisi, F., and Seed, H.B. Simplified Procedure for Estimating Dam and Embankment
Earthquake-Induced Deformations. ASCE Journal of Geotechnical Engineering 1978;
15. Bardet, J.P. and Seed, R.B. Soil Liquefaction, Landslides, and Subsidence. 1999 Kocaeli,
Turkey, Earthquake Reconnaissance Report in Earthquake Spectra 2000; 16(A): 141-162.
16. ZETAS. DEMPORT Liman Yatirimlari ve Isletmeciligi A.S., Izmit Yenikoy Limani Zemin
ve Temel Muhendisligi Etudleri Degerlendirme Raporu. Zetas Zemin Teknolojisi A.S.,
Istanbul, Turkey (in Turkish), 1995.
17. Youd, T.L., Hansen, C.M., and Bartlett, S.F. Revised Multilinear Regression Equations for
Prediction of Lateral Spread Displacement. ASCE Journal of Geotechnical and
Geoenvironmental Engineering 2002; 128(12): 1007-1017.
18. Abrahamson, N.A., and Silva, W.J. Empirical Response Spectral Attenuation Relations for
Shallow Crustal Earthquakes. Seismological Research Letters 1997; 68(1): 94-127.
19. Ambraseys, N.N., Zatopek, A. The Mudurnu Valley, West Anatolia, Turkey, Earthquake of
22 July 1967. Bulletin of the Seismological Society of America 1969; 59(2): 521-589.
20. Youd, T.L., Cetin, K.O., Bray, J.D., Seed, R.B.Durgunoglu, T., and Onalp, A. Geotechnical
Investigation at Lateral Spread Sites. Data accessed from
Table 1. Parameters used in Youd et al. (2002) relationship to predict
lateral spread displacement at the tea garden site in Seymen.
R (km) 1, 20
S (%) 1-3
T15 (m) 5-10
F15 (%) 20
D5015 (mm) 0.4
Table 2. Summary of field investigations of coastal failures.
Maximum Maximum Vertical
inland extent Delta Nose? pre-event tectonic
of inundation slope subsidence
Degirmendere 75 m Yes 18 0.5-1.0 None
300 m Yes 20 ~0.5 1.5 m
Yenikoy 350 m No 4 N/A 1-3 m
70 m Yes 22 <0.5 Minor
<5m No 3 <0.5 Minor
Esme 35 m Yes ? <0.5 None
Hotel Sapanca 50 m Yes ? <0.5 Minor