• Triggered slip: observations of the 17 August 1999 Izmit ...

  • FileName: Wright_Izmit_GRL2001.pdf [read-online]
    • Abstract: Radar interferogram for the Izmit earthquake (data copyright ESA) revealing the surface displacements, measured in ... caused by the change in static stress from the Izmit earth ...

Download the ebook

Triggered slip: observations of the 17 August 1999
Izmit (Turkey) earthquake using radar interferometry
Tim Wright, Eric Fielding1 , Barry Parsons
Department of Earth Sciences, University of Oxford, Parks Road, Oxford, UK.
Abstract. We use Synthetic Aperture Radar interferome- Procedures for determining earthquake source parame-
try (InSAR) to map the displacement field of the 17 August ters from radar interferometric data are well established
1999 Izmit earthquake, which largely conforms to that pre- [Massonnet and Feigl , 1998; B¨rgmann et al., 2000]. We
dicted for an elastic upper crust. We determine the earth- use a hybrid Monte-Carlo, downhill simplex inversion pro-
quake source parameters and show that slip continues far- cedure [Wright et al., 1999] to determine best-fitting model
ther west than the mapped fault ruptures. We also show parameters. This procedure minimises the misfit between
that additional sub-surface displacements occurred on par- our interferometric measurements of range change, sampled
allel strands of the North Anatolian Fault Zone. We argue at discrete locations, and those predicted by an elastic dis-
that this was caused by changes in static stress accompa- location model[Okada, 1985] of the earthquake. We fix the
nying the mainshock, or by the dynamic release of regional fault plane location to coincide with the surface rupture on-
stresses. shore, solving for the offshore location. We aim to fit the
data with as simple a model as possible, because the reduc-
Introduction tion in misfit obtained from more complex models is small.
We perform the inversion procedure twice, first using inter-
The North Anatolian Fault Zone accommodates the west- ferograms a,c,d and a second time using b,c,d. The differ-
ward motion of Turkey relative to Eurasia by right-lateral ence between the two inversion results gives us an indication
shear and regularly generates large earthquakes [Ambraseys, of the errors in our model parameters (Table 2) resulting
1970; Barka, 1996] of which the Izmit earthquake was the from atmospheric and orbital effects: slip magnitudes agree
largest in 60 years. The surface rupture (Fig. 1) was mapped to within 15%, with the ratio of horizontal to vertical dis-
for 100 km from D¨zce in the east to the Gulf of Izmit. No
u placements (rake) being almost identical. The maximum
rupture was observed on or west of the prominent Hersek depth is ∼20 km in the two cases.
delta (29.5o E) [Barka, 1999], but aftershocks continue for
another 50 km beyond it. Estimates of future seismic haz-
ard [Parsons et al., 2000] in the Sea of Marmara depend
crucially on the fault location and magnitude of slip in the 28o 29o 30o 31o 32o
Gulf of Izmit, because if little slip occurred west of Hersek,
faults there were brought closer to failure. 25 mm yr -1 157 50 km
50 km
Source parameters from InSAR
Istanbul Track
41o 64
We have constructed interferograms using ERS1 and GI
LS Nov 99
ERS2 35-day pairs spanning the earthquake (Fig. 2, Ta- Sea of Marmara 1944
Iz 1957
ble 1a,b), as well as two other interferograms on adjacent HD Go
tracks with longer temporal separations (Table 1c,d). The ult
Mudurnu Valley
LI Iznik Fa
35-day interferograms cover essentially the same interval and
we would expect the range changes observed in both to be
identical. There are, however, differences of ±50 mm of 40o
range change, probably resulting from changing atmospheric
conditions, with a possible small (less than 30 mm) contri- Fig 2
bution from orbital errors. The other interferograms are
less coherent but provide valuable information about the 17 August 1999 rupture Iz - Izmit LI - Lake Iznik
lateral extent of faulting. In particular, the combination of Faults with triggered slip Go - Golcuk LS - Lake Sapanca
Previous fault ruptures Du - Duzce GI - Gulf of Izmit
descending interferogram c with ascending a or b gives mea-
Mapped active faults HD - Hersek Delta
surements in two different look directions, and hence two
components of the displacement vector at the western end
Figure 1. Topographic and tectonic map showing the extent of
of the rupture. the 17 August 1999 Izmit earthquake rupture. Seismic activity
(Kandilli Obs.) in the interval between our radar image acquisi-
1 Also at Jet Propulsion Lab, Caltech, USA. tions (13 August to 16 September) is shown as black dots with
events larger than Md 4.5 depicted by circles. The star indicates
Copyright 2001 by the American Geophysical Union. the epicentral location (Kandilli Obs.) with focal mechanism from
the Harvard CMT solution. Arrows show the GPS-determined
Paper number 2000GL011776. interseismic velocities relative to Eurasia[Reilinger et al., 1997].
0094-8276/01/2000GL011776$05.00 The inclined boxes delimit the coverage of our interferograms
29 30' 30 00' 30 30' 31 00' 29 30' 30 00' 30 30' 31 00' 29 30' 30 00' 30 30' 31 00'
a b c
25 km 25 km 25 km
41 00'
40 30' 3a-c
40 00'
Figure 2. a, Radar interferogram for the Izmit earthquake (data copyright ESA) revealing the surface displacements, measured in
the satellite’s line-of-sight, in the 35-day period between the two image acquisitions (Table 1a). Each interference fringe is equivalent
to 28 mm of displacement in the satellite line-of-sight, or approximately 70 mm if caused by pure horizontal motion. A correction
for the small topographic contribution was made with a DEM constructed from ERS tandem interferograms and GTOPO30 data
[Fielding et al., 1999] and the interferogram was smoothed using a power spectrum filtering algorithm [Goldstein and Werner , 1998].
Red lines are the mapped surface rupture [Barka, 1999] and the dashed lines are previously mapped segments of the North Anatolian
Fault [Saroglu et al., 1992]. b, Synthetic interferogram calculated using the elastic dislocation model described in the text that
intersects the surface at the location shown by the thick black lines. c, Residual interferogram, obtained after subtracting b from a.
Our fault model (Table 2, Fig. 2) shows the western limit orbital error evident in the differences between the 35-day
of rupture to be ∼15 km west of the Hersek Delta, where over interferograms. An exception to this is in one region near
1.5 m of slip is required to obtain a good fit to the interfero- Izmit where a large aftershock (13/9/99, Md 5.8, Kandilli
grams. The rupture location passes just north of the delta, Obs.) occurred within the time span of the interferograms,
although the absence of data near the fault means we cannot contaminating the signal of the 17 August event.
be certain of this location to better than a few kilometres.
Slip was a maximum of 4.5–5 m on the segment north of
Gol¸uk and over 4 m between Izmit and Lake Sapanca, but
Triggered Slip
is only 1.5–2 m where the rupture is 5–10 km north of seg- By removing our model of coseismic deformation from the
ments that ruptured in 1967. The overall residual phase interferogram, we are able to detect small deformation sig-
falls within the level of noise caused by atmospheric and nals associated with faulting away from the Izmit rupture.
Distance from South (km) Distance from South (km)
a b c -20 0 20 B’
10 km 15
B’ B’
B 0
d C’ e C’ f C’
Range Change
bar mm
Shear Stress
2 28
C -2 C 0 -20 20 C 0
Range Displacment (mm)
Figure 3. Evidence for triggered fault slip on Mudurnu Valley (a-c) and Iznik (d-f) faults. a,d, Static shear stress changes (vertical,
E-W faults) calculated from our best-fitting elastic dislocation model of the Izmit earthquake, which intersects the surface at the red
lines. Negative shear stress changes imply left-lateral shear, with shading from the DEM. Dashed lines are mapped segments of the
North Anatolian Fault [Saroglu et al., 1992]. b,e Residual interferogram (Fig. 2). The tight color gradients at the center of profiles
B–B’, C–C’ coincide with the location of active faults. c,f Range change profiles along B–B’(c), C-C’(f). Red line corresponds to the
residual interferogram shown in b,e (ERS2 to ERS2, linear trend removed from e); Blue – residual of ERS1 interferogram; Black –
range change calculated using an elastic dislocation model. The dislocation models shown assume pure horizontal (i.e. left-lateral)
slip (Table 3).
Table 1. Details of ERS SAR data ( c ESA). Table 3. Triggered slip fault parameters [0o rake = left-lat;
90o = vertical; 150,160o = oblique right-lat]
ERS Track/Framea Date 1/2 Orbit 1/2c ha c
Rake Slip Depth Range M0 (Nm)
a) 157/815,797(A) 13-8-99/17-9-99 22556(2)/23057(2) 1680
b) 157/815,797(A) 12-8-99/16-9-99 42229(1)/42730(1) 560 Mudurnu Valley 0o 10cm 0.6–1 km 1.7×1016
c) 64/2786(D) 20-3-99/15-10-99 20459(2)/43138(1) 1270 (Str. 280o , Dip 50o , 90o 1.9cm 0.6–3.5 km 2.4×1016
d) 386/810(A) 9-8-98/3-10-99 17274(2)/23286(2) 348 Len. 10 km) 150o 7cm 0.6–15 km 4.5×1017
Iznik Fault 0o 20cm 2.5–3.5 km 2.0×1017
a Ascending(A)/descending(D) pass; b Orbit numbers for ERS-1 (1) (Str. 260o , Dip 90o , 90o 1.4cm 2.5–12.5 km 1.4×1017
or ERS-2 (2); c Scene center altitude of ambiguity (metres). Len. 30 km) 160o 8cm 2.5–20 km 1.4×1018
Generally, the residual interferogram (Fig. 2) varies very
smoothly. However, across the Mudurnu Valley and east :
the right-lateral slip. By varying the rake we find that, to
of Lake Iznik, strong phase gradients remain in the residu- produce the observed range change, rakes must be less than
als that are coincident with mapped active faults (Fig. 3). 165o and 155o at the Iznik and Mudurnu Valley respectively,
These signals are very similar in both 35-day interferograms, and slip must extend over most of the seismogenic layer.
but could not be observed in the other datasets due to tem- Our interferograms do not distinguish between slip oc-
poral decorrelation. Profiles of range change constructed curring in aftershocks and aseismic triggered slip up to a
across the Mudurnu Valley and Iznik faults (Fig. 3c,f) show month after the earthquake. However, earthquakes larger
that there is a range change decrease of 20–25 mm from than Mw 4.7–5 would be required to cause sufficient defor-
south to north across both faults. mation and there are no aftershocks, in these locations, that
These apparent range changes are unlikely to be the re- are sufficiently large (Fig. 1). In addition, the deformation
sult of atmospheric artefacts because of the spatial coin- has a ratio of fault slip to length of ∼ 5 × 10−6 or less, an or-
cidence with active faults, and because atmospheric phase der of magnitude smaller than the typical ratio for seismic
variations observed in the difference between the 35-day in- events [Pegler and Das, 1996]. This suggests that, unless
terferograms have longer wavelengths. Satellite geometry this slip occurred preseismically or at the same time as the
rules out topographic artefacts: topographic errors of over mainshock, it was the result of aseismic triggered slip.
4800m (ERS2) and 660m (ERS1) in the Mudurnu Valley,
and 850m/300m on the Iznik strand, are required to gener-
ate 20 mm of range change.
Discussion and Conclusions
We model the displacements on the Mudurnu Valley and There are several explanations for triggered slip. Coseis-
Iznik faults using elastic models to determine best-fitting mic movement during the 17 August mainshock changed the
fault parameters. In both cases, continuity of the phase static stress in the area [Parsons et al., 2000] by an amount
across the faults implies that slip was subsurface. InSAR that we can calculate using our slip model (Fig. 3a,d). At the
detects only the component of displacement in the satellite locations of our profiles, left-lateral shear stress increases of
line-of-sight, hence a variety of fault rakes can produce the 2.0 and 0.8 bars occurred on the Mudurnu Valley and Iznik
observed range change. Solutions include pure left-lateral faults. For these to create left-lateral slip, they would have
strike-slip, pure vertical (upthrown to north) and right- to be larger than the right-lateral stresses accumulated in-
lateral with some vertical (Table 3), but the range changes terseismically. While this is possible for the Mudurnu Valley
could not have been caused by pure right-lateral slip. If the fault, which ruptured in 1967, it seems unlikely on the Iznik
slip is left-lateral, its depth is shallow (0.6–1 km in Mudurnu fault which has been seismically quiescent for the last 500
Valley). Less slip is required for the purely vertical solutions, years [Ambraseys and Jackson, 2000]. However, it is possible
but over a larger depth range (0.6–3.5 km in Mudurnu Val- that interseismic stress is released by continuous deforma-
ley). Oblique right-lateral slip is also possible, provided that tion in weak shallow layers. Any stress changes due to the
the vertical displacement causes a greater range change than Izmit event would then dominate and cause left-lateral slip.
Table 2. Source parameters of the Izmit Earthquake from InSAR. Where two parameter values appear, they are the result of
separate inversions on interferograms a,c,d and b,c,d (Table 1). Other parameters are held fixed.
Seismica Geodetic (6 segments, starting from West)
Scarp Latitude 41.01o b
40.730o 40.744o 40.739o 40.726o 40.708o 40.728o
Scarp Longitude 29.97o b
29.450o 29.630o 29.812o 30.039o 30.347o 30.813o
Length / km — 20.1 10.5 20.3 18.2 34.2 32.8
Total Length = 136.1
M0 /1018 Nmc 288 Total M0 = 265,253
Slipd/ m — 1.7,1.6 (—)d 2.5,2.3 (—)d 4.9,4.7 (—) 4.6,4.4 (3-4)e 2.1,1.8 (0.5-4.5) 1.7,1.4 (1.5)
Strike 91o 84,264o 91,271o 96,96 o
277,97o 276,96o 249,249o
Dip 87o 88,84o 86,88o 86,87o 88,88o 81,85o 61,81o
Rake 164o 174,-174o 171,-167o 178,178o -178,177o -164,162o -168,-166o
dmin / km 17.0b 0 0 0 0 0 0
dmax / km 20.0,21.6
a Harvard CMT solution; b Centroid Location; c Assuming Lam´ elastic constants µ = 3.43 × 10 10 Pa, λ = 3.22 × 1010 Pa; d Figures in brackets
refer to geological observations of surface slip [Barka, 1999]; e An improved fit is obtained if slip is only 3.7m in the top 2 km
If the slip is predominantly right-lateral and occurs over B¨rgmann, R., P. Rosen, and E. Fielding, Synthetic aperture
most of the seismogenic layer thickness, then it cannot be radar interferometry to measure Earth’s surface topography
caused by the change in static stress from the Izmit earth- and its deformation, Ann. Rev. Earth. Planet. Sci., 28 , 169–
209, 2000.
quake but instead is likely to be the result of release of tec- Fielding, E., T. J. Wright, B. Parsons, P. England, P. Rosen,
tonic stress accumulating in the North Anatolian Fault Zone S. Hensley, and R. Bilham, Topography of northwest Turkey
due to the westward movement of central Turkey with re- from SAR interferometry: Applications to the 1999 Izmit
spect to Eurasia (Fig. 1). These stresses can be released by earthquake geomorphology and co-seismic strain (abstract),
transient dynamic unclamping due to the passage of seismic Eos Trans. AGU , 80 , 663, 1999.
surface waves from the mainshock [Bodin et al., 1994]. How- Goldstein, R. M., and C. L. Werner, Radar interferogram filtering
for geophysical applications, Geophys. Res. Lett., 25 , 4,035–
ever, right-lateral slip requires the release of significantly 4,038, 1998.
more seismic moment (Table 3) than left-lateral or vertical Massonnet, D., and K. L. Feigl, Radar interferometry and its
slip solutions and, in the absence of other other data, the application to changes in the earth’s surface, Rev. Geophys.,
smaller M0 solutions are preferred. 36(4), 441–500, 1998.
Although triggered slip has been observed in only a lim- Okada, Y., Surface deformation due to shear and tensile faults in
ited number of cases [Sylvester , 1986; Bodin et al., 1994; a half-space, Bull. Seismol. Soc. Am., 75 , 1,135–1,154, 1985.
Parsons, T., S. Toda, R. Stein, A. Barka, and J. Dieterich, Height-
Price and Sandwell , 1998], it may well be a common occur- ened odds of large earthquakes near Istanbul: an interaction-
rence. Most of the previous observations have been of small based probability calculation, Science, 228 , 661–665, 2000.
surface ruptures. In these examples, however, the interfer- Pegler, G., and S. Das, Analysis of the relationship between seis-
ometric fringes can be traced continuously across the faults mic moment and fault length for large crustal strike-slip earth-
showing that slip does not break the surface. Although the quakes between 1977-92, Geophys. Res. Lett., 23(9), 905–908,
triggered slip is small compared to that in the main Izmit
Price, E., and D. Sandwell, Small-scale deformations associated
event, it is not a negligible part of the strain budget. For with the 1992 Landers, California earthquake mapped by syn-
example, if slip is right-lateral with a magnitude of ∼10 cm, thetic aperture radar interferometry phase gradients, J. Geo-
it would represent ∼10 years of strain accumulation, assum- phys. Res., 103 , 27,001–27,016, 1998.
ing a 2 cm/yr slip rate partitioned equally onto two faults. Reilinger, R., S. McClusky, M. Oral, R. King, M. Toksoz,
In future, it may be possible to use InSAR to determine A. Barka, I. Kinik, O. Lenk, and I. Sanli, Global Positioning
System measurements of present-day crustal movements in the
displacements for triggered slip events using a number of
Arabia–Africa–Eurasia plate collision zone, J. Geophys. Res.,
different lines of sight, hence resolving the ambiguity in the 102 , 9,983–9,999, 1997.
slip mechanism. ¨ ˙
Saroglu, F., O. Emre, and I. Ku¸cu, Active Fault Map of
¸ s¸
Turkey, General Directorate of Mineral Research and Explo-
Acknowledgments. This work is supported by NERC ration (MTA), Eski¸ehir Yolu, 06520, Ankara, Turkey, 1992.
and the European Space Agency, and a NERC studentship, with Sylvester, A., Near-field tectonic geodesy, in Active Tecton-
Nigel Press Associates, to TJW. We thank Aykut Barka, Philip ics (Studies in Geophysics Series), pp. 164–180, National
England, Kurt Feigl, Mark Haynes, Margaret Moore, Ross Stein, Academy Presss, Washington, D. C., 1986.
Roger Bilham and two anonymous reviewers for helpful discus- Wright, T. J., B. Parsons, J. Jackson, M. Haynes, E. Fielding,
sions and comments; CNES and Kurt Feigl for support with the P. England, and P. Clarke, Source parameters of the 1 October
DIAPASON/PRISME software and Paul Rosen and Scott Hens- 1995 Dinar (Turkey) earthquake from SAR interferometry and
ley of JPL for the use of their ROI pac software. seismic bodywave modelling, Earth Planet. Sci. Lett., 172 , 23–
37, 1999.
Barry Parsons, Tim Wright, Department of Earth Sci-
Ambraseys, N., Some characteristic features of the North Anato- ences, University of Oxford, Parks Road, Oxford, UK. (e-mail:
lian Fault Zone, Tectonophys., 9 , 143–165, 1970. [email protected];[email protected])
Ambraseys, N., and J. Jackson, Seismicity of the Sea of Marmara Eric Fielding, JetPropulsionLaboratory, 4800 Oak Grove Drive,
(Turkey) since 1500, Geophys. J. Int., In Press, 2000. Pasadena, CA, 91109, USA. (e-mail: [email protected])
Barka, A., Slip distribution along the North Anatolian Fault asso-
ciated with the larger earthquakes of the period 1939 to 1967,
Bull. Seismol. Soc. Am., 86 , 1,238–1,254, 1996.
Barka, A., The 17 August 1999 Izmit earthquake, Science, 285 ,
1858–1859, 1999.
Bodin, P., R. Bilham, J. Behr, J. Gomberg, and K. Hud-
nut, Slip triggered on southern California faults by the 1992
Joshua Tree, Landers and Big Bear earthquakes, Bull. Seis- (Received May 16, 2000; revised August 3, 2000;
mol. Soc. Am., 84 , 806–816, 1994. accepted September 25, 2000.)

Use: 0.4312