Content uploaded by Simon Mcclusky
Author content
All content in this area was uploaded by Simon Mcclusky
Content may be subject to copyright.
A preview of the PDF is not available
Deformation during the 12 November 1999 Düzce, Turkey, Earthquake, from GPS and InSAR Data
Abstract and Figures
Only 87 days after the M w 7.5, 17 August 1999 I ˙ zmit earthquake, the Düzce earthquake ruptured a ca. 40-km-long adjoining strand of the North Anatolian fault (NAF) system to the east. We used displacements of 50 Global Positioning System (GPS) sites together with interferometric synthetic aperture radar (InSAR) range-change data spanning the event to estimate the geometry and slip distribution of the coseismic rupture. Postseismic deformation transients from the Düzce earth-quake and the preceding I ˙ zmit event that are included in some of the measurements are corrected for using dislocation models fit to GPS data spanning the various time periods. Nonlinear inversions for fault geometry indicate that the rupture occurred on a ca. 54 north-dipping oblique normal, right-lateral fault. Distributed-slip inver-sions indicate maximum strike slip near the center of the Düzce fault close to the earthquake hypocenter. Slip magnitude and depth of faulting decrease to the west and east of the hypocenter. Both GPS and InSAR data suggest that normal slip is restricted to the shallow portion of the rupture. The Düzce earthquake had the highest slip-to-rupture-length ratio of any historic earthquake along the NAF.
Figures - uploaded by Simon Mcclusky
Author content
All figure content in this area was uploaded by Simon Mcclusky
Content may be subject to copyright.
... The depth of the fault is set within 0-20 km; the rake is set within −30 • -30 • ; and the dip of the fault is set within 30 • -90 • according to the background of the regional structure. Under the above parameter settings, the multiple peak particle swarm optimization (MPSO) algorithm searched for the optimal uniform sliding fault parameters (Table 1), and then the error of the uniform sliding inversion fault geometry parameters was estimated using Monte Carlo correlation noise simulation [19,25,38,39]. During the inversion, the function between the InSAR observations and the fault parameters is first established: ...
... where κ 2 is the smooth factor; L is the Laplace second-order smoothing operator; and S is the sliding amount on each fault sheet. Since the inclination of the uniform sliding fault model obtained by previous inversion is not the optimal inclination [39] corresponding to the distributed sliding model, the method that comprehensively considers the roughness and model fit residual needs to search for the optimal inclination. Based on the InSAR line-of-sight deformation results obtained by inversion, the Levenberg-Marquardt [40] least squares optimization algorithm is used to iterate until the Remote Sens. 2023, 15, 3753 7 of 17 convergence of the objective function at the global minimum. ...
On 5 September 2022, an Ms6.8 magnitude earthquake occurred in Luding County, Sichuan Province, China. Based on Sentinel-1 SAR images, this paper uses the D-InSAR approach to obtain the displacement field of the earthquake, invert the coseismic sliding distribution, and then calculate the static coulomb stress changes of the coseismic deformation on the aftershock distribution and surrounding faults. Further, the seismic structure is analyzed and discussed. The InSAR coseismic deformation field demonstrates that the maximum LoS displacement of the surface deformation caused by the Luding earthquake is about 15 cm. The Luding Ms 6.8 earthquake is dominated by the Moxi fault, which is a left-lateral strike-slip fault that ruptures along the NNW-SSE trend at about 160.3°, and the dip is 81°. The fault depth is mainly 5~15 km, the maximum sliding amount is about 174.8 cm, and the corresponding depth is 8.5 km. The seismic moment tensor obtained by inversion is 1.06 × 1019 Nm, Mw = 6.65. The Coulomb stress generated by the Luding earthquake on the northern end of the Anninghe fault zone exceeded the trigger threshold. The risk of the Anninghe fault’s future earthquake was greater, and continuous monitoring and risk assessment were required.
... Using GNSS and InSAR data to jointly invert the distribution of fault slip has become an important means to understand the mechanism of earthquakes and the process of source rupture (Bürgmann et al., 2002;Motagh et al., 2008). Near-field GNSS observations provide more accurate constraints, whereas InSAR measurements provide spatial continuity compared to GNSS observations, which can give a full view of fault slip in remote regions. ...
We use surface deformation measurements, including Interferometric Synthetic Aperture Radar data acquired by the Sentinel-1A satellites and Global Navigation Satellite System observations, to invert the fault geometry and coseismic slip distribution of the 2022 Mw 6.6 earthquake in Sichuan, China. The dip of the best-fitting model is 68°. The rupture of the 2022 Luding earthquake is dominated by northwest strike-slip movement, mainly concentrated over a length of about 20 km above a depth of 15 km. The maximum slip is at approximately 4 km depth with the maximum displacement of about 2.1 m. The results indicate that the 2022 Luding earthquake ruptured the shallow layer of the seismic zone. The slip distribution indicates that the Moxi–Shimian fault segment is fully locked from the surface down to 15 km, which is consistent with the estimated locking depth. Based on the Coulomb stress analysis and considering the strong locking state of the Anninghe fault, more attention should be paid to the possibility of earthquakes in the Anninghe fault.
... İzmit Körfezi ile Gölyaka (Düzce) arasında yaklaşık 150 km uzunluğundaki Hersek, Karamürsel-Gölcük, İzmit-Sapanca Gölü, Sapanca-Akyazı ve Karadere segmentlerinin kırılmasına neden olmuştur [13][14][15][16][17]. 17 Ağustos 1999 İzmit depreminden yaklaşık 3 ay sonra, KAFZ üzerinde bulunan Düzce havzasının güneyinde 12 Kasım 1999 Düzce depremi (Mw7.2) meydana gelmiştir. Deprem, Bolu'nun batısında; Kaynaşlı'nın doğusu ile Gölyaka arasında 41 km uzunluğunda yüzey kırığı oluşturmuştur [18]. İki büyük deprem sonrası Düzce ve Hendek faylarına büyük bir gerilme enerjisinin aktarıldığı ve bu faylar üzerinde deprem riskinin arttığı ileri sürülmüştür [19][20]. ...
Bu çalışmada 17 Kasım 2021 tarihinde yerel saatle 15.40’da meydana gelen M5.0 büyüklüğündeki Merkez (Düzce) depremi ile M4.3 (15.57) ve M3.9 (21.35) büyüklüklerindeki iki artçışokunun kaynak mekanizma çözümleri moment tensör yöntemiyle bulunmuştur. Anaşokun KD-GB yönelimli doğrultu atımlı kırık üzerinde meydana geldiği tespit edilmiştir. Bu çalışmada elde edilen anaşok moment tensör çözümü, deprem veri merkezlerinden elde edilen fay çözümleri ile karşılaştırılmış ve bulunan sonucun uyumlu olduğu görülmüştür. M4.3 büyüklüğündeki artçışokun kaynak mekanizmasının doğrultu atım bileşeni olan ters faylanma mekanizmasına, M3.9 büyüklüğündeki artçışokun ise KD-GB yönelimli doğrultu atımlı faylanma mekanizmasına sahip olduğu görülmüştür. Anaşokun neden olduğu Coulomb gerilme değişimi hesaplanmış ve K-G, KB-GD ve KD-GB yönlerinde kesitler alınarak gerilmenin 20 km kadar olan derinlikteki değişimi incelenmiştir. Buna göre, Merkez (Düzce) depreminin anaşok ve artçışoklarının bölgedeki KD-GB yönündeki gerilmeyi azalttığı, buna karşın K-G ve DGD-BKB yönündeki yakın faylar üzerine maksimum 0.5 bar gerilme yüklediği tespit edilmiştir.
... Nonetheless, we believe this work represents a step toward failure prediction in the field for the following reasons. First, an approach similar to the one presented here still using lab data might be followed to better constrain the rate and state frictional models and associated parameters that are used in geodetic studies to infer fault slip distribution at depth [59][60][61][62][63][64][65][66][67] . Second, stress and slip rate data might be inferred in the field with ML models using earthquake recurrence as input data, and possibly pre-training on lab data. ...
Predicting failure in solids has broad applications including earthquake prediction which remains an unattainable goal. However, recent machine learning work shows that laboratory earthquakes can be predicted using micro-failure events and temporal evolution of fault zone elastic properties. Remarkably, these results come from purely data-driven models trained with large datasets. Such data are equivalent to centuries of fault motion rendering application to tectonic faulting unclear. In addition, the underlying physics of such predictions is poorly understood. Here, we address scalability using a novel Physics-Informed Neural Network (PINN). Our model encodes fault physics in the deep learning loss function using time-lapse ultrasonic data. PINN models outperform data-driven models and significantly improve transfer learning for small training datasets and conditions outside those used in training. Our work suggests that PINN offers a promising path for machine learning-based failure prediction and, ultimately for improving our understanding of earthquake physics and prediction.
... Under the assumption of a uniform slip, the dip angle is usually the most difficult parameter to determine uniquely in the nonlinear inversion stage (Bürgmann et al., 2002;Fukahata and Wright, 2008). To obtain an optimal dip angle, we apply an iterative approach with a log-function proposed by Feng et al. (2013) in the linear slip inversion step to further refine the dip angle, in which we conduct a series of linear slip inversion with variable dip angles to explore a global minimum solution. ...
... More recent sequences followed the devastating 26 December 1939 Erzincan M w 7.9 quake [43], which likely produced the further destabilization which was the reason for the 1942-1944 seismic activity, continuing with major quakes in 1949,1951,1957,[1966][1967]1992 and two in 1999. The last one culminated with the 17 August 1999 M w 7.4 Izmit earthquake [44] and the following 12 November 1999 M w 7.2 Düzce event [45]. The Aegean region also nucleated large seismic events, such as the 23 July 1949 Chios [46], the 24 April 1957 M w 7.1 Ortaca, and the 30 October 2020 M w 7.0 Izmir Bay earthquake [47]. ...
The Anatolian region is one of the most seismically active tectonic settings in the world. Here, we perform a clustering analysis of Turkish seismicity using an updated version of the Turkish Homogenized Earthquake Catalogue (TURHEC), which contains the recent developments of the still ongoing Kahramanmaraş seismic sequence. We show that some statistical properties of seismic activity are related to the regional seismogenic potential. Mapping the local and global coefficients of variation of inter-event times of crustal seismicity which occurred during the last three decades, we find that territories prone to major seismic events during the last century usually host globally clustered and locally Poissonian seismic activity. We suggest that regions with seismicity associated with higher values of the global coefficient of variation of inter-event times, Cv, are likely to be more prone to hosting large earthquakes in the near future than other regions characterized by lower values, if their largest seismic events have the same magnitude. If our hypothesis is confirmed, clustering properties should be considered as a possible additional information source for the assessment of seismic hazard. We also find positive correlations between global clustering properties, the maximum magnitude and the seismic rate, while the b-value of the Gutenberg-Richter law is weakly correlated with them. Finally, we identify possible changes in such parameters before and during the 2023 Kahramanmaraş seismic sequence.
... Under the assumption of a uniform slip, the dip angle is usually the most difficult parameter to determine uniquely in the nonlinear inversion stage (Bürgmann et al., 2002;Fukahata and Wright, 2008). To obtain an optimal dip angle, we apply an iterative approach with a log-function proposed by Feng et al. (2013) in the linear slip inversion step to further refine the dip angle, in which we conduct a series of linear slip inversion with variable dip angles to explore a global minimum solution. ...
An Mw 6.2 earthquake struck Khōst, Afghanistan on June 22, 2022, killing 21 more than 1,700 people and devastating the region. For studying this earthquake, we 22 computed the coseismic deformation fields of the earthquake using the Sentinel-1 23 Terrain Observation with Progressive Scans (TOPS) Interferometric Synthetic Aperture 24 Radar (InSAR). The InSAR results show that the maximum coseismic displacement in 25 the satellite Line of Sight (LOS) direction reaches up to 39 cm. We determined the 26 geometric parameters of the fault and coseismic slip distribution from these InSAR 27 measurements. The best-fitted fault model shows that the rupture occurred on a right-28 lateral strike-slip fault with a strike of 203.7° and a dip of 68°. The most slip is 29 concentrated at a shallow depth within the upper 10 km with the maximum slip of ~3 30 m at 2.5 km depth. The maximum slip produced by this earthquake is significantly 31 larger than the slip produced by several other similar earthquakes with similar 32 magnitudes, implying that the focused shallow slip is likely the reason for the 33 significant damage in the earthquake. The heavy rainfall was recorded during the 34 earthquake period, which resulted in complicated fringes in coseismic interferograms 35 close to the earthquake in time. As a positive spatial and temporal correlation with the 36 earthquake occurrence can be seen, the rainfall may have potential contributions to the 37 earthquake, which deserves additional analysis in future. Combined with the potential 38 2 effects of the 2015 Mw 7.5 Hindu Kush deep-seated earthquake, the seismicity in 39 Afghanistan is the result of the ongoing subduction of the Indian plate beneath the 40 Eurasian plate along their west boundary. 41 42
This study presents an assessment of the displacements of the Earth’s crust generated by two earthquakes: one in Croatia on 29 th December 2020 and one in Greece on 3 rd March 2021. The main data source is SAR satellite images from Sentinel-1A and B (ascending and descending orbits) in the form of images of the amplitude and phase signals. The data is processed via SNAP software package. The results are obtained applying the differential radar interferometry method (DInSAR). The Line of Sight (LoS) displacements are converted into vertical displacements and compared with modeled deformations of the Earth’s surface using rectangle finite source model proposed by Okada.
The Bolu tunnel is located between the Istanbul-Ankara Highway and its construction took
approximately 13 years. Since the beginning of the tunnel excavation, serious deformations and
stability problems have been encountered. The basis of the problems encountered during the tunnel
excavation is the fact that the geological units through which the tunnel passes are very weak and
fault lines cut the tunnel location. The fault lines pass through secondary faults in the seismically
active North Anatolian Fault. At these fault crossings, deformations occurred continuously so
that revisions had to be made in the support systems. The 12 November 1999 Düzce earthquake
occurred on the Düzce fault and a surface rupture 40 km long was caused. The rupture terminated
1.5 km west of the tunnel; it did not reach the tunnel. Throughout the earthquake, instances of
collapse occurred at the areas excavated on the fault line at the entrance of Elmalık and where the
deformations exceeded 1.0 m; this section of the tunnel had to be abandoned. After these problems, a
new improved design for tunnel excavation was developed. These new support systems, which are
Bernold lining and bench pilot tunnel systems, contain allowance for rigid lining that is the opposite
of the new Austrian tunneling method (NATM) principles. Within the scope of this study, the causes
of the collapse in the tunnel are investigated and the effect of the Düzce earthquake on the tunnel is
discussed. Also, the applicability of the NATM in tunnels excavated on weak zones is also discussed.
On 23 November 2022, a MW 6.0 earthquake occurred in the direct vicinity of the MW 7.1 Düzce earthquake that ruptured a portion of the North Anatolian Fault in 1999. The Mw 6.0 event was attributed to a small portion of the Karadere fault off the main North Anatolian Fault that did not rupture during the 1999 sequence. We analyze the spatiotemporal evolution of the MW 6.0 Gölyaka–Düzce seismic sequence at various scales and resolve the source properties of the mainshock. Modeling the decade-long evolution of the background seismicity of the Karadere fault employing an Epistemic Type Aftershock Sequence model shows that this fault was almost seismically inactive before 1999, while a progressive increase in seismic activity is observed from 2000 onwards. A newly generated high-resolution seismicity catalog from 1 month before the mainshock until 6 d after was created using artificial-intelligence-aided techniques and shows only a few events occurring within the rupture area within the previous month, no spatiotemporal localization process and a lack of immediate foreshocks preceding the rupture. The aftershock hypocenter distribution suggests the activation of both the Karadere fault, which ruptured in this earthquake, and the Düzce fault that ruptured in 1999. First results on the source parameters and the duration of the first P-wave pulse from the mainshock suggest that the mainshock propagated eastwards, which is in agreement with predictions from a bimaterial interface model. The MW 6.0 Gölyaka–Düzce event represents a good example of an earthquake rupture with damaging potential within a fault zone that is in a relatively early stage of the seismic cycle.
We present and interpret the results of postseismic, Global Positioning System monitoring of the first 298 days following
the 17 August 1999 İzmit earthquake. Whereas the data suggest some spatial and temporal complexity in the postseismic motions,
the overall pattern can be characterized by time-dependent relaxation functions and suggests exponential decay with an estimated
57-day relaxation time. The very rapid deformation during the first few weeks following the mainshock suggests rapid afterslip
on and below the coseismic rupture segments. The exponential decay of the postseismic deformations through the end of the
observation period suggests contributions from the lower crustal viscoelastic relaxation.
The use of SAR interferometry is often impeded by decorrelation from thermal noise, temporal change, and baseline geometry. Power spectra of interferograms are typically the sum of a narrow-band component combined with broad-band noise. We describe a new adaptive filtering algorithm that dramatically lowers phase noise, improving both measurement accuracy and phase unwrapping, while demonstrating graceful degradation in regions of pure noise. The performance of the filter is demonstrated with SAR data from the ERS satellites over the Jakobshavns glacier of Greenland.
Two large strike-slip ruptures 11.4 hours apart occurred on intersecting, nearly orthogonal, vertical faults during the November 1987 Superstition Hills earthquake sequence in southern California. This sequence is the latest in a northwestward progression of earthquakes (1979, 1981, and 1987) rupturing a set of parallel left-lateral cross-faults that trend northeast between the Brawley seismic zone and Superstition Hills fault, a northwest trending main strand of the San Jacinto fault zone. The first large event (MS=6.2) in the 1987 sequence ruptured the Elmore Ranch fault, a cross-fault that strikes northeasterly between the Brawley seismic zone and the Superstition Hills main fault. The second event (MS=6.6) initiated its rupture at the intersection of the cross-fault and main fault and propagated towards the southeast along the main fault. The following hypotheses are advanced; (1) slip on the cross-fault locally decreased normal stress on the main fault, and triggered the main fault rupture after a delay; and (2) the delay was caused by fluid diffusion. It is inferred that the observed northwestward progression of ruptures on cross-faults may continue. The next cross-fault expected to rupture intersects both the San Andreas fault and the San Jacinto fault zone. We hypothesize that rupture of this cross-fault may trigger rupture on either of these main faults by a mechanism similar to that which occurred in the Superstition Hills earthquake sequence.
We study the stress transferred by the June 27, 1988, M=5.3 and August 8, 1989, M=5.4 Lake Elsman earthquakes, the largest events to strike within 15 km of the future Loma Prieta rupture zone during 74 years before the 1989 M=6.9 Loma Prieta earthquake. We find that the first Lake Elsman event brought the rupture plane of the second event 0.3-1.6 bars (0.03-0.16 MPa) closer to Coulomb failure but that the Lake Elsman events did not bring the future Loma Prieta hypocentral zone closer to failure. Instead, the Lake Elsman earthquakes are calculated to have reduced the normal stress on (or ``unclamped'') the Loma Prieta rupture surface by 0.5-1.0 bar (0.05-0.10 MPa) at the site where the greatest slip subsequently occurred in the Loma Prieta earthquake. This association between the sites of peak unclamping and slip suggests that the Lake Elsman events did indeed influence the Loma Prieta rupture process. Unclamping the fault would have locally lowered the resistance to sliding. Such an effect could have been enhanced if the lowered normal stress permitted fluid infusion into the unclamped part of the fault. Although less well recorded, the ML=5.0 1964 and ML=5.3 1967 Corralitos events struck within 10 km of the southwest end of the future Loma Prieta rupture. No similar relationship between the normal stress change and subsequent Loma Prieta slip is observed, although the high-slip patch southwest of the Loma Prieta epicenter corresponds roughly to the site of calculated Coulomb stress increase for a low coefficient of friction. The Lake Elsman-Loma Prieta result is similar to that for the 1987 M=6.2 Elmore Ranch and M=6.7 Superstition Hills earthquakes, suggesting that foreshocks might influence the distribution of mainshock slip rather than the site of mainshock nucleation.
A number of techniques are employed to overcome nonuniqueness and instability inherent in linear inverse problems. To test the factors that enter into the selection of an inversion technique for fault slip distribution, penalty function with smoothness (PF+S), a damped least-squares method (DLS), damped least-squares method with a positivity constraint (DLS+P), and a penalty function with smoothness and a positivity constraint (PF+S+P) were used for inverting the elevation changes for slip associated with the 1983 Borah Peak earthquakes. All available geological and geophysical information was used to determine a geophysical deformation model for the earthquake. It is suggested that the PF+S+P solution for a fault length of 75km is the preferred model. -from Authors
A complete set of closed analytical expressions is presented in a unified manner for the internal displacements and strains due to shear and tensile faults in a half-space for both point and finite rectangular sources. Several practical suggestions to avoid mathematical singularities and computational instabilities are presented. -from Author
Strike slip on various scales and on faults of diverse orientations is one of the most prominent modes of deformation in continental convergence zones. Extreme heterogeneity and low shear strength of continental rocks are responsible for creating complex 'escape routes' from nodes of constriction along irregular collision fronts toward free faces formed by subduction zones. The origin of this process is poorly understood. The 2 main models ascribe tectonic escape to buoyancy forces resulting from differences in crustal thickness generated by collision and to forces applied to the boundaries of the escaping wedges. Escape tectonics also creates a complicated geological signature, whose recognition in fossil examples may be difficult. We examine the Neogene to present tectonic escape-dominated evolution of Turkey both to test the models devised to account for tectonic escape and to develop criteria by which fossil escape systems may be recognized.-from Authors
Between 1939 and 1967, six large fault ruptures formed a westwardmigrating sequence of events along a 900-km-long nearly continuous portion of the North Anatolian fault. For these events - the 1939 Erzincan, 1942 Niksar-Erbaa, 1943 Tosya, 1944 Bolu-Gerede, 1957 Abant, and 1967 Mudurnu Valley earthquakes - I have compiled a record of dextral slip, which contains nearly 100 points. These data indicate that the amount of slip is irregularly distributed along the 1939 to 1967 rupture zone. The maximum slip, 7.5 m, occurred during the 1939 earthquake in the eastern 150 km of the 900-km-long rupture zone. Dextral offsets diminish very abruptly eastward but very gradually westward. The rate of westward decrease in the 1939 to 1967 offsets is only slightly greater than the rate of westward decrease of post-Miocene displacement along the North Anatolian fault. This suggests that westward decrease in slip can be expected to be a general characteristic of earthquake ruptures along the North Anatolian fault in the future. Nevertheless, within the 1939 to 1967 slip distribution, there are three regions that had less slip than the neighboring regions. These I interpret as possible sites of large future earthquakes. One of these seismic gaps has already experienced another earthquake (in 1951) and subsequent creep.
We have developed dynamic finite-element models of İzmit earthquake postseismic deformation to evaluate whether this deformation
is better explained by afterslip (via either velocity-strengthening frictional slip or linear viscous creep) or by distributed
linear viscoelastic relaxation of the lower crust. We find that velocity-strengthening frictional afterslip driven by coseismic
shear stress loading can reproduce time-dependent Global Positioning System data better than either linear viscous creep on
a vertical shear zone below the rupture or lower crustal viscoelastic relaxation. Our best frictional afterslip model fits
the main features of postseismic slip inversions, in particular, high slip patches at (and below) the hypocenter and on the
western Karadere segment, and limited afterslip west of the Hersek Delta (Bürgmann et al., 2002). The model requires a weakly velocity-strengthening fault, that is, either low effective normal stress in the slipping regions
or a smaller value for the parameter describing rate-dependence of friction (a-b) than is indicated by laboratory experiments.
Our best afterslip model suggests that the Coulomb stress at the Düzce hypocenter increased by 0.14 MPa (1.4 bars) during
the İzmit earthquake (assuming right-lateral slip on a surface dipping 50° to the north), and by another 0.1 MPa during the
87 days between the İzmit and Düzce earthquakes. In the Marmara Sea region (within about 160 km of the İzmit earthquake rupture),
this model indicates that the Coulomb stresses increased by 15%-25% of the coseismic amount during the first 300 days after
the earthquake. Three hundred days after the earthquake, postseismic contributions to Coulomb stressing rate on the Maramara
region faults had fallen to values equal to or less than the inferred secular stress accumulation rate. Our estimates of postseismic
Coulomb stress are highly model dependent: in the Marmara region, the linear viscous shear zone and viscoelastic lower crust
models predict greater postseismic Coulomb stresses than the frictional afterslip model.
Near-field stress and fault-zone rheology estimates are sensitive to the Earth's elastic structure. When a layered elastic
structure is incorporated in our model, it yields a Coulomb stress of 0.24 MPa at the Düzce hypocenter, significantly more
than the 0.14 MPa estimated from the uniform elastic model. Because of the higher near-field coseismic stresses, the layered
elastic model requires a higher value of velocity-strengthening parameter (A-B) ([a-b] times effective normal stress r′) to produce comparable postseismic slip. (A-B) is estimated at 0.4 and 0.2 MPa, respectively, for the layered and uniform elastic models. These results highlight the importance
of understanding the Earth's elastic structure and the mechanism for postseismic deformation if we wish to accurately model
coseismic and postseismic crustal stresses.
Postseismic deformation in the 5 years following the 1989 Loma Prieta earthquake has been measured with the Global Positioning System and precise leveling. Postearthquake velocities at distances greater than ~20km from the coseismic rupture are not significantly different from those observed in the 20 years prior to the earthquake. However, velocities at stations within ~20km of the rupture exceed preearthquake rates and exhibit unanticipated contraction normal to the strike of the San Andreas fault system. A combination of forward modeling and nonlinear optimization suggests that the observed postseismic deformations were caused by aseismic oblique reverse slip averaging 2.9 cm/yr on the San Andreas fault and/or the Loma Prieta rupture zone and 2.4 cm/yr reverse slip along a buried fault within the Foothills thrust belt. The best fitting sources of postseismic deformation are all located at depths of less than 15 km. We find no evidence for accelerated flow or shear below the Loma Prieta rupture in the first 5 years following the earthquake. The inferred postseismic slip is likely to have been caused by the coseismic stress change updip of the 1989 rupture.