Abstract Submitted to Fall 1998 AGU Meeting

Colloque International Technologies Nouvelles et Amélioration de la Gestion des Risques Majeurs, Kénitra, Morocco, 15-17 March 2004

Seismic Risk Assessment: Space Geodetic Techniques Expand Ability to Measure Elastic Strain Accumulation

James N. Kellogg, University of South Carolina, Dept. of Geological Sciences, Robert Trenkamp, University of South Carolina, Dept. of Geological Sciences, Scott M. White, University of South Carolina, Dept. of Geological Sciences,

Abstract

Seismic risk maps have been traditionally based on historical seismic records. However, long-term seismic hazard estimates based on historical records are often biased by variations in seismic recurrence intervals. Estimates of seismic risk in areas with low strain rates are even more difficult because the seismic recurrence interval may be longer than the historical record. In the last decade, advances in space geodetic techniques have greatly expanded our ability to measure active strain rates in the earth, both seismic and aseismic. It is now possible to routinely measure very low strain rates with Global Positioning System (GPS) observations at sub-centimeter level accuracy. Interseismic elastic strain accumulation or “locking” associated with the earthquake cycle can be differentiated from aseismic slip leading to more reliable estimates of long-term seismic hazard.

The Ecuador-Colombia subduction zone of the northern Andean margin is characterized by a high rate of plate convergence and frequent great earthquakes. The subduction zone has been ruptured repeatedly in 1906 (Mw = 8.8), 1942 (Mw = 7.9), 1958 (Mw = 7.8), and 1979 (Mw = 8.2). However, estimates of seismic slip from the earthquakes are significantly less than the estimates of plate convergence, suggesting an apparent 33% “slip deficit”. GPS measurements of horizontal deformation can be modeled by elastic recoverable strain accumulation associated with the earthquake cycle at a locked fault interface and motion due to permanent deformation (sometimes several hundreds of kilometers from the fault zone or plate boundary). GPS results in the northern Andean margin can be modeled as apparent s locking of the subduction zone, in approximate agreement with the seismic slip estimates at the one sigma level. The apparent “seismic slip deficit” may be produced by permanent deformation, or recently discovered slow afterslip or “silent earthquakes”. Postseismic relaxation of a viscous upper mantle may, however, mask the true total strain rate. Geodetic measurements of seismic strain accumulation may be particularly important in Morocco where strain rates are relatively low and seismic recurrence intervals are long.

1 Colloque International Technologies Nouvelles et Amélioration de la Gestion des Risques Majeurs, Kénitra, Morocco, 15-17 March 2004 Introduction

Damage and death from earthquakes can be substantially reduced with zoning requirements for earthquake resistant building construction based on accurate seismic risk maps. Until recently, seismic risk maps have been based on historical seismic records or paleoseismology studies. However, long-term seismic hazard estimates based on historical records are often biased by inaccurate records and variations in seismic recurrence intervals. Estimates of seismic risk in areas with low strain rates are even more difficult because the seismic recurrence interval may be longer than the historical record (e.g., Cascadia, Dragert et al., 2001). In the last decade, advances in space geodetic techniques have greatly expanded our ability to measure active strain rates in the earth, both seismic and aseismic. It is now possible to routinely measure very low strain rates with Global Positioning System (GPS) observations at sub-centimeter level accuracy. Interseismic elastic strain accumulation or “locking” associated with the earthquake cycle can be estimated from geodetic measurements of horizontal and vertical ground deformation. Geodetic measurements can also be used to estimate permanent deformation produced by aseismic slip. These measurements have resulted in exciting discoveries of large afterslip events and “silent earthquakes.” Accurate models of seismic and aseismic slip are producing more reliable estimates of long-term seismic hazard. For example, preliminary GPS measurements (R. Reilinger, personal communication, 2004, MIT) indicated left lateral strain accumulation on a N-S trending fault near Al Hoceima, Morocco, prior to the magnitude 6.4 February 24, 2004 earthquake.

Elastic Strain Accumulation.

Crustal motion at major fault systems may be characterized as reflecting at least two modes of deformation:

1. Deformation due to elastic recoverable strain accumulation associated with the seismic cycle at a locked or partially locked fault interface, and

2. Motion due to permanent strain associated with aseismic slip.

The first mode of deformation can be modeled as relative deformation in an elastic half- space due to strain accumulation imposed on a defined fault plane (Savage, 1983). Figure 1, for example, shows geodetic data at one standard error compared with elastic half-space models for the Ecuador-Colombia trench (Trenkamp et al., 2002). A model with s locking (29 mm/yr) in the subduction zone plus 6 mm/yr permanent shortening several hundred kilometers to the east in the Andes fits most of the observations well at 1 sigma confidence.

The Ecuador-Colombia subduction zone of the northern Andean margin is characterized by a high rate of plate convergence and frequent great earthquakes. The subduction zone has been ruptured repeatedly in 1906 (Mw = 8.8), 1942 (Mw = 7.9), 1958 (Mw = 7.8), and 1979 (Mw = 8.2). However, our result of 50% elastic locking means that only half of the

2 Colloque International Technologies Nouvelles et Amélioration de la Gestion des Risques Majeurs, Kénitra, Morocco, 15-17 March 2004

Figure 1. Elastic half-space models compared to GPS eastward velocities at the Ecuador-Colombia plate boundary relative to stable South America with 1σ uncertainties. The best fitting model has 50% elastic locking and 50% aseismic slip on the interplate boundary and 6 mm shortening to the east in the Andean mountains (modified from Trenkamp et al., 2002). The interplate fault is locked to a depth of 50 km. total plate motion is stored for eventual release in earthquake (elastic) events. Kanamori and McNally (1982) estimated the slip resulting from each of the twentieth century earthquakes. These estimated values of seismic slip average approximately 67% of the cumulative convergence at the trench since 1906 (Figure 2, White et al., 2003). Relaxation of a viscous upper mantle after a great earthquake has the opposite sense of motion as elastic interseismic strain and may mask the true total elastic strain rate (Cohen, 1984). The large uncertainties associated with both estimates of elastic strain do not rule out that all of the elastic locking is being released as elastic slip (Figure 2).

3 Colloque International Technologies Nouvelles et Amélioration de la Gestion des Risques Majeurs, Kénitra, Morocco, 15-17 March 2004 However, both the estimates of seismic slip from the earthquakes and the estimated elastic locking are significantly less than the estimates of plate convergence, suggesting an apparent 33 to 50% “slip deficit”.

Figure 2. Seismic slip on the Ecuador- Colombia Trench since 1906 plotted against a Nazca-South America convergence rate of 58 mm/yr (White et al., 2003). Estimates of seismic slip for each earthquake are from Kanamori and McNally (1982). The seismic line indicates the seismic slip necessary to release all of the convergence with 1σ error envelope (shaded area). The short dashed line is the estimate of seismic slip necessary to release s of the total accumulated convergence. A linear least-squares fit to the estimates of seismic slip from earthquakes is shown with a long dashed line (error envelope for this regression indicated by the shaded area), and suggests that these earthquakes have been releasing ~67% of the total accumulated convergence. The great earthquakes along the seismogenic areas of the Japan trench and Nankai trough subduction zones account for only 20% of the energy release predicted by plate tectonic models such as NUVEL-1A (e.g., Peterson and Seno, 1984). Mazzotti et al. (2000) inferred that recently discovered aseismic processes such as slow or silent earthquakes accommodate most of the plate convergence on this plate boundary. Along strike-slip fault boundaries as well, fault segments are known to move with stable aseismic creep (Tse and Rice, 1986).

Aseismic Slip and the Seismic Slip Deficit.

Recent geodetic observations suggest that great subduction zone earthquakes do not stop abruptly but rather steadily decay into slow slip downdip of the seismogenic zone. Postseismic subduction zone deformation involves both short-term afterslip on the fault plane and long-term viscoelastic relaxation of the lower crust/upper mantle (Figure 3, Nur and Mavko, 1974; Cohen, 1984; Heki et al., 1997; Masterlark et al., 2001). In the first several months following an earthquake, on-going slip on the rupture plane (afterslip) and poroelastic effects may contribute a significant component of the GPS deformation in combination with the relaxation (Nakano and Hirahara, 1997; Pollitz et al., 1998). In 1997 Heki et al. reported the surprising discovery that silent fault slip continued for a year following an interplate thrust earthquake at the Japan Trench with an equivalent moment magnitude equal to the earthquake.

Figure 3. Map of geodetic signals in terms of spatial and temporal scales (modified after Minster et al., 1990).

In the past few years, GPS observations from Cascadia, Japan, and the Middle America Trench have revealed the existence of aseismic transient slip events not associated with plate boundary earthquakes (Dragert et al., 2001; Lowry et al., 2001; Ozawa et al., 2001). In the Cascadia subduction zone these events occur with remarkably consistent recurrence intervals of 13 to 16 months (Dragert et al., 2001). In the Middle America Trench, equivalent moment magnitudes reach as high as Mw 7.7 (Lowry et al., 2001). To the degree that these slip events increase coseismic loading across locked portion of the plate boundary, they may play a long-term role in triggering future earthquakes.

Klotz et al. (2001) and Freymueller et al. (2000) suggest that postseismic deformation continues for decades following great (Mw>8.5) earthquakes. Physically realistic models of plate boundary deformation involve faulting in an elastic layer overlying a viscoelastic lower crust or upper mantle (Nur and Mavko, 1974; Pollitz, 1992; Segall, 2002) or velocity strengthening slip on a downdip extension of the coseismic fault surface (Tse and Rice, 1986). White et al. (2003) show that a simple viscoelastic model can explain the apparent reduction in elastic locking spatially associated with a 1979 earthquake on the Ecuador-Colombia Trench (Figure 4). Thus postseismic relaxation may mask the true total strain rate.

Discussion.

Geodetic measurements of seismic strain accumulation may be particularly important in Morocco where strain rates are relatively low and seismic recurrence intervals are long. Seismic slip estimates from historic and paleoseismic records together with geodetic estimates of elastic locking provide constraints on seismic risk analyses. For example, preliminary GPS measurements (R. Reilinger, personal communication, 2004, MIT) indicated left lateral strain accumulation on a N-S trending fault near Al Hoceima,

Figure 4. Simulated crustal velocities produced by viscoelastic relaxation after the 1979 earthquake along the Ecuador-Colombia Trench (White et al., 2003). The fault plane of the earthquake is shown as a dashed rectangle. The red arrows show the viscoelastic deformation field, yellow arrows show the modeled viscoelastic deformation at each GPS site, and the black arrows show the observed GPS vectors.

Morocco, prior to the magnitude 6.4 February 24, 2004 earthquake. Postseismic relaxation of a viscous upper mantle may mask the true strain rate. Aseismic processes, including afterslip, and silent earthquakes may explain the difference between the apparent seismic risk and plate boundary strain rates.

References

Cohen, S.C., 1984, Postseismic deformation due to subcrustal viscoelastic relaxation following dip-slip earthquakes, Journal of Geophysical Research, 89, 4538-4544. Dragert, H., K.L. Wang, T.S. James, 2001, A silent slip event on the deeper Cascadia subduction interface, Science, 292, 1525-1528. Freymueller, J.T., S.C. Cohen, H.J. Fletcher, 2000, Spatial variations in present-day deformation, Kenai Peninsula, Alaska, and their implications, Journal of Geophysical Research, 105, 8079-8101. Heki, E.R., S. Miyazaki, H. Tsuji, 1997, Silent fault slip following an interplate thrust earthquake at the Japan Trench, Nature, 386, 595-598. Kanamori, H., K.C. McNally, 1982, Variable rupture mode of the subduction zone along the Ecuador- Colombia coast, Bulletin of the Seismological Society of America, 72, 1241-1253. Klotz, J., G. Kahzaradze, D. Angermann, C. Reigber, R. Perdomo. O. Cifuentes, 2001, Earthquake cycle dominates contemporary crustal deformation in Central and Southern Andes, Earth and Planetary Science Letters, 193, 437-446. Lowry, A.R., K.M. Larson, V. Kostoglodov, R. Bilham, 2001, Transient fault slip in Guerrero, southern Mexico, Geophysical Research Letters, 28, 3753-3756. Masterlark, T., C. DeMets, H.F. Wang, O. Sanchez, J. Stock, 2001, Homogeneous vs heterogeneous subduction zone models: coseismic and postseismic deformation, Geophysical Research Letters, 28, 4047-4050. Mazzotti, S., X. Le Pichon, P. Henry, S. Miyazaki, 2000, Full interseismic locking of the Nankai and Japan-west Kurile subduction zones: An analysis of uniform elastic strain accumulataion in Japan constrained by permanent GPS, Journal of Geophysical Research, 105, 13159-13177. Minster, J.H., T.H. Jordan, B.H. Hager, D.C. Agnew, L.H. Royden, 1990, Implications of precise positioning, in Geodesy in the Year 2000, Committee on Geodesy, National Research Council, National Academy Press, 23-52.

6 Colloque International Technologies Nouvelles et Amélioration de la Gestion des Risques Majeurs, Kénitra, Morocco, 15-17 March 2004 Nakano, T., K. Hirahara, 1997, GPS observations of postseismic deformation for the 1995 Hyogo-ken Nanbu earthquake, Japan, Geophysical Research Letters, 24, 503-506. Nur, A. , Mavko, G., 1974, Postseismic viscoelastic rebound, Science, v. 183, 204-206. Ozawa, S., M. Murakami, T. Tada, 2001, Time-dependent inversion study of the slow thrust event in the Nankai trough subduction zone, southwestern Japan, Journal of Geophysical Research, 106, 787-802. Peterson, E.T., T. Seno, 1984, Factors affecting seismic moment release rates in subduction zones, Journal of Geophysical Research, 89, 233-248. Pollitz, F.F., Sacks, I.S., 1992, Modeling of postseismic relaxation following the great 1857 earthquake, southern California, Bulletin of the Seismological Society of America, v. 82, 454-480. Pollitz, F.F., R. Burgmann, P. Segall, 1998, Joint estimation of afterslip rate and postseismic relaxation following the 1989 Loma Prieta Earthquake, Journal of Geophysical Research, 103, 26975-26992. Savage, J.C., 1983, A dislocation model of strain accumulation and release at a subduction zone, Journal of Geophysical Research, 88, 4984-4996. Segall, P., 2002. Integrating geologic and geodetic estimates of slip on the San Andreas fault system, International Geology Review, v. 44, 62-82. Trenkamp, R., J.N. Kellogg, J.T. Freymueller, H. Mora, 2002, Wide plate margin deformation, southern Central America and northwestern South America, CASA GPS observations, Journal of South American Earth Sciences, 15, 157-171. Tse, S.T., and Rice, J.R., 1986, Crustal earthquake instability in relation to the depth variation of frictional slip properties: Journal of Geophysical Research, v. 91, 9452-9472. White, S.M., R. Trenkamp, J.N. Kellogg, 2003, Recent crustal deformation and the earthquake cycle along the Ecuador-Colombia subduction zone, Earth and Planetary Science Letters, 216, 231-242.