Slow deformation and lower seismic hazard at the New Madrid seismic zone

Science

Volumn 284, Issue 5414, 1999, Pages 619-621

  Newman, Andrew ^a   Stein, Seth ^a   Weber, John ^b   Engeln, Joseph ^c   Mao, Ailin ^d   Dixon, Timothy ^d  

^a NORTHWESTERN UNIVERSITY   (United States)

^b GRAND VALLEY STATE UNIVERSITY   (United States)

^c UNIVERSITY OF MISSOURI   (United States)

^d UNIVERSITY OF MIAMI   (United States)

Author keywords

[No Author keywords available]

Indexed keywords

EARTHQUAKE RECURRENCE; GPS; SEISMIC HAZARD;

ARTICLE; DISASTER PLANNING; EARTHQUAKE; FREQUENCY ANALYSIS; GEOLOGY; GEOMETRY; HAZARD; HISTORY; MEASUREMENT; PALEOANTHROPOLOGY; PRIORITY JOURNAL; TELECOMMUNICATION; UNITED STATES; VELOCITY;

UNITED STATES;

EID: 0033597455     PISSN: 00368075     EISSN: None     Source Type: Journal    
DOI: 10.1126/science.284.5414.619     Document Type: Article

References (35)

1. See the review by A. Johnston and E. Schweig, Annu. Rev. Earth Planet. Sci. 24, 339 (1996).
2. A. Johnston, Geophys. J. Int. 126, 314 (1996);
3. L. Himes, W. Stauder, R. Herrmann, Seism. Res. Lett. 59, 123 (1988); J. Chiu, A. Johnston, Y. Yang, ibid. 63, 375 (1992).
4. L. Himes, W. Stauder, R. Herrmann, Seism. Res. Lett. 59, 123 (1988); J. Chiu, A. Johnston, Y. Yang, ibid. 63, 375 (1992).
5. R. Herrmann and J. Canas, Bull. Seismol. Soc. Am. 68, 1095 (1978); R. Herrmann, J. Geophys. Res. 84, 3543 (1979).
6. R. Herrmann and J. Canas, Bull. Seismol. Soc. Am. 68, 1095 (1978); R. Herrmann, J. Geophys. Res. 84, 3543 (1979).
7. L. Braile et al., Tectonophysics 131, 1 (1986); J. Sexton et al., Geophysics 51, 640 (1986).
8. L. Braile et al., Tectonophysics 131, 1 (1986); J. Sexton et al., Geophysics 51, 640 (1986).
9. M. L. Zoback and M. D. Zoback, Science 213, 96 (1981); J. Granga and R. Richardson, J. Geophys. Res. 101, 5445 (1996).
10. M. L. Zoback and M. D. Zoback, Science 213, 96 (1981); J. Granga and R. Richardson, J. Geophys. Res. 101, 5445 (1996).
11. J. Weber, S. Stein, J. Engeln, Tectonics 17, 250 (1998).
12. The sites occupied are listed in (7), except CRVL, which was buried by the construction of a baseball Field. The 1997 survey was similar to that in 1993 except for shorter (2 day) site occupations.
13. GPS data were analyzed at the University of Miami following the procedures of T. Dixon et al. [J. Geophys. Res. 102, 12017 (1997)]. We used high-precision nonfiducial satellite orbits and the Jet Propulsion Laboratory GIPSY analysis software to estimate site velocities in the ITRF-96 reference frame. Site velocities were estimated from least-squares fits to daily positions, weighted by errors following the method of Mao et al. (11). The 1991 data have higher root mean square (rms) scatters compared with 1993 and later data. Scaling the 1991 position errors with the correlations between weighted rms and white and colored noise (11) de-weights them compared with later data.
14. We subtracted motion predicted by an Euler vector in ITRF-96 for stable North America (1.16°S, 80.2°W, 0.193° per million years) determined by inversion of GPS data from 16 continuous stations (11). This subtraction removes most network-wide motion, although a small (but not statistically significant) east-ward drift remains. Whether this drift is tectonically significant is unclear; it has no effect on our velocity gradient analysis (Fig. 2) because a mean velocity is also removed.
15. Until recently, most GPS studies assumed that geodetic noise is uncorrelated in time. Recent studies show that GPS noise is time-correlated and that assuming noise is uncorrelated can underestimate velocity errors by up to an order of magnitude [A. Mao, C. Harrison, T. Dixon, J. Geophys. Res. 104, 2797 (1999)].
16. J. Savage and R. Burford, J. Geophys. Res. 78, 832 (1973).
17. We compared geodetic data to a model in which steady far-field motion loads a fault before earthquake rupture. This model is commonly used for plate boundaries, where space geodetic data show rates of plate motion consistent with those over millions of years, indicating that steady far-field motions give rise to episodic earthquakes [R. Gordon and S. Stein, Science 256, 333 (1992); S. Stein, in Space Geodesy and Geodynamics, Geodynamics Ser. 23, D. Smith and D. Turcotte, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 5-20]. Similarly, space geodetic data (for example, Fig. 3) show that plates thought to have been rigid on geological time scales are quite rigid on decadal scales. Hence, application of these ideas of steady motion to intraplate settings seems plausible but has not been demonstrated. We did not consider strain transients after the 1811-1812 earthquakes [P. Rydelek and F. Pollitz, Geophys. Res. Lett. 21, 2303 (1994)], which predict motions larger than interseismic motion. Similarly, our geodetic approach implicitly focuses on motions due to platewide rather than locally derived stresses (6).
18. We compared geodetic data to a model in which steady far-field motion loads a fault before earthquake rupture. This model is commonly used for plate boundaries, where space geodetic data show rates of plate motion consistent with those over millions of years, indicating that steady far-field motions give rise to episodic earthquakes [R. Gordon and S. Stein, Science 256, 333 (1992); S. Stein, in Space Geodesy and Geodynamics, Geodynamics Ser. 23, D. Smith and D. Turcotte, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 5-20]. Similarly, space geodetic data (for example, Fig. 3) show that plates thought to have been rigid on geological time scales are quite rigid on decadal scales. Hence, application of these ideas of steady motion to intraplate settings seems plausible but has not been demonstrated. We did not consider strain transients after the 1811-1812 earthquakes [P. Rydelek and F. Pollitz, Geophys. Res. Lett. 21, 2303 (1994)], which predict motions larger than interseismic motion. Similarly, our geodetic approach implicitly focuses on motions due to platewide rather than locally derived stresses (6).
19. We compared geodetic data to a model in which steady far-field motion loads a fault before earthquake rupture. This model is commonly used for plate boundaries, where space geodetic data show rates of plate motion consistent with those over millions of years, indicating that steady far-field motions give rise to episodic earthquakes [R. Gordon and S. Stein, Science 256, 333 (1992); S. Stein, in Space Geodesy and Geodynamics, Geodynamics Ser. 23, D. Smith and D. Turcotte, Eds. (American Geophysical Union, Washington, DC, 1993), pp. 5-20]. Similarly, space geodetic data (for example, Fig. 3) show that plates thought to have been rigid on geological time scales are quite rigid on decadal scales. Hence, application of these ideas of steady motion to intraplate settings seems plausible but has not been demonstrated. We did not consider strain transients after the 1811-1812 earthquakes [P. Rydelek and F. Pollitz, Geophys. Res. Lett. 21, 2303 (1994)], which predict motions larger than interseismic motion. Similarly, our geodetic approach implicitly focuses on motions due to platewide rather than locally derived stresses (6).
20. T. Dixon, A. Mao, S. Stein, Geophys. Res. Lett. 23, 3035 (1996). These results are updated here with additional sites [A. Mao, thesis, University of Miami (1998)].
21. T. Dixon, A. Mao, S. Stein, Geophys. Res. Lett. 23, 3035 (1996). These results are updated here with additional sites [A. Mao, thesis, University of Miami (1998)].
22. L. Liu, M. Zoback, and P. Segall [Science 257, 1666 (1992)] used GPS to remeasure monuments previously measured by triangulation and reported rapid strain accumulation in the southern NMSZ corresponding to 5 to 7 mm/year of slip. A similar study across the northern NMSZ found strain rates indistinguishable from zero [R. Snay, J. Ni, H. Neugebauer, in U.S. Geol. Surv. Prof. Pap. 1538-F (1994), pp. F1-F6]. Our earlier study (7), based on the first two GPS occupations of presumably more stable monuments, found a far-field rate of 3 ± 3 mm/ year (limits are from the approach used in Fig. 2), indistinguishable from zero at 2σ. Hence, improved geodetic techniques and longer measurements generally reveal successively slower motion, presumably because the far-field velocity is small (or zero). Because the data have uncertainties, the first two measurements typically overestimate the velocity, which successive measurements better approximate. Unless the uncertainties are well understood, the estimated velocity may appear unduly significant (11). For New Madrid, the older triangulation data were presumably less accurate than GPS because of limitations of the technique, possibly compounded by instability of shallow-rooted triangulation pillars. Moreover, for low-strain rate areas a few measurements can change triangulation results significantly [R. Snay, J. Geophys. Res. 91, 12695 (1986)]. Our GPS surveys use deeper rooted and presumably more stable monuments and extend outside the embayment, but geodetic GPS technology was still immature in 1991. By 1993, improved GPS receivers and an improved network of global tracking sites yielded better data, as determined by better repeatability of site positions between successive days. GPS velocity analysis is also less sensitive to site-specific errors than the triangulation analysis. Therefore, we consider the GPS results here more accurate than earlier surveys.
23. L. Liu, M. Zoback, and P. Segall [Science 257, 1666 (1992)] used GPS to remeasure monuments previously measured by triangulation and reported rapid strain accumulation in the southern NMSZ corresponding to 5 to 7 mm/year of slip. A similar study across the northern NMSZ found strain rates indistinguishable from zero [R. Snay, J. Ni, H. Neugebauer, in U.S. Geol. Surv. Prof. Pap. 1538-F (1994), pp. F1-F6]. Our earlier study (7), based on the first two GPS occupations of presumably more stable monuments, found a far-field rate of 3 ± 3 mm/ year (limits are from the approach used in Fig. 2), indistinguishable from zero at 2σ. Hence, improved geodetic techniques and longer measurements generally reveal successively slower motion, presumably because the far-field velocity is small (or zero). Because the data have uncertainties, the first two measurements typically overestimate the velocity, which successive measurements better approximate. Unless the uncertainties are well understood, the estimated velocity may appear unduly significant (11). For New Madrid, the older triangulation data were presumably less accurate than GPS because of limitations of the technique, possibly compounded by instability of shallow-rooted triangulation pillars. Moreover, for low-strain rate areas a few measurements can change triangulation results significantly [R. Snay, J. Geophys. Res. 91, 12695 (1986)]. Our GPS surveys use deeper rooted and presumably more stable monuments and extend outside the embayment, but geodetic GPS technology was still immature in 1991. By 1993, improved GPS receivers and an improved network of global tracking sites yielded better data, as determined by better repeatability of site positions between successive days. GPS velocity analysis is also less sensitive to site-specific errors than the triangulation analysis. Therefore, we consider the GPS results here more accurate than earlier surveys.
24. L. Liu, M. Zoback, and P. Segall [Science 257, 1666 (1992)] used GPS to remeasure monuments previously measured by triangulation and reported rapid strain accumulation in the southern NMSZ corresponding to 5 to 7 mm/year of slip. A similar study across the northern NMSZ found strain rates indistinguishable from zero [R. Snay, J. Ni, H. Neugebauer, in U.S. Geol. Surv. Prof. Pap. 1538-F (1994), pp. F1-F6]. Our earlier study (7), based on the first two GPS occupations of presumably more stable monuments, found a far-field rate of 3 ± 3 mm/ year (limits are from the approach used in Fig. 2), indistinguishable from zero at 2σ. Hence, improved geodetic techniques and longer measurements generally reveal successively slower motion, presumably because the far-field velocity is small (or zero). Because the data have uncertainties, the first two measurements typically overestimate the velocity, which successive measurements better approximate. Unless the uncertainties are well understood, the estimated velocity may appear unduly significant (11). For New Madrid, the older triangulation data were presumably less accurate than GPS because of limitations of the technique, possibly compounded by instability of shallow-rooted triangulation pillars. Moreover, for low-strain rate areas a few measurements can change triangulation results significantly [R. Snay, J. Geophys. Res. 91, 12695 (1986)]. Our GPS surveys use deeper rooted and presumably more stable monuments and extend outside the embayment, but geodetic GPS technology was still immature in 1991. By 1993, improved GPS receivers and an improved network of global tracking sites yielded better data, as determined by better repeatability of site positions between successive days. GPS velocity analysis is also less sensitive to site-specific errors than the triangulation analysis. Therefore, we consider the GPS results here more accurate than earlier surveys.
25. Earthquake populations approximately follow log N = a - bM, where N is the number of earthquakes (total or annual) whose magnitude exceeds M. Because b is about 1, earthquakes of a given size are about one-tenth as numerous as those one magnitude unit smaller. Although body and surface wave magnitudes do not exceed about 6.4 and 8.4, respectively, earthquake catalogs typically show b to be about 1 because body wave magnitudes are reported for small earthquakes and surface wave magnitudes are reported for large earthquakes [E. Okal and B. Romanowicz, Phys. Earth. Planet. Int. 87, 55 (1994)]. Hence, if the linear frequency-magnitude relation is used, magnitudes above 6.4 should be treated as surface wave magnitudes.
26. Extrapolation of the recurrence of larger earthquakes from the rate of smaller earthquakes, used because of the limited data available for earthquakes before this century, faces various uncertainties. M. Stirling, S. Wesnousky, and K. Shimazaki [Geophys. J. Int. 124, 883 (1996)] found that such extrapolation overestimates recurrence times inferred from geological data, whereas E. Triep and L. Sykes [J. Geophys. Res. 102, 9923 (1997)] found that this extrapolation underpredicts recurrence times for large intracontinental earthquakes.
27. Extrapolation of the recurrence of larger earthquakes from the rate of smaller earthquakes, used because of the limited data available for earthquakes before this century, faces various uncertainties. M. Stirling, S. Wesnousky, and K. Shimazaki [Geophys. J. Int. 124, 883 (1996)] found that such extrapolation overestimates recurrence times inferred from geological data, whereas E. Triep and L. Sykes [J. Geophys. Res. 102, 9923 (1997)] found that this extrapolation underpredicts recurrence times for large intracontinental earthquakes.
28. For 1974 to 1998, the New Madrid catalog (http:// elwe.ceri.memphis.edu/∼seisadm) of earthquakes with seismologically determined magnitudes yields a and b values of 3.446 ± 0.041 (1σ) and 0.954 ± 0.013. For 1816 to 1984 (beginning at 1816 excludes major aftershocks), Nuttli's catalog (http:// www.eas.slu.edu/Earthquake_Center), in which body wave magnitudes before 1964 are typically inferred from the reported shaking, yields a and b values of 4.537 ± 0.105 and 1.079 ± 0.022. Combining the two lines predicts recurrence times (with 2σ uncertainties) for magnitude 7 and 8 earthquakes of about 1400 ± 600 and 14,000 ± 7000 years. These values seem plausible: since 1816, there have been 16 earthquakes with magnitude greater than 5 (about a 10-year recurrence), and two with magnitude greater than 6 (about a 100-year recurrence), so magnitude 7 and 8 earthquakes should have about 1000- and 10,000-year recurrences, respectively. These estimates do not depend on whether the seismogenic stresses are local or platewide in origin (5, 13).
29. s) > 8.3. This frequent recurrence results from treating magnitude 7 earthquakes as body wave magnitude 7, and equating them to surface wave magnitude 8.3. Recent results (16) indicate that such earthquakes are better treated as surface wave magnitude 7.
30. S. Wesnousky and L. Leffler, Bull. Seismol. Soc. Am. 82, 1756 (1992); M. Tuttle and E. Schweig, Geology 23, 253 (1995); K. Kelson et al., J. Geophys. Res. 101, 6151 (1996).
31. S. Wesnousky and L. Leffler, Bull. Seismol. Soc. Am. 82, 1756 (1992); M. Tuttle and E. Schweig, Geology 23, 253 (1995); K. Kelson et al., J. Geophys. Res. 101, 6151 (1996).
32. S. Wesnousky and L. Leffler, Bull. Seismol. Soc. Am. 82, 1756 (1992); M. Tuttle and E. Schweig, Geology 23, 253 (1995); K. Kelson et al., J. Geophys. Res. 101, 6151 (1996).
33. E. Schweig and M. Ellis, Science 264, 1308 (1994).
34. A. Frankel et al., National Seismic Hazard Maps Documentation: U.S. Geol. Surv. Open-File Rep. 96-532 (1996). These maps assume that a magnitude 8 earthquake occurs every 1000 years at New Madrid, so the predicted peak ground acceleration expected in 50 years at 2% probability for the NMSZ exceeds that in San Francisco, and the predicted highest acceleration (exceeding 1.2g) area for the NMSZ is larger than for Los Angeles or San Francisco.
35. Supported by NASA grants NAGW-2522 and NAG5-6685. We thank R. Graves, K. Wellington, C. Bunker, K. Nolan, and C. Griffen for field assistance in the third survey; F. Farina for assistance with data processing; E. Okal and K. Mueller for helpful discussions; R. Snay and an anonymous reviewer for useful reviews; and University Navstar Consortium for technical support.