The Mechanics of Intraplate Earthquake Generation: Application to the New Madrid Seismic Zone, U.S.A.

S. J. Kenner and P. Segall, Stanford University

Abstract

In contrast to plate boundary earthquakes, the mechanics of intraplate seismicity is poorly understood. The ~200 km long New Madrid Seismic Zone (NMSZ), which experienced three large earthquakes in 1811-1812, is one of the best studied zones of intraplate seismicity. Paleoseismic evidence suggests repeated large events every 500 to 1000 years during the Holocene, yet cumulative fault offsets could be as little as ~100 m. Recent geodetic measurements indicate that the rate of strain accumulation is below the current detection threshold. This has led to the suggestion that seismic hazard in the region has been significantly overstated [Newman et al., 1999]. However, without viable mechanical models for intraplate seismicity, interpretation of these data with respect to seismic hazard is ambiguous.
Here we present a time dependent mechanical model for the generation of repeated, intraplate earthquakes, which incorporates a weak lower crustal zone within an otherwise elastic lithosphere. Initially, the entire body is uniformly stressed and constant, far-field stress boundary conditions are applied. Subsequent relaxation of the weak zone transfers stress to the overlying crust which loads the seismogenic fault, potentially generating a sequence of earthquakes which continues until the weak zone fully relaxes. For an appropriate choice of parameters, the model predicts repeating sequences of large (5-10 meter) slip events with recurrence intervals of 500 to 4000 years and cumulative offsets of order 100 meters. Relative displacements in the far-field are zero, consistent with intraplate environments. In most cases, interseismic strain-rates computed between large, potentially hazardous, slip events would not be detectable given available geodetic data, implying that low observed rates of strain accumulation cannot be used to rule out the occurrence of future damaging earthquakes within the NMSZ.

Figure 5.1: Regional setting and seismicity (filled circles) within the NMSZ. The hatched regions are plutons. Thick lines denote the boundary of the Reelfoot rift. Thinner northwest trending lines denote the approximate lateral extent of the Missouri batholith. Solid triangles denote hypothetical sites in the 5 station continuous network used in the GPS uncertainty calculations. Both solid and open triangles are included in the 28 station campaign network.

Figure 5.2: Model schematic including dimensions and rheological distribution. Light gray areas are elastic. Dark gray areas are composed of time dependent materials. White areas represents the seismogenic fault. To make the numerical calculations tractable a symmetry condition is applied along a plane perpendicular to the fault at its centerline.

Figure 5.3: Cumulative moment vs. time for 1) t^¥ = 60 MPa, t^max = 62 MPa, t^residual = 50 MPa, W_w = 75 km, and d_w = 40 km (thin solid line), 2) t^¥ = 25 MPa, t^max = 27 MPa, t^residual = 15 MPa, W_w = 75 km, and dw = 40 km (thin dashed line), and 3) t^¥ = 60 MPa, t^max = 62 MPa, t^residual = 50 MPa, W_w = 18 km, and d_w = 40 km (thin dash-dotted line). The moment release rate slows with time and recurrence intervals for the largest events increases. Because details of the rupture process are not accurately modeled, only the dominant features of the model, i.e., the largest slip events, are considered (thick gray lines).

Figure 5.4: (a) Engineering shear strain-rates across a 100 km region bracketing the fault at its midpoint assuming a fault oriented N30°E. Numerical results are plotted for Model 1 (Figure 3) at 1360 years, 3850 years, and 8650 years after the initiation of weak zone relaxation. Because the relaxation process starts instantaneously these are upper bounds. After ~2000 years model strain-rates are within the uncertainties of the geodetic measurements. For models with narrower weak zones and/or lower remote stress levels (e.g., Models 2 and 3, Figure 3), calculated strain-rates are even lower. Model results are compared with NMSZ strain-rates (95% confidence ellipses) estimated for two networks from campaign GPS measurements made by Stanford University in 1991, 1993, and 1997 [M. H. Murray, personal communication]. Uncertainties in the strain-rates are scaled by the fit to an assumed spatially uniform strain-rate field. Since individual station velocities within the NMSZ are inconsistent with this assumption, presumably due to local site instabilities, the uncertainty is increased by a factor of 3.3 [M. H. Murray, personal communication]. Positive g1 represents left-lateral shear across a NW-SE striking fault. Positive g2 represents left-lateral shear across a N-S striking fault. (b) The 1991 network is confined to the Reelfoot rift and its northwest boundary. Additional stations (filled triangles) were included in the 1997 network which extends further from the rift boundary and was expanded to the southwest. Open triangles represent new stations added during the 1997 survey.

Figure 5.5: Uncertainty (1s) in engineering strain-rate as a function of time for two candidate networks considered in the GPS uncertainty analysis (Figure 1). Campaign measurements (dashed lines) are annual. Continuous measurements (solid lines) are recorded daily. Y (mm/Öyr) represents the scale of random walk processes associated with local benchmark motion.

Reference: Kenner, S.J., and P. Segall, A Mechanical Model for Intraplate Earthquakes: Application to the New Madrid Seismic Zone, Science, v. 289, p. 2329-2332, September 2000.