The New Madrid Seismic Zone (NMSZ), in the central United States, was the site of large magnitude (M) earthquakes in 1811-12. Paleoseismic studies find events of similar size recurring about once every 500 years. Assessing the causes of deformation and estimating the recurrence of these events are necessary for determining earthquake hazard in the region.

Global Positioning System (GPS) measurements made in 1991, 1993 and 1997 at 23 sites within 300 km of the NMSZ show minimal motion across the fault zone. Using a locked fault model, I find -0.2 +/- 2.4 mm/yr (2 sigma) of right-lateral fault parallel (NE-SW) motion across the network. This result is consistent with continuous GPS results away from the NMSZ which indicate less than 1 mm/yr of differential plate motion. Assuming horizontal slip accumulates at less than 2 mm/yr, the recurrence interval (T_r) for ``New Madrid style'' M 8 earthquakes, with 5-10 m slip, would exceed 2,500-5,000 years, significantly greater than the 500-1,000 years previously estimated. However, these GPS results are consistent with my re-evaluation of the frequency-magnitude relationship, paleoseismic recurrence estimates, and lack of fault-associated topography if the 1811-12 earthquakes and the paleoearthquakes if the events were smaller, low M 7, and recurring every 500-1000 years. Also, smaller M 7 events are consistent with a new earthquake intensity study for the 1811-12 series. Reducing the size of the maximum magnitude events (M_max) has considerable effect on the predicted seismic hazard in New Madrid.

Currently, the U.S.G.S. seismic hazard maps show large accelerations for the central U.S. due to the NMSZ. These maps reflect crucial parameter assumptions, having large uncertainties because little data exists for large events. Since my GPS and seismic recurrence studies, along with other recent results, show that the M_max may be 7 with T_r of 500 years, rather than M_max 8 and T_r of 1000 years used by the U.S.G.S., I explore the effects that these parameters, along with several predicted ground shaking models, have on predicted 1 Hz and peak ground accelerations from the NMSZ. I find that changing these parameters significantly affects the predicted accelerations thus demonstrating the considerable uncertainties in seismic hazard from the NMSZ.