Southeastern U.S. recurrence relations are developed for the region and for various subregions (three geologic provinces; four seismic zones) from a combination of historical (1698--1977) and network (1978--1986) earthquake catalogs. The synthesis of these two data sets was accomplished by the application of MM intensity to mb(Lg) conversions in the form of discrete distributions to the historical catalog. Aftershocks were removed from the resulting catalog, it was tested for completeness, and then a maximum likelihood curve fitting technique for unequal periods of observation was employed. Tests were also made for different size magnitude intervals and different maximum and minimum magnitudes to assess the stability of the various data sets.

The recurrence equations are as follows: Southeastern United States, log NI=(3.13¿0.08)-(0.84¿0.02) mb(Lg), n=861; Valley and Ridge and Blue Ridge, log NI=(2.67¿0.08)-(0.82¿ 0.03) mb(Lg), n=395; Piedmont, log NI=(2.16¿0.18)- (0.81¿0.05) mb(Lg), n=191; Coastal Plain, log NI=(2.19¿ 0.16)-(0.78¿0.05) mb(Lg), n=188; Giles County, Virginia, log NI=(0.94¿0.25)-(0.64¿0.08) mb(Lg), n=28; central Virginia, log NI=(1.06¿0.19)-(0.64¿0.06) mb(Lg), n=94; eastern Tennessee, log NI=(2.75¿0.10)-(0.90¿0.04) mb(Lg), n=239; Charleston, South Carolina, log NI=(1.65¿ 0.25)-(0.77¿0.08) mb(Lg), n=84; where NI=number/yr in ¿0.25 magnitude intervals and n is the number of earthquakes input to the completeness testing and after deletion of aftershocks. The regional recurrence relation is similar to those for the New Madrid area (log Nc= 3.43 -- 0.88 mb) and the New England are (log Nc=2.87--0.84 Mc).

The primary purpose for the determination of recurrence relations for the various subregions is to gain understanding of their relative contributions to the earthquake frequencies for the region as a whole. A problem with this approach is that as smaller subsets of the catalog are considered, the precision of the log N versus m model results decreases. We see such an expected decrease, but invoke recurrence interval estimates from confidence interval (CI) bands, rather than simply point estimates, to assess the variability associated with our calculations. It is in the context of those interval estimates that we derive some insights into the relative contributions to the regional log N versus m. A principal result is that the region as a whole (minus the Charleston zone) has a much shorter recurrence interval (250--560) for a ''Charleston earthquake'' (mb(Lg)=6.75) than does the host Charleston zone (940--14,000). The millennia recurrence estimates are in accord with paleoseismic estimates for the Charleston area. If the Charleston zone is the only zone in the region capable of an mb(Lg)=6.75 shock, then the proper interevent time estimate from the seismicity data is that from the zone proper. If, however, the entire region outside of the Charleston area has a maximum earthquake potentia