The hypocentral distribution of locally recorded aftershocks of the great (Ms=8.1) Michoacan, Mexico, earthquake of September 19, 1985, defines a narrow Wadati‐Benioff zone structure, roughly 10 km thick, dipping 14° at N23°E. This is in good agreement with the source geometry obtained by waveform modeling of the 1985 Michoacan mainshock and the large 1979 Petatlán earthquake in the adjoining region. We inverted for the crustal velocity structure in the epicentral region by applying the Levenberg‐Marquardt non‐linear least squares algorithm to our local aftershock data. The velocity model consists of a layer with linearly increasing velocity in depth overlying a dipping, constant velocity halfspace. Our hypocentral location program uses a velocity model of the same form together with ray tracing. The earthquake hypocentral resolution obtained with this program is significantly better than that from conventional approaches (HYPO) and looks very promising for use in velocity structures with an important dipping interface like subduction zones.

The size and source complexity of the great (Ms -8.1) Michoacan, Mexico earthquake can be attributed to both a restriction of the downdip width of the seismic interface between the overriding and downgoing plates, and a uniform distribution of asperities. The seismogenic evidence of strain accumulation during the 20 years prior to the 19 September, 1985 mainshock closely resembles that observed prior to the smaller Mw--7.6 earthquakes in the adjacent Colima (1973) and Petatl•n (1979) regions: the subsequent rupture zones are seismically quiescent (rob >_4.0) for ~ 2.5 -4 years prior to the mainshocks; the locked thrust interface is loaded by aseismic slip and normal faulting within the downgoing plate below 25-30 km. There is evidence that the 1981 Playa Azul earthquake (Mw = 7.3) is an integral stage of the evolutionary process of strain accumulation and release in the Michoacan area and could thus be considered a foreshock to the great 1985 earthquake.

We found evidence that four episodes of large shallow (h < 30 km) interplate earthquakes ruptured approximately the same segment of the Middle America subduction zone in Ometepec, Guerrero, Mexico, during the last century: in 1890 (M = 7.2), 1937 (MS = 7.5), 1950 (MS = 7.3), and 1982 (MS = 6.9 and 7.0). This segment is about 70 km long, which is near the limit of resolution to define seismic gap dimensions. We synthesized the spatial, temporal, and mechanistic patterns of regional relocated seismicity (mb ≥ 4.0) to investigate the mechanics of rupture within the same seismic gap for the relatively large number of times (four) offered by the Ometepec case history. Despite the similarities in source depths and geometries of the shallow main shocks, we found marked differences in their magnitudes, recurrence intervals, and rupture modes (e.g., degree of source complexity and event multiplicity). The episodic amounts of energy release do not appear to scale with the observed elapsed times between the shallow main shocks (47.1, 13.0, and 31.5 years in chronological order) according to time‐ or slip‐predictable models of earthquake recurrence. We interpret these findings in the context of a variable rupture mode of the Ometepec region. This variable rupture mode suggests that probabilistic forecasts of future large earthquakes in the region, which are based solely on the historic record of main shocks, can be uncertain by at least ±57% of the average local repeat time of 30.5 years and 1/2 unit of magnitude. Specific predictions of future strong ground motions are obviated because earthquakes within the gap do not exhibit a “characteristic” behavior. We also found that all four episodes of interplate main shocks were preceded by large downdip normal fault earthquakes at intermediate depths (M ≥ 6.5, h > 60 km). The time intervals between extensional and subsequent compressional events range from 1.6 to 9.9 years with an average of 6.3±3.7 years. This apparent pattern which initiates at depth was further documented for the final 2/3 of the last seismic cycle. From the spatiotemporal distribution of seismicity during this final period, we identified three general stages of stress release. Stage I, from 1964 to January 1980, was characterized by scattered seismicity. Stage II, from February 1980 to June 6, 1982, showed a significant increase in downdip seismic activity while the shallow zone became quiescent. (There was also a period of apparent seismic quiescence from mid‐1975 to early 1978). Most seismic energy in stage II was released on October 24, 1980, by a downdip normal fault earthquake at intermediate depth (MS = 7.0, h = 65 km). Stage III began on June 7, 1982, with the occurrence of two large interplate thrust fault earthquakes at shallow depths (MS = 6.9, h = 18 km, and MS = 7.0, h = 19 and 23 km). The documentation of stage III through June 1986 shows subsidence of previous downdip activity and an aftershock zone which is spatially coincident with the shallow quiet zone identified in stage II. The spatiotempo...