Paleoseismology of the Gazikoy-Saros segment of the North Anatolia fault, northwestern Turkey: Comparison of the historical and paleoseismic records,implications of regional seismic hazard, and models of earthquake recurrence

We excavated five trenches across the North Anatolia fault zone (NAFZ) alon g the Ganos fault (Gazikoy-Saros segment), which last produced surface rupt ure in 1912, near Kavakkoy where the fault enters the Gulf of Saros. The tr enches exposed faulted sediments in a flood-plain environment with abundant detrital charcoal and scattered land-snail shells. Twenty-two radiocarbon dates place constraints on the ages of the exposed sediments, which range f rom less than a few hundred years to about 6000 years in age. In two closel y spaced trenches, we identified five discrete earthquake event horizons in the upper 2.5 m of stratigraphy based on abrupt upward termination of shea r zones, folding, fissuring, and abrupt stratigraphic thickening, four of w hich may corresponded to historically recorded large regional earthquakes. The earliest of the identified events occurs below an unconformity and date s to about 4 ka B.P. The more recent four events all occurred within the pa st 1000-1200 years and may correspond to large earthquakes in A.D. 824, ca 1354, 1509, 1766 and 1912 (Ambraseys and Finkel, 1987, 1991, 1995). In anot her trench, we identified at least two events that have occurred during the past 500 years and probably correspond to the large events of 1766 and 191 2. These observations support an average return period of about 250-300 yea rs for the Gazikoy-Saros segment of the NAFZ. They also suggest that this s egment, which is bound both to the east and west by large releasing stepove rs, behaves in a quasi-periodic fashion, at least for the past several surf ace ruptures. Most of the 23 mm/yr of dextral shear between Anatolia and Eu rope observed by GPS occurs on the North Anatolian fault. We use 18 mm/yr a nd the similar to 250-300 year recurrence rate, as determined from our tren ching and the historical record, to suggest that each of the earthquakes ob served in our trenches produced several meters of slip, consistent with the ir inferred sizes from the extent of historical damage. Considering that Is tanbul has not suffered a large nearby event in the Marmara Sea since 1766, we suggest that about 4 m of strain has accumulated across faults in the M armara during these past centuries. This is similar to the average slip in many of the large earthquakes on the North Anatolian fault this century. If released seismically, this could result in an earthquake in the M 7.2-M 7. 6 range, similar to the August and November, 1999 earthquakes east of the M armara Sea.