IASPEI - Picture Gallery

IUGG XXIV General Assembly July 2-13, 2007 Perugia, Italy
(S) - IASPEI - International Association of Seismology and Physics of the Earth's
Interior
JSS006 Oral Presentation 1931
Latent bias in temporal constraints of paleoseismic events in coastal
plains: an example from the Shonai-Heiya-Toen Fault in Northeast Japan
Dr. Shinji Toda
Active Fault Research Center Geological Survey of Japan, AIST
Temporal parameters of paleoseismicity to calculate earthquake probabilities on active faults are
commonly influenced by the preservation of datable sediments. The recent rapid accumulation of
paleoseismic data in Japan reminds us that paleo-earthquakes recorded in sediments are mostly
incomplete despite numerous excavations. We are aware of the limitations of geological method on the
faults particularly in coastal plains where sea level change directly controls the local sedimentation and
erosion during the datable time range of the radiocarbon analysis (in the past 50ka). Here we first
present a typical example of temporal bias of datable sedimentary units from the Shonai-heiya-toen
fault in northern Honshu, Japan, which mostly lacks sedimentary units in the period between 15 25 ka.
We then argue how temporal bias of the analyzed radiocarbon ages yielded from trench walls affects
estimates of average recurrence time. To evaluate such latent bias of preservation of paleoseismicity,
we not only examined the actual detail geological data themselves but also performed Monte Carlo
simulations using both synthetic earthquakes and observed radiocarbon ages. We computed the
following procedures: 1) Repeating earthquakes on a fault are synthetically produced by log-normal
probability density function with predetermined recurrence time (Trs) and coefficient of variation 0.5. 2)
Several radiocarbon samples (changeable parameter) are picked up from the actual C14 data recovered
from the Shonai-Heiya-toen fault. 3) Using the picked-up radiocarbon samples, some of the synthetic
earthquakes is constrained by sandwiching two ages. 4) If the number of the paleoseismic events
determined is more than one event, we calculate the average recurrence time (Tro) during the recent
10,000-yr and 50,000-yr periods. 5) We iterate 1-4 procedures 1000 times and make a histogram of the
estimated average recurrence times. We then compare the mean recurrence time (Trom) from
numerous Tro with the synthesized recurrence time (Trs). As the results of the simulations, ten random
C14 sampling from a trench leads to estimate the recurrence time twice as long as the real one for the
50ka case. Even we increase the number of samples up to 30, we still overestimate the recurrence time,
in other words Trom > Trs. It clearly attributes to several missing events originated from the deficiency
of sediments or possible erosion in the last glacial climax (MIS 2) due to sea level regression. For the
10ka case, we tend to overestimate Trom for Trs < 3000 yr together with C14 samples fewer than ~20.
In contrast, we underestimate Trom for Trs > 5000 yr even increasing the number of samples. We
speculate that 10,000-yr time window tricks us that it befalls the shorter intervals produced by the
combination of Trs > 5000 yr and coefficient of variation 0.5.
Keywords: paleoseismicity, recurrence time, earthquake probability