Evaluating variability in coseismic slips of paleo-earthquakes from an incomplete slip history: an example from displaced terrace flights across the Kamishiro fault, central Japan

Abstract

Examining the regularity in slip over seismic cycles leads to an understanding of earthquake recurrence and provides the basis for probabilistic seismic hazard assessment. Systematic analysis of three-dimensional paleoseismic trenches and analysis of offset markers along faults reveal slip history. Flights of displaced terraces have also been used to study slips of paleoearthquakes when the number of earthquakes contributing to the observed displacement of a terrace is known. This study presents a Monte Carlo-based approach to estimating slip variability using displaced terraces when a detailed paleoseismic record is not available. First, we mapped fluvial terraces across the Kamishiro fault, which is an intra-plate reverse fault in central Japan, and systematically measured the cumulative dip slip of the mapped terraces. By combining these measurements with the age of the paleoearthquakes, we estimated the amount of dip slip for the penultimate event (PE) and antepenultimate event (APE) to be 1.6 and 3.4 m, respectively. The APE slip was nearly three times larger than the most recent event of 2014 (Mw 6.2): 1.2 m. This suggests that the rupture length of the APE was much longer than that of the 2014 event and the entire Kamishiro fault ruptured with adjacent faults during the APE. Thereafter, we performed the Monte Carlo simulations to explore the possible range of the coefficient of variation for slip per event (COVs). The simulation considered all the possible rupture histories in terms of the number of events and their slip amounts. The resulting COVs typically ranged between 0.3 and 0.54, indicating a large variation in the slip per event of the Kamishiro fault during the last few thousand years. To test the accuracy of our approach, we performed the same simulation to a fault whose slip per event was well constrained. The result showed that the error in the COVs estimate was less than 0.15 in 86 % of realizations, which was comparable to the uncertainty in COVs derived from a paleoseismic trenching. Based on the accuracy test, we conclude that the Monte Carlo-based approach should help assess the regularity of earthquakes using an incomplete paleoseismic record.