Periodic, random or clustered? Constraining earthquake recurrence in different tectonic settings using lake paleoseismology

Abstract:

Probabilistic seismic hazard analysis requires good knowledge of strong earthquake recurrence parameters that goes beyond selecting an “average” recurrence interval. Due to the typical short temporal span of instrumental and historical data, it is unclear whether quasi-periodic, weakly-periodic, random or clustered recurrence models should be used. In recent years, a fair amount (20-30) of long paleoseismic records in lakes and have been constructed that encompass sufficient events to address this question.

We present a worldwide compilation of published long lacustrine paleoseismic records grouped per tectonic domain, and statistically explore the variability of their recurrence intervals. By resampling these empirical records and synthetic recurrence patterns, we discuss how many events are needed in a record to robustly constrain some basic recurrence parameters, such as the Coefficient of Variation (CoV), and how this value is affected by age uncertainty and the nature of the underlying recurrence model.

We find that plate boundary settings generally exhibit a quasi-to-weakly periodic recurrence behavior, characterized by a rising occurrence probability with elapsed time since the last event. In contrast, intraplate settings are characterized by a Poissonian (time-independent) or clustered model and either a constant probability or an enhanced one shortly after an event. The geometry and distribution of seismic sources modulates these patterns: a CoV of 0.3-0.4 can be interpreted as caused by a simple, isolated seismic source, whereas a CoV of ~1 may indicate the additive effect of several seismic sources capable of leaving a sedimentary fingerprint in the lake. Most plate boundary records show a CoV of 0.5-0.7 and thus a mixture of a dominant source and some minor ones. In both subduction and transform domains, it seems that lower overall slip rates lead to higher recurrence variability (high CoV), with the Dead Sea region as the most extreme example of clustering (CoV 1-1.5).

Besides for seismic hazard analysis, constraining earthquake recurrence is also crucial for evaluating cascading hazards. This is exemplified by a new record from the Eastern Alps where we infer unusually strong and frequent seismic shaking as the main preparatory and triggering factor of massive prehistoric rockslides.