Combining conflicting Bayesian models to develop paleoseismic records—An example from the Wasatch Fault Zone, Utah

Abstract

Bayesian statistical analyses of paleoseismic data result in the probabilistic determination of earthquake times using geochronological data evaluated in the context of a stratigraphic model. However, a fundamental problem in paleoseismology is how to use the Bayesian approach to model sparse and/or conflicting geochronological datasets, such as those derived from sites exhibiting episodic sedimentary and pedogenic processes in moderate‐ to high‐energy environments (e.g., a normal‐faulted alluvial fan). Using paleoseismic data for the Corner Canyon site on the Salt Lake City segment of the Wasatch fault zone (Utah), we develop an approach by which multiple Bayesian models are combined to generate an earthquake history at a site. This approach accommodates mutually exclusive interpretations of the geochronological data and thereby limits the influence of sparse data, stratigraphically inconsistent ages, or a single, subjective model interpretation. For the Corner Canyon site, we integrate four OxCal Bayesian models to generate a chronology of six events between ∼ 4.8 "> ∼ 4.8 and ∼ 0.5    ka "> ∼ 0.5 ka ⁠ . Late Holocene (post‐5 ka) mean recurrence and vertical slip‐rate estimates are ∼ 0.9    ky "> ∼ 0.9 ky (0.7–1.0 ky; 95% confidence) and 1.1    mm / yr "> 1.1 mm / yr ( ⁠ 0.8 – 1.7    mm / yr "> 0.8 – 1.7 mm / yr range), respectively. Although our method increases the uncertainty in the timing of individual earthquakes, it more objectively accounts for potential geochronological errors and different interpretations of stratigraphic age control. By relaxing the need to select a single age model, our approach yields more accurate earthquake‐timing results that will better facilitate evaluations of along‐fault event correlation and earthquake rupture length.