EXTRACTING TECTONIC RATES AND PALEO-EARTHQUAKE RECORDS FROM COSMOGENIC EXPOSURE AGES ALONG BEDROCK FAULT SCARPS

Cosmogenic surface exposure dating is a potentially powerful tool for reconstructing long term slip rates on active faults and may provide evidence for temporal earthquake clustering (e.g. Benedetti et al., GRL, v.29, 2002). Extensional faults in limestone are particularly amenable to this type of study because they commonly produce a striated bedrock scarp, exhumed by faulting, that can be sampled to obtain the cosmogenic ³⁶Cl concentration as a function of height up the scarp. The number of earthquakes, their timing and the magnitude of the associated slip determine the shape of ³⁶Cl profile. Existing methods for extracting paleo-earthquake records from such data use forward modelling and conclude that individual slip events ≥1m (Magnitude ≥ 6.5) may be resolved although a cluster of smaller magnitude events can produce a similar ³⁶Cl variation. Due to uneven scarp preservation, sample spacing in real data sets is variable (up to 10’s cm), further limiting our ability to extract tectonic information. Here we demonstrate a Monte Carlo inversion approach to extract fault slip histories objectively. We transform the ³⁶Cl concentration versus height data into a linear form. For a given set of fault slip events, event timings are determined using least squares inversion of the linearised data. By searching all possible combinations of slip event size (ranging from the sampling interval to the total scarp height), and using a range of different statistical measures for the goodness of fit, we rank the best fit scenario(s). By applying this approach to synthetic ³⁶Cl profiles, generated using a numerical fault growth model, we show how variable sample density and analytical error influence our interpretation of the true slip history (which for the synthetic data is known). We find that determining the timing of specific earthquakes requires (quasi-) uniform as well as dense sampling (5cm spacing), whereas long recurrence intervals (≥2000 yrs) (or periods with low throw rate) followed by a cluster of earthquakes (corresponding to a change to a higher throw rate) are well resolved even if the sampling density is lower (10cm). We use the conclusions drawn from our analysis of the synthetic data to guide our inversion of published real ³⁶Cl concentration data.