COMPARISON OF TRISHEAR AND ELASTIC DISLOCATION FAULT-PROPAGATION FOLD MODELS AND CHARACTERIZATION OF UNCERTAINTY IN MODEL PARAMETERS

Fault-related fold models, such as the trishear kinematic model for fault-propagation folding, provide a valuable means for inferring quantities such as fault slip and fault and fold geometry at depth. Such models are subject to substantial uncertainty, however. Uncertainty in the parameters of a given model can be quantified using stochastic inverse modeling methods, which deliver a range of possible values for the parameters rather than a single unique solution. We demonstrate this approach with simulated annealing and Markov chain Monte Carlo methods applied to trishear models of laboratory and natural examples of fault-propagation folds. A second source of uncertainty not addressed by these methods, however, arises from the choice of model. A variety of kinematic and mechanical models for folding exist, requiring different assumptions about the folding process, and often making different predictions of important quantities such as fault slip. This epistemic uncertainty is more difficult to quantify and requires the testing and comparison of a variety of possible models. To test the effect of this type of uncertainty, we consider models of the Santa Fe Springs anticline and underlying Puente Hills thrust fault – a structure that has been the subject of previous trishear modeling studies and that is a source of significant seismic hazard. Using simulated annealing to fit models to this structure, we compare six previously-published trishear velocity fields as well as an elastic dislocation model. Results suggest that the range of possible values for fault slip, fault tip position, and propagation to slip ratio among the different models significantly exceeds the range for any one model alone. To compare models, we calculate the Akaike and Bayesian information criteria (AIC and BIC), and we use BIC values to calculate averages for model parameters using Bayesian model averaging. AIC and BIC are criteria for model selection, favoring a better fit to data but penalizing model complexity. For the Santa Fe Springs anticline, they suggest that trishear and elastic dislocation models are of similar quality, while the improvement in fit achieved by more complicated trishear models involving asymmetry or center concentration is not sufficient to justify their greater complexity compared to a linear trishear velocity field.