A stress inversion procedure for polyphase fault/slip data sets

Palaeostress estimation from striated fault data is frequently frustrated by the fact that natural fault data are heterogeneous in the sense that they cannot satisfactorily be explained by reactivation brought about by a single stress tensor. In many instances there is clear evidence to show that striation data can record the effects of multiple stress events. Any attempt to find the tensor which best fits (or explains) such dynamically mixed data sets risks determining a spurious stress state which is some form of compromise between the real stress states corresponding to different tectonic phases. To avoid this problem a strategy is proposed here which involves an initial separation of the raw data into coherent sub-sets prior to formal stress inversion. This separation is performed by assigning attributes to each fault which describe the faultʹs compatibility with trial stress tensors. Using these attributes faults can be grouped into dynamically-homogeneous families using the statistical techniques of cluster analysis.