Hard classification of polyphase fault-slip data: Improvement and application

Abstract The quickest approach to the separation of polyphase fault-slip data into single-phase subsets is perhaps the recognition in stress space of the hyperplanes, consisting of their corresponding datum vectors, whose normals are parallel to the solutions of the stress vector. However the uncertainty of the solution sign makes it impossible to distinguish the subsets with respect to the opposite real stress vectors. In order to overcome this drawback the existing hard-classification method was improved by incorporating the slip senses during the separation. A series of artificial two-phase fault-slip datasets with the measurement errors generated under the imposed stresses were taken to test the improved method. There is a high accuracy of the stress solution even when the measurement errors have a range of [-12°, 12°]. The new method is further applied to data of shear bands measured on the limbs of the NNE-trending syncline at the Shihtipin resort, east of Taiwan. Two stress estimates representing the strike-slip or compressional tectonic regimes were obtained on each limb. They indicate an anticlockwise rotation of the maximum principal stress axis from ESE–WNW or SE–NW to E–W during the folding, probably in response to the oblique collision between the Luzon volcanic arc and the South China continent.