Reliability of methods to separate stress tensors from heterogeneous fault-slip data

The reliability of methods for separating palaeostress tensors from heterogeneous fault-slip data is evaluated. The methods of Etchecopar et al. (1981), Yamaji (2000), and the cluster procedure of Nemcok and Lisle (1995) are assessed but the results can probably be extrapolated to other methods based on similar assumptions. Heterogeneous fault-slip data sets, artificially generated by mixing two natural homogeneous data sets, have been used to evaluate both the role of the relative dominance (in number of faults taken from each tensor) and the difference between the parent tensors. The results obtained from a natural heterogeneous data set were compared with additional field data to evaluate and constrain the tensor separation process as well. Results suggest that attempts to devise a fully automatic separation procedure for distinguishing homogeneous data sets from heterogeneous ones will be unsuccessful because the researcher will always be required to take some part in the correct choice of the tensors. In this sense, additional structural data such as geometrical characteristics of the faults (e.g. conjugate or quasi-conjugate Andersonian systems), stylolites or tension gashes will be very useful for the correct separation of stress tensors from fault-slip data.