Reliability of palaeostress analysis from fault striations in near multidirectional extension stress fields. Example from the Ebro Basin, Spain

A standard palaeostress analysis was carried out by applying a sequence of the Right Dihedra, y–R diagram and Etchecopar methods to 37 fault samples collected in Miocene deposits of the central Ebro Basin. This gave rise to 37 standard solutions which show a pervasive multidirectional extension stress regime for this region. Owing to the possible low definition of the horizontal stress axes in this type of stress regime, the results were then submitted to a stability test by subsampling the initial data sets. The level of stability of the solutions is good, as defined by the differences between successive solutions and the standard one. Most individual stress tensors so obtained differ by less than 15° in σ3 azimuth and 0.1 in stress ratio R=(σ2−σ3)/(σ1−σ3), respectively, from the corresponding standard solutions. Furthermore, the application of the Central Limit Theorem by averaging solutions from groups of subsamples demonstrates that the majority of the standard stress tensors are quite exact. The regional homogeneity of the final results is interpreted as a further proof of reliability. ity and reliability increase as the sample size grows. There is no ‘magic’ number of data which separates reliable from unreliable stress solutions, but a quantitative approach allows us to estimate the expected error and the associated probability for each sample size. A sufficient stability may be accomplished using monophase fault samples from 25–30 faults, and it does not increase significantly for samples larger than 40 faults. For samples under 20–25 faults in uniaxial stress regimes (especially in multidirectional extension), the subsampling/averaging technique may be a useful tool. This method also allows testing and, in some special cases (these where subsample mean tensors provide stable results that systematically differ from the initial standard tensor), improving stress solutions.