Large earthquakes often produce huge loss in economy and life, for example, the Mw7.9 earthquake having occurred in Wenchuan, China in 2008 killed more than 90 thousand people and ruined almost all the cities and villages along the over-100km-long seismogenic fault11,12. Over a century, geo-scientists have been seeking for the information that indicates impending of large earthquakes1,2,3,4,5. Only months ago, a process of regional weakening was proposed as a new model for the generation of large earthquakes, during which temporally and spatially multi-scaled deformation was manifested and characterized with clusters or swarms of micro-earthquakes6. This model states that deformation should be the primary information associated with generation of large earthquakes, and micro-earthquakes in styles of clusters or swarms should be the primary manifestation. However, a question was remained: how to build a bridge to physically link the deformation with the manifestation?
For decades, the efforts in seismological and rock-laboratory studies7,8,9,10 have revealed that micro-earthquake seismicity were able to cause stress change, and deformation were able to produce rock anisotropy. Apparently, using anisotropic property to monitor deformation is a good choice, and the interrelation between deformation and stress change makes it possible to extract anisotropic information by observing micro-seismicity.
The anisotropy of the rocks that make up the Earth is simply divided into two types: the LPO (lattice-preferred orientation) formed by the alignment of intrinsically anisotropic mineral grains, and the SPO (shape-preferred orientation) generated by the ordered assembly of individually isotropic materials, and any of them may lead to a speed difference of up to and even exceeding 10% for the waves of different polarization or propagation directions13,14. However, the crust and especially upper crust is almost dominated by the SPO due to the deformation response to long-term stress action15,16,17,18,19.
Reviewing the works associated with seismic anisotropy, we noticed that in most of them shear waves, instead of compressional waves, were employed 20, 14, and that in a small number of them the inclined symmetrical axis were allowed21,22,23. However, compressional waves also contain anisotropy information, and the travel-time measurements of the direct compressional waves are usually more accurate and reliable 24,25. Moreover, symmetrical axis of anisotropy is generally neither vertical nor horizontal in reality23,22. Thus, it will be very significant to extract the information on seismic anisotropy with arbitrarily inclined axis from the direct compressional waves.
Here we present a real example in which we successfully extracted the information on the seismic anisotropy with inclined and vertical axis from a dataset of the travel-time residuals obtained by directly measuring the direct P-wave arrivals of a cluster of micro-earthquakes recorded by a dense seismic array, and in which we adopted an inversion method designed based on a freshly proposed concept, a slowness deviation tensor (SD tensor) describing seismic anisotropy with arbitrarily inclined symmetrical axis. Also, we exhibit a comparison in direction between the stress field inverted from the focal mechanism solutions of the cluster and the seismic anisotropy resolved by the travel-time residuals, to demonstrate that the seismic anisotropy of crustal rocks even under local stress action was detectable. All of these are strongly suggesting that it will be promising and expectable to monitor the weakening process preceding to large earthquakes by densely observing the micro-earthquake seismicity.