S U M M A R Y It is 10 years since shear-wave splitting, thought to be diagnostic of some form of seismic anisotropy, was first positively identified in the Earth's crust. From the beginning it was argued that the splitting was probably associated with the presence of stress-aligned cracks (inclusions) in the crust, and that this would provide the opportunity for monitoring the in situ geometry of cracks and stress in a variety of different circumstances and in a variety of different applications. The early promise was not immediately realized, and the first 10 years were spent mainly in observing the phenomena in a variety of different situations. However, 1990 appeared to mark a turning point for anisotropy. Papers at the Fourth International Workshop on Seismic Anisotropy and elsewhere have announced major progress in understanding, interpreting, and particularly processing shear-wave splitting, with direct applications to hydrocarbon production, and a possible (but disputed) application to monitoring stress changes before earthquakes. However, there is still much that we do not understand about the phenomenon, and we are clearly only just beginning to appreciate the enormous information content of the shear wavetrain and its potential applications to science and engineering in the Earth's crust.This paper briefly reviews the past 10 years, and speculates on how best we can exploit this new window of opportunity for exploring the internal structure of the crust. In particular, what causes the shear-wave splitting? what use can we make of the phenomenon? and what should we do next?

Shear wave splitting has been on the free surface. Outside the shear wave observed in seismograms of two aftershock and one window, the incident polarizations are distorted foreshock sequence recorded over 12 days by a local network of nine three-component digital seismometers during the 1982 Enola, Arkansas, earthquake swarm. Polarizations of the faster split shear wave show alignments within the shear wave window which correlate with the regional stress field. The time delay between split shear waves appears to increase before, and to decrease at or after the time of each main shock.This behavior is similar to observations before and after a M= 6.0 event near the Anza seismic gap, southern California, and may correspond to variations in crack geometry caused by changes in the stress field before and after the earthquakes.There are indications that the decrease in delay may occur a few hours before the main shock and could be a possible precursor for earthquake prediction. Booth, 1985].The time delay, which is due to the velocity difference between the split shear waves, depends on the direction of propagation and the path length through the cracked rock. It is also sensitive to the crack density, crack identified on almost all suitable ray paths, in a dimensions, aspect ratio, pore fluid content and wide variety of tectonic regimes (see Crampin [1987a] for a review, and elsewhere in this issue for more recent examples).By "suitable ray paths", we mean those ray paths which are incident at the Earth's surface within the shear wave window, which is the locus of propagation paths which subtend an incidence angle less than the critical angle, sin-•(Vs/Vp), at a receiver other parameters of the cracked rock.If the observed shear wave splitting is caused by stress-aligned EDA cracks, then changes in the stress field before an impending earthquake will alter the most compliant feature of the rock mass -the EDA crack distribution -and thus modify the characteristic behavior of the split shear waves and, in particular, their polarizations and the time delay between them. Crampin [1987b] has suggested that most earthquake precursors, and the often irregular behavior of precursors, can be explained by indirect effects of temporal variations in the characteristics of EDA cracks, and their anisotropy.

The third occupation (experiment TDP3) of recording sites above a persistent swarm of microearthquakes near the North Anatolian Fault, with a larger seismic network and over a longer period of time, confirms and refines previous observations with greater resolution. The greater resolution in earthquake locations has revealed marked clustering in time and space. Many, perhaps most, of the earthquakes belong to clusters, where successive earthquakes originate in a very small volume and have similar fault mechanisms. Such studies allow the progression of seismic activity of small earthquakes to be followed in some detail, and may reveal features which are hidden in larger and more complex earthquake sequences.

No abstract

Observations of seismic shear-wave splitting (birefringence) along almost all raypaths in the Earth's crust have led to the recognition that the fluid-filled cracks, microcracks and pores known to pervade most rocks in the crust have been preferentially aligned by the prevailing stress-field. This phenomenon is known as extensive-dilatancy anisotropy or EDA, and the fluid-filled microstructures as EDA-cracks because the seismic effects can be modelled by distributions of parallel cracks. The effects of these stress-aligned EDA-cracks have been observed in many different areas, and they appear ubiquitous in the upper 10 to 20 km of the crust. Many details of the crack geometry and the current stress-field can be readily determined from records of the shear-waves which have propagated through these aligned cracks. In separate studies, geochemists have confirmed the presence of microscopic fluid-filled cavities in almost all geological material. These are referred to as fluid inclusions, and provide unequivocal evidence of crustal fluid circulation, often along stress-controlled fractures. Different stress episodes can be recognized in many rock samples and studies of fluid inclusions can be used to interpret the palaeo-stress history of a rock, where the in situ morphology of these inclusions is related to the current regional stress-field. This paper discusses the relationship between the fluid inclusions of the geochemical literature and EDA-cracks, and concludes that the former may be regarded as one of the constituent inclusion types of extensive-dilatancy anisotropy.