A review of the current understanding of seismic shear-wave splitting in the Earth’s crust and common fallacies in interpretation

Abstract

Azimuthally-aligned shear-wave splitting is widely observed in the Earth’s crust. The splitting is diagnostic of some form of seismic anisotropy, although the cause of this anisotropy has been sometimes disputed. The evidence in this review unquestionably indicates cracks, specifically stress-aligned fluid-saturated microcracks, as the predominant cause of the azimuthally-aligned shear-wave splitting in the crust. Although, in principle, shear-wave splitting is simple in concept and easy to interpret in terms of systems of anisotropic symmetry, in practice there are subtle differences from isotropic propagation that make it easy to make errors in interpretation. Unless authors are aware of these differences, misinterpretations are likely which has led to incorrect conclusions and charges of controversy where only misinterpretations exist. As a consequence, stress-aligned fluid-saturated microcracks as the cause of azimuthally-aligned shear-wave splitting in the crust is still not universally accepted despite there being distinguishing features that directly indicate crack-induced anisotropy. This paper reviews observations and interpretations of crack-induced shear-wave splitting and demonstrates that claims for aligned crystals and other sources of shear-wave splitting are due to fallacies in interpretation. This review shows how previous contrary interpretations are resolved and discusses common fallacies and misinterpretations. It is suggested that this new interpretation of shear-wave splitting has such fundamental implications for almost all solid-earth geoscience that it amounts to a New Geophysics with applications to particularly exploration and earthquake geoscience but also to almost to all other branches of solid Earth geoscience.

Introduction

Shear-waves propagating in anisotropic rocks split into two approximately orthogonal polarisations that travel at different velocities and write characteristic easily-identifiable signatures into three-component seismic wave trains [1]. Such shear-wave splitting (seismic birefringence) aligned azimuthally is widely observed in almost all igneous, metamorphic, and sedimentary rocks in the Earth’s crust in almost all geological and tectonic regimes [2], [3], [4], [5], [6]. There are only a few well-understood exceptions where azimuthally-aligned shear-wave splitting has not been observed in the crust [7]. Such azimuthally-aligned shear-wave splitting was first identified in 1981, in both the crust [8] and the upper-mantle [9]; coincidently and independently, both reported in the same volume of Nature. However, despite 25 years of observations, the causes and interpretation of shear-wave splitting in the Earth’s crust are still often misunderstood. This review attempts to resolve some of these misunderstandings.

Note that the phrase ‘azimuthally-aligned shear-wave splitting’ refers to splitting where the faster split shear-waves are approximately parallel as illustrated schematically in Fig. 1. This distinguishes it from the splitting, observed particularly in exploration seismology in finely-layered sedimentary strata, where the polarisations are controlled by the direction of propagation and are strictly SV and SH.

Anisotropic symmetries are discussed in Section 2.1, below. The evidence suggests azimuthally-aligned shear-wave splitting observed in the Earth’s crust is invariably caused by stress-aligned parallel vertical microcracks. This mechanism has the enormous benefit that the crack geometry has comparatively simple anisotropic symmetry that can be specified by three parameters: (1) orientation of the parallel vertical microcracks, imaged by the strike of the nearly-parallel shear-wave polarisations; (2) crack density imaged, by one hundredth of the percentage of shear-wave velocity anisotropy; and (3) changes in crack aspect-ratios, imaged (less easily) by changes in average time-delays between split shear-waves in a particular range of angles of incidence to the free-surface [7]. Changes in crack aspect-ratios are important as we have shown in a recent more formal review of theory and interpretation in Wave Motion [10] that their behaviour in the Earth’s crust demonstrates that the microcracks are so closely-spaced they are critical-systems monitoring the low-level pre-fracturing deformation of in situ rock. Consequently, shear-wave splitting is caused by stress-aligned near-vertical microcracks [2], [3], [4], [5], [6], and is the key diagnostic of the New Geophysics: a new understanding of fluid-rock deformation which has profound implications for many properties of in situ rocks [7], [10], [11]. This review provides the consistent interpretation of shear-wave splitting necessary for understanding and promoting the New Geophysics.

In 1981, one of us published in Wave Motion [1] a review of theoretical and numerical aspects of wave propagation in cracked and anisotropic media, particularly with reference to observations of anisotropy in the Earth’s crust. The most diagnostic feature of azimuthally-aligned anisotropic wave propagation is shear-wave splitting (seismic birefringence), where shear-waves split into differently polarised phases which are azimuthally-aligned, travel at different velocities, and write easily-recognised symbols into polarisation diagrams (PDs), or hodograms, of three-dimensional particle motion. Shear-wave polarisations and time-delays between split shear-waves, the two distinctive parameters in Fig. 1, can easily be identified and measured in PDs.

Azimuthally-aligned shear-wave splitting was positively recognised in in situ rock in observations above small earthquakes in Turkey [8], [12], and later in seismic reflection surveys, and vertical seismic profiles in oil exploration surveys [13], [14] and other reports reviewed by Helbig and Thomsen [15]. Such shear-wave splitting with azimuthally-aligned polarisations is widely observed above small earthquakes (see for example [5], [6], and many papers cited throughout this review). Fig. 2 shows an example of shear-wave splitting above a small earthquake in SW Iceland. The horizontal seismograms are rotated into faster and slower shear-wave polarisation directions showing different arrival times. PDs of the horizontal motion of the two split shear-waves show linear motion with a difference in arrival times of ∼0.1 s. The enlarged PD of the horizontal motion shows the linear motion of the faster split shear-wave where the deviations of linearity between the circled shear-wave arrivals are less than the amplitude of the P-wave coda. It is interesting that, although shear-wave splitting is a second-order phenomenon of small differences in shear-wave velocities, if split shear-waves are rotated into faster and slower polarisations, as in Fig. 2a, the arrival times can often be read with first-order accuracy.

Azimuthally-aligned shear-wave splitting is also observed in exploration seismology in a huge variety of controlled source experiments in reflection profiles, vertical seismic profiles, and well logs as reviewed in Wave Motion [10].

Critical-systems are a New Physics [17]; a New Geophysics. Complex heterogeneous interactive systems initially interact locally, but when they approach singularities, bifurcations, or, in the case of the Earth, fracture-criticality [2], they abruptly display coherent behaviour involving collective organisation of large numbers of degrees of freedom: “It is one of the miracles of nature that huge assemblages of particles subject to the blind forces of nature, are nevertheless capable of organising themselves into patterns of cooperative activity” [18]. Critical-systems and self-organisation are extremely common, including: quantum mechanics; superfluidity; traffic clustering on roads the life cycle of fruit flies; stocks in the New York stock exchange; and a huge number of physical relationships including the Gutenberg–Richter relationship between the logarithm of the cumulative number of earthquakes and the earthquake magnitude, which is linear (self-similar) over at least eight orders of magnitude. Thus, in suggesting that the Earth’s crust is a critical-system of stress-aligned fluid-saturated microcracks [11], we are merely suggesting that the Earth behaves like all other complex heterogeneous interactive phenomena.

Critical-systems have a range of behaviour fundamentally different from conventional sub-critical physics (and geophysics), including calculability (hence predictability), extreme (“butterfly wing’s”) sensitivity to initial conditions leading to deterministic chaos [19]. This implies that fluid-rock deformation in the crack-critical crust of the Earth can be: monitoredby analysing shear-wave splitting; modelled by the anisotropic poro-elasticity (APE) model of fluid-rock deformation (see Section 3); future behaviour predicted (if the changing conditions can be quantified); and in appropriate circumstances, future behaviour controlled by feedback [11].

These properties are fundamentally different from the sub-critical behaviour of the classic conventional brittle-elastic upper crust. Thus, the understanding of the behaviour of deformation in the Earth’s crust has advanced substantially from a conventional brittle-elastic crust to a dynamic compliant self-organised crust. We now know that very small changes in rock mass conditions readily modify the geometry of fluid-saturated microcracks and that these can be monitored by shear-wave splitting. These properties are so different from conventional sub-critical behaviour, that many geoscientists are (understandably) reluctant to accept the idea of a highly compliant crack-critical crust of fluid-saturated stress-aligned microcracks, (see for example, a recent book on rock mechanics [20]). A major difficulty for many geoscientists is that the diagnostic effects are almost entirely confined to shear-wave splitting and shear-waves, and the behaviour of shear-wave splitting is only now becoming understood.

This review of azimuthally-aligned shear-wave splitting (seismic birefringence) shows three things:

• Azimuthally-aligned stress-aligned shear-wave splitting is almost invariably caused by propagation through distributions of stress-aligned fluid-saturated microcracks, which are highly compliant to small changes of stress.

• Stress-induced changes to microcrack geometry can be monitored by variations in shear-wave splitting so that, in particular, at the approach of fracture-criticality, the times and magnitudes of impending larger earthquakes can be estimated by analysing shear-wave splitting.

• Characteristic temporal variations of shear-wave time-delays (typically, increases in normalised time-delays, monitoring stress-accumulation, where the logarithm of the duration of the increase is proportional to the magnitude of the impending earthquake) are now seen before earthquakes worldwide (Table 1a).

We refer to such linear log–log relationships as self-similar. When there is sufficient seismicity before the impending earthquake to provide shear-wave source events within the shear-wave window (9 cases out of 15 in Table 1b), the increasing time-delays show an abrupt precursory decrease (interpreted as stress relaxation due to cracks coalescing onto the eventual fault break) immediately before the larger earthquake [6]. The durations of these decreases of time-delays are also self-similar with respect to the magnitudes of the impending earthquakes [6], [30]. There has been one successful real-time stress-forecast of time, magnitude and fault break of a M 5 earthquake in SW Iceland, when the increase in time-delays was recognised before the earthquake had occurred [25]. There are no contrary observations where adequate data sets have not shown characteristic changes before large earthquakes.

We show in the next section (summarised in Table 2) that azimuthally-aligned shear-wave splitting in the crust is almost invariably caused by propagation through stress-aligned fluid-saturated microcracks as illustrated schematically in Fig. 1. Fluid-saturated microcracks are the most compliant elements of the rock mass and crack geometry will readily respond to changes in stress [1], [7], [41], [42], [43]. Consequently, variations in shear-wave splitting are the most sensitive diagnostic indicators of variations in in situ microcrack geometry, and such changes have been observed whenever there have been sufficient appropriate shear-wave ray paths before larger earthquakes [5], [21], [22], [23], [24], [25], [26], [27], [28], [29], [30], [31]. Changes in shear-wave splitting have also been observed before volcanic eruptions [5], [32], [33], [47], and in hydrocarbon reservoirs following fluid-injection [34], [35] and possible variations with ocean tides [37].

However, claims that the rock mass is sensitive to small changes of stress is contrary to the concept of the conventional brittle-elastic upper crust and some papers suggest aligned crystal mineralisation as the source of the anisotropy. Some papers suggesting mineral alignments as the cause of shear-wave splitting [48], [49], and [50], have been answered by Crampin [51], [52], and [31], respectively. These various exchanges are summarised in Table 3. Papers [57] and [58] also questioned observations and interpretations of shear-wave splitting, and were answered by Crampin et al. [29]. (Note that [59] suggests that all wholly-automatic measurements of shear-wave splitting are likely to be inadequate.) Paper [60] questioned the statistics of the stress-forecast earthquake [25] and was answered by Crampin et al. [61].

Unfortunately, detailed refutations, however well-founded, seldom receive as much attention as the original criticisms, and misunderstandings and misinterpretations still persist [20], [50]. This review discusses a further range of papers reporting shear-wave splitting, and outlines the current understanding. The review demonstrates that the only viable interpretation of azimuthally-aligned shear-wave splitting in the crust is propagation through compliant stress-aligned fluid-saturated microcracks that allow temporal changes in stress to be monitored by changes in shear-wave splitting. This review also establishes that, whenever there are adequate source data, characteristic patterns of behaviour of shear-wave time-delays are observed before larger earthquakes. Again there are no contrary observations.