The 2009 L'Aquila earthquake in Italy, and its foreshock sequence, has been in the news lately for all the wrong reasons (Nosengo, 2010), but this month two papers in Geology use the sequence to suggest new breakthroughs in science: they provide strong evidence that high fluid pressure contributed to the rupture. Fluids and their movement have been increasingly implicated in earthquake generation, yet few studies have quantified the relationship for individual earthquake sequences. In this issue of Geology, Terakawa et al. (2010, p. 995) present a new method of mapping fluid pressure from focal mechanisms, and use the entire suite of earthquakes to show that fluids were highly overpressured near the foreshocks and subsequent mainshock. Lucente et al. (2010, p. 1015) show that earthquakes in the early part of the foreshock sequence yielded different measures of the ratio of compressional-to-shear velocity and of seismic anisotropy than did the earthquakes between a magnitude 4.0 foreshock and the mainshock a week later. Lucente et al. suggest that the different measurements are caused by fluid migrating between different domains within the foreshock region.
The ultimate cause of earthquakes is stress in the Earth exceeding the elastic limits of rock. Unbroken homogeneous and isotropic rock will fail on predictable planes (Anderson, 1951). Yet most regions have rocks that are weaker than others, often because of previous faults. This allows preexisting faults to break at lower stresses and at different angles than those required to break fresh rock (Sibson, 1985). The pressure from the weight of the rocks above faults helps to keep them from slipping. But the pressure of fluids in pores of the rocks counteracts the weight, so that high pore fluid pressures reduce the stress needed to break rocks (Hubbert and Rubey, 1959). This effect of fluid on fault material is similar to the effect of oil on ball bearings. Mineral deposits in exhumed fault zones, and triggering of earthquakes by small changes in fluid pressure, provide abundant evidence for fluid involvement in earthquakes (e.g., Sibson, 1992; Townend, 2006). Another variable that can affect fault orientations is that faults with low coefficients of friction, due for example to weak minerals along the fault plane, will preferentially break compared to faults with higher friction (Byerlee, 1978).
The first motions recorded on seismometers and waveform modeling of seismograms can both yield focal mechanisms, which determine two possible fault planes that may have slipped in an earthquake. In most regions, and particularly for small earthquakes, fault planes for nearby earthquakes vary, and such variation is used in stress inversion studies to determine the correct orientation and relative magnitudes of the regional stress tensor (e.g., Gephart and Forsyth, 1984). The misorientation of stress tensors determined in this way, and the faulting of large earthquakes, has in many cases shown that large faults such as the San Andreas fault in Southern California, and smaller faults such as those in Marlborough, New Zealand, are weak, possibly due to high pore fluid pressures (Balfour et al., 2005; Mount and Suppe, 1987). Studies such as these look at the average orientations of faults and determine friction values or pore fluid pressures consistent with the orientations. Other studies look at the aggregate of fault orientations and conclude that pore fluid pressure and friction are both heterogeneous (Rivera and Kanamori, 2002). The new twist provided by Terakawa et al. is to use the misorientation of individual faults to map pore pressure variation. They assume that variable friction on faults is negligible compared to the pore pressure variation. They can then determine the pore pressure needed to allow a fault of arbitrary orientation to fail. In that way, they map the minimum pore pressures that were necessary to allow each earthquake to occur. The highest pore pressures found in this technique were close to the foreshock and mainshock locations, suggesting that high pore fluid pressures allowed the mainshock rupture. However, this is somewhat surprising since extensional areas such as these are least likely to hold overpressures (Sibson, 2004). The assumption that friction is not variable on the faults could be wrong. One way to test the assumption would be to make the alternative hypothesis that the pore fluid is constant and that variable friction is the only difference, and to map out the locations of low friction. Presumably such an exercise would show that low friction occurs in the vicinity of the fault planes. The Terakawa et al. technique for tomographic images of fluid pressure or friction variation will be useful in other studies of earthquake generation, and also in volcanic areas where earthquakes are related to movement of fluids and magmas.
Variations in the ratio of the P wave speed to the S wave speed (Vp/Vs) (e.g., Nur and Simmons, 1969a; Sadovsky and Nersesov, 1974) and in seismic anisotropy (e.g., Gupta, 1973), among other seismic properties, were heralded in the 1970s and 1980s as methods to examine the buildup of stress that should occur before large earthquakes. The idea was that high pressures could cause rocks to “dilate,” increasing crack numbers or dimensions, which in turn decreased Vs velocity more than Vp, increasing the Vp/Vs ratio (e.g., Griggs et al., 1975). Oriented cracks can cause seismic anisotropy by making waves traveling or polarized parallel to the cracks travel at different speeds from those that must cross the cracks (e.g., Nur and Simmons, 1969b). Numerous studies reported changes between properties recorded prior to or after mainshock occurrences. It was hoped that these kinds of studies would allow earthquake prediction to be “just around the corner” (e.g., review by Geller, 1997). However, several careful studies showed that such variations could be explained simply by phase conversions or by lateral variations in properties such as Vp/Vs ratios or anisotropy of materials, together with the change in earthquake locations over time; aftershocks are usually located in different areas than the background seismicity before the mainshock (e.g., Lindh et al., 1978; McEvilly and Johnson, 1974). Thus, seismologists became discouraged and studies of time variations of seismic properties became unfashionable. However, recent increases in instrumentation and new techniques have allowed much better determination of locations, and have allowed some properties to be measured through methods with more repeatable sources, such as families of earthquakes with highly similar waveforms, and noise sources (e.g., Bokelmann and Harjes, 2000; Brenguier et al., 2008). Moreover, seismic exploration companies have used controlled sources to determine time variations in seismic properties due to fluid extraction or injection (e.g., Duffaut and Landro, 2007). Thus, time-varying properties can now be more carefully measured and these previously discarded techniques can be resurrected.
Lucente et al. performed careful relocations of earthquakes in the L'Aquila sequence using modern techniques, and their results show a strong change in Vp/Vs and in anisotropy immediately after the 30 March 2009 magnitude 4.0 foreshock. They suggest a scenario by which the mainshock fault plane initially acted as a barrier to fluid flow, so that one side of the fault had higher fluid pressures than the other. The 4.0 foreshock broke that barrier, allowing fluids to migrate across the fault and to change the Vp/Vs ratios by means of changing the fluid content of the pores and cracks. The movement of the fluids onto both sides of the fault presumably lubricated the fault enough to allow it to fail in the mainshock. Lucente et al. present a forward model that explains the Vp/Vs ratios and anisotropy measurements. There is still the fundamental problem that the earthquakes in the two sets of time periods—before and after the magnitude 4.0 foreshock—have, on average, different locations. However, there was a small group of earthquakes that had similar locations both before and after the magnitude 4.0 foreshock. To show that the change in properties was truly in time and not in space, they show that most of the earthquakes in that group also exhibited changes. Therefore, they have done almost all they can to prove that the change is a time, rather than a spatial, variation. The remarkable coincidence of the Lucente et al. forward model of areas with high fluid involvement, with the areas determined to have high pore fluid pressure in the Terakawa et al. study, help to enhance the credibility of both hypotheses (Fig. 1).
There are three relatively simple ways to test the hypotheses of time and spatial variation in pore fluid pressure. (1) The Terakawa et al. technique should be applied separately to the three different time periods examined in the Lucente et al. study to determine if the fluid pressures have indeed changed over time in the different regions. If such a time variation in pore fluid pressures has occurred, then not only will the Lucente et al. hypothesis be supported, but the hypothesis that the changes in fault plane orientation are caused by friction variations instead of pore fluid variations will have less weight. This is because it is much harder to imagine a process that would change friction over time, than one that would change the pore fluid pressure over time. (2) Determining quantitative relations between Vp/Vs ratios and fluid overpressures would help to test how well the two models fit with each other. (3) The catalogue used in the Lucente et al. study could be searched for “repeating events” whose P waves are similar to each other, and those events could be examined to be sure that the S arrivals vary in the manner that they are supposed to vary to give the reported changes in Vp/Vs ratios and anisotropy.
If these methods can be compared quantitatively to each other and can be shown to yield similar results in other earthquake sequences, they bear the potential to be used real-time in developing earthquake sequences, possibly leading to a method of forecasting whether a sequence will lead to a large event. This ability would help to prevent the kind of political turmoil that has arisen from the response of public officials and scientists over the L'Aquila earthquake sequence.
I thank Rick Sibson for an informal review of an early version of this article.