May/June 2002

Mechanical Pollution

Some things are "in" and others are "out." Fashion plays a role in most of our choices, from clothes to topics in seismology. A certain amount of collective irrationality is good; it holds us together and simplifies our decisions. But culture must keep up with the rapidly changing environment and we need to poke at it relentlessly with radical thought. The emphasis is on "radical" because thought can also be wasted on rationalizations. Even in science, rationalizations built on wrong assumptions can be a cover for outdated cultural values or, worse, for narrow private agendas. A glaring example of a research topic that seems to be unjustifiably "out" is anthropogenic seismicity, earthquakes triggered by human activities.

The concept that humans can cause earthquakes is more than a century old, yet it is still astonishing. Reservoir-induced earthquakes were not necessarily the first anthropogenic earthquakes, but they were widely recognized and studied in the mid-20th century because they could be explained by straightforward and physically plausible models. Perturbations of subsurface poromechanical conditions from the load of water in large reservoirs are clearly substantial, even at seismogenic depths, and were calculated to be of sufficient amplitude to cause failure on faults already loaded by tectonic stress. The important role of pore pressure and interstitial fluid flow became obvious when earthquakes were triggered by deep fluid injection near Denver, Colorado in the early 1960's. The propagation of seismicity as far as 5 km away from the well was interpreted as the effect of an expanding fluid pressure front. Injection was halted to reduce earthquake hazard, yet seismicity continued, producing the largest earthquake a year later.

Once recognized, anthropogenic earthquakes have been associated with many engineering activities, and the minimum perturbation considered significant has become progressively smaller. But the interest in anthropogenic seismicity was short-lived, probably because the seismological community concluded, erroneously, that the hazard from anthropogenic earthquakes was negligible and, correctly, that recognition of this hazard would engender strong resistance.

In spite of the current interest in earthquake-earthquake triggering, anthropogenic earthquakes remain largely off our profession's radar screen. While old but persistent misconceptions reveal why we have not been interested in anthropogenic earthquakes, new results make a strong case that we should, particularly in stable continental regions (SCR). SCR are characterized by relatively low seismicity, but they cover much of the Earth's land surface and have experienced a number of nasty earthquake disasters in the last decade.

Hazard is the first concern. Anthropogenic earthquakes are common, but fortunately potential sources of anthropogenic earthquakes are not only relatively easy to identify but can also be controlled. Another reason to study anthropogenic earthquakes is that they may offer useful engineering applications. In spite of their bad reputation, earthquakes can be helpful in illuminating faults and folds and in monitoring stress change and fluid flow. Abundant, low-magnitude anthropogenic seismicity is being studied in hydrocarbon and hydrothermal fields as a way to monitor physical properties. Finally, anthropogenic seismicity offers unique laboratory conditions to study seismogenesis in general. We have only begun to take advantage of research opportunities offered by the ability to monitor stress, pore pressure, and fluid flow as earthquakes are turned on and off.

One of the main reasons that anthropogenic earthquakes are out of fashion is that they are relatively unimportant in active areas where natural seismicity is high and where most of seismology happens. In SCR, seismogenesis and our understanding of it are relatively low and thus we tend to import notions from California and assume that any distinctions are only quantitative. I suggest here some of the areas where this assumption is incorrect and why it is particularly misleading in terms of the relative levels of anthropogenic and natural seismicity.

The probability that an anthropogenic perturbation triggers an earthquake depends on ambient stress-how close a fault was to failure-and not on strain rate, provided the perturbation is short-lived relative to the earthquake loading cycle. In situ stress measurements have shown that the upper crust is in a critical stress state in both active and stable continental regions. Assuming pure elastic loading and a similar range of stress drop, faults in either environment would be equally close to failure, on average. Therefore, anthropogenic earthquakes would be equally likely in high and low strain-rate environments, if other factors were the same (they are not; see below). In contrast, natural seismicity is proportional to tectonic strain rate and is vastly higher along plate boundaries. On this basis alone, the ratio between anthropogenic and natural seismicity is expected to be much higher in SCR than at plate boundaries.

Recent data support this first-order prediction. In peninsular India, for example, a rapid expansion of the irrigation system in the 1960's coincided with a factor of 3 seismicity increase, including several well known reservoir-induced events, such as the 1967 Koyna earthquake. In South Africa, official bulletins of recent seismicity differentiate more numerous and easily identifiable mining-related earthquakes from natural ones. In the area of the northeastern U.S. covered by the Lamont-Doherty regional seismic network, the ratio between anthropogenic and natural earthquakes is about 1 to 3 (more below). Compare these and other examples in SCR with the San Andreas plate boundary, where, despite intense seismicity and seismology, only a few and mostly controversial cases of anthropogenic earthquakes have been reported. This juxtaposition is particularly striking in light of the current habit of considering an earthquake to be natural until proven anthropogenic. This practice biases the anthropogenic/natural ratio in favor of intensely studied areas.

"Anthropogenic earthquakes tend to be shallow and small and thus of little concern." This concept derives also from California and other active areas where most important earthquakes nucleate near the bottom of the seismogenic layer and out of range from anthropogenic perturbations. The upper 3-5 km of the crust in active areas may be too fractured to sustain large coherent stresses, and the relatively scarce seismicity in this depth range seems to be magnitude-limited below damage threshold. In contrast, many of the damaging or potentially damaging SCR earthquakes are very shallow. Their ruptures are mostly confined to the upper 5 km and often reach the surface. Well documented examples in the M 6-7 range include 1990 Ungava, northern Québec; 1993 Killari, central India; and 1968 Meckering, 1970 Calingiri, 1979 Cadoux, 1986 Marriat Creek, and 1988 Tennent Creek, all in Australia. Although the largest SCR events, such as the 1811-1812 New Madrid events and the recent one in Gujarat, India, seem to nucleate deep in the crust, the class of very shallow SCR earthquakes clearly reaches magnitudes well above the damage threshold. Furthermore, their proximity to the surface lowers the damage threshold, and relatively abundant small events can have severe consequences. The Mb 5.3, 1995 earthquake centered below New Castle, Australia is an example.

Earthquake depth distribution suggests that, unlike the crust in a tectonically active zone, SCR crust can store plenty of elastic energy, well within reach of anthropogenic perturbations. In SCR, therefore, the magnitude ranges of anthropogenic and natural shallow earthquakes are expected to be similar. Furthermore, if upper magnitude limits for very shallow earthquakes in SCR and at plate boundaries are below and above damage thresholds, respectively, the likelihood of damaging anthropogenic earthquakes is greater in SCR than in California in absolute terms, not just in comparison with natural seismicity. Finally, depth and magnitude limits should be applied to anthropogenic earthquakes with great caution, considering, for example, the midcrustal hypocenters below Lake Nasser in Egypt and the three M 7 events in the proximity of the Gazli gas field in central Asia.

About one third of the sequences I have co-investigated with temporary local networks in the northeastern U.S. since 1980 were demonstrably triggered by a variety of engineering activities, including deep fluid injection, large quarries, and deep mines. Five of the main shocks were in the magnitude range 4.3-5.2 and caused MMI VI-VII damage, mostly in rural areas. The largest one was 8 km deep and was natural. The others were no more than 5 km deep; two were anthropogenic, one was natural, and one is still uncertain. In the relatively small area of the northeastern U.S. covered by our observations, anthropogenic seismicity is up there with natural seismicity in terms of number of events and potential for damage. Many other areas are not as closely monitored, and a reliable comprehensive evaluation of current anthropogenic seismogenesis may be difficult. Yet even a cursory survey should merit attention. The last two M > 4 U.S. events east of the Mississippi, for example, were in Ashtabula, January 2001, and in Alabama, January 1999; both were very shallow and likely triggered by fluid injection (see below for more on Ashtabula). In summary, earthquakes in the upper few kilometers of the crust account for a large share of the hazard in SCR. This depth range is within reach of many anthropogenic perturbations which have been triggering a substantial portion of shallow SCR earthquakes. The hazard implication of any shallow SCR earthquake may not depend a priori on whether it is natural or anthropogenic. But we know where and how we are altering underground conditions and the resulting earthquake hazard is relatively circumscribed and intense. It is thus potentially better known and easier to deal with from the engineering and economic standpoints. Finally, anthropogenic earthquakes tend to occur near population centers and thus impose particularly high risk.

"Triggering only shifts the hazard in time." When triggered, an earthquake occurs earlier than it would have naturally because a perturbation raised the stress to failure prematurely. An area undergoing industrialization may experience a burst of anthropogenic earthquakes and thus an increase in seismicity and hazard. The long-term budget of moment release, however, must match the average tectonic rate, and this increase is therefore compensated later by a seismicity decrease. The time between increase and decrease is proportional to tectonic rate; here the difference between California and SCR lies in the meaning of "long term." Compensation time along the San Andreas fault may be on the order of a lifetime: Pay now, cash in later. In SCR, however, one may have to wait ten thousand years or more before full compensation: a bad deal. While anthropogenic may trade off with triggered seismicity in California, these two types of seismicity are largely independent and additive in SCR. Our California-based seismological culture has once again led us to underestimate the significance of anthropogenic earthquakes for hazard in SCR.

Finally, let's face it: Fear of liability and bad publicity by those who trigger earthquakes is hindering research in anthropogenic earthquakes. Ashtabula, Ohio offers a sad example. In 1987 we monitored a Mb 3.5 aftershock sequence, and we showed the source to be a vertical fault below that town. This fault was also 0.7 km from a deep high-pressure waste-disposal well that had begun operation a year earlier. In response, the well operator invested about $0.7 million for a reflection survey that found no fault and thus, allegedly, showed no risk from earthquakes. This result published in the local press satisfied the population and apparently also the Ohio EPA. We were left to pull out our stations and to end our self-financed experiment. But seismicity in Ashtabula continued, even after the well shut down in 1994 and the operating company disbanded. Another fault 4 km from the well produced a Mb 4.3 main shock January 2001 that caused some damage. After a felt Mb 3.0 in June, instrumentally recorded events continued through the rest of 2001. In remarkable analogy to the seismological consequences of the 1962-1966 waste-fluid injection near Denver, Colorado, the injection in Ashtabula apparently left a perturbation that is still spreading from the well and is triggering earthquakes half a decade later. Such a delayed reaction suggests the notion of mechanical pollution.

We lost the opportunity to monitor the complete time-space development of seismicity in Ashtabula, but the data illuminate two seismogenic faults with minimum dimensions in the 1-to-several kilometer range below the city. Clearly, this situation deserves scientific and regulatory attention. Yet the Ohio EPA is still apparently objecting to the notion that earthquakes were triggered by fluid injection in Ashtabula. Such a notion may be a problem for routine flooding by several gas-recovery operations in Ashtabula County. Ashtabula-like cases, where research in anthropogenic earthquakes is being discouraged, abound worldwide and point to a serious problem: Humans pollute the upper crust and trigger earthquakes but often refuse to admit it, thereby losing an opportunity to learn about them and reduce hazard.

Effective pollution control is unlikely until the public demands it. History suggests that heads will remain buried in sand until a pollution disaster educates and galvanizes public opinion. These reactions, however, will likely be not only late, but also adversarial and unlikely to promote cost-effective solutions. The U.S. will eventually experience an earthquake disaster where human triggering is undeniable. Among the many polluters, one will get caught, chastised, and financially drained-a lottery in reverse. Waiting for such a disaster might make sense for litigators but not for society and, above all, not for seismologists. We can read the writing on the wall; after a destructive anthropogenic earthquake, it may be awkward explaining why earthquake hazard maps of the eastern U.S. account for natural earthquakes with great sophistication but ignore anthropogenic ones. We need to stand up to the polluters, realistically assess the hazard from anthropogenic seismicity, and propose ways to control mechanical pollution. While pursuing these tasks, we should also acknowledge that these polluting endeavors are generally not frivolous. Their products are critical and we all share in the benefit and thus in the responsibility. We need to promote a regulatory environment that emphasizes risk-sharing rather than passing the buck. Many and diverse engineering operations contribute mechanical pollution, but only a subset actually trigger earthquakes and only a few of these events are damaging. The regulatory environment will probably continue to protect polluters, but only as long as their pollution can be hidden from the public. This system is not only unfair, but it obstructs the use of science to seek and promote cost-effective ways to minimize hazard. We need to be more forceful in the study of anthropogenic earthquakes and in alerting society to this hazard.

Nano Seeber

To send a letter to the editor regarding this opinion or to write your own opinion, contact the SRL editor by e-mail.

Posted: 15 November 2002