- © 2014 by the Seismological Society of America
Online Material: Complete descriptions of the earthquakes and luminosities discussed in this paper.
With the beginning of seismology as a science in the 19th century, many scholars devoted time to reporting luminosities associated with earthquake activity. To name a few, the Irish engineer Robert Mallet, the “founder of seismology”, published a five part catalog entitled “On the Facts of Earthquake Phenomena” (Mallet, 1851, 1852, 1853, 1854, 1855), in which numerous reports on earthquake luminosities can be found. His catalog, first presented to the British Association of Science, covers the years 1606 B.C. to 1842 A.D.. Ignazio Galli, an Italian priest who graduated in Natural Sciences, published in the early 1900s a catalog of 148 seismic events associated with different types of luminosities. His catalog covers the years 89 B.C. to 1910 A.D. and focuses mainly, but not exclusively, on European events (Galli, 1910). Other early researchers on the subject of earthquake lights (EQL) include the work of Taramelli and Mercalli (1888), De Ballore (1913), Terada (1931), Musya (1932), and Montandon (1948).
More recently, numerous studies have been published dealing with descriptions of EQL, some of them offering a possible explanation with regards to the light‐producing mechanisms (e.g., Yasui, 1973; Tributsch, 1978, 1982; Devereux et al., 1983; Gold and Soter, 1984; Hedervari and Noszticzius, 1985; Derr, 1986; Freund, 2003a; St‐Laurent et al., 2006; Derr et al., 2011).
From the detailed study of 39 large earthquakes (magnitude of 5 or greater) that occurred in 20 different intraplate regions spread over 6 continents, Gangopadhyay and Talwani (2003) have shown that many intraplate earthquakes are associated with intracratonic rift environments. More specifically, intraplate seismicity occurs frequently in close association with extensional rift basins, grabens, and/or aulacogens (i.e., failed rifts), which in turn are often associated with the presence of mafic intrusions or dikes. To carry this idea further, the main topic of the paper presented here is (i) to determine the tectonic environments of the best‐documented earthquakes that were preceded, accompanied, or followed by luminosities, and (ii) to evaluate the possibility that EQL may be associated predominantly with intraplate earthquakes located within or nearby rift‐related structures.
Hence, the present study is based on a detailed investigation of 65 earthquakes in various geological settings for which sufficient information about associated luminous phenomena is available (i.e., 27 cases from the Americas and 38 cases from Europe). It pertains to the gathering and interpretation of geological, structural, and seismological data as well as EQL reports and other earthquake precursor phenomena.
A model is proposed for the origin of EQL associated with both intraplate and interplate (i.e., subduction zone) earthquakes, which is based on the generation of electronic charge carriers under high‐stress conditions (Freund et al., 1994, 2009; Freund, 2002, 2003a,b, 2010; St‐Laurent et al., 2006), their migration within the crust, and their electrical effect (or luminous expression) at the ground/air interface near specific types of faults.
Gathering and selecting data from the literature is a critical step in producing a meticulous and unbiased compilation of luminosities observed in association with earthquakes. We discarded from this study all cases where the reported luminosities were described as flames accompanied by smoke issuing from ground fissures, or were suggestive of moon or sun halos, of meteor‐like bolides passing over the sky at the time of the earthquake, or of luminous fog or cloud, as long as no other type of light emission was also included in the report. Events during stormy or unsettled weather with the possibility of lightning were also discounted, as well as cases where the earthquakes coincide with periods of intense aurora activity or where the description of the light phenomena might fit aurora qualities and effects. In summary, we elected to discard some possible EQL reports rather than keeping uncertain cases in our data compilation.
To cover a similar time span for EQL described from the Americas as from Europe, it was arbitrarily decided to select reports dating from about 1600 to present, considering that no documented historic earthquake was reported before this time in North America (Ebel, 1996; Gouin, 2001). The only listed earthquake that is significantly older is the Aquilano (Italy) earthquake of 1461. Furthermore, for the same area, no more than two observed earthquake/EQL events were described. This paper presents a list of 65 American and European earthquakes associated with reliable EQL reports (Tables 1 and 2; Figs. 1 and 2), along with detailed descriptions of 5 of these events. Descriptions of the other 60 events are provided in the electronic supplement to this paper. Note that all times cited in the text are given in local time (LT).
DESCRIPTIONS OF EARTHQUAKES ASSOCIATED WITH LUMINOSITIES
Saguenay Earthquake—1988 November 25, 18h46
From November 23, 1988, until the end of January, 1989, the Saguenay region experienced a total of 67 earthquakes (M>0). A relatively strong 4.8 foreshock occurred on November 23 at 04h12. Two days later, at 18h46 LT, more than two hours after sunset, the 6.5 (5.9 mb) mainshock caused strong shaking near the epicenter and was felt over much of northeastern North America. The hypocenter of this event was at a depth of 29 km in an area with no known previous seismic activity. The epicenter was located about 15 km (Du Berger et al., 1991) to the southwest of the south wall of the Saguenay Graben, a west‐southwest–east‐southeast oriented 50 km wide by 200 km long aborted rift structure that is oriented perpendicular to the genetically related St. Lawrence Rift system further to the east (Figs. 1 and 3; North et al., 1989; Du Berger et al., 1991; Roy et al., 1993; Thériault et al., 2005). Both of these rifts were formed during the break‐up of the supercontinent Rodinia and the ensuing opening of the Iapetus Ocean during the Late Proterozoic to Early Cambrian (Thomas, 2006). The earthquake occurred within the northern end of the Jacques Cartier Tectonic Block, a horst structure that is bounded to the north by the Saguenay Graben, to the west by the Saint‐Maurice Lineament and to the southeast by the St. Lawrence Rift system (Du Berger et al., 1991; Roy et al., 1993). Based on radar images, the south wall of the graben is poorly defined from the area of the epicenter toward the east up to the St. Lawrence Rift, as it becomes segmented into several subparallel lineaments (Fig. 3; Roy et al., 1993). Figure 3 shows the general area of the Saguenay Graben, along with the location of the seismic events of November 1988 to January 1989 and the observed EQL. Most of the luminous phenomena were seen either along the north and south margin of the Saguenay Graben, or near southwest–northeast oriented faults that transect the Saguenay Graben in the area of Chicoutimi, Jonquière, and Laterrière. Some of these faults are injected by lamprophyre dykes (Gittins et al., 1975; Perron, 1990), suggesting subvertical, deeply penetrating structures.
A total of 46 luminous phenomena reports, including 8 seen before the seismic sequence, were compiled by Ouellet in January 1989 (Ouellet, 1990). Most were sighted within 125 km of the epicenter, between the north and south bounding faults of the Saguenay Graben (Fig. 3). Three other cases were reported from around the city of Québec, located 150 km to the south of the epicenter and along the St. Lawrence Rift, while two EQL sightings came from the Beauce region, about 225 km to the south of the epicenter. The November 23 foreshock was associated with an atmospheric illumination, with most witnesses reporting a loud explosion‐like sound coinciding with the illumination just before the start of the shaking (St‐Laurent, 2000). Much further to the southwest, in northern Pennsylvania near the village of Goshen, located over 950 km from the epicenter; five people witnessed a brilliant light to the east at approximately 18h50 on November 25, about at the same time as the seismic waves of the mainshock arrived at this location. It is noteworthy to mention that Goshen is located along the southeast margin of the Rome Trough, a major Cambrian graben structure that extends for over 800 km toward the St. Lawrence Rift (Fig. 1). Other types of phenomena were also reported in association with the Saguenay earthquake. For example, at Chicoutimi, located about 30 km northeast of the epicenter, distinct radio interferences were noticed on an ordinary household radio a few days and hours before the mainshock (St‐Laurent, 2000).
New Madrid Earthquake, 1811 December 16, 02h15
A series of powerful earthquakes struck the mid‐Mississippi River Valley over a 3‐month period between December 1811 and February 1812. The general epicentral area was located near the city of New Madrid. The sequence included at least 18 earthquakes having a moment magnitude ranging between 5.8 and 8.0 (Johnston and Schweig, 1996). These earthquakes occurred within the broad New Madrid seismic zone, which extends for approximately 250 km along a southwest–northeast direction (Fig. 1). This zone, the seismically most active area of the United States east of the Rocky Mountains, is located along the northern part of the Mississippi River Valley graben, which is part of the Reelfoot Rift. This Late Proterozoic to early Cambrian failed intracontinental rift dates back to the breakup of the supercontinent Rodinia (Thomas, 1991; Csontos et al., 2008). About 100 km to the northeast of New Madrid, the rift subdivides into 3 separate arms, which are the St. Louis arm (oriented northwest–southeast), the Southern Indiana arm (oriented northeast–southwest, i.e., along the northeast extension of the Reelfoot Rift), and the Rough Creek Graben (oriented east–west; Braile et al., 1982).
Several EQL were documented in relation with the New Madrid earthquake series. Some luminosities were observed near the epicentral area, while others were seen much further away, for example (Fuller, 1912; Corliss, 2001):
Prior to the shock of 8 February 1812 in Livingston County, Kentucky (115 km to the northeast).
In St. Louis, Missouri (235 km to the north‐northwest), near the St. Louis arm of the Reelfoot Rift (Braile et al., 1982).
Following the first shock of 16 December 1811in Bardstown, Kentucky (365 km to the east‐northeast), within the Rough Creek Graben.
Following the first shock, in Knoxville, Tennessee (500 km to the east).
Following the first shock, in Hot Springs, North Carolina (600 km to the east), along the northwest margin of the Blue Ridge rift (Hough, 2000).
Pisco, Peru Earthquake, 2007 August 15, 18h41
The Mw 8.0 Pisco, Peru earthquake (also known as the Ica earthquake) occurred 25 km offshore in the Pacific, about 60 km to the northwest of the coastal city of Pisco and 150 km to the south‐southeast of Lima (EERI, 2007). Two rupture events were associated with this earthquake, the first near the location of the reported epicenter, and the second (60 seconds later) about 60 km to the south and just west of the Paracas Peninsula (Sladen et al., 2010). The seismic event is seemingly related to the subduction of the Nazca plate under the South American plate. The location of the second rupture event coincides more or less with the southwestern termination of the Pisco‐Juruá fault, a southwest–northeast oriented, 3000 km long intraplate structure that extends from the Paracas Peninsula to eastern Guyana, near the Atlantic coast. This continental scale structure is interpreted to have acted as a normal fault in the early Paleozoic, and to have been later reactivated as a sinistral strike‐slip fault during rifting of the Atlantic in the Mesozoic (Szatmari, 1983; James, 2007). Furthermore, the earthquake is located about 100 km to the northwest of the termination of the Nazca Ridge, which is interpreted to represent the trace of the Easter hotspot, located several thousand kilometers to the west in the Pacific (Schissel and Smail, 2001; Smith, 2007). Note that the Nazca Ridge is located more or less along the trend of the Pisco‐Juruá fault (Fig. 1).
During the Pisco, Peru earthquake, a large number of luminous phenomena were witnessed by people along the Pacific coast, from Ica to Lima, respectively, about 100 km to the southeast and 150 km to the north‐northwest of the epicenter (Ocola and Torres, 2007). Heraud and Lira (2011) report more specifically on coseismic luminescence observed in the Lima area, including records of security cameras operating (i) in a mall overlooking the shoreline and (ii) on the campus of the Pontificia Universidad Católica del Perú (PUCP), as well as eyewitness reports considered to be of high quality. Together with seismic records obtained on the PUCP campus, the automatic security camera records allow for an exact timing of light flashes that illuminated a large portion of the night sky. The light flashes identified as EQL coincided with the passage of the S waves. The video records and eyewitness reports provided the exact location of the EQL. For example, as described by Heraud and Lira (2011), a navy officer on the pier on San Lorenzo Island, while looking toward the coast of Lima, saw light blue columns of light bursting four times in succession seemingly out of the water between the El Fronton Island and the coast (Fig. 4). The exact location from where the EQL were emitted between the El Fronton Island and the coast was a cluster of rocky islets sticking out of the shallow water. Other luminous phenomena associated with the Pisco earthquake were reported by Ocola and Torres (2007), some near Lima, but most of them closer to the epicenter, in the Pisco region.
Ebingen, Germany Earthquake—1911 November 16, 22h25
The Mw 5.8 Ebingen earthquake, also called the Württemberg earthquake, occurred on the western outskirts of Ebingen in the Swabian Jura of Germany (ECOS, 2013). The earthquake took place approximately 8 km to the southwest of the Hohenzollern Graben, a 2 km wide by 30 km long, southeast–northwest oriented rift structure that is located approximately 80 km to the east of the Upper Rhine Graben (Figs. 2 and 5). This region is characterized by the presence of abundant alkalic volcanic rocks, which are exposed at the Kaiserstuhl, Hegau, and Urach volcanic fields (Fig. 5). Tertiary volcanic activity is interpreted as being related to rifting of the Rhine Graben system (Keller, 1985). The area in the vicinity of the Hohenzollern Graben is seismically active, as indicated by >25 seismic events of magnitude greater than 4.0 having occurred in this region since about 1850 (Baumann, 1984).
Originally, a total of 110 light sightings were reported for the Ebingen earthquake (De Ballore, 1913). Of those, 43 observations were described by von Schmidt and Mack (1912). As shown in Figure 5, EQL were observed as far as 110 km from the epicenter, and appear to have had a tendency to occur along the eastern margin of the Upper Rhine Graben and in the vicinity of the Hohenzollern Graben. Several light sightings were also reported from the margins of the Tertiary volcanic fields. One of the most detailed EQL observations was made near Ebingen, in which two people, after hearing a distant noise accompanied by a faint vibration, saw a bright flash emitted from the ground, which then, at a considerable height, turned into a ball of light and eventually divided itself like lightning in the direction of Ebingen (Fig. 6). The tremors began with the appearance of the ball of light. After the seismic waves had rolled past, toward the town, the two witnesses observed a second sphere of light, while the entire surroundings were brightly illuminated (von Schmidt and Mack, 1912).
L’Aquila, Italy Earthquake—2009 April 06, 03h32
The Mw 6.3 L’Aquila earthquake occurred approximately 5 km to the southwest of the city center of L’Aquila, near the southwestern edge of the Aterno River Basin. It was followed by two major aftershocks one day (April 7; Mw 5.6) and 3 days (April 9; Mw 5.4) after the mainshock, respectively. Surface faulting took place along the southwest dipping, northwest–southeast trending Paganica normal fault (EMERGEO Working Group, 2010). The general area surrounding L’Aquila is transected by several northwest–southeast oriented normal faults that form a series of grabens and half‐grabens spread over a width of about 40 km (Fig. 7; Fidani, 2010). These normal faults occur within the Central Apennines mountain chain, which has been undergoing crustal extension in the recent geological past (D’Agostino et al., 2008).
EQL were seen up to 45 km from L’Aquila (Fig. 7). As reported by Fidani (2010), among 1057 reported anomalies of various types associated with the L’Aquila earthquake, 241 were classified as anomalous atmospheric luminosities. Of these, 136 conform to the EQL classification by Montandon (1948), which does not take into account luminous clouds, vapor, and streamers. As with the seismicity pattern (Pondrelli et al., 2010), some anomalous luminous phenomena were reported months before the April 6 mainshock. Marked seismicity started around March, as did the majority of EQL sightings (Fidani, 2010). Most of the EQL were seen along or close to the numerous northwest–southeast oriented normal faults present in the epicentral area.
The general area surrounding L’Aquila was the site of two historic earthquakes for which associated EQL were also witnessed. These earthquakes occurred on 2 February 1703 (Mw 6.65) and on 28 June 1898 (Mw 5.48; Galli, 1910).
Due in part to restrictions regarding the length of the paper, earthquakes with associated luminosities from seismically active regions in the Asian continent were not included in this paper. However, using preliminary data, we have noted that out of seven identified events in China, a total of five EQL sightings (Geotimes, 1977; Huang and Deng, 1979; Wallace and Teng, 1980; King, 1983) were associated with earthquakes occurring in intraplate rift environments, namely: (1) the Mw 7.2 Xingtai 1966 and Ms 7.8 Tangshan 1976 earthquakes of the North China Rift (Replumaz and Tapponnier, 2003; Liu et al., 2011); (2) the M 6–7 Qishan BC780 and M 6.75 Pinglu 1815 earthquakes of the Shanxi Graben System (Huang and Deng, 1979; Liu et al., 2011); and (3) the M 7.5 Longling 1976 earthquake of the Tengchong Rift (Socquet and Pubellier, 2005). To a lesser extent, the occurrence of EQL in association with a rift setting was also identified in the course of our research from other parts of Asia. For example, in India, the well‐reported M 7.8 Rann of Kachchh 1819 earthquake, which was accompanied by EQL, occurred within the Kachchh Rift Zone (Macmurdo, 1821; Jain et al., 2002). Similarly, in Japan, we have identified a few earthquakes associated with EQL that occurred within a rift environment, for example, the M 8.1 Nankaido 1946 earthquake of the Kyushu Rift Valley (Yasui, 1973; Huang and Deng, 1979; Okubo et al., 2006) and the M 5.1 (maximum magnitude) Matsushiro 1965–1966 earthquake swarm of the Nagano Basin (Yasui, 1971; Derr, 1977; Tsuneishi, 1978). Furthermore, the M 7.2 Hyogo‐ken Nanbu (Kobe) 1995 earthquake, where a total of 25 eyewitness reports of EQL were documented (Tsukuda, 1997; Kamogawa et al., 2005), was associated with an important system of subvertical faults, with the main fault undergoing dextral strike‐slip displacement over a length of 30–50 km (Somerville, 1995).
Magnitude of Earthquakes Associated with Luminosities
As can be seen from Tables 1 and 2, which list the main earthquakes associated with luminosities compiled respectively from the Americas and Europe, EQL are generated in association with earthquakes over a wide range of magnitude from 3.6 to 9.5. It can hence be concluded that EQL may occur regardless of the earthquake magnitude, although the majority of the listed cases (i.e., 80%) were observed for events with magnitudes greater than 5.0.
As already noted by Hedervari and Noszticzius (1985), our compilation also indicates that the maximum distance at which EQL are observed tends to increase with the magnitude of the event. For example, EQL have been reported for distances up to 600 km from the epicenter in the case of the New Madrid earthquake, which had a magnitude of about 8 (Table 1).
Timing of EQL Relative to Associated Earthquakes
A characteristic feature of seismic luminosities (and other earthquake‐associated phenomena) is the observation that most EQL are seen before and/or during an earthquake, but rarely after the release and dissipation of the seismic‐stress energy in the crust. This strongly suggests that the processes responsible for EQL formation are related to a rapid build‐up of stress prior to fault rupture and rapid stress changes during the actual fault movement.
Based on our compilation, most pre‐earthquake luminosities have been observed from a few seconds to up to 3–4 weeks prior to seismic events, such as the Saguenay earthquake (St‐Laurent, 2000). The duration of EQL varies from less than a second to several minutes. They vary in shape and extent, the most frequent occurrences being globular luminous masses, stationary in the air or moving. Light emission during the time of seismic activity, termed coseismic luminosity, is most frequently observed as either short flashes of light shooting high up into the air, diffuse but relatively bright atmospheric illuminations that last from seconds to a few minutes, or flame‐like luminosities coming out of the ground.
Distance between EQL and Earthquake Epicenter
At rare occasions, EQL have been seen as far as 600 km from any given epicenter, as our compiled list of earthquakes shows (e.g., New Madrid earthquake, Table 1). More typically, EQL have been observed at distances not more than about 300 km from an epicenter. Pre‐earthquake luminosities were generally seen closer to the epicenter relative to coseismic luminosities, a few at 200 km but the majority of them occurring at 150 km or less.
It is important to note that when EQL were seen far away from the epicenter, as some reports for the New Madrid earthquake suggest, they seem to be always time correlated with the passing of the seismic waves. The most definite evidence comes from Lima, Peru, in which the passage of the seismic wavetrain associated with the 2007 Mw 8.0 Pisco earthquake (coming from a distance of 150 km) was recorded by a seismometer on the PUCP university campus, while the EQL were recorded simultaneously by automated surveillance cameras (Heraud and Lira, 2011). In this case, it was clear that the outbursts of light did not occur during the passage of the compressional (P) waves but during the passage of the shear (S) waves. This implies a direct coupling between the crustal rocks and the very rapid, high‐amplitude change in shear stress caused by the S wavetrains (Gharibi et al., 2003).
The dynamic nature of this process is supported by the successive appearance of EQL at an increasing distance away from the epicenter as reported for the New Madrid earthquake. This mechanism is additionally supported by direct observations of the instant generation of an electric field in an underground laboratory in Japan during the arrival of the S waves from a moderate M 4.6 earthquake, which occurred 75 km away (Takeuchi et al., 2010).
Frequency of Intraplate Versus Interplate Earthquakes Associated with Luminosities
More than 95% of the world’s earthquakes are interplate earthquakes occurring along active plate boundaries. The three most seismically active interplate regions are: (1) the circum‐Pacific belt (90% of all earthquakes), termed the “Ring of Fire”; (2) the Alpide belt (5%–6% of all earthquakes), termed Alpine–Himalayan belt, which stretches from the Mediterranean region eastward through Turkey, Iran, Pakistan, northern India, Indochina to Sumatra; and (3) the Mid‐Ocean ridges, circling the globe (USGS, 2013). Hence, less than 5% of the world’s earthquakes occur in intraplate tectonic settings, but they are the ones for which most EQL are reported. A model is presented below that explains this dichotomy. Although earthquakes that occur in interplate tectonic settings (subduction zones) can occasionally generate EQL, the latter are most often seen along subvertical, normal, or strike‐slip faults located at a fair distance (≥80 km) from the epicenter and associated subduction zones, where they are often related to a paleorift.
A recent worldwide survey of the seismicity in intraplate regions revealed that 64% of the earthquakes in these seemingly stable continental tectonic environments were associated with failed rift basins or grabens (Gangopadhyay and Talwani, 2003). Intracontinental rifts and grabens are characterized by subvertical crustal faults that generally extend downward to depths of up to at least 40–50 km, that is, down to the Moho, which marks the boundary between the crust and the upper mantle. In our study, we included in a similar category not only earthquakes associated with old failed rifts and grabens located in tectonically stable intraplate environments, but also earthquakes that were associated with clearly identifiable paleorifts along the margins of presently active or recently active orogenic belts such as the Andes, Pyrenees, and the Alpine–Himalayan belt. In the latter tectonic settings, the steeply dipping and deeply penetrating crustal faults associated with paleorifts are typically reactivated into strike‐slip faults, less commonly into reverse faults.
Based on our compilation of 65 earthquakes in the Americas and Europe, which were associated with well‐documented EQL, a total of 37 earthquakes and/or their associated luminosities occurred in or near an intraplate rift or graben, 19 occurred in or near a back‐arc or pull‐apart rift (or paleorift) structure within an orogenic system, 7 occurred in or near a transform or strike‐slip fault, and only 2 occurred within an orogenic system with no recognized, rift‐related structures. Hence, 56 out of the total of 65 (85%) of the investigated earthquakes produced luminosities along a rift or paleorift structure. Furthermore, 63 out of the 65 (97%) occurred at or adjacent to regional subvertical faults (e.g., a rift, graben, or strike‐slip or transform fault). The remaining 2 earthquakes that produced luminosities were associated with shallow‐dipping thrust faults in subduction zone settings (Table 3).
From a compilation of 30 cases of luminous phenomena associated with earthquakes in the East Mediterranean dating from the 5th century B.C. to the recent past, Papadopoulos (1999) also noted a clear prevalence of cases associated with normal and strike‐slip faulting environments. In his compilation, Papadopoulos (1999) mentions that luminosities were reported for about 5%–6% of all major earthquakes. This number agrees very well with what was previously reported by De Ballore (1913) and with what was evaluated from the work of Mallet (1855) for similar regional EQL studies. If one considers that EQL will be hardly visible during daytime, and hence unobservable and unreported, it can then be argued that, because EQL certainly also occur during daytime, the 5%–6% number may well translate to about 10% of all such earthquakes. This percentage represents a lower limit considering that many EQL observed are never reported and published in the scientific literature, or are interpreted as “unidentified flying objects (UFOs)” (e.g., Persinger and Derr, 1984).
Model for the Generation of Earth Currents and EQL
Our preferred model for the generation and propagation of earth currents and ensuing EQL formation is based on work by Freund et al. (1994, 2006, 2007, 2009), Freund (2002, 2007, 2010), and Freund and Pilorz (2012) that describes experiments stressing igneous rocks (quartz‐bearing and quartz‐free), limestone, marble, and others. These experiments demonstrate that electronic charge carriers are activated in the high‐grade metamorphic and igneous rocks (in particular mafic and ultramafic rocks) when subjected to deviatoric stresses, turning them into semiconductors. The charge carriers derive from pre‐existing defects in the matrix of the minerals, electrically inactive in their dormant state as peroxy bonds or links (i.e., O3Si/OO∖SiO3), and are introduced into the matrix of minerals during cooling at high temperatures when two oxygen anions convert from their normal valence state 2− to the valence state 1−, that is, O2− to O−. When subjected to stress, mineral grains slide along grain boundaries or dislocations sweep through, causing peroxy links to break. The O− states thus formed represent defect electrons in the oxygen anion sublattice, which turn into highly mobile electronic charge carriers, referred to as positive holes or pholes. These previously unrecognized charge carriers have the remarkable ability to flow out of the stressed rock volume and to move away from where they have been generated.
Invariably, several types of pholes are generated during stressing of rocks, characterized by different lifetimes ranging from less than a second to longer than days. As the long‐lived pholes diffuse outward, they can reach the Earth’s surface. There, they form surface/subsurface charge layers, which cause locally high electric fields, often strong enough to ionize the air and even trigger corona discharges. The corona discharges are associated with the emission of visible light close to the ground and with the formation of ozone.
There is yet another aspect of the same basic process of stress activation of pholes: the highest charge carrier densities can be achieved if stresses increase so rapidly that even short‐lived pholes do not have the time to recombine. This implies that, if tectonic stresses deep in the Earth’s crust increase very rapidly in any given rock volume, the number densities of pholes can reach a critical value beyond which the electronic wave functions of both the pholes and the coactivated electrons begin to overlap. This is expected to create a plasma‐like state, that is, a volume of rock with a very high mobile‐charge density and high conductivity. It has been suggested (St‐Laurent et al., 2006) that, inside the Earth’s crust, this plasma state will become unstable and will rapidly expand outward. When such an intense charge state reaches the Earth’s surface and crosses the ground–air interface, it is expected to cause a dielectric breakdown of the air and, hence, an outburst of light. This process is suspected to be responsible for flashes of light coming out of the ground and expanding to considerable heights at the time when seismic waves from a large earthquake pass by. Those waves, especially S waves, subject the rocks to very rapid shear forces, causing mineral grains to move relative to each other, possibly even generating dislocation movements within the grains. This activates peroxy defects and creates the capability to momentarily generate high concentrations of pholes (Heraud and Lira, 2011). Igneous rocks, in particular mafic igneous rocks, have much higher concentrations of pre‐existing peroxy defects than sedimentary rocks. Hence, the processes that seem to be responsible for the generation of EQL can be expected to occur preferentially in those rocks, providing a possible explanation for the often reported close association of EQL with mafic dikes and intrusions (e.g., Saguenay, Ebingen, and Pisco Peru earthquakes).
The positive hole theory can account not only for EQL but also for other pre‐earthquake phenomena, such as:
Air ionization at the ground‐to‐air interface.
Changes in the electrical conductivity of the soil.
Geo‐electric and geomagnetic anomalies in the Earth’s crust.
Ultralow and extremely low frequency (ULF/ELF) and radio frequency (RF) emissions.
Anomalous infrared emissions from around a future epicentral area.
Anomalous fog/haze/cloud formation and unusual animal behavior (Derr et al., 2011).
As mentioned previously, our compilation of earthquakes associated with EQL for the Americas and Europe show that a high proportion of the documented EQL are spatially associated with deeply penetrating and steeply dipping crustal faults, and that these structures are often located a significant distance away from the associated epicenter (e.g., 80 km or more). The EQL events occur not only coseismically but are spread in time over periods of days, weeks, and sometime months before and after given earthquakes. This suggests that the processes responsible for the generation of EQL either cause rock volumes of different composition along the same tectonic structures to separately undergo rapid stressing ahead of an impeding earthquake, or may involve the focusing of pholes, stress activated in a larger volume, into rocks with inherently higher electrical conductivity. Such a focused flow of electronic charges may explain the observed luminous discharges at the ground/air interface in close proximity to deeply penetrating subvertical faults typical of many rift structures.
Figure 8 shows a simplified tectonic model explaining the occurrence of EQL within an intraplate rift or graben setting, in this case pertaining to the Saguenay earthquake. Stress‐induced charge carriers are interpreted as having been generated prior to and during the earthquake series. According to our model, pholes and electrons are coactivated when, due to strain acting on a given rock volume at a constant rate, the stresses will increase exponentially (Freund and Sornette, 2007). In this simplified scenario, probably applicable to the depth of the Saguenay hypocentre, the positive holes would propagate outward from the stressed source volume, flowing through the p‐type upper crust and toward the surface, whereas electrons might migrate downward within the n‐type lower crust (Freund, 2007; Balk et al., 2009). Assuming a plasma‐like state, instability can develop, causing the plasma to “explode” through the surface, leading to a visible light flash. The luminous discharge into the air could also be favored at sites with high topographic relief, if the conditions below the surface are conducive to the formation of plasma instability. Note that enhanced electric fields associated with luminescence have for a long time been observed at elevated areas and/or around or above pointed objects (Knoche, 1912; Montandon, 1948; Robson, 1955; Markson and Nelson, 1970; Derr, 1973; Yasui, 1973; St‐Laurent, 1991, 2000).
Figure 9 illustrates a tectonic model where EQL are considered to occur within an Andean‐type, interplate orogenic setting (i.e., subduction zone environment). As explained in the model for the rift setting, stress‐induced charges may be generated prior to and/or during an earthquake, with the source area being usually located either within or adjacent to a subduction zone (region A), or more inland within a back‐arc rift, paleorift, or graben (region B). These charges, while propagating away from the hypocenter, will be subjected to scattering and/or focusing along their path and, eventually, will reach the surface predominantly in areas of high topographic relief or adjacent to subvertical faults. In these areas, the charge carrier densities will be highest, leading to ionization of the air and generation of luminosities. Note that stress‐induced charges may propagate over long distances from a subduction zone setting to inland rift settings and lead to EQL, possibly without the need for a localized source of rapidly increasing stress in the latter area. One example of this setting is the Santiago earthquake of 1851, where the epicenter was located in the general area of the subduction zone near the Pacific coast, whereas the EQL were observed about 100 km inland within the Central Valley Graben paleorift structure (see electronic supplement). The recent devastating M 9.0 Tohoku earthquake of March 2011 (Honshu, Japan) resulted from thrust faulting along a subduction zone (USGS, 2013). The fact that no associated luminosities from the surrounding area have come to our attention yet seems to be in accordance with our results. One could argue that, because it was a daytime earthquake (i.e., 14h46 local time), light observations would have been difficult to note. However, considering the strength of this earthquake, precursor luminous phenomena should, at least, have been observed during the previous night or nights, as was often the case for large intraplate earthquakes discussed previously. Based on our model, we believe that visible electrical discharges probably did not take place because the active fault related to the Tohoku earthquake was a shallow‐dipping thrust fault and, importantly, this very large earthquake occurred 130 km offshore under about 1500 m of ocean water, hence without direct connection to the atmosphere. If our tectonic model is correct, potential areas where corona discharges or other disruptive discharge phenomena could preferentially take place along the Japanese coast would be in association with regional subvertical faults and/or islets protruding from the sea, similar to those from which EQL were seen during the M 8.0 Pisco, Peru earthquake of 2007 (Heraud and Lira, 2011).
Earthquake lights may be classified into two different groups based on their time of appearance: (1) preseismic EQL, which generally occur a few seconds to up to a few weeks prior to an earthquake, and are generally observed closer to the epicenter and (2) coseismic EQL, which can occur either near the epicenter (“earthquake‐induced stress”), or at significant distances away from the epicenter during the passage of the seismic wavetrain, in particular during the passage of S waves (“wave‐induced stress”). EQL during the lower magnitude aftershock series seem to be rare. It is also worth mentioning that luminous phenomena with the same characteristics as EQL have been documented from areas where no temporally associated earthquake could be identified (e.g., Derr and Persinger, 1990). The process responsible for the generation of such luminosities has been explained in terms of the so‐called “tectonic strain theory” (Persinger, 1983; Persinger and Derr, 1984). Within the framework of the work presented here, such observations are consistent with the fact that not every stress build‐up in the Earth’s crust will be followed by catastrophic rock rupture leading to an earthquake. We suggest that this broader category of luminous phenomena be referred as tectonic stress lights.
When studied individually, some EQL reports may appear questionable. However, a large number of eyewitness reports from certain areas (e.g., Saguenay, Pisco, L’Aquila), coupled with similarities with respect to shapes and colors (e.g., globes, flame‐like luminosities) for incidences in very different regions of the world should be taken as evidence that EQL occurrences are real and a widespread phenomenon. A unifying theory explaining the origin of different types of nonseismic, pre‐earthquake signals (including EQL) has recently been proposed, which is based on the generation of stress‐activated electric currents in rocks (Freund, 2010). Mobile charge carriers, termed positive holes or pholes, flow along stress gradients and eventually accumulate at the surface, ionizing air molecules and leading to the generation of luminosities, among other phenomena.
Our study has shown that the vast majority of EQL (i.e., 97%) have been observed in the following three tectonic environments: (1) intraplate rifts or grabens; (2) back‐arc or pull‐apart rifts or grabens (or paleorifts) located inland from subduction zones (orogenic settings); and (3) strike‐slip or transform faults, irrespective of the tectonic setting. The common characteristic of these three geological settings appears to be the presence of deeply penetrating subvertical faults, which exact role, passive or active, in phole propagation and EQL formation has yet to be resolved.
We would like to thank Eliane de Nicolini and Richard Thériault for having drawn our attention to the EQL they observed over Quebec City at the time of the Saguenay earthquake, and for providing us with a preliminary sketch of their sighting (see electronic supplement). Thanks are extended to Simona Derr for later redrafting this sketch. We are also grateful to Nathalie Asselin for her description and drafting of the EQL she observed in Saint‐Prime, also at the time of the Saguenay earthquake (see electronic supplement). Finally, we wish to acknowledge Dr. John Ebel for having brought to our attention the EQL sightings associated with the Newbury earthquake.
We also wish to thank Massimo Silvestri, Nico Conti, Roberto Labanti, Maurizio Morini, and Renzo Cabassi, from the Italian Committee for Project Hessdalen and Smart Optical Sensors Observatory (ICPH‐SOSO), who helped with some Italian translation and historic documents. Steven Soter kindly translated many German reports of the Ebingen earthquake. We would also like to express our gratitude to Marc Hallet for the translation of a Latin document and for providing old reports of Robert Mallet. Special thanks are extended to Tibor Zsiros for providing us with a Latin document for the Mór earthquake.
Finally, John Ebel and an anonymous reviewer are thanked for their critical review that significantly improved the original manuscript.