The search for more reliable earthquake precursors continues, however, and for some time there has been interest in electromagnetic (EM) signals within the Earth’s crust that have been reported prior to earthquakes. In a paper published in Geophysical Research Letters, Minoru Tsutsui ⁷¹ of Kyoto Sangyo University (Japan), reports the development of a boreholebased system capable of detecting extremely low frequency (ELF) EM pulses, and its use in detecting an intense EM pulse at the time of a magnitude 5.5 earthquake, and originating at the quake’s epicentre 130 km distant. Given the near simultaneous occurrence of the quake and the pulse, there would seem to be limited scope for using such a system as a predictive tool. Tsutui points out, however, that more research needs to be undertaken into the relationship between EM pulses and earthquakes. As it seems likely that the pulses represent a so-called ‘piezoelectric effect’, caused by the pressurisation or shearing of rock, the detection of EM pulses may prove useful in monitoring movements within the crust that precede an earthquake, providing at least the potential for useful prediction.

Earthquake forecasting, as opposed to prediction, involves a broad assessment of the likelihood of a forthcoming quake. Earthquake forecasts are often probabilistically-based, for example in relation to the average return period of a particular size quake. A forecast may also be predicated or modified on the basis of observations that might point to the possibility of a coming quake, but without providing any information on timing. Examples include deformation of the ground surface or patterns of small earthquakes. In the journal Nature, Joanne Bourgeois ¹⁰ of the University of Washington (Washington State, US) addresses tell-tale surface deformation, which seems to precede past great earthquakes in the Alaska and Cascadia Subduction Zones off the west coast of North America. It has been known for some time that major subduction zone earthquakes are accompanied by significant co-seismic subsidence. For example, the 1964 Moment Magnitude 9.2 Alaska earthquake resulted in between 1 - 2 m of subsidence along the coast, with similar effects recorded along the Washington coastline following the 1700 c. magnitude 8.7 - 9.2 Cascadia earthquake. Bourgeois, however, reviews new studies that seem to have picked up a degree of subsidence before each of these great seismic events. The amount of subsidence, determined on the basis of examination of key species of microfossils in the coastal environments, appears to be on the order of 10 - 30 cm, and - prior to the Alaska 1964 event - seems to have started about a decade before the quake. Although the results and their significance remain a matter for debate, one possible cause of the subsidence is the occurrence of so-called ‘slow earthquakes’ deeper down in the subduction zones, which release their energies over periods of weeks, rather than minutes, and are thus difficult to detect using seismometers. Worryingly, such events have been detected in the Cascadia Subduction Zone since 1992, using GPS. Could this mean that an imminent great Cascadia earthquake is on its way?

Seismic hazard in California

One hundred years after the Great San Francisco earthquake, and 12 years on from Northridge, Californians are getting jumpy about when and where the next ‘big one’ will happen. Far more is now known about Californian earthquakes than just 12 years ago, let alone the beginning of the last century, and much of this is down to what is known as the Parkfield Earthquake Prediction Experiment (PEPE). Twenty years ago the 40 km long Parkfield section of the San Andreas Fault was recognised as a promising seismic laboratory where a future quake was almost certain to happen within decades. As a consequence, the section was intensively instrumented so that when the quake did finally occur, detailed observations could be made. The magnitude 6.0 Parkfield earthquake happened on September 28th, 2004 (figure 6), providing a wealth of data and pointing up major implications for earthquake prediction and hazard assessment in the state. William Bakun ⁵ of the United States Geological Survey, and colleagues, summarise the findings of the PEPE in an article in Nature. One of the most striking features of the Parkfield quake was the lack of detectable precursors, despite the fault rupture occurring in the middle of a dense network of instruments. While identical foreshocks occurred prior to Parkfield earthquakes in 1934 and 1966, none preceded the 1901 and 1922 events. Neither did any foreshocks precede the 2004 event, with not even any micro-seismicity being recorded in the six days up to the quake. Several strain meters recorded subtle strain changes in the 24 hours before the quake, but although these provided information on the physics of the earthquake, they were too small to form a basis for warning that a potentially damaging quake was imminent. Notwithstanding an absence of sufficiently robust precursors to make a prediction of the Parkfield 2004 event, Bakun and his co-authors stress that a huge amount of useful data was gathered. This is addressed in some detail in the paper, with particular attention paid to the structure of the fault, prediction of damaging ground motion, the long-term randomness of earthquakes, and implications for future research.



Figure 6. The Parkfield earthquake of September 28th 2004, showing the main Moment Magnitude 6.0 shock, and aftershocks that occurred over in the following 33 hours.

Courtesy: United States Geological Survey.

Southern California appears to have been a focus of attention for researchers over the past 12 months, and two papers highlight concerns that a large earthquake is due in the region. Writing in Nature, Yuri Fialko ²³ of the University of California at San Diego, reports the results of a satellite-based survey of deformation along the southern part of the San Andreas Fault, where no large earthquake has occurred for at least 250 years. Assuming an average slip rate of a few centimetres a year, which characterises the rest of the San Andreas Fault, Fialko estimates that between 7 and 10 m of slip deficit has been accrued along the southern section of the fault. In other words, deep within the crust, this much movement has occurred along the fault, but at shallower depths the fault has been locked, leading to an enormous build-up in strain. Fialco estimates that this degree of slip deficit is on a par with the maximum experienced between quakes, and is sufficient to trigger a magnitude 8 earthquake should all strain be released in one go.

The seismic threat to southern California is also highlighted in a second paper, published in Geophysical Research Letters by Kim Olsen ⁴⁹ of San Diego State University, and co-workers. Olsen and his colleagues note that the southernmost section of the San Andreas Fault has a high probability of rupturing in a large earthquake (> magnitude 7.5) in the next few decades. To constrain better the expected strong ground shaking that is likely to result, the authors have simulated the effect of the underlying geology, and have come up with worrying results. A chain of sediment-filled basins between San Bernardino and downtown Los Angeles act as waveguides, which channel surface seismic waves along the southern edge of the San Bernardino and San Gabriel Mountains. During a future earthquake, this will lead to unusually high, long-period ground motions over much of the greater Los Angeles region. While major damage can be expected, the authors note that it is fortunate that the locations of many high-rises, such as downtown Los Angeles, Long Beach and Santa Monica, are outside the predicted areas of strongest shaking, as tall buildings are most susceptible to such long-period waves.

Latest news from the New Madrid Seismic Zone

Earthquakes in the New Madrid Seismic Zone (NMSZ) (figure 7), in the central United States, are far less frequent then those of California, but they have the potential to be at least as damaging and lethal. A prerequisite for large, frequent earthquakes is the rapid accumulation in the crust of a significant amount of strain energy. This process readily occurs along the boundaries between the Earth’s tectonic plates, for example along California’s San Andreas Fault and Turkey’s North Anatolian Fault. Away from plate margins, in intra-plate settings, rapid strain accumulation is more difficult, and at New Madrid its occurrence has been disputed. Bob Smalley ⁶³, and co-workers, at the University of Memphis, however,


Figure 7. (Left). Shake map for the Moment Magnitude 7.3 New Madrid earthquake of December 15th 1812, showing the enormous extent of ground shaking compared to the samesized Landers (California) quake of May 28th, 1992 (a).

(Right). Map of the New Madrid and neighbouring Wabash Valley seismic zones shows earthquakes as circles. Red circles indicate earthquakes that occurred from 1974 to 2002 with magnitudes larger than 2.5 located using modern instruments (University of Memphis). Green circles denote earthquakes that occurred prior to 1974. Larger earthquakes are represented by larger circles.

Courtesy: United States Geological Survey


announce in Nature, evidence for rapid strain rates in the NMSZ. Smalley and his team report the detection, using a continuously-recording GPS network, of surface movements of up to a few millimetres a year. Such movements reflect strain rates that are comparable in magnitude to those observed at plate boundaries, and which are consistent with known faults in the region. The authors note that the explanation for the surface movements is problematical. The movements, for example, are spatially confined, possibly suggesting that the driving force of New Madrid earthquakes is local rather than regional, but the source of which remains to be unequivocally determined. One possibility is that the movements represent some long-term post-seismic ‘settling’ process following the major 1811-12 event, although the time span since is unusually lengthy. Whatever the cause of the surface movements, and associated strain accumulation, Smalley and his colleagues rightly note that their detection makes an important contribution to the ongoing debate in relation to seismic hazard and risk in the NMSZ. As a footnote, the results of Smalley’s team have been disputed by Eric Calais ¹² of Purdue University, and colleagues, who address alternative interpretations of the deformation data, and their comments and Smalley’s reply are also published in Nature.

New probabilistic seismic hazard assessment for greater Tokyo

Writing in the Extreme Natural Hazards special issue of Philosophical Transactions of the Royal Society (A), Ross Stein ⁶⁶ of the United States Geological Survey, and colleagues, present a new probabilistic seismic hazard assessment for greater Tokyo, based upon a joint Japanese - US study. With major, destructive earthquakes striking the Japanese capital in 1703, 1855 and 1923, concern is growing about the nature, timing, size and impact of the next one. To try and answer these questions, Stein and his group used the prehistoric record of great earthquakes preserved by uplifted marine terraces and tsunami deposits, a newly digitized dataset of historical shaking, the dense modern seismic network, and Japan’s GeoNet array to comprehensively reinterpret the tectonic structure, identify active faults and their slip rates and estimate their earthquake frequency. On the basis of these new data, the authors determine that the likelihood for severe shaking (ca 0.9g peak ground acceleration (PGA)) in Tokyo, Kawasaki and Yokohama for the next 30 years is c. 29 percent, which corresponds to an annual probability of about 1.2 percent. Based upon a US$1 trillion estimate for the cost of a future Magnitude 7.3 shock beneath Tokyo, Stein and his co-researchers note that the annual probable loss for the Japanese government is US$12 billion.

National studies of seismic hazard and risk

Three studies, published in the journal Natural Hazards, and focusing on aspects of country-wide seismic hazard and risk are summarised here. The first, by Jochen Schwarz ⁶¹, and colleagues, at the Bauhaus - Universität, Weimar, involves comparative seismic risk assessments in Germany, using the European Macroseismic Scale (EMS-98). The authors apply a GIS-based seismic risk assessment to two communities, Schmölln in eastern Thuringia, representing one of the country’s many small towns, and the city of Cologne, as representative of the larger urban centres. One of the paper’s conclusions is that, for stronger earthquakes, damage will be greater in small towns such as Schmölln, due to increased vulnerability of the building stock and other factors, than in larger cities exemplified by Cologne. The second, by Roberto Romeo ⁵⁶, of the University of Urbino, presents an assessment of earthquake hazard in Italy for the period 2001 - 2030, on the basis of timedependent earthquake probabilities determined from the historical catalogue. The main findings of the study are: (i) seismic source zones in southern Italy are the most prone to experience damaging quakes in the next 30 years, with conditional probabilities as high as 10 percent; (ii) any influence exerted by earthquake interaction (for example stress transfer) in increasing such probabilities, is too small to be of importance. In the third paper, Wen-Yu Jean ³³ of Taiwan’s National Center for Research on Earthquake Engineering, and co-researchers, describe a system designed to produce shake maps that provide rapid estimates of strong ground shaking due to earthquakes in Taiwan. The high resolution, near-real time, output incorporates local site effect influences on ground shaking, and generates early estimates of potential damage for emergency response services.

Earthquake loss estimation and damage assessment

Catastrophe models of earthquakes now underpin and inform most decisions made by insurers and reinsurers who have vested interests in areas of high seismic hazard and risk. They also have an important role to play, however, in designing building codes and guiding emergency planners and response agencies. Building a loss model, whether for a city, a region or a country, involves compiling databases of past earthquake activity, ground conditions, attenuation equations, building stocks, infrastructure exposure, and vulnerability characteristics. Inevitably, large uncertainties are involved that result in any model constructed being poorly constrained. Many uncertainties, however, can be regarded as epistemic, in the sense that they can be reduced - in theory at least - through a better understanding of the physical processes involved or the acquisition of better data. In a paper published in Earthquake Engineering and Structural Dynamics, Helen Crowley ¹⁷ of the European School for Advanced Studies in Reduction of Seismic Risk (Pavia, Italy), and colleagues, examine the possible impact of epistemic effects on earthquake loss models, by introducing systematic variations into a test case of buildings along the northern shore of the Marmara Sea in Turkey.

In a second paper, in the journal Bulletin of Earthquake Engineering, that continues the uncertainty theme, Crowley teams up with Julian Bommer ⁹ of London’s Imperial College, to examine the influence of ground-motion variability in earthquake loss modelling. Bommer and Crowley note the often large uncertainties associated with input parameters, such as the seismicity itself, ground motion, and the exposure and vulnerability characteristics of the building stock, have to be reduced if any meaningful result is to be derived from an earthquake catastrophe loss model. This, propose the authors, can only be accomplished by first correctly identifying and characterising these uncertainties, and then by incorporating them into the calculations and interpreting the output taking into account their influence. One important element of uncertainty, in all earthquake loss models, will be the aleatory variability in ground motion prediction. Spurning more traditional approaches to this uncertainty, the authors propose that the only approach that is consistent with the real nature of ground-motion variability, is to model the shaking component of the loss model through generating large numbers of synthetic earthquake scenarios, which sample the magnitude and spatial distributions of the seismicity, and also the distribution of ground motions for each event, as defined by the aleatory variability.

Crowley and Bommer ¹⁸ also collaborate in another Bulletin of Earthquake Engineering paper, which examines the modelling of seismic hazard in earthquake loss models with spatially distributed exposure. The prediction of possible future losses from earthquakes that can affect structures over a wide area is of critical importance to national and local governments, emergency planners, and to insurers and reinsurers. The authors propose two ways of tackling the problem; one involves generating loss exceedance curves by performing independent probabilistic seismic hazard assessment calculations at several locations simultaneously, and then combining the losses at each site for each annual frequency of exceedance. The other entails using multiple earthquake scenarios to generate ground motions at all sites of interest, defined through Monte Carlo simulations based on the seismicity model.

Where the underlying geology is soft sediment or landfill, liquefaction can play a major role in building loss and damage. In Soil Dynamics and Earthquake Engineering, Juliet Bird ⁸ of Arup Geotechnics, and colleagues, address the issue of modelling liquefaction in earthquake loss models, wherein large variations in building stock and ground conditions need to be considered. The authors discuss the main modes of building response to a range of uniform and differential ground movements, and examine the uncertainties involved. They propose a unified damage scale with application to both reconnaissance and assessment of all modes of buildings damage, and examine the interaction of ground shaking and liquefaction in relation to induced structural damage. Bird and her co-workers conclude that incorporating liquefaction damage into an earthquake loss model is complex; failing to do this, however, may result in significant errors in resulting damage estimations.

Tsunami hazard and risk in the major ocean basins

Less than two years after the December 26th 2004 Indian Ocean tsunami, and with more than 500 lives lost due to another tsunami that struck the Indonesian island of Java in July 2006, tsunami risk and hazard continues to be a major research area. In a new approach to the subject, Eric Geist and Tom Parsons ²⁷, of the United States Geological Survey, propose - in the journal Natural Hazards - the application of probabilistic analysis to tsunami hazard. The authors note that most tsunami hazard assessments tend to rely upon deterministic or scenario-type models, for example based upon inundation maps derived from the maximum credible tsunami for a particular location or region. While Geist and Parsons accept that some intuitive measure of probability is involved in assigning a maximum credible tsunami, they feel that a more overt probabilistic model is more appropriate, not only in explicitly defining the probability associated with any individual scenario, but also because this enables the determination of an annualised risk, which is of particular use to insurers. In this context, the authors present in the paper a methodology for Probabilistic Tsunami Hazard Analysis (PTHA), along the lines of the Probabilistic Seismic Hazard Analysis (PSHA) that is now in common use. Geist and Parsons present two-case studies; firstly a site-specific hazard analysis applied to Acapulco (Mexico), and secondly, a region-wide analysis of the tsunami threat presented to the US west coast by tsunamis generated in the Cascadia Subduction Zone.

Two further papers in Natural Hazards zero in on tsunami risk in the Pacific Basin, which hosts most of the world’s lethal and destructive tsunamis. In the first, Yoshiki Yamazaki ⁷⁷ and colleagues at the University of Hawaii, present a forecasting model for earthquake-generated tsunamis generated in the Japan-Kurile-Kanchatka Subduction Zone of the NW Pacific, with particular emphasis on the threat posed to the Hawaiian Islands. On the basis of seismic data over the past century, the authors divide the subduction zone into more than 200 separate ‘sub-faults’, along which synthetic tsunami waveforms or ‘mareograms’ are generated for existing tide gauges and tsunami warning buoys in the region. The authors ran the model to simulate two historical, tsunamigenic events in the subduction zone, the 1944 Tonankai earthquake and the 1994 Kurile earthquake, and report good agreement between predicted and actual tsunami heights at tide gauges around the Hawaiian Islands. In the second paper, Byung Ho Choi ¹⁶ of Sungkyunkwan University in South Korea, and co-workers, use a synthetic catalogue to estimate tsunami risk along coastlines adjacent to the Japan Sea. The synthetic catalogue is adopted as a way of getting around a paucity of observational data, and is used to develop a range of possible tsunami scenarios. The catalogue includes just four, real (20th century) events, together with 24 hypothetical tsunamigenic earthquakes, located in ‘seismic gaps’, and a further 76 earthquake tsunami sources distributed across the eastern part of the Japan Sea, which is the main source of distant tsunamis affecting Korea and Russia. Tsunami wave-heights are calculated and used to determine tsunami risk zones in the region.

Switching to the Atlantic Basin, which hosts just two percent of recorded tsunamis, but where the potential for rare but severely damaging events remains real, César Andrade ¹ of the University of Lisbon, and colleagues, focus on evidence for historical tsunamis in the Azores. Like the Hawaiian Islands, the Azores - by virtue of their mid-ocean location, are particularly vulnerable to tsunamis, either generated locally or remotely. Andrade and his fellow researchers report that since the archipelago was settled in the 15th century, its coasts have been struck by 23 tsunamis. The most damaging resulted from the great Lisbon earthquake in 1755, which generated tsunamis with an 11 - 15 m run-up on the island of Terceira. The authors also point out that a number of ‘floods’ attributed to intense Atlantic storms, may have been tsunamis arising from large sub-marine landslides in the region, and warn that a future event on the scale of the 1755 tsunami, would be extremely damaging and costly.

Tsunamis generated by landslides and other non-seismic sources


Figure 8. Tsunami damage to shore properties caused by the landslides at
Stromboli volcano on December 30th 2003.

Courtesy: Tom Pfeiffer

The great majority of tsunamis result from large, shallow, submarine earthquakes. Others, however, arise from submarine landslides, from exploding or collapsing volcanoes or, even, from very rare asteroid or comet impacts in the ocean. The most recent damaging landslide-generated tsunamis, were sourced as recently as December 2002, when masses of rock totalling up to 30 million cubic metres, slid from the flanks of the Stromboli volcano, north of the island of Sicily in the Mediterranean. The events are described in a pair of papers by Stefano Tinti ^{67, 68} of the University of Bologna, and published in Bulletin of Volcanology, one addressing the physical effects, and the other presenting the results of numerical simulation of the events. Tinti and his colleagues describe the formation of two tsunamis, a couple of hours apart, which caused severe damage to property around the Stromboli coast, and also damaged buildings on the island of Panarea, 20 km distant. The waves were also detected across the region, including along the north coast of Sicily and on the Calabrian and Campanian coasts of the Italian mainland. For Stromboli, the authors report run-up heights of up to 10.9 m, which led to the complete destruction of some buildings close to the sea and severe damage to others (figure 8). At Panarea, a 2.3 m run-up resulted in minor damage to coastal properties and structures. Although only three injuries resulted from the waves, a much higher toll would probably have resulted if the waves had struck at the height of the holiday season, when many more sea-front properties would have bee occupied. Tinti and his colleagues, point to the fact that similar landslide-generated tsunamis may have occurred on as many as six occasions since 1919, and caution that this continues to be a threat to those living around the coasts of Stromboli island.

Tsunami-producing landslides from active volcanoes can be far larger than those generated at Stromboli, and Eli Silver of the University of California, Santa Cruz, Simon Day ⁶² (currently at BUHRC), and others, describe the largest such event in historical times in EOS. In 1888, the Ritter Island volcano, in Papua New Guinea, catastrophically shed around five cubic km of rock into the sea, resulting in tsunamis up to 20 m high that may have taken more than 2,000 lives. Silver and colleagues, describe the results of a recent scientific cruise to the Ritter Island area, including submarine imagery of the landslide deposits. They also report up to a dozen debris field adjacent to other volcanoes in the Bismarck Volcanic Arc, providing testament to a history of past landslides in the region. The authors flag their study as the first step in estimating the regional risk from volcanogenic tsunamis, but also stress its importance in constraining computer simulations for future tsunami-generating volcano collapses. Tsunamis also result from giant sediments slides in the marine environment, often triggered by the ground shaking associated with large earthquakes. This is a hazard addressed by Koji Minoura ⁴⁸ of Tohoko University (Japan) and co-researchers, for the Marmara Sea immediately south of Istanbul. The authors report the discovery of a layer of sand, shell debris and charcoal, along the northern coast of the Marmara Sea, which they interpret as providing evidence of an 11th century tsunami that inundated the adjacent coastal plain to a distance of several hundred metres. Minoura and colleagues relate the event to coastal or submarine sediments slides caused by an earthquake on the North Anatolian Fault (NAF). They note that a similar mechanism may have generated tsunamis following an earthquake in 1912, and warn that the awaited NAF earthquake, below the eastern Marmara Sea could trigger a tsunami capable of impacting on the heavily populated and industrialised areas around its shores (figure 9).


Figure 9. During the August 17th 1999 Izmit earthquake, the town of Degirmendere on Izmit Bay (Marmara Sea) was inundated by tsunamis due to a 2 m sudden and permanent subsidence of the area. Minoura and colleagues (2005) report similar behaviour during historic earthquakes on the North Anatolian Fault, and warn of a similar threat during the forthcoming Istanbul quake.

Courtesy: NOAA National Geophysical Data Centre.

On a broader scale, Angus Best ⁷ of the UK’s National Oceanography Centre (Southampton), and colleagues, writing in EOS, draw attention to the growing threat to coastal zones - over the next 100 years - arising from submarine landslides and their potential tsunamigenic potential. The problem lies with seafloor sediments that contain abnormally high levels of methane gas, generated by the breakdown of organic detritus. Higher levels of methane in such sediments are being promoted by human activities, which increase the supply of sediments and organic matter through intensive farming, deforestation and the disposal of human waste at sea. The authors point out that shallow, methane-rich, sediments have the potential to destabilise seabed slopes quite close to shore, with collapse triggered by heavy rainfall, earthquakes or even human activities. Resulting tsunamis, Best and his co-authors note, pose a serious threat not only to coastal settlements but also to offshore drilling operations, pipelines, power and communications cables, and wind farms. Best and colleagues warn that a dramatic increase in seafloor methane, resulting in instability of the continental slopes, will occur when the current high levels of human activity-related organic matter reaching the sea floor, become buried to ~ 1m depth, probably within the next 100 years.

While extremely rare, our planet is certain to be struck by a comet or asteroid at some point in the future. More likely than not, the impact will occur in the ocean, where it will have the potential to trigger devastating tsunamis. This is a scenario addressed in Natural Hazards by Steven Chesley of NASA’s Jet Propulsion Laboratory in California, and Steven Ward ¹⁵, of the University of California at Santa Cruz. In a quantitative and economic assessment of the threat from asteroids less than 2 km across, they determine that averaged over time, around 180 people annually will be affected by impact-induced waves, with corresponding infrastructure loss estimated at US$18 per M/y. Chesley and Ward define a ‘mean generic scenario’, within which tsunamis arising due to the impact will affect 1.1 million people and destroy US$110 billion of infrastructure. A generic impact of a 400 m diameter asteroid (such as Asteroid Apophis, which has a 1 in 38,000 probability of striking the Earth in 2036), is estimated by the authors to be capable of generating tsunamis that would cause a US$400 billion loss.

Predicting volcanic eruptions

Unlike earthquakes, volcanic eruptions never occur without some or other warning sign. These arise because rising magma needs to break rock to get to the surface, generating swarms of small earthquakes, and also needs to make space for itself - causing the ground surface above to swell. While the latter can be detected using a range of technologies that include the Global Positioning System, the former can be monitored using a system of seismographs. In a paper published in Geophysical Research Letters, Chris Kilburn and Peter Sammonds ³⁶ of the BUHRC, show how measuring the daily numbers of earthquakes can provide an important eruption precursor at volcanoes reawakening after an extended period of quiescence. The authors concentrate on andesitic-dacitic volcanoes that have been in a quiescent state for a century or more. Such volcanoes are characterised by moderate to very large explosive eruptions, and have the potential to seriously threaten adjacent communities. Kilburn and Sammonds link accelerating seismic activity to fracturing in the crust surrounding a pressurised magma reservoir (figure 10), and show that the final acceleration to eruption develops over a period of 2 - 3 weeks. They point out, however, that as a week or longer is needed to identify an accelerating trend, seismic forecasts of eruptions after a long repose period are unlikely to provide more than a few days warning, which is probably not sufficient to permit effective mitigation plans, such as large-scale evacuation. The authors suggest that the reliable warning interval may be increased by means of future improvements to their model, or by incorporating additional geophysical or geochemical precursors, such as ground deformation and changes in gas chemistry.



Figure 10. (a) Precursory seismicity develops within a restricted volume (grey) between the magma reservoir (black) and the surface. During a Preparation Stage (P), seismicity is controlled by an increasing number of activated fractures. These eventually unite to form a connected fracture across the seismic volume (Unificatioin stage, U).

(b) Unification is characterised by a hyperbolic increase in peak seismic event rate (black circles), corresponding in

(c) to a linear decrease in the inverse-rate minima (black circles and broken line). The event rate data are for precursors to the 1995 eruption of Soufriere Hills volcano, Montserrat, measured from 00.00 h on November 1st 1995; lava emerged on November 15th.

Courtesy: Chris Kilburn.

Volcanic hazard and loss models

Due to the absence of any major loss due to volcanic activity, the development of volcanic hazard and loss models attracts little more than passing interest within the insurance industry. Models useful to the market continue to be constructed, however, and will - undoubtedly - attract considerable interest once a large volcanic loss is sustained or appears imminent. HRSR2006, included a volcanic risk ranking for the New Zealand city of Auckland. In the last 12 months, one of the papers co-authors, Christina McGill ^{43, 44} at Macquarrie University (Sydney), along with various colleagues, has published two new papers in the Journal of Volcanology & Geothermal Research, addressing volcanic hazard and loss as it relates to Auckland. The first presents a probabilistic tephra-fall (~ ash-fall) simulation for the city, which faces such a threat both from a number of distant, large-volume volcanic centres in the North Island, and also from smaller eruptions in the Auckland Volcanic Field (AVF) itself. The authors note that less than one percent of simulations from distant volcanoes reached the Auckland region, and that, at just 1.6 mm, the mean modelled tephra thickness for eruptions within the AVF, were very small. McGill and her colleagues also stress, however, that larger thicknesses are also possible, with simulated maximum thicknesses from within the AVF of 150 mm, and from distant volcanoes, 830 mm. Clearly, the realisation of such events, although very rare, would have the potential to cause serious damage in the Auckland region. In a second paper, McGill and co-authors present VolcaNZ, a volcanic loss model for Auckland based upon the predicted impact of the tephra fall addressed in the first paper. The model calculates both structural and non-structural damage to residential buildings and associated clean-up costs using Monte Carlo simulation. Losses from all simulations are plotted against calculated Average Recurrence Intervals (ARIs) to produce loss curves. Structural damage does not become apparent until ARIs of around 8,000 years, with a NZ$1 billion loss due to structural damage expected every 35,000 years and a NZ$26 billion loss every 1 million years. Total loss, including clean-up, is estimated at approximately NZ$210 for ARIs of 600 to 3,000 years, rising to NZ$10 billion at 100,000 years, and NZ$26 billion at 1 million years. McGill and her colleagues warn, however, that as the model only considers residential building damage and clean-up, these values greatly underestimate total loss from the next volcanic event to affect the Auckland region.

Large-scale hazard implications of volcanic super-eruptions

Volcanic super-eruptions occur every 50,000 years or more, and eject in excess of 1,000 cubic kilometres of debris. Much of this takes the form of volcanic ash that can cover continent-sized areas, as demonstrated by two super-eruptions at Yellowstone (Wyoming, US) in the last two million years, and another at Toba (Sumatra, Indonesia), around 74,000 years ago. Until recently, no explanation existed for how single volcanic eruptions - however big - could deposit ash across such huge areas. In Geophysical Research Letters, however, Peter Baines and Steve Sparks ⁴ of Bristol University provide a possible answer. They propose that supereruption ash columns are large enough to be affected by the Earth’s rotation via the Coriolis Force. As a consequence they form gigantic, spinning ash clouds that may reach diameters as great as 6,000 km in as little as a few days, enabling ash to be deposited at the surface across an entire continent. Furthermore, the authors predict that rates of radial expansion and spinning may be as fast as tens of metres a second, sufficient to carry larger particles to greater distances from the erupting volcano.

In addition to their ability to deposit ash across vast distances, volcanic super-eruptions are also implicated in triggering severe episodes of global cooling. Such action is believed to be a function of the loading of the stratosphere with sulphur aerosols, which are particularly effective at reducing the amount of solar radiation reaching the surface. In the journal Climate Dynamics, Gareth Jones ³⁴ of the UK Met Office Hadley Centre, and colleagues, describe the climate changes that ensue following a hypothetical super-eruption, simulated using an appropriately-adapted general circulation model (figure 11). The predicted changes to the global climate are dramatic to say the least, and include a near-surface temperature fall of as much as 10 degrees C globally for a few months, and a considerable deviation from normal temperatures lasting for several decades. While the cooling appears to be insufficient to trigger an ice age, at least for the parameters used in the model, it does increase the extent of snow and ice cover to a maximum of more than one third of the Earth’s surface.


Figure 11. Summary graphic of the major impacts to ocean and nearsurface climatology, due to a super-eruption (as simulated by Jones and others, 2005). Globally, temperatures fall by up to 15ºC below normal for a while. Bold numbers on the continents and oceans represent the maximum deviations from normal annual temperatures for that area as a whole.

Courtesy: Gareth Jones.

Recently restless volcanoes

On average, around 50 volcanoes are in eruption every year. Some, such as Etna in Sicily and Kilauea on Hawaii, are almost continually active, while others may be erupting for the first time following periods of quiescence lasting for decades, centuries or millennia. Papers addressing three such awakening volcanoes are highlighted here, with the intention of perhaps providing advance warning of forthcoming eruptions. In EOS, Alicia Garcia ²⁶ of the National Museum of Natural Sciences in Madrid, and co-workers, focus on monitoring the reawakening of Teide volcano on the Canary Island of Tenerife. With its summit located 7 km above the sea floor, Teide is the third largest volcano on the planet. In recent times, the volcano has erupted every century or so, the last time in 1909. At the end of 2003 and into 2004, the authors point out that Teide began to show signs of restlessness, including increased numbers of earthquakes, more fumarolic (steam vent) activity, and carbon dioxide gas releases. While it is perfectly possible that any forthcoming erruption will be effusive, in other words will involve the relatively quiet production of lava flows, it could also be explosive. Garcia and her colleagues note that the last explosive eruption here occurred around 1,500 years ago, and a similar event today could have the potential to affect more than 400,000 people.

Heading north to Iceland, one of the most volcanically-active countries on the planet, Heidi Sousalu ⁶⁴ of the Nordid Volcanological Center, and colleagues, report in the Journal of Volcanology & Geothermal Research, similar signs of a possible impending eruption at Katla in the south of the island. The authors report that the volcano has been restless since 1999, when abnormal seismic activity accompanied a glacial flood from the vicinity of the volcano, which may have been caused by a small sub-glacial eruption. In 2002, the level of seismic activity rose dramatically and earthquakes have occurred continuously ever since. Sousalu and her team interpret the seismic activity in terms of a shallow, growing dome of sticky ‘acid’ (silica-rich) magma that now threatens an explosive eruption. This, say the researchers, could pose a serious threat in this popular tourist region, particularly if the eruption occurs during the summer season.

In the Cascades Range of the western United States, Mount St. Helens remains active following its reawakening in October 2004, and there are also concerns that another volcano in the range may shortly erupt. This possibility is discussed by Dan Dzurisin ¹⁹ and colleagues at the United States Geological Survey, who report, in the Journal of Volcanology & Geothermal Research, recent signs of unrest at the Three Sisters volcanic centre in central Oregon. Inflation of the ground surface at South Sister volcano, which has not erupted for 2,000 years or more, has been observed since 1996, and is summarised by Dzurisin and his team. Their interpretation is that observed ground surface movements are the result of an inclined slab of basaltic magma intruding the volcano at a depth of around 6.5 km. Dzurisin and his co-workers estimate that the magma accumulation rate is about five million cubic metres a year, and suggest that such events are not that unusual, probably occurring every few decades to centuries. They feel that there is a low probability of a resulting eruption, but warn that - should one occur - the consequences could be severe. This is reflected in the fact that the USGS has updated its volcanic assessment for the Three Sisters region, notified the appropriate agencies and the public, and is working to put together an emergency coordination and communication plan.

«back to top«