Tsunami hazard and risk in the major ocean basins
On December 26th, 2004, the second largest earthquake of the last 100 years triggered a tsunami that led to the greatest natural catastrophe of modern times. Waves up to 30m high took more than 300,000 lives in seven countries; over 400,000 buildings were destroyed (figure 5) and Figure 5. Tsunami damage in the Thai resort of Phuket. The waves took approximately two hours to reach the country from the source of the earthquake off the west coast of Sumatra. Courtesy: Sebastian Bordage.

close to eight million people displaced, unemployed or impoverished. Losses are estimated at up to US$13.4 billion economic, and between US$2.5 and 4 billion insured. While numerous papers on the science of the Magnitude 9.15 earthquake and resulting tsunami can be expected in the coming months, the most worthwhile reports about the event itself are those published online in the US Earthquake Engineering Research Institute (EERI) newsletter. The first, written by Jose Borrero¹⁶ of the University of Southern California and co-workers, summarises the results of post –tsunami surveys in northern Sumatra and the coast of south-east India. Succeeding reports cover effects in mainland India and the Andaman-Nicobar Islands, an assessment of societal impacts and consequences, a post-tsunami survey in Sri Lanka, and an examination of the effects on Sri Lankan buildings and lifelines.

Just two months after the Asian tsunami, a special volume of Marine Geology was devoted to examination of the tsunami risk in the Pacific and Atlantic Basins and in Europe, focusing on case histories and evaluating risk. Of these, the risk is by far the highest in the Pacific Basin, where 400 tsunamis have taken 50,000 lives in the last century alone. V. K. Gusiakov²⁴ reviews tsunami generation potential in the Pacific, concentrating on the different tsunami-sourcing regions. The paper draws on data from the Historical Tsunami Database Project, initiated in 1995, and which – for the Pacific Basin – contains the most complete record of tsunami events from 47 BC to present. Gusiakov uses the data to determine, for each region of the Pacific, the tsunami efficiency, which is a measure of the efficiency with which earthquakes equal to or larger than magnitude 7 and shallower than 100 km, generate tsunamis. The author shows that this varies considerably, and is highest off the South American coast, off Indonesia, and the Philippines. The variation is attributed to differences in the amount of submarine sedimentation; those regions where sedimentation rates are highest providing the best opportunity for earthquakes to trigger submarine landslides, which considerably increases the efficiency of tsunami generation.

Japan has been hugely affected by tsunamis throughout the country’s history, and a recent paper in Eos highlights a new threat. Shu-Kun Hsu of Taiwan’s National Central University and Jean-Claude Sibuet²⁸ of Ifremer Centre de Brest (France) draw attention to the potential for a strong tsunami generated by a great earthquake off Japan’s west coast. The authors liken the tectonic situation here to that prevailing at the contact between the Indo-Australian Plate and Burma Plates, which ruptured to trigger the Indian Ocean tsunami. Hsu and Sibuet warn that the release of accumulating strain might well result in a major earthquake (the so-called Tokai quake) in the Nankai Subduction Zone, leading to a severe tsunami, which will strike Japan’s west coast and may also affect other states including mainland China and Korea.

While Japan has a highly developed tsunami warning system that can be expected to keep casualties to a minimum, even in the event of a large tsunami generated close by, the same cannot be said of the Caribbean. Also in the journal Eos, Nancy Grindlay²³ of the University of North Carolina, and colleagues, highlight concerns over tsunami potential in this small, but tectonically active sea basin. The authors review potential tsunami sources in the northern Caribbean, which threaten the Gulf and east coasts of the US as well as the Caribbean Islands. Six lethal tsunamis have been documented in the region since the arrival of Europeans in 1492, with waves up to 12 m high taking up to 4,000 lives in total, including 1,800 as recently as 1946. While all six tsunamis were related to submarine earthquakes, Grindlay and co-workers point to the threat of future tsunamis being generated by large submarine landslides, particularly south of Puerto Rico where huge slump features attest to past major tsunami-forming events.

While not as frequent, tsunamis also occur in the Atlantic Basin, which hosts around one percent of recorded tsunamis. The last major event occurred in 1929, when a magnitude 7.2 submarine earthquake shook the southern edge of the Grand Banks, 280 km south of Newfoundland (Canada) As I. V. Fine¹⁹ of the Heat and Mass Transfer Institute in Minsk (Belarus) and colleagues, report the quake triggered a large (~ 200 km3) submarine sediment slide that destroyed 12 submarine telegraph cables and, in turn, generated a tsunami (figure 6). Waves up to 13m high killed 28 people in Newfoundland and Nova Scotia, and smaller waves were observed as far away as Portugal and the Azores. Fine and co-authors present a model that successfully simulates the 1929 slide and tsunami, providing both travel times and wave amplitudes in reasonable agreement with observations and tide gauge records.

Figure 6. Estimated tsunami travel times (in hours) for waves generated by the 1929 Grand Banks earthquake-triggered submarine landslide. Wave heights reached 13m in Newfoundland and Nova Scotia and were recorded in the Azores and Portugal. Courtesy: Richard Thomson.

One of the most surprising and controversial tsunami papers published in recent years also deals with an Atlantic Basin event, this time affecting parts of the south-west UK. Although published in 2002, it was felt that inclusion here was justified following a recent high profile BBC television programme and associated publicity. Ted Bryant of the University of Wollongong (Australia) and Simon Haslett⁸ of Bath Spa College in the UK believe that both documentary and geological evidence points to a major tsunami striking the coasts of the Severn Estuary and Bristol Channel on January 30th, 1607. In a paper published in Archaeology in the Severn Estuary, they propose that waves up to 7.5 m high took more than 2000 lives while devastating a 570 km stretch of the coast of South Wales, Somerset, and North Devon. While the event was sufficiently memorable to be commemorated by plaques at a number of churches, and while it appears to have left behind deposits of sand and broken shells, the existence of the Severn Estuary tsunami is still far from proven. In particular, there is another body of opinion that argues that a major storm surge could have had a similar destructive and lethal outcome. On the other hand, a large, tsunami-producing earthquake on a known ancient – but active – fault off the coast of south-west Ireland is not entirely fanciful. The fault generated a magnitude 4.8 quake as recently as 1980.

Earthquake forecasting and prediction
Seismic science has been dominated over the past nine months by the Great (magnitude 9.3) Sumatra Earthquake of December 26th, 2004, which was addressed in the previous section on tsunami hazard and risk. The December 26th quake was, however, followed by another great (magnitude 8.7) earthquake three months later, and the two events were almost certainly linked. Just prior to the second quake, John McCloskey⁴³ and colleagues at the University of Ulster (Northern Ireland) published a short paper in Nature warning of the likelihood of another large quake in the region. McCloskey and co-workers had determined that the first quake had increased the stress on neighbouring faults, such that one could rupture imminently to produce another major quake. Just eleven days after the paper was published, the second quake shook Sumatra just 150 km south-east of the first. More than a thousand people were killed by building collapse but the tsunamis generated were less than 3m high and of only local extent. In a second paper published in Nature following the second quake, McCloskey⁴⁶ and co-workers update their warning, revealing that the subsequent event has brought closer to failure a neighbouring megathrust fault as well as expanding the area of increased stress on the major Sumatra Fault (figure 7). The authors warn that a major quake on the Sunda megathrust may result in up to 10m of slip. They also show that the probability of a magnitude 7.0 – 7.5 earthquake on the Sumatra Fault has not receded as a consequence of the March 2005 quake. Figure 7. Stress perturbations arising from the December 26th 2004 Sumatra-Andaman and the March 28th 2005 Simeulue-Nias earthquakes. Colour scale represents the Coulomb stress change resolved; directly onto the mega-thrust plane in the case of the Sunda Trench and onto planes parallel to the Sumatran Fault on mainland Sumatra. Also shown are the epicentres and approximate rupture areas of the two causative events. Courtesy: John McCloskey.

Effective earthquake prediction remains the sought-after grail for seismic science and last year’s review addressed a prediction made by Vladimir Keilis-Borok³² of UCLA’s Institute of Geophysics and Planetary Physics and his team. At the beginning of 2004, Keilis-Borok and his colleagues predicted that southern California would be struck by a Richter Magnitude 6.4 or larger earthquake by September 5th of that year. The prediction, which involved identifying clusters (‘chains’) of small earthquakes in the same region as precursors to a larger one, was not realised, so further work is clearly required. Improved understanding of the clustering properties of earthquakes do, however, hold some promise for improving predictions, and this possibility is revisited in a Nature paper by Matthew Gerstenberger²² of the United States Geological Survey (USGS) and co-authors. Gerstenberger and colleagues have developed a model that generates time-dependent maps (figure 8) showing the probability of Figure 8. Maps of California, showing the probability of exceeding MMI (Modified Mercalli Intensity) VI shaking over the next 24 hour period. The period starts at 14.07 Pacific Daylight Time on 28 July 2004. From left to right: Time-independent hazard based on the 1996 USGS hazard maps for California (SF = San Francisco; LA = Los Angeles); the time-dependent hazard that exceeds the background, including contributions from the 22 Dec 2004 San Simeon (SS) magnitude 6.5 event, a magnitude 4.3 quake four days earlier near Ventura (VB), a magnitude 3.8 quake near San Bernardino (FN) 30 minutes before the map was made, the magnitude 7.1 1999 Hector Mine quake, and the magnitude 6.9 Loma Prieta (LP) quake; the combination of these two contributions, representing the total forecast of the likelihood of ground shaking in the next 24 hours; the ratio of time-dependent contributions to the background. Courtesy: Matthew Gerstenberger.

strong seismic shaking anywhere in California within the next 24 hours. The model combines an existing time-independent earthquake occurrence model based upon fault data and historical earthquakes, with increasingly complex models that describe the local time-dependent earthquake clustering – for example due to aftershocks that follow a main shock. For most of the time, and across the majority of the state, a recent earthquake large enough to generate any appreciable hazard will not have occurred, so the model assumes that the background model provides an adequate measure of the hazard. Following a quake of magnitude three or greater, however, the additional hazard associated with possible aftershocks is added to the background model, until this time-dependent contributions decays below the background level. Given that the first hours to days after a main shock are the most hazardous, with respect to aftershocks capable of causing strong shaking, the new model provides a powerful new way of quantifying hazard at the critical time when emergency response activities are at their height.

Seismic hazard assessment
Whatever success the future brings in relation to improved prediction of earthquakes, the level of time-averaged seismic hazard in the world’s tectonically active regions will remain the same. Determining, evaluating and analysing this hazard will continue, therefore, to be a critical element in reducing death, injury and damage during future events. In an opinion paper published in Earthquake Spectra Norman Abrahamson of the University of California at Berkeley and Julian Bommer¹ of Imperial College London tackle the broad issue of probability and uncertainty in seismic hazard analysis. It is standard practice in logic-tree based, probabilistic seismic hazard analysis (PSHA) to use a mean hazard curve to determine ground motions for engineering design. Abrahamson and Bommer, however, argue vigorously against its use and propose both the discontinuation of the practice and its removal from anti-seismic regulations. They point out that hazard curves may be drawn for any confidence level so that for engineering applications a choice needs to be made about which curve to use. The hazard curve selected as the basis for design, they say, should be chosen on the basis of a confidence level that reflects the desired degree of confidence that the safety level implied by the selected annual frequency of exceedance (return period) is being achieved in light of the uncertainty in the estimation of the hazard. Concerns over the current PSHA methodology, based on logic trees, are also expressed by J – U. Klügel³⁵ of Kernkraftwerk Goesgen-Daeniken, in a paper published in Engineering Geology. In a study that looks at the application of the method to assessing earthquake hazards at Swiss nuclear power plants, Klügel shows that, due to the accumulation of uncertainties, the logic tree PSHA methodology in combination with the procedures of the Senior Seismic Hazard Analysis Committee (SSHAC) led to a significant over-estimation of the seismic hazard in areas of low seismic activity. As a consequence, the author reveals, further careful study will be required before any decision can be made about whether or not the Swiss nuclear power industry adopts – without modification - the recommended use of SSHAC procedures as a basis for the evaluation of seismic hazard at individual power plants.

Adopting a different approach to the appraisal of seismic hazard, M. Holschneider²⁷ of the University of Potsdam (Germany) and co-authors, propose a new methodology, which incorporates consideration of the spatial and temporal distribution of earthquakes and their effect on the maximum values of expected intensities. Holschneider and his colleagues show that the standard approach, which involves uniformly distributing epicentres inside each seismogenic zone, is likely to be biased towards lower expected intensity values. They also reveal that the estimated hazard at a fixed point is very sensitive to the assumed spatial distribution of epicentres and their estimators. Holschneider and co-workers present an alternative model that takes into account observed clustering of seismic events.

Earthquake loss estimation and damage assessment
With the prospect of damaging earthquakes striking a number of major cities in moderate/high income countries, including Tokyo, Istanbul, San Francisco, Los Angeles and Seattle, ways and means of determining and improving earthquake loss estimation continue to attract considerable interest, particularly from the insurance market. Earthquake loss estimation studies require predictions to be made of the proportion of a building class falling within discrete damage bands from a specified earthquake demand. In a paper published in the Bulletin of Earthquake Engineering, Helen Crowley¹² of the University of Pavia (Italy) and colleagues stress that earthquake effects on a building should be represented by a parameter that shows good correlation to damage and that accounts for the relationship between the frequency characteristics of the ground motion and the building’s fundamental period (the reciprocal of its natural frequency). Traditionally, damage assessment for loss estimation studies have been based on macroseismic intensity or peak ground acceleration. Although both parameters are directly related to building damage, however, they have their shortcomings. Crowley and her co-authors propose, in contrast, a loss estimation method that is based upon the displacement of a building during seismic shaking and its capacity to resist this. Vulnerability curves derived using the method are, according to the authors, realistic and consistent with field observations. They are also better at predicting damage between various classes of building than those curves generated by the HAZUS seismic loss estimation model.

In another earthquake loss estimation study, J. F. Bird⁵ of Imperial College London and co-workers compare loss estimation with observed damage in a zone of ground failure due to the 1999 Kocaeli (Turkey) quake. The results of the study, published in the Bulletin of Earthquake Engineering, look at the implications of ground failure (due to liquefaction) for loss estimation and discuss whether or not the mechanism is worth incorporating into an earthquake loss model that would normally be concerned only with ground shaking. They conclude that this would only be worthwhile if a detailed approach using in situ geotechnical data as well as adequate representation of building foundation is adopted. Modellers, Bird and her colleagues write, should make a choice between neglecting liquefaction and its effects entirely or incorporating them rigorously into a loss model on the basis of good quality data. Such a decision should be influenced by the size of the area of interest, with liquefaction having the potential to dominate damage distribution in small areas, where the geology is particularly suited to ground failure, but having a less significant influence across larger and geologically heterogeneous tracts.

Some form of loss estimation is invariably undertaken by or on behalf of any buyer of commercial property in a seismically-active region, and traditionally the outcome is the probable maximum loss (PML). This concept, however, is of little use to property investors as it has no place in standard financial analysis and covers too long a planning period. In a paper in Earthquake Spectra, Keith Porter⁴⁹ and fellow researchers of the California Institute of Technology propose an alternative, the probable frequent loss (PFL), which is defined as the mean loss resulting from shaking with 10 percent exceedance probability in five years, and which is approximately related to an expected annualised loss (EAL). Porter and his colleagues highlight a number of advantages that PFL has over PML, including a meaningful planning period, its applicability to financial analysis, and its straightforward estimation. In order to demonstrate the worth of PFL, the authors show that, with respect to a case study of a non-ductile reinforced-concrete, moment-frame building, probabilistic repair costs can be dominated by small, frequent, events, as opposed to large PML-level losses. In a related second paper in Earthquake Spectra, Keith Porter⁵⁰ and colleagues again address the issue of the sale and resale of commercial buildings in seismically-active regions, this time analysing the effect of seismic risk on lifetime property value. In this paper, the authors present a methodology designed to estimate the uncertain net asset value (NAV) of an investment opportunity, considering both market risk and seismic risk. Porter and co-workers show that uncertainty in market value appears greatly to exceed uncertainty in repair costs and that seismic risk produces a modest mean reduction in the NAV. Because market uncertainties relating to factors such as future rental rates and vacancy rates so dominate uncertainty in value, the authors determine that seismic risk (as reflected in earthquake repair costs) can effectively be ignored in relation to property investment decision making.

Continuing the theme of earthquake building losses, another paper in Earthquake Spectra by Robert Wesson⁶⁴ of the United States Geological Survey and co-authors, examines the distribution of insured losses to single-family housing following the 1994 Northridge earthquake. Through combining insurance loss data with both observed and estimated ground motions, Wesson and his colleagues construct empirical probabilistic loss curves as a function of ground motion. These have the potential better to estimate potential losses to single-family housing from scenario quakes, and also to make probabilistic estimates of annualised loss.

Regional studies of earthquake impact and hazard
In addition to the papers on Indonesian seismic hazard, introduced at the start of this section and in the section on tsunami, a number of excellent regional seismic studies have been published over the last twelve months and four are summarised here.

One of the largest earthquakes to strike North America in the last hundred years was the 2002 magnitude 7.9 Denali Fault earthquake, which occurred 330 km west of the Yukon-Alaska border. One of the most astonishing elements of the event was the length of ground rupture, which stretched over a distance of 341 km. The importance of the event, however, lies in its location, with the Denali Fault crossing the Trans-Alaskan pipeline. Although the consequences for the local environment and for energy supply could have been devastating, the pipeline held (figure 9) – a testimony to geological and seismological assessments undertaken prior to construction and to the engineering design. The Denali Fault earthquake forms the focus of a special issue of the Bulletin of the Seismological Society of America, edited by Charlotte Rowe⁵⁴ of the University of New Mexico and colleagues, which contains 27 papers on different aspects of the event.
Figure 9. Aerial photo of the Trans-Alaska Pipeline System (TAPS) line near the Denali fault, looking west. This is where the line is supported by rails on which it can move freely in the event of fault offset. Here the line has moved toward the west end of the rails. Alyeska Pipeline Service Company reported no breaks to the line and therefore no loss of oil. Note the transverse crack on the Richardson Highway in lower left. Out of view to the left (south) of this photo is a 2.5 m right-lateral offset of the highway where it crosses the fault. Photo credit: Rod Combellick, Alaska State, Division of Geophysical & Geological Surveys.

Greece is the most seismically active country in Europe and although earthquakes have been recorded here since classical times, the need for a comprehensive catalogue spanning the 20th century represented a gap in our knowledge. This has now been filled by Paul Burton⁹ of the University of East Anglia and colleagues. In a paper in Tectonophysics, Burton and co-workers present a magnitude homogeneous earthquake catalogue spanning the 20th century and covering both Greece and adjacent areas. The product contains 5,198 events, having been truncated at magnitude 4 so as to rule out large numbers of poorly reported smaller events. An estimate of both the moment magnitude (Mw) and the surface magnitude (Ms) is provided for all the quakes, the latter having important application to seismic hazard assessment. The catalogue complements an existing one that covers the period from 550 BC to 1899.

Staying in the same part of the world, a new study published in Soil Dynamics and Earthquake Engineering by M. Erdik¹⁸ and colleagues of Istanbul’s Kandilli Observatory and Earthquake Research Institute, evaluates earthquake hazard in the Marmara region to the south of the city. The probability of a magnitude 7+ quake in the region is on the order of 70 percent in the next three decades and concern for major damage and serious loss of life in Istanbul and its surroundings is high. Erdik and his co-researchers use historical seismicity, tectonic models and known slip rates on faults in the region to construct hazard maps of the Marmara region. These depict peak ground acceleration (PGA) for the Istanbul area of 0.4 to 1.0 g for a 10 percent probability of exceedance in 50 years. A two percent probability of exceedance in 50 years sees PGA values rise to between 0.6 and 1.5 g.

Switzerland, like much of western Europe, is generally regarded to have low seismic hazard, and certainly far less than Greece or Turkey. Damaging and lethal earthquakes are, however, possible, and the city of Basle was famously destroyed by a magnitude 6.0 – 6.5 quake in 1356. For this reason, a new seismic hazard assessment of the country has recently been compiled by Domenico Giardini²⁰ and many others at the Swiss Seismological Service. The report draws attention to the fact that earthquakes in Switzerland are a serious issue, with 28 events exceeding magnitude 5.5 in the last 700 years and twelve of these causing severe damage as a result of shaking of intensity VIII or higher. Of particular concern is the seismic hazard as it relates to critical installations such as large dams, nuclear power facilities and chemical plants. The assessment, which is based on an earthquake catalogue covering Switzerland and adjacent regions for the period 1300 – 2003, includes hazard curves for selected Swiss cities and hazard maps for return periods of 100, 475, 2,500 and 10,000 years. The highest hazard is identified in the Basel region and in the Valais/Wallis canton in the south-west of the country.

Landslides in the terrestrial and marine environment
Landslides are ubiquitous, occurring both on land and beneath the ocean in response to a large range of destabilising and triggering factors. On land, severe precipitation and earthquakes are two of the most common triggers of large scale mass movement. Hong Kong, for example, is particularly susceptible to the former, combining – as it does – dense and rapid urban development in hilly terrain with heavy rainfall associated with frequent tropical storms and typhoons. In a paper in Engineering Geology, H. Chen and C. F. Lee¹¹ evaluate a number of case histories of disastrous landslides in Hong Kong, and possible ways of preventing such events in the future. More than 90 percent of Hong Kong’s landslides occur as a direct result of major rainstorms, with a significant event – leading to numerous landslides within a 24 h period – occurs, on average, every two years. The authors summarise the ground conditions relating to landslide formation, the range of landslide types and the mechanisms by which rainfall-induced landslides are triggered. In regard to prevention, they focus – in particular – on how computer finite element modelling (FEM) can be used to simulate the extent of slope failure and on landslide prevention strategies adopted in Hong Kong.

The strong ground motion associated with large earthquakes is particularly effective at triggering landslides, which – for example - are held responsible for more than half of all earthquake-related deaths in Japan. The link between earthquakes and landslides was once again clearly demonstrated when a magnitude 6.6 quake struck Japan’s Niigata prefecture in October 2004, killing 40, injuring more than 3,000 and requiring the evacuation of more than 100,000 people. The earthquake also triggered thousands of damaging landslides, which are evaluated in a paper in Eos by Roy Sidle⁵⁶ and colleagues at Kyoto University. The authors determine that the landslides caused considerable damage to roads, farmland, residential areas and water bodies. They determine that the unusual level of landslide damage (for a moderate earthquake) was related to a number of factors, including the steep terrain, the soft nature of the regolith (unconsolidated sand and silt), the proximity and shallow depth of the epicentre, high rainfall, and land use. They also point out that the occurrence of three separate moderate quakes within 40 minutes was a critical factor in the degree of landsliding, with early shocks destabilising slopes that were brought down in later ones.

Primarily as a result of their potential tsunami threat, there is considerable interest in landslides occurring in the marine environment, and the issue is addressed in a series of excellent papers on Continental Slope Stability in the journal Marine Geology. One of these, by V. Hühnerbach²⁹ and co-workers, provides a comprehensive inventory of submarine landslides in the North Atlantic, a total of around 260 in all. The authors recognise that while the eastern North Atlantic is characterised by a few large slope failures, abundant small landslides typify the western North Atlantic. They also recognise a peak in landslide headwall location at depths of between 1000 and 3000 m on both sides of the ocean, suggesting failure in a specific sediment rheology (a weak layer or layers). This depth range suggests that the breakdown within the sediments of the solid methane + water deposits known as gas hydrates may play an important role in the triggering of many of the landslides. [It is worthy of note that rising sea temperatures associated with global warming favour gas hydrate breakdown and may increase the probability of future submarine landslides]. Landslides in the marine environment are also examined by Barbara Keating of the University of Hawaii and Bill McGuire³¹ of UCL’s Benfield UCL Hazard Research Centre. Here, however, the focus is on giant landslides formed by the collapse of island volcanoes. In a review paper in Advances in Geophysics the authors review those factors that affect instability and collapse at volcanic ocean islands that make up the Canary, Hawaii and other volcanic archipelagos. In particular, they focus on climate change as a trigger of volcanic island collapse, in relation to changes in sea level and precipitation. The authors concede that although some evidence does exist to link changing environmental parameters to the incidence of collapse, the precise mechanism remains to be pinned down. Inevitably, Keating and McGuire also address the potential tsunami risk arising from major volcanic landslides in the marine environment, and here again note that there is as yet no consensus on the scale of the risk. For example, tsunami run-up heights on the east coast of the US, resulting from a future collapse of the Cumbre Vieja volcano on the Canary Island of La Palma, range from just 3m up to 20m.

New loss models for volcanic eruptions
In the absence of a large insured volcanic loss, there has been little emphasis on the development of catastrophe models that address volcanic hazards. This is beginning to change, however, with three new papers that focus on ash-fall; an unspectacular volcanic hazard, but one that can have a multitude of destructive and damaging effects, ranging from building collapse to human health problems. All concentrate on the threat in New Zealand’s North Island, an area of particularly high volcanic activity. In the Journal of Volcanology & Geothermal Research, Tony Hurst and Warwick Smith³⁰ of New Zealand’s Institute of Geological and Nuclear Sciences present a methodology developed using the Monte Carlo statistical technique, which enables the probability of a particular thickness of ash to be quantified for any location. The methodology is based upon an established program (ASHFALL) designed to model individual eruptions, where the likelihood thickness of ash deposited at a particular site depends upon the volcano’s location, the volume of material erupted, the height of the eruption column, the size of the ash, and the prevailing wind direction. Using the Monte Carlo procedure, variations in erupted volume and wind conditions are simulated by analysing repeat eruptions, each time allowing the parameters to vary according to known or assumed distributions. The method is able to handle the effects of multiple volcanic sources, each with its own characteristics. Ash thicknesses from all sources are accumulated to determine the combined ash-fall hazard, which is expressed as the frequency with which any given depth of ash is likely to be deposited at selected sites. These numbers are expressed as annual probabilities or mean return periods.

Zeroing in on the Auckland region of New Zealand’s North Island, Christina Magill and Russell Blong40, 41 at Macquarrie University (Sydney) have developed a volcanic risk ranking for the city that incorporates all likely volcanic hazards, including falling ash. The results are published in two papers in Bulletin of Volcanology, the first dealing with methodology and investigation of the potential hazards, and the second with hazard consequences and risk calculation. With a population of more than 1.2 million, Auckland is New Zealand’s most populous region (figure 10). It is also located within a so-called monogenetic volcanic field (the Auckland Volcanic Field or AVF), which contains 49 small-volume basaltic volcanoes, and within which a volcanic eruption could occur with little notice. In addition, the region is threatened by eruptions from five other more distant volcanic centres – including the Tongariro centre that includes Ruapehu - some of which have hosted major to cataclysmic eruptions in the past. Magill and Blong calculate individually the risk of each hazard from each source. In order to take account of multiple events and outcomes, they also multiply the results by the relative probability of the event occurring, and the relative importance of the outcome, ending up with a total risk calculation that enables ranking to be carried out. In terms of building damage, ash-fall was determined to constitute the highest hazard by an order of magnitude, while in terms of loss of life, the explosive phenomenon known as base surge topped the risk table, again by an order of magnitude over other hazardous volcanic phenomena. In terms of the most threatening volcanic centre, an eruption of the Taranaki (Egmont) volcano, 280km from Auckland’s centre, was determined to pose the biggest risk in terms of building damage, while in terms of human loss, an eruption on land (as opposed to offshore) in the AVF presented the greatest risk.

Figure 10. With a population of 1.2 million, Auckland is New Zealand’s most populous regions. It is under threat, both from small, short-duration eruptions within the Auckland Volcanic Field (AVF) - see volcanic cone in foreground of top image - and from larger, explosive eruptions at five other volcanic centres (bottom).