Eos, Vol. 78, No. 38, September 23, 1997

VOLUME 78 NUMBER 38

SEPTEMBER 23,1997

PAGES 401-416

Eruption of Soufriere Hills Volcano in Montserrat Continues PAGES 4 0 1 , 4 0 8 - 4 0 9

Simon Young, Stephen Sparks, Richard Robertson, Lloyd Lynch, and Will Aspinall

On July 18 ,1995 , the Soufriere Hills vol­cano in Montserrat erupted for the first time in recorded history. The eruption began with intense fumarolic venting and phreatic explo­sions following 3 years of elevated seismicity. An andesite lava dome emerged on Novem­ber 15, 1995, and continued to grow, with sev­eral periods of copious pyroclastic flow generation and an explosive eruption on Sep­tember 17 ,1996. The largest pyroclastic flows to date (on June 2 5 , 1 9 9 7 ) and a period of vul-canian explosions with fountain collapse (during early August 1997) indicate contin­ued escalation of the eruption.

The population in the southern part of the island was evacuated in April 1996 and vil­lages and farmland on all flanks of the vol­cano have been devastated. Many of Montserrat's people lost their homes and live­lihoods.

Simon Young, British Geological Survey, Edin­burgh, Scotland; Stephen Sparks, Bristol Univer­sity, Bristol BS8 1RJ, England; Richard Robertson and Lloyd Lynch, Seismic Research Unit, University of the West Indies, Trinidad; and Will Aspinall, Aspinall and Associates, Beaconsfield, Bucks, HP9 1JQ England

The eruption is being monitored by the new Montserrat Volcano Observatory (MVO), which advises the public and the gov­ernments of Montserrat and the United King-

Fig. 1. a) Pyroclastic flow from late July 1996 entering the sea at the mouth of the Tar River valley, b) Photo­graph of the Soufriere Hills volcano from the east on Oc­tobers, 1996. The new dome in English's Crater has a large crater open to the east within it due to the collapse and explosive eruption on September 17, 1996. A new dome appears on the crater floor. The Tar River valley is inundated by pyroclastic flow deposits and a new delta of these deposits progrades into the sea. Original color im­ages appear at the back of this volume.

dom on the volcano's activity and the haz­ards it poses. The proximity of the island's main service, agriculture, and housing infra­structure to the volcano calls for highly re­sponsive volcano emergency management to sustain living conditions on the island. In late July, pyroclastic flows devastated parts of the capital, Plymouth.

Innovative applications of new technol­ogy are being used to monitor the eruption and document dome growth, ground defor­mation, seismic activity, gas release, and pe­trology. New technologies are also being used to observe pyroclastic flows entering the sea (Figure l a ) , measure pulsations in dome growth and volume of new deposits (Figure l b ) , and observe remarkable alterna-

This page may be freely copied.

E o s , T R A N S A C T I O N S , A M E R I C A N G E O P H Y S I C A L U N I O N

Eos, Vol. 78, No. 38, September 23, 1997

4 k m 1 1

Fig. 2. Topographic map of Montserrat show­ing main locations, position of active dome in English's Crater and risk zones. This map and its predecessors have been used byMVO to provide a basis for response to the crisis. Con­tour interval is 500 ft. [This map was super­seded after events of June 25, 199 7J.

tions of intrusion and extrusion during the eruption. Study of the Soufriere Hills eruption is also spawning new ideas on the processes that control lava dome activity and on the ap­proaches to evaluating risk on an island where life continues c lose to a dangerous volcano.

H i s t o r y o f t h e V o l c a n o

The Soufriere Hills volcano is a medium-sized andesite dome complex that lies at the southern end of the island of Montserrat (Fig­ure 2 ) in the northern sector of the Lesser An­tilles island arc. The volcano is composed of several andesite lava domes flanked by aprons of pyroclastic flow deposits formed by predominantly nonexplosive collapse of the growing domes [Rea, 1974] . Radiocar­bon dates indicate that the volcano's history extends back at least 26,000 years with many of the deposits yielding dates of 26,000 to 16,000 years ago [ Wadge and Isaacs, 1988] .

The volcano's summit crater, English's Crater, has a diameter of about 1 km. The cra­ter is open to the east into the Tar River val­ley, and this is interpreted as a sector collapse feature. Prior to the eruption, Eng­lish's Crater was partly filled in with the young Castle Peak lava dome. The associ­ated pyroclastic flow deposits were exposed in the Tar River valley. Radiocarbon ages from these deposits average between 350 and 400 years, indicating that this eruption likely occurred just before colonization in 1642 [Youngetal., 1996] .

Seismic crises occurred on Montserrat in 1897-1898, 1933 -1937 ,1966 -1967 , and 1985. Although these crises did not result in erup­tion, the 1933 and 1966 crises involved in­creased thermal and gas activity at fumaroles. The 1966-1967 earthquake hypo-centers were located in a west-northwest-trending belt beneath the Soufriere Hills at depths of less than 15 km [Shepherd et al., 1971] . Hypocenter depths decreased during the crisis, becoming as shallow as 2.8 km in July and August of 1966. The crisis was ac­companied by increased hot spring activity and changes in ground tilt.

The MVO was established soon after the onset of the current eruption. The observa­tory is as an entity of the government of Mont­serrat and an extension of the Seismic Research Unit (SRU) of the University of the West Indies, Trinidad, which monitors earth­quake and volcanic activity in the Common­wealth island territories of the eastern Caribbean.

M o n i t o r i n g Risks, H a z a r d s

The MVO monitors the volcano and ad­vises the civil authorities on how to protect the public. The most hazardous zones around the volcano encompass about 40% of the island and include Plymouth, the capital town, and important parts of the country's commercia l base (Figure 2 ) . The situation poses particular difficulties because much of the citizenry and government wish to sustain Montserrat as a community despite the prox­imity of the now active volcano. A system of

zoning and alert levels allows for a flexible re­sponse and the maintenance of essential serv­ices, while alternative infrastructure is developed in the north. Hazards maps were developed for different eruption scenarios and have been amalgamated into a risk map (Figure 2 ) that accounts for community vul­nerability as well as scientific criteria. Escala­tion in activity during August has resulted in expansion of areas of vulnerability and a new risk map is now being developed by the authorities.

For the first time ever in a volcanic crisis, MVO is using a formal methodology to evalu­ate the status of the alert level and guide the authorities in planning during the crisis. Mem­bers of the MVO staff are asked to give prob­abilities of various outcomes; for example, the probability of an explosive eruption in the next 3 months. Participants' opinions are weighted and combined according to experi­ence , knowledge, and judgment. The applica­tion on Montserrat is promising, but it is regarded as an experimental project that will need a postmortem when the eruption is over. For more information, see the Montser­rat web page (http://www.geo.mtu.edu/ volcano es/west. i n d i es/so u f r i e re/g o vt/).

M o n i t o r i n g M e t h o d s

Before the eruption, seismic activity was monitored by a short-period station located about 3 km from the summit and stations on neighboring islands that constitute the north­ern part of the regional network run by SRU in Trinidad. One triaxial and five vertical seis-

E

E 3 O

>

110

1 0 0

9 0

8 0

7 0

6 0

5 0

4 0

3 0

2 0

10

Rapid d o m e growth

First big col lapse to southwest £

Pumiceous pyroclastic flows to the sea

1 7 / 1 8 S e p t e m b e r explosion ' Total erupted

, volume

Rapid d o m e growth a n d collapses

First pyroclastic flows to the s e a

0 1-Nov-95

M • 30-Jan-96 29-Apr-96 28-Jul -96

Date 26 -Oct -96 24-Jan-97 24-Apr-97

Fig. 3. Plots of the total volume of magma erupted and the volume of the dome as dense rock equivalent, assuming a density of2650 kg/m3 for nonvesicular magma. The dome volume de­creases during collapses that generate pyroclastic flows (for example, the collapse of September 17, 1996).

This page may be freely copied.

Eos, Vol. 78, No. 38, September 23, 1997

I I I | M I | M I I | I I I I | I I I I | I I I I | I I I I I | I I I |

L Log m a x h y b r i d a m p l i t u d e

160

120

2 4

I I I I I I I I I I I I I I I I I j .

N u m b e r o f h y b r i d e a r t h q u a k e s pe r 4 h r s

N u m b e r o f d o m e r o c k f a l l s pe r 4 h r s

N u m b e r o f Ga lway ' s lans l ides p e r 4 h r s

ilm .a. i ill Hi BUI HI I I I iHfll I NOV 0 5 10 15 2 0 2 5 DEC 0 5 10 15 2 0 2 5 START DAY IS OCT 2 6 1 9 9 6 GMT - JULIAN DAY 3 0 0

Fig. 4. Plot of seismic data at Soufriere Hills Volcano from October to December 1996. The upper plot shows the relative amplitude of the hybrid earthquakes plotted against time, illustrating their increasing size and the sharp transitions between seismic and aseismic periods. The three lower plots demonstrate some anticorrelation of hybrid earthquake activity with dome rockfalls and good correlation between hybrid earthquakes and landslides from Galway's Wall, especially until mid-December. Rockfalls and landslides are identified and distinguished on seismic criteria.

mometers were established on Montserrat by June 1995. Shortly after the eruption began, the Volcano Disaster Assistance Program team of the USGS expanded the system to 10 stations and refined the data acquisition sys­tem to improve the processing capability. This system has been the backbone of the monitoring effort and has allowed the rapid real-time detection of a wide range of vol­cano-seismic phenomena to be monitored.

A broadband system was installed in Octo­ber 1996 and consists of five three-compo­nent seismographs with 24-bit digital telemetry and data acquisition. This occa­sion marks the first time a permanent broad­band array of this type has been installed on an active volcano. Data from the array show a rich diversity of seismic phenomena, in­cluding volcano-tectonic earthquakes with distinct S and F w a v e phases, long-period earthquakes with nearly monochromat ic spectra mainly below 2 Hz and a peak around 1 Hz, hybrid events with mixed long-and short-period components and no recog­nizable S phases, periods of repetitive hybrid events with remarkably stable periodicities for hours at a time, banded tremor, and rock fall signals.

Ground deformation is studied princi­pally by electronic distance meter (EDM) and Global Positioning System (GPS) surveys

supplemented by measurements across ac­tive fractures and electronic and dry tilt meas­urements. GPS surveys include both rapid-static and long occupat ion techniques, and two continuous geodetic GPS stations te­lemeter data to the observatory in real time.

The deformation studies show that the most marked movements occurred in the early stages when magma was approaching the surface. Much of the deformation has been confined to the immediate vicinity of the dome. Tilt meters on the crater rim have allowed the recognition of cycl ic patterns of dome deformation with intervals typically of 6 to 20 hours. A well-studied EDM line to a point c lose to the dome growth site showed increased rates of deformation in the few days prior to increases in surface dome growth rate. The station that obtained these data was buried by the expanding lava dome by the end of 1996. Other EDM lines have shown mild deformation cycles while a sig­nificant switch in deformation pattern on the eastern flank of the volcano was measured by GPS in June 1997. These deformation data are not compatible with the existence of a large swelling magma chamber deforming an elastic crust.

An innovative technique for measuring the volume of the dome uses a helicopter, range-finding binoculars, and GPS and is sup­

plemented with still photographs from fixed points. A kinematic survey involves creating a topographic map of the dome by tracking precise positions of the helicopter with GPS and the distance and bearing of points on the dome from the helicopter using the laser range-finding binoculars. About 30 to 50 points can

be col lected in 1 to 2 hours and the data are processed in a day. Errors on volumes calcu­lated in this way are estimated to be less than 10%. Much of the data in Figure 3 was ob­tained through this technique. The frequency of observations was largely determined by weather conditions. Tests of longer-range bin­oculars and automated dome mapping soft­ware are underway at MVO, which should enable ground-based surveys to be com­pleted in a matter of hours.

SO2 fluxes are monitored by correlation spectrometer observations of the plume. The fluxes are moderate compared to many other volcanoes with values usually between 50 and 500 tons per day (average April 1996 to June 1997 is 350 tons per day). For example, the average at Mt. Unzen in Japan was 180 tons per day from July 1991 to January 1992 [Hirabayashi et al., 1995] . Occasional values exceed 1000 tons/day. Higher fluxes are asso­ciated with periods of enhanced dome growth and with pyroclastic flow emplace­ment.

MVO also samples fumaroles for hy-drothermal geochemistry and heat flow; tests rain, spring, and surface water for acidity, dis­solved solids, and chemistry; and studies ash mineralogy, grain size, and gas concentra­tions for human health purposes.

C h r o n o l o g y

Seismicity above background levels can be traced back to January 1992. Eighteen earthquake swarms of A-type events were re­corded prior to the start of the eruption, with the largest swarms occurring in June, Novem­ber, and December of 1994. When surface ac­tivity c o m m e n c e d on July 18 ,1995 , phreatic eruptions were accompanied by two clusters of volcano-tectonic earthquakes, one be­neath the volcano and extending out in a northeast direction beyond the coast and the other focused beneath St George's Hill to the northwest of English's Crater (Figure 2 ) . Hy-pocenter locations were estimated from near the surface to depths exceeding 6 km. The first phreatic eruption to produce a cold, ash surge cloud—which engulfed Plymouth—oc­curred on August 21 , 1995, and resulted in the first major evacuation.

Magma reached the surface sometime be­tween November 14 and 16 ,1995 , when an andesite lava dome began to extrude. The volume of the dome and the amount of magma extruded have been carefully moni­tored (Figure 3 ) . The size of the dome and the extrusion rate progressively increased over the early months, and the first substan­tial pyroclastic flow due to gravitational col-

This page may be freely copied.

Eos, Vol. 78, No. 38, September 23, 1997

lapse from the dome occurred on April 3, 1996. Growth spurts in late July and mid-Au­gust exceeded 500,000 m 3 /day, with a back­ground rate of around 150,000 m 3 /day. Large pyroclastic flows reached the sea in five sepa­rate collapse events between late July and early September, forming a large delta (Fig­ure l b ) . Pyroclastic flows were observed en­tering the sea (Figure l a ) . Seismicity during this period was almost entirely shallow.

On September 17 ,1996, a major dome col­lapse generated pyroclastic flows for nearly 9 hours. Many of these flows reached the sea, and about 12 x 1 0 6 m 3 of material was removed from the dome, more than one third of its total volume. After a brief quiescence , explosive activity began at 1142 LT and con­tinued for 48 minutes. Blocks up to 1.2 m in diameter (2.4 metric tons) were ejected in a northeast direction to the village of Long Ground, which is 2 km from the dome, and caused much damage. The eruption column ascended - 1 4 km and approximately 8 x 1 0 6

m 3 of magma was ejected, distributing pum­ice and lithic tephra over southern Montserrat.

A new dome appeared in the explosion scar on October 1,1996. The new dome grew rapidly (Figure 3) and growth was accompa­nied by high SO2 production. The volcano then entered a period of activity during which intense swarms of shallow hybrid earthquakes lasting several hours to a few days alternated with almost aseismic periods (Figure 4 ) . Overall extrusion rates slowed dur­ing this period (Figure 3 ) . In late October the southwest wall of English's Crater (the Gal-way's Wall) showed signs of instability with the appearance of large cracks and many small avalanches. The wall slowly disinte­grated during November and early Decem­ber with a c lear correlation between earthquake swarms and rock avalanches (Figure 4) . Concerns were raised over the possi­bility of tsunami generation should a cata­strophic failure of the wall occur. An energetic collapse of a sector of the dome on December 19 following renewed dome growth produced pumiceous pyroclastic flows and marked the end of this phase of ac­tivity.

Growth of the dome accelerated through late December into January 1997 with extru­sion rates again reaching 500,000 m 3 /day. This enhanced extrusion rate led to relatively short-lived collapse episodes and pyroclastic flow generation in January. Further alterna­tions of hybrid earthquake swarms and aseis­mic periods and further disintegration of the Galway's Wall have occurred, allowing pyro­clastic flows to overtop the wall. In late March of 1997 the volcano once again be­c a m e very active: the southern active part of the dome collapsed, producing major pyro­clastic flows to the southwest for the first time, extending 3.6 km, and heavy ashfall over the island. In mid-May, pyroclastic flows overtopped the northern crater wall.

Prolonged high growth rates accompa­nied by high SO2 flux, swarms of hybrid earth­quakes, and strong cycl ic inflation and deflation of the dome measured by elec­tronic tiltmeters led to the largest pyroclastic flows to date on June 25, 1997. These flows descended the northern flank of the volcano, devastating several villages close to Montser-rat's airport. Ash cloud surges damaged a wide area peripheral to the flows. Nine con­firmed and 11 suspected fatalities occurred within the evacuated zone, and 200 houses in seven villages were destroyed. This event changed the dome morphology, channelling later pyroclastic flows toward the west. By the end of June, the first houses in Plymouth had been set alight by flows. Strongly cycl ic activity continued at an elevated level throughout July, and culminated in a series of vulcanian explosions in early August that produced pyroclastic surges to 1 km and fountain collapse pumice flows to 4 km as gas-charged magma reached the dome.

P e t r o l o g y a n d G e o c h e m i s t r y

The dome lava is a porphyritic andesite (58 -60% Si02) with limited compositional variation throughout the eruption. The phe-nocryst assemblage of hornblende, plagio-clase, orthopyroxene, titanomagnetite, and minor quartz can be reproduced underwater-saturated conditions in experimental studies at between 820° and 850°C at pressures of 125 MPa. Melt inclusions in quartz and plagio-clase indicate that before the eruption, water content was 4 to 5% and SO2 contents were low (<150 ppm). The lava contains abundant mafic inclusions with quench textures, indi­cating invasions of mafic magma into the sys­tem during its evolution. Resorbtion features and reversely zoned plagioclase indicate heating events, which are interpreted as the influence of mafic magma influxes.

The lava shows the petrological effects of extensive degassing and decompression dur­ing ascent. Hornblende phenocrysts show de­compression reaction rims [Rutherford and Hill, 1993] . The rims decreased in thickness over time as the extrusion rate and inferred ascent rate increased. Application of the re­sults of Rutherford and Hill [ 1993] imply as­cent rates ranging from only a few meters per day at the beginning of the eruption to 1000 m/day when extrusion rates were high in the summer of 1996. The lava often erupts in an advanced state of crystallization with py­roxene replacing hornblende in the ground-mass. The lava typically contains less than 30% melt and some samples contain less than 10% melt. The advanced state of crystal­lization is attributed to extensive degassing of magma during slow ascent. Magma erupted during the explosive events has a less crystal-rich ground mass, indicative of more rapid as­cent and reduced degassing.

T h e Or ig in o f P r e s s u r i z a t i o n

As in other lava dome eruptions—for ex­ample, Mt. Unzen [Nakada, 1993]—the Sou­friere Hills eruption shows abundant evidence of anomalously high pressures de­veloping in the deeper parts of the dome and the uppermost parts of the conduit. Evidence for high pressures include the explosive erup­tion and the occur rence of shallow, long-pe­riod and hybrid earthquakes, which suggest that pressurized fluid is moving along frac­tures [Chouet, 1996] , jets of ash and gas emerging from large fractures across the dome surface, explosive disintegration of blocks in pyroclastic flows, observations of tuffisite veins in ejecta and forceful intrusions at shallow levels beneath the dome as evi­denced by the hybrid earthquake swarms, and ground deformation.

Nonlinear pressure gradients 5 to 6 orders of magnitude [Sparks, 1997] due to near-sur­face changes in viscosity during degassing and increases in pressure due to groundmass crystallization estimated to be ~1 MPa per 1% crystallization [Stixetal., 1997; Sparks, 1997] both suggest generation of high pressures.

The insights gained from the monitoring work and research at Montserrat promise to advance understanding of the processes in­volved in these eruptions, in particular on the relationships between degassing, shallow seismicity, associated intrusions, ground de­formation, dome growth rates, pyroclastic flow generation, and explosive activity.

A c k n o w l e d g m e n t s

The results reported in the article are the work of a large number of colleagues who have staffed MVO over the last 2 years. Their dedication and contributions have been out­standing. The dedication and skill of the tech­nical and administrative staff of the Montserrat Volcano Observatory, Seismic Re­search Unit of the University of the West In­dies, and British Geological Survey have been outstanding. Funding for the monitor­ing effort is provided by the Department for International Development (UK) and the gov­ernment of Montserrat. Some aspects of the research were supported by the National En­vironment Research Council (UK), the U.S. National Sc ience Foundation, and the Leverhulme Trust. The U.S. Geological Sur­vey Volcano Disaster Assistance Program team, supported by U.S. AID, also contrib­uted to the monitoring work. Published with permission of the Director, British Geological Survey. MVO Contribution No. 2.

R e f e r e n c e s Chouet, B. , Long-period volcano seismicity: Its

source and use in eruption forecasting, Na­ture, 380, 3 0 9 - 3 1 5 , 1996.

Hirabayashi, J . , T. Ohba, K. Nogami, and M. Yoshida, Discharge rate of SO2 from Unzen volcano, Kyushu, Japan, Geophys. Res. Lett., 22, 1709-1712 , 1995.

This page may be freely copied.

Eos, Vol. 78, No. 38, September 23, 1997

Nakada, S., Lava domes and pyroclastic flows of the 1991-1992 eruption at Mount Unzen volcano, in Unzen Volcano: The 1990-1992 Eruption, edited by T. Yanagi, H. Okada and T. Ohta, pp. 5 5 - 6 6 , Nishinippon and Kyusku University Press, 1993.

Rea, W. J . , The volcanic geology and petrol­ogy of Montserrat, West Indies, J . Geol. Soc. London, 130, 3 4 1 - 3 6 6 , 1974.

Rutherford, M. J . , and P. M. Hill, Magma as­cent rates from amphibole breakdown: an experimental study applied to the 1 9 8 0 -1986 Mount St Helens eruptions, J . Geo-phys. Res., 98, 19 ,667-19 ,685 , 1993.

Shepherd, J . B . , J . F. Tomblin, and D. A. Woo, Volcano-seismic crisis on Montserrat, West Indies, 1966-67 , Bull. VolcanoL, 35, 143-163 , 1971.

Sparks, R. S. J . , Causes and consequences of pressurisation in lava dome eruptions, Earth Planet. Sci. Lett., in press, 1997.

Stix, J . , C. Torres, M. Narvaez, G. P. Cortes, J . Raigosa, M. Gomez, and R. Castonguay, A model of vulcanian eruptions at Galeras Volcano, Colombia, J . VolcanoL Geotherm. Res., 77, 2 8 5 - 3 0 4 , 1997.

Wadge, G., and M. C. Isaacs, Mapping the hazards from the Soufriere Hills volcano,

Montserrat, West Indies using an image proc­essor, J . Geol. Soc. London, 145, 5 4 1 - 5 5 1 , 1988.

Young, S.R., R. P. Hoblitt, A. L. Smith, J . D. Devine, G. Wadge, and J . B . Shepherd, Dat­ing of explosive volcanic events associa ted with dome growth at Soufriere Hills vol­cano , Montserrat, Proceedings of the Spe­cial Symposium on Volcanism in Montserrat, S e c o n d Caribbean Conference on Natural Hazards and Hazard Manage­ment, Kingston, J ama ica , Oc tober 1996, MVO Open File Report 96/22, 1996.

1920s Prediction Reveals Some Pitfalls of Earthquake Forecasting PAGES 4 0 1 , 4 1 0 , 4 1 2

Carl-Henry Geschwind

A number of seismologists are pursuing the possibility of making scientifically cred­ible earthquake forecasts or predictions. Oth­ers, however, are concerned about how the public might react to such forecasts [Na­tional Research Council, 1978; Geller, 1997] . In the 1920s, the president of the Seismologi-cal Society of America publicly predicted that southern California would suffer a severe earthquake within the next ten years. Only passing mention has been given to this epi­sode in historical discussions of earthquake seismology [Richter, 1958, p. 388; Meltsner, 1979, pp. 343, 3 4 6 - 3 4 7 ] . Three related points are illustrated by this episode which may have some application to other instances of earthquake forecasting. Even credible predic­tions can evoke opposition from groups who see their e conomic interests threatened; pre­dictions about future events may be under­mined as scientific data are revised; and actual moderate earthquakes often spur more people to take hazard mitigation efforts than predictions of large earthquakes in the future.

Rise o f B a i l e y Wil l i s a s E a r t h q u a k e P r o p h e t

The earthquake prediction considered here was made not by a crank, but by a re­spected member of the geological commu­nity. Bailey Willis had spent 3 decades with the U.S. Geological Survey, rising to the rank of chief geologist, before joining the faculty of Stanford University in 1915. He was a mem­ber of the National Academy of Sc iences and later served as president of the Geological So­ciety of America in 1928. In 1921, Willis also assumed the presidency of the Seismological Society of America, which had been founded in the aftermath of the catastrophic 1906 San Francisco earthquake. Under Willis's prede­

cessors, this society had warned the Califor­nia public that strong earthquakes would re­cur in the state. In the early 1920s, though, this message was still ignored by the public, mostly because the Seismological Society had not yet discovered how to capture the at­tention of the news media [Geschwind, 1996] .

Willis embarked on a concer ted public re­lations campaign to bring about wider knowl­edge of earthquake hazards in California. Besides speaking to a variety of business and civic organizations, Willis produced a large-scale map showing all known active faults in the state. In 1924, he also published a five-part series summarizing, in easily understood language, all that was known about earth­quake risks in California. Here he first an­nounced that" [a] great shock may c o m e soon [to the Los Angeles area] , or within a decade, or not till after more than a decade . But it will come" [Willis, 1924] . Willis based this rather vague forecast on a number of fac­tors. Southern California had suffered severe earthquakes in 1769 ,1812 , and 1857, suggest­ing a recurrence interval of about 45 years. Consequently, Willis considered the absence of a strong earthquake in the nearly seven decades s ince 1857 to be an ominous quies­c e n c e . Moreover, southern California was rid­dled with many active faults. Finally, Willis thought that the series of moderate earth­quakes that struck the Imperial Valley in 1915, San Jacinto in 1918, and Inglewood in 1920 might be precursor events for a much more significant regional stress release.

Willis received considerable additional support for his forecast from the preliminary results of work by the United States Coast and Geodetic Survey in California. This survey had triangulated the state in the late 19th cen­tury to fix the latitude and longitude of many prominent points. At the urging of seismolo­gists, the survey repeated this triangulation in

the mid-1920s, running lines from near Lake Tahoe to the San Francisco Bay area and then south along the California coast to Santa Barbara County. Although the retriangulation was not yet complete, the survey released a map in early 1924 showing preliminary re­sults from this work. Geographic positions in northern California had shifted rather little from the 1880s to 1922, but in southern Cali­fornia, some points west of the San Andreas Fault near Santa Barbara had moved north­ward as much as 24 feet [Bowie, 1924] .

S ince no large earthquake had occurred in the Santa Barbara area in this time, the measured surface deformation meant that an immense amount of strain had built up along the San Andreas Fault, ready to be released in a catastrophic quake. In late 1925, Willis is­sued a statement to a northern California newspaper that summarized yet again the seismic history of southern California as well as the results of the retriangulation. He con­cluded that" [n] o one knows whether it will be one year or ten before a severe earth­quake comes , but when it does c o m e it will c o m e suddenly, and those who are not pre­pared will suffer."

The newspaper's editor understood this statement to mean that the strain along the southern San Andreas Fault "will probably re­sult in an earthquake shock in from one to ten years." This interpretation was soon picked up by national news media. The New York Times proclaimed that, according to Willis, "Los Angeles, or its immediate vicinity, will exper ience a severe earthquake, prob­ably more violent than that at San Francisco in 1906, in from one to ten years." Several weeks later Time magazine made the predic­tion appear even more certain by asserting that "within the next ten years Los Angeles will be wrenched by a tremor worse than that of San Francisco."

Willis's initial notoriety for predicting im­minent doom for Los Angeles c a m e from a misrepresentation of the actual statement he had made to a local newspaper. In May 1926, though, Willis himself positively asserted that the expected catastrophe would c o m e within the next decade . The occas ion was the annual meeting of the National Board of Fire Underwriters, a group of insurance ex­ecutives, in New York City, at which Willis gave the keynote address. During the ques-tion-and-answer period following his speech,

This page may be freely copied.

Eos, Vol. 78, No. 38, September 23, 1997

Fig. 1. a) Pyroclastic flow from late July 1996 entering the sea at the mouth of the Tar River valley. b) Photograph of the Soufriere Hills volcano from the east on October 3,

1996. The new dome in English's Crater has a large crater open to the east within it due to the' collapse and explosive eruption on September 17, 1996. A new dome appears on the crater floor. The Tar River valley is inundated by pyro­clastic flow deposits and a new delta of these deposits progrades into the sea.

Page 401