GEOPHYSICAL RESEARCH LETTERS, VOL. 28, NO. 12, PAGES 2349-2352, JUNE 15,2001

Inflation or deflation? New results for Mayon volcanoapplying elastic-gravitational modeling

J. Fernández,' K. F. Tiampo,2 G. Jentzsch;' M. Charco,' and J. B. Rundle2

Abstract. Volcanic activity produces deformation and gravitychanges that many times can be used as precursors of futureeruptions. Applying geodetic techniques to monitoring activityinvolves interpretation using deformation models. Usually, theobserved changes of the deformation and gravity fields areinterpreted seperately, not in a joint inversion. It can bedifficult, if not impossible, to interpret the data coherently orcorrectly in terms of the characteristics of the intrusion or thedeflation derived from the gravity changes with purely elasticmodels, as in the case of Mayon Volcano, Phillipines. Weshow that elastic-gravitational models can be used to interpretthese cases simultaneously leading to a result that is moreplausible on the basis of the available information. Thus, wemay need to change the philosophy normally used to interpretgeodetic observations. Interpretation as proposed in this workcan significantly improve the possibility of predicting futureeruptions.

1. IntroductionVolcanic activity almost inevitably produces geodetic

effects such as deformation and gravity changes before,during, and after the activity, as well as between events. Onthe basis of this fact and the high levels of precisionattainable, different geodetic techniques are proving to be apowerful tool in the monitoring ofvolcanic activity, making itpossible to detect ground motion and gravity changes thatreflect magma rising from depth, sometimes months or weeksbefore the magma flow leads to earthquakes or other eruptionprecursors [e.g., Delaney and McTigue, 1994; Rymer, 1996;Dvorak and Dzurisin, 1997; Massonnet and Feigl, 1998;Dzurisin et al., 1999; Fernández et al., 1999; Stein el al.,2000; Rymer and Williams-Jones, 2000]. Geodetic monitoringthus complements seismic monitoring by extending the studyof volcanic phenomena from seconds to years and providingdetails on the growth of magma bodies within the volcano[Stein el al., 2000]. Applying such longer-terrn monitoringtechniques to volcanically active zones inevitably involvesdata processing and the subsequent final interpretation ofobserved deformations and gravity changes. Mathematicaldeformation models [e.g., Rundle, 1982; Fernández andRundle, 1994; De Natale and Pingue, 1996; Dvorak andDzurisin, 1997] are a basic, essential tool for the latter task.

Quantitative mathematical models cannot purport to coverall different physical and chemical aspects of volcanoes and

I Instituto de Astronomía y Geodesia (CSIC-UCM), Madrid, SpainZ CIRES, University of Colorado, Boulder, Colorado3 Institute for Geosciences, FSU Jena, Jena, Germany

Copyright 2001 by the American Geophysical Union.

Paper number 2000GL012656.0094-8276/0 112000GLO12656$05.00

we must select and focus on key phenomena. Presentknowledge on critical stages of volcanoes prior to eruption ismostly based on elastostatic views; by studying volcanicunrest in terms of mechanical models involving overpressurein magma chambers and conduits. Reaching eruptiveconditions is interpreted, in this framework, as overcomingthe mechanical rock strength in large volumes, from the top ofa magma chamber to the surface. Mogi [1958] applied acentre of dilatation (point pressure source) in an elastic half-space to interpret the ground deformation produced involcanic areas. Mogi's model has been extensively applied inmodelling ground deformation and has been successful inexplaining primarily the vertical component of thedeformation. However, this model often poses difficulties insimultaneously modelling observed displacements and gravitychanges [Rymer el al., 1993; Rymer, 1996], and there is alarge body of evidence for ground deformation and seismicityat calderas and other volcanic areas that cannot be modelledby these purely elastic effects [e.g., Bonafede, 1991; DeNatale el al., 1997; Gaeta el al., 1998; Jahr el al., 1998;Baitaglia el al., 1999; Jentzsch el al., 2001].

Rundle [1980, 1982] presented an elastic-gravitationalmodel that considers a stratified half-space of homogeneouslayers and takes into account the interaction between the massof the intrusion and the ambient gravity field and the effectcaused by the change of pressure in the magmatic system. Ithas been shown theoretically [Fernández and Rundle, 1994;Fernández el al., 1997; Fernández el al., 1999] that conside-ration of gravity effects can be fundamental for adjusting andproperly explaining gravity changes measured in active zones.Vertical displacements and gravity changes produced bypressure increases and the mass ofthe intrusion have differentsigns. This fact is very important as it can serve to explainobservations in active zones where major gravity changesappear without any significant deformation, or vice versa.

The following is a practical comparison ofthese theoreticalresults with the modeling of geodetic data observed at Mayonvolcano in the Philippines, where there are gravity changeswithout significant deformation. Furthermore, while gravitychange data and displacement data generally are interpretedseparately, we will see in the modeling of deformation andgravity change data that this can lead to different and maybeincorrect interpretations. The results obtained will show thatelastic-gravitational models can be a far more appropriateapproximation to problems of volcanic load in the crust thanthe more commonly used purely elastic models [e.g., Mogi,1958; McTigue, 1987; Davis, 1986; Yang el al., 1988].

2. Background & DataMayon volcano is part of the Bicol volcanic chain

southeast of the island of Luzon, Philippines (see Figure 1).Part of the Legaspi Lineament of the central Philippine faultsystem runs NW-SE across Legaspi City, southwest of

2349

2350 FERNÁNDEZ ET AL.: NEW RESUL TS FOR MA YON VOLCANO

the summit. The differential GPS measurements carried outwith three simultaneous receivers revealed no significantchanges of elevation within the accuracy range ±3 to 4 cmbetween all campaigns.

The connection between the variation of gravity andelevation due to mass and/or density changes is usuallyinterpreted in relation with two gradients: if gravity followsthe free air gradient (FAG), no subsurface change in the masshas occurred, while data following the gradient after thestandard Bouguer correction (BCF AG) implies mass changes[see e.g., Brown and Rymer, 1991; Rymer el al., 1993; Rymer,1996]. Any departures from these gradients are used to modelvolcanic processes. In this case, however, the results atMayon cannot be related to the gravity gradients FAG andBCF AG because the gravity changes drawn against heightdifferences from one campaign to the next show no gradient[Jenlzsch el al., 200 1]. Furthermore, it is unusual for gravityto increase with decreasing activity, as it does during thisparticular time period at Mayon.

The estimated maximum effect due to water level changesin the zone on gravity is about 50 ~Gal. Since some of thepoints with strong gravity changes are located in consolidatedrock, ground water level changes cannot be responsible for

124' 125' 126< the observed gravity variations. Further, the steep slope ofthe:L -121' 122< 123·119' 120'

Figure 1. Location ofMayon Volcano.

Mayon. The shape ofthe volcano is a nearly perfect con e withan altitude of 2462 m. Nearly one million people live in thevicinity of Mayon vol cano , and Legaspi City is situated only12 km away from the crater. The activity of Mayon isquasiperiodic: during the last century it erupted nearly every10 years [Jenlzsch el al., 2001]. The last eruption was in early2000.

A small eruption took place in February/March 1993, justafter the start of gravity and GPS measurements in December1992 [VOIksen and Seeber, 1995; Jahr el al., 1998; Jentzsch elal., 2001]. Therefore, a second campaign was carried out inMay 1993 to monitor the de crease in activity. Height controlis provided by simultaneous Global Positional System (GPS)measurements; in all, the network consists of25 points. Thereare two profiles towards the summit (one of them up to anelevation of 850 m), that are connected to a local and aregional network around the volcano with an extension of 40km by 50 km. Two points at the opposite side of the Legaspilineament are al so connected to the reference network. Theerrors of the gravity differences [Jenlzsch el al., 2001]between the points derived from a least squares adjustment ofthe data of each campaign are around ± 12 uGal.

Five microgravity and differential GPS campaigns werecarried out over the next four years. There was no significantgravity change between the first and the second campaign(December 1992, May 1993), although Mayon was active inFebruary/March 1993. The increase in gravity along theprofiles in the slope was significant between May 1993 andDecember 1996, reaching around 150 ~Gal (± 14 ~Gal),increasing with elevation and as the distance from the craterdecreases. The variation is 30 ~Gal per year if a continuousprocess is assumed. It should be noted that the volcanicactivity .at Mayon subsided during that time and the volcanoremained inactive up untillate 1999.

All the data show that gravity changes are restricted to thearea around the volcano within a radius of about 8 km from

a

ruUPA-LAHAR-CHANNEL PROflLE. EPOCH 4-1

600

Ueosured: _

MOdeled, GA Inver,¡on: - - -Yodeled, Disk Source:

400

200

"'-:. :-<0- •••••••

-200

o

b

2000

1500

1000

'¡:<,

g500

o

-500

o

10 15kitornetera

TUMPA-LAHAR-CHANNEL PROfILE. EPOCH 5-1

,,,Meosured: _l.4odeled, GA tnveraice: - - -Modele:d, Disk Source:

5 10 15kilometers

Figure 2. 3. Gravity changes for Tumpa-Lahar-Channel profile,observed VS. Modelled for epochs 4-1. b. The same for epochs 5-1. Dotted lines represent gravity changes modelled consideringdensity changes within a vent system [Jahr el al., 1998, Jentzschel al., 2001]. Dashed lines show results of modelling usingelastic-gravitational deformation model via genetic algorithm(GA) inversion technique.

FERNÁNDEZ ET AL.: NEW RESUL TS FOR MA YON VOLCANO 2351

Mayon Resthouse profile does not seem to permit anydiscussion of a prevailing water table. Only the small gravitychange at the foot ofthe Mayon Resthouse profile ofup to 30~IGal can be associated to changes in the local water level[Jenlzsch el a/., 2001].

As a result, the observed gravity increase, unaccompaniedby significant changes in elevation, cannot be explained withthe c1assical Mogi model, which primarily models surfacedeformation as a result of inflation. In a first step, it isexpIained [Jahr el a/., 1998; Jenlzsch el a/., 2001] by densitychanges within a vent system. This model adjusts about 50%of the observed signaI and expIains the pattern of observedgravity changes (see the computed resuIts shown in dottedIines in Figure 2). This model is based on a redistribution ofmass down the vent system. However, the inabiIity of thismodeI to account for a Iarge part of the observed gravitychange prompted further analysis using a more compIicatedelastic-gravitational model.

3. ResultsThe recent eruption of Mayon volcano and the intermediate

small activities give rise to the assumption that the oppositeprocess might have taken place: injection instead of deflation.We test this hypothesis by modeling the observed gravitychanges without deformation using a genetic algorithm (GA)inversion technique [e.g., Michalewi:c. 1994; Tiampo el a/.,2000] and considering elastic-gravitational Earth models[Rund/e, 1980, 1982; Fernánde: and Rundle, 1994;Fernánde: el a/., 1997; Fernánde: el al., 1999]. The elastic-gravitational model provides a much better solution thaneither the simple Mogi source or the vent magma model,emphasizing the importance of considering gravitationaleffects in modeling magmatic sources. The characteristicsobtained for the intrusion are: a depth of 1.95 km below thebase of the volcano, 0.21 kbar pressure, a radius of 1.65 km,and 0.474 MU for mass (1 MU = 1 Mass Unit = 1012 kg) forepoch 4-1; and a depth of 1.82 km, 0.31 kbar pressure, aradius of 1.71 km, and 0.841 MU for mass for epoch 5-1.Figure 2 compares the results for this elastic-gravitationalinversion (dashed line) with the actual measurements for theTurnpa-Lahar-Channel profile (sol id line), as well as that forthe model which redistributes magma down the volcanic vent(dotted line). The results for the Mayon-Resthouse profile forthe same epochs reproduce the total magnitude and characterof the measured signal, but do not extend as far from the vent.Further study is needed to determine if this is due toasymmetries in the source itself, or baseline inaccuraciesalong that particular profile [Jahr el a/., 1998; Jentzsch el a/.,2001]. Interpretation of these results suggests that the gravitychanges at Mayon are better explained by reinjection ofmagma below the volcano following the 1993 eruption ratherthan mass redistribution in the volcanic vent.

More importantly, the elastic-gravitational model, whichaccounts for the interaction of the mass of the intrusion withthe ambient gravity field and redistribution of densities insidethe crust, is demonstrated to be a very powerfu1 tool forinterpreting geodetic measurements in volcanic zones anddistinguishing between pressure and mass effects, unlike thepurely elastic models which are primarily used to analyzevolcanic systems, but which are unable to explain gravitychanges without displacements or vice-versa. It is also clearthat gravity and displacement data must be interpreted at thesame time in order to model magmatic sources correctly.

4. ConclusionsThe first conclusion from the results obtained above is that

interpreting deformations and gravity changes by using purely~Iastic m~dels sometimes can lead to results that are clearlymcorrect m terms of the location and characteristics of theintrusion. On the other hand, the interpretation of gravitychanges alone may be not sufficient, as can be seen at Mayonvolcano, where only about 50% of the observed changes canbe explained [Jahr el a/., 1998; Jentzsch el al., 2001]. Here,additional constraints are needed. Taking into accountdifferent processes in the magma system like pressurechanges in the intrusion or magma recharge, elastic-gravitational models should be used. Especially in caseswhere gravity changes exist without deforrnation, and viceversa, the interpretation should include adequate mass andpressure combinations for the intrusion. This approachaccounts for the effect of opposite signs for both types ofsource.

Finally, our results c1early show the need to change thephilosophy normally used to interpret geodetic observationsin volcanism.' in which displacement data and gravity changedata are not mterpreted together. Very often the interpretationusing purely elastic models that are both simplistic andinappropriate, gives rise to false alarms based on insufficientassumptions. Displacements and gravity changes must beinterpreted together wherever possible. This will allow us todistinguish between the different cases and to discriminatebetween the roles played by pressure changes and by massdisplacements, which are otherwise very hard to interpretcorrectly. A correct interpretation of the observed geodeticsignals will have implications in the improving anddevelopment of monitoring and alert systems for themitigation of hazards, as well as direct socio-econornicimplications as it affects urban and industrial infrastructureplanning. Remembering that many sources can be representedas masses or pressure changes, similar to the volcanic case,our results and conclusions are generally applicable to themonitoring and interpretation of other kinds of natural hazardsand rnan-rnade activities (mining, pumping, tunneling, etc.)which induce both displacements and gravity changes andpose a clear risk to human activity.

Acknowlcdgments. This research has been supported by projectsAMB96-0498-C04-04 and AMB99-1015-C02 of C1CYT, Spain;NASA grants NAG5-3054 and NGT5-30025, and European Unionresearch contract ENV4-CT96-0259. This reseach has been alsopartially supported with founds from a New Del Amo Prozramproject. the German Research Soco (DFG) and the German So;' forTechnical Cooperation (GTZ). We thank F. Luzón and twoanonirnous reviewers for their comments and suggestions.

ReferencesBattaglia, M.. Roberts, C. & Segall. P. Magma intrusion beneath

Long Valley caldera confirmed by temporal changes in gravity.Science, 285, 2119-2122 (1999).

Bonafede. M. Hot fluid migration:an efficient source of grounddeformation: application to the 1982-1985 crisis at Campi Flegrei-Italy: J. Vo/canol. Geother. Res., 48,187-198 (1991).

Brown, G.c. & Rymer, H., 1991. Microgravity monitoring at activevolcanoes: a review of theorie and practice. Cahiers du CentreEuropéen de Ceodynamique el Seisrnologie, 4, 279-304 (1991).

Davis P.M. Surface deforrnation due lo inflation of an arbitrarilyoriented lriaxial ellipsoidal cavity in an elastic half-space, withreference to Kilauea volcano, Hawaii. J. Ceophys. Res., 91, 7429-7438 (1986).

De Natale, G., Petrazzuoli, S.M. & Pingue, F. The effect of collapse

2352 FERNÁNDEZ ET AL.: NEW RESUL TS FOR MA YON VOLCANO

structure on ground deformations in calderas. Geophys. Res. Leu.,24, 1555-1558 (1997).

De Nata1e, G. & Pingue, F. Ground deformation modelling involcanic areas In R. ScarpalR. l. Tilling (Eds): Monitoring andMitigation of Vo/cano Hazards. Springer, 365-388 (1996).

De1aney, P.T. & McTigue, D.F. Vo1ume of magma accumu1ation orwithdrawa1 estimated from surface uplift or subsidence, withapp1ication to the 1960 collapse of Kilauea Volcano. Bul/.Volcanol., 56, 417-424 (1994).

Dvorak, 1.1.& Dzurisin, D. Volcano geodesy: the search for magmareservoirs and the formation of eruptive vents. Rev. Geophys., 35,343-384 (1997).

Dzurisin, D., Wicks, C. Jr. & Thatcher, W. Renewed uplift at theYellowstone caldera measured by leveling surveys and satelliteradar interferometry. Bu/!. Volcano/., 6/, 349-355 (1999).

Fernández, J., Carrasco, J.M., Rundle, J.B. & Araña, V. Geodeticmethods for detecting volcanic unrest: A theoretical approach.Bull. Vo/cano/., 60, 534-544 (1999).

Fernández, J. & Rundle, J.B. Gravity Changes and deformation dueto a magmatic intrusion in a two-Iayered crustal model: JGeophys. Res., 99, 2737-2746 (1994).

Fernández, J., Rundle, J.B., Granell, R.d.R & Yu, T.-T. Programs tocompute deformation due to a magma intrusion in elastic-gravitationallayered Earth models. Comp. & Geosc., 23, 231-249(1997).

Gaeta, F.S., De Natale, G. & Rossano, S.. Genesis and evolution ofunrest episodes at Campi Flegrei caldera: the role of the thermalfluid-dynamical processes in the geothermal system. 1. Geophys.Res., /03, 20921-20933 (1998).

Jahr, T., Jentzsch, G., Punongbayan, R.S., Schreiber, U., Seeber, G.,Volksen, C. and Weise, A., Mayon Volcano, Philippines:Improvement of hazard assessment by microgravity and GPS?Proc. Int. Symp. On Current Crusta/ Movement and HazardAssessment (IUGG, lAG), Wuhan, nov. 1997, SeismologicalPress, Beijin, 599-608 (1998).

Jentzsch, G., Punongbayan, R.S., Schreiber, U., Seeber, G., Volksen,C. and Weise, A., Mayon Volcano, Philippines: change ofmonitoring strategy after microgravity and GPS measurements. 1.Vo/canol. Geotherm. Res. (2001, In press).

Massonnet, D. & Feigl, K.L. Radar lnterferometry and its applicationto changes in the Earth's surface. Rev. Geophys., 36, 441-500( 1998).

McTigue, D.F. Elastic stress and deformation near a finite sphericalmagma body: resolution of the point source paradox. 1. Geophys.Res., 92, 12931-12940 (1987).

Michalewizc, Z. Genetic Algorithms+Data structures=EvolutionPrograms. Springer, Berlin Heidelberg New York (1994).

Mogi, K. Relations between the eruptions of various volcanoes andthe deformations of the ground sur faces around them. Bul!.Earthquake Res. lnst. Univ. Tokyo, 36, 99-134 (1958).

Rundle, J.B. Static elastic-gravitacional deformation ofa layered halfspace by point couple sources. 1. Geophys. Res., 85, 5355-5363(1980).

Rundle, J.B. Deformation, gravity, and potencial changes due tovolcanic loading of the crust. 1. Geophys. Res., 87, 10729-10744(1982). (Correction, 1. Geophy. Res., 88, 10647-10652, (1983»

Rymer, H. Microgravity monitoring. In R. ScarpalR. 1. Tilling (Eds):Monitoring and Mitigation of Vo/cano Hazards. Springer, 169-197 (1996).

Rymer, H., Murray, J.B., Brown, G.C., Ferruci, F. & McGuire, W.J.Mechanism of magma eruption and emplacement at Mt Etnabetween 1989 and 1992. Nature, 361, 439-441 (1993).

Rymer, H. & Williams-Jones, G., Volcanic eruption prediction:magma chamber physics from gravity and deformationmeasurements. Geophys. Res. Lett., 27, 2389-2392 (2000).

Stein, S., Hamburger, M., Meertens, C., Dixon, T. & Owen, S.,UNAVCO Conference Explores Advance in Volcanic Geodesy.EOS Trans, AGU, 81,121,126 (2000).

Tiampo, K.F., Rundle, J.B., Femández, J. & Langbein, J. Sphericaland ellipsoidal volcanic sources at Long Valley Caldera,California, using a Genetic AIgorithrn inversion technique. 1.Volcanol. Geother. Res., /02, 189-206 (2000)

Volksen, C. & Seeber, G. Establishment of a GPS based controlnetwork at Mayon volcano. Cahiers du centre Européen deGeodynamique el de Seismologie, 8, 99-113 (J 995).

Yang, X.M., Davis, P.M. & Dieterich, J.H. Deformation frominflation of a dipping finite prolate spheroid in an elastic half-space as a model for volcanic stressing. 1. Geophys. Res., 93,4249-4257 (1988).

J. Femández and M. Charco, Instituto de Astronomía y Geodesia(CSIC-UCM). Fac. C. Matemáticas. Ciudad Universitaria. 28040-Madrid, Spain. (e-mail: jft@iagmatl.mat.ucm.es)

G. Jentzsch, Department of Geology and Geophysics, 1nstitute forGeosciences, FSU Jena, Burgweg 11, D-07749 Jena, Germany.

K. F. Tiampo and J. B. Rundle, Cooperative Institute for Researchin Enviromental Sciences, University of Colorado, Boulder, CO80309.

(Received November 18, 2000; accepted March 19, 200 l.)