A-F

The extrusive phase of the eruption of the Soufriere Hills volcano on Montserrat has been dominated by low frequency volcanic earthquakes. These earthquakes have distinctive peaked spectra and commonly occur in swarms related to the pressurization of the upper part of the magma conduit. We use data from an array of broadband seismometers to examine spatial and temporal variation in the spectral properties of these earthquakes, between January and August 1997. Although spectra are generally stable over long periods of time at a given reference point, we also find evidence changes in the spectra with time and with event magnitude, which may be attributed to changes in the source. In general, spectra are not coherent across the array. This leads to the conclusion that the wave-field is a combination of both source and propagation effects. However, during certain tremor .episodes we observe harmonic spectra, with shifting spectral peaks which are coherent across the whole array. In some cases this behaviour can be modelled by repetitive triggering of low frequency events, where the harmonics are controlled by the trigger frequency or by the harmonic spectra of individual earthquakes. However, other occurrences of similar behaviour cannot be so easily explained in this way. We suggest that the shifting spectral lines may be due to the changing behaviour of interface waves, resulting from short term changes in conduit properties.

The characteristics of volcanic seismicity at Ruapehu changed noticeably during the 1995-1996 eruptions. During the 1970's, 1980's, and early 1990's, volcanic tremor and volcanic earthquakes at Ruapehu were both characterized by a dominant frequency of c. 2 Hz when recorded at the summit and upper-mountain seismic sites and by frequencies of 1.2 - 2 Hz at lower-elevation flank sites. The characteristics of the seismicity remained unchanged through the eruptions of September and early-October 1995. However, following the eruptions of mid-October, volcanic earthquakes recorded at the summit site became more broadband and were similar to the hybrid events described at Redoubt, Galeras, Montserrat, and other volcanoes. Furthermore, the dominant frequencies of the volcanic tremor and the volcanic earthquakes began to differ. Whereas volcanic earthquakes remained predominantly 2 Hz at the summit and upper-mountain seismic sites, the frequency of the background tremor increased to 4 - 7 Hz. However, periods of strong tremor, continued to be dominated by frequencies of c. 2 Hz.

We illustrate these changes in the characteristics of the volcanic seismicity using data from three deployments of broadband seismographs on Ruapehu as well as data from the permanent, short-period, vertical-component seismograph network that monitors Ruapehu. A six-week deployment during 1994 of 14 broadband seismographs on Ruapehu including four stations surrounding the crater lake and two on the upper mountain occurred during a time of volcanic quiescence. Although only a few volcanic earthquakes were recorded and there were no episodes of strong tremor during this deployment, the volcanic earthquakes

showed identical frequency characteristics to the background tremor. A deployment of three broadband seismographs on the upper mountain from December 1995 to January 1996 also occurred during a period of relative quiescence between the September - November 1995 and June - August 1996 eruptions. A reoccupation of the two upper-mountain 1995 - 1996 sites and the deployment of a third broadband seismograph on the upper mountain for 6 months in 1998 also occurred during a time of volcanic quiescence, but followed a 3-month period of sporadic, discrete, small eruptions lasting from October 1997 until January 1998. This deployment spanned numerous periods of strong volcanic tremor and the occurrence of several large volcanic earthquakes. Both of the latter deployments revealed significant differences between the characteristics of the volcanic earthquakes and the volcanic tremor.

A model for the volcanic plumbing system of Ruapehu prior to the 1995 eruptive activity consists of four regions: (1) degassing magma, (2) single-phase vapor, (3) two-phase vapor-liquid, and (4) single-phase liquid. The single-phase vapor zone, which extends to a depth of approximately 1000 m beneath the surface of the crater lake, contains the source regions of the 2 Hz tremor and the volcanic earthquakes. In contrast, higher frequency tremor is thought to originate from a separate source at the base of crater lake whereas volcano-tectonic earthquakes occur outside of the conduit. We propose that changes in the seismicity during the 1995 - 1996 eruptions resulted from both the disruption to the source region of the 2 Hz tremor and volcanic earthquakes and the presence of magma at a very shallow depth within the volcano.

Three ski fields are situated on the upper slopes of Mt. Ruapehu, an active andesite volcano located in the center of the North Island of New Zealand. Lahars originating from the summit crater lake on Mt. Ruapehu can enter the upper ski fields within 90 seconds of their generation and move through the ski fields with speeds of up to 25 m/sec. Thus, they pose a major hazard to skiers and other users of the mountain. In contrast, the eruption of large rocks is restricted to within about 2 km of the crater, and the resulting hazard can be easily managed by restricting access to this area during periods of enhanced volcanic activity. The eruption of large volumes of ash, although possibly affecting large areas, is not likely to injure users of the mountain in the short-term. Lahars are therefore the main volcanic hazard to safe ski field operation.

Prior to the 1995 eruptions of Ruapehu, the Department of Conservation (DOC), Ruapehu Alpine Lifts (RAL) and the Institute of Geological & Nuclear Sciences (GNS) instituted a review of the existing Lahar Warning System (LWS). This review was brought about by a realization that the LWS was 10 years old and in need of reassessment. It was made more urgent by the failure of the LWS to trigger during the eruption of the afternoon of September 23, 1995, when lahars swept down the upper slopes of the volcano and through the recently vacated Far West T-Bar area. We report on the design of the upgraded system that, in light of the way it is designed to function, has been named the Eruption Detection System (EDS).

Unlike the existing Lahar Warning System, the Eruption Detection System does not attempt to detect a lahar; rather, it uses the occurrence of an eruption, together with the size of the accompanying volcanic earthquake, as a guide to whether lahar generation is likely. The algorithm consists of three parts: the detection of volcanic earthquakes, the estimation of the size of those earthquakes, and the detection of airwaves. The earthquake detection component performs analyses of the frequency content of the seismic signal from up to six seismographs to distinguish volcanic earthquakes from other types of earthquakes and non-seismic sources. The airwave detection component is used to differentiate volcanic earthquakes that are associated with an eruption from those that are not. Once it has been determined that an eruption has occurred, the size of the volcanic earthquake is then used as an indication of the likelihood of lahar generation.

The EDS was installed on Mt. Ruapehu during 1998 and 1999 and became operational in August 1999. The system acquires data from six seismographs and two acoustic microphones located on and around the volcano. The detection algorithm runs on a Quanterra Q4120 data acquisition system at the GNS volcano observatory 9 km from the summit of Mt. Ruapehu. The EDS interfaces with existing equipment to trigger sirens and play a prerecorded warning message through speakers on the Whakapapa ski field and uses an auto-dialer to notify DOC staff of possible lahars when a system trigger occurs. The EDS is remotely accessible and fully programmable using TCP/IP protocol.

The origin of volcanic tremor is still the subject of vigorous scientific debate. One of the most recent models, by Ripepe and Gordeev, suggests that volcanic tremor is related to the coalescence of bigger gas bubbles from a layer of smaller ones in the magma foam, forced by a structural barrier. The evidence of this model is mainly related to the relationship between the seismic signal and infrasonic waves. These must be generated by the same dynamical process, although the latter are related to a successive phase, i.e. the bursting of the bubbles at the magma-air interface.

Previous studies by Carniel and Di Cecca furnished evidence that volcanic tremor can be considered as a state variable of a deterministic dynamical system of "low" dimension. If the pressure signal is related to the same dynamical process, the two time series should not show too different dynamical behaviour.

On 19 June 1999, an experiment was carried out, recording simultaneously not only seismic and pressure waves, but also thermal data, using a Minolta/Land Cyclops 330 infrared thermometer aimed at one of the vents in NE Crater. Preliminary results from the dynamical analysis of the resulting time series are presented and discussed.

G-M

Over an 18 hours interval the tremor at Lascar volcano, Chile, was characterized by a spectrum with narrow peaks at a fundamental freqency of about 0.63 Hz and up to 30 overtones at exact integer multiples. This "harmonic tremor" was recorded at four three-component, high-dynamic range stations during the deployment of the Proyecto de Investigación Sismológica de la Cordillera Occidental 94 (PISCO'94).

Three source models from fluid dynamics produce repetitive, nonsinusoidal waveforms: The release of gas through a very small outlet (the soda bottle model), slug flow in a narrow conduit, and von Kármán vortices produced at obstacles. These models represent different flow regimes, each with its own characteristic range of Reynold's numbers. Combining the Reynold's numbers for each model with typical kinematic viscosities for fluids encountered in volcanoes -- magma, water, steam, air or some combination, at appropriate temperatures and pressures -- provides limits on such physical parameters of the volcano as the dimensions of the flow conduit and the flow velocity of the fluid generating the tremor.

For each of these models, I calculate the motion at the source using reasonable parameters for the geometry and flow properties and compare it with the amplitude of the seismograms recorded at the stations of the Lascar network.

Active volcanoes represent severe hazards with potential risk for the neighbouring population. Therefore the availability of methods for the forecasting of volcanic activity is of paramount importance. A stochastic approach has been developed for the identification of precursors announcing the imminence of eruptive events. The method has been applied to seismic data monitored at Stromboli volcano over a seven years period. Stochastic estimation of time components for the volcanic tremor were performed. Their behaviour is analysed in relation to paroxysmal (explosive) phases occurring at Stromboli.

This stochastic technique shows forecasting potential and should contribute to the assessment of hazard level (with respect to explosive phases) at short term. In order to reduce forecast uncertainty, additional non-seismic variables are required; some experiments in this direction are presently undertaken.

Analysis of the data issued from a broadband seismic network operating on the ASO volcano in Japan shows that well individualized, long period events may have very similar forms and may occur continuously, even when the volcano is not in an eruptive phase. Therefore, controlling the activity of a volcano by counting automatically the number of such events and measuring different parameters such as duration, amplitude, dissimilarity between each event of these long period tremors is of the greatest interest.

These analysis can be done by using artificial intelligency: we applied the algorithms developped for earthquakes and strong motions classification (the so called SPARS method, i.e Syntactic PAttern Recognition Scheme) to organize software allowing a real-time monitoring of pre-eruptive activity. It is also interesting to see the non linear behavior of the temporal distribution of these events that often happen in a small region, that can be caracterized by multi-fractal dimensions. The temporal variations of such multi- fractal dimensions may be of some interest to control the continuous activity of a volcano.

One of the most common types of seismic event recorded during the eruption of Soufriere Hills volcano is known as a rockfall signal because signals recorded when rockfalls are observed on the dome are of this type. There is more to these signals than purely the seismicity generated by falling debris, however. Evidence is presented that a second seismic source also contributes to these events and the recent deployment of a pressure sensor near the volcano has shown that this seismicity is associated with de-gassing at the surface of the dome.

No abstract

Many different models have been proposed for the source of low-frequency or long-period volcanic earthquakes. Almost all such models invoke a liquid or gas phase to provide the compliance in the source process needed to generate the lower frequency and peaked spectra of these characteristic events. Because of the large variety of source spectra and large variety of magma types from different volcanos, there is likely no one model which applies to all observations. Rather, different models likely apply to different situations. In this presentation I provide evidence for yet a different type of source model; one which does not involve a fluid phase. I propose that stick-slip type failure may occur within a magma with high viscosity under conditions of a high strain rate. The seismograms of the resulting radiated seismic energy may have a much lower frequency content that those of a similar sized conventional tectonic earthquake because the stress-drop of such events will be small and the elastic rigidity of the fracturing medium is much lower than rocks around a tectonic fault.

Seismograms from several eruptions of Mount St. Helens during the mid-1980's exhibit characteristics consistent with this model. In particular, eruptions during 1984 and 1985 were dome-building eruptions of viscous, low-gas content, dacitic magmas in which slickenside-like striations were sometimes visible on the actively growing lob of the dome. Seismic events occuring before and during the early stages of these eruptions had characteristic low-frequency content, but not overly peaked spectra. They often occurred in regular sequences of nearly identical events, ie multiplets which evolved slowly in character and size over periods of hours. In at least one case a decrease in inter-event times between individual isolated earthquakes progressed until they were close enough together to blend into what appeared to be continuous volcanic tremor. The transition of event type and size and the energetics of the whole sequence is consistent with the stick-slip source mechanism for seismic energy radiation in these cases.

N-Q

Recent observations from Soufriere Hills volcano in Montserrat reveal a wide variety of low-frequency seismic signals. We discuss similarities and differences of hybrid earthquakes and long-period events, and their role in explosions and rockfall events. These events occur usually in swarms, and occasionally merge into tremor, an observation which can shed further light on the generation and composition of harmonic tremor. We use a 2-D finite difference method to model major features of low-frequency seismic signatures and compare them with the observations. A depth dependent velocity model for a fluid-filled conduit is introduced which accounts for the varying gas-content in the magma, and the impact on the seismic signals is discussed.

We analyse carefully episodes of tremor which show shifting spectral lines and model those in terms of changes in the gas content of the magma as well as in terms of a time-dependent triggering mechanism of low-frequency resonances. In this way we explain the simultaneous occurrence of low-frequency events and tremor with a spectral content comprising integer harmonics.

A steady homogeneous magma model is used to derive the relationship between gas mass and volume fraction, density, temperature, and pressure as a function of depth. In turn, these parameters are used to obtain the vertical seismic velocity profile in the gas-charged magma. A finite difference model of a magma-filled conduit embedded in an elastic medium is then employed to generate the seismic wavefield in and around the conduit. The high impedance contrast between gas-rich magma and surrounding rock results in the seismic energy being efficiently trapped in the conduit; this leads to the generation of a long-lived resonance commonly observed as low-frequency earthquakes and harmonic tremor. During a single event a variety of different seismic radiation patterns along the conduit is observed leading to the occurrence of several distinct seismic phases in the synthetic seismograms. Observations from several volcanoes show peaked amplitude spectra with integer harmonic overtones which exhibit a time-dependent gliding. These features are successfully modelled by varying the excess pressure and, consequently, the gas volume fraction and the seismic velocity with time. This simulates sudden degassing events such as the reduction of excess pressure by ash-venting.

Time and again seismo-volcanic activity that shows a diurnal and semidiurnal modulation is related to the earth's tidal stress field as an external triggering mechanism. Several seismic data sets from different volcanoes are examined, including Stromboli volcano, Italy, Mt Ruapehu, New Zealand, and Merapi and Batur, Indonesia, where in some cases also weather data are incorporated. A comparison with theoretical body tides in the spectral domain reveals strong evidence against a correlation between seismo-volcanic activity and tidal stress. However, other parameters such as temperature and barometric presssure show the same diurnal and semidiurnal modulation and are, therefore, candidates for a possible external modulation of volcanic activity. Several mechanisms are discussed. For Stromboli a clear correlation between volcanic activity and barometric pressure is shown.

Ballistic blocks are a common feature of volcanic eruptions, but so far there have been few studies of their ballistic behaviour. The Minoan eruption on Santorini, Greece, was a major Plinian caldera-forming event that occurred at ca. 1640 BC. The eruption is generally divided into 4 distinct phases:
1. Plinian phase,
2. phreatomagmatic base-surge,
3. ash-flow phase and
4. non-welded ignimbrite.
Many ballistic blocks, mostly consisting of older lava-fragments with diameters up to several meters, were ejected during phases 2 and 3. As caldera collapse followed the eruption, the precise vent location is unknown, but its position during phase 1 has been inferred from the pumice isopachs of the Plinian deposits. Location of the vent during the following phases has been estimated by measuring thickness variations, flow directions within the pyroclastic flows and impact directions of ballistic blocks.

In this study, the vent location has been estimated by analysis of the size distribution of the ballistic blocks. Phase 2 blocks show no clear size-dependant distribution pattern. Blocks ranging between 0.15 and 1.60 m in diameter are found almost throughout the entire deposit, but are concentrated in the south-eastern parts and within a circle of 14 km diameter, thus suggesting a maximum throw range of at least 6-7 km. The centre of that circle was regarded as the most probable ejection region. It is proposed that the subaerial vent of phase 1 developed into an opening fissure to the southwest along the general tectonic trend. Very large blocks, up to 3 m in diameter (or possibly more) were also ejected during phase 3, with the largest ones being found only in the northern half of the island group near the caldera rim. This indicates that either the existing vent extended to the northern part of the present-day caldera, or that another vent opened, possibly along radial fractures with beginning caldera collapse.

Minimum initial velocities of the ejected blocks have been calculated. Assuming that the blocks were ejected into still air, parameters such as range, air-drag coefficients, size and shape of the blocks, specific weight, and ejection angles are very critical for blocks thrown to ranges significantly more than 3-4 km. The ballistic trajectories of blocks from the second phase were numerically calculated and the minimum initial velocities were obtained by trial and error methods using optimum launch angles (typically between 20 and 30 degrees). For the biggest blocks, these velocities are in the order of 300-450 m/s, whereas many of the smaller blocks theoretically had to be launched at least at 400-600 m/s. These estimations are based on "very favourable" conditions, but velocities of at least 600 m/s are likely using "normal" parameters.

These high velocities seem unrealistic because
a) no reports could be found in the literature of blocks ejected at significantly more than 400-600 m/s,
b) no plausible explosion model could be established that might explain such high ejection velocities which require even higher gas exit velocities at the vent.

The archipelago of Santorini in the South Aegean Sea consists of the islands of Thera (76 km2), Therasia (9 km2) and Aspronisi (0.1 km2) enclosing a sea-filled caldera of ca. 60 km3 volume and two young volcanic islands Palea and Nea Kameni (0.5 and 3.4 km2) inside the caldera.

Santorini is an active volcano of the Hellenic Volcanic Arc which is related to subduction of the African beneath the Aegean plate and stretches from the mainland of Greece to the Bodrum peninsula in Turkey. Santorini lies on the southern margin of a relatively stable crustal block ca. 200 km behind the subduction trench. Its geologic and volcanic evolution has been determined by a system of NE-SW trending tectonic lines.

Basement rocks of Santorini are mainly schists and marbles from Triassic time belonging to the metamorphic Cycladic massif. They are a part of a tectonic ridge and only crop out in the SE parts of Santorini where they form prominent heights.

Volcanic activity started ca. 2 million years ago and is generally divided into six main stages:
1) early centres of the Akrotiri peninsula,
2) cinder cones of the Akrotiri peninsula,
3) the northern stratocones (Megalo Vouno and Mikro Profitis Elias),
4) the first explosive eruptive cycle,
5) the second explosive eruptive cycle and
6) the formation of the Kameni shield during historic time.

So far, 12 Plinian eruptions with erupted magma volumes from few to tens of km3 have been recognised. The present-day caldera is a composite structure created by at least 4 major collapses.

The explosive activity last culminated with the Minoan eruption at ca. 1640 BC, a caldera-forming Plinian event that discharged around 30 km3 of rhyodacitic magma. The eruption, like most other large explosive events on Santorini, occurred in distinct phases, starting with pumice fall-out followed by phreatomagmatic surge and massive pyroclastic flows.

On Santorini, the Minoan eruption destroyed a high-standing Bronze Age civilisation. Rests of a rich Cycladic town with well-preserved buildings and magnificent wall-paintings are being excavated near the village of Akrotiri since 1967. The probably catastrophic impact of the Minoan eruption on the Eastern Mediterranean is an important subject of research. It is proposed that tsunamis, ash-fall and climatic changes by emission of aerosols into the stratosphere might have led to the decline of the Minoan civilisation on Crete.

R-Z

Pressure oscillation data from the laboratory simulation of magma-steam foam flows using natural gum rosin and a volatile organic solvent are discussed. These experiments were carried out over a range of volatile supersaturations generating conditions ranging from gentle unfragmented to violent fragmenting flow. Pressure oscillations were detected over the whole range of flow conditions. Spectral analysis of the oscillations showed that the dominant oscillation frequencies could be attributed to resonant oscillation in the foam. Flows where foam fragmentation occurred showed a broader, higher frequency and more energetic spectrum than unfragmented flows. Pressure oscillations were generated within different regions of the flows. Some of the experimental pressure data show qualitative similarity with volcano-seismic data from a number of volcanoes.

It is assumed that seismic broadband signals in the nearfield of Strombolian explosions well reflect the pressure-time function acting at the source. However, caution is advised when source models are established from the sole use of these data. Quite similar time series can be obtained when a pressure transducer, f.i., is put into boiling and puffing semolina pudding. It only indicates that the information content with seismic signals induced by eruptive activity might not be sufficient to draw too detailed conclusions of the source mechanism. Many different loading-deloading processes can lead to comparable seismograms.

Movement of fluids in the network of channels, tubes and cavities inside an active volcano radiate a large variety of random and impulsive seismic tremor signals. One very interesting class of tremor wavelets, called tornillos at Galeras volcano in Colombia, have screw-like profiles on seismograms and can last up to several minutes. The seismic record at Galeras volcano includes many tornillos recorded both on short-period and broadband seismometers.

The amplitude modulated, narrow-band tornillo wavelets can be described ,by a characteristic function, f_T = p×a(t)×exp(2pif₀t) of the tornillo's parameters, polarization, p, envelope, a(t), peak frequency and amplitude, f₀ and max(f_T) and the tornillo energy, E_T.

While their envelope properties, peak amplitude, rise time and decay constant exhibit strong variations even over short intervals, tornillos peak frequencies fall predominantly in one of several narrow bands between 0.5 and 8 Hz. Thus, all tornillos conform to a common pattern and we can consider them as an ensemble of signals radiated by one or more physically similar sources. The members of the ensemble have different states of excitation and may be described by distribution and correlation functions of and among the individual parameters and their variations with the state of activitiy.

Cavity resonators offer a model for the tornillo source, for which fluctuations in the physical and chemical boundary conditions of the fluid rather than the geometry explain both the high stability of the peak frequency during a single tornillo as well as its variation from one tornillo to another.

Although it is natural to consider pressure variations within a cavity as the source of the seismic signals associated with strombolian explosions, there is no way of determining the pressure, the volume, or the rigidity of the surrounding rock independently from seismic observations. The only quantity that can be inferred is the volume of rock that is displaced when the pressure in the cavity changes. This leads to the concept of a "volume source" whose only significant parameter is the displaced volume. The mathematical relationship between the source volume (as a function of time) and the seismic displacement is much simpler than that between pressure and displacement. The presentation will outline the basic properties of volume sources and use this concept to interpret long-period seismic observations on Stromboli.