Tuesday 18 September - Session 1

The persistent seismic signals recorded on basaltic volcanoes, known as volcanic tremor, have proven to be one of the most significant measures for the volcano surveillance. Nevertheless, the continuous acquisition of the tremor signal produces masses of data (something like 50 MB per day and component) which are cumbersome to handle. Data condensation and parameter extraction, which can be carried out automatically, is highly recommended for the study of the long-term behavior of a volcano. In this context it is necessary to identify parameters which can be correlated with volcanic activity. The surveillance can be thus reduced to the monitoring of key parameters which speeds up the data processing as well.

In the light of these goals we present the analysis of volcanic tremor recorded on Stromboli between 1990 and 1998. We describe how we obtained a drastic data reduction carrying out multivariate statistical analysis, and our choice of the parameters to monitor. Finally, we discuss the relationship between these parameters and the volcanic activity visible at the surface. We believe that the basic concepts outlined here could be generalized in a sense that they may be used for the definition of a monitoring system based on data reduction and parameterization.

Broad-band seismic networks provide a powerfull tool for the observation and analysis of volcanic earthquakes. Observed amplitude spectrograms of long-period volcanic earthquakes display distinct spectral lines sometimes varying by several Hertz over time spans of minutes to hours.

We use 2D finite-difference modelling methods to compute the propagation of seismic waves in simplified volcanic structures. This modelling technique provides synthetic earthquakes, which frequency content is the same as the observed one. Moreover the change in magma properties (like the velocity) produces a change in the frequency content of the recorded signal, which we model continuously, i.e., as the wave propagation is computed. We model the gliding spectral lines by introducing continuously changing magma properties during the wavefield computation. We explore the resulting pressure distribution within the conduit.

We have observed quasi-continuous seismic tremor phenomena at the surface above hydrocarbon reservoirs. The spectral peaks are in the frequency interval [1-8]Hz. This unique signature diminishes at the rim of reservoirs and is absent at non-reservoir locations. The results of our experiments are remarkably uniform for more than ten major oil producing areas worldwide - regardless if producing or virgin reservoirs were examined. Phenomenologically and structurally, the observed signals are very similar to some previously observed tremor signals originating e.g. from volcanoes, but weaker by several orders of magnitude. At present, the source mechanism of this hydrocarbon reservoir tremor is open to speculations. It might both be similar to some volcanic tremor mechanisms (e.g. excitation by fluid flow), or completely different, such as microscopic surface tension effects or nonlinear filtering of ambient background noise. The talk will focus on the presentation of data from several fields, results of linear and nonlinear analysis, similarities to volcanic tremor signals, and possible source models.

Observation of spectrograms of harmonic tremor from Montserrat, W.I. found the phenomena of gliding spectral lines. These lines are seen in harmonic tremor which follows the build up of repeated hybrid events. These hybrids appear to get so close together that they are no longer distinguishable as separate events.

Repeated events in the time domain produce spectral peaks in the frequency domain, with the relationship dt = 1/dw.These peaks are unaffected by external factors, unless the noise levels get extremely high, or amplitude variation is periodic and on a scale that can be resolved in the amplitude spectrum.

The accuracy of the time separation of the events must be less than or equal to 2%, to allow the resolution of all spikes in the frequency range of 0.5-10 Hz. For peaks spaced at 0.5 seconds, this is an error of only 1/100 second.

>When a spectrogram is taken of period doubling (halving) events, they are clearly recognisable by the appearance (disappearance) of peaks due to the doubling (halving) of the period. Gliding spectral lines can be recreated by gradual reduction of the time gaps between the events.

The gliding spectral lines found in the real data can be directly reproduced by changing the time gap from 1 second to 1/3 second over 12 minutes.

Long period and hybrid events, seen at the Soufriere Hills Volcano, Montserrat, show dominant low frequency content suggesting the seismic wavefield is formed as a result of interface waves at the boundary between a fluid and a solid medium. This wavefield will depend on the impedance contrast between the two media and therefore the difference in seismic velocity. For a gas-charged magma, increasing pressure with depth reduces the volume of gas exsolved, increasing the seismic velocity with depth in the conduit. The seismic radiation pattern along the conduit can then be modelled. Where single events merge into tremor, gliding lines can sometimes be seen in the spectra and indicates either changes in the seismic parameters with time or varying triggering rates of single events.

The differential equation describing the time dependence of bubble growth by diffusion is solved numerically for a stationary magma column undergoing a decompression event. From the volume of gas exsolved, densities and seismic velocities with depth within the conduit and their evolution with time can be obtained. These are used within a finite-difference scheme to model the seismic radiation pattern. The effect of varying velocity can be seen in the synthetic seismograms and spectrograms and compared to models with constant velocity and real data. The time dependent velocity is used to model the gliding lines seen in the spectral content of real data.

The existence of gas bubbles in a magmatic melt introduces damping-related phenomena which can be broken down into three mechanisms, all of which occur in daily life when handling carbonated beverages in closed containers (e.g. champagne bottles!).

(i) rectified diffusion acts when seismic noise/tremor introduces pressure fluctuation in the magma. The inflating and deflating is not symmetrical resulting in a net growth of the bubble radius, and in case of a closed system in a pressure increase of the magma (analogue: shaking of the champagne bottle).

(ii) in case of a sudden unloading (e.g. by lava dome failure) a decompressional wave propagates through the magma which becomes now supersaturated. Diffusion of gas into the bubbles leads to an exponential bubble growth which is in general frequency dependent. Such a sysytem can be represented by a negative bulk viscosity which results in a net amplification rather than damping of the decompressional wave (analogue: letting the cork pop).

(iii) seismic waves in a bubbly liquid experience frequency dependent damping, which is less severe than intuitively expected, provided the seismic frequency (1Hz) is far off the eigenfrequency of the gas bubbles (kHz) (analogue: glasses filled with champagne don't cling).

Volcanic processes operate over a wide range of time scales that require a variety of instruments and techniques for their study. The inversion of broadband data at periods longer than a few seconds may be used to image the forces operating at a source and infer the fluid path geometry and the mass transport balance in a volcano. Analysis of long-period events and tremors at periods shorter than one second offers further constraints on short-term pressure fluctuations resulting from unsteady mass transport, and observations of volcano- tectonic activity tell us about hydraulic/magmatic fracturing processes, as well as brittle response of the rock matrix to stress changes induced by rapid injection and/or withdrawal of fluids.

Small-aperture seismic arrays are required to track tremor and separate source, path, and site effects in tremor wavefields. The new array system is formed by modules of three to six channels of short period geophones, 4.5 Hz natural frequency with electronic response expansion to 1 Hz. The array operate at 200 samples per second at resolution of 16 bit. The data adquisition is provided by a laptop running with Windows operate system. The data analysis provides the azimut and slowness of the input signal.

This system allows the identification of the seismic signal source area and it easily solves the question about the volcanic or non-volcanic origin of several registered signals.

The adquisition and analysis software is writted in C++ using the Borland Builder 5. The analysis of synthetic signals and records obtained in Arenal volcano (Costa Rica) are shown.

Tuesday 18 September - Session 2

In the past decade several of the ash eruptions at Galeras volcano (Colombia) have been preceded by tornillos. These unusual seismic events of unknown origin have screw-like profiles on seismograms and can last up to several minutes. Since 1997, a joint project between the Bundesanstalt für Geowissenschaften und Rohstoffe (BGR) and the Instituto de Investigación e Información Geocientífica, Minero-Ambiental y Nuclear (INGEOMINAS) has supplemented the shortperiod network of the Observatorio Vulcanologico de Pasto with four broadband, three-component seismometer stations along with continuous fumarole gas chemistry measurments, electromagnetic sensors and weather observations in the hopes to learn more about the physical or chemical process which generates tornillos and their significance in the sequence leading to ash explosions.

A suite of tornillos occurred at Galeras Volcano between 08 December 1999 and 11 February 2000. These tornillos appear to be somewhat more complex than the tornillos previously recorded on the broadband instruments or those recorded with the short-period network. They are multichromatic having narrow spectral peaks at up to nine, not necessarily harmonic frequencies. Some of these peaks persist throughout the entire tornillo while others are only contribute to the turn-on transient at the tornillo's beginning.

We investigate this suite of tornillos to characterize the wavefield before proceeding to source modelling. For the lowest frequency, the particle motion is nearly horizontal and fairly linear. Like the higher modes, however, the waves producing the particle motion exhibit some characteristcs of Rayleigh waves, namely ellipticity of varying degrees and an "R*Z" product that oscillates with twice the frequency of the mode. The lack of change of fundamental wavefield parameters such as the polarization for low frequencies during any given tornillo and from one tornillo to the next suggests that the location of the source is stable for this suite of tornillos. The triggering mechanism is probably encoded in the initial episode with many frequencies, while the longlasting coda with few frequencies suggests a resonance.

The many events in the suite of tornillos which occurred at Galeras Volcano between 08 December 1999 and 11 February 2000 were recorded well at both of the crater rim broadband stations, Anganoy (ANG) and Achalay (ACH).

These tornillos appear to be more complex than the tornillos previously recorded on the broadband instruments or those recorded with the short-period network. They are multichromatic, having narrow spectral peaks at up to 9 frequencies, some of which last throughout the entire tornillo. Other peaks only contribute to the turn-on transient at the tornillo's beginning.

We compare parameters measured from this suite of tornillos in each frequency band at the stations ANG and ACH in an effort to determine parameters which may be associated with the source. For most of the tornillos in this suite, the polarization of peaks with frequencies below 5 Hz is consistent with a single source location, although the polarization does not necessarily indicate the direction to the source. For peaks with higher frequencies, the polarization varies from one tornillo to the next. As with harmonic tremor at Lascar Volcano, Chile, this, as well as the lack of a clear relationship between the polarization ,direction and a possible source location, may be explained by scattering in the volcanic medium.

The frequencies present in each tornillo are well correlated between the two stations, indicating that they are not a path or site effect at the stations. Correlation plots of the maximum ground velocity measured at each station and the maximum spectral amplitude show much more scatter than do the frequencies. However the points for the peaks above 2 Hz lie close to a line of slope one, indicating a good correlation. For the 1.9 Hz peak, most points for ACH lie well above this line, nearly double the amplitude of the measurments at ANG. This may indicate that at this frequency, a tornillo has a distinct radiation pattern. We hope to investigate this further using data from more stations. The decay of amplitude measured at both stations also correlates well, indicating that this parameter may be attributed to the source. While none of these observations give us a clear picture of the source process of tornillos, they provide additional pieces we can add to the puzzle.

The Torfajökull central volcano in south Iceland is an elongate rhyolitic complex rising about 500 m above the surrounding basaltic landscape, 450 km² in area. It is the largest silicic centre in Iceland with a caldera, 12 km in diameter, and an outstanding high-temperature geothermal field. During the last 1100 years there have been three eruptions in the Torfajökull area, the latest at the end of the 15th century.

Torfajökull is a source of high and persistent seismicity, both high frequency tectonic earthquakes and low frequency volcanic earthquakes occur. The different types of seismicities seem to be caused by different physical processes and do not have any apparent spatial or temporal correlation. The high frequency events concentrate to the western part of the caldera and the low frequency events to the southern part of it. We have studied the both forms of seismicity with the data gathered by SIL, the digital Icelandic seismograph network maintained by the Icelandic Meteorological Office, added with two local analogue seismograph stations.

In July 1991 - October 1995 we located about 160 high frequency earthquakes in the Torfajökull area, with magnitudes (M_L) between 0.4 and 2.8. The epicentres were mainly in the western part of the caldera and west of it, in the east there was almost no activity of this sort. These hypocentres were located at all depths from the surface down to 14 km, with highest activity at 5 - 12 km. Inside this cluster, in the northwest part of the caldera, is a spherical volume void of earthquakes, about 4 km in diameter and centered at 8 km depth. This we have interpreted to be a cooling magma body. Thus the high-frequency earthquakes at Torfajökull are a result of thermal cracking around this body, which is being cooled by percolating groundwater. The aseismic volume at the centre of the hypocentral cluster is at a relatively high temperature, above the brittle - ductile transition, and cooling at a low rate.

Low frequency earthquakes occur typically in swarms at Torfajökull, lasting from one to a few days and consisting of up to a couple of hundreds events per day, mainly rather small in size. The swarms show no clear pattern in their temporal occurrence. A common feature is their relatively abrupt beginning; the activity rises typically from rather low levels to its highest peak in a single day. Between the swarms occasional single events can occur, or there can be quiescences up to several days or weeks.

The low frequency seismicity is best visible on the paper seismograms of the nearby analogue seismograph station, but the SIL network can at present detect and locate the largest ones of these events. We have gathered a data file of about 200 low frequency volcanic earthquakes at Torfajökull during the years 1996 to 2001. It is difficult to obtain good locations for these events. Their P waves are characteristically emergent; the S waves are typically somewhat clearer and the phase picks seem to fit more or less with the location. According to our preliminary analysis these events seem to be focused in the southern part of the caldera, and thus point to a location of an active magma chamber there. The exact process causing these events is still unclear. process of tornillos, they provide additional pieces we can add to the puzzle.

In the Tjörnes Fracture Zone (TFZ) north of Iceland two strong anomalies of high seismic b-values have been found, which coincide with the locations of recent volcanism or hydrothermalism. The connection to magma chambers in a depth of 5-10 km was expected owing to the results of similar b-value anomalies benath volcanic edifices such as Mt. St Helens.

On first view, the existence of a magma chamber between Kolbeinsey and Hóll seamount appears to be confirmed by reflection seismic surveys conducted in the area, which show a clear deep reflector at around 3.5 km beneath the seafloor. However, this reflector is best imaged in a migrated record at 1500 m/s, i.e. the speed of sound in water, and it can be modelled by single scattering from the bathymetry in the area, which leads us to the conclusion, that the apparent reflection images the seafloor rather than a magma chamber.

The existence of a southern magma chamber south of Grimsey appears to trigger hydrothermalism in the Grimsey hydrothermal field and relatively recent volcanism along the Manareyjar Ridge with the last eruption in 1868. However, the presence of hydrocarbon homologues, large production of talc and high silica contents in the sea water hint on a serpentinization process typical for a transform zone along the Mid Atlantic Ridge. Serpentinite prevents the build-up of large stresses and is likely responsible for the high b-values in the subsurface of the region. Even in a volcanic area as the TFZ, neither the existence of high b-values nor an apparent deep reflector conclude the existence of a magma chamber. Instead, the seismic imaging of magma chambers in ridge areas appears to be questionable.

A seismic swarm of tectonic earthquakes occurred in a crustal region close to the island of Stromboli from 6 to 17 June 1999. Eighty-four earthquakes with duration magnitude M_d of between 1.5 and 3.2 were recorded at the Aeolian Island Seismic Network stations, deployed in Southern Tyrrhenian Sea. The locus of the activity was Northeast of the island at a depth of less than 12 km. This source region was identified using standard methods of hypocentral location integrated with azimuth analysis. We propose a classification in families of these earthquakes on the basis of their waveforms and frequency content.

The occurrence of the swarm interrupted a period of relative quiescence of tectonic seismicity lasting about five years. In this time span, we recorded volcanic tremor and explosion quakes, which represent the most typical form of seismic activity on this volcano. Indeed, seismic rate and energy release of earthquakes related to brittle failure of rocks are usually low in comparison to other basaltic volcanoes with persistent eruptive activity. We highlighted such evidence in previous studies, according to both the continuous seismic monitoring since 1985 and data from historical catalogs of earthquakes over the last 100 years. Consequently, we postulate that the peculiar high number of earthquakes recorded during the 1999 swarm and the total energy release associated may represent a case history. In light of this hypothesis, we also analyze the state of volcanic activity and discuss its possible temporal relationship with the June 1999 seismic swarm.

Thursday 20 September - Session 3

In the present work we show results of seismic experiments carried out in three different volcanic areas, Deception Island (Antarctica), Mt Vesuvius (Italy) and Kilauea (Hawaii), using seismic antennas, where a strong influence of the surface and topography structures has been observed over the propagation of the seismic waves.

Data from Deception island volcano were collected by two short-period seismic arrays, and during the experiment the instruments recorded several earthquakes related to both local volcano-tectonic activity and regional dynamics. Seismograms of regional earthquakes reveal a marked difference in the apparent velocities measured at the two array sites. The analysis of the data with different procedures indicates a marker difference in the seismic velocities for the shallower 200 m beneath the two array sites. Based on the morpho-structural characteristics of the volcano, the inferred velocity discontinuity maybe associated with the ring-fracture system bordering the collapsed caldera structure that extends over the inner part of the island.

Data from Vesuvius volcano belong for a deep seismic sounding experiment and recorded by a dense short-period three-component seismic array. The aim of the data analysis was to study the portion of the seismograms long after the first P onset in order to obtain the slowness and back azimuth distribution of correlated secondary phases. Results show that correlated phases detected in the coda of the seismogram shots are composed of surface waves. Location of the scatters position shows a high concentration of them around the volcanic edifice of Mt. Vesuvius, confirming that topographical structure plays an important role in the generation of the high-frequency scattering waves.

Data from Kilauea volcano were collected using three seismic arrays deployed on the caldera floor. Apparent slowness and azimuth of very local events are inconsistent among the three seismic antennas: no unique source position and depth fits the three antenna solutions. The introduction of the effects of shallow structure and the caldera topography on synthetic seismograms can explain these inconsistencies and provide an unique solution for the source hypocenter. The systematic use of this procedure for volcanic tremor and events of Kilauea volcano permitted to obtain a three dimensional image of the focal volume.

Villarrica, one of Chile's most active volcanoes, rises above the lake and town of the same name. It is the westernmost of three large stratovolcanoes that trend perpendicular to the Andean chain. A 6-km wide caldera formed during the late Pleistocene, >0.9 million years ago. A 2-km-wide postglacial caldera is located at the base of the presently active, dominantly basaltic-to-andesitic cone at the NW margin of the Pleistocene caldera. About 25 scoria cones dot Villarica's flanks. Plinian eruptions and pyroclastic flows have been produced during the Holocene from this dominantly basaltic volcano, but historical eruptions have consisted largely of mild-to-moderate explosive activity with occasional lava effusion. Lahars from the glacier-covered volcano have damaged towns on its flanks.

Seismic activity shows frequent events associated to strombolian activity and tremor periods, sometimes very intenses. The activity under volcano is reduced, occurring less than an event per month. Frequent tectonic events are detected at 200-300 km deep

As of 20 September 2000, an increase in seismicity occurred while gas and steam emissions ceased, and the usually persistent fumarole disappeared. During 25-28 September, seismic stations VNVI, 4 km from the crater, recorded anomalous seismicity characterized by clusters of low-frequency earthquakes, harmonic tremor, and tremor bands. OVDAS indicates that the zone near the crater was dangerous. A flyover of the summit was performed on 4 October to observe any changes in activity. Fumarolic emissions remained absent, and the crater floor was obstructed by black rock. The bottom crater and walls were covered with patches of snow, indicating relatively low temperature. That day seismicity showed tremor bands again. By means of the analysis of the seismic activity evolution it was possible to predict the explosive event occurred on october 29th, 2000.

In the present work we show preliminary results of the analysis of seismic attenuation in Deception Island volcano (Antarctica) using local volcano-tectonic earthquakes. Data base used in the present work corresponds to events of a seismic swarm occurred in Deception Island in January-March 1999. Also, the determination of their magnitudes using local magnitude, M_l, and momentum-magnitude M_w (for P and S waves), the calibration of a magnitude-duration scale, M_d, for the volcano and comparison among all scales are studied. The objective is not only to quantify the size of the selected events, but also to study how the different magnitude scales can be applied to volcanic environments and the special consideration have to be take into account to apply such a scale for the volcanic areas. Also alternative ways to estimate the magnitude (manually and automatically) of the events will be presented and the relations among them will be discussed.

The attenuation study was performed in the high-frequency range (between 10 and 35 Hz) using different techniques and seismic waves. For P-waves we used the broadening P-wave pulse and the spectral methods to estimate the Qp factor. For S-waves we apply the above methods and also the Code-normalization technique, and for coda waves we use the single scattering theory to estimate the attenuation factor. Using the information of S and coda -waves attenuation we apply a procedure to separate the intrinsic and scattering attenuation effects on the seismic waves.

Results reveals a very low Q values, also at high frequencies, similar to those observed in other volcanic areas. For example, Qp value ranges between 13 and 28, Qc follows laws as Qc = 13 f ^0.86 or Q_d ranges between 15 and 60. The separation of intrinsic and scattering effects indicates that the scattering phenomena at high frequencies seem to be the most important mechanisms of attenuation of Deception island volcano.

One of the major problems that confronts volcanologists when working with active volcanoes which have been quiescent for some hundreds to a few thousand years, is to try and identify the relevant precursory signals that will precede the next phase of eruptive activity. This is the case of Fogo volcano, located on the island of São Miguel, in the Azores archipelago.

Fogo Volcano is the largest of three active central volcanoes (Sete Cidades, Fogo and Furnas) and it dominates the central part of the island, located at the intersection of E-W and NW-SE fault systems. It has a complex morphology with a summit caldera that appears to have formed as a result of numerous collapse events, the most recent one occurring as a result of the intracaldera eruption of 1563. The volcano during its history has given rise to a number of different styles of eruptive activity. This is illustrated by the historic activity in the years of 1563 and 1564, that ranged from an effusive basaltic event, on its flanks, to the plinian explosive trachytic eruption, inside the caldera. Historical records have also numerous accounts of important seismic activity, in particular the 1522 earthquake that triggered a major landslide, killing almost 5000 people.

Analysis of the frequency of eruptions and evaluation of the hazard posed by the eruptive behaviour of Fogo volcano, degassing through the flanks, failure of the crater lake, other aspects of edifice instability (in particular associated with the heavy rainfall) and seismic activity, has enabled a preliminary assessment of the volcanic risk to the surrounding communities and infrastructures to be undertaken.

>Despite the knowledge of the volcanic geology and the programme of monitoring that is currently in place, the important question that, in common with other volcanoes for which there is no past experience of instrumental monitoring of an eruptive event,: ..are we prepared for the next eruption? remains.

A new macroseismic catalogue and database of earthquakes on Mt. Etna from 1832 to 1998 was prepared and recorded on CD-ROM. The need for creating a catalogue specifically for this region comes out from the observation that the traditional parametric catalogues do not allow a detailed characterisation of seismicity, for instance in terms of seismic sequences or seismotectonic analyses. On the other hand, the huge quantity and the quality of available historical information allow investigating these aspects. The performed analysis has permitted to associate about 550 earthquakes with seismogenic structures so that it is possible to investigate fault behaviour in time. The catalogue, compiled through the analysis of some 200 contemporaneous sources, lists 1735 earthquakes, 168 of which produced damage. 7325 macroseismic observations are available, the intensities was assessed using the EMS 98 macroseismic scale.

For each event the focal parameters (epicentral location, maximum and epicentral intensities, macroseismic magnitude) are defined. The seismogenic fault, the co-seismic surface faulting, as well as the instrumental values of magnitude and depth are also reported when available. The CD ROM is a searchable catalogue for both epicentres and intensity maps of all the events. In addition, queries about some different possibility of earthquake selection (by date, intensity and coordinates) can be performed. The seismic histories for 426 Etnean localities are also available.

Most of existing catalogues show lack of continuity between the instrumental and pre-instrumental period. For Etna area, instrumental records of earthquakes and both low-frequency events and tremor, are available. The data collected in past 25 years by different institutions need to be preserved by creating a digital record having a homogeneous standard. Aim of this work is to integrate both macroseismic and instrumental datasets achieving the continuity between them and creating an overlap of about ten years

A list of earthquakes (3085) located in the Etnean area from 1976 up to 1989 is reported adopting the hypo71 standard. The input arrival time lists of p and s (when available) phases are included as well. Located shocks can be selected from the database and represented either subdivided by year or by different swarms.

The database also contains the features of volcanic tremor recorded in the time interval 1977-2000. The tremor catalogue file shows the trend of the overall spectral amplitude as well as the spectral amplitude integrated in four different frequency ranges in order to set into evidence possible differences in the spectral shape. Moreover, a list of the first six dominant spectral peaks and the corresponding amplitudes is reported. Examples of diagrams summarizing such information are shown.

Furthermore, the instrumental database CD will include a file where all the volcanic phenomena occurred on Mt. Etna from 1975 till 2000 are collected and classified. The data come from the reports of the visual observations described in the official bulletins.

The integrated database so far obtained should be a powerful tool for understanding the features of both volcanic and seismic activity on Mt. Etna. It spans indeed over a considerably large time interval, so that many different case histories can be analysed and better compared.

Logical modeling in volcanology opens new opportunities of forecast and reconstruction of volcanic activity. Presented is the draft of conceptual model of formation of subaerial erupted unit in terms of propositional logic. The model uses observations of different types of eruptions published in the literature. Successions of events that occur during an eruption can be represented as a formula deducible from special axioms, and the geologic result (an assemblage of deposits or landforms) as a formula that needs verification. In terms of this model, forecast is making inferences from the formula of eruption, and reconstruction is the search of appropriate inference for existing formula. In the latter case, the model should tell how many and what eruptions occurred and, desirably, assess the probability of different variants of reconstructed history.

In the future the model will be elaborated in the following directions:

Involvement of axioms describing observations of non-eruptive processes that influence volcanic products;

Involvement of physical models of eruptions and post-eruptive petrogenesis and landform evolution;

Passage from propositional to predicate logic that allows more accurate description of processes;

Passage from classical to non-classical (possibly, fuzzy) logic for better account of ill-defined processes and objects;<

Adjustment for individual eruptive centers and/or particular volcanic provinces.

Logical modeling suggest similar language for observation and modeling results. This would hopefully introduce explicit structure into the world of concepts of volcanology, help a researcher navigate in it and bring closer "traditional" and "physical" volcanologists, experts and students, scientists and public.

Thursday 20 September - Session 4

Pyroclastic flows from the Soufriere Hills Volcano, Montserrat, present a hazard to populated areas if they travel down the northern or western sides of the volcano. However, due to persistent low-cloud and winds from the east, it is frequently impossible to observe in which direction the dome is failing. What is needed is some method of determining the direction of dome failure based on continuous near-real-time data.

Our method is to use seismic amplitude data from 8 high-dynamic-range seismic stations in order to determine the trajectories of rockfall and pyroclastic flow signals. A grid search algorithm is used. At each grid position the reduced displacement (corrected for attenuation) for each station is calculated. The final location is chosen as the one which has the smallest standard deviation of reduced displacement. There is a good agreement between trajectories determined using this method and visual observations.

The products of this method will include hazard maps and estimates of run-out distance and collapse volume. Our ultimate goal is to be able to determine pyroclastic flow trajectories in near-real-time, which could be linked directly to an alarm system.

Previously, comparison of tilt with the seismic data has been done using RSAM (seismic amplitude) data. RSAM peaks have been matched with the peaks in the tilt cycles. When these cycles are looked at more closely, it can be seen that sometimes the peaks match and at othertimes the RSAM precedes the tilt and vice versa.

The RSAM data contains all seismic data, including rockfall, hybrids, long period events, etc., so really needs to be compared with the raw seismic data, to see how the individual events match up with the cycles.

In the raw data, some cycles fit beautifully to the ideal build up of hybrids before an eruption, a big eruption at the peak of the tilt cycles, with little seismicity during deflation. Others show activity all through the cycle, and others no activity at the peak of the cycle.

Parameters from the seismic data, dt, duration and maximum amplitude, show a little of how that types of events relate to each other. Large events usually come close after small events and small events come after large time gaps.

Separating out the different types of events, and their parameters, and comparing them to the tilt cycles, and each other, will allow greater understanding of the different seismic events and how they relate to the activity of the volcano. Statistical analysis of large volumes of this data will help constrain the physical models that are being developed.

The possibilities of using three-dimensional finite difference (FD) methods for numerical simulation of the seismic wave field at active volcanoes have been investigated. Special emphasis has been laid on the implementation of the boundary conditions for free surface topography.

We have compared two different approaches to solve the boundary conditions. The algorithms have been implemented on parallel hardware and have been tested for correctness and stability. They have been applied to smooth artificial topographies and to the real topography of Mount Merapi, Indonesia. We conclude, that grid stretching type methods are not well suited for realistic topography as they tend to be unstable. The representation of topography through staircase shaped grids results in stable calculations.

The simulations show the effects of a three-dimensional surface topography on elastic wave propagation. Numerical studies like this can help to understand wave propagation phenomena observed on field recordings in volcano seismology. Future studies will aim at separating the effects of media scattering, topography and source effects.

I use the data of the TOMOVES active seismic experiment to study scattering effects in the shallow heterogeneous structure of Vesuvius volcano. Seismograms from shots located on the volcano itself are characterized by spindle-like envelopes, small or missing P-onsets, missing S-onsets, and long codas. Seismograms from shots in the surroundings of the volcano, on the contrary, show clear and impulsive P- and S-onsets and short codas.

These different shapes of the envelopes can be explained by strong multiple scattering in the shallow heterogeneous eruptive material and single scattering in the surrounding crust, respectively. I model the seismograms of shots located on the volcanic mountain using the diffusion model. As a result I obtain a diffusivity of d = 0.1 km**2/s for the frequency range of 2 to 16 Hz used in this study.

Assuming the dominance of S waves in the coda and a typical S-wave velocity of around 1.5 km/s this corresponds to a mean free path of only 200 m. These results show, that multiple scattering is an important effect, which cannot be neglected in the modeling of seismic wave propagation in volcanic environments.

Much effort as been laid on the interdisciplinary interpretation of monitoring parameters within the Indonesian-German MERAPI-Project. I.e., seismicity and gas exhalation are regarded as useful parameters for the evaluation of the activity state of a volcano. The correlation between geochemical parameters of volcanic gases and seismic activity has often been postulated, mostly motivated by widely accepted ideas about the seismic source processes at volcanoes and their relation to magma/gas movements in the active volcanic feeding systems. However, it is difficult to prove a direct correlation between volcanic gas measurements and the occurrence of single seismic events which is mostly due to the sparseness of gas measurements in time (occasional sampling with intervals of days or weeks).

The data set of monitoring parameters collected at Merapi from August 2000 to January 2001 contains both continuous records of the gas temperature with a sufficiently high sampling rate (every minute) and parallel recorded broadband seismic records. The temperature sensor is located in a fumarole at Woro gas field around 150 m from Merapi's active lava dome. The seismic broadband sensor is installed at a distance of 600 m from the active dome region.

The results of the correlation analysis between these two time series show a positive correlation between the increase of around 5-10 K of the fumarole temperature within less than 10 minutes after the occurrence of clusters of seismic events. Additionally, in order to estimate the influence of meteorological conditions on the individual parameters, both the rain amount and the barometric pressure from several stations at the flanks of Merapi are analyzed. The rain induced changes in fumarole temperature and seismicity are shown and discussed.

The 1991 eruption of Hekla occurred in January 17 - March 11, and produced about 0.15 km³ of lava. The 2000 eruption took place in February 26 - March 8, and its preliminary lava volume estimate is 0.11 km³. This study is based on digital data from the SIL seismograph network of the Icelandic Meteorological Office, supplemented by three analogue stations near Hekla.

Both these eruptions were similar in many senses: small in scale, and occurring about ten years after the previous event. No long-term preceding seismic activity was observed. In both cases the seismicity started with tiny local earthquakes shortly before the eruption. In 1991 the first ones were observed about 30 minutes before the onset, coinciding with strain observations of the start of an intrusion. In 2000 the first, very small earthquakes were observed some 80 minutes before the onset, on the paper seismograms of an analogue station on the flank of the volcano. Similarly to 1991, the strain signal started about 30 minutes before the onset. Apparently the seismicity had started earlier also in 1991 but was not seen, as the station at Hekla was out of order then and the nearest operating seismograph station was 30 km away.

The earthquakes at the onset are small but numerous, the time between subsequent events may be less than a minute. The magnitudes grow in size from the beginning until the onset of the eruption, from about M_L magnitude 0 towards M_L 1.5 - 2. The volcanic tremor enters to the seismograms at the very onset of the eruption. During the first couple of hours the eruption is violent and the tremor amplitude is high. It thus sets a threshold to the observable earthquakes. The numerous and frequent earthquakes keep on continuing, being up to 2.5 - 3 in magnitude.

The locations of the 1991 earthquakes were not good enough to make conclusions of the spatial development of the seismic activity. The preliminary analysis of the 2000 earthquakes shows that there was no proceeding front of earthquakes ahead of an intrusion, but the earthquakes rather occurred diffusely in a larger volume.

After a couple of hours the tremor starts to decrease, together with the settling eruptive activity. When the explosive onset is over, the eruptive activity consists mainly of peaceful streaming of the lava and occasional gas bursts. The intensive earthquake activity calms down, as well, and in later phases of the eruption only very few earthquakes occur at Hekla. However, the tremor continues with minor amplitude throughout the eruption and dies down together with it.

The tremor consists of low frequencies. We have analyzed the tremor during the first hours of the 1991 eruption, observed by a digital station in 30 km distance. It has a dominant peak at about 0.7 - 0.75 Hz during the first one and a half hours. Then it wanders to about 0.85 Hz there it remains the next few hours. Tentative observations of the frequency content of the tremor at the onset of the 2000 eruption point to similar frequencies.

Friday 21 September - Session 5

Canarian Volcanism is one of the most interesting of the planet. Taking into account its variety and complexity, this conference presents a brief summary of its geodynamic framework, lithological variety, eruptive mechanism and associated structures etc. The relation between the Canarian population and its natural volcanic environment will be discussed both as a cultural element and potential hazard. The present surveillance network and the strategies for the mitigation of volcanic risk will also be described.

The Instituto Geográfico Nacional (IGN) is the institution in charge for providing the geodetical and geophysical infrastructure at regional scale to monitor geophysical activity. This work presents the different geophysical and geodetical networks existing in the archipelago.

Seismology: The seismological network is composed of 7 short period vertical component seismometers linked by telephone dedicated lines (mainly to the Centro Geofísico de Canarias and to the Seismic Data Reception Centre at the IGN in Madrid). This network is nowadays being improved with broadband three-component seismic stations. At the moment, two stations are operating (Tenerife, El Hierro) using real time satellite transmission and other four are being installed in other four islands.

Geomagnetism: The Güímar Observatory (GUI) in Tenerife provides absolute and relative measurements. This Observatory, integrated in the INTERMAGNET Project, disseminates data in near real time basis. There is also a network composed by 15 stations to control the secular variation of the geomagnetic field.

There is a multi-institutional tide gauge network composed by 6 stations using a common format to interchange data.

GPS Network: The main goal of the REGENTE Project is to cover all the spanish territory with a high precision three dimensional geodetic network. Every sheet of the Spanish map at the scale 1/50.000 contains, at least, one REGENTE point, belonging to the Conventional Geodetic Network (ROI) or to High Precision Levelling Network (NAP). The total number of stations (1.200) covers all spanish territory, including Canary and Balearic Islands. Observations began in 1994 and were performed with 9 GPS receivers (double frequency and 6th observable) during three hours and recording data every 15 seconds.

Precision Levelling: All the Canary islands have already precision levelling lines with ortometric altitude, GPS and gravity observations. New geoids have been also calculated for each island.

Differential GPS: The main objective of the DGPS projects is to establish and to implement a public service to provide terrestrial positioning to GPS users of the spanish community with criteria of precision, integrity and availability. The differential corrections are delivered following international accepted formats, compressed in RASANT, with free access to all users.

Permanent GPS Stations:here are two stations running in Canaries: Maspalomas (Gran Canaria) and Roque de los Muchachos (La Palma). The IGN is since September 2001 one of the Local Analysis Centres of EUREF.

Data analysed in the present work correspond to a 40 days field experiment carried out in Teide volcano (Canary Islands, Spain) with two short-period small-aperture dense seismic antennas in 1994. The objective of this experiment was to detect, analyse and locate the local seismicity. We analysed also the background seismic noise to investigate the possible presence of volcanic tremor.

From a set of 76 events, we selected 21 of them in base of their good signal to noise ratio and their possibility to locate their seismic source by using the seismic antennas. A visual classification in base of the S-P time and seismogram shape has permitted us to establish three groups of events: local seismicity (S-P time between 3 and 5 seconds), very local earthquakes (S-P time smaller than 3 seconds) and artificial explosions.

These earthquakes have been located by applying the Zero Lag Cross-Correlation technique and the inverse ray-tracing procedure. Those earthquakes that were recorded simultaneously by both seismic antennas were also located by intersecting both back-azimuths. The analysis of the seismicity has revealed that the amount of seismicity in Teide volcano is moderate. This seismicity could be distributed in three main areas: Inside of the Caldera edifice (below the Teide-Pico Viejo complex), in the border of the Caldera Edifice (possibly related with the 1704-05 eruption) and in the border of the island.

At the present, this activity is the only indicator of the volcano dynamics. The analysis of the back-ground seismic noise has revealed that, at frequencies lower than 2 Hz, the Oceanic Load signal is predominant over other signals, even over local earthquakes with magnitude 2.0. Due to that, although if in the Teide area were present a week volcanic tremor, or other volcanic signals with predominant peaks below 2 Hz, to observe them could be a very difficult task.

Ambrym is a large basaltic volcano with a 12-km-wide, 1,900 years old caldera, sited in the Vanuatu archipelago, formerly known as New Hebrides. Two cones are currently active in the caldera, named Marum and Benbow, each with several active vents that hosted visible lava lakes until 26 November 1999, when an earthquake (Mw = 7.5) located few km S of the volcano caused important changes to the vents morphology.

During July 2000, we installed a short period seismic station on the Eastern rim of Benbow crater in the Western side of the caldera. The station acquired continuous data for about one month. During this period volcanic tremor was recorded together with discrete events. In this talk first results from the analysis of these data will be discussed, focusing the attention in particular on the existence of transitions in the long term volcano dynamics. During the observation period, on 9 August 2000, another significant earthquake was recorded, (mb= 6.3, about 65 km N of the volcano). Possible influences of this event on the volcano dynamics are investigated.

The Jovian dust streams are high-speed bursts of submicron-sized particles traveling in the same direction from a source in the Jovian system. Since their discovery in 1992, the Jovian dust streams have been observed by three spacecraft: Ulysses, Galileo and Cassini. The work presented here describes an emerging electrodynamical picture of the Jovian dust streams as they appear inside and outside of the Jupiter environment. The source of the Jovian dust streams is Jupiter's moon, Io, in particular, dust from Io's volcanoes. Charged Io dust, traveling on trajectories from Io's location, is shown to have some particular signatures in real space and in frequency space.

The Jovian dust stream dynamics in the frequency-transformed Galileo spacecraft dust measurements show different signatures, varying, orbit-to-orbit during Galileo's last 29 orbits around Jupiter. The presence of Io's orbital rotational frequency demonstrates that Io is a localized source of charged dust particles and a confirmation of Io's role as a localized charged dust source arises through the modulation effects. This time-frequency analysis is the first direct evidence that Io is the source of the Jovian dust streams. I provide additional frequency evidence of Io recorded by the Cassini and Galileo spacecraft during an August 2000 Jovian dust streams `storm'.

Several aspects of the dust stream particles' dynamics in real space can be understood if the particle's charge is varying, via the different currents generated as the dust particle samples the plasma while traveling. One application I show is matching the travel time of a stream particle during the December~2000 joint Galileo-Cassini dust stream measurements, where the two spacecraft were located inside (Galileo) and outside (Cassini) of the Jovian magnetosphere. To match these measurements, the smallest dust particles could have the following range of parameters: radius: 6 nanometers, density: 1.35--1.75~g/cm^3, secondary electron emission yield for a submicron sulfur particle: 3.0, dependent on a maximum electron energy 300 eV and a photoelectron emission yield: 0.1--1.0, which produce dust particle speeds: 220/450 km-s^{-1} (Galileo/Cassini) and charge potentials: 5.5/6.3 V (Galileo/Cassini).

The monitoring of active volcanoes using automatic recording devices is particularly prone to data losses because of the inherent geological and environmental conditions encountered in situ. However, most analysis methods applied in the investigation of medium-to-long term volcano dynamics, e.g. statistical methods, require time series without gaps in the observations. An appropriate reconstruction of the missing data in the recorded time series is, therefore, essential.

Classical interpolation methods are not suitable for filling gaps in time series; their smoothing properties do not preserve the observed variability of the data. Furthermore, these methods of dominant polynomial nature can usually not be calibrated to the analysed observations. In response to these requirements, an approach based on geostatistical concepts is proposed that allows time series reconstruction by stochastic simulation with the following capabilities: (a) honouring temporal auto-correlation, (b) conservation of the observed variability and (c) conditioning of the data and of their histogram.

The developed approach is applied to seismic data monitored at the Stromboli volcano. The obtained results demonstrate the possibilities of this approach; a time series missing 20 % of its values can be reconstructed without gaps while preserving its temporal behaviour in a statistical sense.

Friday 21 September - Session 6

The photogrammetry can be defined as a set of operations which allows to get three-dimensional metric informations of an object from the analysis of frames or digital images of the same object taken by different points of view. The digital photogrammetry is very useful to study the volcanic areas, giving products as Digital Terrain Models (DTM) and orthophotos.

In 1996 a photogrammetric flight was performed on Vulcano island (Aeolian islands - south Italy); the DTM using the middle scale 1:10000 frames was realized. The Socet Set software is used to orientate the stereoscopic model in the digital photogrammetric workstation using 19 ground control points. However, the full automated DTM has some bugs, where the vegetables and buildings cover the ground. Therefore, some complex editing operations must be performed to eliminate trees and buildings, allowing the comparison among different DTMs (the time scale can change too).

The orthophotos too can be generated automatically but, again, editing operations are necessary to correct some effects due to clouds and long shades.

The repetition of photogrammetric flights allows to compare DTMs of the same zone referred at different periods in order to monitor soil and crustal movements. The DTMs accuracy depends on flight altitude, on scanning resolution and on density of ground control points.

Orthophoto of Vulcano island ("La Fossa")

DTM of Vulcano island (shaded relief)

This research is related to Global Positioning Systems applications on volcanic areas in order to monitor small ground deformations. We stress out the GPS performances in continuous static acquisition. The Survey and Geomatic Laboratory of the Padua University monitor two GPS networks on italian volcanic areas: the first one, on the Euganean Hills (NE of Italy) with 9 vertexes which cover a 300 Km² area, the second one on the Aeolian Islands (SW of Italy) with 8 vertexes on 1500 Km² area.

We approach this study from the network establishment to the data processing, in order to compare different temporal sessions. Double frequency geodetic GPS receivers were adopted with relative-static positioning using carrier phase measurements. In the processing phase, data coming from the GPS permanent stations are used.

To compare results of different and not directly comparable sessions, also using "historical" data, we adopted the seven parameters Helmert's transformation (3 translations, 3 rotations and 1 scale factor). Using this approach, we present the results related to the Euganean Hills network (1993 and 2000/2001 survey campaigns) and Aeolian Islands network (2000 and 2001 campaigns).

The airborne Laser Scanning system is a powerful tool to obtain data for DTM (Digital Terrain Model generation). It uses a laser ray to scan the ground and capture data. This technique has been used for Neapolitan volcanic area which is considered one of the active volcanic zone in Italy and has monitored for long years.

The laser scanning system mainly consist of two sensor group: the first is the Laser Range Finder used for measuring the distance from itself to the terrain surface. GPS and INS constitute the second sensor group which determines the position and attitude of the Laser Range Finder at the time of measurements.

TopEye laser system has been used during the experiments; this system is mounted on an helicopter to perform different types of surveying at different altitude with different features (density of points per meter squared, different scan angles, etc..). The system scans the ground across the track of the Helicopter and measure the distance to the ground with up to 7000 laser pulse per second and it is able to record four echoes for each laser pulse.

The experiments have been done on two areas: the first survey has been performed effecting 10 flight lines on Phlegrean Fields, at flight height 195m above ground; the second experiment has been done effecting 7 flight lines on Ischia Island (helicopter altitude at 750m and 195m).

In the Laser Scanning data processing, the first step is to obtain the helicopter trajectory by using GPS processing software such as GeoGenius software. Subsequently the data are elaborated from the software of the TopEye system which combines GPS solution (position of helicopter at 1 Hz), INS data (attitude at 50Hz), laser range and mirror scan angles to calculate the coordinates of the laser foot print on the ground for each laser pulse. The final position is given in WGS84 geocentric coordinates. To distinguish between the ground and non-ground data, a software should be used to classify captured data (Terrascan by Terra Solid).

For these experiments five reference stations have been used: MORT and MEZZ at Ischia Island, RITE and POSO at Phlegrean Fields, and then all four stations have been connected to ACAE permanent station which has known coordinates in WGS84 system. The four reference stations have been processed in static mode respect to ACAE permanent station to obtain an high quality set of coordinates in WGS84.

Finally the contour lines and the triangular irregolar network have been created using tools of the TerraModeler software package. The rendering Phong's algorithm has been used to generate the final DTM.