Surveillance of Thera Volcano, Greece: Microseismicity Monitoring

The results obtained indicate that most recorded events are of volcanic origin. However, when tested against the established classification parameters, they exhibit a mixture of type A and type B earthquake characteristics (Minakami 1960). This indicates that the sources of these events exist in domains with characteristics similar to those of a volcano in the middle eruptive stage. The boundaries of these domains were approximately defined using shear wave attenuation data. In this way, two magma chambers were located, one underneath the area of the Kameni islands and one, hitherto undetected chamber, underneath the Kolombos submarine volcano off the north-east coast of Thera.

INTRODUCTION

The volcano of Thera has been very active throughout its history. It is mainly known for its enormous explosion of the 15th century BC, destroying the Minoan civilization and creating the large caldera we observe today. Geological studies (Ktenas 1927; Reck 1930) place the first eruptive activities of the volcano approximately 1.5 mA ago. A lot of hypotheses, based on historical, archaeological and geological data, have been put forward to describe its pre-Minoan history. They deal with the displacement of volcanic centres, as well as changes in the shape of the island (Reck 1930; Heiken and McCoy 1984; Georgalas 1940). An outline of the post-Minoan history of Thera produced from ancient and recent chronicles is given in Table I.

It is evident that such an active volcano requires systematic and detailed study in order to understand its behaviour and assess associated volcanic hazard. Surveillance of volcanoes was first proposed in the early days of eruption prediction studies as the method of investigating precursory phenomena. Among the first methods used was the establishment of microseismic networks (Sassa 1936; Minakami 1960) in the area of volcanoes.

Volcanic eruptions and the accumulation of volcanic signs have always been considered to have a close relationship (Shick 1988). It is well known that volcanic activity is connected closely with pre-and post-seismic activity.

In this paper we present the results of a microseismic experiment in the area of the Thera volcano, using data recorded from the seismological network established in 1985 and operated sporadically until 1988.

SEISMOLOGICAL NETWORK

In order to study the microseismic activity in the vicinity of Thera, a small network (Fig. 1) of four vertical seismometers was used, from April 1985 to January 1986. The network was re-used at intervals over the next three years, during the times shown in Table II. As can be seen in Fig. 1, the network is roughly circular, with its centre near Nea Kameni island. This was considered to be appropriate for the best possible estimation of hypocentral parameters of earthquakes generated in the area of Nea Kameni. Stations OIA and AKR were placed at the ends of Thera island in a NE-SW direction. This is identical to the direction of major faulting and great lava flows. With such a network, it would be improbable to miss micro-earthquakes occurring along this axis. Although the stations were placed away from inhabited areas, the level of ground roll (industrial or natural) was quite high. Thus it was not possible to use high gain in order to capture small amplitude events. Almost all seismographs worked with a gain of 60-66 db, except for KAM, which sometimes worked with 72 db. For the better location of hypocentres, especially of earthquakes occurring outside the perimeter of the network, we also used data from the (3 component) station at Aperanthos (APE) on Naxos (approximately 65 km to the north of Thera), which is part of the National Seismological Network.

During the periods of network operation a total of 109 micro-earthquakes were recorded, of which only 49 events were recorded at three stations simultaneously. The rest were recorded at one, or at most two stations. For the estimation of the hypocentral parameters of the above 49 earthquakes, the program HYPOCENT (HYPO-71 version, for use on IBM-PC compatibles) was used. The velocity model used is shown in Table III and was taken from Makris (1977) (seismic profile along the line Evia-Amorgos). The estimated parameters of these earthquakes are given in Table IV. In the same Table are included the maximum amplitudes of P and S waves, in mm, and their period content for earthquakes recorded at AKR.

Because of the small number of available seismographs, we implemented the following procedure in order to determine the epicentral locations of small or noisy events with a degree of confidence. The available seismograms were grouped according to their shape, because this depends on the path followed by seismic waves from the source (hypocentre) to the receiver (seismometer). It is, therefore, evident that earthquakes from the same area will display similar characteristics, when received at a given station. We identified three groups of earthquakes with similar shapes and assumed that each group came from a different area. Based on this assumption, we accepted some estimates of earthquake parameters with relatively large error bounds, because they were similar to well-estimated corresponding parameters of earthquakes belonging to the same group.

Another classification attempt was made on the basis of S-P arrival times, as recorded at AKR. This can be seen in Fig. 2, where the S-P arrival difference (ordinate) is plotted against the frequency of occurrence of earthquakes (coordinate) in histogram form. According to this criterion, no evident differentiation is observable in the earthquake population we have studied. It appears that the S-P arrival difference of most events (21) falls in the interval 2.5-4 sec; this corresponds to a range of 20-30 km.

As can be seen for Table IV the duration-magnitude (magnitude calculated from the duration of an earthquake) values of micro-earthquakes range between 0.5-3.0 on the Richter scale.

Although the geographical distribution of the stations could only cover a limited area, the precision of the estimated hypocentral parameters can be considered to be satisfactory, because most earthquakes occurring outside the perimeter of the network had also been recorded at APE. As can be seen in Fig. 3a, the epicentres are distributed off the north-east coast of Thera, exhibiting higher concentration around the Kolombos submarine volcano. A cross section along the SW-NE direction (Fig. 3b), shows that hypocentres are located at depths of approximately 5 km, and become shallower in the area of Kolombos.

Even though our field campaigns were few and the network did not operate for long time intervals, we can assert, with a degree of confidence, that the micro-earthquake tremor of the volcano is low.

CLASSIFICATION OF EARTHQUAKES

According to Minakami (1960), McNutt and Beavan (1984) and McNutt (1986), volcanic earthquakes are classified in four types:

     A)     A-type: These originate in the base of volcanoes, at depths ranging from 1-10 km and usually precede the eruptive phase of the volcano. Phases P and S are distinct, with amplitude ratios, in general, S:P > 3:1 and frequency content greater than 10 Hz.

     B)     B-type: This type is very shallow, with hypocentral depths ranging between 0-1 km and, located within 2 km of the crater. They appear in earthquake swarms. Because of their shallow depth the surface waves predominate and the S-waves cannot be clearly distinguished. S:P wave amplitude ratios range between 1 and 3 and their frequency content is between 1-5 Hz.

     C)     Explosion earthquakes: These accompany volcanic explosions and their initial motion is 'push' in every direction. They are very shallow earthquakes and are dissimilar to type B.

     D)     Volcanic pulsation, or continuous volcanic microtremors: These earthquakes belong to the previous type. They take place in volcanoes displaying Strombolian or Hawaiian type eruptions, and appear as continuous pulsation. In this type P waves cannot be distinguished from S waves. They are mainly surface wave strains.

During the early stages of a period of volcanic activity only A-type earthquakes appear. B-type earthquakes appear together with the first signs of uplift in the area of the volcano. The occurrence of B-type earthquake increases as the corresponding occurrence of A-type earthquakes decreases. From the above definitions it is evident that A- and B-type earthquakes are more significant in monitoring a volcano.

The recorded events (Table IV) cannot be unambiguously classified because: i) using the criterion of maximum amplitude ratio (A_s/ A_p) we observe earthquakes belonging to both types A and B as well as a number of events with A_s/A_p < 1, which cannot be classified in either of the above types; ii) using the criterion of frequency content, most events are classified in the B-type, because the dominant frequency appears to be approximately 5 Hz.

DETERMINATION OF GUTTEMBERG-RICHTER'S CONSTANT

It is well-known that the lesser the magnitude of earthquakes, the higher is the probability of occurrence. This can also be seen in the well-known empirical Guttemberg-Richter's (1944) relationship

 logΣN = a - bM                                                         (1)

where ΣN is the cumulative frequency of earthquakes according to magnitude M. Ishimoto and Iida (1939) give the relationships:

 NA^m = c                                                                  (2a)

 or

 ΣN = c [ (1 - m)Α ^1_m]^A                                                        (2b)

where A is the maximum recorded amplitude, and c is a constant; (m) is a dimensionless scalar, the Ishimoto-Iida constant. By taking the logarithms of both sides of equations (2a) and (2b), we have: 

logN = mlogA + c                                                   (3a)

or 

log(ΣN) = [ (1 - m) logA]^A + log [1 - m) c]              (3b)

It is obvious that relationships (1) and (3b) are equivalent, so that

-b _G-R = (1 - m)_I-I

According to Ishimoto and Iida, the constant m ranges within 1.8-2.2. Minakami (1960) and Minakami et al. (1969), studying volcanic earthquakes, found that:

1. For A-type earthquakes m ranges within the same bounds as for tectonic earthquakes.
2. For volcanic earthquakes m ranges within 2.8-4.5.

Using the above indirect procedure, we calculated the Ishimoto-Iida and the Guttemberg-Richter's constants.

The maximum trace amplitude A was measured from the seismograms of station AKR. We plotted a log-log diagram of ΣN (A_max) (Fig. 4). The slope of the graph, which gives the value of 1-m or the value of b, was calculated with least squares and was found to be equal to 1.34 ± 0.175. This confirms that the earthquakes we studied are volcanic. On the other hand we can assert that the estimated value of m (2.34) lies approximately midway between the ranges established by the above researchers for A- and B-type volcanic earthquakes.

AN APPROXIMATE DETERMINATION OF MAGMA CHAMBERS UNDERNEITH THE THERA AREA

The analysis of available seismograms revealed the existence of shear wave attenuation phenomena. As a result, some S waves appeared with amplitudes smaller than those of P waves. Utilizing such seismograms we attempted to define the boundaries of the domains causing the attenuation effects. The projections of direct S-wave ray paths on a horizontal plane are shown in Fig. 5.

Rays exhibiting anomalous behaviour are depicted with broken lines, rays with normal behaviour with solid lines. In this way we can outline the areas containing only parts of rays with anomalous behaviour (Fig. 5, hatched areas).

Firstly, in order to estimate the depth at which the observed absorption phenomena occurred (i.e. in the crust or in the mantle), we need to consider the epicentral distribution of earthquakes in relation to their corresponding hypocentral depths. It is observed that the hypocentres are generally shallow (< 5 km) and the epicentral ranges small. The depth to the Moho discontinuity in the region is approximately 28 km (Makris 1978). Therefore, we may infer that attenuation of shear waves took place in crustal depths rather than in the mantle. Furthermore, it appears that the observed waves arrive directly from the source (i.e. they are not reflected or refracted waves).

Moreover, because attenuation of S waves occurs when they pass through low viscosity materials, we can hypothesize the presence of viscous material in relatively shallow depths.

In Fig, 2 we observe that both domains of anomalous attenuation are related to the volcanic centres of Kameni and Kolombos respectively, which have been active in historic times (Table I). Moreover, considering thermal flow data (Fytikas and Kolios 1977; Stobbe 1980) we may put forward the assertion that the aforementioned low viscosity domains consist of partially molten material in shallow depths. Finally, considering the satisfactory precision of hypocentral parameter estimation, and the relatively significant number of earthquakes exhibiting anomalous S wave attenuation, we suggest that the boundaries depicted in Fig. 5 should, approximately correspond to the true boundaries of the speculated molten bodies. A systematic study of the region, with more and better distributed seismographs should, in the future, determine the true dimensions of these bodies with higher precision.

CONCLUSION

Although we could use only a limited amount of data, the hypocentral parameters were determined satisfactorily. Thus, we can make a few comments on the micro-earthquake activity of the region.

Volcanic activity is limited and scattered off the north-east coast of the island. The recorded events are definitely of volcanic origin, as indicated by the value of Ishimoto-Iida (1939) constant m, (= 2.34). However, the various classification parameters quoted in Minakami (1960) reveal the existence of a mixture of A- and B-types.

These results would appear to suggest that the observed events originate in domains with characteristics similar to those of volcanoes in the middle eruptive stage.

------------------------------------------------

---------------------------------------------