Seismic swarms, fault plane solutions, and stress tensors for
Transcrição
Seismic swarms, fault plane solutions, and stress tensors for
Seismic swarms, fault plane solutions, and stress tensors for São Miguel Island central region (Azores) Rita Silva, Jens Havskov, Chris Bean & Nicolau Wallenstein Journal of Seismology ISSN 1383-4649 J Seismol DOI 10.1007/s10950-012-9275-x 1 23 Your article is protected by copyright and all rights are held exclusively by Springer Science+Business Media B.V.. This e-offprint is for personal use only and shall not be selfarchived in electronic repositories. If you wish to self-archive your work, please use the accepted author’s version for posting to your own website or your institution’s repository. You may further deposit the accepted author’s version on a funder’s repository at a funder’s request, provided it is not made publicly available until 12 months after publication. 1 23 Author's personal copy J Seismol DOI 10.1007/s10950-012-9275-x ORIGINAL ARTICLE Seismic swarms, fault plane solutions, and stress tensors for São Miguel Island central region (Azores) Rita Silva & Jens Havskov & Chris Bean & Nicolau Wallenstein Received: 14 June 2011 / Accepted: 5 January 2012 # Springer Science+Business Media B.V. 2012 Abstract The central region of São Miguel Island is one of the most seismically active areas of the Azores archipelago. A revised analysis of the seismicity distribution at this region has, for the first time, shown that the seismicity is clustered in two distinct areas: the area around Fogo Volcano (Fogo) and the area around Congro maar (Congro), with each area having a highly localized swarm activity. From a total of about 15,000 events in the period from 2002 to 2010, 78 best located events were selected to make fault plane solutions using P-wave polarities and amplitude ratios. This set of fault plane solutions, and another six subsets derived from it, were inverted for the best fitting stress tensor. The stress tensor using all the 78 fault plane solutions is characterized by a subhorizontal σ1 striking WNW-ESE and a σ3 striking NNE-SSW, consistent with the regional stress field for this region. A similar result, using only the fault plane solutions R. Silva (*) : N. Wallenstein Centro de Vulcanologia e Avaliação de Riscos Geológicos, Universidade dos Açores, Ponta Delgada, Portugal e-mail: [email protected] J. Havskov Department of Earth Science, University of Bergen, Bergen, Norway C. Bean School of Geological Sciences, University College Dublin, Dublin, Ireland located in the Fogo area, was obtained. On the other hand, for the Congro area, a local stress field seems to be superimposed on the regional field: subhorizontal σ3, striking NNE-SSW, and a near-vertical σ1. The same stress regime persists in the first 5 km depth, probably related to the upwelling of thermal fluids. The rising fluids generate horizontal extension at shallow depths, which favour the opening of cracks and the circulation and ascension of hydrothermal fluids. The stress regime deeper than 5 km is more uncertain; however, it is indicative of a compressional regime. Thus, it can be conclude that the smaller Fogo area appears to be dominated by the normal regional stress field while the high active Congro area seems to have a different, highly heterogeneous stress field dominated by local conditions. Keywords Azores . São Miguel Island . Fault plane solutions . Seismic swarms . Stress inversion 1 Introduction The Azores archipelago is located at the Triple Junction, defined by the intersection of the Mid-Atlantic Ridge (MAR) with the North American (NA), Eurasian (EU), and African (AF) plates boundary (Fig. 1). The Azorean Islands are all of volcanic origin and several historical eruptions occurred inland and offshore. São Miguel is the largest island of the archipelago, and the last highly explosive eruption Author's personal copy J Seismol Fig. 1 Map of Azores archipelago showing the main morphotectonic features and seismicity distribution. The gray-shaded area represents the different segments of the Terceira Rift, and the dashed line represents the sheared western segment of the African-Eurasian plate boundary. Tectonic structures: EAFZ East Azores Fracture Zone, NAFZ North Azores Fracture Zone, GF Gloria Fault, FFZ Faial Fracture Zone, AFZ Açor Fracture Zone, PAFZ Princesa Alice Fracture Zone, PFZ Pico Fracture Zone (morphotectonics features adapted from Hipólito 2009). The Azores Plateau is limited by the 2,000 m bathymetric curve (bathymetric data from Lourenço et al. 1997). The seismic data are from 1997 to 2010 and contains all epicentres available in the CIVISA database. The red solid line defines São Miguel Island (subplinian) inland occurred in 1563 at Fogo caldera (Fig. 2). In the last 100 years, there have been recurrent swarm activity in the Fogo-Congro area (Nunes and Oliveira 1999) and, in 2005, there was an exceptionally high seismic swarm activity, with more than 6,000 events recorded in 6 months (Figs. 3 and 4). Recent studies indicate that the Fogo-Congro seismicity might be related to crustal deformation (Trota 2008), and there has been a real concern that renewed volcanic activity might occur. So far no studies have been made of the stress field in the area. In this work some of the best recorded events were used to analyze seismic swarm patterns, determine fault plane solutions, invert for the stress field, and investigate a possible relationship between the seismicity, stress field, and deformation. We use data from both the permanent seismic network operated by the Center for Seismo-volcanic Information and Surveillance of the Azores (CIVISA) and a temporary network of portable stations deployed between April and July, 2003. 1.1 Tectonics The triangular shaped Azores Plateau is, in the West, defined by the MAR, in the North, defined by the North Azores Fracture Zone (NAFZ), and by a complex alignment with WNW-ESE direction, running from the MAR to the western limit of Gloria Fault (GF), designated as Terceira Rift (TR),and in the South, defined by the East Azores Fracture Zone (EAFZ) (e.g., Luís et al. 1994; Borges 2003) (Fig. 1). The MAR and TR are detectable in the archipelago seismicity distribution, as well as the Faial Fracture Author's personal copy J Seismol Fig. 2 Tectonic sketch of São Miguel Island (adapted from Gaspar et al. 1995; Wallenstein 1999; Ferreira 2000; Carmo et al. 2009). Observable faults are shown as solid lines and inferred faults as dashed lines. The solid line box outlines the study area Zone (Fig. 1), where a high seismicity concentration is observed. There is no significant seismicity associated with the EAFZ. São Miguel Island is located in the alignment of the TR, and its most important tectonic structures (faults and lineaments) follow the general pattern exhibited Fig. 3 Left, the 1989 swarm (Nunes and Oliveira 1999); right, the best located events (see text) from the time period 2002–2010. The Fogo and Congro areas are indicated with dotted lines Author's personal copy J Seismol Fig. 4 Monthly number of earthquakes with Md≥1.5, as well as magnitude and depth as a function of time (top right). The times of some of the major swarms are indicated. Top left, monthly number of the 1,336 selected earthquakes (shown in Fig. 3) by the Azores Plateau, with main directions NW-SE and WNW-ESE (Fig. 2), as shown by Gaspar et al. (1995), Wallenstein (1999), Ferreira (2000), and Carmo et al. (2009). The emplacement of the polygenetic volcanic edifices and the distribution of secondary volcanic phenomena, like fumaroles (Notcutt and Davies 1999) and thermal springs (Cruz et al. 1999) are related to the main structures referred above. The central region of São Miguel Island, where recent increase in seismic activity has been observed, consists of the Fogo and Furnas volcanoes and the Achada das Furnas Plateau (Fig. 2). The Fogo Volcano, also known in the literature as Água do Pau Volcanic Massif, is placed in the central region of São Miguel Island. Like other polygenetic volcanoes in the archipelago, its emplacement seems to be due to the intersection of the dominant NW-SE and E-W fracture systems, thought to be associated with the existence of deep transform faults (e.g., Queiroz 1997; Wallenstein 1999). At the northern flank of the volcano, it is possible to infer the existence of the Ribeira Grande Graben by the alignment of the scoria cones and domes, reflecting the NW-SE and NNW-SSE directions. It is also possible to observe NE-SW alignments and a circular system of faults, which might be responsible for the emplacement of trachytic domes on the upper part of the volcanic edifice (Wallenstein 1999). The Achada das Furnas region corresponds to a basaltic plateau, which links the Fogo Volcano to the Furnas Volcano. This plateau consists of cinder/spatter cones and lava flows, and includes a small young trachytic centre known as Congro maar (Cruz et al. 1999). Furnas Volcano is the youngest of the three polygenetic volcanoes. The main fracture system crosses all volcanic Author's personal copy J Seismol systems with a dominant WNW-ESE direction, probably associated with a right-lateral strike-slip fault (Gaspar et al. 1995). On the southern and western flanks of the volcano, a distinctive fault system with N-S to N20°–50° E direction is present, as well as NW-SE directions (only on the S flank) coincident with the TR regional direction. The NNE-SSW system seems to control some structural depressions, calderas outlines, valleys alignments (Gaspar et al. 1995) and plays a major role in the transfer of magma to the surface (Guest et al. 1999). 1.2 Seismicity During the last century several seismic swarms were recorded in the study area. The first swarm accurately located occurred in 1989 (e.g., Tryggvason et al. 1989; Nunes and Oliveira 1999). Over 17,000 earthquakes have been recorded at this central region since 1997, the majority with duration magnitude, Md≤2, and located in the depth range between 0 and 10 km. Since the seismic network configuration has changed over this time period, only the statistics for events larger than the overall detections threshold (Md01.5) will be shown. More than 4,000 events with Md≥1.5 were recorded in this time period (Fig. 4). Duration magnitudes were calculated using the empirical relation of Lee et al. (1972). Since 2002, the central region of São Miguel Island has experienced a higher seismic activity than in the previous decades (Fig. 4). This seismicity occurs mainly as seismic swarms. Figure 4 indicates some swarms recorded in this region. It is seen (Fig. 3) that the 1989 swarm was located in the same area as the latest swarms in Congro area. In the previous centuries, at least since 1891, several other swarms have been reported (Nunes and Oliveira 1999), showing that seismic swarm activity in this area is a recurrent phenomenon, being the one occurred in 2005 probably the largest instrumentally recorded. Earthquake swarms are characterised as a sequence of small magnitude events with a strong time and space clustering, and without a dominant earthquake (Mogi 1963). This type of seismicity occurs normally in volcanic and/or geothermal regions, and presents high b values of magnitude-frequency distribution since the swarms consist of a larger number of events with low magnitude values, compared with the ones with high magnitude (e.g., Fischer 2005). The occurrence of earthquake swarms in volcanic active regions has been attributed to subsurface magma movement at shallow depths (e.g., Spicák and Horálek 2001; Fischer and Horálek 2005; Hensch et al. 2008) and/or associated with hydrothermal fluids circulation through new or pre-existent fractures (e.g., Husen et al. 2004; Farrel et al. 2009). Therefore, the swarms are commonly limited to regions with exceptionally heterogeneous distributions of material properties and stress concentrations (Hill 1977). The seismicity located at the central region of São Miguel Island has earlier been seen without a clear separation into distinct seismicity areas, probably because earlier raw results presented important global location errors. In order to look only to the best data, a selection of the events occurred between 2002 and 2010 was made. First, the original events were relocated using only stations with less than 13 km of epicentral distance and the standard crustal velocity model for the region (Senos and Nunes 1978 in Senos et al. 1992) (Table 1). The Hypocenter program (Lienert and Havskov 1995) was used for the locations. Then a second selection was made requiring a minimum number of eight stations, the root mean square of travel time residuals (RMS) less than 0.1 sec, and estimated errors (within 90 % confidence) with less than 1.5 km in all directions, being selected 1336 events (Fig. 3). Some revision of the phase arrivals was performed, and so the phase picking data are not identical to what was available in the CIVISA database. The data are clearly clustered into two areas according to their locations: Fogo and Congro areas (Fig. 3). The Fogo area is located around Fogo Volcano, and Congro area is located nearby Congro Maar. This separation was not so clear prior to the selection of the high-quality data, and has not been reported earlier, probably due to the absence of restrict selection criteria. In fact, Zandomeneghi et al. (2008) Table 1 P-wave velocity model— TAC model (Senos and Nunes 1978 in Senos et al. 1992). See section on velocity models later Depth (km) P-wave velocity (km/s) 0 2.4 0.5 4.1 2.2 5.4 5.4 6.8 11.9 7.8 Author's personal copy J Seismol elaborated a tomographic study of the area, and the 3D locations obtained did not show the two clustered areas. In order to better identify swarms in different localities, the data will be analyzed separately for Fogo and Congro areas. 1.3 Seismic swarm periods From 2002 to 2010, four main swarm episodes (April 2003, October 2004, May/September 2005, and September 2006) were identified (Fig. 4), where hundreds to thousands of small earthquakes occurred within days or over a few months. The statistic for the data selected for Fogo and Congro regions is also shown in Fig. 4, despite that these two subsets do not correctly reflect the overall seismicity since not all events are included. However, the comparison gives an indication if the seismicity occurs simultaneously within one month in the two areas. As it can be seen, most of the seismicity in the two areas does not quite coincide in time. In Fig. 5, none of the 15 identified swarms, with duration of 1–2 days, occur simultaneously in the two areas, which indicate that the source of the activity in the two areas is not directly related. The Congro area is the area with the largest number of swarms, and with the highest activity. Generally, it is seen that the seismicity has gradually declined since the high activity in 2005. Nevertheless, the central region is still the most active zone in São Miguel Island, and one of the most active of the Azores archipelago. In order to identify the best epicentral locations of the swarms, periods with higher activity were identified within the selected dataset of high-quality events (Fig. 3). This limited dataset might reduced the number of events in the swarms when compared with the whole dataset of the 17,000 events, and might even have prevented the detection of some swarms. A total of 15 swarms, with a total 381 events (Table 2; Fig. 5), were identified within the 1,336 high accurately located events. The epicentres for the 15 swarms are shown in Fig. 5, as well as some individual fault plane solutions occurring during the swarm periods (see text below). In addition, the even higher accurately located epicentres of the events used for individual fault plane solutions are shown (see later). The events in the individual swarms are generally located within areas less than 2×2 km, and considering the location accuracy estimated to 1.5 km, the enclosed area might be even smaller. The similarity among some seismograms of different events within one swarm seems to point to the existence of multiplets, which reflect a restricted area of influence. The swarms seem to move around and cover different parts of the seismically active area at different times. The two largest Congro swarms are the exception since they cover a much larger area. No systematic migration with time was found within these two large swarms. The events look to jump around randomly in the whole area during the occurrence of the swarm, indicating that the stress changes that are in the origin of these large swarms cover large areas when compared with the small swarms. The hypocentral depths of the routine picked events are more uncertain than the depths of the events used for fault plane solutions, since the calculated errors for these last are less than 1 km, in contrast with the range of 0.5 to 2 km of the former ones. Table 2 shows the depth range for the different swarms. All events have depths ranging from 1 to 7 km. Within one swarm, the events can be at very similar depths (like swarms #1 to 3), or spread out in depth (like swarm #5). Considering swarm #5, the depths of the 11 events used for fault plane solutions are very well determined (less than 1 km error), so it is clear that there is a considerable spread in depth for some seismic swarms. There is a significant difference in depths between Fogo and Congro seismic events. In the Fogo area, the majority of the events are above 5 km depth while in the Congro area there are several events that are located deeper than 5 km. High seismic activity and swarms generally occur either in Fogo or Congro areas. However, they do not seem to occur simultaneously in the two areas. The use of only the best located events has enabled the identification of more restricted swarm source areas, which have not been seen so clearly previously, since earlier it was thought that the swarm events were occurring randomly over the whole seismically active Fogo-Congro area (Silva 2004). 2 Data for fault plane solutions The data used for fault plane solutions were obtained from the 16 permanent stations operated by CIVISA at São Miguel Island and from 11 temporarily deployed stations on the behalf of the European project, e-Ruption, from April to July, 2003 (e.g., Bonagura et al. 2004; Saccorotti et al. 2004). The earthquake locations were obtained using SEISAN software (Havskov and Ottemöller 1999) and a five-layer, 1D velocity Author's personal copy J Seismol Fig. 5 The 15 swarms identified in this study. In each panel, the 78 best located events (grey squares) used for fault plane solutions are shown together with the swarm events (black squares). Individual fault plane solutions available for the swarm (bottom right corner) are shown. The seismic activity is concentrated either near the Fogo lake (outline shown to the left) or to the east, in the Congro maar area model (Table 1). In order to get reliable fault plane solutions, the selected events had to be recorded on, at least, 10 stations, and with Md ≥ 1.5. The RMS values should also be lower than 0.2 s. In Author's personal copy J Seismol Table 2 Swarm depth ranges Swarm # Year Month Day Area 1 2003 2 2003 3 4 Number of events/swarm Depth range (km) 04 26 Congro 7 5 09 07 Congro 14 5 2003 11 08 Congro 17 6 2004 01 22 Fogo 20 1–5 5 2004 10 30–31 Fogo 15 1–7 6 2005 05 12–14 Congro 76 1–6 7 2005 07 08 Congro 21 2–5 8 2005 07 20–21 Congro 25 1–5 9 2005 09 20–23 Congro 92 1–7 10 2006 05 27 Congro 10 4–5 1–5 11 2006 06 03–04 Congro 30 12 2006 09 17 Congro 21 5 13 2006 10 12 Fogo 11 4–5 14 2007 06 16 Fogo 12 4–5 15 2008 04 03 Congro 10 2–5 For swarms references, see Fig. 5 the total, 100 events were selected, 78 of which were used in determining the fault plane solutions (Fig. 5). The remaining seismic events were exclude due to the absent of clear polarities and/or amplitudes to get reliable solutions. 2.1 Velocity models and locations The first seismic study of crustal structure in Azores occurred on 1974, with the main scope to define the lithospheric structure near the Mid-Atlantic Ridge (Senos et al. 1992). Later, in 1977 and 1978, other seismic studies focusing on the crustal structure of Terceira and São Miguel islands were carried out in order to define geothermal areas of interest (Senos et al. 1992). Two 1D seismic velocity models were deduced: one for Azores in general (MAC model) and one for São Miguel Island in particular (TAC model; Table 1) (Senos and Nunes 1978 in Senos et al. 1992). These two models have been used since then in the routine location at CIVISA. There have been several attempts to improve the velocity model for the central region of São Miguel without much success (Escuer 2006). Later, Zandomeneghi (2007) and Zandomeneghi et al. (2008), using data from the e-Ruption project, calculated a 3D velocity model for the same region. With the obtained 3D model, the average shift in location was less than 0.07 km in any direction relative to using the TAC model, and therefore the TAC model will still be used as the prime model for this study. To get accurate fault plane solutions, the hypocentral location is critical, and in particularly its depth is very important due to the dependence on angle of incidence. Since the 1D model used for location assumes a layered structure, errors have to be assumed in the location in addition to the ones associated to the uncertainties in the phase readings. In order to minimize the errors caused by the velocity model structure, events were located using only stations within 10 km epicentral distance. In general, the effect of this limitation on the location was an increase in depth of 1 to 2 km, with a virtually unchanged epicentre, as compared with using all stations. All phase readings and depths were checked (RMS vs. depth was also checked). Formal errors were generally less than 1 km, but the real errors can probably be larger. It was observed that it was often difficult to get very low residuals (RMS generally between 0.1 and 0.2 s using 10 arrivals) for the S-arrivals, despite the arrivals being very clear (example in Fig. 6). This fact can be attributed to the existence of a heterogeneous structure, which will clearly increase the uncertainty of the Author's personal copy J Seismol Fig. 6 Seismograms from station MESC. The event origin time is 8 August 2003 at 22:00, with Md01.8, and hypocentral depth of 5.3 km results. It was also investigated if the station corrections could improve the location results. However, it was not possible to determine consistent stations residuals, probably because the stations were located too close to the heterogeneous structure. 2.2 Determination of fault plane solutions Although the hypocenters locations were calculated with stations at less than 10 km epicentral distances, data used for the fault plane solutions included stations up to 15 km distance (only polarity). The fault plane solutions could be found, or at least the P- and T-axes estimated, for about 20 % of the data with only polarity data. However, for the majority of the data, SH/P amplitude ratios were also needed in order to get a solution, and in practice at least five polarities and three amplitude ratios were necessary. The actual calculations were made using two different programs: FOCMEC and HASH. FOCMEC (Snoke et al. 1984) can use polarities and amplitude ratios of S- and P-phases for all components. However, only the P-phase on the Z component, and SH on the T component were used due to the uncertainty of correcting SV for the free surface effect (Snoke et al. 1984). The input amplitudes were corrected for the free surface effect, and could also be corrected for Q. Due to the short hypocentral distances, no correction for Q was made. The program HASH (Hardebeck and Shearer 2002, 2003) also uses polarities and amplitude ratios of S- and P-phases. The P-amplitude is given by: pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi Ap ¼ A2r þ A2z and the S-amplitude is given by: qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi As ¼ A2sv þ A2sh , where A is amplitude, r is radial, z is vertical, sv is SV, and sh is SH. The free surface correction was not built in, but replaced by a fixed factor per station, which had to be determined independently. In order to simplify the input, the free surface corrected amplitude ratios from FOCMEC were used as input for HASH. The program was modified to use only SH, and by using the free surface corrected P-phase on the Z-component, the assumption was made that Ap 0Az(P). Amplitude ratios were only used for the stations within 10 km epicentral distance, where clear direct arrivals (example in Fig. 6) could be observed while polarities were used for stations up to 15 km’s distance. To test the data, FPFIT (Reasenberg and Oppenheimer 1985) and PINV (Suetsugu 1998) programs were also used. However, these programs do not use amplitude ratios, and they were used only for comparison. FOCMEC operates on the principle that the user gives a maximum number of polarities and amplitude ratios errors and the amplitude ratio error should be less than a given ratio error. This is useful for getting an idea of all the possible solutions. However, if there are few or badly distributed data, changing one parameter (like allowing one more polarity error) can give distinct results. HASH returns solutions with less than a given number of polarity errors and averages the amplitude ratio error less than a given limit. If no solutions are found, error limits can be increased and normally several solutions are returned. Using this, an estimate of the best solution is made and likely errors calculated. The advantage in using HASH is that it finds one best solution while with FOCMEC the user must select one among many. Also HASH will not change completely the solution by one wrong amplitude ratio, once the amplitude ratio errors average is used as selection criteria, instead of a single amplitude ratio. On the other hand, FOCMEC does not give any estimate of the errors in the solution, while HASH calculates an estimated error. However, that requires an input in which each event has been located with, e.g., ten different likely input models, and all data are used as input, in order to get an estimate of the model uncertainties. This was not done in the present study, but an estimate of the quality of the solutions was performed, with the attribution of quality codes ranging from A to C. The codes were attributed according to the following criteria: A. The fault plane solution could be obtained with polarities only, and both FOCMEC and HASH Author's personal copy J Seismol gave the same result when adding amplitude ratios; B. There was not a unique fault plane solution using polarities only, and only an indication of dominant P- and T-axes was possible. Both FOCMEC and HASH results coincided when adding amplitude ratios; C. No possible solution was found with polarities only. Fault plane solutions for HASH and FOCMEC were different (<40°) when adding amplitude ratios. The evolution of the seismicity from 2003 to 2008 (Fig. 5) shows two distinct areas acting independently are visible: Fogo and Congro areas. With four significant swarms in the 6-year period, the Fogo area was less active than the Congro area, where 11 swarms occurred in the same period. For 10 of the 15 swarms, it was possible to find some individual fault plane solutions. In some cases, just one individual solution (swarms #4, 6, 12, and 13) was found, but in other cases there were two or more individual solutions (swarms of #5 and 7) identified for each swarm. In general, there is a mixture of fault plane solution types both in the Fogo and Congro areas. The most representative fault style in both areas is strike-slip to oblique strike-slip, with an important thrust component. 3 Crustal deformation Ground motions using GPS measurements in the area of the Fogo Volcano has been carried out since 1994 (Jónsson et al. 1999). Their study indicated a slight deflation of the volcano and they suggested that it could be related to a pressure decrease in a shallow magma chamber or due to the extraction of water and steam from the ground by the nearby geothermal power plant (Jónsson et al. 1999). Recent crustal deformation studies (Trota 2008), using GPS datasets from the end of 1999 to 2007, suggest that the central region of São Miguel has experienced a complex pattern of deformation, which when correlated with the seismic data recorded during the same period, lead Trota (2008) to divide the recorded deformation into three phases. The first phase, from 1999 to the beginning of 2005, corresponds to a period with both horizontal and vertical components of deformation, with a general uplift with the maximum value at the northern slope of Fogo Volcano (Fig. 7a). The author suggests that this is due to pressure increase caused by some intrusion at superficial crustal levels. The second phase was complex in terms of deformation since deflation and inflation occurred at the same time (Fig. 7b). Regarding the horizontal displacement, this phase presents the same radial pattern as the first, while the vertical motion shows uplift and subsidence, respectively at the South and North of the centre of deformation (marked as a pink star on the figure). According to the author, this is probably related to partially draining magma from a magmatic chamber along new and/or reactivated fissures. The last deformation phase started in October 2005 and included the 19th September 2006 swarm. It was marked by a low deformation rate, which the author suggests to be related with the decrease of pressure due to magma migration from the reservoir. During this last phase the radial outward pattern was still present at the Fogo and Congro areas, but the North sector showed a general uplift, whereas for the South sector exhibited a strong subsidence (Fig. 7c). We will try to compare this information on deformation with the seismicity and the possible changes in the stress field. Comparing the crustal deformation with seismicity, it can be observed that there is, in general, an overall correlation between the area of deformation and the area of seismicity, so they are clearly related to each other. The modelled centre of deformation is located in the Congro area and coincides with the higher seismicity zone of the area. This might explain the higher rate of seismicity in the Congro area when compared with the one of Fogo area. Since crustal deformation is affecting mainly the Congro, the seismicity in the two areas might be decoupled as it is also indicated by the occurrence of non-simultaneous swarms in the two areas. The seismicity in the Fogo area might be representative of the average long term seismicity while the Congro seismicity is exceptional. 4 Stress tensor analysis The stress tensor analysis consists in finding the directions of the three principal stress axes (σ1, σ2 and σ3, from most compressional to most extensional) that are responsible for a set of earthquakes. The stress tensor can be calculated from different fault plane solutions under the assumption that they represent a uniform stress field. Considering a homogeneous, elastic and Author's personal copy J Seismol Fig. 7 Top, horizontal velocity field for all datasets (from 1999.9 to 2007.7 (decimal years)); bottom, vertical components of the velocity field for the three deformation phases and the location of the identified swarms (same as in Fig. 5) occurring in each phase: a) from 1999 to the beginning of 2005, b) from the beginning of 2005 to September 2005, and c) from September 2005 to 2007. The pink stars indicate the centre of deformation after modelling. Contour lines of 10 mm. Scale bar units, mm/ year (Trota 2008) isotropic medium without friction and without preexisting faults, the stress tensor should have a shear stress (τ) of (σ1+σ3)/2, where σ1 and σ3 correspond to the largest and smallest eigenvalues of the stress tensor, respectively (Legrand et al. 2002). According to this ideal scenario, the P- and the Taxes can approximately correspond to the principal stress orientations. However, the medium is often prefractured and friction exists on the faults, and therefore the movement along pre-existing faults might imply that σ1 and σ3 are not oriented at 45° to the fault surface (Ramsey and Huber 1987) and therefore, they will not match the P- and T-axes. In fact, the only restriction that can be made is that σ1 lies within the dilatational quadrant of the focal mechanism. Therefore, a given stress field can induce movement of pre-existing faults with different orientations (e.g., Legrand et al. 2002), which suggests that a better approach is to find the stress tensor that best fit the overall fault plane solutions. To obtain the principal stress axes, the method described by Michael (1984) was used. In this method, to quantify the difference between the obtained stress tensor and the available data, the β parameter, or misfit angle, is computed. β is the angle between the “observable” slip direction on each plane and the maximum shear stress (τ) predicted by the stress tensor (Michael 1984), which ideally should be parallel to the slip direction. If β>40º, it means that the stress field is spatially heterogeneous, since it cannot be accurately represented by only one stress tensor (Michael et al. 1990). Michael’s stress inversion code uses a bootstrap resampling technique (Michael 1987a, b) in order to Author's personal copy J Seismol define the confidence regions. This technique consists in resampling the original dataset randomly and several times (which means that the same mechanisms can be repeated two or more times, while others can be absent) to simulate having new earthquakes in the same area and at the same period. According to Michael (1987a, b), to get a confidence level of 95%, 2000 repetitions are required, which was the number used in this work. Each bootstrap-resample is inverted for its best stress tensor, and the tensors distribution gives the confidence limits on the best tensor for the original dataset. This inversion process, as tested by Michael et al. (1990), was shown to be robust for datasets with 15 or more events, and at least 15 events should ideally be used. When inverting fault-plane solutions, two nodal planes are obtained. However, as discussed by Michael (1987a, b), it is advisable to select one of the nodal planes as the fault plane for the purposes of inversions, and for estimating confidence limits by bootstrap resampling, but since this information is not known a priori, the Michael (1987a) algorithm uses both nodal planes, trying at the same time to choose the correct fault plane and to find the best stress tensor. The stress tensor inversion gives the orientation (strike/dip) of the three principal stress axes σ1, σ2, and σ3, and the ϕ parameter (ϕ0(σ2–σ3)/(σ1–σ3)), which defines the relative size of the intermediate principal stress with respect to the maximum and minimum principal stress. 5 Stress inversion results 5.1 Regional stress field The regional stress field near São Miguel Island has been inferred from fault plane solutions of large earthquakes to be approximately normal (ca N40° E) to the Terceira Rift (Fig. 1) deformation zone (Borges et al. 2007). The nearest earthquake to São Miguel Island used in that study was the event of 21st January 1989, Mw05.7, located North of the island. It was a pure normal fault with a T-axis strike of 39°. For the São Miguel area, there are no other published fault plane solutions for large or for small earthquakes. Therefore, the average stress field, using the data from the present work, will be estimated. Table 3 Different regional stress directions and fault plane orientations inferred from different methods Strike Dip Rake σ3 σ1 1989 earthquakea 131 41 −87 39/5 193/86 Composite solution 104 69 −59 171/18 53/55 Source ABC solutions 19/34 113/5 A solutions 15/30 −95/31 B solutions 21/19 118/22 C solutions 24/46 167/38 T P 15/− 105/− Rose diagram For the two fault plane solutions, it is assumed the T- and P-axes correspond to σ3 and σ1, respectively. For stress inversion, σ3, σ1, consist of strike/dip. The composite solution is obtained with all the 78 events. Rose diagram corresponds to the dominant direction from rose diagrams of P- and T-axes strikes. All values in degrees ABC, A, B, and C solutions are stress inversions results using: all quality types, only A solutions, only B solutions, or only C solutions, respectively a According to Borges et al. (2007) In this study, 736 polarities for 78 events were used. Of these, 79 % were the same for all events indicating that, although the fault plane solutions look rather random (Fig. 5), they are the result of a non-random stress field. A better way of getting an idea of the average stress field is doing a stress inversion, assuming that the stress field is uniform. The average stress field was calculated with the software SLICK (Michael 1984), using the 78 fault plane solutions (Table 3; Fig. 8). Since we consider three quality types of fault plane solutions (A, B, and C), the stress analysis was carried out for each of the three groups. With this basis, the results (Table 3) indicate similar solutions for the three quality types, being performed an average inversion using all the data. A rose diagram was also made with the P- and Taxes directions from the 78 best located events Fig. 8 Stress tensor solutions obtained from all the inversions b performed. Stress tensor solutions: a) for the 78 best located events, b) for the events located within Fogo area, c) for the events located within Congro area, d) for the events located below 5 km, e) for the events located above 5 km, f) for the October 2004 swarm, and g) for the 2005 swarm. The square represents the σ1 axis, the triangle the σ2, and the circle the σ3 axis. The numerous dots define the 95% confidence region for each stress axis. The rose diagrams of P- and T-axis azimuths are also seen in the figure Author's personal copy J Seismol Author's personal copy J Seismol (Fig. 8a). It can be observed that the well-defined dominant T-axis is orientated ~15° N while the P-axis direction is more random. However, because of the circular distribution of the P- and T-axes, and the fact that the dip is ignored, when averaging directions, a careful treatment and interpretation of the rose diagrams is necessary. The minimum compressive stress axis (σ3) strike is also similar to the one inferred from the rose diagram. The results obtained inverting all of the 78 fault plane solutions (Fig. 8, all) show a nearly subhorizontal WNW-ESE maximum compressive stress axis (σ1) and a minimum compressive stress axis (σ3) striking NNE-SSW, with a dip <45° (see Table 4). The rose diagrams indicate that the P- and the T-axes are orientated in agreement with the stress tensor. The inversion of the entire dataset showed that all fault plane solutions could not be explained by a single uniform stress tensor, since the misfit angle (β) was too large to assume a near constant stress field (Fig. 9). However, the analysis shows a consistent T/σ3 strike of 15° to 19°, indicating that the dominant regional stress field is tensional with a direction as expected from the regional tectonics, although it is not completely normal to the Terceira Rift zone. This could indicate some local modification in the regional stress field. Looking at the variation of the misfit angle (β) as a function of event number (Fig. 9), it is observed that there is a smaller oscillation of the values between event number 18 and 41, which could indicate a more uniform stress condition during this time period. The October 2004 swarm, located at the Fogo area (swarm #4 in Fig. 5), occurred at the beginning of this sequence and it is also the swarm with the largest number (11) of individual fault plane solutions. The stress tensor inversion performed for this group (Fig. 8f), shows a β value of ~24° (Table 4), emphasizing the hypothesis of a more homogeneous stress field, when compared with the results from the others inversions. The result from the stress inversion for the October 2004 swarm shows σ1 and σ3 subhorizontal, striking WNW-ESE and NNE-SSW, respectively, and σ2 nearvertical, indicating a strike-slip mechanism. Again, the analysis shows σ3 strike of ~20°, consistent with the regional dominant tension. The 2005 swarm is considered the most representative of the seismic swarms that occurred in the Congro area, and it was used to perform the stress inversion which showed a normal faulting style (Fig. 8c). However, and as it was verified for the entire Congro dataset, the β value is high (Table 4), showing one more time the existence of a heterogeneous local stress field in this area. A stress inversion analysis was done for different depth ranges and using the focal mechanisms of the 78 best located events. The focal mechanisms of events shallower than 5 km were inverted and the results showed a normal faulting style, with σ1 near-vertical and the other two subhorizontal axes, with σ3 striking NNE-SSW (Fig. 8d). The uncertainties in σ1 and σ2 are quite large while σ3 seems to be well determined and in agreement with the regional stress field. The Table 4 Stress inversion results for the datasets analyzed in the present study, with strike/dip of the principal stress axes Dataset # FPS σ1 σ2 −149.3/55.6 All 78 112.7/5.3 Fogo 20 −72.5/11.1 Congro 51 −171/46.6 −71.9/8.6 Depth<5 61 −156/62.4 111.3/1.3 Depth>5 17 −80.7/0.5 −170.7/6.1 October 2004 11 −71.9/13.3 164.9/66.5 2005 swarm 20 −133.4/54.8 123.6/8.9 All values in degrees SS strike-slip, NF normal fault, TF thrust fault a The magnitude ratio of principal stress axes (see text) b The misfit angle c According to Zoback (1992) 91.3/78.4 σ3 19.2/33.8 ϕa βb Stress regimec 0.94 59 SS 0.39 50 SS 25.9/42.1 0.56 58 Unknown 20.6/27.5 0.85 54 NF 14.3/83.8 0.31 63 TF 22.7/19 0.22 24 SS 27.6/33.6 0.53 47 NF −163.1/3.1 Author's personal copy J Seismol Fig. 9 Misfit angle (β) variation along the entire dataset stress solution obtained with events deeper than 5 km shows reverse mechanisms (Fig. 8e). However, this solution is more uncertain than the solution obtained for shallower events, and should be used carefully. Despite the uncertainties, there is a clear difference in the stress field for deep and shallow events. but with a lower β value (Table 4). The stress tensors are different for the two groups and could indicate that there is a real difference in the average stress fields between Congro and Fogo. The inconsistencies for Congro could be caused by a larger extension of the area, indicating that the stress field is not homogeneous over the whole region. 5.2 Comparing Fogo and Congro areas Since the seismicity splits naturally into two different areas (Fogo and Congro), stress inversion, and rose diagrams were also made for these two groups (see Fig. 8b, c). Considering both groups individually, it can be observed that the stress tensors are better constrained than in the other inversions. The dominant T-axis direction, in the rose diagrams, still follows the commonly observed for the region in both groups, although it is much better defined for Fogo than Congro. Even the P-axis direction is well defined for Fogo while quite random for Congro, reflecting the mixture of different focal mechanisms (Fig. 10). The stress tensor found for the entire Fogo area dataset is similar to the stress tensor obtained with all the 78 events, as well as for the October 2004 swarm, which seems to represent a more homogeneous stress field, since it is the dataset with the lowest β value. For the Congro area, σ1 is subvertical, σ2 subhorizontal, striking WNW-ESE and σ3 subhorizontal, striking NNESSW. The 2005 swarm, which occurred in the Congro area, presents a similar stress field as the whole area 6 Discussion/conclusions This study has shown that significant swarm activity has occurred over a long period at São Miguel Island central region, being particularly representative the one that took place in 2005. It was observed, for the first time, that the seismicity is clearly divided into two areas, Fogo and Congro, and with the exception of the 2005 swarm, most of them were constrained in small localized areas. A new set of fault plane solutions was obtained and it was demonstrated that reliable fault plane solutions can be obtained using signal amplitudes, even in a heterogeneous structure. From all the events used to build fault plane solutions, 35% occurred during one of the identified swarms (Fig. 5), being 74% of the solutions obtained for the seismicity of Congro area while the remaining 26% were related with the Fogo area. The stress solution found with all the 78 fault plane solutions shows a minimum compressive stress axis (σ3) striking NNE-SSW, consistent with the regional stress field, with a nearly subhorizontal WNW-ESE maximum compressive Author's personal copy J Seismol Fig. 10 Focal mechanisms of the best located events from each area stress axis (σ1). Studies based on geological surveys at São Miguel (NE region) by Carmo (2004) and at Graciosa Island by Hipólito (2009), found a similar regional stress field acting at these distinct zones, with σ1 ranging from NW-SE to N-S, subhorizontal and σ3 from E-W to NE-SW, also subhorizontal. Taken together, all studies point to a minimum compressive stress axis (σ3) subhorizontal and oriented approximately NE-SW, parallel to the direction of motion of the Eurasian plate relative to the African plate. There is a significant difference between the Fogo and Congro areas seismicity, where the Fogo area seems to have a more consistent stress field than the whole study area, and coincident with the regional tectonic directions (Fig. 11). The Congro area shows more seismic activity that covers a larger area. The former has deeper events and a different stress field, with a permutation between σ2 and σ1, when compared either with the Fogo area or with the regional directions, suggesting the existence of a very local heterogeneous stress regime. In some cases, the complex geology and the highly fractured medium, along with the ascendant hydrothermal fluids, lead to conditions that favour normal faulting (e.g., Giampiccolo et al. 1999). Therefore, the extensional regime at Congro can be seen as a response to a compressive regime Fig. 11 Left, sketch map summarizing the main stress directions found for the Fogo and Congro swarms. The black arrows represent the directions of the regional minimum compressive (σ3) and maximum compressive stress axes (σ1). The green arrows are the principal stress axes found for the October 2004 Fogo swarm (shadowed area to the W) and the red arrows for the 2005 Congro swarm (shadowed area to the E). These two directions are also consistent with the direction found for the whole Fogo and Congro areas, respectively. The red dot represents the near-vertical σ1 axis. Right, sketch of the main directions found for the two ranges of depths with a schematic interpretation. The σ3 subhorizontal and σ1 subvertical opens mainly subvertical fractures, which allows the upward flow Author's personal copy J Seismol at larger depths. This could promote the ascension of thermal fluids to crustal lower levels thereby weakening pre-existent fault planes which, due to gravitational stress, become potential slip surfaces. The seismicity is clearly related to the observed deformation level changes. However there is no direct relation between the maximum level changes and the rate and location of seismicity. This might be due to the depth of the events, since surface level changes might not necessarily be related to corresponding level changes at depth. Various authors have concluded (Zandomeneghi 2007; Zandomeneghi et al. 2008) that the main driving force for the level changes is not the presence of magma but hydrothermal activity. Evidence of the hydrothermal activity is seen in the Furnas Volcano area (Fig. 2), and it is also well documented North of Fogo, where the hydrothermal field is located (Johnston 1979 in Silva et al. 1985; Duffield and Muffler 1984; Zandomeneghi 2007; Zandomeneghi et al. 2008). A local earthquake tomography study (Zandomeneghi 2007; Zandomeneghi et al. 2008) of the central region of São Miguel indicated a positive Vp anomaly near the Fogo Volcano with a high Vp/Vs ratio, which seems to indicate that this zone is marked by low Vs, due to fluids contained in fractures. Thus, this area might correspond to a transport zone and/or a zone of crystallization of fluids through a solidified intrusion (Zandomeneghi 2007; Zandomeneghi et al. 2008). Seismic tomography studies for the area also related the velocity anomalies to hydrothermal activity (Zandomeneghi 2007; Zandomeneghi et al. 2008). This seems reasonable considering that there has not been any magmatic eruption in the area since 1563, but frequent and intense seismic activity. Also the typical low-frequency earthquakes related to magma movements has not been observed, at least in significant amount. The magmatic bodies that are in the origin of the secondary volcanic activity of the Fogo Volcano (such as fumaroles and thermal springs) appear to induce upflow of the thermal fluids, acting as a heat source and promoting the convection of the hydrothermal system (Duffield and Muffler 1984; Gandino et al. 1985; Silva et al. 1985). For shallow depths, the predominance of subhorizontal σ3 (D < 5 km) is compatible with what is expected beneath volcanic areas (e.g., Legrand et al. 2002; Sánchez et al. 2004; Konstantinou et al. 2009). It is thought that rising fluids can generate horizontal extensional, which would favour the opening of cracks and the ascent and circulation of hydrothermal fluids. This might explain the observed fumarolic activity around Fogo Volcano. The results obtained in the inversion for events deeper than 5 km are not reliable; however the stress regime is clearly different below 5 km as compared to above 5 km, indicating different deformation regimes. The main importance of this study is that it has been shown that the seismic swarm’s area on São Miguel central region is not a uniform area of activity, but clearly can be divided into two main areas, Congro and Fogo, each with localized swarm activity. The smaller Fogo area appears dominated by the normal regional stress field, while the high activity at Congro area seems to have a different, highly heterogeneous stress field. This could indicate that the Congro area is affected by abnormal local stresses caused by deformations and/or large pressure changes in the first 10 km below the surface. Acknowledgments Rita Silva was supported by a Ph.D. grant from Direcção Regional da Ciência e Tecnologia, Região Autónoma das Açores (Azores Government) during this work. CIVISA partially supported the stay of Jens Havskov at the Azores. References Bonagura M, Damiano N, Saccorotti G (2004) Seismicity at Fogo and Furnas Volcanoes (São Miguel Is., Azores): evidences for fluid involvement. Atti del 23°Convegno Nazionale, 5 pp Borges JF (2003) Fonte sísmica em Portugal—algumas implicações na Geodinâmica da região Açores-Gibraltar. Ph.D. thesis, University of Évora, 305 pp Borges J, Bezzeghoud M, Buforn E, Pro C, Fitas A (2007) The 1980, 1997 and 1998 Azores earthquakes and some seismotectonic implications. Tectonophysics 435:37–54 Carmo R (2004) Geologia estrutural da região PovoaçãoNordeste (Ilha de S. Miguel, Açores). M.Sc. thesis, University of Azores, 121 pp Carmo R, Ferreira T, Gaspar JL, Madeira J (2009) Stress-fields at S. Miguel island (Azores): first approach (poster). International Summer School “Field volcanological laboratory: the Nisyros and the adjoining volcanoes, Greece—a window on the pre-eruptive magma processes”. IAV, Greece, 25–30 Setembro Cruz V, Coutinho R, Carvalho M, Oskarsson N, Gislason S (1999) Chemistry of waters from Furnas volcano, São Miguel, Azores: fluxes of volcanic carbon dioxide and leached material. J Volcanol Geotherm Res 92:151–167 Duffield W, Muffler L (1984) Geothermal Resources of São Miguel Island, Azores, Portugal. US Geol Surv Open-file report 84–287, pp. 21 Author's personal copy J Seismol Escuer M (2006) Improving seismic Vp and Vs models of the Azores Archipelago. M.Sc. thesis, University of Azores, 99 pp Farrel J, Husen S, Smith R (2009) Earthquake swarm and b-value characterization of the Yellowstone volcano-tectonic system. J Volcanol Geotherm Res 188:260–276 Ferreira T (2000) Caracterização da actividade vulcânica da ilha de S. Miguel (Açores): vulcanismo basáltico recente e zonas de desgaseificação. Avaliação de riscos. Ph.D. thesis, University of Azores, 248 pp Fischer T (2005) West-Bohemian earthquake swarms and their dynamics. Ph.D. thesis, Faculty of Mathematics and Physics, Charles University (Prague), 37 pp Fischer T, Horálek J (2005) Slip-generated patterns of swarm microearthquakes from West Bohemia//Vogtland (central Europe): evidence of their triggering mechanism? J Geophys Res 110, (B05S21) Gandino A, Guidi M, Merlo C, Mete L, Rossi R, Zan L (1985) Preliminary model of the Ribeira Grande geothermal field (Azores Islands). Geothermics 14(1):91–105 Gaspar J, Ferreira T, Queiroz G, Wallenstein N, Pacheco J, Guest J, Duncan A, Cole P (1995) Evolução morfoestrutural do vulcão das Furnas (ilha de S. Miguel, Açores). Universidade do Porto, Faculdade de Ciências, Museu e Laboratório Mineralógico e Geológico. IV Congresso Nacional de Geologia, Memória no. 4, pp. 999–1003 Giampiccolo E, Musumeci C, Malone S, Gresta S, Privitera E (1999) Seismicity and stress-tensor inversion in the Central Washington Cascade Mountains. Bull Seism Soc Am 89:811– 821 Guest J, Gaspar J, Cole P, Queiroz G, Duncan A, Wallenstein N, Ferreira T, Pacheco J (1999) Volcanic Geology of Furnas Volcano, São Miguel, Azores. J Volcanol Geotherm Res 92:1–29 Hardebeck J, Shearer P (2002) A new method for determining first-motion focal mechanisms. Bull Seism Soc Am 92 (6):2264–2276 Hardebeck J, Shearer P (2003) Using S/P amplitude ratios to constrain the focal mechanisms of small earthquakes. Bull Seism Soc Am 93(6):2434–2444 Havskov J, Ottemöller L (1999) SeisAn earthquake analysis software. Seism Res Lett 70:532–534 Hensch M, Riedel C, Reinhardt J, Dahm T, The NICE-People (2008) Hypocenter migration of fluid-induced earthquake swarms in the Tjörnes Fracture Zone (North Iceland). Tectonophysics 447:80–94 Hill D (1977) A model for earthquake swarms. J Geophys Res 82(8):1347–1352 Hipólito A (2009) Geologia estrutural da Ilha Graciosa—enquadramento no âmbito da Junção Tripla dos Açores. M.Sc. thesis, University of Azores, 155 pp Husen S, Taylor R, Smith R, Healser H (2004) Changes in geyser eruption behavior and remotely triggered seismicity in Yellowstone National Park produced by the 2002 M 7.9 Denali fault earthquake, Alaska. Geology 32(6):537–540 Jónsson S, Alves M, Sigmundsson F (1999) Low rates of deformation of the Furnas and Fogo Volcanoes, São Miguel, Azores, observed with the Global Positioning System, 1993–1997. J Volcanol Geotherm Res 92:83–94 Konstantinou K, Lin C, Liang W, Chan Y (2009) Seismogenic stress field beneath the Tatun Volcano Group, northern Taiwan. J Volcanol Geotherm Res 187:261–271 Lee WHK, Bennett R, Meagher L (1972) A method for estimating magnitude of local earthquakes from signal duration. US Geol Surv Open-file report, 28 pp Legrand D, Calahorrano A, Guillier B, Rivera L, Ruiz M, Villagómez D, Yepes H (2002) Stress tensor analysis of the 1998–1999 tectonic swarm of northern Quito related to the volcanic swarm of Guagua Pichincha volcano, Ecuador. Tectonophysics 344:15–36 Lienert BR, Havskov J (1995) A computer program for locating earthquakes both locally and globally. Seism Res Lett 66:26– 36 Lourenço N, Luís J, Miranda M (1997) Azores triple junction bathymetry. Edited map by University of Lisboa (Faculty of Sciences) and by University of Algarve Luís J, Miranda J, Galdeano A, Patriat P, Rossignol J, Mendes Victor L (1994) The Açores Triple Junction evolution since 10 Ma from an aeromagnetic survey of the Mid-Atlantic Ridge. Earth Planet Sci Lett 125:439–460 Michael A (1984) Determination of stress from slip data: faults and folds. J Geophys Res 89(B13):11517–11526 Michael AJ (1987a) The use of focal mechanisms to determine stress: a control study. J Geophys Res 92:357–368 Michael AJ (1987b) Stress rotation during the Coalinga aftershock sequence. J Geophys Res 92:7963–7979 Michael A, Ellsworth W, Oppenheimer D (1990) Coseismics stress changes induced by the 1989 Loma Prieta, California earthquake. Geophys Res Lett 17(9):1441–1444 Mogi K (1963) Some discussions on aftershocks, foreshocks and earthquake swarms—the fracture of a semi-infinite body caused by an inner stress origin and its relation to the earthquake phenomena (third paper). Bull Earthq Res Inst 41:615–658 Notcutt G, Davies F (1999) Biomonitoring of volcanogenic fluoride, Furnas Caldera, São Miguel, Azores. J Volcanol Geotherm Res 92:209–214 Nunes J, Oliveira C (1999) Actividade sísmica do Vulcão do Fogo, Ilha de São Miguel (Açores). Actas do I Simpósio de Meteorologia e Geofísica-Comunicações de Geofísica, pp 55–60 Queiroz G (1997) Vulcão das Sete Cidades (S. Miguel, Açores) —História eruptiva e avaliação do hazard. Ph.D. thesis, University of Azores, 226 pp Ramsey J, Huber M (1987) The tectonics of modern structural geology, volume 2: folds and fractures. Academic, London, pp 309–700 Reasenberg P, Oppenheimer D (1985) FPFIT, FPPLOT, and FPPAGE: FORTRAN computer programs for calculating and displaying earthquake fault-plane solutions. US Geol Surv Open-file report, 85–739, 109 pp Saccorotti G, Wallenstein N, Ibáñez J, Bonagura M, Damiano N, La Rocca M, Quadrio A, Silva R, Zandomeneghi D (2004) A seismic field survey at Fogo-Furnas volcanoes, Sao Miguel Island, Azores. Geophy Res Abstr 6:04493 Sánchez J, Wyss M, McNutt S (2004) Temporal-spatial variations of stress at Redoubt Volcano, Alaska, inferred from inversion of fault plane solutions. J Volcanol Geotherm Res 130:1–30 Senos M, Nunes J, Moreira V (1992) Estudos da estrutura da crosta e manto superior nos Açores. In: Oliveira CS, Lucas ARA, Correia Guedes JH (eds) Monografia 10 anos após o sismo dos Açores de 1 de Janeiro de 1980—aspectos técnico-científicos, vol. I. LNEC, Lisboa, pp 205–214 Author's personal copy J Seismol Silva R (2004) Características da sismicidade na região central da Ilha de São Miguel (Açores): identificação e análise de famílias sísmicas. M.Sc. thesis, University of Azores, 136 pp Silva A, Duffield W, Muffler L (1985) Geothermal studies of Agua de Pau Volcano, São Miguel, Azores. Geotherm Resour Counc Trans 9(II):395–399 Snoke JA, Munsay JW, Teague AG, Bollinger GA (1984) A program for focal mechanism determination by combined use of polarity and SV-P amplitude ratio data. Earth Notes 55(3):15 Spicák A, Horálek J (2001) Possible role of fluids in the process of earthquake swarm generation in the West Bohemia/Vogtland seismoactive region. Tectonophysics 336:151–161 Suetsugu D (1998) Practice on source mechanism, IISEE Lecture note, Tsukuba, Japan, 104 pp Trota A (2008) Crustal deformation studies in S. Miguel and Terceira Islands (Azores). Volcanic unrest evaluation in Fogo/Congro area (S. Miguel). Ph.D. thesis, University of Azores, 281 pp Tryggvason E, Dibble R, Okada H (1989) Água de Pau volcano/ seismic crisis of 1989. Final report on the UNESCO/ WOVO Volcanology mission to São Miguel, Azores, 2 to 8 August 1989. Relatório final da missão UNESCO/ WOVO, CV-INIC, Universidade dos Açores. Not published documents, 26 pp Wallenstein N (1999) Estudo da história recente e do comportamento eruptivo do Vulcão do Fogo (S. Miguel, Açores). Avaliação preliminar do hazard. Ph.D. thesis, University of Azores, 266 pp Zandomeneghi D (2007) Passive and active seismic tomography of volcanic islands: São Miguel (Portugal) and Deception (Antarctica). Ph.D. thesis, University of Granada, 335 pp Zandomeneghi D, Almendros J, Ibáñez J, Saccorotti G (2008) Seismic tomography of Central São Miguel, Azores. Phys Earth Planet Inter 167:8–18 Zoback M (1992) First- and second-order patterns of stress in the lithosphere: the world stress map project. J Geophys Res 97(B8):11703–11728