Intraplate seismicity across the Cape Verde swell: A contribution from a temporary seismic network

Tectonophysics 636 (2014) 325–337

Contents lists available at ScienceDirect

Tectonophysics journal homepage: www.elsevier.com/locate/tecto

Intraplate seismicity across the Cape Verde swell: A contribution from a temporary seismic network Dina Vales a, Nuno A. Dias b,c,⁎, Inês Rio c, Luís Matias c,d, Graça Silveira b,c, José Madeira c,d, Michael Weber e,f, Fernando Carrilho a, Christian Haberland e a

Instituto Português do Mar e da Atmosfera — IPMA, Lisbon, Portugal Instituto Superior de Engenharia de Lisboa — ISEL, Lisbon, Portugal Instituto Dom Luiz — IDL, Lisbon, Portugal d Faculdade de Ciências da Universidade de Lisboa, Lisbon, Portugal e GeoForschungsZentrum — GFZ, Potsdam, Germany f University of Potsdam, Postdam, Germany b c

a r t i c l e

i n f o

Article history: Received 3 February 2014 Received in revised form 9 September 2014 Accepted 19 September 2014 Available online 12 October 2014 Keywords: Intraplate seismicity Clustering Local magnitude scale Active volcanism Cape Verde Atlantic Ocean

a b s t r a c t We present an analysis and characterization of the regional seismicity recorded by a temporary broadband seismic network deployed in the Cape Verde archipelago between November 2007 and September 2008. The detection of earthquakes was based on spectrograms, allowing the discrimination from low-frequency volcanic signals, resulting in 358 events of which 265 were located, the magnitudes usually being smaller than 3. For the location, a new 1-D P-velocity model was derived for the region showing a crust consistent with an oceanic crustal structure. The seismicity is located mostly offshore the westernmost and geologically youngest areas of the archipelago, near the islands of Santo Antão and São Vicente in the NW and Brava and Fogo in the SW. The SW cluster has a lower occurrence rate and corresponds to seismicity concentrated mainly along an alignment between Brava and the Cadamosto seamount presenting normal faulting mechanisms. The existence of the NW cluster, located offshore SW of Santo Antão, was so far unknown and concentrates around a recently recognized submarine cone field; this cluster presents focal depths extending from the crust to the upper mantle and suggests volcanic unrest. No evident temporal behaviour could be perceived, although the events tend to occur in bursts of activity lasting a few days. In this recording period, no significant activity was detected at Fogo volcano, the most active volcanic edifice in Cape Verde. The seismicity characteristics point mainly to a volcanic origin. The correlation of the recorded seismicity with active volcanic structures agrees with the tendency for a westward migration of volcanic activity in the archipelago as indicated by the geologic record. © 2014 Elsevier B.V. All rights reserved.

1. Introduction The Cape Verde archipelago comprises ten volcanic islands and several islets located 500 km off the coast of Senegal, West Africa. The archipelago stands on the southwest flank of the Cape Verde Rise, an elevated region of the ocean floor approximately 1200 km in diameter. The fact that the islands do not form a linear chain as observed in other hotspots is interpreted as the effect of a long-lived mantle plume underlying an almost stationary plate (Fig. 1). The seismicity in Cape Verde is sparse and poorly known; most of the information is available only in old reports from 1947 to 1973. Previous data correspond to reports of felt earthquakes of low intensity ⁎ Corresponding author at: Department of Physics, Instituto Superior de Engenharia de Lisboa, Rua Conselheiro Emidio Navarro, 1, 1959-007 Lisbon, Portugal. Tel.: + 351 218317135; fax: +351 218317138. E-mail addresses: [email protected], [email protected] (N.A. Dias).

http://dx.doi.org/10.1016/j.tecto.2014.09.014 0040-1951/© 2014 Elsevier B.V. All rights reserved.

mainly in Fogo and Brava, in the southwest of the archipelago and typically related to volcanic eruptions (Ferreira, 1956; Mendes, 1956; Neves, 1981), providing intensities and number of events but not locations. Local temporary networks (Fonseca et al., 2003; Heleno, 2001; Heleno and Fonseca, 1999; Helffrich et al., 2006; Matias et al., 1997) previously deployed in the archipelago had very focused objectives, such as local seismo-volcanic activity monitoring or teleseismic tomography, with limited results concerning the local seismicity in Cape Verde. Only very recently a permanent seismic network was installed in the archipelago by the Cape Verdean authorities (National Institute for Meteorology and Geophysics — INMG; Faria et al., 2013; Fonseca et al., 2013; Faria and Fonseca, 2014), but their results are still limited to the assessment of volcanic hazard. Within the framework of project CV-Plume, a cooperation between GeoForschungsZentrum — Potsdam and Instituto Dom Luiz — Lisbon, 39 Broad-band (BB) stations were deployed on nine of the ten major islands of the Cape Verde archipelago between 2007 and 2008

326

D. Vales et al. / Tectonophysics 636 (2014) 325–337

Fig. 1. The Cape Verde archipelago showing the location of the CV-Plume temporary broadband seismic network stations deployed in 2007–2008. Seismicity available in the ISC catalogue 1938–2014 is resumed to 24 events (International Seismological Centre, 2011) and registered earthquakes felt in the islands during the 20th century. Top right: location of Cape Verde in the Atlantic. Bottom right: presumed epicentral area of the 1980–81 seismic crisis and epicentres of the 1995 Fogo volcano eruption.

(cf. Fig. 1). While aimed mainly to image the seismic structure beneath the archipelago, this dense seismic network covering it entirely, provided a rare opportunity to characterize the seismicity in the region. Therefore, this work presents the analysis of the data collected by the CV-Plume BB seismic network during the ten months of the project. From the numerous events detected during the processing stages described below several low-frequency signals, eventually associated with volcanic tremor or other similar sources, were detected but were discarded for future analysis. We have focused on the signals that could be attributed to earthquakes in order to assess the nature of the recorded seismicity, ignoring other signals of possible volcanic origin. Thus, the analysis scheme followed is a typical one for local seismicity: detection stage, phase and amplitude picking, relocation tests with simultaneous inversion of a new 1D model, waveform cross-correlation to assess the similarity between events, eventual location refinement, and calculation of focal mechanisms for source assessment and correlation with geological features. 2. Geological setting The Cape Verde archipelago, located off the West African coast between 14.5° and 17.5° N, is a volcanic archipelago composed of ten islands (nine of which inhabited) and a few islets. The islands, presenting a horseshoe shaped distribution opened to the west, are built on top of the largest oceanic swell in the Atlantic Ocean – the Cape Verde Rise – and formed by persistent volcanic activity since the Oligocene (Torres et al., 2002) until present times. The swell and the long-lived volcanism are attributed to a persistent mantle plume underlying an almost stationary plate producing underplating and dynamic uplift (Pim et al., 2008; Ramalho, 2011; Ramalho et al., 2010c). The islands stand on 150–122 Ma-old oceanic crust (M2–M16 anomalies; Heyes and Rabinowitz, 1975; Stillman et al., 1982). In terms of their state of evolution, the islands can be divided into two main groups: some islands still present youthful strong relief (Santo Antão, São Nicolau, Santiago and Fogo) while at others topography has been levelled by both subaerial erosion and marine erosion (São Vicente, Santa Luzia, Sal, Boavista, and Maio). Brava, although

presenting a strong relief, has been deeply eroded in a way that its southern half corresponds to exhumed magma chambers represented by a plutonic complex of pyroxenites, syenites and carbonatites that contacts to the north with pillow lavas and hyaloclastites forming an uplifted submarine sequence. Volcanic rejuvenation in Brava is represented by a Middle Pleistocene to Holocene post-erosive volcanic sequence (Madeira et al., 2010; Mourão et al., 2010). Historical volcanism (postdating settlement in mid-15th century) is reported solely for the island of Fogo, for which coeval texts describe the occurrence of 26 on-going eruptions (Ribeiro, 1960). Eleven of these descriptions present sufficient detail to allow mapping of the correlative cones and lava flows (including the most recent events in 1951 and 1995; Torres et al., 1997, 1998), while the earliest references (since 1500) are very laconic and in some cases may refer to different observations of the same eruptive episode by ships passing off the island, or even to an almost continuous state of eruption of Pico do Fogo volcano (Torres et al., 1997; 1998). Geological evidence from the islands of Brava and Santo Antão, where very recent (probably Holocene) volcanic events occurred, also points to the presence of active volcanic systems (Holm et al., 2008; Madeira et al., 2010). The remaining islands, with the exceptions of Boavista and Maio, present post-erosive volcanism of Pleistocene age (Duprat et al., 2007; Holm et al., 2008; Plesner et al., 2003; Torres et al., 2002). Bathymetric multibeam surveys of sea-floor areas around the islands of Santo Antão, São Vicente, Fogo and Brava recognized the presence of youthful cone fields to the northwest and southwest of Santo Antão, south of São Nicolau, between the islands of Santiago and Fogo, and between Fogo and Brava (Masson et al., 2008), indicating apparent active volcanic fields on the submarine flanks of these volcanic edifices. A growing seamount – Cadamosto – is located southwest of Brava (Grevemeyer et al., 2010). Most islands present important but variable uplift resulting in the subaerial exposure of sedimentary and volcanic submarine sequences (Madeira et al., 2010; Ramalho et al., 2010a, b; Ramalho, 2011). The different uplift rates of each island are explained by local processes with a small contribution from a swell-related episodic component (Ramalho, 2011).

D. Vales et al. / Tectonophysics 636 (2014) 325–337

Seismicity is scarce in the archipelago and has been reported mostly in the islands of Fogo and Brava. According to published reports (Ferreira, 1956; Mendes, 1956) increased seismic activity occurred in the 1941–1954 period, with 33 events with maximum macroseismic intensities of VI–VII (Modified Mercalli, MM) felt on the islands of Fogo and Brava in the SW and Santo Antão and São Vicente in the NW. The National Meteorological Service database (Serviço Meteorológico Nacional, 1947–1973) also refers to the occurrence of 88 earthquakes in the Fogo–Brava and São Vicente–Santo Antão areas, between 1955 and 1971. Most of this seismicity corresponds to a seismic swarm in 1963, during which 59 earthquakes were felt in Brava and seven in Fogo. Instrumental data obtained during another seismic crisis from December 1980 to May 1981 around the islands of Brava and Fogo, recorded 135 earthquakes presenting magnitudes up to 2.9 (Neves, 1981). During the 1995 eruption, 58 events were recorded from April 9 to 18, two of which were felt with intensities III–IV (MM) in Fogo (Matias et al., 1997). Other seismic swarms have been attributed to volcanic activity in the Cadamosto seamount southwest of Brava (Grevemeyer et al., 2010). In summary, since 1941 there are reports of five earthquakes felt in São Vicente (Max. Int. IV–V), 14 in Santo Antão (Max. Int. V), 20 in Fogo (V–VI), and 83 in Brava (V). The data available in the ISC database (International Seismological Centre, 2011), is very scarce, since they rely on the global network due to the absence of a permanent local seismic network, thus recording only some of the stronger events (cf. Fig. 1). Neotectonic deformation represented by fault scarps displacing Holocene and/or Late Pleistocene volcanic sequences in Brava, suggests that tectonic events cannot be precluded (Madeira et al., 2010). Structural data from the southern group of islands indicate the prevalence of tectonic or tectono-volcanic structures trending NW–SE, NE–SW, N–S and E–W (Madeira et al., 2010; Represas et al., 2012). Tectonovolcanic structures trending NW–SE are present in the islands of Santiago and Maio, including the alignment of volcanic edifices that shapes the island of Santiago and the alignment of plutonic intrusions in the island of Maio (Represas et al., 2012). The northwest flank of the island of Fogo presents conspicuous NW–SE scarps interpreted as a graben structure (Brum da Silveira et al., 1997). The southernmost graben fault is exposed in the caldera wall. NW–SE trending faults are also present in Brava Island, namely the Cachaço and Vigia normal faults defining a graben that displaces the southern central area of the island (Madeira et al., 2010). Among other structures in the same island, a major NE–SW fault zone separates the northwest half of Brava, where the submarine volcanic sequence crops out, from the southeast region characterized by outcrops of exhumed magma chambers. Geological mapping suggests uplift of the southeast region relatively to the northwest, but fault outcrops present slickensides indicating dominant dextral strike slip. These structures displace Middle Pleistocene to Holocene volcanic deposits indicating active faulting. In Fogo Island, a major N–S trending structure (the Sambango–Monte Vermelho fault; Brum da Silveira et al., 1997; Torres et al., 1998) controls the location of several aligned Holocene to Upper Pleistocene volcanic centres, including the 2829 m-tall Pico do Fogo, and probably the head of the flank collapse that opened the caldera to the East. E–W faults were observed in the northwest coast of Brava (Sorno area) presenting dominant left-lateral displacement.

3. Broad-band seismic network and data recorded Depending on the conditions at each site, stations of the CV-Plume network were installed on open-field or in shelters (such as those protecting communications antennas or electrical power distribution points). Some stations were installed near urban areas or near the coast, whereas others were installed in the more central areas of the islands; thus, the recorded noise level in each station shows considerable variability.

327

Each station was equipped with an Earth Data PR6-24 data logger and a Guralp CMG-3ESP (60 s) seismometer that recorded data continuously with a sampling frequency of 100 Hz. This temporary network operated for ten months, from November 2007 to September 2008. During this period, instrumental problems such as timing or failing components affected some stations, mainly in the southern islands, which may partially explain the smaller number of events recorded or locatable in that area of the archipelago. In order to detect the local small earthquakes, the continuous data stream was scrutinized using spectrograms: a spectrogram represents, for each seismic trace, an estimate of the average spectral content around each sample. Centred in each sample to be analysed, the application extracts 1024 points (10 s) and calculates the magnitude of the Fourier transform of this series, resulting in 512 values between 0.1 Hz and 50 Hz. The linear component of the series is removed before the spectral calculation. The output file is an image in bitmap format in which each strip of 50 pixels represents the desired frequency band. In our case we analysed the frequencies between 1 and 25 Hz as being the most representative for the detection of local earthquakes, since it contains their typical frequency content (Fig. 2, top). For each station, the spectrograms were calculated over one day time-windows; each daily spectrogram was then assembled by the concatenation of each station spectrum with the simultaneous plotting aiming to ease the search for coherent signals (Fig. 2, bottom). Although it is possible to simultaneously plot the entire network, to simplify the analysis we divided the 39 stations in two overlapping groups of 24, one focused on the northern islands and the other on the southern, the overlapping occurring for the eastern stations on Sal and Boavista islands. The visual correlation of the records of several stations provides the enhancement of coherent signals; Fig. 2 (bottom) shows the typical features that can be found on a spectrogram. A local earthquake, or an active source like an explosion, usually generates a signal with a large frequency content and duration of a few to tens of seconds, which will be recorded in several stations. On a spectrogram, this will result on a very marked vertical tight bar that is visible in several stations depending only on their distance to the source. This narrow band in the spectrum is usually more or less clear due to enhancement of its energy; the analysis of the raw seismograms usually proves to be much harder and quite often not allowing the identification of enough seismic phases for an event location. Thus, while one can detect several events on a spectrogram, the actual number of locatable earthquakes is usually smaller; the triangles in Fig. 2 (bottom) mark 3 detected events, but only one effectively locatable (red triangle). Nevertheless, this proved to be a very robust technique that helped overcome the high shortperiod noise present at many stations, particularly during daytime. Other features also visible in the spectrogram correspond to local noise produced near or at the station: several blurs, having limited frequency content and durations longer than seismic signals and usually confined to a single station or nearby stations, may correspond to atmospheric phenomena or have an anthropic origin. During the ten month period, 358 events were manually detected from the visual inspection of the spectrograms calculated for the entire network, from which it was possible to calculate 265 hypocentre solutions. The manual phase picking was performed by visual inspection of a time window of about 10 s around the time determined from the spectrograms, with weights attributed to each picking on an empirical basis according to the quality of the signals. Most signals had a low signal-to-noise ratio, usually the S-wave packet being easier to identify than the P, resulting in several events being located mainly by the S-arrivals. Therefore, the average accuracy of the P-arrivals picks was equivalent to the S-arrivals (∼ 0.1–0.2 s). First-motion polarities were determined only for signals with high signal-to-noise ratios, but the problems with the P-arrivals resulted that very few polarities were really available in the dataset. From the recorded events 19 were identified as corresponding to explosions,

328

D. Vales et al. / Tectonophysics 636 (2014) 325–337

Fig. 2. Top: Example of a spectrogram segment of 8th January 2008, for station CVSA1 located in Santo Antão, corresponding to a zooming marked with a green rectangle on the spectrogram below. The horizontal axis displays the time and the vertical displays the frequency in the range of 1–25 Hz. The colour scale is proportional to the amplitude of the signal, adjusted and normalized according to the noise level of each station, blue corresponding to lower and red to the highest amplitude values. Bottom: Spectogram segment for a set of stations in the northern group, where each station is represented by a horizontal coloured strip. A coherent common trace around 08:13, corresponding to the detection of a local earthquake, is marked by a red triangle; other weaker coherent traces are marked by yellow triangles.

218 earthquakes were located inside the study area whereas 28 were located outside. Several records present signals of low-frequency content below 5 Hz, typical of volcanic events and may correspond to volcanic tremor. They were stored and left for a future analysis. 4. Seismicity analysis methods

as explosions were located on São Vicente Island and that the main seismicity was located offshore, they were excluded. Based on these conditions, two datasets emerged: one composed of 88 earthquakes located in the NW part of the archipelago, near Santo Antão, and 30 events in the SW near Brava. The smaller size of the SW dataset coupled with instability of some of its hypocentres, which tended to further reduce the number of available events, led us to concentrate only in the group of 88 events from the northwest area (Fig. 3).

Following the event detection and identification stage, several 1D P-velocity models were tested for hypocentre location, some of which adapted from studies in the region: a model derived for the Azores crust (Hirn et al., 1980), the model derived for the 1995 Fogo eruption (Matias et al., 1997); a model adapted from joint inversion of Ps and Sp receiver functions for the north-western islands (Vinnik et al., 2012); two models adapted from 2D wide-angle models (Pim et al., 2008; Wilson et al., 2010). The model that, in first approximation, had a lower root mean square (RMS) residuals corresponded to the 1D model for the Azores crust (Hirn et al., 1980), with 0.386 s while these values for the other models varied between 0.514 s and 0.662 s. Therefore, this was the model used to obtain the initial hypocentral locations (cf. Fig. 3) using the hypocentre programme (Lienert and Havskov, 1995). 4.1. 1D inversion and event relocation The image provided by the preliminary locations in Fig. 3 already gives an indication about the main seismogenic areas of the archipelago, despite having been obtained from a 1D model that was not specifically derived for this region. The fact that most of the seismicity is very deep and located already in the mantle raised the question on whether this model was the most adequate for the region. Thus, in order to refine the hypocentral locations and increase the knowledge on the crustal structure in this region, we attempted to derive a new 1-D velocity model using the typical approach of VELEST (Kissling et al., 1994). This method performs a simultaneous inversion for the 1D seismic velocity model with associated station correction terms and hypocentral locations. Since the 1D inversion method relies on stable hypocentre solutions as input, the criteria adopted for data selection were the choice of earthquakes recorded by at least five stations and having RMS lower than 0.6 (black events in Fig. 3). Considering that all events identified

Fig. 3. Hypocentres of all locatable events registered between Nov. 2007–Sep. 2008 in the Cape Verde archipelago (red); final area selected for derivation of the 1D model and events complying with the selection criteria (black). To the bottom and left of the epicentral map are W–E and S–N vertical cross-sections.

D. Vales et al. / Tectonophysics 636 (2014) 325–337

The optimal Vp/Vs ratio was derived from a Wadati diagram (Havskov and Ottemöller, 1999) applied to this set of 88 earthquakes. For a minimum correlation of 0.90, a Vp/Vs value of 1.75 was obtained for each event while for a minimum correlation of 0.80 a Vp/Vs ratio of 1.74 was obtained. We chose the last value, in spite the lower correlation, since a bigger number of events adjusted to it and also because the numerical response of the several 1D models tested usually produced smaller RMS. For the 1D inversion, several initial velocity models were tested, based on the ones used for the preliminary location, and adapted when needed (cf. Fig. 4a, left). The strategy followed for the inversion was strongly conditioned by the large azimuthal gaps resulting from the seismicity spatial distribution and the seismic network layout, since the majority of the seismicity

329

is outside the network (cf. Fig. 3). Moreover, since there are practically no events beneath the stations, and they are mainly concentrated SW of Santo Antão Island, the 1D model will correspond to an oblique average of the crustal structure between the seismogenic areas and the stations. We started the inversion process with a total of 1537 readings (1185 P and 352 S) and a fixed Vp/Vs of 1.74. The conventional 1D inversion procedure of VELEST resulted in unstable numerical solutions, being hard to achieve a stable reduction of the RMS, a consequence of the relative low number of S-readings, essential to constrain focal depths and therefore the vertical velocity distribution. Since the majority of the events were deep (Fig. 3; 15 to 40 km), the inverted models were relatively insensitive to the mid-crustal input structure, frequently showing numerical divergence of the RMS values on the 2nd or 3rd iterations.

Fig. 4. a) 1D inversion models. Left: initial input models, with the model used for the initial location (dashed red); Right: all output models from the simultaneous inversion, together with the final adopted model (red); b) Spatial distributions of P-wave travel time delay computed from the simultaneous inversion for the CV-Plume 1-D model. Stations with less than four phase readings were not considered. CVSA6 was used as the reference station with a P-wave station correction fixed to zero.

330

D. Vales et al. / Tectonophysics 636 (2014) 325–337

Therefore, we adopted a procedure that is slightly different from the typical methodology of Kissling et al. (1994). Instead of testing each input model with several varying parameterizations and a high number of iterations (each model resulting from the inversion also serving as an input for further iterations), we tested a limited number of variations, in some cases limiting the number of inversions for each initial model to be performed. For the numerical stabilization of the output models, besides the introduction of shallow layers to accommodate the island structures, the damping factor of each layer increased with depth. The numerous models resulting from the inversion are plotted in Fig. 4a (right), with the final model adopted in red. The evaluation of the obtained models was based on the lower RMS and visual inspection of the event distribution, favouring the models that simultaneously decrease the RMS and that spatially concentrate the events. The initial RMS values for the input models (cf. Fig. 4a, left) varied between 0.5 and 0.7 s, decreasing by about 20%–30% to values around 0.2–0.3 s following the inversion; the model chosen as final (CV-Plume model in Fig. 4a, right; RMS = 0.199 s) already includes the phase delay for each station (Fig. 4b). The comparison of the several models resulting from the inversion (plotted in Fig. 4a, right), shows larger variations at the upper and lower crustal levels and relatively smoother values in the mid-crust. The final model has a shallow 1-km thick sedimentary layer with a Vp of 3.2 km/s, superimposed on a 1.5 km thick layer with 4.9 km/s; the mid-crust is modelled by a layer with thickness of 5.5 km and a Vp of 5.9 km/s, while the lower crust corresponds to a 6 km thick layer having a Vp of 7.1 km/s, overlying the top of the mantle which has a typical velocity of 8.0 km/s. These velocity values are consistent and within the range of values predicted for oceanic crusts (Karson, 1998), the model being composed respectively by the superposition of layers 2A, 2B, 2C and 3B, although the thickness values for the mid-crust layers exceed the expected values. While this inconsistency may be due to a real feature, more probably is a result of the complexity of the 1D inversion and its difficulty in fixing a stable final model. The new velocity model CV-Plume of Fig. 4a (right) was used to relocate the entire recorded seismicity. However, when applied to the events recorded in the Cadamosto–Brava–Fogo sector it revealed to be inadequate: the RMS of the relocated events increased about 56% compared with the ones provided by the model used in the preliminary location. Also, the epicentral distribution tended to be much sparser and less consistent with the type of locations reported by Grevemeyer et al. (2010). As described below, the attempt to use a location based on waveform clustering was not particularly successful, and as a consequence we opted to use two different velocity models for the final relocations: the seismicity recorded on the NW sector of Cape Verde, including the furthermost western events was relocated with the new CV-Plume model (see above), whereas for the events recorded on the SW and the few events dispersed in other areas of the archipelago we have kept the initial locations provided by the model derived from Hirn et al. (1980). 4.2. Waveform cross-correlation and event clustering It is widely recognized that earthquakes associated with the same tectonic structure usually present similar waveforms at all stations recording those events. To find “identical” events and to correlate them with the known geological structures, a classification scheme was applied, quantifying the similarity and coupling the events into “clusters”. Such scheme was developed within our group and successfully applied in the Azores (Dias et al., under review). The basic measurement for the event resemblance is based on waveform cross-correlation of the seismograms for pairs of events recorded at each station. While this could be performed over the complete recording, to better control the output we considered preferable to apply the cross-correlation to time windows centred on specific seismic

phases. Considering the energy distribution between the three components of a seismogram, we used specific tapered time-window lengths for each seismic phase: 2.56 s for the P- and 5.12 s for the S-phases, starting respectively 0.5 and 0.8 s before the phase onset. The records were band-pass filtered between 2 and 10 Hz for the P- and 1–6 Hz for the S-phases; in the case of the P-phase the time window was applied only to the vertical component, whereas for the S-wave it was applied to both horizontal components. Since the correlation may be affected by several problems, such as data gaps, low signal-to-noise ratios, and spikes, the final similarity measure considered was based on a weighted average of the cross-correlations computed for the whole network, building a matrix such as the one shown in Fig. 5a. The partition of the events into clusters was based on that similarity coefficient, with the exact value being selected upon a trial-and-error process: several values were tested, ranging from 0.5 to 0.9 (only the comparison of an event with itself reaching 1.0), with inspection of the resulting matrix aided also by dendrograms (see e.g. Dias et al., under review, and references therein) and visual inspection of the waveforms of the resulting clusters. In similar clustering studies conducted on tectonically generated events (e.g. Dias et al., under review, and references therein), this type of analysis is based on a sequential clustering S–P procedure, i.e., a first partition based on the S-wave similarities followed by a refinement based on the P-wave. The basic reason is that S-waves, being more complex and dependent on the fault structure, are the main characterization property to be used to discriminate between events; the P-wave is then used merely for a refinement of the S-clusters, namely for the source type. In our study of the Cape Verde seismic data, we found that the use of the S-wave as primary discriminator tended to split the events into several small clusters, mainly doublets or single events. Therefore, the best results were obtained by inverting the sequence analysis, i.e., by first computing the similarity of the P-phase, defining a first partition of P-clusters, followed by a refinement in each P-cluster by computing the S-phase similarity. All final clusters were thus obtained from a P–S sequence analysis, ensuring that all the events in each cluster belong to the same family. In Fig. 5a to c we present the stepwise output of the P–S clustering procedure. Fig. 5a shows the first step of the clustering process, with a graphical representation of the P-wave similarity analysis, with a minimum similarity cut-off value of 0.6 obtained by a trial-and-error process. The “hot-colour” squares, matching events with similar P-waves, correspond to preliminary clusters that are already close to the sequences obtained in the final clustering process; however, since the P-wave has a much smaller complexity than the S-wave, there are still some difficulties in discriminating events at this stage, resulting in some mixture between clusters. Fig. 5b presents the final sequences after the S-phase similarity analysis was applied to the preliminary P-clusters: the majority of the P–S sequences dispersed, mainly into doublets and single events, remaining only ten sequences with at least 3 events. Of those, only four sequences remained intact through the P–S clustering process. The two sequences corresponding to explosions in São Vicente, S1 and S5, due to its size and identical P and S waveforms presented by all events, were used as an output quality control and validation of the discriminating capability of the method. Sequence S2 gives an example of the changes occurring during the sequential P–S clustering process. Following the P-stage, this sequence contained 19 events mainly concentrated near Brava Island, but presenting some scattered around the archipelago; following the sequential S-stage, the number of events was significantly reduced from 19 to 7 events with practically identical waveforms, eliminating the outliers, but at the cost of also eliminating a few close events (cf. Fig. 5b). This final sequence, S2, corresponds also to the more numerous final cluster, thus deserving special attention. Fig. 5c shows the P-waveforms for the events of this sequence recorded by a station installed in Brava Island (CVBR1), all showing a compression first motion polarity. This

D. Vales et al. / Tectonophysics 636 (2014) 325–337 Fig. 5. Waveform similarity analysis and event clustering procedure: a) plot of the similarity matrix for the P-wave of the 88 events analysed in Cape Verde. The matrix was built by cross-correlating P-pulses, over tapered time windows on the vertical component, between pairs of events and later averaging for the entire network. The nodes were later re-ordered according to the hierarchical classification obtained by the clustering procedure, with a minimum similarity value of 0.6 obtained by trial-and-error; b) Epicentral map of the 10 clusters obtained from the P–S sequential clustering procedure, excluding all doublets and single events. The numbers between brackets correspond to the number of events contained in each sequence; c) P-waveforms for the best P–S final cluster located near Brava Island (S2), aligned on the P-onset.

331

332

D. Vales et al. / Tectonophysics 636 (2014) 325–337

set allowed the calculation of a composite focal mechanism. The general low coherence of the waveforms even of neighbouring events (see above) did not allow using high-precision relative relocation based on waveform cross-correlations. 4.3. Focal mechanisms To compute focal mechanisms we used the FOCMEC programme, included in the SEISAN software package (Havskov and Ottemöller, 1999), in which all the compatible solutions with a given P-arrival polarity distribution are obtained. These best solutions were later used as initial solutions in the MECSTA programme (Brillinger et al., 1980) to determine the most viable solution for the focal mechanism (nodal plane orientations and also the orientation of the principal stress axes, P and T). For the calculation of the focal mechanisms, the selection criterion followed was to use only events with a minimum of nine polarities reported, so a selection of just four events was obtained due to the difficulty in marking the polarities. In order to increase the number of source mechanisms, we tried to calculate composite focal mechanisms for the final clusters obtained from waveform cross-correlation process; however, these clusters suffered from the same problem of reduced number of determined polarities, and only the biggest cluster S2 located near Brava had enough polarities to allow a composite focal mechanism to be calculated. The four earthquake locations and their corresponding focal mechanism parameters are listed in Table 1, together with the composite focal mechanism parameters. The nodal planes and the distribution of P-wave polarities are represented in Fig. 6, where we present the calculated single fault-plane solutions and a composite solution of the best cluster obtained from the waveform similarity analysis. The composite focal mechanism position corresponds to the cluster centroid of S2 of Fig. 5b. 4.4. New local magnitude scale The local magnitude (ML) scale was calibrated using the approach described in Hutton and Boore (1987) and Carrilho and Vales (2009). According to this method, an ML = 2 earthquake is defined as a seismic event that produces a maximum amplitude of 1 mm on a Wood– Anderson seismometer located 17 km from the epicentre. Therefore, to calculate the local event magnitudes for m stations, we used the following formula (1): ML ¼

   m      Δi j 1X þ k Δi j −17 þ 2:0 þ S j log Ai j þ n log m j¼1 17

ð1Þ

where A is the maximum amplitude in μm/s, Δ is the hypocentral distance in km, n and k are respectively geometric and anelastic attenuation coefficients, S is a station correction, i is the earthquake number and j the station number. The attenuation coefficients are determined, using data recorded by the CV-Plume network, by minimizing Eq. (2) with constraints on the parameter ranges: 2

σ ¼

n X m  X i¼1 j¼1

  2     Δi j þ k Δi j −17 þ S j þ 2:0−ML log Ai j þ n log : ð2Þ 17

From this joint inversion, performed with σ = 0.266, the obtained attenuation coefficients were n = 1.685 (geometric spreading) and k = 0.00 (anelastic attenuation). Thus, anelastic attenuation is not relevant in the computation of local magnitude for this region. The coefficient of geometrical spreading is higher than the one inferred by Hutton and Boore (1987) for California (n = 1.110) and anelastic attenuation is lower (k = 0.00189). Fig. 7a shows a comparison of the ML attenuation correction, represented as −Log(A0), for the two regions. The local-magnitude station correction values are listed in Table 2. Within the distance range of the data from Cape Verde, the major difference in the attenuation correction occurs within the first few kilometres, although not very well constrained due to few available event data for such distances. The impact of the new attenuation curve on the ML magnitudes can be observed in the histogram presented in Fig. 7b. Most events have higher ML values (average correction ML ~ 0.25). The few exceptions correspond to the few local events located in Fogo Island. With the exception of Fig. 1, all other figures with magnitude scale indication were performed already with this new magnitude scale.

5. Discussion The obtained results show that the seismicity recorded during the ten month period in 2007–2008 was mainly concentrated in two offshore areas, roughly coincident with the western tips of the northern and southern groups of islands of the Cape Verde archipelago. No significant seismicity was recorded in other areas of the archipelago, which is consistent with the activity previously reported for the region. The most abundant seismicity occurred in the NW area around the islands of Santo Antão and São Vicente, with a high number of events distributed over a wider area. In the south-western area, near the islands of Brava and Fogo, the seismicity presents fewer events with a greater concentration in time and space (cf. Fig. 3 and Graphical abstract). Due to the horseshoe shaped disposition of the islands, to the fact that only land stations were available, and because the majority of the events occurred outside the archipelago, the azimuthal coverage of the recorded seismicity was not optimal and disturbed the relocation procedure. Therefore, several methodologies were conducted in order to characterize the recorded seismicity as best as possible. The first step was to develop a new 1D velocity model suitable for the region and simultaneously increase the quality of the hypocentral solutions. Such a model could only be derived for the north-western area, due to the previously referred constraints and to the lower seismicity rate recorded in the south-western area. Considering the uncertainties involved in the inversion process, the newly derived 1D model – CV-Plume – should be considered a mathematical response to the spatial disposition of the north-western seismicity, and its interpretation in terms of the crustal structure should be done with caution. Nevertheless, considering the obtained values and following Karson (1998) the values are consistent with the expected structure of an oceanic crustal type: an upper crust composed of two layers of mainly basaltic composition, layers 2A and 2B, and a lower crust of gabbroic composition eventually interlayed with ultramafic rocks, layer 3B; the middle-crust corresponds to a layer marking the transition between these two domains, eventually more complex than the simple correspondence to layer 2C. While the CV-Plume model has the best

Table 1 Location and source parameters for four earthquakes and one composite focal mechanism. Date

Time

Latitude (° N)

Longitude (° E)

Depth (km)

ML

Strike (°)

Dip (°)

Rake (°)

2008-04-15 2008-05-01 2008-05-14 2008-08-28 Composite

00:33:54 15:10:41 06:51:06 07:36:19

14.201 16.672 16.702 14.978

−24.176 −25.492 −25.410 −24.993

48 30.5 32.3 23

3.9 3.3 3.6 3.2

163 182 353 188 164

18 15 46 80 39

65 −21 −10 2 −83

D. Vales et al. / Tectonophysics 636 (2014) 325–337

333

Fig. 6. Top: P wave polarities and focal mechanism solutions obtained for the four single events as well as for the composite (in blue), which corresponds to a cluster near Brava. Bottom: location of the computed fault-plane solutions and the composite focal mechanism.

Fig. 7. a) Comparison of the ML attenuation correction for Cape Verde and California; b) Distribution of the ML changes after applying the new set of parameters compared to Hutton and Boore (1987).

334

D. Vales et al. / Tectonophysics 636 (2014) 325–337

Table 2 Station corrections coefficients derived for the local magnitude scale of Cape Verde. Santo Antão

São Vicente

São Nicolau

Sal

Station code

Mag correction

Station code

Mag correction

Station code

Mag correction

Station code

Mag correction

CVSA1 CVSA2 CVSA3 CVSA4 CVSA5 CVSA6

0.26 0.14 0.05 0.28 0.18 −0.04

CVSV1 CVSV2 CVSV3 CVSV4

0.22 0.17 0.00 0.17

CVSN1 CVSN2 CVSN3 CVSN4

0.16 −0.40 −0.06 0.14

CVSL1 CVSL2 CVSL3 CVSL4

0.29 0.13 0.01 0.40

Brava

Fogo

Maio

Santiago

Boavista

Station code

Mag correction

Station code

Mag correction

Station code

Mag correction

Station code

Mag correction

Station code

Mag correction

CVBR1 CVBR2

−0.18 –

CVFG1 CVFG2 CVFG3 CVFG4 CVFG5

−0.24 −0.40 −0.21 −0.40 −0.40

CVMA1 CVMA2 CVMA3 CVMA4

−0.17 −0.04 0.04 0.01

CVST1 CVST2 CVST3 CVST4 CVST5

−0.18 0.26 0.11 0.22 0.20

CVBV1 CVBV2 CVBV4 CVBV5 CVBV6

0.29 −0.28 0.22 0.26 –

numerical performance for the events located in the northern area, for the southern seismicity the model derived by Hirn et al. (1980) from the Azores had the best numerical results in terms of RMS values and produces a smaller spatial dispersion. Therefore, the final locations of the events were obtained using two different models, one for each area: CV-Plume in the north and the model from the Azores for the south. Our attempt to discriminate similar events based on waveform cross-correlation and to try to associate the resulting groups or families with specific structures had ambiguous results. Unlike tectonic environments, where each structure presents a constant behaviour, i.e., a given fault tends to present the same type of slip producing waves with the same basic characteristics (notwithstanding variations due to the event magnitude), volcanic events may display considerable variation. For example, events associated with magma transfer within the magmatic plumbing system produce very different signals in the same structure due to expansion and contraction. An example of different waveforms associated with two P–S sequences roughly at the same location, S8 and S10 in Fig 5b, can be observed in the supplementary material provided (cf. Supp. Fig. 1). The S-waveforms shown were recorded at station CVSA1 and band-pass filtered between 1 and 6 Hz which is the same used in the similarity analysis. The difficulty in aggregating events into similar clusters, together with the fact that different clusters located in a given position usually show different waveforms, suggest a possible volcanic origin for some of the recorded seismicity. The four individual focal mechanisms and the composite solution computed for one of the sequences (cf. Fig. 6) show diversity in faulting styles: normal, reverse and strike-slip. Given the large geographical dispersion of the events, their low-magnitudes and the non-optimal data coverage reliable conclusions on local and regional tectonics cannot be inferred from the results. The reverse solution is a particular case, since it corresponds to an isolated event located SSE of Fogo Island where no relevant information, such as high-resolution bathymetry, is available and therefore we discarded it. Fig. 8 summarizes the results obtained in this work, showing both the final locations of the seismicity in the NW (left) and the SW (right) areas, and corresponding focal mechanisms. The north-western area (Fig. 8, left) is dominated by a major cluster located south-west of Santo Antão and São Vicente islands, with the majority of the focal depths between 10 and 30 km including P–S sequences S8 and S10, as well as two focal mechanisms calculated for the NW area. Both solutions present oblique normal and strike-slip movement, but it is unclear which is the dominant component. The main earthquake cluster coincides with a recently identified volcanic cone field in the multibeam bathymetry covering the slopes of the western Cape Verde Islands by Masson et al. (2008); this coincidence (stars in Fig. 8, left) may represent ongoing magmatic activity in this

volcanic system. This activity seems to be continuous, since it is also reported since 2008 (Faria et al., 2013), which may be an indicator of volcanic unrest at this submarine cone field. In addition, some of the events with higher magnitude were also located in this cluster, the remaining corresponding to the westernmost recorded events. Since we excluded from the analysis all low-frequency signals, phenomena like volcanic tremor are excluded and more probably these events are associated with fault slip affecting the volcanic system. The great depths of the hypocentres can, therefore, be due to effects induced by a magma chamber or by deep volcanic intrusions. The alignments presented by some events inside this cluster are probably due to the low azimuthal coverage and probably do not correspond to real features; alternatively, since the orientation of the nodal planes of both focal mechanisms seems to be coherent with some of these alignments, they may represent events at aligned volcanic edifices or planar feeding structures. Additional smaller clusters can be perceived to the west, north-west and north of Santo Antão (Fig. 8, left and Supp. Fig. 2), which may also be related to volcanic edifices, and coincident with P–S sequences S9, S6 and S3 respectively (cf. Fig. 5b); however, since they correspond to some of the cases of worst instrumental azimuth coverage, it is hard to be certain. Their focal depths of zero clearly show the uncertainty of their location. The only certainty is that they do not seem to be associated with the Nola seamount nor to the submarine volcanic cones in its vicinity. The tight clusters located on São Vicente correspond to a series of blasts associated with ongoing works for the enlargement of the airport. To the south of São Vicente there is a smaller group of events, corresponding to sequence S4, which is not related with either of the previous clusters; they may also be associated with submarine volcanic cones, but they are already outside of the area surveyed by Masson et al. (2008). In the south-western area (Fig. 8, right), besides the major NE–SW trending cluster between Cadamosto Seamount and Brava, coincident with sequences S2 and S7, a conspicuous transverse NW–SE alignment is evident immediately to the north of Cadamosto. The NE–SW alignment may represent a volcano-tectonic structure, possibly a common magmatic plumbing system shared by three aligned volcanic edifices (Seco and Rombo Islets, Brava, and Cadamosto). The elongated shape of Cadamosto seamount also trends NE–SW, corroborating a probable magmatic system linking those volcanic edifices, which present ages progressively younger towards the south-west. The transverse alignment could correspond to a NW–SE trending fault, possibly belonging to the same fault system as Vigia Fault in Brava (see Section 2), a tectonic direction common to Brava and Fogo Islands. Furthermore, several epicentres just west of Brava seem to be aligned with the most prominent fault scarp (the Vigia Fault; Madeira et al., 2010; Fig. 8b). The composite focal mechanism suggests a pure normal movement, which is in

D. Vales et al. / Tectonophysics 636 (2014) 325–337

335

Fig. 8. Epicentral maps of the seismicity detected in Cape Verde during the CV-Plume experiment and associated focal mechanisms together with vertical W–E profiles with the depth distribution. Left: NW area with the events located with the new CV-Plume model derived in this study. The green stars correspond to the submarine volcanic cones identified by Masson et al. (2008). NS, Nola seamount; SA, Santo Antão; SV, São Vicente; SL, Santa Luzia; SN, São Nicolau; Right: SW area with the events located with the original 1D model of Hirn et al. (1980), overlying the bathymetry of Grevemeyer et al. (2010). CS, Cadamosto seamount; B, Brava, and F, Fogo. Island topography from ASTER Global Digital Elevation Model v001 (NASA, 2009).

agreement with field evidence for normal faulting in the Vigia Fault. The focal depths also suggest two different sources for this seismicity: the NE–SW events associated with a shallower volcano-tectonic structure whereas the deeper events around Cadamosto seamount may have a pure volcanic origin. Surprisingly, no significant activity was recorded at Fogo volcano, the most active volcano in the archipelago. The single event source information available in this area points to a pure strikeslip movement; and due to its location outside the bathymetric surveyed area by Grevemeyer et al. (2010) and to the fact that this focal mechanism solution corresponds to the one with fewer polarities, no final conclusion about its nature can be made. An attempt was made to check whether this seismicity, and in particularly the final sequences obtained, was clustered in time besides being clustered in space. Based on the spatial distribution of the relocated seismicity (cf. Figs. 3 and 8) and on the position of the P–S sequences (cf. Fig. 5b) several seismogenic sectors were empirically defined and the temporal evolution of the contained events analysed. While there are some bursts of activity, no obvious temporal evolution can be perceived. The results are summarized in the supplementary material that accompanies this work (cf. Supp. Fig. 2). 6. Conclusions In this work, we present the first evaluation of the seismicity across the Cape Verde archipelago at a regional scale, using data recorded between 2007 and 2008 by a temporary network of 39 BB stations. For the detection of seismic events, we applied a spectrogram technique, allowing a thorough screening of all seismicity detected in the region during that period. A standard approach was followed, based on

Geiger's method and inverting for a new 1D crustal model, called CV-Plume, for the analysis of the detected events around the northwestern islands. For a better characterization of the aggregation of events in time and space, a waveform similarity method was applied to define clusters of similar events which, coupled with determination of single and composite focal mechanisms, allowed a better constraint of the associated sources. In addition, attenuation coefficients required for local magnitude evaluation were also derived, allowing the determination of event magnitudes with a scale specific for this region. During the survey period it is shown that the majority of the seismicity was concentrated in the westernmost areas of the archipelago, at the terminations of the northern and southern groups of islands. No significant activity was detected associated with the most active volcano of Fogo or the island of Santiago. As previous reports suggested, no relevant seismic activity is registered in the easternmost islands of Sal, Boavista and Maio. In terms of signal and occurrence, the recorded seismicity presents characteristics that suggest a dominant volcanic origin. In the south-western area of the archipelago, the seismicity presents a lower occurrence rate and is concentrated mainly along an alignment between the Cadamosto seamount and Brava. The distribution of epicentres in this area suggests the presence of active linear tectonovolcanic structures trending NE–SW and NW–SE. These are interpreted as indicating the presence of an active volcanic system shared by Brava and Cadamosto seamount and NW–SE trending active faults. According to the composite focal mechanism, the movement associated with this structure points to a normal faulting type. In the north-western area of Cape Verde, the recorded seismicity presents a greater activity rate but is dispersed over a wider geographical area. The most important clusters are concentrated south–

336

D. Vales et al. / Tectonophysics 636 (2014) 325–337

southwest of Santo Antão and São Vicente islands and the comparison with recent high resolution bathymetric surveys shows that they are associated with a recently recognized submarine cone field, discovered and named “Charles Darwin Volcanic Field” during a RRS Charles Darwin 2005 cruise (Masson et al., 2008; Hansteen and Kwasnitschka, 2013; Kwasnitschka et al., 2013). The clustering of seismicity in time and space, suggests that the events in this cluster are the manifestation of unrest in that volcanic structure, with the focal depths indicating an origin deep within the magmatic system. Additional activity is occurring west of Santo Antão, probably a manifestation of similar volcanic signals, but no activity was observed related to the Nola seamount. This seismic activity in the NW had been previously described, the few reports referring to rarely felt events in the volcanic edifices of Santo Antão and São Vicente islands. The correlation with a volcanic origin suggests on-going magmatic activity in the western tips of the two branches of archipelago, associated with the westernmost islands and active seamounts or submarine cone fields. This is in agreement with the location of the youngest volcanic manifestations in the westernmost islands, while the eastern islands are dominantly older (Mitchell et al., 1983; Plesner et al., 2003; Duprat et al., 2007; Holm et al., 2006, 2008; Dyhr and Holm, 2010; Madeira et al., 2010). Supplementary data to this article can be found online at http://dx. doi.org/10.1016/j.tecto.2014.09.014. Acknowledgements This work was supported by the Portuguese Science and Technology Foundation (FCT/Fundação para a Ciência e Tecnologia), under research project “CV-PLUME: An investigation on the geometry and deep signature of the Cape Verde mantle plume” (PTDC/CTE-GIN/64330/2006). This is a contribution to PEST – OE/CTE/LA0019/2013–2014 – IDL. The instruments for the deployment were provided by the Geophysical Instrument Pool Potsdam, Germany. The figures were produced using the software packages GMT (Wessel and Smith, 1998) and SEISAN (Havskov and Ottemöller, 1999). We would like to acknowledge the effort of all the elements of the CV-Plume and COBO teams involved in the field operations of the seismic experiment that took place in the Cape Verde in 2007–2008. We are in debt with ELECTRA, the Cape Verde electrical company, by the support provided and the invaluable aid of Professors António Lobo de Pina, Sónia Silva and Vera Alfama from the University of Cape Verde. In addition, the authors would also like to thank Agust Gudmundsson and an anonymous reviewer for their valuable revision and suggestions to improve the quality of the initial manuscript. References Brillinger, D.R., Udias, A., Bolt, B.A., 1980. Probability model for regional focal mechanism solutions. Bull. Seismol. Soc. Am. 70 (1), 149–170. Brum da Silveira, A., Madeira, J., Serralheiro, A., 1997. A estrutura da ilha do Fogo, Cabo Verde. Proceedings of the 1st International Symposium “A erupção vulcânica de 1995 na ilha do Fogo, Cabo Verde”, edited by Instituto de Investigação Científica Tropical and Ministério da Ciência e Tecnologia 972-672-863-0,pp. 63–78, (in Portuguese). Carrilho, F., Vales, D., 2009. Calibration of magnitude ML for Portugal mainland and adjacent areas. 6th Symposium of the Portuguese Association of Meteorology and, Geophysics. 48. Dias, N.A., Matias, L., Téllez, J., 2014. Insight on the crustal strain in Faial and Pico Islands (Azores), from the 1998 aftershocks waveform similarities and shear-wave splitting. subm. Geophys. J. Int. (under review). Duprat, H.I., Friis, J., Holm, P.M., Grandvuinet, T., Sørensen, R.V., 2007. The volcanic and geochemical development of São Nicolau, Cape Verde Islands: constraints from field and 40Ar/39Ar evidence. J. Volcanol. Geotherm. Res. 162 (1–2), 1–19. http://dx. doi.org/10.1016/j.jvolgeores.2007.01.001. Dyhr, C.T., Holm, P.M., 2010. A volcanological and geochemical investigation of Boa Vista, Cape Verde Islands; 40Ar/39Ar geochronology and field constraints. J. Volcanol. Geotherm. Res. 189, 19–32. http://dx.doi.org/10.1016/j.jvolgeores.2009.10.010. Faria, B., Fonseca, J.F.B.D., 2014. Investigating volcanic hazard in Cape Verde Islands through geophysical monitoring: network description and first results. Nat. Hazards Earth Syst. Sci. 14, 485–499. http://dx.doi.org/10.5194/nhess-14-485-2014.

Faria, B.V.E., Day, S., Fonseca, J.F.B.D., 2013. Insights from Geophysical Monitoring into the Volcano Structure and Magma Supply Systems at Three Very Different Oceanic Islands in Cape Verde Archipelago. AGU2013 Fall Meeting, San Francisco, California, USA, pp. 9–13 (December, Abstract V41B-2786). Ferreira, H.A., 1956. Macrosismos sentidos em Portugal no período de 1901–1945. Boletim da Ordem dos Engenheiros. 5(5), pp. 1–10, Lisboa. Fonseca, J.F.B.D., Faria, B.V.E., Lima, N.P., Heleno, S.I.N., Lazaro, C., d'Oreye, N.F., Ferreira, A.M.G., Barros, I.J.M., Santos, P., Bandomo, Z., Day, S.J., Osorio, J.P., Baio, M., Matos, J.L.G., 2003. Multiparameter monitoring of Fogo Island, Cape Verde, for volcanic risk mitigation. J. Volcanol. Geotherm. Res. 125, 39–56. Fonseca, J.F.B.D., Faria, B.V.E., Trindade, J., Cruz, G., Chambel, A., Silva, F.M., Pereira, R.L., Vazão, T., 2013. “Last mile” challenges to in situ volcanic data transmission. Nat. Hazards Earth Syst. Sci. 13, 3419–3428. Grevemeyer, I., Helffrich, G., Faria, B., Booth-Rea, G., Schnabel, M., Weinrebe, W., 2010. Seismic activity at Cadamosto seamount near Fogo Island, Cape Verdes-formation of a new ocean island? Geophys. J. Int. 180 (2), 552–558. http://dx.doi.org/10.1111/ j.1365-246X.2009.04440.x. Hansteen, T.H., Kwasnitschka, T., Klügel, A. (Eds.), 2013. Cape Verde Seamounts — R/ V Meteor cruise No. 80 Leg 3, 29 December–1 February 2010, unpublished report. Havskov, J., Ottemöller, L., 1999. Seisan earthquake analysis software. Seismol. Res. Lett. 70, 532–534. http://dx.doi.org/10.1785/gssrl.70.5.532. Heleno, S.I.N., 2001. Estudo Sismológico do Vulcão do Fogo(PhD Thesis) IST, Lisbon, (in Portuguese). Heleno, S.I.N., Fonseca, J.F.B.D., 1999. A seismological investigation of the Fogo Volcano, Cape Verde Islands: preliminary results. Volcanol. Seismol. 20, 199–217. Helffrich, G., Heleno, S.I.N., Fonseca, J.F.B.D., 2006. Hydroacoustic detection of volcanic ocean-island earthquakes. Geophys. J. Int. 167 (3), 1529–1536. http://dx.doi.org/10. 1111/j.1365-246X.2006.03228.x. Heyes, D.E., Rabinowitz, P.D., 1975. Mesozoic magnetic lineations and the magnetic quiet zone off northwest Africa. Earth Planet. Sci. Lett. 28, 105–115. http://dx.doi.org/10. 1016/0012-821X(75)90217-4. Hirn, A., Haessler, H., Hoangtrong, P., Wittlinger, G., Mendes-Victor, L.A., 1980. Aftershock sequence of the January 1 st , 1980, earthquake and present-day tectonics in the Azores. Geophys. Res. Lett. 7 (7), 501–504. http://dx.doi.org/10.1029/ GL007i007p00501. Holm, P.M., Wilson, J.R., Christensen, B.P., Hansen, L., Hansen, S.L., Hein, K.M., Mortensen, A.K., Pedersen, R., Plesner, S., Runge, M.K., 2006. Sampling the Cape Verde mantle plume: evolution of melt compositions on Santo Antão, Cape Verde Islands. J. Petrol. 47, 145–189. http://dx.doi.org/10.1093/petrology/egi071. Holm, P.M., Grandvuinet, T., Friis, J., Wilson, J.R., Barker, A.K., Plesner, S., 2008. An 40 Ar– 39 Ar study of the Cape Verde hot spot: temporal evolution in a semistationary plate environment. J. Geophys. Res. 113, B08201. http://dx.doi. org/10.1029/2007JB005339. Hutton, L.K., Boore, D.M., 1987. The ML scale in southern California. Bull. Seismol. Soc. Am. 77 (6), 2074–2094. International Seismological Centre, 2011. Internatl. Seis. Cent. On-line Bulletin. Thatcham, United Kingdom (http://www.isc.ac.uk). Karson, J.A., 1998. Internal structure of oceanic lithosphere: a perspective from tectonic windows. In: Buck, W.R., Delaney, P.T., Jeffery, J.A., Karson, A., Lagabrielle, Y. (Eds.), Faulting and Magmatism at Mid-Ocean Ridges. American Geophysical Union, Washington, pp. 177–218. Kissling, E., Ellsworth, W.L., Eberhart-Phillips, D., Kradolfer, U., 1994. Initial reference models in local earthquake tomography. J. Geophys. Res. 99 (B10), 19635–19646. http://dx.doi.org/10.1029/93JB03138. Kwasnitschka, T., Hansteen, T.H., Devey, C.W., Kutterolf, S., 2013. Doing fieldwork on the seafloor: photogrammetric techniques to yield 3D visual models from ROV video. Comput. Geosci. 52, 218–226. http://dx.doi.org/10.1016/j.cageo.2012.10.008. Lienert, B.R.E., Havskov, J., 1995. A computer program for locating earthquakes both locally and globally. Seismol. Res. Lett. 66, 26–36. Madeira, J., Mata, J., Mourão, C., Brum da Silveira, A., Martins, S., Ramalho, R., Hoffmann, D., Masson, D.G., Le Bas, T.P., Grevemeyer, I., Weinrebe, W., 2010. Volcano-stratigraphic and structural evolution of Brava Island (Cape Verde) based on 40Ar/39Ar, U/Th and field constraints. J. Volcanol. Geotherm. Res. 9, Q07015–Q07235. http://dx.doi.org/ 10.1029/2008GC001983. Masson, D.G., Le Bas, T.P., Grevemeyer, I., Weinrebe, W., 2008. Flank collapse and largescale landsliding in the Cape Verde Islands, off West Africa. Geochem. Geophys. Geosyst. 9, Q07015. http://dx.doi.org/10.1029/2008GC001983. Matias, L., Cruz, J., Pena, J.A., Martins, I., Senos, M.L., 1997. A sismicidade registada na Ilha do Fogo durante os primeiros dias de actividade do vulcão na erupção de Abril de 1995. In: Réffega, A., Figueiredo, M.O., Silva, L.C., Costa, F.L., Mendes, M.H., Torres, P.C., Silva, T.P., Correia, E. (Eds.), A erupção vulcânica de 1995 na ilha do Fogo, Cabo Verde. Instituto de Investigação Científica Tropical and Ministério da Ciência e Tecnologia, Lisboa. ISBN: 972-672-863-0, pp. 13–32 (in Portuguese). Mendes, A.S., 1956. Contribuição para o conhecimento da sismicidade dos territórios Portugueses. Boletim da Ordem dos Engenheiros. 5(3), pp. 1–4, (Lisboa. (in Portuguese)). Mitchell, J.G., Le Bas, M.J., Zielonka, J., Furnes, H., 1983. On dating the magmatism of Maio, Cape Verde Islands, Earth Planet. Sci. Lett. 64, 61–76. Mourão, C., Mata, J., Doucelance, R., Madeira, J., Brum da Silveira, A., Silva, L.C., Moreira, M., 2010. Quaternary extrusive calcio-carbonatite volcanism in Brava Island (Cape Verde): a nephelinite–carbonatite immiscibility product. J. Afr. Earth Sci. 56, 59–74. http://dx.doi.org/10.1016/j.afrearsci. 2009.06.003. NASA, 2009. ASTER Global Digital Elevation Model v001, (at http://gcmd.nasa.gov/ KeywordSearch/Metadata.do?Portal = GCMD&MetadataType = 0&MetadataView= Full&KeywordPath = &EntryId = ASTGTM1 (accessed in January 2014)).

D. Vales et al. / Tectonophysics 636 (2014) 325–337 Neves, J.V., 1981. Relatório preliminar da missão geofísica a Cabo Verde no período de I0 de Abril a I I de Maio de 1981. Unpublished report, Instituto Nacional de Meteorologia e Geofísica, Lisboa. (in Portuguese) Pim, J., Peirce, C., Watts, A.B., Grevemeyer, I., Krabbenhoeft, A., 2008. Crustal structure and origin of the Cape Verde Rise. Earth Planet. Sci. Lett. 272 (1–2), 422–428. http://dx. doi.org/10.1016/j.epsl.2008.05.012. Plesner, S., Holm, P.M., Wilson, J.R., 2003. 40Ar/39Ar geochronology of Santo Antão, Cape Verde Islands. J. Volcanol. Geotherm. Res. 120, 103–121. http://dx.doi.org/10.1016/ S0377-0273(02)00367-0. Ramalho, R.A.S., 2011. Building the Cape Verde Islands(Springer Theses) 1st ed. SpringerVerlag, Heidelberg, Dordrecht 978-3-642-19102-2 http://dx.doi.org/10.1007/978-3642-19103-9 (207 pp.). Ramalho, R., Helffrich, G., Schmidt, D.N., Vance, D., 2010a. Tracers of uplift and subsidence in the Cape Verde Archipelago. J. Geol. Soc. Lond. 167, 519–538. http://dx.doi.org/10. 1144/0016-76492009-056. Ramalho, R.S., Helffrich, G., Cosca, M., Vance, D., Hoffmann, D., Schmidt, D.N., 2010b. Vertical movements of ocean island volcanoes: insights from a stationary plate environment. Mar. Geol. 275, 84–95. http://dx.doi.org/10.1016/j.margeo.2010.04.009. Ramalho, R., Helffrich, G., Cosca, M., Vance, D., Hoffmann, D., Schmidt, D.N., 2010c. Episodic swell growth inferred from variable uplift of the Cape Verde hotspot islands. Nat. Geosci. 3 (11), 774–777. http://dx.doi.org/10.1038/NGEO982. Represas, P., Catalão, J., Montesinos, F.G., Madeira, J., Mata, J., Antunes, C., Moreira, M., 2012. Constraints on the structure of Maio Island (Cape Verde) by a 3D gravity model: imaging partially exhumed magma chambers. Geophys. J. Int. 190 (2), 931–940. http://dx.doi.org/10.1111/j.1365-246X.2012.05536.x. Ribeiro, O., 1960. A Ilha do Fogo e as suas erupções, Memórias (série geográfica I), Junta de Investigações do Ultramar2nd ed. (Lisboa, Portugal, 319 pp. (in Portuguese)).

337

Serviço Meteorológico Nacional, 1947–1973. Anuário Sismológico de Portugal, Year Book from Serviço Meteorológico Nacional, No 1–N°27, (Lisboa. (in Portuguese)). Stillman, C.J., Furnes, H., LeBas, M.J., Robertson, A.H.F., Zielonka, J., 1982. The geological history of Maio, Cape Verde Islands. J. Geol. Soc. Lond. 139, 347–361. Torres, P.C., Madeira, J., Silva, L.C., Brum da Silveira, A., Serralheiro, A., Mota Gomes, A., 1997. Carta geológica das erupções históricas da ilha do Fogo: revisão e actualização. In: Réffega, A., Figueiredo, M.O., Silva, L.C., Costa, F.L., Mendes, M.H., Torres, P.C., Silva, T.P., Correia, E. (Eds.), A erupção vulcânica de 1995 na ilha do Fogo, Cabo Verde. Instituto de Investigação Científica Tropical and Ministério da Ciência e Tecnologia, Lisboa. ISBN: 972-672-863-0, pp. 119–132 (in Portuguese). Torres, P.C., Madeira, J., Silva, L.C., Brum da Silveira, A., Serralheiro, A., Mota Gomes, A., 1998. Carta geológica da ilha do Fogo (República de Cabo Verde): erupções históricas e formações enquadrantes. Edited by Laboratório de Tectonofísica e Tectónica Experimental (LATTEX), University of Lisbon, 1 sheet at the 1:25.000 scale. (in Portuguese) Torres, P.C., Silva, L.C., Serralheiro, A., Tassinari, C., Munhá, J., 2002. Enquadramento geocronológico pelo método K/Ar das principais sequências vulcano-estratigráficas da Ilha do Sal – Cabo Verde. Garcia de Orta. 18 (1–2), pp. 9–13, (in Portuguese). Vinnik, L., Silveira, G., Kiselev, S., Farra, V., Weber, M., Stutzmann, E., 2012. Cape Verde hotspot from the upper crust to the top of the lower mantle. Earth Planet. Sci. Lett. 319–320, 259–268. http://dx.doi.org/10.1016/j.epsl.2011.12.017. Wessel, P., Smith, W.H.F., 1998. New, improved version of Generic Mapping Tools released. Trans. Am. Geophys. Union 79, 579. Wilson, D.J., Peirce, C., Watts, A.B., Grevemeyer, I., Krabbenhoeft, A., 2010. Uplift at lithospheric swells-I: seismic and gravity constraints on the crust and uppermost mantle structure of the Cape Verde mid-plate swell. Geophys. J. Int. 182 (2), 531–550. http://dx.doi.org/10.1111/j.1365-246X.2010.04641.x.