Monitoring the volcanic unrest of El Hierro (Canary Islands) before the onset of the 2011–2012 submarine eruption

Authors


Abstract

[1] On 10 October 2011, a submarine volcanic eruption started 2 km south from El Hierro Island (Spain). Since July 2011 a dense multiparametric monitoring network was deployed all over the island by Instituto Geográfico Nacional (IGN). By the time the eruption started, almost 10000 earthquakes had been located and the deformation analyses showed a maximum deformation of more than 5 cm. Earthquake migration from the north to the south of the island and acceleration of seismicity are in good correlation with changes in the deformation pattern as well as with some anomalies in geochemical and geomagnetic parameters. An earthquake of local magnitude 4.3 at 12 km depth (8 October 2011) and shallower seismicity a day after, preceded the onset of the eruption. This is the first time that a volcanic eruption is fully monitored in the Canary Islands. Data recorded during this unrest episode at El Hierro will contribute to understand reawakening of volcanic activity in this region and others of similar characteristics.

1. Introduction

[2] Knowledge of precursors of long dormant volcanoes is limited due to the low number of well documented cases [Sigmundsson et al., 2010]. The seismicity is one of the symptoms associated to volcanic processes that may lead to an eruption and could indicate either an injection of magma from the source region or movement of the melt in a storage zone [Chouet, 1996; McNutt, 1996]. Surface deformation is another precursor related to magma migration that may become critical to monitor a volcano [Cabral-Cano et al., 2008; Sigmundsson et al., 2010]. The geochemical analysis of gases released by the ascending magma can be also useful to forecast a possible eruption [Tedesco, 1995]. Many other techniques have been tested in the last decades as potential monitoring tools, like gravimetry [Carbone et al., 2007; Williams-Jones et al., 2008] or geomagnetism [Zlotnicki and Le Mouel, 1988; Zlotnicki, 1995], both able to provide information about magma migration.

[3] Starting on 7 July 2011, different precursory signals were exhaustively monitored during 96 days until the 10 October, when a submarine volcanic eruption started in El Hierro (Canary Islands, Spain) after, at least, 200 years of quiescence. Little is known about precursors of the historical eruptions in the Canaries. All documented eruptions were preceded by seismicity felt by the inhabitants, lasting between several months (Chinyero, 1909) to a few days (Teneguia, 1971) [Romero Ruiz, 1991]. There are also few reports suggesting reawakening with no eruption, the last one occurred in Tenerife in 2004 [Martí et al., 2009; Domínguez Cerdeña et al., 2011]. El Hierro eruption was the first one in the Canary Islands to be instrumentally monitored from the initial unrest, being also the first submarine eruption reported in about 600 years of historical record in the archipelago. The volcano monitoring system was deployed by Instituto Geográfico Nacional (IGN) which is responsible for volcano monitoring in Spain and has been involved in the management of the crisis since the very beginning, along with the regional government emergency committee (PEVOLCA).

[4] This paper focuses on the volcano monitoring of El Hierro Island during the unrest episode of 2011. First results obtained from many different geophysical and geochemical techniques are presented in order to bring new insight to the body of knowledge on the precursory activity of eruptive processes in the Canary Islands.

2. Geological Setting

[5] El Hierro is located on the southwesternmost edge of the Canary Islands and is the youngest island of the archipelago. The oldest subaerial rocks are dated at 1.12 Ma [Guillou et al., 1996]. The island is the emergent summit of a volcanic shield which rises from a 3800–4000 m depth and grows up to 1500 m a.s.l. Its subaerial part (280 km2) shows the characteristic shape of three convergent rifts, separated by at least three gigantic landslides that formed wide, horseshoe embayments [Masson, 1996; Masson et al., 2002; Gee et al., 2001]. The construction of the island can be divided into three main volcanic phases or edifices (Figure 1): 1) Tiñor volcano in the NE of 0.8–1.2 Ma 2) El Golfo edifice in the NW, 130 ka–550 ka and 3) the rift volcanism with simultaneous activity on the three rifts of the island (NE-Rift, NW-Rift and S-Ridge) [Carracedo et al., 2001; Acosta et al., 2003]. This last stage currently continues active, although in the historical record prior to 2011 there is only one possible eruption at Lomo Negro, AD1793(?) [Hernández-Pacheco, 1982].

Figure 1.

Top left corner: Location of the Canary Islands. Main figure: DEM of El Hierro and its surrounding bathymetry. The three main rift zones (yellow) and the three main collapse scars of the island (red lines) are shown together with the location of the last inland eruption (blue dot) and the 2011 eruption (red cross).

3. Volcano Monitoring Network

[6] The volcano monitoring network of the IGN at El Hierro Island prior to the unrest consisted of two seismic stations plus public data of a GPS station, FRON, belonging to the Canarian Regional Government that has been included in IGN processing since summer of 2010. The two seismic stations are part of the IGN National Seismic Network: a three-component broadband station (CTIG) and a vertical short period station (CHIE). As soon as an increase in the seismic activity was detected by mid July, a dense monitoring network was deployed in order to obtain real time data to assist authorities in emergency management. This network involved not only deformation and seismicity, but also other techniques that could reveal any eruption precursor.

3.1. Seismicity

[7] Seven new seismic stations were installed during the unrest (Figure 2a). On 20 July (the day after the beginning of the seismic unrest) two 3 C short period stations of natural period of 5 s where installed in the north of the island, in El Golfo region, close to the ongoing seismicity (CTAB and CTAN). In the second week of August, three new seismic stations were deployed when the events started to migrate to the southern inland. Two of them were 3 C short period stations with a natural period of 1 s (CCUM and CJUL) and one vertical component short period station with a natural period of 1 s (CMCL). Finally, on 1 October 2011, the seismic network was completed with two other 3 C short period stations with a natural period of 1 s which were placed on the southern and western points of the island (CORC and CRST) (Figure 2a). Data was transmitted on real time to the IGN National Seismic Network, where earthquake locations and local magnitudes were calculated. In order to obtain a better constrain of their focal depths, P and S wave arrivals were manually picked and hypocentral solutions were computed using a 3 layer velocity model based on a previous study of the Canary Islands [Dañobeitia, 1980]. Mean location errors of the whole series are 3.7 ± 1.5 km in horizontal and 5.1 ± 2.2 km in depth.

Figure 2.

(a) Seismicity located in El Hierro before the eruption (circles). Seismic stations are shown as triangles. Depth of the events is shown for vertical cross-sections in N-S direction (top right panel) and E-W direction (bottom panel). White star shows the epicenter of the ML = 4.3 event. (b) Daily horizontal deformation (colored squares and diamonds) with respect to the location of each GPS station (inversed triangles). Seismicity is plotted as grey circles. The bottom panel shows the vertical deformation versus the horizontal displacement in the E-W direction. The color code is equivalent for both seismicity and deformation plots. Both panels show the position of the 2011 eruption (white cross).

3.2. Geodesy

[8] Since the unrest started, the IGN implemented a GPS continuous station densification network, with dual frequency receivers (Figure 2b). The first GPS station (HI01) was placed outside the seismicity area in the last week of July 2011. At the same time, two other stations were deployed close to and on the maximum occurrence of the seismicity in El Golfo region (HI02 and HI03). In the first week of September, another station was added at El Golfo just on the western part of the seismogenic area (HI04). Three of the stations transmitted data in real time while the other two stored them on field and were recovered every two days. GPS measurements were used to quantify surface deformation and to obtain three-dimensional displacements as a function of time. Data were processed using Bernese software version 5.0 [Dach et al., 2007] to obtain daily coordinate solutions. Ocean-loading model FES2004 was applied and the IGS (International GNSS Service) absolute antenna phase centre models were used for satellite and stations antennas. Precise satellite orbits from the IGS were also considered (J. Kouba, unpublished data, 2009, available athttp://igscb.jpl.nasa.gov/components/usage.html). Daily north, east and vertical deformation components were calculated from the coordinates obtained in the process. The reference frame used was ITRF2008 and no correction of plate movement was considered (about 2.4 cm/year NE for this area).

[9] Two TerraSAR-X images were acquired on 11 August and 5 October 2011 over El Hierro Island. An interferogram was processed with DORIS software [Kampes et al., 2003], using a IGN 5 m resolution DEM resampled to 15 m, applying DEM co-registration and precise orbits. Finally, it was unwrapped using Snaphu software [Chen and Zebker, 2001] to obtain the absolute value of the phase between the images. A coherence mask was applied before unwrapping to eliminate bad quality pixels.

3.3. Geochemistry

[10] Several geochemical parameters were measured in order to improve the volcanic monitoring network at El Hierro. Measurements of CO2 flux through the soil were carried out in some areas of the island selected on the basis of the existence of previous data [Barrancos et al., 2008], the location of earthquake epicentres and the presence of particular geological structures [Carracedo et al., 2001]. According to these criteria, CO2 diffuse flux was determined in the central zone of the island, El Golfo and the western rift. Data were obtained using a portable instrument based on the accumulation chamber method [Chiodini et al., 1998]. More than 450 measurements were accomplished during different periods between July and September 2011. Four geochemical continuous stations were installed in three sub-horizontal water galleries (HVER, HTIN, HTIG) and also one well (RSIM) where air and soil temperature and222Rn concentration in air were measured. The first two galleries were also equipped with a sensor for measuring CO2 concentration in air (Figure 3a). Other physicochemical parameters of groundwater such as temperature, pH, electric conductivity and total dissolved solids were also measured at least once a week in four wells in the area of El Golfo since July 2011.

Figure 3.

(a) The remainder IGN monitoring network of El Hierro Island. The map shows the position of the magnetometers (plus signs), the geochemical stations (squares), gravimeters (diamonds) and the wells sampled (small circles). (b) CO2 flux measured in El Hierro. Black dots show measurement locations and the gray circles show the earthquakes located before 13 August. (c) InSAR results. The GPS stations inside the InSAR measurements are shown (triangles). Black orthogonal arrows show the satellite flight path and look direction.

3.4. Geomagnetism and Gravimetry

[11] Four magnetic stations were installed on the island in order to monitor the temporal behaviour of the total geomagnetic field intensity, F, and its possible relation to volcanic processes. Two Overhauser magnetometers were installed in the first and second week of September (MARB and MLLA). A third magnetic station was a proton magnetometer (MMEN) that was initially set in the western slopes of El Golfo in the third week of September, but it was moved to El Julan area the first days of October (MJUL), in response to the southward migration of the earthquake epicentres (Figure 3a). Data treatment and analysis were performed according to Zlotnicki and Le Mouel [1988] and Zlotnicki [1995]. MARB station was used as reference for calculating inter-station F-differences, sincea priori it seemed to be the most distant from the areas with higher probability of eruption as indicated by the location of earthquake epicentres.

[12] Finally, complementing the monitoring network, a gPhone#054 gravimeter was installed in early August for continuous gravity record at El Golfo area (GH01) and moved to a more appropriate location in the center of the island by the end of September (GH02) (Figure 3a).

4. Results: Pre-eruptive Phases

[13] Five distinct pre-eruptive phases can be established according to significant changes in the evolution of the geophysical and geochemical processes that preceded the onset of the eruption:

4.1. Phase I (7–18 July 2011): Slight Deformation

[14] The first sign of reactivation was a northeastward displacement of FRON GPS station which started on 7 July. On 16 July 2011, automatic detectors for individual seismic phases on CHIE and CTIG waveforms showed few tens of local events, whereas the previous mean value was less than one event per day. This value decreased considerably in the following two days.

4.2. Phase II (19 July–3 September 2011): Beginning of the Unrest

[15] On 19 July an important increase up to hundreds of events per day was detected, with earthquakes located on the north of the island (Figure 2a). During the first part of the reactivation, HI01, HI02 and FRON started moving northeastward and HI03 moved to the north for some days and then to the northwest. Seismic activity alternated between relatively calm periods and high energy periods with a maximum rate of 450 located events per day. During this stage 9,8% of the total seismic energy was released. Most of the earthquakes were located in the El Golfo area at 10–15 km depth. By that time, the maximum local magnitude (ML) was 2.7 and only one earthquake had been slightly felt by the population, with a maximum intensity value of II (EMS-98).

[16] Measurements of CO2 flux carried out between 22 July 2011 and 14 August 2011 in several parts of the island revealed the existence of a spatial anomaly in a particular zone at the south of El Golfo near Sabinosa (Figure 3b). Although the abnormal flux was limited to a relative small zone (0.36 km2 approx.), the maximum value obtained reached 620 g·m−2·d−1 which is substantially higher than fluxes measured in previous studies along the island by Barrancos et al. [2008] which show a maximum CO2 diffuse flux of 393.6 g·m−2·d−1 between 1998 and 2006. However, it is possible that previous works had not accomplished measurements exactly on the anomalous zone due to the high slope of the terrain and the fact that the anomaly was rather local.

4.3. Phase III (4–26 September 2011): Migration to the South

[17] Starting in the first week of September, GPS data showed a clear northward movement in every station (Figure 2b) and a significant increase in the deformation rates compared to previous weeks (Figure 4). On the following days, seismicity started to migrate to the southern inland (Figure 2a) and seismic energy release increased significantly. Simultaneously to southern migration, deeper hypocenters were observed (12–17 km in the south of the island). During this period, 6% of the total energy was released and eight of the earthquakes were felt by residents. Maximum MLrecorded was 3.3 which corresponded to the maximum intensity value of this period, IV (EMS-98). A considerable increase in the baseline of222Rn concentrations in air were also detected at RSIM coinciding with the increase in deformation rates (Figure 4). However no other geochemical parameter in the galleries seemed to show any anomalous behavior.

Figure 4.

Temporal evolution of different monitored parameters at El Hierro from July to October 2011. Gray shading shows the accumulated seismic energy. The red and blue points show the northing displacement of two GPS stations with its correspondent error bars (1σ). The F-differences between MLLA and MARB magnetic stations is plotted as green dots. The222Rn temporal variation in RSIM is shown in magenta line. The five phases are separated by gray dashed lines. The upper colored line corresponds to the time color code of Figure 2.

[18] Between 22 and 24 September, measurements of CO2 flux were accomplished in the same zone as the previous month. Data obtained showed a significant decrease in the anomalous area, were values comparable to those measured in the rest of the island were obtained. Despite the impossibility of repeating the measurements at the most inaccessible points, the maximum value detected during the fieldwork was 43 g·m−2·d−1 in an area where the flux obtained a month before had been 318 g·m−2·d−1.

4.4. Phase IV (27 September–7 October 2011): Acceleration of the Process

[19] Starting on the 27 September there was a drastic acceleration of the seismicity. The 69,7% of the total energy release was liberated during these 11 days and 135 of the earthquakes were felt by the population. A maximum intensity value of V using EMS-98 was obtained for a ML = 3.8 event on 7 October 2011. At the same time, magnitudes of the earthquakes became larger, and their location migrated to the south, most of the hypocenters at this period were located offshore, SW of the island, at around 12–14 km depth (Figure 2a). The energy increase, deformation variation, evolution of F-differences between two geomagnetic stations and222Rn concentrations during this stage are shown in Figure 4. Deformation showed acceleration to the north during the first days of this period. On 1 October, and lasting 5 days, a sudden deflation and reinflation process was observed in all the stations (Figure 4). Coinciding with the acceleration in deformation, mean day F-differences between MLLA and MARB magnetic stations showed a clear increase in the variation rate. At the same time that the seismic acceleration started, a peak of222Rn was observed at RSIM with amplitude twice the size of the background and which lasted two hours. This peak did not seem to be an instrumental artifact because it was observed as a soft increase of amplitude during several samples. A new 222Rn peak was observed just after the deformation deflation-inflation process and was followed by a slight decrease in the baseline level.

4.5. Phase V (8–10 October 2011): Forthcoming Eruption

[20] On 8 October, 20:34 h (GMT), a ML = 4.3 earthquake (the highest magnitude up to that moment) occurred 1.5 km from the SW coast of the island at 12 km depth. From that day on, a change in the trend of the superficial deformation suggested the start of the system stabilization. During the night of 8 October and all through the following day, a swarm of 30 shallow seismic events occurred offshore at the south of the island, about 5 km from the coast. Their depths varied from 1 to 6 km and their local magnitudes did not exceed the value of 1.8. This activity continued simultaneously with the deeper and larger events at 15 km depth in the Mar de las Calmas area as the previous days. The energy released during these two days was 14,5% of the total energy release and four of the earthquakes were felt by the population. Finally, at 05:15 h (GMT) of 10 October 2011, a clear volcanic tremor started to be recorded by all of the seismic stations on the island, with the highest amplitudes recorded in the southernmost station, CRST. All data suggested the beginning of a submarine eruption, though evidence of it was not observed on the sea surface until a couple of days later. By the time the tremor signal started, almost 10000 earthquakes had been located and a maximum deformation of more than 5 cm had been measured.

4.6. Additional Results

[21] Temperature, pH and electric conductivity measurements accomplished regularly in four wells of El Golfo during the months prior to the eruption, did not show significant changes that could be interpreted as directly related to the volcanic activity. The highest temperature (26.2°C) and the lowest pH value (7.17) were obtained at RSIM, which was the sampled well closest to the area where the maximum value of CO2 flux was measured.

[22] Figure 3cshows results of SAR interferometry, where the phase differences in the scene were measured in the Line Of Sight direction (LOS) in cm. Direct interpretation of negative values suggests the distance between satellite and ground was shorter in 5 October than in 11 August 2011; while positive values mean the opposite. According to GPS measurements, as vertical movements were larger than east-west movements in those days, and taking into account that TerraSAR-X is almost insensitive to north-south displacements, an increase or decrease in the phase was translated into subsidence (positive values) or uplift (negative values) respectively. A different atmospheric effect between east-west sides of the island is present in the interferogram. It makes it difficult to quantify the magnitude of the deformation without further analysis, but allows identifying changes in the deformation state between images. So, it can be stated that the apparent uplift in El Golfo area reached its local maximum in the western part, involving HI02 GPS permanent station. However the highest values of uplift are found in the area where the seismicity moved from north to south and where no GPS was available. The east of the island seemed to show a subsidence trend but it is very likely to be the consequence of the mentioned atmospheric effect.

5. Discussion and Conclusions

[23] The first sign of the stress caused by the magma intrusion was the deformation on FRON GPS station (Phase I). It was followed by the earthquake swarm that was initially focused on the north of the island (Phase II). The anomaly of CO2 flux, also placed at El Golfo area during this stage, could be related to a higher degree of permeability of the volcanic edifice caused by the increase in fracturation induced by magma intrusion.

[24] During Phase III, both seismicity and ground deformation seemed to trace a magma migration towards the south. The earthquake swarm started to migrate at the same time that the horizontal displacement of El Golfo GPS stations began to rotate to the North, suggesting a southern source of deformation. InSAR analysis corroborated the deformation occurring at El Golfo area but, unfortunately, due to atmospheric conditions, it was not so conclusive in the rest of the island and as far as the magnitude of the deformation is concerned, it requires further analysis. The increase on the 222Rn baseline at the beginning of Phase III could be linked both to the increase in fracturation of the whole volcanic edifice and the magma ascent. It is also remarkable a slight increase in the depth of the located seismicity between phases II and III, although this fact is not clearly explained yet and could be a result of poorer network coverage, a limitation on the location velocity earth model or geometric irregularities of the crust/mantle boundary. However, most of the events were located at 12–15 km depth which corresponded to Mohorovicic discontinuity in the area [Watts, 1994; Ranero et al., 1995]. All these data point to magma accumulation around this discontinuity with a lateral expansion to the South along September (a similar conclusion is also proposed in Carracedo et al. [2012]).

[25] An acceleration of the process was clearly observed during Phase IV, both in the seismic energy release and in the rapid reinflation (Figure 4), which can be interpreted as an overpressurization of the system due to magma accumulation in the south. At the same time, differences in daily mean geomagnetic total intensity between stations showed strong variations that seem to be correlated with deformation trends and suggest piezoelectric or electrokinetic effects linked to ground deformation. Nevertheless, not a definitive causal relationship can be established between both signals, since time series are not long enough to eliminate the contributions of long-period ionospheric, magnetospheric, telluric or oceanic magnetic variations.

[26] The ML = 4.3 earthquake that marked the start of Phase V took place at the W of the bathymetric high in the south of the island, in the area where the highest magnetizations presently occur [Blanco-Montenegro et al., 2008]. This area has been interpreted by these authors as the feeding system of the southern ridge. The depth of the event (12 km) is also compatible with locations of magma pockets below El Hierro computed by Stroncik et al. [2009]. This event could have triggered the ascent of the magma to the surface as the existence of dike swarms in this part of the island suggests that it is a preferential area for magma ascent.

[27] Regarding the usefulness of the various monitoring techniques, it is important to point out that seismological and deformation data, which have been transmitted and interpreted in real time, have been essential to follow the pre-eruptive process and assist in the management of the volcanic crisis. Geochemistry, geomagnetism and InSAR data have been useful to support the other results. However, we have not been able to find a clear correlation between pre-eruptive volcanic activity and gravimetric and groundwater geochemical data. In conclusion, all these data will be very useful to understand future reawakening processes in the Canarian Archipelago and may also help to interpret those occurring in other volcanic areas of similar characteristics.

[28] As a final remark, it should be kept in mind that the main objective of the monitoring network deployed by IGN at El Hierro was to assist authorities in emergency management. This objective required the best possible characterization of the anomalous activity with real-time data transmission, that was really difficult to achieve in the scarce populated areas like those to the west and south to the island. The initial strategy of network deployment was therefore focused in El Golfo area. When the coverage of this area was considered to be good enough, and more equipments were available, efforts were focused on extending the network to the whole island, coinciding with the migration of the seismic activity to the south (Phase III). Therefore, the main lesson learned from post-eruption critical analysis is that global coverage of the area should be accomplished as soon as possible in order to have a real-time characterization of possible lateral magma migrations or new areas of magma injection.

Acknowledgments

[29] The authors wish to thank the National Seismic Network (IGN) and the Geodesy Department (IGN) for their excellent work, help and support throughout the volcanic crisis of El Hierro. We also want to thank the CSIC and UCA scientists Ramón Ortiz, Alicia García, Joan Martí and Manuel Berrocoso for their assistance in the interpretation of the monitored signals. We are indebted as well to every resident and authority of El Hierro Island that provided support and care both to the stations and personnel of the IGN during these months. This work was funded by the Instituto Geográfico Nacional of the Spanish Ministerio de Fomento. Finally, we would like to thank the two anonymous reviewers and the Editor Andrew Newman for their constructive reviews.

[30] The Editor thanks Andrea Di Muro for assisting in the evaluation of this paper.

Ancillary