Soil gas Radon ( 222Rn) and Thoron ( 220Rn) activities have been continuously measured during the period of the recent volcanic unrest that occurred at El Hierro, Canary Islands, at two different monitoring stations, namely HIE02 and HIE03. Significant increases in soil 222Rn activity and 222Rn/220Rn ratio from the soil were observed at both stations prior to the 2011-2012 submarine eruption off the coast of El Hierro, showing the highest increases before the eruption onset and before the occurrence of the strongest seismic event (M = 4.6). Statistical analysis showed that the long-term temporal trend of filtered radon data matched closely that of seismic energy release during the volcanic unrest. The observed increases of 222Rn are related to the rock fracturing processes (seismic activity) and the magmatic CO2 outflow increase, as observed in HIE03 station.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Radon is continuously emitted from any rocks and soils containing uranium in concentrations which are directly dependent of the amount of 238U in the parent rocks [Tilsley, 1992]. The concentration of 238U is relatively small in most of soils and rocks, whereas in granitic rocks its concentration can reach tens of ppm. Decay of 238U through a series of shorter-lived radionuclides eventually produces 226Ra, which has a half-life of 1,620 years. 226Ra decays directly to 222Rn by alpha-particle emission. Two other radon isotopes can be produced by natural decay, 219Rn (action) from 235U and 220Rn (thoron) from 232Th. Actinon (219Rn) has a half-life of 4 seconds, which is too short to be used as a useful geochemical tracer in volcanological investigations. Radon (222Rn) and thoron (220Rn) are two radioactive isotopes of radon with a half-life of 3.8 days and 54.4 seconds, respectively. Owing to its geochemical properties, radon cannot transport over long-distances by diffusion alone. The transport of radon over long-distances occurs advectively and needs the existence of a naturally occurring flux of a carrier gas [Kristiansson and Malmqvist, 1982; Etiope and Martinelli, 2002]
 Monitoring of 222Rn and 220Rn in ground and surface waters has become in the last decades an important tool to forecast earthquakes and volcanic eruptions [Chirkov, 1975; King, 1986; Ui et al., 1988; Igarashi and Wakita, 1990; Igarashi et al., 1995; Ohno and Wakita, 1996; Heiligmann et al., 1997; Toutain and Baubron, 1999; Virk, 1986; Virk et al., 2001; Hernández et al., 2004; Zmazek et al., 2005; Neri et al., 2006; Giammanco et al., 2007; Pérez et al., 2007]. The spatial distribution of surface radon anomalies suggests that radon ascends towards the upper part of earth's crust mainly through cracks or faults [Nishimura and Katsura, 1990; Etiope and Martinelli, 2002; Walia et al., 2005; Yang et al., 2005]. Other authors have monitored the activity of radon and radon/thoron ratios to detect and delimitate uranium ore bodies in the subsurface [Fleischer, 1980] and to find hidden faults/fractures in the ground [Guerra and Lombardi, 2001; Baubron et al., 2002].
 It is commonly accepted that gas migration at a large scale ten to hundreds of meters is supported by advection (pressure changes), prevailing over diffusion (concentration changes) [Etiope and Martinelli, 2002]. Gas migration is related to the existence of a gas source itself (fluid reservoirs such as hydrocarbon pools in sedimentary basins, geothermal fluid in high heat flow regions or fluids linked to magmatic and metamorphic phenomena), and to the existence of preferential routes for degassing [Hernández et al., 2004; Burton et al., 2004; Yang et al., 2005; Neri et al., 2006; Giammanco et al., 2007]. The nature of the driving force can change during gas ascent, depending on the physical-geological conditions in the environment that the gas encounters. Finally, variations in temperature, pressure, mechanical stresses, chemical reactions and mineral precipitation can change the gas-bearing properties of geological formations.
 The 2011-2012 submarine eruption that occurred at El Hierro Island in the Canaries, Spain, has been an exceptional opportunity to monitor volcanic activity using a multidisciplinary approach, with soil gas radon as one of the most useful geochemical parameters to understand the dynamics of this submarine eruption. Here we describe the results obtained during our monitoring of the 2011-2012 submarine eruptive activity of El Hierro by means of a discrete soil radon survey performed at the entire surface environment of El Hierro Island and by using two soil radon monitors installed at 2 and 18 km distance from the eruptive vent. In order to have a better interpretation of soil radon data, we have compared them with seismic data recorded by the network held by the National Geographical Institute (IGN), in terms of the seismic energy release, calculated using the Gutenberg-Richter law [Gutenberg and Richter, 1954].
 The purposes of this article were: (1) to evaluate soil radon distribution at El Hierro Island during the period of volcanic unrest; (2) to evaluate the temporal evolution of soil radon measured by continuous monitoring stations and (3) to discuss the correlation between the observed spatial and temporal patterns of soil radon and the eruptive and seismic activity. To achieve these objectives, we have measured soil gas radon because: (1) the results can be compared with those from other volcanoes; (2) soil radon is relatively easy and cheap to measure and (3) it has been demonstrated that radon responds rapidly to stress changes within the volcano, mainly during periods of volcanic unrest. [Chirkov, 1975; Connors et al., 1996; Cigolini et al., 2005; Federico et al., 2008; Pérez et al., 2007; Cigolini et al., 2009].
2 Geological Background
 El Hierro is the smallest and most south-western Island of the Canarian archipelago with an area of 278 km2 (Figure 1a). It represents the summit of a volcanic shield elevating from the surrounding seafloor at depth of 4000 m to up to 1501 m above sea level. The main volcano-tectonic features of the island are three main volcano-tectonic axis (Figure 1b). El Hierro Island is the youngest in the Canarian archipelago, with the oldest subaerial rocks dated at 1.12 Ma [Guillou et al., 1996]. Recent volcanic activity occurred mainly along the three volcanic ridges bearing NE, S and SW with respect to the centre of the island. Before the 2011-2012 submarine eruption and for the last 500 years, Hernández Pacheco  reported a single volcanic eruption during 1793 at Lomo Negro, although its occurrence is questionable (carbon-14 indirect dating) and most probably consisted of an offshore eruption [Romero and Guillén, 2012].
 In the last 37 ka, El Hierro has been covered with lavas erupted in the last stage of its volcanic evolution, and deep embayment has been produced by giant landslides between the three rift zones. The most recent one was the El Golfo failure on the northwest flank of El Hierro, which took place approximately 15 ka ago [Masson, 1996]. Stroncik et al.  carried out a thermobarometric and petrologic study on basanites erupted from young volcanic cones along the submarine portions of the three El Hierro rift zones to reconstruct magma plumbing and storage beneath the island. They concluded that small, intermittent magma chambers might be a common feature of oceanic islands fed by plumes with relatively low magma fluxes, which results in limited and discontinuous magma supply.
2.1 Chronology of the 2011-2012 Submarine Eruption
 After more than two hundred years of repose, at the end of July 2011, the two seismic stations deployed by the IGN since the beginning of the 1990s registered a large number of relatively small earthquakes (M < 2.5). The quakes were mainly located in the north part of the island at depths between 8 and 15 km, indicating the onset of a volcanic-seismic unrest at El Hierro. The earthquake catalogue during the period of unrest includes more than 12,000 events (Fig. 2), the largest had magnitude of 4.6 and occurred on November 11, 2011. After almost three months of intense seismic activity, on October 10, 2011, at 05:15 (UTC), the dominant character of seismicity changed from discrete earthquakes to continuous tremor. An underwater eruption was confirmed on October 12, 2011 by visual observations off the coast of El Hierro, about 2 km south of the small village of La Restinga, when a brown patch of warm water appeared on the sea surface. During the submarine eruption, a large water discoloration area was observed on the sea surface, ranging from light-green to dark brown color, due to the intense discharge of high temperature hydrothermal fluids as well as magmatic gases and steamy lava fragments. Preliminary petrological analysis of the eruptive products evidenced a substantial presence of low-SiO2 magma, and indicated that the eruption tapped a magmatic source in the upper mantle [Castro et al., 2011]. GPS measurements [Sagiya et al., 2012] indicated inflation directly underneath El Hierro between August and late September, which is in good agreement with the epicenter location of seismicity over the same period. Before the occurrence of the submarine eruption on October 12, an increase of the diffuse CO2 emission was recorded in a geochemical station located at Llanos de Guillén [Pérez et al., 2012], together with an increase in the total soil He emission, probably as response to stress/strain changes at depth [Padrón et al., 2012]. On 5 March, the Scientific Committee stated that the submarine eruption was over (Smithsonian report, 29 February-6 March 2012), although the volcanic process that started on mid July 2011 had not finished (at least two new intrusion related events have occurred with hundreds of earthquakes and clear deformation up to date).
3 Material and Methods
 With the aim of providing a multiparametric geochemical approach, continuous measurements of 222Rn and 220Rn have been performed since 2005 and discrete soil 222Rn and 220Rn survey in the entire island was performed between July 25 and August 3, 2011. Due to the volcanic unrest which occurred from June 2011, a special effort was made to improve the surveillance program at this volcanic island, consisting of increasing the number of GPS antennas (from 2 to 7) and geochemical stations (from 3 to 8) as well as the frequency of the soil gas and hydrochemical surveys.
3.1 Discrete Soil Radon Monitoring
 Between July 25 and August 3, 2011, a soil radon survey was carried out on the entire surface of El Hierro Island consisting of selecting 601 sampling sites after taking into consideration the main geological and volcanic features of the study area as well as accessibility. This survey was performed a few days after the start of the seismicity and two months before the onset of the submarine eruption. Soil radon was measured at each sampling site with a SARAD radon monitor, model RTM-2010-2 (produced by SARAD GmbH, Dresden, Germany) connected to a stainless steel probe inserted at 40 cm depth in the soil. Between the probe and the instrument a water trap was placed to avoid water entry in the instrument. The 220Rn and 222Rn activities were determined by high-resolution alpha spectroscopy of the deposited decay products. The 222Rn activity is obtained from the recorded number of 218Po alpha decays, and the 220Rn activity from the recorded number of 216Po alpha decays. The SARAD radon monitor has a short sampling cycle and responds rapidly to changes in the gas concentration. The accuracy of measurements depends on two factors: gas concentration and integration time. The integration time used was 10 min. The sampling device is also provided with built-in temperature and humidity sensors.
 Soil radon concentration data was used to produce a spatial distribution map using sequential Gaussian simulations (sGs) provided by the sgsim program [Cardellini et al., 2003]. The procedure of the sGs program is composed of the following steps: (1) normal score transformation of the original data to transform the data into a normal population, (2) experimental variogram computation of the normal score transformed data, (3) variogram model assignation to the experimental variogram, (4) sequential Gaussian simulation of N equiprobable realizations and (5) back-transforming the normal score data into simulated values of the original variable. The final map is produced as an average of 100 equiprobable realizations performed over a grid of 27,149 squared cells (100 m × 100 m).
3.2 Continuous Soil Radon Monitoring
 In August of 2005, two geochemical stations (HIE02 and HIE03) were installed at San Simon well (Fig. 2, N 27º 45.28’; W 18º 6.41’; 26 m a.s.l.), and at La Restinga Village (Fig. 2, N 27º 38.54’; W 17º 58.97’; 30 m a.s.l.), to measure 222Rn and 220Rn activities by means of a SARAD RTM-2010-2 radon monitors during a period of volcanic quiescence. HIE02 and HIE03 stations are located 18 and 2.2 km away from the 2011 submarine eruptive vent (Fig. 2). Soil gas was collected by a cylindrical chamber, 75 cm long and 15 cm in diameter, made of a PVC pipe inserted at depth of 75 cm in the ground with its walls being thermally insulated to avoid the influence of air temperature fluctuations on soil radon emissions (Fig. 3). Soil gas is pumped (3 L/min) continuously through a polyamide pipe to the place where the instrumentation is installed and then sent back to the PVC pipe. A water trap is placed before the radon monitor to avoid water entering into the instrument. At both stations, soil radon was measured on an hourly basis. Since the beginning of the seismo-volcanic unrest period, a time series of 7,075 observations of soil radon activity was recorded at HIE02. A total of 8.2 % of radon data were missing due to radon monitor failures, with the largest lack of data occurring on February 14th and March 26th, 2012. Regarding the data recorded from HIE03 station, a total of 7,400 observations of soil radon were recorded in the same period, with 16.7 % of missing data. The effect of the missing values on the overall data set is not likely to produce spurious correlation results. Barometric pressure data was obtained from a meteorological station installed at station for continuous monitoring of soil CO2 and H2S efflux, HIE01 (N 27º 42.92’; W 18º 1.20’, 1,154 m a.s.l.) and was used to filter its potential influence on the soil radon discharge process [Pinault and Baubron, 1996]. Data was stored in an internal memory module of the RTM-2010-2 and automatically downloaded via GSM from Instituto Volcanológico de Canarias (INVOLCAN) facilities. Also, at HIE03 station, soil CO2 concentration was measured on an hourly basis by means of a NDIR (non-dispersive infrared) spectrophotometer (Dräger Polytron IR transmitter) with a double-beam IR detector compensated in temperature. Measuring range was adjusted between 0 and 1% by vol. Measurement accuracy of 3% was acquired for a reading at 350 ppm.
4 Results and Discussion
 Table 1 shows a summary of the results and uncertainties of the soil radon survey performed in July-August 2011. Soil 222Rn values ranged from non-detectable values up to 23,758 Bq m-3, whereas soil 220Rn values ranged from non-detectable values up to 24,668 Bq m-3. On the other hand, the time series of soil 222Rn showed variations ranging between 0 to 16,459 Bq m-3, and 0 to 1,646 Bq m-3 for HIE02 and HIE03 geochemical stations, respectively. Time series of soil 220Rn showed variations between 0 and 14,864 Bq m-3, and between 0 and 803 Bq m-3 for HIE02 and HIE03 geochemical stations, respectively. Table 2 shows a summary of the soil 222Rn and 220Rn data as well as barometric pressure measured at HIE01 geochemical station.
Table 1. Descriptive Statistic of the Recorded Variables by the Automatic Geochemical Station
Number of measurement
Soil 222Rn (Bq m-3)
0.0 – 23,758
Soil 220Rn (Bq m-3)
0.0 – 24,668
0.0 – 189
Table 2. Descriptive Statistic of the Recorded Variables by the Automatic Geochemical Station
No. of measurements
Soil 222Rn (Bq m-3)
0.0 – 16,459
Soil 220Rn (Bq m-3)
0.0 – 14,864
0.0 – 222
Soil 222Rn (Bq m-3)
0.0 – 1,646
Soil 220Rn (Bq m-3)
0.0 – 802
0.0 – 155
Soil CO2 concentration (ppm)
306.9 – 1,380
Barometric Pressure (HPa)
881 – 901
4.1 Spatial Distribution of Soil Radon
 In order to distinguish possible different contributions for soil radon enrichments, the probability-plot technique [Sinclair, 1974] was applied to 222Rn data to check whether their Log values come either from unimodal or polymodal distributions (Figure 4a). Two distinct modes were found: population I or background (97.1% of the total data), with a mean of 602 Bq m-3, and population II or peak (0.8%), with a mean of 11,707 Bq m-3. The existence of the peak population suggests a contribution from deep magmatic degassing of the El Hierro volcanic system to the soil radon. The background soil 222Rn values could be mainly due to a shallow soil degassing, in which it is assumed that the 222Rn is coming from shallow volcanic rocks and minerals [Giammanco et al., 2007]. To check this assumption, we plotted soil 222Rn vs soil 220Rn activities to distinguish between deep and shallow contribution to radon in the soil atmosphere. Fig. 4b shows that most of the soil gas samples plot along the shallow soil degassing trend and only few samples plot along the deep magma degassing trend, strengthening the results of the probability-plot method.
 Soil 222Rn data was used to construct spatial distribution maps for El Hierro Island using Sequential Gaussian Simulation (sGs). The experimental variogram for soil 222Rn activities was fitted with a spherical model with the following parameters: nugget 0.25 and range of 1000 m.. A distribution map (Fig. 5) was constructed for each studied area using the average of the different values simulated at each cell. Most of the study area showed background values of soil 222Rn activity. However, relative high soil 222Rn activities were clearly identified along the three volcanic rifts of the island, indicating a structural control on soil radon degassing. These volcanic structures are characterized by high vertical permeability, favoring the ascent of gases to the surface. Etiope and Martinelli  suggested that carrier gases, e.g., CH4 and CO2, may play an important role in controlling the migration and transport of trace gases such as Rn and He towards the surface. They concluded that advective transport movement in fissured rocks is an effective means of rapid, long-distance gas migration [Yang et al., 2003]. Melián et al.  carried out a soil CO2 efflux survey on the entire surface of El Hierro Island and found a distribution of high values along the main rift zones of El Hierro. This result is in agreement with the distribution of radon activities found with the present study. The good spatial correlation observed when radon data is compared with those of soil CO2 efflux from Melián et al.  in the studied area supports this hypothesis. The short half-life of 222Rn, implies that much of the 222Rn measured is derived from a shallow source unless it is rapidly transported to the surface. Its limited migration indicates that a significant portion of the radon must have a 226Ra/238U source in the shallow local aquifers and/or in the weathered bedrock of El Hierro Island (range of U (ppm) and Th (ppm) content in recent lavas between 0.75-2.29 and 3.01-8.74, respectively; Lundstrom et al., 2003). Soil 222Rn activity and CO2 efflux [Melián et al., 2012] spatial distribution maps showed significant correlation coefficient of r2 = 0.31. The relatively good spatial correlation observed between soil 222Rn and CO2 efflux suggests that CO2 may act as a 222Rn carrier to the surface, indicating that during the survey, the mechanisms controlling the transport of CO2 and 222Rn to the surface environment seems to be similar for both gases. In addition to this, spatial distribution of soil 222Rn/220Rn ratio anomalies (Fig. 6) showed also a similar behaviour as that of soil 222Rn, with higher 222Rn/220Rn ratios observed along the three rift zones of El Hierro. This support the hypothesis that these geological structures act as preferential pathways to the ascent of gases from depth.
Giammanco et al.  investigated the behaviour of CO2 and radon gases during their transport to the surface, taking into account the marked difference in the half-lives of the 222Rn and 220Rn. They considered a two-component degassing model, assuming that the most likely mechanism for the transport of these gases was mixing of deep magmatic gases with gases equilibrated in shallow porous soils. As the weight fraction of the 222Rn in the system that comes from the “old” component increases, the (222Rn/220Rn) ratios increases. In the case of El Hierro, the observed correlation between soil 222Rn activities and CO2 efflux values is in agreement with this mixing model of deep magmatic gases with gases equilibrated in shallow porous soils.
4.2 Continuous Monitoring of Soil Radon and Seismic-Volcanic Activity
 Figure 7 shows the soil 222Rn time series recorded at HIE02 and HIE03 stations, between July 01, 2012, and June 01, 2012. Since the beginning of the selected observation time until the middle of August, 2011, soil 222Rn activity time series at both stations were characterized by a low variance, with soil 222Rn activity values at HIE02 and HIE03 showing a mean value of 121 and 98 Bq m-3, respectively. From August 15 to October 4 a sharp increase in soil 222Rn activity was recorded at HIE02, reaching a maximum value of 16,460 Bq m-3 on October 4, 2011. Regarding HIE03, soil 222Rn activity increase reached its highest value (1,595 Bq m-3) on October 3, 2011. These two peaks of soil 222Rn activities were observed, respectively, eight and nine days before the eruption onset on October 12, 2011. These two geochemical precursory signals are in good agreement with that recorded by Pérez et al.  for CO2 fluxes during the same span of time. The second largest peak of soil 222Rn activity at HIE02 and HIE03 stations occurred 14 and 13 days before the strongest seismic event (M = 4.6; November 11) during all the volcanic unrest period. At HIE03 station, this anomalous increase on soil 222Rn activity was also accompanied by a high CO2 increase (Fig. 8) likely related to the strongest seismic energy release associated with the impending eruption. The above observations highlight the importance of carrier effect of CO2 (whose advective flux is controlled by pressure gradients) on the movement of radon towards the surface. A positive correlation (r2 = 0.65) and a similar behaviour are also observed between the temporal variations of radon activity and CO2, demonstrating the carrier effect of CO2. However, sometimes 222Rn changes seemingly precede those of CO2, as observed in the second half of October 2011. Indeed, just after the eruption onset, high radon emission occurred, suggesting relatively shallow rock fracturing with consequent soil permeability increase. In this case, when CO2 does not act as a carrier, radon escapes first followed by CO2, permeability increase deepens so much as to drain deep CO2 to the surface. Giammanco et al. , proposed a model of two end-member components to explain the behaviour of CO2 efflux and radon isotopes (222Rn, 220Rn) in volcanic gas emissions: (1) a shallow soil gas with low efflux CO2 flux and low (222Rn/220Rn) and (2) a deep, magmatic gas with high CO2 efflux and high (222Rn/220Rn), concluding that when the flux rate of the magmatic gas component is high, the CO2 efflux increases and the abundance of magmatic 222Rn increases significantly as well. This is clearly observed in Fig. 8, with highest CO2 and 222Rn values measured during October-November 2011, when both seismicity and volcanic activity were more intense. This indicates that the flux rate of magmatic gas component during that period was the highest.
 According to Ibáñez et al. , we have also grouped the seismic sequence into three distinct phases, which correspond to well-separated spatial clusters and distinct earthquake regimes (A, B and C in Figure 9). They interpreted that the early phase of activity (A) was characterized by high rates of seismic energy released and relatively low magnitudes (elevated b-value: 2.25), thus reflecting the initial intrusion of magma from the upper mantle into the crust. In contrast, a b-value of 1.25 was observed during phase B, which is closer to what one would expect for ordinary tectonic activity. However, the eruption occurred during phase B. Therefore, Ibáñez et al.  concluded that the earthquakes occurred in the period July-September reflected magma migration from the upper mantle to shallower crustal levels through a highly fractured zone, whereas the larger magnitude earthquakes recorded in late 2011 may reflect crustal relaxation around the magma reservoir that had fed the submarine eruption. Figure 9a and 9b show the 168 h moving average of soil 222Rn measured at HIE02 and HIE03 stations, respectively, together with hourly barometric pressure and daily released seismic energy. Phases A, B and C are also shown. Both couples of time series show a good temporal correlation.
 Since the relationships between radon and atmospheric pressure time-series are complex [Cigolini et al., 2009], a signal processing analysis was performed. To do so, we applied a frequency analysis to the data recorded on an hourly basis. This analysis allowed us to investigate the presence of periodic components in the 222Rn and 220Rn time series by applying the Fourier transform. The spectrograms of Fig. 10a and Fig. 10b for HIE02 and HIE03 time series, respectively highlight major frequencies that match the combined effects of daily earth tides and temperature variations (every 12 h) and their multiples, as well as the alternation of moon phases (~14.5 days and its multiple 29 days) together with moon-quarters (~7 days). However, barometric pressure and tidal forces do not seem to actively modulate radon degassing at both time series since there was no correlation between these parameters.
 In order to measure the similarity of both time series with the seismic energy release, a cross-correlation analysis as a function of a time-lag was applied to both time series (Figures 11 and 12). Maximum Pearson Correlation Coefficients (0.35 and 0.59) were estimated for soil 222Rn considering a time lag of 13 and 31 days (at HIE02 and HIE03, respectively) relative to the released seismic energy. Therefore, by lagging the time series of soil 222Rn for HIE02 and HIE03 stations 13 and 31 days, respectively, the correlations between soil 222Rn activities at both stations with released seismic energy is found to be maximum. The observed correlations, which are above confidence limit, suggest that the observed increases on soil 222Rn activity are related to the onset of the submarine eruption and could be useful to forecast in the future new episodes of seismic energy release. However, cross correlation analysis between the ratio of 222Rn/220Rn and seismic energy release show the Maximum Pearson Correlation Coefficient (0.44 and 0.27) a time-lag of 4 and -5 days at HIE02 and HIE03, respectively. These observed correlations practically without time-lag it's in agreement with the high emissions during October-November 2011 and due to short half-life of 220Rn.
 Time series of 222Rn/220Rn ratios measured at HIE02 and HIE03 stations (Fig. 13a and 13b, respectively) also showed a good temporal correlation with the released seismic energy during the period of observation. In particular, at HIE02, the highest 222Rn/220Rn ratios were observed before the eruption onset and before the occurrence of the M = 4.6 earthquake (Fig. 13a). The peak in the radon ratio recorded during October-November 2011 is the highest in the whole time series, suggesting a deeper source of the gas. This is corroborated by the hypocentral location of seismicity during November 2011 and the timing of the presumed injection of new magma beneath El Hierro volcanic system [Ibáñez et al., 2012]. Two more peaks in the radon ratio were recorded in January and May 2012, both of them preceding the increase in seismicity that occurred in February-March 2012 and the beginning of a new seismic-deformational crisis on June 24, 2012, respectively. The time series of 222Rn/220Rn ratios at HIE03 showed the highest values before the eruption onset, although this parameter does not seems to be as sensitive to changes in the pressure conditions of gas sources as HIE02 site (Fig. 13b).
 In addition to the observed increases of soil 222Rn activities and 222Rn/220Rn ratios recorded at the HIE02 and HIE03 stations, other relevant geochemical changes were also observed. Padrón et al. , reported a drastic increase of the diffuse helium emission from the surface of El Hierro Island several days before the onset of the submarine eruption, followed by a drastic decrease between October 20 and 26. According to Padrón et al. , at the beginning of the period of volcanic unrest, magma movement beneath El Hierro Island may have produced new fractures and microfractures in the shallow crust, thus favouring the ascent of deep seated gases to the surface. Among volcanic gases, SO2, H2S and CO2, due to their relative high solubility in water, were almost completely dissolved in aquifers or they underwent chemical reactions that hindered their detection at the surface. However, helium can ascend through an aquifer easily and thence scape to the surface from deep areas through structures with high vertical soil permeability, thus enhancing the helium content in soil gases.
 Finally, another complementary analysis was to study the correlation between 222Rn and 220Rn. Figure 14 shows distinct behaviours in the linear correlation between these two gases during the seismic phases A, B and C for HIE02. From July to September 20, 2011 (phase A on Figure 14), we identify two trends (slopes i and ii) which correspond to different sources of radon as is observed in Figure 4b, a deep and shallow contribution. During most of phase A, radon degassing is mainly characterized by a deep magma contribution reflecting the initial intrusion of magma from the upper mantle into the crust, with the highest deep contribution (slope i) occurring at the end of phase A, just before the start of the first significant seismic energy release (phase B) which ended in the submarine eruption. However, between 14-16 September (slope ii), a decrease on the 222Rn/220Rn ratio was observed, suggesting a shallow contribution to the radon degassing. From September 21 to October 10, 2011 (phase B on Figure 14), just before the eruption onset on October 12, a good correlation (r2 = 0.8) is observed with a trend suggesting a mixing between different gas sources. During this phase, highest 222Rn and 220Rn activities were recorded during all observation periods, or all of the observation period indicating strong deep and shallow contributions due to the magma movement to the surface. Changes in stress/strain due to the increase of seismic energy release might have significantly increased the level of fracturation in the surrounding rocks favouring the emission of radon gas to the surface. Finally, from October 11, 2011, (phase C on Figure 14), when the largest magnitude earthquakes were recorded reflecting crustal relaxation around the magma reservoir that had fed the submarine eruption, the shallow degassing component for radon was predominant. However, during this phase, the highest 222Rn/220Rn ratio occurred a few days before the occurrence of the strongest seismic event (M = 4.6; November 11) during all the volcanic unrest period, which favoured the released of radon from a deeper source.
 Although high radon activity in soil is usually associated with faults and active seismic areas, the slow velocity for diffusive transport and the short half-life of radon make it highly unlikely for deep radon to reach the surface only through diffusion [Heiligmann et al., 1997]. Therefore, reasonable transport mechanisms to explain the observed high radon activities at El Hierro could be advective transport and/or greater availability of radon in the soil near the surface. However, the spatial distribution of soil radon suggests that advection is the main mechanism controlling the transport of radon, since anomalies at the surface were found along the three volcanic rifts (areas with higher vertical permeability and greater release of carrier gases such as CO2). Seismic/volcanic activity during the period of volcanic unrest may have produced pressure increases along existing faults and fractures, thus enhancing the pressure gradient of magmatic gases and hence channelling more radon to the surface. This phenomenon has also been observed at other volcanic systems, such as Galeras (Colombia) during periods of volcanic activity [Heiligmann et al., 1997]. Alparone et al.  explained the peaks in 222Rn activity before the paroxysmal summit activity at Mt Etna due to temporary increases in gas pressure within the shallow plumbing system of the volcano, leading to increased gas leakage through fracture systems intersecting the main volcanic conduit. However, Neri et al. , observed an increase by several orders of magnitude in radon activity shortly before the phase of eruptive activity of Mount Etna volcano in July 2006 due to micro-fracturing of uranium-bearing rocks, induced by the geodynamic stress of the volcanic-hydrothermal system. The injection of magma at El Hierro during phases A and B described above produced deformation along the entire volcanic edifice of the Island [Sagiya et al., 2012], causing probably changes in pore spaces and favouring the pushing of gases out to the surface. This possibility is corroborated by the observed increases of soil radon at both geochemical stations before the eruption onset and the occurrence of the strongest earthquake during the period of unrest. Increases in gas pressure along existing faults and fractures within the shallow plumbing system of El Hierro during the pre- and sin-eruptive periods seem a plausible explanation for the observed increase in the 222Rn and 220Rn activities at HIE02 and HIE03 stations. The observed increase of magmatic CO2 flux due to the magmatic intrusion and the behaviour of CO2 and radon gases during their transport to the surface (carrier effect), also supports the observed increases in the 222Rn/220Rn ratio before the eruption onset.
 The geochemical monitoring of soil gas 222Rn and 220Rn carried out at El Hierro during the 2011-2012 volcanic-seismic crisis has shown to be of particular interest to understand the dynamics and evolution of this eruptive process. The submarine eruption onset and the occurrence of the strongest earthquake (M = 4.6), occurred on November 11, were preceded by clear geochemical anomalies in the soil 222Rn and 222Rn/220Rn ratios measured by the geochemical stations HIE02 and HIE03, indicating input of magmatic gases from deeper levels. During approximately 6 years of continuous geochemical monitoring at HIE02 and HIE03, no such anomalies had been observed, with the volcano at quiescence state. The increase of magmatic-gas pressure due to magma movement towards the surface is the most plausible explanation for the increases in radon activities observed both at HIE02 and at HIE03.
 During the period of study, recorded time series did not show any temporal correlation between soil radon emissions and seasonal barometric pressure variations. Signal processing analysis showed that radon emissions at both stations were not modulated by tidal forces, and cross correlation analysis performed on both of the time series showed a positive relationship between soil 222Rn and soil 222Rn/220Rn ratios emissions and seismic energy release, supporting the hypothesis that soil 222Rn activity variations acted as a precursory signals of the onset of the submarine eruption. Assuming that anomalous 222Rn originated from hydrofracturing of rock, from direct magma degassing, or from both, depending the soil 222Rn/220Rn ratio, then these processes almost ceased or continued at much reduced rate through November 2011.
 Since, strong seismicity is normally concentrated at the onset of eruptions, whereas strong gas emissions can be detected also afterwards. As soil gas radon activities and soil CO2 increased prior to the occurrence of major seismic volcanic events and prior to the eruption onset, these gases can be efficiently used as an initial warning sign of the pressurization of magma beneath El Hierro Island, together with other geochemical and geophysical data.
 This research was financially supported by project MAKAVOL from MAC 2007-2013 Transnational Cooperation Program of the European Union, by project ProID20100158 from Canarian Agency for Research, Innovation and Information Society (ACIISI), Canary Islands Government, as well by the projects ALERTA and ALERTA II (financially supported by INTERREG IIIB Azores-Canaries-Madeira), Dirección General de Universidades e Investigación of the Canary Islands Government under the project PI2001/025, and by the Cabildo Insular de Tenerife (Spain). We are also grateful to the Water Community San Simón ad Cultural Center of La Restinga for providing logistic support as well as to Cabildo de El Hierro and the staff of the Security and Emergency Coordination Center of El Hierro Island (CECOI). Roma Burhop is warmly thanked for improving the English.