The measurement of soil CO2 flux variations is a well−established practice in many volcanic areas around the world. Until recently, however, most of these were made using direct sampling methods. These days, a variety of automatic devices providing real-time data now make the continuous monitoring of volcanic areas possible. A network of automatic geochemical monitoring stations (EtnaGas network) was developed by INGV Palermo and installed at various sites on the flanks of Mt. Etna. Here, we present a large set of soil CO2 flux data recorded by the network, dating back 10 years, a period in which several noteworthy eruptive phenomena occurred. Our statistical analysis strongly suggests that anomalous measurements of soil CO2 flux are attributable to volcanic origin and in almost all cases precede volcanic activity. Here, we present the actual data series recorded by EtnaGAS and an interpretative model of the expected behavior of soil CO2 flux (in terms of increase-decrease cycles), which corresponded well with the volcanic activity during this period. Through the use of a comparative approach, incorporating both volcanological and geochemical data, the global soil CO2 flux trends are put into a coherent framework, highlighting close links between the time flux variations and volcanic activities. These insights, made possible from 10 years of uninterrupted data, confirm the importance of continuous monitoring of volcanic soil degassing, and may contribute in the forecasting of imminent eruptive activity or the temporal evolution of an in-progress eruption, therefore facilitating Civil Defense planning in volcanic areas under high-hazard conditions.
 Throughout the Mediterranean area, Mt. Etna is well known for its frequent eruptions and extensive lava flows, being, among all of the basaltic volcanoes, one of the most active in the world. The repeated eruptive activities from the flank of the volcano and from the craters on the summit area have produced different types of volcanic phenomena, varying from weak, passive lava flows to strong paroxysms [Behncke and Neri, 2003; Burton et al., 2005]. The most violent paroxysms and the most extensive lava flows have presented, and continue to present, a serious threat to both local economic activities and to the very high density of people living around the volcano [Andronico et al., 2009].
 The frequent activity of the last two decades has induced the scientific community and Civil Defense to focus more attention upon the surveillance of the volcano, resulting in a variety of volcanic monitoring systems which have been developed over the years, now making Mt. Etna one of the most intensively monitored volcanoes in the world [Bonaccorso et al., 2011a].
 In terms of seismic monitoring, instruments are now very efficient and reliable; in fact, the network for the continuous monitoring of seismic activity was the first one to be set up on Mt. Etna. With regard to geochemical monitoring, however, this was previously only possible using discrete sampling methods, and, whilst well established in several areas of the volcano, was generally carried out at weekly intervals at most. It is only recently that suitable technologies have become available which permit the continuous automated monitoring of any meaningful geochemical parameters on Mt. Etna. The first network of continuous monitoring of geochemical parameters in this region was developed by the Istituto Nazionale di Geofisica e Vulcanologia (INGV) of Palermo [Gurrieri et al., 2008] to monitor soil CO2 flux (EtnaGAS network). CO2 was selected as it is recognized as one of the most important gas species for monitoring volcanic activity [Symonds et al., 2001]. Mt. Etna is one of the world's most degassing volcanoes, particularly in terms of the total rate of magmatic soil CO2 release [Allard et al., 1991; Symonds, 1998]. The first of the EtnaGAS network stations to be installed was in December 2002 (P78), with further additions made over the following years. Today, EtnaGAS comprises 18 automatic stations which are located close to the main volcanic structures of Mt. Etna in areas of the volcano characterized by strong soil CO2 emissions.
 The data sets acquired from the network allow us to investigate the role of passive CO2 soil degassing and its relationship with the dynamics of the volcano itself. A previous lack of insight into this has been particularly due to the fact that EtnaGAS is the first automated geochemical network for continuous monitoring which widely and homogeneously covers a volcano. Following the acquisition of raw data, we use an interpretative method which takes into account the continuous variations of CO2 degassing from all the network stations, evaluating their anomalies in relation to volcanic activities. In this paper, we show the results of 10 years of CO2 soil flux variation, describing different cycles of increase-decrease. We have included data from published literature regarding the plume CO2/SO2 ratio variations and the volcanic activities, showing how, in these contexts, the cycles of CO2 variations can provide an extremely useful aid in the understanding of volcanic activity on Mt. Etna and, moreover, lead to an enhanced ability to predict the course of further activity.
2. Volcanic Activity During 2004–2012
 Mt. Etna is one of the most active volcanoes in the world, erupting every 1.5 years on average over the last century. This study focuses on five distinct eruptive periods, as described below.
2.1. Period 2004–2005
 The first starting on 9 September 2004, in which a new fracture appeared on the southeast flank of the SEC, producing a lava flow that lasted until 8 March 2005; this eruption was characterized by the absence of any prior seismic signal or any initial paroxysmal activities. For these reasons, several authors consider the 2004–2005 volcanic activity to be the first recorded example of a passive eruption caused by sliding on the eastern flank.
2.2. Period 2006
 This eruptive period can be split into two phases: The first began on the night of 14 July 2006, and lasted until 23 July 2006; this phase was characterized by Strombolian activity with strong lava fountaining which reached heights in excess of 250 m and was clearly visible from the city of Catania. No evident geophysical changes preceded this phase [Bonforte et al., 2008] and it was only upon the onset of the eruption that an increase in volcanic tremor amplitude was observed [Neri et al., 2006].
 The second phase started on 31 August and ended on 15 December, with a brief interruption of a few days. This eruption created a field of fractures on the eastern flank of the Southeast Crater (SEC) at about 3000 m asl, where a lava flow was emitted. This second phase involved new episodes of Strombolian activity: (i) at the beginning; (ii) on 13 October, when a new fracture opened up on the eastern flank of the SEC; and (iii) on 26 October, also involving the south flank of the Bocca Nuova Crater (BNC), which began erupting simultaneously with the continuing activity on the Southeast crater. Minor episodes of Strombolian activity occurred at SEC before the end of the eruption, some of which were particularly strong although short in duration (16, 19, and 24 November) [Behncke et al., 2009].
2.3. Period 2007
 The volcanic activity during 2007 was also characterized by very short, although, powerful events. Each of these began with normal Strombolian activity and culminated with lava fountains and lava flow. The first occurred on 31 March and finished just some hours later. The others occurred on 11 and 29 April and on 7 May. Again, in the middle of August, new Strombolian explosions occurred at the SEC, culminating in lava fountains and a lava effusion during the night of 4 September and the early hours of 5 September. The last event occurred on 23 November on the same site as that of 4 September, with the same characteristics.
2.4. Period 2008–2009
 During the first month of 2008, small ash emissions (e.g., 14, 15, 20, 25 February) were frequently observed from the summit area. On 10 May, a strong explosive phase accompanied by a lava fountain and a large lava flow began in the afternoon and lasted until the night of the same day. On 13 May, there was new paroxystic activity, accompanied by strong tremor signals. This marked the onset of a new eruption which, after the initial paroxysms, continued with very little other than effusive activity, interrupted only by brief Strombolian activity between 11 and 14 March, with a consequent increased lava output. The 2008–2009 eruption lasted more then a year, finishing in mid July 2009.
 The remaining part of 2009 proceeded without further eruptive activities, with the important exception, however, of 6 November, when a small degassing “hole” opened up on the flank of the SEC from which a red glowing light could be seen by all the villages around the southern slopes of the volcano. This continued without further consequence until the end of the year.
2.5. Period 2010–2012
 This is the last of the five eruptive periods discussed in this paper. In this period, a new crater (Pit Crater) was formed, caused by some intracrateric collapses that occurred during the summer of 2010, which considerably enlarged the small hole that had opened up at the base of the Southeast Crater on 6 November 2009, increasing its diameter from just a few meters to a few hundred meters. At the beginning of the new year (during the night between 11 and 12 January 2011), a huge paroxystic phase began at the new Pit Crater, accompanied by a lava fountain and a lava flow which lasted for several hours. The violence of this paroxysm erupted a volume of magma close to 1.55 × 106 m3 in only 4 h [Ganci et al., 2012]. Although the type of activity of this eruption is similar to the paroxysms that occurred during 2007, the volume of magma erupted was four times higher than in 2007 [Harris et al., 2011]. The paroxysm of 12 January was the first of 18 paroxysms that took place over 2011, all of which bore the same characteristics: (i) short in duration; (ii) strong in energy, as demonstrated by the very large lava fountaining; and (iii) emitting a comparable volume of erupted lava. During the first 4 months of 2012, a further seven paroxysms with similar features followed (5 January; 9 February; 4 and 18 March; 1, 12, and 24 April), reaching a total of 25 paroxysms since the first in 2011.
 It is interesting to compare the total amount of lava erupted during the 2011–2012 period with that of the previous eruptions of 2008–2009: Harris et al.  estimated 68 × 106 m3 for the 2008–2009 eruptions, which lasted 420 days, while Ganci et al. , using a comparable method, estimated the total amount of lava erupted for all the 18 events in 2011, including the first of the 2012 (5 January), at about 25.4 × 106 m3. This is remarkable if we consider that the huge amount of lava erupted over these 19 episodes was emitted in the equivalent of less than 9 days, and that this value might have been underestimated as the weather conditions during the events on 18 February and 19 September did not permit any measurement survey. Furthermore, we do not know the contributions of lava erupted from the subsequent six episodes (9 February to 24 April) as there are no published data. This indicates that the volcanic activities during 2011–2012 were very different to those observed in the previous years.
3. Network Features
 EtnaGas network consists of 18 automatic stations located close to the main volcanic structures of Mt. Etna and sited in areas of the volcano (Figure 1) characterized by significant soil CO2 emissions. The sites were selected according to the volcano tectonic structures present, which are known to facilitate the movement of volcanic gases across the crust toward the surface [De Gregorio et al., 2002; Gurrieri et al., 2008].
 The first station of the network to be installed was in 2002, and the final configuration was reached in 2008. The monitoring stations of the network were entirely developed by INGV Palermo and are able to monitor parameters such as CO2 and CH4 soil flux, atmospheric T, P, and RH, rain, wind speed, and wind direction. Data are acquired at hourly sampling periods, which is a good compromise between detail of information (as a strong variation in CO2 flux at higher frequency is improbable) and energy consumption.
 CO2 flux measurements are carried out by the dynamic (or dilution) method [Gurrieri and Valenza, 1988], which is based on the CO2 content in a mixture of air and soil gas (Cd) obtained using a probe inserted into the soil at a depth of 50 cm; soil gases enter through the base of the probe and are diluted with air above; these gases are pumped to an IR spectrophotometer, which measures the CO2 content in the mixture. As deduced by Gurrieri and Valenza, the dynamic concentration is proportional to the soil CO2 flux through an empirical relationship (1) found experimentally in the laboratory for a range of applicable permeability 0.36–123 µm2 and pumping flux 0.4–4.0 L/min [Camarda et al., 2006a, 2006b]:
where ΦCO2 is the soil CO2 flux expressed in kg m−2 d−1, k is the numerical values of the gas permeability (µm2), and Cd is the numerical value of molar fraction of the diluted CO2 concentration. In this work, ΦCO2 is converted into g m−2 d−1. For more details of the method, see Camarda et al. [2006a, 2006b].
4. Filtering Data From Atmospheric Influences
 When environmental factors affect soil degassing, it is necessary to filter original data to minimize the effect on endogenous CO2 release. With this in mind, soil CO2 flux data are analyzed and filtered before investigating the relationship between them and volcanic activity.
 The filtering model shown in this paper consisted of two stages: First, we reduced the time series, taking one value per day at noon; then, any lack of data (usually not more than a few days) was reconstructed by linear interpolation; finally, all the data sets were processed by FFT analysis and the periodic components, if any, were removed by Matlab Ideal filter. This filter does not introduce any phase rotation. In this filtering stage, all the frequencies higher than 7 day−1 were suppressed.
 In the second stage, the resampled data series were compared with atmospheric data in order to recognize which of them needed to be further filtered. This was done in two different ways: (i) applying an FFT filter which suppresses all the frequencies between 300 and 400 day−1 in order to determine and eliminate the seasonal variations and (ii) using a multiple linear regression to evaluate the relationship between soil CO2 gas flux and atmospheric parameters and normalize to a constant T and P value.
 The graphs in Figure 2 show the results for the SV1 station, which, among all the stations of the EtnaGas network, is most influenced by temperature variation. As shown in the figure, the second filtering method gave the best results (Figure 2d) and was applied to the stations Sv1, Agro, N1, N2, Passop, Fondachello, and Parcoetna, which were also found to be influenced by atmospheric temperature and/or pressure. The other signals did not show traces of atmospheric influence so for these, only the first stage of filtering was used in order to avoid introducing further data manipulations.
 Besides temperature and pressure, soil moisture can also appreciably influence soil degassing and is strongly related to atmospheric humidity and rain. The influence of this upon soil degassing, however, is relatively short in terms of time. As shown in Figure 3, 3C is one of the monitoring stations that seems to be most affected by rainfall and humidity (note the positive correlation between this parameter and CO2 flux raw data), although this effect visibly decreases a few days after an episode of rain. Similar behavior has also been observed at the Solfatara of Pozzuoli and Vesuvius volcanoes [Granieri et al., 2003] and Vulcano [Camarda et al., 2006a, 2006b]. In all cases, the authors observed a fast decrease of the rainfall effect on soil degassing, which disappeared within a few days. As shown in Figure 3, this component is totally negligible in the resampled data.
 Finally, the possible interference related to groundwater is not considered here as the modest scrubbing effect on CO2 flux [Symonds et al., 2001] may possibly reduce the total magnitude of the CO2 flux emitted, but cannot perceptibility modify the trend over time.
5. Statistical Method to Identify Anomalous Soil Degassing
 Soil CO2 flux in volcanic areas originates from two main sources: Volcanic (mainly located along active tectonic structures) and/or organic. During quiescent periods, either of these two factors may generate soil degassing, which also varies in quantity according to soil type and thickness, climate conditions, residual magma degassing, and the connection of the site with the magmatic CO2 source. When new magma enters into the volcanic system, the magmatic CO2 component in soil degassing increases and, in some cases, the organic component becomes totally negligible. Many authors have used different methods to distinguish the contributions of the two different sources of CO2 [e.g., Kanemasu et al., 1974; Chiodini et al., 1998, 2008; Giammanco et al., 1998, 2006]. The effects of site in relation to the surrounding areas have also been investigated, giving results which indicate that it is not possible to use one single threshold for stations in different locations [Aiuppa et al., 2004]. In agreement with these last authors, in this work we do not consider a constant threshold of CO2 flux for all the investigated areas as the monitoring stations are distributed over a wide area of the volcano and vary considerably in terms of soil thickness and type, vegetation cover, and presence of tectonic structures, with differences in magmatic CO2 migration toward the surface. We, therefore, adopted a statistical method based on a normal probability plot (NPP), focusing on the data acquired by each station. As an example in Figure 4, CO2 flux data series of SML1 station is shown, resampled as described above. Different distributions of values were obtained for each site (different populations identifying different ranges), but at least two different populations were recognized. In this paper, we assume that the lower populations (if more than one) represent a typical quiescent period (organic production and residual CO2 magmatic degassing), while the higher values are considered anomalous, the lowest of which assumed as the anomaly threshold for the site. For the SML1 station, this value was equal to 44 g m−2 d−1 (Figure 4b).
 Previous studies on P78 area confirm these assumptions. Pecoraino and Giammanco , working on a long discontinuous data series, found a clear positive correlation between CO2 flux and the isotopic composition of carbon in which δ13C becomes significantly more positive and magmatic. This evidence validates the NPP method as a way to isolate the part of the signal which is more related to volcanic activity.
 Table 1 shows all the thresholds computed by NPPs for each monitoring station. In this paper, we assume that values higher than the anomaly thresholds indicate in-progress volcanic activity. Figure 4c shows the time series of CO2 flux, where the anomalies are highlighted using the threshold values.
Table 1. Main Features of Each Monitoring Stationa
Min and max CO2 flux (g m−2 d−1) threshold and standard deviation for each EtnaGAS station. The value of the threshold is computed as described in the text and in Figure 4.
6. Local Anomalous Degassing Analysis
 Soil CO2 flux anomalies recorded by the EtnaGas network can be explained by previous studies on gas solubility in basaltic magmas [Moretti and Papale, 2004], and on the typical CO2 content in fluid inclusions in Etnean basalts, [Kamenetsky and Clocchiatti, 1996; Spilliaert et al., 2006]. These studies suggest that CO2 content in a basaltic magma is released starting from pressures ≤500 MPa (15–20 km bsl). From this perspective, it is reasonable to attribute the strong increase in CO2 flux recorded by the monitoring stations either to a new input of volatile-rich magma entering a deep storage area or to its ascent from the lower volcanic conduits. The excess of CO2, no longer in equilibrium with the magma owing to the changed conditions, will be released from the early stage of this hypothesis, reaching the surface of the volcano before and/or during a possible subsequent volcanic activity. The variation in CO2 flux, as described above, may be related to an in-progress deep magmatic activity that could evolve into a volcanic event observed at the surface; however, it cannot be assumed that this is always the case. In some situations, in fact, this phenomenon might even determine “false positive” cases with regard to CO2 flux variations.
 Considering the above analysis carried out to identify the volcanic anomalies, we can see in Figures 5a–5c how the CO2 flux anomalies in all the 18 monitoring stations of EtnaGAS network are distributed. As these were installed over a period of approximately 5 years, the length of data recorded is different for each site so in order to simplify an initial qualitative analysis, the stations were divided into three different groups of comparable life length. The first group consists of five monitoring stations which have been in operation since 2002 (Figure 5a), at the end of the violent 2002 eruption. Signals from this group regarding levels of intensity, time durations of anomalous degassing, and time of occurrences are not the same for all sites owing to differences that are specific to each location; however, time densities of CO2 flux anomalies increase appreciably before and during volcanic events. This is particularly evident during the 2004–2005 eruption, in which a peak is recorded by all these stations. In the second group, the oldest stations were installed during the second half of 2005 (Figure 5b). For all of these, the same considerations as those of the previous group are valid. In particular, as shown in Figure 5b, the maximum correlations between anomalous flux and volcanic activities occurred during 2006–2007, and a secondary relevant relationship was observed during 2011–2012. The third group consists of the most recently installed monitoring stations, the first of which became operative at end of 2007. Although few of these sites recorded anomalous flux levels during the eruption of 2008–2009, the measurements from this group did show a particularly high correlation with the eruptive activities of 2011–2012 (Figure 5c).
 As a general summary of the qualitative analysis given above, our investigations into the relationships between volcanic activity and CO2 soil degassing on Mt. Etna can be outlined thus: EtnaGAS network recorded several periods of anomalous soil CO2 flux levels, and all locations showed anomalies that can be correlated with volcanic events. Although these are remarkable results, it is not easy to correlate the anomalous measurements with volcanic activity in a simple, intelligible manner, owing to the large number of stations and the aforementioned variations in terms of their specific locations. In addition, a lack of synchronicity is recognizable in some cases between different sites. These incongruities can be easily explained, however, considering the size of the area covered by the EtnaGAS network (more than 1.4 × 103 km2) and that each volcanic event has been characterized by different magmatic dynamics involving different tectonic systems and heterogeneous geologic features of the volcano, each affecting the site in different way.
7. Data Discussion
 In accordance with the previous consideration regarding the uneven distributions of anomalies, it was important to use a different data processing approach to obtain a unified signal of CO2 flux from all the monitoring stations. Therefore, each soil CO2 flux series was normalized in the range 0–1 and calculated using the equation:
 where n is the number of stations, Φi(t) is the measured flux at the site i and time t, Φimin and Φimax are, respectively, the minimum and the maximum values of the i flux series. In this way, each station has the same weight and ΦnNorm does not have the flux physical quantity, but can be seen as a type of variable representing the total soil CO2 flux at a specific time.
 As shown in Figure 6, we applied this model to two different groups of stations: First to the original group of the older 11 stations, the last of which was installed in 2006, and then to the total group of 18 stations (the original 11 plus the further seven more recently installed stations, the last of which was set up in 2008). Obviously, when the analysis focused on the larger group of stations, the data series available were shorter in time, though covered larger sectors of the volcano. The group of older 11 stations covered a smaller area but provided an opportunity to analyze the data within a greater time window. The difference in value totals between the two groups is clearly related to the number of stations considered; however, it is still possible to see that the curves accord very well (R2 > 0.77), albeit with some modest difference (March 2009 to April 2009 and July 2010). In this paper, therefore, we present the ΦnNorm calculated using data from the group of 11 stations (Φ11Norm), which is a good compromise between the number of stations evenly distributed on Mt. Etna and the data series length, which can be backdated to July 2006.
 Before comparing the results of Φ11Norm time variation against other geochemical and geophysical parameters, some considerations need to be made on the geographical distribution of the EtnaGAS network and the models of gas solubility in basaltic magmas. The first of these regards the location of the network's monitoring stations, which have been sited in peripheral areas of the volcanic edifice (Figure 1) up to a level of 1900 m asl (this is because Mt. Etna has a rather long winter season with a snow and ice covering a large part of the top of the edifice). Furthermore, Φ11Norm variations must be considered in a volcanic context, where a new input of volatile-rich magma enters a deep storage area and moves toward the shallowest part of the volcano, releasing its gas content as a function of the pressure and of the solubility of each gas species in the magma melt. These considerations suggest two possible scenarios regarding magma degassing, degassing processes of the overall edifice, and CO2/SO2 ratios in the plume gases (EtnaPlume network).
 Figure 7a shows a schematic cross section of Etna where the position of the geochemical monitoring stations are approximated, differentiating between those dedicated to plume monitoring and those for CO2 soil flux, and also indicating the different zones of temporary magmatic storage. Figure 7b describes the first scenario, where a magmatic batch rises from a depth of about 10 km bsl to the intermediate storage zone. As a consequence of depressurization, CO2 is released from the magma and subsequently detected by the monitoring stations. Between approximately 5 and 2 km, the gas of the plume will be characterized by an increase in the CO2/SO2 ratio [Aiuppa et al., 2007]. If the batch of magma continues to move toward the surface, the depressurization determines a rapid decrease in the SO2 solubility in magma and a consequent decrease of the CO2/SO2 ratio in the plume gases; this trend is shown with a red line in Figure 7b and is consistent with the cycles of the plume CO2/SO2 ratios during the 2007–2009 volcanic activity on Mt. Etna [Aiuppa et al., 2010]; soil CO2 flux also follows the same trend (green line of Figure 7b). Although CO2 is continuously released from magma, the source (magma) moves to the top of the edifice and does not supply the peripheral areas where the CO2 monitoring stations are located. This first scenario suggests that in the initial phase, soil CO2 flux and plume CO2/SO2 ratios are positively correlated.
 A second scenario is shown in Figure 7c, where a new batch of volatile-rich magma feeds the deep storage chamber without moving toward the surface. In this case, thermodynamic conditions persist in which the less soluble CO2 is liberated, while the more soluble species (HCl, HF, SO2) remain stored in the magma melt. For the reasons described above, an increase-decreasing cycle of soil CO2 flux can be predicted which anticipates the increase-decreasing plume cycle, decoupling the two signals.
 These two scenarios can be considered as two end members of the degassing process on Mt. Etna, although it is reasonable to expect that the magmatic activity will also be affected by a dynamic exchange between both conditions, especially during an eruption. In this case, the trend of the two signals (CO2 flux and CO2/SO2) could be affected, resulting in a loss of their predicted correlation.
 In order to verify our hypothesis, Φ11Norm is shown in Figure 8a, together with some data from EtnaPlume network, and other previously published data. In Figure 8a, 11 intervals have been highlighted (indicated by Roman numerals) where several increase-decrease soil CO2 flux cycles are visible.
7.1. Interval I
 The first relevant increase-decrease cycles of Φ11Norm occurred during the 2006 eruption. This first interval shows a clear cycle previous to the onset of the eruption on 31 August, and can be interpreted as a new input of volatile-rich magma before the reawakening of the Southeast crater and after the previous July paroxysm.
7.2. Interval II
 The second interval considered spans the period between August and December 2006. Here, there are two recognizable cycles, in which the first reached the highest value on 22 October and then rapidly decreased just before the opening of the new vent on the South flank of the Bocca Nuova Crater, on 26 October, a separate eruption from the coexisting Southeast Crater eruption. We believe that this cycle is particularly important as it is the only case during the entire period analyzed that involved both the Southeast craters and the Bocca Nuova craters simultaneously. The smaller second cycle began after 26 October and finished just after the 2006 eruption. It is not possible to link this second cycle to any definite sequence of events, being either possibly related to the later stages of the 2006 eruption, or to a new contribution of a deep batch of magma that anticipated the third interval, where, in fact, there is a lack of correlation with the increase-decrease CO2/SO2 plume ratio, as we can see in detail in the next interval.
7.3. Intervals III–VII
 The following five intervals (III–VII in black), with the exception of the third, coincided with increase-decrease CO2/SO2 plume ratio cycles, described in Aiuppa et al. . Of these, it is particularly clear that intervals IV, V, and VI show a good correlation between the plume and the Φ11Norm signal. As outlined above, the lack of correlation with the third interval cannot be explained simply, although it is plausible to conjecture that the absence of a clear cycle of Φ11Norm in this interval is due to the fact that it is integrated within the very high variations of the previous second interval. As an alternative explanation, it is also possible to suppose that the plume variation cycle in interval III is shifted to a later time compared to the second Φ11Norm cycle in interval II, probably because the magma feeding the first eruptive activity of 2007 was stored in an intermediate reservoir for a time, before continuing toward the upper part of the volcano.
 Interval VII also shows a low correlation between the two signals, plume and Φ11Norm. This is, however, an expected result within the model as it coincides with the 2008–2009 eruption in which there was most probably a dynamic passage between the two scenarios described above (Figures 7b and 7c). Regarding the 2008–2009 eruption, the Φ11Norm cycle in interval VII can be considered concluded just before the brief Strombolian activity on 11–14 March 2009, after which the eruption continued to decrease steadily until finishing entirely.
 These first seven intervals show, on the whole, a slight decline of the Φ11Norm, which finishes at the end of the 2008–2009 eruption. In order to explain this decreasing trend, we have to consider that: (i) the 2008–2009 eruption is recognized as a weak eruption which, following the initial paroxysm of the first day, continued with a mild effusion; (ii) compared with the previous 2006 eruption, and in relationship with the length of each eruption, the daily rate of magma erupted in 2006 is 3.5 times greater than in 2008–2009 (39 m3 in 69 days for 2006 eruption to 68 m3 in 420 days for 2008–2009 eruption); (iii) the petrographic analysis of the lava samples between 2007 and 2008 show a well-defined trend toward a more evolved magma in the products erupted and are, therefore, less volatile rich. This trend was interrupted on 17 June 2008, in which the products of the effusive activity were being fed by a more primitive magma [Bonaccorso et al., 2011b]. After 17 June, the analysis of the subsequent effusive products again showed a less primitive bulk rock composition (Corsaro R.A. INGV—Catania personal communication); and (iv) in addition, the CO2/SO2 ratio cycles from Voragine crater show a moderate decreasing trend of peaks between 2007 and 2008, which is more evident if we also consider that during the 2006 eruption, CO2/SO2 ratios reached much higher values, reaching a peak of 45 on 26 October [Aiuppa et al., 2007], than in the following cycles highlighted in this paper. These considerations clearly indicate, therefore, that the 2008–2009 eruption was sustained by a less volatile-rich magma, compared to the previous 2006 eruption.
7.4. Interval VIII
 The more stable degassing phase, which began at the end of 2008 and lasted until the first day of July 2009, was interrupted again by a new cycle of the Φ11Norm variation. This increase-decrease phase can be located between the first days of July 2009 and 6 November, immediately before the opening of the small “hole” on the eastern flank of the Southeast Crater, which started to degas without interruption. After this cycle, the subsequent degassing rate continued with a relatively high Φ11Norm average. To understand better the Φ11Norm variations from this interval, it is rather important to consider the evolution of this new structural feature over time; after its initial formation of 6 November 2009, the structure began to collapse during the following summer of 2010, generating a pit crater which continued to degas. The subsequent paroxysms starting on 2011, from what was then referred to as “Pit Crater,” were responsible for the formation of a new cone, which is now referred to as “New South-East Crater,” which is, at the time of writing, comparable in size to the present Southeast crater. This evidence of a possible change in volcanic behavior seems to correspond with the later phases of eruptions between 2011 and the early part of 2012, which were characterized by strong Strombolian episodes and notable fountaining.
7.5. Intervals IX–XI
 In these last intervals, the Φ11Norm variations showed three other major increase-decrease cycles. A first increase of Φ11Norm (interval IX) starts at mid June, reaching the highest values around October, followed by a fast decrease phase, finishing at the end of 2010. On 12 January, strong Strombolian activity from the Pit Crater marked the beginning of a “new season” of activity on Etna, which was distinguished by numerous episodes of quite spectacular paroxysms of lava fountaining, with jets of incandescent lava reaching over 500 m in height.
 Interval X began abruptly with a renewed input of fresh volatile-rich magma. It is noteworthy that the decreasing phase of the Φ11Norm begins before the most intensive period of Strombolian activity between June and October 2011.
 Finally, interval XI shows a strong Φ11Norm variation in a relatively briefer time than the previous. These last major increase-decrease cycles end before the final period of fountaining that occurred throughout the first 4 months of 2012.
 We can observe that the Φ11Norm cycles between intervals VIII and IX correlate well with the CO2/SO2 ratios at Bocca Nuova crater (BNC), which, during this period, was the most active crater of the volcano (unfortunately, other BNC data are not available for this period, particularly during the winter, owing to station maintenance difficulties, and no significant data were recorded by the other two EtnaPlume network monitoring stations).
 As pointed out in Aiuppa et al. , cycles of increase-decrease of CO2/SO2 plume ratios are very closely linked to volcanic activity. These authors, using a multidisciplinary approach, showed that the resumption of volcanic activity at SEC was systematically anticipated by a complete CO2/SO2 plume cycle (increasing-decreasing) and marked by sharp increases in tremor. In agreement with geophysical data, this model is consistent with preeruptive magma storage at 1–2.8 km reservoir below the summit crater.
 All the decreasing phases of the Φ11Norm cycles depicted in Figure 8a correlate very well with the fast increase of tremor occurring before each eruptive event, as deducible from INGV-CT reports (www.ct.ingv.it). These are not included in the graph in order to simplify Figure 8a. These observations are coherent with those made in Aiuppa et al.  and Cannata et al. . Furthermore, Φ11Norm increase-decrease cycles can be put in relationship with the eruptive activity and the volumes of erupted magma [Harris et al., 2011; Ganci et al., 2012], even if, as explained in section 'Local Anomalous Degassing Analysis', a simple direct relationship cannot be expected between them. Nevertheless, it is notable that the eruptions during 2011–2012, although very short in duration, produced a quantity of erupted lava of approximately 25 × 106 m3 (the time amounting to only about 9 days in total, calculated until the first paroxysm in January 2012), which is more comparable with the 2006 eruption (in total 39 × 106 m3) than with the small 2007 eruptions (in total 2.5 × 106 m3). This aspect is rather important as it explains the large Φ11Norm time variations during the intervals IX, X, and XI (detail in Figure 8b).
 Additional information can be obtained using a more detailed analysis of the Φ11Norm over a shorter time scale. As shown in Figure 8c, Φ11Norm time variation allows us to distinguish some other minor increase-decrease cycles during the final phases of the 2011–2012 eruptions. The very powerful fountaining episodes that characterized this period occurred at the final decreasing phase of these minor cycles (CminI–CminV). In all of these, there persists a remarkably good correspondence with the above described model and the following paroxysms on Mt. Etna of the last year. Similarly, other minor cycles between the last three major cycles during intervals IX–XI are also recognizable if not always as easily distinguishable. A possible explanation for this is that the major cycles are sustained by a strong flux of gas related to the ongoing volcanic activity and subsequent rapid series of paroxysms, which then produced a masking effect upon the minor cycles. As is visible in Figure 8a, in the last three intervals, the Φ11Norm signal appears to be affected by a sort of “noise,” the effect of the overlapping minor cycles, which is absent before interval IX. It is remarkable to note that this noise, and its interference with the signal, coincides with the periods of strong fountaining (2011–2012). In contrast, we can see that the minor cycles shown in Figure 8c, which mark the final eruptive episodes, are far less masked as no new input of magma could support an increasing flux of gas.
 These analyses of soil CO2 flux from EtnaGAS network give a new insight into the volcanic activities during the 2004–2012 period, and contribute to a better understanding of the degassing processes on Mt. Etna. The findings discussed above suggest the following considerations: (i) The results of Φ11Norm presented here, which have been integrated within a multidisciplinary context, shed new light on our understanding and interpretation of CO2 flux degassing cycles in relation to volcanic activities, through which it has been possible to distinguish major cycles of degassing that are clearly connected to the main volcanic activities occurring over the last 6 years. Furthermore, the minor cycles which, when not masked by the major cycles, show the same cycling trends that anticipated the explosive activity and exceptional fountaining episodes which took place between 2011 and 2012. (ii) The results obtained are clearly due to the improved system of data acquisition made possible by the network, which has taken us from the previous periodic sampling surveys to the now almost real-time acquisition of high-resolution data. In addition, the EtnaGAS network demonstrates a real robustness in terms of its maintenance and, therefore, reliability, comparable to the existing well-tested seismological network, thus guarantying a continuity of solid data acquisition. (iii) Finally, it must be considered that the analysis of Mt. Etna volcanic activities in relation to the Φ11Norm signals has focused upon a period of time in which only the summit craters were involved and are possibly very different from lateral eruptions, where the main conduit of the central craters would perhaps play a different role in the migration of gas toward the surface. For example, the first four monitoring stations of the EtnaGas network recorded high rates of CO2 soil flux during the 2004–2005 lateral eruption despite the fact that this eruption was not characterized by any explosive behavior. Future lateral eruptions may, therefore, contribute in a different way to the total rate of CO2 degassing on Etna. With the expansion of the network, we will be able strengthen further our ability to identify the different regions of the volcano that are affected by anomalous soil degassing, which may be particularly useful in the case of lateral eruptions and their future evolution over a short-time period, especially considering their closer proximity to populated areas.
 These results not only confirm the importance of the continuous monitoring of soil CO2 emissions in the surveillance of volcanic activity on Mt. Etna, but, moreover, open up an interesting scenario for the surveillance of other active volcanoes around the world.
 The authors would like to thank the Editor for his support, and the reviewers for their important and carefully considered suggestions, which have no doubt fundamentally improved the content of this work. The authors would also like to thank S. Caffo (volcanologist of Parco dell'Etna) for his indispensable help in the setting up of the network inside the park. This study has benefited from funding provided by the Italian Presidenza del Consiglio dei Ministri—Dipartimento della Protezione Civile (DPC). This paper does not necessarily represent DPC official opinion and policies.