Long term variations of the Campi Flegrei, Italy, volcanic system as revealed by the monitoring of hydrothermal activity

Transcription

1 Click Here for Full Article JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 115,, doi: /2008jb006258, 2010 Long term variations of the Campi Flegrei, Italy, volcanic system as revealed by the monitoring of hydrothermal activity G. Chiodini, 1 S. Caliro, 1 C. Cardellini, 2 D. Granieri, 3 R. Avino, 1 A. Baldini, 2 M. Donnini, 2 and C. Minopoli 1 Received 17 December 2008; revised 26 September 2009; accepted 16 October 2009; published 27 March [1] Long duration time series of the chemical composition of fumaroles and of soil CO 2 flux reveal that important variations in the activity of the Solfatara fumarolic field, the most important hydrothermal site of Campi Flegrei, occurred in the period. A continuous increase of the CO 2 concentrations and a general decrease of the CH 4 concentrations are interpreted to be the consequence of the increment of the relative amount of magmatic fluids, rich in CO 2 and poor in CH 4, hosted by the hydrothermal system. Contemporaneously, the H 2 O CO 2 He N 2 gas system shows remarkable compositional variations in the samples collected after July 2000 with respect to the previous ones, indicating the progressive arrival at the surface of a magmatic component different from that involved in the episode of volcanic unrest ( bradyseism). The change starts in 2000, concurrently with the occurrence of relatively deep, long period seismic events which were the indicator of the opening of an easy ascent pathway for the transfer of magmatic fluids toward the shallower, brittle domain hosting the hydrothermal system. Since 2000, this magmatic gas source is active and causes ground deformations and seismicity as well as the expansion of the area affected by soil degassing of deeply derived CO 2. Even though the activity will most probably be limited to the expulsion of large amounts of gases and thermal energy, as observed in other volcanoes and in the past activity of Campi Flegrei, the behavior of the system in the future is, at the moment, unpredictable. Citation: Chiodini, G., S. Caliro, C. Cardellini, D. Granieri, R. Avino, A. Baldini, M. Donnini, and C. Minopoli (2010), Longterm variations of the Campi Flegrei, Italy, volcanic system as revealed by the monitoring of hydrothermal activity, J. Geophys. Res., 115,, doi: /2008jb Introduction [2] Campi Flegrei caldera is one of the most dangerous volcanic areas in the world. It includes part of the city of Napoli, the town of Pozzuoli and numerous densely inhabited villages. The last eruption occurred in 1538 at Monte Nuovo. The eruption followed a period of ground uplift, which suddenly interrupted a secular subsidence which continued with the same rate after the eruption [Dvorak and Mastrolorenzo, 1991; Troise et al., 2007]. [3] Campi Flegrei caldera is known for frequent episodes of ground uplift and subsidence. The uplift phases are systematically accompanied by seismic activity. Issel [1883] coined the word bradyseism (from the Greek bradi = slow and seism = movement) to mean slow subsidence or uplift of the ground. The word is in use in the volcanological literature concerning the Campi Flegrei caldera since the 1 INGV, Naples, Italy. 2 Dipartimento di Scienze della Terra, Università di Perugia, Perugia, Italy. 3 INGV, Pisa, Italy. Copyright 2010 by the American Geophysical Union /10/2008JB beginning of the 20th century, to mean the vertical ground movements in historical times [Orsi et al., 1999, and references therein]. [4] The most recent uplift started in 1968 and reached a maximum value of about 3.5 m in 1985 near the town of Pozzuoli. The process occurred during two major bradyseisms, the first in and the second in [Barberi et al., 1984]. Since that time, subsidence has started, which generally followed an exponential decay until ; since then, an apparent stable period began, leaving a maximum uplift of 2.5 m with respect to the onset of the crises. This general behavior was interrupted by four mini uplift periods which occurred systematically every 5 6 years (i.e., 1989, 1994, 2000, and 2006). All the uplift periods of the area were accompanied by seismicity, and, in particular, many thousands of earthquakes occurred during the two major crises in and in [5] The maximum ground uplift and the location of the earthquakes during the crises are focused in the center of the Campi Flegrei caldera, near the fumarolic field of Solfatara crater (e.g., crisis, Figure 1). Solfatara is a tuff cone (180 m above sea level) that consists of hydrothermally 1of17

2 Figure 1. (top left) The earthquake locations and the ground deformation (uplift in centimeters) that occurred during the bradyseism. (bottom) Location map and main geological features of Solfatara crater. BN, Bocca Nuova; BG, Bocca Grande. altered breccias and dune bedded ash to lapilli beds [Rosi and Sbrana, 1987]. The tuff cone was formed between 3.8 and 4.1 ka [Di Vito et al., 1999], and historical chronicles report that a phreatic event occurred in the 12th century [Rosi et al., 1984]. [6] At the present moment, the Solfatara area is affected by intense, diffuse degassing and fumarolic activity. Detailed geochemical study of the fumaroles of Solfatara [Caliro et al., 2007], coupled with the measurements of soil diffuse degassing [Chiodini et al., 2001] and with physical numerical simulations of the hydrothermal system [Todesco et al., 2003], suggests that magma degassing episodes have a relevant role in triggering the volcanic unrest periods that periodically affect the area [Chiodini et al., 2003; Bodnar et al., 2007]. [7] This important, probably primary role of fluid release at depth in generating volcanic unrest periods at Campi Flegrei is not surprising. According to Chiodini et al. [2001], the thermal energy release associated with the gas emission at Solfatara ( 100 MW associated with a total output of 5000 t d 1 of a CO 2 H 2 O gas mixture) is the most important term in the energetic balance of the whole Campi Flegrei caldera. The thermal energy released by fluids at Solfatara is (1) orders of magnitude greater than the elastic energy released during seismic crisis (2) higher than the energy associated with the ground deformation, and (3) about 10 times greater than the conductive heat flux over the 90 km 2 surface of the Campi Flegrei caldera. According to previous geochemical interpretations [Caliro et al., 2007] as well as the results of physical numerical simulations [Todesco et al., 2003], the core of the hydrothermal system is a gas plume, i.e., a vapor liquid biphase zone, fed by magmatic and meteoric sources. The mixing between meteoric and magmatic fluids is thought to occur at the base of the system at T conditions close to the critical point of water (T > 360 C) [Caliro et al., 2007]. [8] Pulses of the magmatic source, relatively rich in CO 2, would have caused increases in the pressure (and temperature) of the hydrothermal plume which is constituted by a compressible single phase gas zone and by a biphase liquid gas zone, and compositional variations of the fumaroles toward CO 2 richer compositions [Chiodini et al., 2003]. These rapid fluid pressure increases are thought to cause ground deformation [Todesco et al., 2004] and possibly to trigger seismicity by significantly reducing the effective normal stress acting on preexisting (or newly formed) slip planes [Miller et al., 2004]. [9] The aim of this work is to investigate the behavior of the Solfatara system over the last few years on the basis of the many data acquired in the frame of the volcanic surveillance of Campi Flegrei. These data include (1) chemical and isotopic compositions of Solfatara fumaroles (Figure 1) from 1983 up to 2008; (2) 12 detailed soil CO 2 surveys performed from 1998 until 2008 over an area including the Solfatara crater and surroundings (survey area in Figure 1); and (3) the data obtained over 1 year (October 2005 to December 2006) of continuous monitoring of soil CO 2 flux in a site located immediately east of Solfatara crater, in correspondence of a seismogenetic fault (FLXOV3 in Figure 1). [10] Because various techniques were used to obtain the data, in order to simplify the reading of the work, our methods and analytical techniques are presented in Appendix A. Appendix A also reports details of the code used to perform specific physical simulations aimed to evaluate the feasibility of the processes hypothesized to have occurred at Solfatara hydrothermal system during the last several years. [11] We will show that important modifications in the hydrothermal system occurred after the crisis of 2000, implying that a new source of fluids has been activated at Solfatara, causing remarkable variations both in the composition of fumarolic effluents and in the pattern of diffuse soil degassing of deeply derived fluids. The new behavior of the system should be carefully evaluated within the frame of the volcanological surveillance of this very dangerous area. 2. Geophysical Signals in [12] This section reviews the last geophysical signals registered at Campi Flegrei (i.e., ground deformation and seismicity), and offers a brief synthesis of the structural setting of the area. Ground deformations at Campi Flegrei are here illustrated on the basis of the ground elevation measured at the bench mark 25 of the Campi Flegrei leveling network [Ricco et al., 2007], where the maximum displacements were observed in each of the various bradyseismic events 2of17

3 1989. These points follow an exponential decay in time given by h ¼ 5:28 þ 0:726 e t=1504 ð1þ Figure 2. Chronograms of ground deformation and seismicity at Solfatara. (a) Elevation of bench mark 25 of the Campi Flegrei leveling network (black dots [Ricco et al., 2007]) and elevation of bench mark 25 after the removal of the effect of secular subsidence (gray dots); (b) ground elevation filtered signal of bench mark 25 (GEFS, see text); and (c) seismicity of the Campi Flegrei (the number of events for the has been divided by 10). The gray vertical bands (bradyseism) indicate the periods of ground uplift. (Figure 2a). The most recent miniuplift periods (few centimeters) occurred in 2000 and in , interrupting the subsidence trend which followed the last large bradyseism of Recently, the ground movement at Campi Flegrei has been interpreted as the results of two processes [Chiodini, 2009]: (1) on the one hand, it derives from volcanic sources, i.e., from a change in the pressure of the hydrothermal system due to the input of magmatic gases (and/or from magma movement at depth), and (2) on the other hand, the Campi Flegrei caldera is affected by a secular subsidence deriving from the compaction of the thick layer of recent volcanic pyroclastic products (secular subsidence) [Dvorak and Mastrolorenzo, 1991]. The removal of the effects of the secular subsidence by adding m yr 1 [Troise et al., 2007] to each elevation measure produced the corrected curve reported in Figure 2a (gray curve [Chiodini, 2009]). The corrected curve shows an exponential decay since 1985 connected to the decline in the time of the effects of the large bradyseismic event that occurred in Chiodini [2009] computed the residuals of the corrected curve with respect to the best fit exponential decay equation of all the data from 1985 up to The resulting obtained signal was considered the most representative of the minor bradyseismic events that occurred after 1985 [Chiodini, 2009]. Here we apply a similar approach but, in the regression analysis, we consider only the measurement obtained before the first miniuplift of where h is the elevation in m, t is the time in days after 1 January 1985 (beginning of the deflation), and the value 1504 (days) is the time constant ( 4.2 years) of the decay. Interestingly, the value of the time constant is practically the same as that computed by Chiodini [2009] considering all the measurements (1512 days). Equation (1) gives an estimation of the elevation change of the bench mark 25 in absence of both secular subsidence and the minor bradyseisms which occurred after The robustness of the model is demonstrated by the almost perfect correspondence between the data before the first miniuplift of 1989 and those computed by equation (1) (R 2 = ). The deviations of observed data from those computed with the model after 1989 (best fit residuals) represent a good parameter for the geochemical signals to be compared with because these deviations are controlled, in our interpretation, by variations in the volcanic sources during the same period. Figure 2b (chronogram of the ground elevation filtered signal, GEFS) suggests that volcanic sources would have caused a total inflation of 20 cm from 1989 to The curve in Figure 2b shows three periods of rapid uplift in 1994, in 1996, and in 2000 and a general inflation trend from 2003 to 2008 (total uplift of 15 cm) with an acceleration of the process during [13] The uplift events of 1994, 2000, and were accompanied by earthquakes (Figure 2c). In particular the seismicity associated with the event of represents the numerically largest seismic swarms ever recorded since 1985 [Saccorotti et al., 2007]. It is significant that in the 2000 and crisis the seismic activity was characterized by the occurrence of both volcano tectonic events (VT, 70 events in 2000 and 260 events in ) and long period events (LP, 10 events in 2000 and >750 events in 2006), suggesting a source mechanism associated with fluid rock interaction [Saccorotti et al., 2001, 2007; Bianco et al., 2004; Cusano et al. 2008] and bulk fluid migration. [14] The 2000 swarm began with the most energetic and deeper LP event (M d = 2.1 and depth is 4 km ± 0.3 below sea level) located 1 km SE of Solfatara crater [Bianco et al., 2004]. In the following, we show that starting from this moment the Solfatara fumaroles began to change composition and important modifications affected the pattern of soil diffuse degassing. The LP events of 2006 were less energetic and shallower (depths of m) and entirely confined beneath Solfatara crater. These events followed a swarm of VT events which occurred in October VT events of both unrest periods (2000 and ) are mainly located east of Solfatara crater in correspondence with a NW SE and NE SW fault system (Figure 3). NW SE and NE SW faults follow the main regional lineaments affecting the Campi Flegrei caldera [Hippolyte et al., 1994; Orsi et al., 1996; Piochi et al., 2005]. The same NW SE and NE SW directions affect the Solfatara cone as suggested by specific structural surveys [Chiodini et al., 2001]. In particular the crater rims are cut by two NW SE striking faults, which dip toward the northeast 3of17

4 Figure 3. Location of earthquakes at Campi Flegrei in the period Data from Centro di Monitoraggio of INGV, Sezione di Napoli, Osservatorio Vesuviano. Earthquake magnitudes given by circle diameters. (Figure 1). The outer, northeastern flank of the tuff cone is affected by two subparallel NW SE (N135 E+30 ) striking faults, dipping toward the northeast. In the following we will refer to these last faults as the Pisciarelli faults. 3. Results and Discussions 3.1. Chemical Composition of Solfatara Fumaroles and Conceptual Model [15] This section aims to highlight both the correlations between fumarole chemical compositions and geophysical signals, and the compositional changes observed since For most of the geochemical aspects we will refer to a detailed geochemical model of the fluids that has been recently published [Caliro et al., 2007]. [16] We consider here 450 chemical and isotopic compositions observed at Bocca Grande (BG, T 161 C), Bocca Nuova (BN, T 147 C), and Pisciarelli (T 97 C) fumaroles (Figure 1), from 1983 up to 2008 within the framework of the volcanological surveillance of Campi Flegrei (auxiliary material). 1 The main component of the fumaroles is H 2 O followed by CO 2,H 2 S, N 2,H 2,CH 4, He, Ar, and CO. The absence of acidic gases (SO 2, HCl, and HF) was interpreted as an indicator of the storage and reequilibration of the fluids in a hydrothermal environment [Cioni et al., 1984; Caliro et al., 2007; Allard et al., 1991; Chiodini et al., 1996]. The main features of the hydrothermal system are synthesized in the conceptual model of Figure 4, which was derived both from geochemical investigations [Caliro et al., 2007] and from physical simulations [Todesco et al., 2003]. Figure 4 shows that the core of the system is constituted by a plume of vapor liquid biphase zones and single vapor zones fed by a mixture of magmatic and meteoric fluids. The magmatic component enters the base of the system in a high temperature zone 1 Auxiliary materials are available at ftp://ftp.agu.org/apend/jb/ 2008jb Figure 4. Geochemical conceptual model of Solfatara modified after Caliro et al. [2007]. Gas/liquid mass fractions (Xg), and low T isotherms are reported, as returned by the physical numerical simulations described in detail by Chiodini et al. [2003] and Todesco et al. [2003]. Gas fractions indicated by gray shading scale. (T > 360 C) where it mixes with and vaporizes the liquid of meteoric origin H 2 O, CO 2,CH 4 : Geochemical Evidence of Magmatic Fluids During Bradyseisms [17] Water (gas) and CO 2, the main components of Solfatara fumaroles, constitute 99.8% of the total discharge. In the following discussions we will refer only to CO 2, because H 2 O concentration (in molar fraction) is complementary to CO 2 (X H20 1 X CO2 ). [18] The comparison between time series of CO 2 content at Solfatara fumaroles and the occurrence of bradyseisms (Figure 5) previously suggested that magma degassing episodes had an important role in the bradyseismic events [Chiodini et al., 2003]. Energetic magma degassing events Figure 5. Chronogram of CH 4 and CO 2 composition of Solfatara fumaroles. The gray vertical bands mark the bradyseismic crises (see Figure 2). 4of17

5 Figure 6. Chronogram of ground elevation, and CO 2 and CH 4 concentrations of BG fumarole. (a) Ground elevation filtered signal (GEFS) is compared with CO 2 and CH 4 concentrations; and (b) GEFS is compared with the CO 2 /CH 4 ratio. in 1983, 1989, and 1994 would have favored the bradyseismic phenomena and the CO 2 peaks observed at the fumaroles some time later than the maximum ground uplift. Contrary to CO 2, the CH 4 concentration manifests an opposite behavior showing negative peaks systematically with the occurrence of each bradyseismic event. [19] A detail of this period ( ) is shown in Figures 6a where CO 2 and CH 4 concentrations are compared with the signal of ground elevation filtered from secular subsidence and from the effects of the bradyseism (ground elevation filtered signal, GEFS). Figure 6a shows that GEFS correlates positively with CO 2 concentrations while it is anticorrelative with CH 4 concentrations. The opposite behavior of CO 2 and CH 4, and their evident correlations with ground elevation, are consistent with the hypothesized input of magmatic fluids during the bradyseisms. [20] On one hand, according to previous geochemical models (Figure 4), an increment of the CO 2 content in the fumaroles is indicative of an increase in the fraction of the magmatic component. In fact the molar fraction of CO 2 in the magmatic component has been estimated to be much higher than the concentration in the hydrothermal fluids ( 0.38 versus 0.12, respectively [Caliro et al., 2007]). Data on melt inclusions of the Campi Flegrei volcanic products suggested even higher CO 2 molar fractions of the magmatic gases (from 0.5 to 0.8 [Mangiacapra et al., 2008]). [21] On the other hand, CH 4 is a gas species which differentiates hydrothermal gases, where it is present in relatively high concentrations, from magmatic fluids, where it is normally absent or present in very low concentrations [Chiodini, 2009]. According to Caliro et al. [2007], at Solfatara CH 4 is formed in the deeper and hottest zones of the hydrothermal system by CO 2 reduction as suggested by the agreement between isotopic and chemical temperature estimations. The d 13 C CO2 CH4 isotopic geothermometer indicates temperatures from 360 C to 420 C, while the chemical geothermometer CO 2 CH 4 gives temperatures >360 C. These estimations were made assuming that H 2 O fugacity (f H2O ) is fixed by the presence of liquid water, that f CO2 is controlled by the full equilibrium function [Giggenbach, 1988], and that f O2 is fixed by a redox buffer specific for Campanian volcanoes [Chiodini and Marini, 1998; Caliro et al., 2007]. The model uniquely defines the theoretical H 2 O CO 2 CH 4 composition of hydrothermal liquid and vapor phases at various temperatures (see diagram log CO 2 /CH 4 versus log CO 2 /H 2 O, Figure 7). Figure 7, which compares theoretical and measured values, shows that Solfatara fumaroles plot on the whole close to the expected compositions for high temperature hydrothermal fluids (>360 C). However, a mixing trend from hydrothermal compositions toward high CO 2 /CH 4 and CO 2 /H 2 O ratios is also evident. CO 2 /CH 4 ratios higher than 10 6, i.e., compatible with magmatic fluids, are extrapolated at the magmatic CO 2 /H 2 O ratio suggested by Caliro et al. [2007]. [22] Because of the large difference between magmatic and hydrothermal fluids, the CO 2 /CH 4 ratio results as a powerful geochemical indicator of the arrival of magmatic fluids in the shallow levels of volcanoes with extensive hydrothermal systems [Chiodini, 2009]. Excluding few very high values observed during the crisis of 1994 and Figure 7. Diagram of log CO 2 /CH 4 versus log CO 2 /H 2 O. The compositions of Solfatara fumaroles are compared with the theoretical compositions of hydrothermal vapor and liquid phases computed at various temperatures (see text). Solfatara fumaroles plot along a mixing trend between high temperature hydrothermal fluids (T > 360 C) and magmatic fluids (see text). 5of17

6 Figure 8. He Ar N 2 triangular diagram of the samples (note that routine analytical determination of Ar started in 1998). The arrows depict the mixing trends of deep gases with an atmospheric component (air or ASW). The samples follow the mixing trend of the atmospheric component with the deep component estimated on the basis of the argon isotopes (N 2 /Ar = 1750, He/Ar = 8.04, see Figure 9) , the measured CO 2 /CH 4 ratios varied in time practically overlapping the GEFS (Figure 6b). This agreement suggests, in our opinion, the strict control played by the dynamics of the hydrothermal system on ground movements at Campi Flegrei. The input of magmatic gases causes uplift (i.e., a response to fluid pressure increases) and compositional variations in the fumarole. It is worth noting that the process of fluid pressure increase as a consequence of the input of magmatic fluids in the hydrothermal system should be enhanced by the presence of zones of single gas phase and gas liquid biphase (Figure 4). An increasing input of magmatic gases would, in fact, cause a direct increase of the fluid pressure in the part of the system containing a compressible gas phase, potentially triggering ground deformation and seismicity. [23] Finally, Figure 5 shows a change in the behavior of Solfatara fumaroles after From that time the CO 2 content of the fumaroles started to increase continuously instead of showing peaks as observed in the previous years. This change is connected with an important modification of the fumaroles composition H 2 O CO 2 Ar N 2 He: Evidence of the Input of a New Gas After 2000 [24] Main gas species (CO 2 and H 2 O), unreactive gases (He, Ar) and N 2, which can be considered practically unreactive because of the sluggishness of equilibration with NH 3, are conservative species and can be usefully used to investigate the source(s) of the fluids. [25] Argon is a good tracer of the input of atmospherically derived gases because of the large difference between its concentrations in air and that in deeply derived fluids, which usually are characterized by very low Ar content. The argon content of Solfatara fumaroles is very low ( ppm, mean of 0.6 ppm) suggesting limited air, or air saturated water (ASW), contamination. However, the air (or ASW) derivation of part of the Ar is evident in the diagram He Ar N 2 of Figure 8 where the points of BG, BN, and Pisciarelli fumaroles plot along mixing lines between a component that is poor in Ar, and characterized by high N 2 /Ar and He/Ar ratios, and the atmospheric component. A more constrained indication of the source of the Ar is furnished by the 40 Ar/ 36 Ar isotopic ratio which has been determined in 6 samples of (Table 1). The 40 Ar/ 36 Ar ratio ranges from 460 to 620, values always higher than the atmospheric 40 Ar/ 36 Ar ratio of [Ozima and Podosek, 1983], indicating in all the samples the contribution of some Ar of nonatmospheric, deep origin. Therefore, we used suitable ternary plots of H 2 O Ar 36 Ar N 2 Ar 36 Ar, and He Ar 36 Ar (Figure 9) to extrapolate Ar/ H 2 O, N 2 /Ar, and He/Ar in the nonatmospheric component. We assumed that the 36 Ar content of the nonatmospheric component is 0, an assumption justified by the very high values of 40 Ar/ 36 Ar in the upper mantle (up to 40,000 [Burnard et al., 1997]) and by the absence of nonatmospheric crustal source of 36 Ar [Sano et al., 2001]. The N 2 /Ar, He/Ar, Ar/H 2 O ratios for the air (or ASW) uncontaminated deep components are estimated to be 1750, 8.04, and , respectively. It is significant that the estimated Ar/H 2 O for the deep component agrees well with the Ar/H 2 O ratio of ASW ( ), that seems to hold also for magmatic vapors, as suggested by Giggenbach [1987]. Assuming that the Ar concentration of the deep component(s) remained constant in all the observation time, it is possible to estimate the atmospheric contamination (here considered air) of the samples in the range from 0.001% to 0.03%. Such limited contamination will strongly affect the Ar contents, in a minor proportion the N 2 (generally < 10%), while the other gas species will be practically unaffected. The nonatmospheric origin of most of the N 2 is confirmed by the d 15 N isotopic Table 1. Chemical and Isotopic Compositions of the Solfatara Fumaroles a Sample Date T ( C) H 2 O CO 2 H 2 S Ar N 2 CH 4 H 2 He CO d 15 N 40 Ar/ 36 Ar BG 17 Oct , , BG 3 Dec , , BG 5 Feb , , BG 17 Mar , , BN 5 Feb , , BN 17 Mar , , a Concentrations are given in mmol mol 1. The concentration of nitrogen isotopes, d 15 N, are expressed in per mil ( ) units, relative to atmospheric N 2 standard (ATM). 6of17

7 scattered in time, while since July 2000 the three fumaroles have homogenized and show a well defined decreasing trend of the N 2 /He ratio. In particular the points follow an exponential decay in time (from 2000 to 2008, Figure 10b) interpretable as the substitution in the hydrothermal system of fluids originally with a high N 2 /He ratio (composition A) with fluids of low N 2 /He ratio (composition B) passing through a mixture progressively enriched in component B. The best fitting of the data suggests N 2 /He ratios of 370 and 176 for the pure components A and B. The mixing between these two end members is evident also in the diagram N 2 versus He of Figure 11a where the samples collected after 2000 follow a linear trend between the component A, with high N 2 /He ratio, and the component B with low N 2 /He ratio. The oldest samples depart from this mixing trend toward N 2 richer compositions, indicating that the component B was not involved in the recharge of the hydrothermal system before However, this evidence is not definitive because He determinations started in 1989, and a comparison with the fluid involved in the last big bradyseism of is not possible. [27] In Figure 11b the binary plot N 2 versus CO 2 is considered in order to better investigate this matter and Figure 9. Triangular plots of (a) Ar H 2 O 36 Ar, (b) N 2 Ar 36 Ar, and (c) He Ar 36 Ar of Solfatara fumaroles. data ranging from 6.0 to 6.7 (Table 1), values which characterize volcanic fluids in subduction zones (d 15 N=7± 4 [Sano et al., 2001]) and which are far from the atmospheric value (d 15 N atm =0 ). [26] As a final observation on the He Ar N 2 plot, Figure 8 shows a significant decrease of the N 2 /He ratio of the post 2000 samples with respect to the samples collected in the period. The variation of the N 2 /He ratio over time is shown in the chronogram of Figure 10a. From 1988 to June 2000 the N 2 /He ratios were relatively high and Figure 10. (a) Chronogram of N 2 /He ratio of Solfatara fumaroles for the entire data set and (b) for the samples of BG and BN fumaroles collected after July The exponential decay best fit equation is reported (see text). 7of17

8 Figure 11. (a) N 2 versus He and (b) N 2 versus CO 2 of Solfatara fumaroles. The arrows depict the mixing trend between hydrothermal and magmatic components. characterize the different components. The oldest samples ( , i.e., before the effects of the bradyseism became evident in the fumarolic composition) were characterized by relatively low CO 2 and N 2 contents. In 1985, immediately after the bradyseism, the fluids quickly moved to a CO 2 and N 2 rich composition at the extreme of the trend in Figure 11b ( 1985 magmatic fluids ). During the subsequent crises (1989 and 1994) the composition showed similar enrichment in CO 2 and N 2 and moved along the trend. In 2000 the samples, representative of component A, were characterized by the lowest CO 2 concentrations, i.e., they were close to the hydrothermal composition [Caliro et al., 2007]. The most recent samples progressively moved toward a CO 2 richer composition along the mixing trend which is very different from that of This behavior indicates the arrival in 2000 of a new magmatic component, i.e., component B. [28] Following the method described by Caliro et al. [2007], it was possible to estimate the CO 2 content of these magmatic fluids on the basis of the oxygen isotopic composition of the H 2 O CO 2 gas system (Figure 12). The d 18 Oof the whole fumarolic H 2 O CO 2 gas system (d 18 O H2O+CO2 ) is computed on the basis of the measured d 18 O values in the steam condensates and of the computed d 18 O values of the CO 2, assuming oxygen isotopic equilibrium between H 2 O and CO 2 at the temperature of the fumaroles [Chiodini et al., 2000]. These values, plotted in Figure 12 (d 18 O H2O+CO2 versus c CO2 diagram, where c CO2 is the CO 2 oxygen atomic fraction in the system H 2 O CO 2 ), define a mixing line between the compositions observed in 2000 and CO 2 rich magmatic compositions (theoretical composition of magmatic fluid curve in Figure 12). The theoretical oxygen isotopic composition of the magmatic fluids is derived for fluids at different CO 2 molar fractions considering the isotopic equilibrium with Campi Flegrei magmatic melt [Caliro et al., 2007]. The intersection between the mixing line and the theoretical composition of magmatic fluid curve gives a c CO2 of 0.51 (X CO2 of 0.34) expected for the magmatic fluids entering the Solfatara system after [29] Original magmatic He and N 2 contents of 4 ppm and 640 ppm, respectively, are estimated by extrapolating the mixing trends of Figure 11 to the magmatic CO 2 value (i.e., 340,000 ppm). It is worth noting that the estimated N 2 /He ratio (160) is close to the value independently Figure 12. Diagram of oxygen isotopic composition of the (H 2 O+CO 2 ) gas system, d 18 O (H2O CO2), versus the relative oxygen atom fraction of CO 2 (c CO2 ). In the upper axis the corresponding CO 2 molar fraction (X CO2 ) is reported. The isotopic variation on c CO2 of the computed magmatic fluids is also reported as the theoretical composition of magmatic fluids line [after Caliro et al., 2007]. Solfatara fumaroles show a trend representative of mixing between a hydrothermal component and a magmatic component characterized by a c CO2 of 0.51 (X CO2 of 0.38). 8of17

9 Table 2. Description of the CO 2 Flux Surveys, of the Statistical Parameters of CO 2 Flux Partitioned Populations, and of the Size of the Solfatara DDS Surface From 1998 to 2008 a Survey Period Number of Measures CO 2 Flux Population Fraction (%) Mean CO 2 Flux Mean CO 2 95% Confidence Interval Minimum Maximum Average LF HF LF HF LF HF DDS Surface b (10 5 m 2 ) Dec , Jul , Feb , nd 5.46 Jul , Apr , Aug , Mar , Oct , May , Oct , Mar , Jun , a CO 2 fluxes and 95% confidence interval are reported in g m 2 d 1 ; nd, not determinable. b Area where the probability that estimated soil CO 2 flux is larger than 50 g m 2 d 1 is above 0.5 (see text). estimated on the basis of the exponential best fit decay (176, Figure 10b) Soil CO 2 Diffuse Degassing [30] In this section, we consider 4856 CO 2 flux measurements obtained by means of the accumulation chamber technique, during 12 different surveys carried out in the period 1998 to A synthesis of the results of the 12 surveys, which roughly covered the same area of 1 km 2 (Figure 1) with 400 measurements for each survey, is reported in Table 2. In the March 2007 survey, 360 isotopic composition of the CO 2 efflux (d 13 C CO2 ) were measured together with soil CO 2 flux. Details of the used methods are reported by Chiodini et al. [2008] and in Appendix A. [31] Moreover, time series of CO 2 flux performed by an automatic station (FLXOV3) at Pisciarelli site (Figure 1) are discussed in relation to the seismic activity Changes in the Pattern of Soil Diffuse Degassing and CO 2 Flux Anomalies [32] The results of the twelve surveys are reported in the probability plots of log CO 2 flux (Figure 13). The soil CO 2 fluxes distribute in a wide range of values (Table 2) and the data of each survey plot along quasi bimodal curves which were modeled as the combination of two lognormal CO 2 flux populations [Sinclair, 1974]: the HF population includes the highest CO 2 flux values and the LF population includes the lowest CO 2 flux values. The statistical parameters of the partitioned HF and LF populations, estimated by the Sichel t estimator [David, 1977], are reported in Table 2. [33] The elevated mean CO 2 flux values of the HF populations (Table 2) clearly indicate that these populations are representative of CO 2 fluxes fed by an endogenous source, i.e., by the Solfatara hydrothermal system [e.g., Chiodini et al., 1998, 2001, 2008; Cardellini et al., 2003]. On the contrary, the interpretation of the LF population is less obvious. In the period , the mean CO 2 flux values of LF populations were relatively low (14 24 g m 2 d 1 ) while, after 2003, the CO 2 flux mean values of LF populations underwent a sudden increase of up to about 88 g m 2 d 1 in October The relatively low mean LF CO 2 fluxes before 2003 are in agreement with a pure biogenic source of the CO 2. In fact the mean biogenic CO 2 flux from soils in various ecosystems normally ranges from 0.2 to 21 g m 2 d 1 reaching higher values (up to g m 2 d 1 ) only in particular environments, such as agricultural fields or grasslands [e.g., Raich and Schlesinger, 1992; Raich and Tufekcioglu, 2000; Reth et al., 2005; Yazaki et al. 2004; Cardellini et al., 2007, and references therein]. [34] The relatively high soil CO 2 flux values of the LF populations observed after 2003 could depend either on some environmental changes that caused an increase in the biogenic soil CO 2 production, or on a generalized increment of soil CO 2 fluxes from the hydrothermal source which should have affected large areas of Solfatara and its surroundings. In order to answer this question, in March 2007 a detailed survey of both soil CO 2 fluxes and carbon isotopic composition of the CO 2 efflux was performed at Solfatara [Chiodini et al., 2008]. The combined interpretation of flux and CO 2 efflux isotopic composition data allowed us to identify and characterize the hydrothermal and the biogenic sources. The soil CO 2 from the hydrothermal source was characterized by a mean d 13 C CO2 of 2.3 ± 0.9, close to the isotopic composition of the fumarolic CO 2 Figure 13. Probability plots of log CO 2 fluxes of the 12 surveys from 1998 to of17

10 Figure 14. Probability maps of CO 2 flux from 1998 to The maps report the probability that at any location the soil CO 2 flux is higher than 50 g m 2 d 1 which was selected as maximum threshold for the occurrence of the solely biogenic CO 2 flux (see text). Probability higher than 0.5 (yellow) defines the Solfatara diffuse degassing structure (DDS). The dashed box indicates the area where in 2003 the most evident increase in soil CO 2 flux occurred (CB box, see Figure 18a). (d 13 C CO2 = 1.48 ± 0.22 ) and by a mean CO 2 flux of 2875 g m 2 d 1. The CO 2 from the biogenic source was characterized by a d 13 C CO2 of 19.4 ± 2.1, and by a mean CO 2 flux of 26 g m 2 d 1, which are both in the range of typical values for biogenic CO 2 soil degassing [Chiodini et al., 2008]. Therefore, the isotopically determined mean CO 2 flux from the biogenic source was significantly lower than the mean value of the LF population (47 g m 2 d 1 ) indicating that a consistent part of the gas was derived from a low level anomaly of the deep source. [35] Assuming that the mean CO 2 flux from biogenic background sources at Solfatara was always in the range of typical values (<20 30 g m 2 d 1 ), as demonstrated for the March 2007 survey, we infer that after 2003 LF populations represent a mixture of biogenic and deeply derived CO 2. This implies that in 2003, large areas, previously not affected by an anomalous CO 2 degassing, started to release deeply derived CO 2. In other words, the first arrival of deeply derived CO 2 caused new CO 2 flux anomalies where both the hydrothermal and the biogenic CO 2 fluxes are in the same order of magnitude. [36] In order to highlight these areas we produced for each data set a map of the probability that at any location the soil CO 2 flux is higher than a reasonable maximum flux value for the biogenic CO 2 flux (Figure 14). This cutoff value was selected at 50 g m 2 d 1 which is the 95th percentile of the biogenic soil CO 2 fluxes derived on the basis of the isotopic compositions of the CO 2 efflux of the March 2007 survey [Chiodini et al., 2008]. Yellow and red in Figure 14, i.e., probability >0.5 that CO 2 flux is above the cutoff value, define the area which is degassing deeply derived CO 2 (diffuse degassing structure, DDS). The maps show a marked increase in the DDS extension from February 2003 to July 2003 (Figure 14). The enlargement of Solfatara DDS after 2003 is confirmed by the isotopic data of March The map of the distribution of the d 13 C CO2 of the CO 2 efflux (Figure 15), elaborated from the data of the March 2007 survey [Chiodini et al., 2008], highlights that the CO 2 efflux over most of the surveyed area is affected by CO 2 contribution from the hydrothermal source (e.g., d 13 C CO2 values > 10, yellow and red in Figure 15). The DDS enlargement, qualitatively pointed out by Figures 14 and 15, corresponds to an expansion from m 2 of February 2003 to m 2 of June This expansion of the DDS has been a sustained modification because the DSS extension remained in the range to m 2 from April 2004 until the last survey of June 2008 (Table 2). [37] Figures 14 and 15 show that the soil degassing of CO 2 from the hydrothermal source occurs both inside the Solfatara crater and along a narrow band about 1 km long and 0.2 km wide located to the east of the crater, which matches well the NW SE striking Pisciarelli faults. In the southern tip of the Pisciarelli faults (dashed white box in Figures 14 and 15, check box (CB)) the most evident 10 of 17

11 Figure 15. Map of distribution of the isotopic composition of the CO 2 efflux in March The dashed box indicates the CB box (see Figure 14). Deeply derived CO 2 is released through the soil mainly within the Solfatara crater and along a narrow band about 1 km long and 0.2 km wide located to the east of the crater, which well matches the Pisciarelli faults. increase of soil CO 2 degassing after 2003 was observed. Considering the CO 2 flux measurements obtained in the CB box in the 12 surveys, the average of the CO 2 flux passed from typical background values in the pre 2003 period (10 15 g m 2 d 1 ) to higher values of 1 order of magnitude ( g m 2 d 1 ) in the period [38] The occurrence of active fracturing/faulting processes along the Pisciarelli faults is highlighted by the epicentral locations of earthquakes that occurred in the area over the last 8 years (Figure 3). It is worth noting that the highest spatial probability for epicentral location, stacked over the whole locations of earthquakes of 2000 [Bianco et al., 2004], practically coincides with the CB box of Figures 14 and 15. [39] During the period other significant anomalies were observed in the main fumaroles located near the Pisciarelli faults (site Pisciarelli in Figure 1). In April 2006 the fumaroles had an anomalous activity consisting in the expulsion of dark mud; the flow rate of the fumaroles visibly increased during that period; and the automatic soil CO 2 flux station sited near the fumaroles (FLXOV3 station, Appendix A) registered evident positive peaks (Figure 16a). In particular, the CO 2 flux series is characterized by a constant value (about 10,000 g m 2 d 1 ) from October 2005 to September 2006, and by positive peaks in September 2006, when the flux increased by a factor of 1.5, and in November 2006, when the flux increased by a factor of 2. The occurrence of this last anomaly is confirmed by a similar and simultaneous temperature anomaly registered by an automatic IR station in the soil surrounding the main fumarolic emissions of Solfatara crater (BG area, Figure 16b [Chiodini et al., 2007]). The anomaly of November 2006 followed of few days 150 VT earthquakes, which mostly affected the area of the Pisciarelli faults, and the October 2006 swarm of LP events (Figure 16) Conceptual Model of the Processes and Physical Simulations [40] Different processes might have caused the relatively short anomaly registered by the continuous monitoring and the modification in the CO 2 flux pattern observed since Our hypotheses are that (1) the evident peak in the FLXOV3 time series was most probably caused by permeability changes induced by the 2006 earthquakes in the Pisciarelli faults; and (2) the enlargement toward to east of the soil CO 2 degassing pattern observed in 2003 reflects the arrival at the surface of deep, CO 2 rich fluids. In particular we suggest that the opening of deep fractures occurred in 2000 could have favored the ascent from depth of CO 2 rich, magmatic fluids along the Pisciarelli faults. In order to asses the physical feasibility of such conceptual models, a physical modeling approach, based on the TOUGH2 geothermal simulator, has been attempted (Appendix A). [41] Previous applications of TOUGH2 at Solfatara were aimed at simulating the hydrothermal system as a whole [Todesco et al., 2003; Chiodini et al., 2003]. On the other hand, in the application illustrated here, we focus on the processes occurring along the Pisciarelli fault; that is, we do not consider the entire system but just the eastern sector where the fault is located. The aim is to explain how a contrast of permeability and/or a variation in the characteristics of the deep system is able to vary CO 2 flux in the presence of a preexisting fault. We considered two types of rocks: one with properties equal to the previous applications Figure 16. (a) Chronogram of CO 2 fluxes measured by the automatic station FLXOV3. The highly scattered behavior of the row data (gray line, 2 h series) is due to a daily effect of the meteorological parameters on the CO 2 flux (see Appendix A). Considering the daily mean of the CO 2 values (black line) this effect disappears. (b) Infrared signal of BG area (temperature of BG area line, redrawn from Chiodini et al. [2007]). Inspection of these time series makes it evident that the peak in both BG area temperature and the CO 2 fluxes in November 2006 is associated with intense seismic activity. 11 of 17

12 Figure 17. TOUGH2 simulation of a fault controlled hydrothermal system (see text). (a) Parameters of the computational domain and rock properties; (b) simulated CO 2 flux at the surface at steady conditions; and (c) temperature and gas/liquid volume fraction (SG) in the simulated domain at steady state conditions. [Todesco et al., 2003; Chiodini et al., 2003] and the second with a higher permeability (from to m 2 ), ideally representing the damage zone of the Pisciarelli fault. The computational domain is a 2 D domain, not axisymmetric, 4980 m wide and 1980 m deep which has been discretized in 11,122 square elements of m (Figure 17a). The lateral and the bottom boundaries of the model are impermeable and adiabatic while the top boundary, fixed at atmospheric temperature and pressure conditions, is open to heat and fluid flow. This plane represents a SW NE cross section orthogonal to the Pisciarelli fault. As the initial conditions, we assigned to each element a total fluid pressure equal to hydrostatic, a temperature computed according to a thermal gradient of 0.13 C m 1 [Todesco et al., 2003, and reference therein], and a CO 2 partial pressure fixed at any temperature by hydrothermal reactions (i.e., full equilibrium equation [Giggenbach, 1988]). [42] Steady state conditions were reached after 1000 years of injection at the bottom boundary, in correspondence to the fault, of kg s 1 of a vapor phase (molar fraction of CO 2 = 0.14) at 350 C. Figure 17b shows the temperature and the gas/liquid volume fraction (SG) which are returned at steady state conditions. An asymmetric plume, whose shape is controlled by the presence of the fault, and a nonsymmetric pattern of CO 2 flux at the surface, is simulated by the physical model. CO 2 fluxes are high on the hanging wall of the fault while they are low on the footwall of the fault. The simulated pattern of CO 2 flux agrees in some ways with the one measured. In fact, the observed CO 2 anomaly extends to the east, on the hanging wall of the fault, while it disappears to the west, i.e., over the footwall of the fault (Figures 14 and 15). This similarity reinforces the idea that CO 2 fluxes in this zone are fed by hydrothermal fluids migrating along the Pisciarelli fault. [43] The steady state condition (Figure 17) was varied in order to investigate the time evolution of CO 2 flux at the surface in relation to different phenomena possibly changing the feeding system. In particular we investigated the modification induced in the CO 2 fluxes caused by (1) the input of a certain amount of CO 2 rich fluids in the deeper part of the fault and (2) a change in permeability along the fault. [44] In the first exercise we increased permanently from 0.22 kg s 1 to 0.44 kg s 1 the amount of fluids injected at the base of the fault, maintaining the original boundary conditions, source location, rock properties and fluid temperature and composition. The model predicts that the increase of CO 2 flux at the surface, in correspondence with the fault, occurs 2 3 years after the modification. The comparison of the simulated signal with the CO 2 flux anomaly observed in 2003 in the CB box (Figure 18a) highlights the similitude between the timing and the shape of observed and simulated variations. This similarity, which is suggestive, but not definitive, testifies, however, to the physical feasibility of the process and generally supports the idea that the injection of CO 2 rich fluid in the deeper part of the Pisciarelli fault during the seismic swarm occurred in summer 2000 caused the change in the CO 2 flux pattern observed since [45] In the second application, we investigate whether the relatively short period CO 2 anomaly registered by FLXOV3 station could have been caused by a permeability increase along the NW SE fault, possibly as a consequence of the earthquake swarm of October The results highlight that soil CO 2 fluxes respond immediately to permeability variations, i.e., an increase of permeability causes a simultaneous CO 2 flux increase and, to the contrary, a decrease in permeability causes a simultaneous CO 2 flux decrease. Considering this direct link between permeability and CO 2 fluxes, a very good fitting of the shape of the anomaly is obtained by a sudden increase (by a factor of 2) of the permeability of the damage zone (from m 2 to m 2 ) and the returning to the initial permeability conditions in the following 40 days as shown by the chronogram of Figure 18b, where simulated and observed signals (i.e., CO 2 fluxes) are normalized for the comparison. We stress here that we were interested in simulating the shapes and the arrival times of the different signals rather then their absolute value, a parameter which can be difficult to evaluate and which is strongly affected by local heterogeneities in real 12 of 17