Imaging the Salinelle Mud Volcanoes (Sicily, Italy) using integrated geophysical and geochemical surveys

Geochemical and geophysical prospecting methods (including measurements of soil heat flux and soil CO2 flux, gravimetry, self-potential and geomagnetism) are used to produce an integrated data set aimed at imaging the migration of fluids in the sub-surface at the Salinelle mud volcanoes, located on the lower southwestern flank of Mt Etna (Sicily, Italy). This area was affected by magmatic eruptions from local volcanic centers between about 48 and 27 ka. Today, only pseudo-volcanic phenomena due to over-pressured multiphase pore fluids there occur. Carbon dioxide of magmatic origin, mixed with biogenic hydrocarbons, warm hypersaline waters and mud, are constantly released at the surface through the main conduits of mud volcanoes, whose activity is characterized by alternation of mild gas bubbling periods and strong paroxysmal phases. The latter produce violent gas eruptions that eject warm water (T ≈ 50° C) to a height up to about 1 m. Surface distribution of the geophysical and geochemical parameters have been investigated to detect the main pathways through which fluids move toward the shallow crust. Integration of geochemical, geophysical and geological maps allowed for the tracing of the fluid flow in the shallowest (a few tens of meters below the surface) part of the local hydrothermal system. Our results showed that the rising of fluids from a deep reservoir is controlled by the main structural and geological features of the area and their temporal and spatial evolution depends on pressure conditions inside the hydrothermal system.


Introduction
Mud volcanoes are geological structures formed by pseudo-volcanic phenomena caused by over-pressured multiphase pore fluids, generally high-salinity water and methane gas, trapped in sedimentary basins by an impermeable top layer of rock. They are commonly associated with compressive tectonics coupled with sediment accretion at convergent margins [Kopf, 2002], where large amounts of organic material were buried at high sedimentation rates by relatively young sedimentary rocks, thus forming hydrocarbon reservoirs. Over-pressured multiphase pore fluids escape along either lithologic or structural discontinuities and through permeable rocks until eventually erupt muddy liquid and gas at the surface. Mud volcano activity is generally characterized by alternating low-emission periods and strong paroxysmal phases. This activity produces typical morphological structures that vary both in their shape (ranging from conic edifices to sub-circular crater-like depressions), and in their size [Carveni et al., 2012], these features being ephemeral and easily modifiable by following emissions. Generally, mud volcano fields are covered with clays, they show no vegetation, and have diffuse salt precipitates on their surface, deposited as incrustations from evaporation of the emitted water [Etiope et al., 2002]. For this reason, in certain areas of Italy they are named Salinelle or Salse [Carveni et al., 2012], from the Italian words that mean "salt". In some cases, when the conduits of mud volcanoes get clogged by solidified material, methane accumulates pressure until exploding violently, with emission of hot, acid and/or poisonous gases. This poses a high potential risk for the people who live nearby. The most recent tragic episode of gas explosions at mud volcanoes occurred in 2014 at the Macalube of Aragona, the largest mud volcano in Sicily (with a total surface of about 1.4 km 2 ) located near Agrigento (southern Sicily). A powerful explosion of mud caused by methane over-pressure killed two kids, who were buried under a thick (about 20 m) cover of mud. In Sicily, mud volcanoes are quite widespread both onshore and offshore [Etiope et al., 2002;Cangemi and Madonia, 2014]. The main vents are located in the central-southern part of the island, with the exception of three groups of mud volcanoes located on the southwest flank of Mt Etna volcano, the main of which is named "Salinelle di Paternò" (Figure 1). These lay at the contact between the sedimentary rocks of the eastern margin of the Sicilian foredeep and the volcanic rocks of Mt. Etna's edifice, where an actively growing anticline is rooted at shallow depth and related to the gravitational loading and spreading of the southern flank of the volcano [Bonforte et al., 2011]. In this area, the released multiphase fluids consist of a muddy liquid and a separate gas phase. The former derives from interaction between warm hyper-saline water (originally a geothermal brine) and local clays that form the rock cover overlying the hydrocarbon reservoir [Chiodini et al., 1996;Parello et al., 2001].
The latter is a mixture, in variable proportions, of hydrocarbons (mostly methane) that accumulate in shallow pockets/reservoirs and magmatic gases (mostly carbon dioxide) that are released from the deepest levels of Mt.
Etna's feeder system [Chiodini et al., 1996;Caracausi et al., 2003]. Emitted fluids have normally a temperature slightly higher than that of air, but at times, during stronger eruptions, the temperature can rise up to almost 50° C . Geothermometric estimates, made using both the gas and the water chemistry, gave equilibrium temperature between 100 and 150 °C [Chiodini et al., 1996]. In this area the temporal variations in the flux of magmatic gases as well as in the temperature of water and mud would be caused by changes in the gas/magma pressure at depth beneath Mt. Etna volcano, which often precede long-standing periods of volcanic activity at Mt. Etna [Pecoraino and Giammanco, 2005;Giammanco and Bonfanti, 2009;Paonita, 2010]. The latest intense mud eruption occurred from January to June 2016. It was particularly strong, as new mud vents opened inside the courtyard of a private house located on the southern edge of the main Salinelle area and impressive mud flows invaded the surrounding streets for several hundreds of meters around. In that occasion, the regional civil protection worked to divert the mud flows into the Salinelle area in order to avoid the mud invasion in urbanized zones of the town of Paternò. Although this area has already been studied either from geochemical or from a geophysical point of view [Chiodini et al., 1986;Pecoraino and Giammanco, 2005;Giammanco et al., 2007;Panzera et al., 2016], until now no combined study was performed. Each geophysical or geochemical method has clearly different accuracy and reliability for defining different structural properties in diverse conditions and the interpretation of the phenomena includes an inherent degree of ambiguity due to the limit of each discipline. However, the ambiguity can be reduced when data are analyzed jointly. Multidisciplinary approaches of combined gas studies and geophysical surveys are well suited in the research of diffuse degassing structures [Chiodini et al., 2001;Nickschick, et al., 2015] like those of the Salinelle di Paternò, as they help defining both the mechanism of gas/fluids transport from depth to the surface, both at regional and at local scale, and the geometry of the pathways used by fluids to move through the shallow crust. Geophysical methods have greatly contributed to a better understanding of the internal structure of volcanic systems [Napoli et al 2007;Schiavone and Loddo, 2007;Blaikie et al., 2014;Maucourant et al., 2014;Napoli and Currenti, 2016] and therefore may help in the detection and investigation of the main structures (dykes, shallow fractures, cavities) that at different scales drive deep fluids in their rise towards the surface [Mauri et al., 2012]. On the other hand, both discrete and continuous geochemical measurements can provide important perspective on the dynamic state of the fluid circulation in a hydrothermal system and are well suited to monitor its spatial/temporal evolution [Pecoraino and Giammanco, 2005;Maucourant et al., 2014;Inguaggiato et al., 2018].

Rosalba Napoli et al.
4 The studied area is characterized by localized density and magnetization contrasts due to the presence of sedimentary formations in contact with the Mt. Etna volcanics, and by fluids circulation through the shallow crust. Considering this context, we have chosen, among the geophysical methods, gravimetry and magnetic prospectings, since they are appropriate to characterize the main structural features, and the self-potential (SP) approach that is particularly suitable to monitor subsurface fluid movement [Hashimoto and Tanaka, 1995]. Therefore, geochemical, gravimetry, self-potential (SP) and magnetic surveys were simultaneously carried out in 2015 for the first time at the Salinelle di Paternò. Detailed soil CO 2 flux and heat flux maps were integrated with gravimetry, SP, magnetic and geological maps, revealing the main structural features of the area and providing an overall picture of the fluid flow processes in the shallow part of the local hydrothermal system. The results encourage the design of multidisciplinary geophysical networks with high spatial/temporal resolution, in order to provide the opportunity to obtain much better comprehensive understanding of the phenomena under study and to follow their evolution with the necessary frequency of data acquisition.

Geological and morphological settings
The Salinelle mud volcanoes lay on the lower southwest flank of Mt. Etna (Figure 1), about 30 km away from the volcano summit craters and within the urbanized areas of Paternò village. The area rests on volcanic products of old Etnean eruptive centers (350-15 ka), in proximity of the Comiso-Messina regional fault system, that is one of the main Sicilian structural features that plays a major role in driving the Etnean magmas from the deepest reservoirs to the shallowest levels of crust [Etiope et al., 2002]. In this area, the Apennines front composed of Pleistocene sedimentary formations is in contact with the Mt. Etna volcanics [Giammanco et al., 2007]. Here, structural traps in the shallow sedimentary rocks allow for the formation of pockets of pressurized natural gas [Caracausi et al., 2003]. The gas phase is mostly composed of deep CO 2 of magmatic origin, with a minor contribution of shallow hydrocarbon gases, chiefly methane [Chiodini et al., 1996;Aiuppa et al., 2004;Pecoraino and Giammanco, 2005]. The aquifer that supplies water to the mud volcanoes is part of the larger hydrogeological system of the southern flank of Mt. Etna [Aiuppa et al., 2004;Ferrara and Pappalardo 2008], where the shallow impermeable sedimentary basement prevents groundwater from reaching a considerable depth. Thinning of the impermeable Pleistocene sediments underlying Etna's volcanic rocks, as consequence of the intense local tectonics, induces upward motion of gases, with consequent mixing with mud and warm hypersaline water until reaching the surface [Aiuppa et al., 2004]. Surface emission of gas takes place from a relatively small area, located on the northern slope The emitted waters show fairly uniform and constant chemical compositions, with sodium and chlorides dominant over the other dissolved ions [Chiodini et al., 1996;D'Alessandro et al., 1996;Aiuppa et al., 2004]. They generally have an electrical conductivity of about 88 mS/cm, thus higher than that of sea-water, and their pH is about 6.0.
Water temperature at outlet generally ranges between 10 and 20° C, but during paroxysmal phase, it may increase up to about 50° C. Periods with higher water temperature values generally precede and/or accompany Etna eruptive activity [Giammanco et al., 1995;Giammanco and Bonfanti, 2009;Paonita, 2010]. During those anomalous periods, both the magmatic gas efflux [Giammanco et al., 2007] and the CO 2 /CH 4 ratio in the emitted gas phase [Giammanco et al., 1998;Greco et al., 2016] increase as well. Geothermometric estimates indicate that the temperature of fluids, at a depth of about 1000 m, is in the range 100-150° C [Chiodini et al., 1996].

Soil heat flux
In order to assess both the spatial distribution and the magnitude of the heat flux in the Salinelle area, we carried out a survey for the measurement of shallow thermal conductivity and thermal gradient in soil in the same sites of the measurements of soil CO 2 effluxes ( Figure 1). The thermal gradient (T z ) that is the soil temperature gradient along the vertical direction z (ΔT/Δz), was calculated measuring the temperature difference between air and soil at each site and dividing it by the depth where soil temperature measurements were carried out (in our case, 6 cm). Soil thermal conductivity was measured using the non-steady-state method [Bristow et al., 1994;Bruijn et al., 1983;van Haneghem et al., 1983;van Loon et al., 1989]. For the scopes of our investigation, we used a probe (Thermal Properties Analyzer, mod. KD2, Decagon Devices, Inc., USA) consisting of a hand-held readout device and a 6-cm-long needleshaped sensor. Each measurement cycle lasts 90 sec, at the end of which a controller computes the thermal conductivity of soil based on the data acquisition during the heating and cooling periods of the probe. Based on the above input, the one-dimensional heat flux (in W m -2 ) at each site was computed following the basic Fourier's law

Soil CO 2 efflux
Gas emissions in the Salinelle area are mostly focused at the many degassing vents, but a significant part of total degassing occurs also in diffuse form through soil principally in the areas surrounding the main vents, though with a lesser magnitude. Soil CO 2 effluxes were measured in 40 sites distributed along two NNE-SSW parallel profiles over an area of ~ 0.12 km 2 (Figure 1). The proximity of the survey to inhabited or cultivated areas prevented the extension of the survey to the south-east and north-west areas, so we could not have a regularly spaced grid of measurement points. Anyway, the profiles were chosen so as to intersect the area affected by the main emission vents, whose location can be considered fairly stable over time [Federico et al., 2019]. The method used was that of the accumulation chamber [Parkinson, 1981;Tonani and Miele, 1991;Chiodini et al., 1998]. Details on the instrumental setup used in the studied area can be found in Greco et al. [2016].
Measured CO 2 effluxes ranged from 0.3 to 299.3 g m -2 d -1 (Table 1), with average value of 37.5 g m -2 d -1 and standard deviation of 55.9 g m -2 d -1 (Figure 3b). Due to the nature of this parameter, whose spatial dispersion follows a lognormal distribution [Ahrens, 1954], all CO 2 data were transformed into their corresponding log 10 values before being processed and mapped. The distribution map of the CO 2 efflux values shows some similarities to that of soil heat fluxes, because the highest degassing values were found near the main eruptive vents in the central part of the study area and in some points of its SW portion. However, some other points located just south of the main vents and at the NE corner of the area showed high CO 2 effluxes but, conversely, very low heat fluxes.
Rosalba Napoli et al.

Geophysical surveys 4.1 Gravity
Gravity data were simultaneously gathered with the measurements of soil CO 2 effluxes and heat flux in the same sites (Figure 1). The instrument used was a portable Scintrex CG5 gravimeter. We also established a reference station in the investigated area, linked with the absolute gravity station in Catania, located about 20 km away, where the FG5#238 absolute gravimeter is routinely used [Pistorio et al., 2011;Greco et al., 2012]. In order to achieve a reliable daily instrumental drift, we performed measurements at the reference station every 2 hours, so that we could get at least four readings during a working day. Measurements started from the reference station and at least three readings were taken and averaged at each measurement point of the surveyed area, obtaining an accuracy of 10 μGal. Spatial coordinates of each gravity points were determined using traditional terrestrial measurements and data analysis. The precision achieved for both horizontal and vertical coordinates is in the order of few centimeters. Consequently, due to the uncertainty in the elevation data, the maximum error in gravity determinations is lower than 0.5 mGal.
Gravity readings were corrected for instrumental drift and adjusted for tidal effects using the Eterna 3.4 software [Wenzel, 1996]. Subsequently, the corrected data were processed to eliminate the latitude effect using the International Gravity Formula [GRS, 1980]: gt = 978032.677 (1+0.005302224 sin 2 φ -0.000005824 sin 2 2φ) where φ is the latitude and gt is in mGal.
In order to obtain both the free air (fa) and the Bouguer (B) anomalies we used the standard formulas: where h is the station's elevation in meters and ρ is the density of the layer above the reference datum, in kg/m 3 . Using the method of Nettleton [1976], we found an average medium density of 2300 kg/m 3 for the Bouguer density reduction.

7
Imaging the Salinelle Mud Volcanoes  Data were also corrected for the local topography using a digital elevation model with horizontal and vertical resolution of 30 and 5 m, respectively. The correction was carried out following the procedure developed by Hammer [1939].
The Bouguer gravity anomaly values (Figure 4) in this small area vary within 2.0 mGal (from 48 to 49.5 mGal). The lower gravity values are observed at the Conetto dei Cappuccini, where a well-defined, though small, anomaly is evident, while higher gravity values characterize the largest gas exhalation zone and the water/mud vents.
No other significant anomalies were observed.

Magnetic survey
The magnetic survey was performed by a GSM19 Overhauser Effect magnetometer with a resolution of 0.01 nT, whose sensor was set vertically on an aluminum pole, 2 m above the ground surface to reduce noise. More than 2,300 measurements were gathered and georeferenced, by GPS data simultaneously collected, on a surface of about 0.45 km 2 (Figure 1). To remove the time variations of external origin we used magnetic data continuously recorded by a reference station temporarily installed near the investigated region, in an area of low magnetic gradient. It is worth noting that the magnetic survey was executed during quiet days (K index values were less than 2). This allowed for a sufficient removal of transient variations from external sources, so that a suitable accuracy could be achieved.
The observed magnetic field was not reduced with respect to the IGRF reference field, because of the limited extent of the investigated area. In fact, in a relatively small survey area the removal of IGRF is not significant because of its low resolution and spatial uniformity [Kearey and Brooks, 1991]. The total-intensity anomaly field, obtained after data reduction, is characterized by anomalies with variable intensity and spatial extension. In particular, small magnetic anomalies are detected within the area affected by mud volcanoes, but wider and more intense anomalies are located in the southeast and northeast parts of the investigated area. Taking advantage of the very short sampling step adopted (2 m) with respect to the real resolution of the survey, a low-pass filter with a cutoff wavelength of 70 m was applied to reduce the high frequency noise and to enhance the effects either of deep-seated bodies or of broad shallow sources. The resulting filtered magnetic map ( Figure 5)

Self-Potential measurements
Self-potential (SP) measurements were acquired at the same sites as the CO 2 efflux, heat flux and gravity measurements ( Figure 1) using a pair of Cu/CuSO 4 non-polarizing electrodes and an insulated electric cable. The SP method consists of measuring the difference of electrical potential between a reference electrode (arbitrarily placed at the beginning of the profile, several tens of meters away from the mud volcanoes) and a mobile electrode, using a high-impedance voltmeter (sensitivity of 0.1 mV, internal impedance of 100 MW). In order to improve the electrical contact between the electrode and the ground, a small hole, generally 10 cm deep, was dug at each site and the electrode inserted in it. Closures of the profiles were made as frequently as possible to warrant an error lower than 10 mV. The Kirchoff law is used to remove the drift of electrodes inside a loop of measurements [e.g., Revil and Jardani, 2013]. Considering the low temperature of the soil when SP measurements were gathered, we rule out that drops due to a Rapid Fluid Disruption effect [Johnston et al., 2001] affected the measurements, which, indeed, appeared stable. Finally, SP measurements in the survey area were not affected by the elevation effect because differences in the elevations among the benchmarks is negligible.

11
Imaging the Salinelle Mud Volcanoes The SP anomaly map (Figure 6) was obtained by automatic interpolation using a kriging method and a 5-m square mesh for gridding. The SP amplitudes range from -45 mV to +35 mV (Table 1) with an average of -10.2 mV and standard deviation of 0.6 mV. The generally low values likely reflect the low eruptive activity of the Salinelle at the time of surveys. The SP anomaly map shows the highest negative values (down to -45 mV) in the north-west part of the investigated area and the highest positive values south of the negative anomaly. Such distribution describes a clear electrical dipole extending over a distance of some tens of meters, very close to the active emissive area at the base of the Conetto dei Cappuccini. On moving southward, a larger negative anomaly is observed. Similar dipole patterns were observed in other studies focused on SP anomalies in volcanic areas Lenat, 2007;Maucourant et al., 2014] and were interpreted as due to convective cells of hot fluids [Michel and Zlotnicki, 1998]. Positive anomalies of smaller magnitude characterize the eastern area, near the old football stadium, in correspondence with the old emissive area.

Discussion and Conclusions
Comparison of the results from geochemical and geophysical surveys in the Salinelle area, although quite complex, allowed us to infer the overall spatial pattern of the fluid flow processes that are acting in the study area. As a general remark, an interesting spatial correspondence was evident among soil CO 2 efflux, heat flux and gravity variations.
Actually, all three parameters show no significant anomaly in the northern half of the study area, whereas a greater variability with marked anomalies occurred in the central and southern half. Some of the highest values both of soil CO 2 efflux and heat flux were measured at the southwestern edge of the main active mud vents area, in correspondence of the northwest slope of the Conetto dei Cappuccini. A similar pattern is evident also from the gravity anomaly map, showing the lowest gravity values in correspondence of the Conetto dei Cappuccini. These most likely are related to the presence of mass deficit or decreased rock density due to a high percentage of porosities/voids in the outcropped lava flows. Furthermore, this area corresponds remarkably well with the largest negative SP anomaly. Generally, in volcanic areas, negative SP values indicate downward motion of fluids, whereas positive values indicate areas of preferential up-rise of fluids [Jackson and Kauahikaua, 1987]. Considering the correspondence between anomalous low gravity and negative SP values, we suppose that this area marks the position of a preferential pathway for meteoric water infiltration, which contributes to the local circulation of underground fluids. In this sense, the geologic structure of the Conetto dei Cappuccini may constitute a high permeability crustal zone acting as preferential shallow pathway for the circulation of fluids.
Conversely, the zone corresponding with the main vents of gas and mud/water emission is characterized by higher gravity values, which seems in contrast to what it is normally expected. Actually, in such a context the gas emission would enlarge the pores in soft near-surface sediments, as well as contribute to widening of the gas channeling fractures. Therefore, these phenomena should produce zones with lower density values within the sediment rock and hence low gravity values at the surface. On the basis of the high gravity values, which could be ascribable to the presence of local higher density subsurface bodies (e.g., compact mudstones with low porosity), we exclude the occurrence of such phenomena and suppose the gas rises from deep through channeling fractures in the shallow impermeable layers. This is supported also by the spatial distribution of soil CO 2 efflux and soil heat flux. Their similar patterns, indeed, strengthen the role of high-enthalpy fluids in carrying both mass and heat up to the surface in the Salinelle area. However, in some points, like P57 to P59 in Figure 3, high CO 2 flux correspond to low heat flux.
This discrepancy is likely the result of the low permeability of the soil at relatively shallow depth. The shallowest layer of soil is mostly made of low-permeability clays that are locally affected by desiccation cracking. Therefore, the hydrothermal fluid rich in CO 2 propagates toward the surface following all possible high-permeability paths formed by cracks and/or other discontinuities. During this process, the fluid cools down, possibly condensate and reaches the surface as "dry" CO 2 -rich gas at places that sometimes can be quite far from the main deeper pathways. This "dispersion effect" of soil gas was already observed on Mt. Etna and discussed as regards soil CO 2 emissions across and along the Pernicana fault system on the northeast flank of Mt. Etna [Giammanco et al., 1997;Azzaro et al., 1998].
Considering the spatial distribution of the geophysical anomalies, the investigated area appears divided into two main sectors, whose sharp contact may delineate a geologic or tectonic structure with an approximately North West-South East direction, that could justify the elongation in the same direction of the area affected by the main eruptive vents (Figure 1). This could be related to the actively growing anticline, rooted at shallow depth, that uplifts Rosalba Napoli et al.
at a rate of about 10 mm/yr producing ground deformation evident in the whole southern periphery of the volcano, including the Salinelle area and that is related to the gravitational spreading of the east and southeast flanks of the volcano [Bonforte et al., 2011].

13
Imaging the Salinelle Mud Volcanoes It is reasonable to suppose, therefore, that the active local tectonics may play a major role in governing the circulation of fluids in the study area and in particular favoring the escape of deep fluids toward the surface.
Probably, those tectonic structures are the same deep faults that allowed magma to erupt at the surface at the time of the formation of the volcanic centers of the Conetto dei Cappuccini, Cono di Paternò (outside the investigated area) and the Ellittico Eruptive Centers lava flows. In the eastern part of the investigated area, although no temperature and gravity anomalies were detected, several small positive SP anomalies appear, corresponding to upward-migrating fluids, and overlap quite well with magnetic lows. Since the spatial distribution of these anomalies is coincident with the old area of emissive vents close to the old football stadium, the reduced magnetization zone can be ascribable to strong hydrothermal alteration of the host rocks. This is supported by the lower gravity values observed in the same area in correspondence of P9 and P11 benchmarks, probably related to clay minerals and leaching produced by hydrothermal alteration. The magnetic anomaly map does not reveal other magnetic lows in the area of the active mud volcano. In agreement with the relatively low temperature values (100-150° C) calculated for the equilibration of the fluids emitted at the Salinelle, based on gas and liquid geothermometry at depth of about 1000 m [Chiodini et al., 1996], and measured at water outlet (40-50° C), this result seems to rule out demagnetization of local rocks induced by high temperature, because this process requires temperature values higher than 500° C. On the other hand, the observed magnetic highs can be related to the presence of lava flows of the Ellittico Eruptive Centers erupted from the Conetto dei Cappuccini [Romano et al., 1979].
Despite the presence of hot fluids in the study area, the amplitudes of the SP anomalies are remarkably low.
Amplitudes ranged between -45 and +40 mV, whereas generally in volcanic environments SP anomalies show amplitude greater than several hundreds of mV [Maucourant et al., 2014]. In the case of Salinelle area, the low SP values could be reasonably due either to the weak eruptive activity of the mud volcanoes when surveys were performed or to the own characteristics of the hydrothermal system related to the local properties of medium and fluids. Considering the small size of SP dipoles depicted at the surface in the studied area, the depth of the corresponding hydrothermal system is probably very shallow [Revil et al., 1999;.
In order to better identify the depth location of SP sources, a dipolar tomography [Patella, 1997;Revil et al., 2014] was applied to the SP profiles AA' and BB'. The concept of probability of charge occurrence is generally used for the tomographic imaging of the charge distribution responsible for the electric current circulation in conductive rocks [Patella, 1997]. In our case, this concept allowed us to recover the dipolar occurrence probability of the source of the dipole responsible for the observed anomalies (Figure 7). The most evident Rosalba Napoli et al. effect is the well-defined separation between zones with negative and positive charge occurrence probability, which appears in profile AA' at depth of about 100 m below the ground surface. The shallow negative source would outline the position of a main path for meteoric water infiltration, located just north of the active emissive vents. The positive source would instead indicate the sites of thermal fluids up-rise. The remarkable wide and deep positive anomaly along the BB' profile would correspond closely to the area of old emissive vents. No other occurrence probability zones are evident at depth greater than 200 m.
The geophysical features outlined by our surveys might thus represent the signature of a very shallow convective hydrothermal system located just a few tens of meters below the ground surface. Geophysical surveys did not reveal a clear preferential pathway through which gases composed of deep CO 2 of magmatic origin, revealed by geochemical surveys, can rise toward the surface. Therefore, it is likely that the hydrothermal uprise system, which regulates the activity of the Salinelle mud volcanoes, is not composed of a single feeding conduit but rather of a combination of several interconnected channels that branch at different depths, according to the stratigraphic sequence described by Panzera et al. [2016]. This interpretation is supported by the observed frequent migration of the emission vents over time, especially during paroxysmal events, although within the geological limits of the Ancient alkaline centers lavas. In fact, the main degassing vents are usually located on the northern slope of the Conetto dei Cappuccini, but emissions may occasionally migrate some hundred meters away. For example, in the past, the emission vents moved toward the old football stadium, which was damaged because of corrosion and weakening of concrete in its structure and it was consequently abandoned. More recently, in early 2016 new vents opened beneath and around a private house located at the southwest corner of the study area, producing a violent and long-standing mud eruption (Figure 2). In a similar way, in early 1990 a strong paroxysmal eruption built the largest mud cone in the area, located close to the 2016 vents [S. Giammanco, unpublished data]. In these occasions, normally the activity in the main degassing vents is strongly reduced, thus supporting the hypothesis of a change in the pathways of fluids coming from the deep feeding system. In support of this observation, soil CO 2 efflux measurements previously carried out in or near our study area both in July 2005 [Giammanco et al., 2007] and in June 2006 (Messina, 2006)  is mostly methane and hence where the chance of gas explosions is higher (e.g., the Macalube of Aragona in southern Sicily, where two kids were recently killed by an explosion of methane and mud). This type of study contributes to the assessment and possibly mitigation of the hazard posed by this type of natural gas emissions.