Active degassing across the Maltese Islands (Mediterranean Sea) and implications for its neotectonics

The Maltese Islands, located in the central Mediterranean Sea, are intersected by two normal fault systems associated with continental rifting to the south. Due to a lack of evidence for offshore displacement and insignificant historical seismicity, the systems are thought to be inactive and the rift-related deformation is believed to have ceased. In this study we integrate aerial, marine and onshore geological, geophysical and geochemical data from the Maltese Islands to demonstrate that the majority of faults offshore the archipelago underwent extensional to transtensional deformation during the last 20 ka. We also document an active fluid flow system responsible for degassing of CH 4 and CO 2 . The gases migrate through carbonate bedrock and overlying sedi- mentary layers via focused pathways, such as faults and pipe structures, and possibly via diffuse pathways, such as fractures. Where the gases seep offshore, they form pockmarks and rise through the water column into the atmosphere. Gas migration and seepage implies that the onshore and offshore faults systems are permeable and that they were active recently and simultaneously. The latter can be explained by a transtensional system in- volving two right-stepping, right-lateral NW-SE trending faults, either binding a pull-apart basin between the islands of Malta and Gozo or associated with minor connecting antitethic structures. Such a configuration may be responsible for the generation or reactivation of faults onshore and offshore the Maltese Islands, and fits into the modern divergent strain-stress regime inferred from geodetic data.

Active fluid seeps have generally been detected and characterised via direct observation and sampling. Offshore, echo-sounders have been used to record acoustic anomalies due to gas bubbles in the water column (Gentz et al., 2014;Jerram et al., 2015;Schneider von Deimling et al., 2007). Near-bottom visual observations by divers or platforms equipped with a camera, such as towed frames or Remotely Operated Vehicles, have enabled identification of active gas bubble seepage (Gasperini et al., 2012). Probes and sensors, mounted on these platforms and using gasextraction step, biosensing or solid-state optical measurements, have been used to detect and measure dissolved methane in situ over long periods of time (Boulart et al., 2010;Gasperini et al., 2012). Discrete water samples collected with Niskin bottles have been analysed for gas chromatography, molecular composition and stable isotopes to determine the concentration and genesis of the fluids (Etiope et al., 2014;Gentz et al., 2014;Savini et al., 2009). Geochemical analyses of the sub-stratum and its interstitial waters have yielded information on rate and timing of fluid seepage (Abrams et al., 2001;Toki et al., 2004;Panieri, 2006). Satellite remote sensing imagery has been used to detect floating oil over active seeps (Espedal and Wahl, 1999) and provide estimates of seepage rates (MacDonald et al., 1993). Onshore, detection of active gas seeps and inference of their source have been based on direct soil gas measurements (Cicerone and Shetter, 1981;Maljanen et al., 2004), and on the chemical and isotopic analyses of the dissolved gas phase in groundwater bodies and springs (Italiano et al., 2009).
The Maltese Islands, located in the central Mediterranean Sea, are intersected by two systems of faults. The ages assigned to these fault systems span from the Early Miocene to mid-Pliocene and are mainly based on interpretation of poor quality offshore seismic reflection profiles (Dart et al., 1993;Illies, 1981). Due to the lack of seafloor evidence of fault displacement and insignificant historical seismicity, the fault systems are thought to be inactive and rift-related deformation is believed to have largely ceased (Dart et al., 1993;Illies, 1981). There are no reports of neotectonic movements along onshore faults (Bonson et al., 2007), and the most recent fault activity has been related to displacement of late Pleistocene-Holocene age alluvial fanglomerates along one fault in the south of Malta (Government of Malta, 1993). In our study, we integrate aerial, marine and onshore geological, geophysical and geochemical data from the Maltese Islands to: (i) document an active fluid flow system responsible for degassing of CH 4 and CO 2 , (ii) demonstrate that the two fault systems were recently active, and (iii) suggest that onshore and offshore faults were either formed or reactivated by an active transtensional system.

The Pelagian Platform
The Pelagian Platform, located in the central Mediterranean Sea (Fig. 1), forms part of the African continental plate and consists of a 25-30 km thick continental crust. It extends from southern Sicily to Tunisia in the west, northern Libya in the south and the Malta Escarpment to the east (Finetti, 1982). The upper sedimentary units on the Pelagian Platform comprise Plio-Pleistocene units of terrigenous, pelagic and hemipelagic sediments that are up to 300 m thick (Max et al., 1993;Osler and Algan, 1999). Underlying these units are > 4 km thick sedimentary sequences of Miocene to Cretaceous shelf edge carbonate build ups, and Cretaceous to Triassic shallow platform carbonates (Jongsma et al., 1985;Scandone, 1981;Torelli et al., 1998). The carbonate sequences are punctuated by extensive depositional hiatuses, as well as tuffs and pillow lavas deposited during a number of volcanic episodes (Bosellini, 2002;Scandone, 1981).
The Pelagian Platform is a foreland domain that, between the Mesozoic and Cenozoic, underwent NeS compression due to the convergence of the African and Eurasian plates (Ben-Avraham and Grasso, 1991;Burollet et al., 1978;Carminati et al., 2004;Goes et al., 2004;Gueguen et al., 1998). Shortening waned during the Miocene, when deformation became mainly extensional. During the Late Miocene, three NW-SE trending grabens (Pantelleria, Malta and Linosa) started forming in the central part of the Pelagian Platform, leading to the development of the 600 km long Sicily Channel Rift Zone (Civile et al., 2008(Civile et al., , 2010Dart et al., 1993;Finetti, 1984;Finetti et al., 2005;Jongsma et al., 1985) (Fig. 1). The grabens have a water depth of up to 1700 m; they are filled by up to 2000 m thick Lower Pliocene-Pleistocene turbidites and are bound by sub-vertical normal and oblique faults (Civile et al., 2010). Rifting has resulted in a thinned continental crust and has been accompanied by widespread subaerial and submarine volcanic activity (e.g. Pantelleria, Linosa, Graham and Nameless banks) (Argnani, 1990;Civile et al., 2008). Three explanations have been put forward for the origin of the Sicily Channel Rift Zone: (i) The grabens are pull-apart basins developed along a major dextral wrench zone (Finetti et al., 2005;Jongsma et al., 1985;Reuther and Eisbacher, 1985); (ii) The rifting is associated with mantle convections developed during the roll-back of the African lithospheric slab beneath the Tyrrhenian Basin (Argnani, 1990;Reuther et al., 1993); (iii) Intraplate rifting related to the north-eastern displacement of Sicily away from the African continent (Illies, 1981;Winnock, 1981).

The Maltese Islands
The Maltese Islands are located in the central eastern part of the Pelagian Platform, along the northern rim of the Malta Graben (Fig. 1a). The archipelago consists of the islands of Malta, Gozo and Comino, which comprise a shallow water, Oligo-Miocene sedimentary succession of five main formations (Pedley et al., 1976). This sequence is disrupted by two normal fault systems, which have a predominant control on the subaerial geomorphology of the archipelago (Alexander, 1988) and which are closely related to the development of the Malta Graben (Illies, 1981;Reuther and Eisbacher, 1985). The most widespread system consists of ENE-WSW trending faults dipping at 55-75°. Faults belonging to this system mainly occur between Malta and Gozo, where they form a ∼15 km wide horst and graben structure. The most prominent fault of this system is the Great/Victoria Fault (Fig. 1b). The second system consists of NW-SE trending, oblique flank-rift faults. The most impressive of these faults is the Maghlaq Fault, which runs along the southern coastline of Malta (Bonson et al., 2007) (Fig. 1b). The offshore faults tend to be oriented parallel to these two fault systems (Fig. 1a). Fig. 1c shows the locations and focal mechanisms of earthquakes that occurred on, and in the vicinity of, the Maltese Islands between 2010 and 2018. Only nine focal mechanisms could be determined, and these are dextral strike-slip with an extensional component. Two of these coincide with the Maghlaq Fault and a NW-SE trending escarpment located north of Malta.

Aerial data
In November 2017, an aerial survey was carried out to determine CH 4 and CO 2 concentrations in the atmosphere across the Gozo Channel (Fig. 2a). Data were acquired with a Los Gatos Research Ultraportable Greenhouse Gas Analyser carried on-board of a Cessna 172 aircraft. This instrument provides accurate CH 4 and CO 2 measurements at levels up to 10% mole fraction (without dilution) and without reducing precision and sensitivity at typical ambient level. The error margin of the instrument is < 2 ppb for CH 4 and < 300 ppb for CO 2 . The instrument inlet pipe was placed on the wing leading edge near a fresh air inlet used for the conditioning of the cockpit. This allowed us to sample undisturbed and uncontaminated air. The flight path entailed 22 parallel lines over an area of 30 km 2 between Malta and Gozo. The height of the flight was set to 500 feet and the speed of the aircraft was maintained at 70 knots. The flight was performed in good weather conditions, with clear sky and low wind velocities.
We also acquired one air sample at sea level in the Gozo Channel (Fig. 2b) using a two-valve glass bottle. The chemical composition and concentration of the gases was measured using the method explained in section 3.2.5.

Marine data
Our marine data were acquired during seven research cruises: HMS Roebuck (

Bathymetric data
An overall area of ∼330 km 2 of seafloor was surveyed using hullmounted multibeam echo-sounder systems (Fig. 2a). The instruments used to acquire each multibeam echo-sounder data set and the cell size of the processed grids are listed below: Bathymetry was derived from the MBES surveys by accounting for sound velocity variations and tides, and by implementing basic quality control. Bathymetric data were exported as 32-bit rasters. Our data sets were complemented by ∼430 km 2 of bathymetric data acquired with a Hawkeye IIb bathymetric LiDAR sensor and an interferometric swath system (Kongsberg GeoSwath), collected as part of the project "Development of Environmental Monitoring Strategy and Environmental Monitoring Baseline Surveys" funded by ERDF-156. The cell size for these data sets is 10 m.

Acoustic profiles
980 km of acoustic profiles were collected using three systems ( Fig. 2a): (i) a pole-mounted Innomar SES-2000 compact sub-bottom profiler, with primary and secondary frequencies of 100 kHz and 8 kHz, respectively; (ii) hull-mounted 16 transducer CHIRP-II profiler (BEN-THOS DATASONICS Mod. CAP-6600) with an operating frequency of 3.5 kHz; (iii) a Boomer seismic source with an operating frequency of 3.5 kHz and a 50 m long Geo-Sense mono-channel streamer with a sensitivity of ± 2 dB (Fig. 2b). The Boomer data were processed with a conventional mono-channel processing sequence that included band pass filtering, gain recovery, seismic signal enhancement, attenuation of noise eddy currents, and seawater column mute.  (Zhu and Helmberger, 1996) using the permanent Malta Seismic Network and INGV stations located in Sicily. Earthquake location is obtained by automatically analysing the 3-component single-station polarization data within 10-15 km from WDD station may be affected by large location error since backazimuth estimations are not always accurate (Agius and Galea, 2011). Faults are denoted in black. Escarpments identified in this study (4.1.2) are shown in red. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.)

Multi-channel seismic reflection profiles
240 km of profiles have been acquired from the Gozo Channel using a mini GI gun with a total volume of 60 cu. in. (1 l) (Fig. 2c). A shot point distance of 15.625-18.750 m and a recording length of 2 s were used. Data were recorded using a 300 m long digital streamer with 96 channels, with a channel distance of 3.125 m. The fold coverage ranged between 8 and 9.6 traces per CDP. The processing sequence included amplitude recovery, bandpass filtering, CDP sorting, pre-stack deconvolution, velocity analyses, normal move out correction, and stacking.

Seafloor sediment samples
Seventeen superficial sediment samples were collected using a 5 l modified Van-Veen Grab (Fig. 2b). These samples were described, photographed and sub-sampled on-board. Grain size distribution of the samples was analysed using sieves, following the ASTM D0422, and a Malvern Mastersizer 3000. Five sediment samples (Site 1, Site 2, Site 8, Site 9, and MH9) were analysed using X-ray diffraction (XRD) to determine the qualitative and semi-quantitative mineralogical composition of the carbonate fraction. Mineral identification was performed on dried and ground sub-samples using a Rigaku MiniFlex XRD (a) Spatial coverage of the aerial data, marine data (multibeam echo-sounder data, sub-bottom profiles, water samples) and onshore data (groundwater and soil gas samples; LiDAR data). (b) Spatial coverage of the Boomer acoustic profiles, water and sediment samples, and air sample. c) Spatial coverage of the multichannel seismic reflection profiles. diffractometer.

Water column samples
Twenty four water column samples were collected with a 5 l Niskin bottle ( Fig. 2a-b). Immediately after the Niskin bottle returned on deck, a drawtube was pre-rinsed with sample water and attached to the Niskin bottle's spigot. Glass and PET flasks were then filled and overflowed, avoiding the formation of air bubbles in order to prevent air contaminations. In the PET flasks, a headspace was created and one drop of saturated HCl was added to the sample. All the flasks were stored in the dark. Geochemical analysis were performed in the laboratories of the Istituto Nazionale di Geofisica e Volcanologia (sezione di Palermo). The chemical composition and concentration of the gases dissolved in seawater samples were determined by using the method in Capasso and Inguaggiato (1998). For the gas chromatography analyses, the sample was split in two aliquots. The first was analysed for O 2 , N 2 , CH 4 and CO with an Agilent 7890B with two columns in series (Poraplot U 25 m × 0.53 mm and Molsieve 5A 25 m × 0.53 mm) fluxed by Ar (detectors TDC and FID with methaniser). The second aliquot was analysed for CO 2 and H 2 S by a microGC module (MicroGC 3000) equipped with Poraplot U column (15 m) fluxed by He (detector TCD). Calibration was made with certified gas mixtures. Analytical precision was always better than ± 3%. The detection limit was ∼0.3 ppm for CO and CH 4 , 30 ppm for CO 2 , and 200 ppm for O 2 and N 2 . The isotopic ratio of oxygen (δ 18 O) was measured using a mass spectrometer Thermo Delta V Plus coupled to a GasBench II that exploits the principle of the head space. For the determination of hydrogen isotopic ratio (δD), we utilised a mass spectrometer Delta Plus XP coupled with a TC/EA reactor. The analytical precision is better than ± 0.1% and ± 1% for δ 18 O and δD, respectively. Isotope ratios are expressed using delta notation as relative differences in parts per mil (δ values %) from Standard Mean Ocean Water (SMOW).

Topographic LiDAR data
Topographic LiDAR data were acquired during a 5.5 h flight in February 2012 using an IGI LiteMapper 6800 system (Fig. 2a). The data have a cell size of 1 m and were collected as part of the project "Development of Environmental Monitoring Strategy and Environmental Monitoring Baseline Surveys" funded by ERDF-156.

Groundwater samples
Water samples were collected at eighteen onshore sites (Fig. 2a), from either boreholes or galleries (Ta' Kandja, Malta), in January 2014. The depth of the sampled water ranges between 29 and 234 m below the surface. Samples for isotopic measurements were collected and stored in PET bottles and in triplicate: one sample as-is, one filtered through Millipore 0.45 μm filters, and one filtered and acidified with suprapure-grade HNO 3 . Samples collected for the extraction of the dissolved gases were stored in 240 ml pyrex bottles sealed in the field using silicon/Teflon septa and purpose-built pliers. Details on the methodology and instrumentation are described in Italiano et al. (2013Italiano et al. ( , 2014.
The water isotopic composition (δ 18 O and δD) of unfiltered samples was determined by mass-spectrometry and expressed in ‰ with respect to the international Vienna Standard Mean Ocean Water (V-SMOW). The dissolved gases were extracted following the equilibration method (Capasso and Inguaggiato, 1998). To prevent atmospheric contamination, all samples were stored upside down with necks submerged in water until the laboratory procedures were initiated. The chemical composition of the dissolved gas phase was obtained from the gaschromatographic analyses by taking into account the solubility coefficients of each gas species (Bunsen coefficient "β", cc gas STP/ml water (cubic centimetres at standard temperature and pressure)), the volume of gas extracted and the volume of the water sample (more details are available in Italiano et al. (2009Italiano et al. ( , 2014. Measurements of water isotopic composition and dissolved gas chemistry and concentration were carried out by using the same methodology described in section 3.2.5.

Soil gas measurements
Seventy nine soil gas measurements were carried out across the Maltese Islands (Fig. 2a). The active method was adopted, whereby soil gas is collected by a pump at a constant flow rate and driven through a pipe, 2.5 cm 2 in cross-section, which is inserted at a depth of about 0.5 m in the soil. The CO 2 concentration value of the analysed gas mixture (soil CO 2 and air) was taken when it reached steady state (constant value). The concentration is related to the CO 2 flux by the relationship: where Φt is the CO 2 flux, CD is the measured CO 2 concentration, and F is the flow rate of the pump. Since our objective was to identify the presence of a positive CO 2 flux towards the atmosphere, only the flow rate of the pump and the CO 2 concentration were taken into consideration. A ± 10% uncertainty was estimated by repeated measurements at the same site.

Sub-seafloor seismic facies and anomalies
The acoustic profiles exhibit two man facies (Fig. 3, 4a-b). The shallower one, facies A, reaches a maximum thickness of 0.05 s (TWTT) and consists of sub-parallel, high amplitude reflectors with good lateral continuity. We correlate this facies with post-Last Glacial Maximum medium to fine sands (Micallef et al., 2013). The deeper facies, facies B, shows very high amplitudes along the top with seemingly insufficient signal penetration beneath; it correlates with Late Miocene carbonates. Correlations are based on information from boreholes drilled offshore Comino (Infrastructure Malta, personal communication, October 2018). These two facies can also be identified in the multi-channel seismic reflection profiles (Fig. 5). Reflectors in these two facies, as well as those underlying them, are offset vertically in the Gozo Channel or to the east and south of Malta (Figs. 3, 4b and 5). By correlating offsets that showed the same dip and displacement across parallel seismic reflection profiles, we were able to map the main trend of the offsets as 38 lineaments. Sub-seafloor offsets coincide with escarpments on the seafloor (see section 4.1.2). The offsets reach the maximum penetration of the seismic signal (∼1 s TWTT).
Facies A and B are also affected by two types of acoustic anomalies that disturb the lateral continuity of the seismic reflectors. The first type are seismic chimneys (up to 15 m wide) (Figs. 3 and 4a), whereas the second type comprise acoustic turbidity that extends laterally up to 2 km (Fig. 5). Both anomalies reach the maximum penetration of the seismic signal (∼1 s TWTT). Acoustic turbidity can be correlated across parallel seismic reflection profiles.

Seafloor morphology
The seafloor predominantly consists of flat to gently sloping terrain that comprises 85 rectilinear escarpments oriented NW-SE, W-E or SW-NE (Figs. 3, 4b-c). The escarpments to the east and west of Malta are predominantly oriented NW-SE; they are up to 13 km long and 50 m high with slope gradients of > 60°. To the north and south of the Maltese Islands, the escarpments are up to 10 km long, 100 m high, up to 80°in slope gradient, and predominantly oriented SW-NE. The Gozo Channel hosts 37 quasi-linear escarpments that are predominantly oriented NW-SE, up to 700 m long, 45 m high, and 80°in slope gradient.
The seafloor also hosts 1026 circular to ellipsoidal depressions ( Fig. 4c-d). Of these, 63% are located to the north-east of the Maltese Islands and 21% to the north of Gozo; the remaining 3% and 13% are located to the south-east of Malta and in the Gozo Channel, respectively. The highest depression densities (up to 35 per km 2 ) are observed offshore NE Gozo and in the Gozo Channel. The depressions occur at depths of 20-205 m, and they are up to 100 m wide and 3 m deep. The depressions generally cluster in NW-SE or WSW-ENE orientations, and appear to coincide with fine to medium sand cover (see Micallef et al. (2013)).

Seafloor sediments
Sediments collected from the seafloor depressions in the Gozo Channel predominantly consist of poorly to moderately well sorted fine to coarse sands. The samples that were analysed with XRD are mainly composed of carbonates associated with detrital minerals (clays) and minor authigenic phases (mostly barite and pyrite) (Fig. 6b). Calcite represents the dominant carbonate phases and is associated with minor quantities of dolomite (Fig. 6b).

Dissolved gases and water chemistry
Gas concentrations in the water samples are shown in Fig. 6a and Table 1 (Supplementary materials). Concentrations of CH 4 and CO 2 dissolved in seawater are above 4.79 × 10 −5 ccSTP/l and 0.43 ccSTP/l, respectively, showing that both the volatiles are more enriched than seawater in equilibrium with the atmosphere (ASSW = Atmosphere Saturated Seawater; Capasso and Inguaggiato, 1998). The highest concentrations were recorded to the north-east of Malta (Fig. 6a). The isotopic signatures of oxygen and hydrogen for the collected seawater samples show a composition that is comparable to the Modern Mediterranean Seawater (Fig. 6c).

Water column amplitude anomalies
In the acoustic profiles we identified 43 water column amplitude anomalies. These anomalies consist of straight or inclined high-amplitude acoustic scattering that extend upwards from the seafloor (Fig. 3). None of the acoustic anomalies appear to reach the sea level. The majority of these anomalies were recorded in the Gozo Channel, with the rest located to the NE of Malta.

Dissolved gases and water chemistry
Atmospheric gases (O 2 and N 2 ) dominate the composition of the gas phase dissolved in shallow groundwater samples (Table 2 -Supplementary materials). A substantial amount of volatiles of different origin than the atmosphere, derived from gas/water interaction processes within the water bodies, is recognised as a common feature of almost all the sampled waters. After the atmospheric gases, the most Fig. 7. Onshore water and gas chemistry. (a) Measured concentration of CH 4 and CO 2 in the onshore groundwater samples. (b) Concentration of CO 2 derived from soil gas measurements (adapted from Birhane (2016) and Gauci (2015)). abundant gas in the shallow groundwater samples is CO 2 , with concentrations of 2.61-10.78 ccSTP/l (Fig. 7a). CH 4 was only detected in two samples (Fig. 7a).
The isotopic composition (Fig. 8) shows that most of the groundwater samples fall on the Local Mediterranean Water Line (LMWL).

Soil gas chemistry
Soil degassing measurements carried out in Malta and Gozo highlight the presence of active CO 2 emission, the concentration of which ranges between 540 and 40080 ppm for Malta, and between 660 and 8780 ppm for Gozo (Fig. 7b). The highest CO 2 concentrations were recorded along the Great and Maghlaq Faults.

Aerial data
The spatial distribution of CH 4 and CO 2 concentrations across the Gozo Channel is shown in Fig. 9. Values range between 1.850 and 1.991 ppm for CH 4 , and between 403 and 406 ppm for CO 2 . The dispersion of a single measurement value is in the range of 4.036-4.050 × 10 −2 ppm, and the precision of the instrument is of the order of 1.5 × 10 −3 ppm. The highest concentrations of these gases occur as SW to NE elongated anomalies located between Gozo and Comino. The anomalies appear to partly correlate with the flight path. However, in view of the low line spacing used and the fact that the same point was covered more than once by different fly overs, the correlation is only geometrical, the generation of artefacts has been reduced and interpolation artefacts were minimised. Our measurements are also not correlated with the wind direction, which during the day of the survey blew from WNW at a velocity of 5-7 knots. The gridding algorithm used was a conservative one and based on a moving average. The concentrations of CH 4 and CO 2 from the ship-borne air measurement are 2.970 ppm and 1600 ppm, respectively.

Spatial correlation between observations
Onshore, sites with groundwater CO 2 and CH 4 anomalies and CO 2 degassing coincide with known faults (Fig. 7). Offshore, atmospheric CO 2 and CH 4 anomalies in the Gozo Channel are spatially associated with water column amplitude anomalies, offsets in seismic reflection profiles, escarpments, depressions, seismic chimneys, and acoustic turbidity zones (Fig. 10). CO 2 and CH 4 anomalies in the water column are spatially associated with CO 2 and CH 4 anomalies in the atmosphere and seafloor depressions. The seafloor depressions are aligned with escarpments and offsets in seismic reflection profiles, particularly to the NE of Gozo and Malta, and spatially correlate with seismic chimneys and acoustic turbidity zones (Fig. 10).

Type and age of offshore faults
We interpret the offset seismic reflectors in the sub-seafloor marine data as faults, and the escarpments in the multibeam bathymetry data as their topographic expressions (Figs. 3, 4b and 5, 10). The sub-seafloor marine data suggest that the majority of these faults are extensional and belong to either the WSW-ENE or NW-SE fault systems onshore (Fig. 11). The focal mechanisms of earthquakes in the vicinity of the faults offshore north Malta and in the south of Malta indicate dextral transtensional deformation (Fig. 1c). 90% of the mapped offshore faults are likely structures that have recently been active (last 20 ka) because they displace facies A (we assume that sedimentation covered all existing topography) and are associated with seafloor escarpments. Only one-fifth of the offshore faults constitute extensions of onshore faults; 75% of these are oriented WSW-ENE, whereas the other 25% are oriented NW-SE. The fault inferred from the offset shown in Fig. 3a connects with the Maghlaq Fault and exhibits strike-slip deformation.

Nature and activity of the fluid flow system
Six types of evidence indicate that the degassing across the Maltese Islands is recent or presently active: (i) amplitude anomalies in the water column (Figs. 3 and 10), which we interpret as acoustic scattering by bubbles in the water associated with active seepage from the seafloor (e.g. Ceramicola et al., 2018;Sauter et al., 2006); (ii) the occurrence of authigenic phases (barite and pyrite) in seafloor sediment samples (Figs. 6b and 10), which suggest active, slow-flux gas systems (e.g. Bezrodnykh et al., 2013;Roberts et al., 2006); (iii) CO 2 and CH 4 anomalies in the seawater samples (Fig. 6), which are higher than airsatured seawater values (Coltelli et al., 2016;Correale et al., 2012;Gasperini et al., 2012;Savini et al., 2009); (iv) CO 2 and CH 4 anomalies in the atmosphere above the Gozo Channel (Fig. 9b-c); (v) CO 2 and CH 4 anomalies in onshore groundwater samples (Fig. 7a); CO 2 values are well above the equilibrium with ASW (0.24 ccSTP/l H 2 O, and CH 4 concentrations are 2-3 magnitudes above ASW (Weiss, 1974); (vi) CO 2 degassing in soils onshore (Fig. 7b).
The seafloor depressions, which we interpret as pockmarks, provide additional evidence of escaping gases (Hovland et al., 2002;Hovland, 1992, 2007;King and MacLean, 1970). We interpret the subseafloor acoustic anomalies observed in the seismic reflection profiles as fluid flow pathways -the seismic chimneys as pipe structures and the wide acoustic turbidity zones as a result of the attenuation of acoustic energy by gas in the pore space, which causes chaotic reflections (Aloisi et al., 2004;Bünz et al., 2012;Goswami et al., 2017;Holland et al., 2003).
We therefore propose that the Maltese Islands host an active fluid flow system. This involves the upward migration of gases through the carbonate bedrock and overlying sedimentary layers via focused pathways, such as faults and pipe structures (e.g. Gasperini et al., 2012;Gay et al., 2012;Saffer, 2015). Diffuse pathways are also important for gas migration and are likely associated with fractures in the carbonate bedrock (e.g. Dimmen et al., 2017). Where gases are expelled in sandcovered areas offshore, they form pockmarks. They do not form in other parts of the study area, which comprise bedrock or gravel-sized maërl (Micallef et al., 2013). Where water depths are lower than a few tens of metres, the gases rise through the water column and reach the atmosphere. Fluid escape is spatially variable, with the highest fluxes (based on concentrations and density) occurring to the NE of Gozo and between south Gozo and S Malta.
The sources of the gases are difficult to determine without carrying out more detailed geochemical analyses. CO 2 is the most abundant gas species in crustal environments. It originates from the degradation of organic matter or is degassed by active volcanic and active tectonic systems through faults and fractures (e.g., Evans and Wong, 1992;Kennedy et al., 1997;Italiano et al., 2000;Caracausi et al., 2005). CH 4 seepage is generally associated with bacterial degradation of organic Values range between 1.850 and 1.991 ppm for CH 4 , and between 403 and 406 ppm for CO 2 . The highest concentrations of these gases occur as SW to NE elongated anomalies located between Gozo and Comino.

Implications for the neotectonics of the Maltese Islands
The occurrence of active gas seepage leads us to propose that the faults intersecting, or in the vicinity of, the Maltese Islands are permeable and thus recently displaced. To explain our observations of gas seeping from both ENE-WSW and NW-SE trending structures, we need a scenario where both fault systems on the Maltese Islands are simultaneously active. One such scenario entails a transtensional fault system (Fig. 11). This could comprises two right-stepping, right-lateral NW-SE trending faults (fault offshore the eastern coasts of Gozo, Comino and Malta; the Maghlaq fault and its offshore extension), binding a pull-apart basin between the islands of Malta and Gozo (Fig. 11).
Alternatively, it could comprise two parallel right-lateral strike-slip faults and associated minor connecting antitethic (left-lateral) structures. Our proposed scenario is supported by the dextral strike-slip focal mechanisms of recent earthquakes in the vicinity of faults offshore north Gozo (Fig. 1c), as well as by geodetic data (Fig. 11 -inset). Horizontal velocity fields obtained by Global Navigation Satellite System and Global Positioning System stations reveal that the strainrate field across the eastern Sicily Channel is extensional and that stretching is taking place between Lampedusa and the Maltese Islands at a rate of 1.4 mm per year in a SW-NE direction (Meccariello et al., 2017;Palano et al., 2012). This implies an active, diverging strain-stress regime across the Maltese Islands, albeit with a low deformation rate. It is likely that faults onshore and offshore the Maltese Islands were formed, or are being reactivated, by this modern stress field. The proposed transtensional fault system likely forms part of a more extensive fault system to the south (Illies, 1981;Jongsma et al., 1985;Dart et al., 1993) and represents one of the best configurations for fluid ascent and   11. Proposed transtensional fault scenario and its relation to the present strain-stress regime across the Maltese Islands (derived from available Global Navigation Satellite System velocity field (Meccariello et al., 2017)). Inset on the left shows a cartoon of a scenario with two right-stepping, right-lateral NW-SE trending faults binding a pull-apart basin. Inset on the right shows vectors decomposition (LAMP: Lampedusa, MALT: Malta), which indicate that the area is undergoing stretching in a N50E direction (blue arrows). Onshore faults are shown in black whereas offshore faults, inferred from multibeam echo-sounder and seismic reflection data, are shown in dark red. (For interpretation of the references to colour in this figure legend, the reader is referred to the Web version of this article.) venting (Evans and Wong, 1992;Kennedy et al., 1997;Maloney et al., 2015).

Conclusions
We have integrated aerial, marine and onshore geological, geophysical and geochemical data to document an active fluid flow system onshore and offshore the Maltese Islands. The system entails the upward migration of CH 4 and CO 2 through carbonate bedrock and overlying sedimentary layers via focused pathways, such as faults and pipe structures, and possibly via diffuse pathways such as fractures. Where the gases seep offshore, they form pockmarks in sand-covered seafloor. Where water depths are lower than a few tens of metres, the gases reach the atmosphere.
We also mapped 85 faults offshore the Maltese Islands that belong to either the WSW-ENE or NW-SE fault systems onshore. Geophysical data indicate that the majority of these faults were active during the last 20 ka and underwent extensional to transtensional deformation. The migration of gases implies that the onshore and offshore faults are permeable, and that the two fault systems were active recently and simultaneously. The best structural configuration that can explain the latter is a transtensional system involving two right-stepping, rightlateral NW-SE trending faults, either binding a pull-apart basin between the islands of Malta and Gozo or associated with minor connecting antitethic structures. Such a configuration, which may be responsible for the formation or reactivation of faults onshore and offshore the Maltese Islands, fits into the modern divergent strain-stress regime derived from Global Navigation Satellite System and Global Positioning System station data.

Data statement
The data used are listed in the tables and figures. The acoustic profiles are available from the corresponding author upon reasonable request. Access to the Boomer acoustic profiles and multichannel seismic reflection data from the Gozo Channel is restricted for privacy purposes; readers can apply to sarah.a.pace@infrastructuremalta.com for access.