This study explores, for the first time, the response of the Mediterranean seafloor to desiccation and its affect on climate during the Messinian lowstand. New high-resolution 3-D pre-stack depth migrated seismic reflection data show evidence for gas outflow stemming from pre-Messinian sources. Our results indicate that giant pockmarks formed during this lowstand. Emission continued throughout the Messinian and persisted after it ended as evident by pockmark arrays on the then-seafloor. High reflectivity between the top-Messinian and overlying Pliocene sediments indicates significant gas accumulation immediately below the latter. Attribute analysis show minor chaotic paths through the Plio-Pleistocene, which do not reach the present-day seafloor. Our data indicate that as long as sea level was low there was massive gas escape to the shallow sea and atmosphere. We suggest that this probably resulted in the mid-Messinian climatic shift. Major emissions identified here indicate an indirect cause to negative climatic feedback during this period.
 Natural gas seeps in the marine environment play a crucial role in the global methane budget, releasing some 6.6–19.5 Tg per year into the atmosphere [Judd et al., 2002]. Pockmarks and mud volcanoes, which are observed on the seafloor globally, are associated with venting of gas-rich fluids [e.g.,Dimitrov, 2002; Loncke et al., 2004; Lykousis et al., 2009; Dupré et al., 2010]. The presence of such gas escape features from within the deep sub-surface sedimentary section therefore indicate gas release activity in the geological past [Moss et al., 2012].
 Studies have suggested that the flux of gas escaping from mud volcanoes and through the water column may have a somewhat significant contribution to atmospheric greenhouse gases [e.g., Dimitrov, 2002; Kopf, 2002]. However, this semi-direct contribution may be limited to water depths shallower than 75 m [Judd et al., 1997] as gas bubbles from underwater seeps will undergo almost total solution in greater depths.
 At the end of the Miocene tectonic forces led to the closure of the Betic and Rifan Corridors [e.g., Krijgsman et al., 1999; Jolivet et al., 2006]. As a result, the Mediterranean began dried out, during the Messinian Salinity Crisis (5.96–5.33 Ma), leading to the deposition of 1–2 km of sediments up to the lowstand sea-level of the end of the Messinian (∼1 km below present [Garcia-Castellanos and Villaseñor, 2011]). The composition of these sediments within the eastern Mediterranean is still debated. Several studies interpreted this unit as halite and anhydrite series [Tibor et al., 1992; Netzeband et al., 2006] and others suggested the interbedding of clastics [Garfunkel and Almagor, 1984], indicating a change of evaporite facies, possibly deposited during temporal sea level rises [Netzeband et al., 2006]. This interpretation is more feasible since Pre Stack Depth Migration-based velocity analysis shows a relatively slow velocity with respect to what is typical for salt units.
 It is commonly thought that the Messinian unit forms a vast hydrocarbon seal that covers most of the Mediterranean basin [e.g., Loncke et al., 2004; Pawlewicz, 2004]. This acts as a barrier for hydrocarbon migration from the deeper, pre-Messinian section, where most of the hydrocarbon reservoirs are thought to occur [Pawlewicz, 2004]. Pre-Messinian deep-water turbidite sands form the principal reservoir of the recent Tamar, Dalit and Leviathan gas field discoveries offshore Israel [Bowman, 2011]. In areas where the Messinian evaporites are faulted, thin or absent (towards the shore), hydrocarbons can escape and penetrate through the Messinian evaporitic column [Loncke et al., 2004; Gradmann et al., 2005; Bowman, 2011].
 As a result of desiccation, large parts of the then seafloor were exposed, while other parts remained covered by shallow water. Towards the mid-Messinian (5.55 Ma) European temperatures, humidity and runoff from the Alps and Trodos increased and the Messinian entered what is known as the Lago Mare stage [Willett et al., 2006; Murphy et al., 2009]. Clauzon et al.  and Bache et al. distinguish between Lago Mare deposits (1) in the peripheral basins caused by Mediterranean-Paratethys high sea-level connections (both in the Messinian and Zanclean times) and (2) in the main basins of the Mediterranean, the cause of which is still debated. Here we focus on the latter setting. While it is accepted that climate did not induce the initiation of the Salinity Crisis [e.g.,Fauquette et al., 2006], the mid-Messinian climatic shift has not been explained. Similarly, it is not clear whether the shift represents a transition to the Lago Mare stage. Our data present new evidence for major gas escape through a relatively thick section of the Messinian unit within the deep eastern Mediterranean basin. We examine its migration paths and release mechanism to propose that this release triggered a climatic shift in correlation with data presented by Shackleton et al. and others. Results are based on 3-D pre-stack depth migrated seismic volume in the eastern Mediterranean combined with 2-D industrial seismic reflection data (Figure 1a).
 A 60 × 24 km high-resolution 3-D pre-stack depth migrated [Reshef, 1991] seismic reflection volume is used in this study. The volume extends from water depths of 1200 m to 1600 m. The survey was acquired with a 50 m trace interval on both inline and cross-line. Vertical and lateral resolutions of the data are ∼6 m. Since data were collected for industrial purposes, the volume was cut off at just below the base of the Messinian section. 2-D multi channel time migrated seismic reflection data were compiled across the entire area offshore Israel (Figure 1). Average penetration is 8–10 seconds two-way-time, allowing for examination of pre-Messinian stratigraphy (Figure 1b).
3. Evidence for Gas Migration
 In general, the presence of gas was inferred to by the occurrence of pockmarks, volumes of acoustic blanking, possible mud volcanoes and bright spots, which were mapped by the analysis of various seismic attributes, such as variance and chaos (for detection of discontinuities such as faults and chimneys), envelope (bright spots, major lithological changes and sequence boundaries detection) and sweetness (changes in overall energy signal). Extraction of 3-D structures was carried out using volume and horizon probes.
 The pre-Messinian section was examined using 2-D industrial data only. Seismic data show clear indication for horst structures with associated blanking and bright spots onlapping along the flanks (Figure 1b). Several chimney-like acoustic blankings were detected above the top of this structure, as well as above its flanks.
 Messinian stratigraphy was examined in the 3-D reflection data. The base of this unit (Figure 2) forms a relatively smooth surface that slopes westwards and contains multiple fractures in an NW-SE direction. Along this surface, a group of circular depressions (pockmarks) were found directly above the acoustic blanking and horst structure mapped in the pre-Messinian section (Figures 1b and 2). The pockmarks appear as near perfect circles and are located at depths ranging from ∼3600–3200 m below the present-day sea level and are spread out in an area of ∼130 km2. These features occur in groups, along the western flank of a NW trending, ∼3.5 km wide lineament, are in the range of 1000 m wide by 150 m deep (Figure 2), and seem to be among the largest ever reported globally [Newman et al., 2008, and references therein]. The NW trending lineament along which the pockmarks appear could be interpreted as a fault zone, however, no such structural element was visible deeper in the seismic data (both 2-D and 3-D).
 Within the Messinian section, three prominent groups of low frequency reflectors, separated by acoustic blanking, are evident across the data. In some areas, amplitudes of these reflectors become significantly brighter than their surroundings. These were interpreted as bright spots, possibly indicating the presence of gas (Figure 1b). The three groups are somewhat continuous and become blanked towards the section directly above the largest group of pockmarks and the pre-Messinian horst. Some of the reflectors exhibit areas of low amplitudes that may result from poor illumination of unbalanced seismic imaging where the sedimentary unit above scatters most of the energy. However, the 3-D structure within these low amplitude areas appears as large (<1 km) circular depressions interpreted as dim spots. As such, they are possibly correlated with hydrocarbons that reduce the contrast in acoustic impedance between the reservoir and the surrounding stratigraphy. Between the groups, data is chaotic and discontinuous as evident by chaos and variance seismic attributes.
 The top-Messinian surface is composed of a series of short wavelength folds, extensional and lateral faults as well as several channel incisions (Figure 3). In particular, this surface is rich with gas-escape features. A set of seven concentric pockmarks (possibly mud volcanoes) was identified. These elevated structures rise ∼50 m above their surrounding top Messinian surface, mainly along a NW trend and can be followed in the data below the top Messinian reflector (Figure 3). Their diameter ranges around a kilometer.
4. Climatic Effect of the Messinian Salinity Crisis
 Data presented here show a variety of gas escape features throughout the pre-Messinian and Messinian section in the deep eastern Mediterranean basin. While it is generally accepted that environmental changes did not contribute to theonset of the Messinian Salinity Crisis [Suc and Bessais, 1990; Clauzon et al., 1996; Krijgsman et al., 1999; Fauquette et al., 2006], the timing of gas/fluid emission and migration, and their contribution to the ocean/atmosphere system and hence, the climatic effect of the Messinian Salinity Crisis, are yet to be addressed.
 The presence of pockmarks at the base Messinian (N reflector, Figure 2) indicates that emission began, or was occurring on the seafloor at the time of desiccation. Water depth at the time of pockmark formation is unknown, as is the flux of emitted gas/fluid. During the Messinian, Mediterranean sea levels dropped by a kilometer or more [e.g., Clauzon et al., 1996; Gargani and Rigollet, 2007; Garcia-Castellanos et al., 2009]. Calculating for massive subsidence that followed the end of the Messinian [e.g., Ben-Gai et al., 2005], the study area presented here would have been exposed or in very shallow water (<75 m). Either way, rapid unloading of the water column would decrease compression of the subsurface and of the effective stress of pores within the sediment. As a result, it is conceivable that there was an increase in gas/fluid escape into the shallow water column, and hence, into the atmosphere.
 As desiccation progressed, changes in brine salinity (resulting from temporal sea level rises [Gargani and Rigollet, 2007]) led to deposition of evaporites of different composition. Oxygen isotope records of benthic foraminifera [Hodell et al., 2001; Vidal et al., 2002] point to a clear global driver to European climate change at this time. Global ice volume expansion ended precisely at 5.55 Ma coinciding with the initiation of the warmer and more humid Lago Mare stage of the Messinian Salinity Crisis [Willett et al., 2006; Murphy et al., 2009]. Enhanced precipitation across the Mediterranean margins, led to an increase in runoff into the desiccating basin from around the area, e.g., the Alps [Fauquette et al., 1999; Fortelius et al., 2002; Willett, 2010] and the Trodos [Rouchy et al., 2001]. Interpretation presented here in the context of Messinian climatic fluctuations, supports the possibility that this terrigenous supply is responsible for the three internal layers visible in the Messinian section.
 While Messinian sediments (evaporites and terrigenous) could have sealed emission, data show evidence for continuous migration of gas/fluid through the section and its discharge at the then-seafloor. If gas/fluid started to migrate only after sediments were deposited and faulted, thus providing clear conduits for gas/fluid escape, then acoustic blanking should have been more concentrated around lineaments and less diffusive. In addition, pockmarks within the Messinian section strengthen the argument for concurrent emission from below during this period.
 Seafloor escape features are last identified on the top Messinian horizon (Figure 3). A rapid rise in sea levels refilled the Mediterranean within a few years [Blanc, 2002; Garcia-Castellanos et al., 2009]. Data show a marked change in seismic characteristics above the M reflector, with horizons becoming continuous and with higher frequencies (Figure 1b). Within these sediments, evidence for gas/fluid migration (conduits, chimneys, etc.) is present without any escape-features. Furthermore, lack of pockmarks or mud volcanoes on the present-day seafloor in the study area indicates that the system described here is sealed or no longer active. Thus, we conclude that the gas/fluid emission to the Messinian water column and the atmosphere was blocked as sea level rose and marine sediments refilled the basin.
 A number of features discovered in 2-D and 3-D seismic data provide evidence for continuous gas emission throughout the Messinian. These include large areas of acoustic blanking through the Cenozoic onlapped and toplapped by several bright spots; chimneys that lead up to base Messinian; giant pockmarks and pockmark fields on base Messinian N unit; blanked volumes, bright spots and pockmarks within the Messinian sediments; very high reflectivity of the top Messinian surface and the presence of large pockmarks (possibly mud volcanoes). Minor chaotic acoustic blanking and lack of clear gas chimneys in the overlying Plio-Pleistocene sedimentary cover coupled with the lack of pockmarks on the present-day seafloor implies that emission ceased or is contained.
 Pockmarks at the base Messinian, throughout the Messinian section and at the top of the Messinian indicate that gas emission began in the Cenozoic and continued throughout the near-desiccation of the Mediterranean despite massive deposition of evaporites thought to act as a good caprock. We assume that since the lowstand affected the entire Mediterranean, gas escape occurred across the basin. As long as sea levels were low, there was massive gas escape into the shallow sea and atmosphere. We suggest that this probably resulted in the increase in temperature, humidity and runoff at around 5.55 Ma. While the cause for sea level drops during the Messinian is tectonic, the major emission identified here indicate an indirect cause to negative climatic feedback during themid-Messinian climatic shift, which may have led to the Lago Mare stage.
 The authors would like to thank ILDC, Modi'in Energy and GGR for their permission to show the seismic data, and Schlumberger for granting academic licenses for Petrel Seismic Interpretation software. We thank the editor of GRL as well as J.P. Suc and an anonymous reviewer for their invaluable remarks and comments.
 The Editor thanks Jean-Pierre Suc and an anonymous reviewer for their assistance in evaluating this paper.