The presence of sandstone intrusions of seismic scale indicates the existence of a parent sand body and therefore that a flow of sand from this sand body took place over several tens or hundreds of meters (Hurst et al. 2003; Cartwright et al. 2008). As mentioned earlier, the conical and saucer-shaped sand intrusions are aligned with the northwest edge of Lobe 1, which is the highest point of this lobe (Fig. 4). We therefore suggest that vertical migration of fluids through the argillaceous sediments transported sand from the silt-sand pinchout of the lobe toward the paleosurface. It means that the sand in the lobe fringe is presumably ‘clean’, in contrast to lobe fringes that are rich in hybrid beds (e.g., Ito 2008; Hodgson 2009). The seismic acquisition does not allow vertical reflectors to be imaged, and so, it is not possible to see whether there are conduits that could connect the cone and saucer-shaped bodies to the lobe. Therefore, we suggest two possible connections: (i) Pipes or columnar intrusions (e.g., Huuse et al. 2004; Chan et al. 2007) and (ii) Planar dykes as wing structures developed on the margins of depositional sand bodies (e.g., Lonergan & Cartwright 1999; Jackson 2007). Large-scale sand injection requires high volumes of sand and their transport through the argillaceous cover, which necessitates more than 50 % of fluids to carry the sediments (Maltman 1994; Hurst et al. 2003; Duranti 2007; Cartwright et al. 2008).
The forced ascending intrusion of sand into the sedimentary column is usually attributed to hydraulic fracturing (Lonergan et al. 2001; Hurst et al. 2011; Mourgues et al. 2012). Hydraulic fracturing may occur in response to an increase in pore pressure if the fluid pressure (Pf) in the sand exceeds the minimum principal stress (S3) plus the tensile strength (T) of the host rock (Price & Cosgrove 1990; Cosgrove 2001). Both tensile and shear hydrofracturing of host strata can occur (Hurst et al. 2011). Hydraulic fractures propagate parallel to the maximum compressive stress direction (S1) and perpendicular to the minimum compressive stress direction (S3) (Anderson 1951; Delaney et al. 1986). This mode of fracturing occurs if Pf > S3 + T and S3 + T < S2 < S1 (Hurst et al. 2011). The creation of this space, which was necessary for the emplacement of the sand intrusions studied in the study area, caused pronounced uplifts in the paleosurface T (Figs 4-7). These uplifts are commonly called forced folds or domal forced folds (Cosgrove & Hillier 2000; Hansen & Cartwright 2006; Cartwright et al. 2008).
Process of formation of the sand injectites
The forceful intrusion of remobilized clastic sediment form by injection of fluidized sand (Duranti & Hurst 2004; Ross et al. 2011) from an overpressured sand unit into hydraulically fractured low-permeability sediments (Cosgrove 2001; Jolly & Lonergan 2002). Overpressures in deep-sea environments can occur for a large range of reasons but are mainly due to the disequilibrium compaction and hydrocarbon (gas) generation and the lateral transfer of pressure (Osborne & Swarbrick 1997; Swarbrick & Osborne 1998; Grauls 1999; Swarbrick et al. 2002). For the injection and fluidization of unconsolidated sand to occur, a trigger mechanism is commonly required (Jolly & Lonergan 2002; Oliveira et al. 2009) such as (i) earthquake (e.g., Obermeier 1996, 1998; Boehm & Moore 2002; Huuse & Mickelson 2004; Levi et al. 2011), (ii) tectonic stress (e.g., Vitanage 1954; Harms 1965; Scholz et al. 2009), (iii) localized excess pore fluid pressures generated by deposition-related processes (e.g., Truswell 1972; Taylor 1982; Rijsdijk et al. 1999; Rowe et al. 2002; Callot et al. 2008), and (iv) the influx of an overpressured fluid from deeper within the basin into a shallow sand body (e.g., Jenkins 1930; Brooke et al. 1995; Jolly & Lonergan 2002; Molyneux et al. 2002; Duranti & Mazzini 2005; Jonk et al. 2005; Andresen et al. 2009). A less common trigger process cited in the literature is the mechanical failure of hydrocarbon reservoirs in the shallow subsurface caused by the buoyancy effect of hydrocarbons (Sales 1993; Jonk 2010).
In the study area, the onset of the maturation and migration of hydrocarbons (oil and gas) took place during the late Early Miocene (approximately 18 Ma), while filling of the Miocene reservoirs occurred during the Late Miocene to Pleistocene (approximately 5 Ma to the present day; based on a 3D basin modeling study carried out by Total, 2000), which coincides with the timing of the sand injection. Consequently, the Upper Miocene Lobe 1 was able to accumulate hydrocarbons rapidly after its burial. This hydrocarbon column was trapped in the northwest portion of Lobe 1 following local subsidence beneath the axis of the lobe, which caused its topographic inversion (distal portion higher than the central portion) (Figs 4 and 8). In view of the tectono-sedimentary analysis and on the hydrocarbon migration and filling history, the most likely process for the generation of overpressures in the shallow submarine fan of the study area, that is, in the parent unit, is attributed to the hydrocarbon buoyancy associated with the lateral transfer of (overpressured) fluids. The lateral transfer is a mechanism that can locally further enhance pore pressures at structural crests due to the transmission of pore fluid overpressure by water flow along laterally inclined sand reservoirs encased in a seal lithology (Mann & Mackenzie 1990; Yardley & Swarbrick 2000; Mourgues et al. 2011). The buoyancy effect is the pressure difference between two immiscible phases, that is, formation water and hydrocarbons, generated by density contrast (Swarbrick et al. 2002). The overpressure due to hydrocarbon buoyancy is in addition to the overpressure induced by the lateral transfer of fluids (Jonk 2010).
The seismic scale sand injectites studied are evidence of hydraulic fracturing of the cover rocks; it means that the leaks of fluids from the Miocene reservoirs took place under a seal of ‘hydraulic’ type (poorly permeable), as defined by Watts (1987). Cover rocks called ‘hydraulic’ have a very high capillary entry pressure Pe (very fine-grained clays, anhydrite, halite, etc.) such that capillary leaks of hydrocarbons cannot occur. Capillary leaking is the normal mode of seal failure under hydrostatic conditions or moderate overpressures (Clayton & Hay 1994). Rupture by fracturing generally takes place only in environments under high overpressures but is possible at shallow depths (Fig. 11). The conical and saucer-shaped intrusions in the study area were very probably initiated from the pinchout northwest of Lobe 1, at a depth of 160 m (assumptions of decompacted thickness) beneath an effective argillaceous cover. At this burial depth, the overpressure in the parent sand body could not have been induced by the disequilibrium compaction. We therefore suggest that the overpressure began at shallow depth and mainly under the effect of buoyancy pressure Phc caused by a column of hydrocarbons with a minor contribution from lateral transfer (Fig. 11) (Sales 1993; Grauls 1997, 1999; Osborne & Swarbrick 1997). The validity of this hypothesis is supported by a series of calculations based on theoretical equations.
Figure 11. Schematic diagrams of Pressure-Depth illustrating two initiations of hydraulic fractures in function of the pressure regime in traps (Modified from D. Grauls communication). Note that the capillary entry pressure Pe is higher than the minimum compressive stress S3 in both cases.
Download figure to PowerPoint
Based on Pascal's principle, that is, the principle of transmission of fluid pressure, we calculated the necessary thickness of a hydrocarbon column to fracture its seal below 160 m of marine sediments, that is, the estimated burial of the source body of sand (NW margin of Lobe 1) during the formation of the injectites and under a 1000-m column of seawater. In our study area, the hydrocarbons may consist of oil and gas during sand injection. Therefore, the results of calculations will be given for oil and gas to define a range of thickness from a maximum value (only oil) to a minimum value (only gas). Pascal's principle is defined in the following equation:
where Δp is the pressure difference (Pa); ρ is the density of the fluid (kg m−3); g is the gravitational acceleration (m s−2); Δh is the height of fluid above the point of measurement (m).
The major horizontal stress S3 is determined graphically on Fig. 11A by the following equations:
where ρw is the density of the seawater (1030 kg m−3); Zsf the depth of the sea bottom below sea level (m); K0 is the neutral earth pressure coefficient or the ratio of the horizontal stress to the vertical stress (K0 ≈ 0.85); ρlitho is the bulk density of sediment (1800 kg m−3); b is the height of the overburden (Z – Zsf); ρhc is the density of the hydrocarbon (ρgas ≈ 200 kg m−3, and ρoil ≈ 800 kg m−3); and h is the thickness of the hydrocarbon column. All density values are extracted from Cathles et al. (2010). From equations 2 and 3, we deduced the value of the thickness of the hydrocarbon column (h) during the fracture initiated in the overburden:
In conclusion, at the time when the studied sand intrusions were being formed, Lobe 1 was buried under 160 m of argillaceous sediment. These sediments, which had a high capillary entry pressure (Pe), could have been hydraulically fractured by the buoyancy effect (Phc) produced by (i) a column of oil at least 350 m high, (ii) a column of gas at least 100 m high, and (iii) a column of oil and gas ranging from 100 m to 350 m high, trapped in the northwest margin of Lobe 1. The maximum case of hydrocarbon filling in the lobe is geometrically estimated at up to about 100 m during sand injection. This estimate corresponds to the actual thickness at the time when Horizon T was deposited (Fig. 8), or half the thickness currently observed on seismic. Therefore, unless we consider this maximum case and filling by gas only, the values of hydrocarbon thickness obtained seem too high to consider the buoyancy effect as the trigger mechanism of sand injection. In addition, the presence of many sandstone intrusions means that the fracture pressure (Pf > σ3 + T) was reached simultaneously in many points of the northwest margin of the lobe.
This situation is thought to be conceivable only if a sudden trigger event occurred just before sand injection. As a result, we estimate that the simpler trigger event of sand injection in our study area is the activity of a nearby fault in relation to diapiric movements, which can have induced seismic shaking (e.g., Boehm & Moore 2002) or allowed the rapid flow of overpressured fluid into Lobe 1 from a deeper geobody (e.g., Sibson 1981; Grauls & Baleix 1994).
Propagation mechanisms for sand injectites
The initiation of fracturing creates a hydraulic gradient between the tip of the fracture and the source body of sand, which can entrain particles of sand if the flow velocity (υ) exceeds the fluidization velocity (υfl) of the injected granular material (see fig. 11 in Vigorito & Hurst 2010). The low density and viscosity of gas are not favorable for fluidization; therefore, it may be possible that the driving force for the fluidization is the movement of the aqueous fluids accompanied by significant quantities of dissolved hydrocarbon gas (Jonk 2010). The role of hydrocarbon gas as a support to sand fluidization has been widely mentioned in the literature (Brooke et al. 1995; Hubbard et al. 2007).
As mentioned earlier, hydraulic fracturing follows planes perpendicular to the minor compressive stress plus the tensile strength (Pp = S3 + T). As a result, injectites usually have simple geometries like dykes (cutting across the stratigraphy) or sills (parallel to the stratigraphy), which may locally be reoriented by planes of weakness (Delaney et al. 1986; Grauls 1999). Conical and saucer-shaped geometries are less common and known only at large scale (Huuse & Mickelson 2004; Shoulders & Cartwright 2004; Huuse et al. 2007; Shoulders et al. 2007). They result from the interaction between the propagation of a fracture and the proximity of a free surface, which is deformed. These conical and saucer-shaped structures are also known from magmatic intrusions (Hansen & Cartwright 2006) or from fluid-escape structures of pockmark type (Gay et al. 2006b, 2012). The formation of the cones and the controlling parameters had recently been studied by Cartwright et al. (2008), and Mourgues et al. (2012). Cartwright et al. (2008) have proposed a simple model of apical cones formation and suggested that a small laccolith of sand forming at the top of a feeder dyke induced the rotation of σ1, thereby allowing the development of low-angle dykes. Mourgues et al. (2012) have used analog and numerical modeling to show in 2D that the formation of vertical fracturing (from a body of sand) requires a sufficiently strong effective vertical stress (fairly thick cover and low level of overpressure in the cover). At a critical depth, vertical propagation stops and the fracture splits into two dilatant branches, forming a ‘V’ shape (Mourgues et al. 2012). These two dilatant branches are taken over by shear zones which extend to the surface and strongly accentuate the amplitude of the forced folds (Fig. 5B). The depth at which the cones initiate is mainly controlled by the mechanical parameters of the host rocks, for example, their cohesion, and the level of fluid overpressure. Mourgues et al. (2012) have also demonstrated that the presence of distributed overpressures in the host rocks (related to under-compaction phenomena, for example) promotes the formation of cones at greater depth. In addition, the same authors have shown that the pressure field induced by the diffusion of overpressures around the sand body was the source of a stress rotation (Mourgues & Cobbold 2003), which also favored the formation of inclined fractures. The geometrical configuration of conical and saucer-shaped intrusions was previously discussed by Pollard & Holzhausen (1979), for igneous intrusions. These authors predicted that once the dimensions of the sill reached a critical value, the fracture interacted with the free surface, and the sill turned upward toward the surface. The formation of saucer-shaped sandstone intrusions may be explained by a greater competence contrast at a boundary (Cartwright et al. 2008). It is consistent with our own observations regarding the laccoliths of saucer-shaped intrusions, which are always located at the same stratigraphic level in the affected interval of sediments (Fig. 5C).
At the top of the conical injectite branches, smaller cones may form (Fig. 5A,B). Their passage deforms and remobilizes the sediments of the surrounding rocks, and the overlying domal forced fold is collapsed. Consequently, they are not interpreted as sand intrusions (e.g., Shoulders et al. 2007; Cartwright et al. 2008) but as deformation cones resulting from the migration of fluids (e.g., Gay et al. 2012). Thus, the sand intrusions are evidence of a propagation of fluids localized in the host rocks, whereas at shallower depths, deformation cones may form as a continuation of the sand cones, indicating that the migration is transforming into a flow of fluid distributed throughout the unconsolidated sediments (Gay et al. 2012). The expulsion cones of fluid formed at the top of the injectite branches lead to a collapse of the sediments up to the paleosurface (Horizon T), which indicates that they are coeval with emplacement of the injectites. Lastly, fluid-expulsion chimneys located above some of the sand injectites revealed a new phase of leakage after burial of Horizon T (Fig. 7). These hydrocarbon (gas) leaks are evidenced by the presence of pockmarks above Channel 2 (Figs 3 and 4). We therefore suggest that hydrocarbons continued to migrate within the reservoirs and along the injectites after additional burial. Some authors have already shown from fluid inclusion and stable isotope data that large-scale sand injectites act as long-term fluid conduits (Hurst et al. 2003; Jonk et al. 2003, 2005).