Page 12 - Antonellini_2013

12 M. Antonellini et al. / Marine and Petroleum Geology xxx (2013) 1e16

Figure 12. Single well steady-state draw-down numerical experiments with MODFLOW 2005 on a model with constant head (20 m) on the boundaries. (a) Draw-down shownby
3
the equipotential lines around a single producer well ( 20 m /day) in homogenous grainstone. (b) Single producer well draw-down in the model of San Vito Lo Capo containing the
strike-slip faults. Note the refraction of the equipotential lines at the fault lines. (c) Single producer well draw-down where the well is located in a ZB. Note the strong reduction in
area of the cone with respect to (a) and (b). (d) Same situation as in (c) but showing the K x hydraulic conductivities (legend in Fig. 9b). (e) Single producer well draw-down where
the well is located at a fault junction (DF and ZB). (f) Single producer well draw-down where the well is located in a compartment in between zones of compactive shear bands (ZB).
with the exception for the draw-down of the wells located within computed from the deterministic model decreases to 79% and the
the structures (see comparison in Table 6 and Fig. 14). The average K y to 93% of the K in the un-deformed matrix. The K x computed
draw-down from a shear structure in a DFN model is more than 50% from the DFN model decreases to 90% and the K y to 89% of the K in
less than in a deterministic model (Table 6). The structures (and the un-deformed matrix. The K z is basically unaffected by the SSRF
specifically the ZB and DF) in the DFN model (Fig. 14) tend to be in both models at this scale, because the faults are vertical. It is
thinner than in the deterministic model (Fig. 9). This is due to the also interesting to note that the up-scaled hydraulic conductivities
inability of the DFN model to mimic the phenomenon of strain obtained from the deterministic and the DFN models are similar;
localization causing the development of ZB. they differ at most by 15%. In this situation, when up-scaling to a
large cell size for fluid flow simulation, the use of a DFN-derived
4.3. Up-scaling up-scaled hydraulic conductivity is as good as the deterministic
analogue-derived hydraulic conductivity. At this scale, however,
Table 7 shows the results of up-scaling the hydraulic conduc- the effect of the SSRF is almost negligible and this is in accordance
tivity to a cell size including the whole outcrop (45 32 1 m); with what reported in the literature. Walsh et al. (1998) examined
this has been done by using the program MODPATH (Pollock, the effects of SSRF on the bulk permeabilities of reservoir se-
1994) and a lagrangian approach based on tracking fluid particle quences at a 3 km 3 km scale and concluded that the signif-
velocities and inverting for hydraulic conductivities under a uni- icance of these structures on fluid flow is strongly dependent on
form gradient in the x, y, and z directions both in the deterministic the fault transmissivities that need to be very low to significantly
and DFN models. It is interesting to note that the effect of the SSRF affect the hydraulic properties of the reservoir. These conclusions,
in view of our experiments, seem to apply to large-scale fluid flow
on K x and K y within this large block is small (Table 7). The K x
Please cite this article in press as: Antonellini, M., et al., Fluid flow numerical experiments of faulted porous carbonates, Northwest Sicily (Italy),
Marine and Petroleum Geology (2013), http://dx.doi.org/10.1016/j.marpetgeo.2013.12.003