Straining at the Leash: Understanding the Full Potential of the Deep-Water, Subsalt Mad Dog Field, from Appraisal through Early Production
Published:December 01, 2012
- PDF LinkChapter PDF
Christopher Walker, Paul Belvedere, Jennifer Petersen, Shalina Warrior, Andrew Cunningham, George Clemenceau, Christina Huenink, Robert Meltz, 2012. "Straining at the Leash: Understanding the Full Potential of the Deep-Water, Subsalt Mad Dog Field, from Appraisal through Early Production", New Understanding of the Petroleum Systems of Continental Margins of the World, Norman C. Rosen, Paul Weimer, Sylvia Maria Coutes dos Anjos, Sverre Henrickson, Edmundo Marques, Mike Mayall, Richard Fillon, Tony D’Agostino, Art Saller, Kurt Campion, Tim Huang, Rick Sarg, Fred Schroeder
Download citation file:
Mad Dog is a giant, subsalt, deep-water oil field that will be producing hydrocarbons for the co-owners BP, BHP Billiton Petroleum, and Chevron in the Gulf of Mexico for many years to come. The field was discovered in 1998 by the GC0826#1 well and sidetracks. Four appraisal wells proved up a material resource but also showed evidence of compartmentalization and imperfect subsalt seismic imaging. A spar development was selected and first oil was achieved in 2005. Further appraisal drilling continued from the spar rig and mobile offshore drilling units, ending with the Mad Dog North appraisal program in December 2011. The original spar rig was lost during Hurricane Ike in September, 2008, and a rig replacement project is currently underway. In 2009, the Mad Dog South appraisal well proved up large volumes of hydrocarbons beyond the drilling radius of the original spar rig, necessitating the construction of a second production facility.
Mad Dog hydrocarbons are predominantly contained in deep-water turbidite sandstones of early Miocene age that can be correlated over hundreds of square miles. The turbidites are interpreted as a series of individual lobes in a submarine fan complex that was deposited in an unconfined basin floor environment. Deformation of the rocks commenced shortly after deposition and continued through the Plio-Pleistocene boundary. The reservoir is divided into several large compartments that are identified by differences in pressure, fluid composition, and oil-water contacts. These large compartments are interpreted to be bounded by seismically visible faults. Smaller seismically visible and subseismic faults act as baffles to fluid flow in the field, and have been identified through logs, dynamic data, and reservoir simulation.
The phased development of the Mad Dog Field has enabled BP and co-owners to mitigate project risk during full-field development. By starting small, developing the known hydrocarbons, investing in technology to improve the imaging of the field, and continuing appraisal drilling, the team was able to evaluate the resources while simultaneously unlocking their value. The integration of dynamic production data with the improved seismic image and appraisal well results has allowed the second phase of the development to proceed with significantly reduced subsurface uncertainty. This has enabled the team to unleash the full potential of the Mad Dog field to be a large deep-water producer for the next 40 years.
The Mad Dog Field is located in the Gulf of Mexico southern Green Canyon protraction area, approximately 140 miles (225 kilometers) south of New Orleans, Louisiana (Fig. 1). The field straddles the Sigsbee Escarpment, a bathymetric feature that represents the southern limit of shallow allochthonous salt, where water depths change from 4,500 to >7,000 feet (1,370 – 2,130 m) (Fig. 2). The history of exploration activity in the area began in the mid-1980s with lease acquisition along the lower Atwater fold belt trend. The first well to be drilled in the trend was BP's Neptune discovery (AT0575 #1) in 1995 (Fig. 3). The Mad Dog discovery well (GC0826 #1_ST01) was drilled in May, 1998, by Amoco, along with co-owners BP, BHP Billiton Petroleum, and Unocal (Fig. 4).
Four additional appraisal wells and sidetracks have been drilled to test the east and west flanks of the structure prior to project sanction in December, 2001. BP currently operates the field with a 60.5% working interest (WI) in the unit which presently includes Green Canyon Blocks 738, 781, 782, 825, and 826. BHP Billiton Petroleum and Chevron are co-owners, with 23.9% and 15.6% WI respectively. Mad Dog was sanctioned as a 14-well, single spar development having facility capacity of 80 mmbo/d, 60 mmscf/d, and 50 mbw/d, based on oil-in-place estimates of approximately 450 million barrels (BHP Billiton Petroleum, 2002).
The estimates of field size have octupled in recent years due to three key appraisal wells. In 2007, the GC0781 #11 well extended the oil column on the west flank by > 1,000 feet (305 m); in 2009, the BHP Billiton Petroleum-operated Mad Dog South appraisal well (GC0826 #5) found a > 4,000 foot (1,220 m) oil column in the south section of the field. Finally, in 2011 the BHP Billiton Petroleum-operated Mad Dog North appraisal well (GC0738 #1 and sidetracks) found over 166 feet (51 m) of net pay and proved up an oil column on the north flank of the structure. These appraisal results have increased estimates of oil in-place volumes to over 4 billion barrels (Brenner, 2011). Based upon these resource additions, a second phase of development has been initiated which will involve the design and construction of a new production facility.
Hydrocarbons have been discovered at multiple stratigraphic intervals over the Mad Dog structure (Fig. 5), but the bulk of the resource is contained within the early Miocene sands that have been penetrated by 29 wells across the development area. These sands have been deposited in a basin floor fan setting and are laterally continuous over a distance of > 60 miles (~100 km; Apps et al., 2002). Correlations between well logs throughout the field also allow recognition of multiple faults, with missing sections ranging from 15 to 1,500 feet (5 – 460 m). The largest of these faults compartmentalize the field into several fault blocks that are not in communication over a production timescale.
The history of development in the Mad Dog Field has been intimately tied to advances made in seismic data acquisition and processing. The field is difficult to image seismically for several reasons; the water is deep, from 4000 to 7000 feet (1,200 to 2,100 m); there is a 2000 foot (610 m) high sea floor bathymetric feature called the Sigsbee Escarpment; the depth to the reservoir is greater than 19,000 feet (5,800 m); and most importantly, the field lies beneath a thick layer of allochthonous salt. The salt has a rugose top (Fig. 6), large variations in thickness, steeply dipping edges, overhangs, and a rugose base (pink in the Fig. 5 cross-sections). These irregularities in salt geometry create illumination holes where reflected seismic energy cannot be captured by receivers at the surface.
Mad Dog discovery
The discovery well (GC0826 #1_ST01) was drilled in 1998 based on imaging from a 1996 vintage 3D speculative seismic survey (Fig. 7A). This data showed good seismic reflectors outboard of salt to the south (right side of Fig. 7A), and fair reflectors beneath the thin salt to the north (left side of Fig. 7A). However, beneath the main body of the salt, over the crest of the deep structure, reflectivity was poor to non-existent. The discovery well was drilled seaward of the Sigsbee Escarpment, down through an area of good reflectivity before penetrating the reservoir section. Over 215 feet (66 m) of measured depth net pay was encountered in the early Miocene sands.
Mad Dog early appraisal
Appraisal wells confirmed the presence of hydrocarbons on the east side of block GC0782 in 1999 (GC0782 #1), on block GC0783 in 2000 (GC0783 #1) and on the west side of block GC0782 in 2001 (GC0782 #4). These well results allowed the subsurface team to conclude that an economically developable resource was present in the area, and in December, 2001, the Mad Dog project was sanctioned. The results of the appraisal wells also demonstrated the limitations of the seismic data set and highlighted the need for an improved seismic image to define more accurately the structure and efficiently develop the field resources. Sidetracks encountered additional fault blocks that could not be seen in the seismic data and were not charged with hydrocarbon. This introduced the risk that the crest of the structure might resemble “shattered glass;” a series of small fault blocks that would not support cost-effective wells. Also, significant amounts of mobile tar deposits were encountered in the overburden section, causing drilling problems.
Advances in seismic processing allowed the original seismic data to be merged with a new narrow azimuth towed streamer (NATS) data that was acquired in 2000 along a different orientation (Michell et al., 2004). This dual-azimuth data set also incorporated new subsurface velocity information from the wells and resulted in improved image quality (Fig. 7B). North flank reflectors were more coherent and some reflectivity was achieved from the crest of the structure beneath the steeply dipping salt wall. However, when the appraisal and development wells for the new facility were drilled (2001–2005), further subsurface surprises were encountered, including the discovery of the western-graben bounding fault having over 1000 feet (305 m) of missing section, by the GC0782 #4_ST01. In addition to finding the largest fault in the field, this well also proved the presence of hydrocarbons on the west side of the structure. The increase in the resource base associated with this well gave BP management the confidence needed to approve a major new seismic acquisition program across the Mad Dog Field, designed to improve further the seismic image and reduce the oil in place uncertainty.
Southwest ridge and Paleogene appraisal
First oil was achieved from the Mad Dog spar on January 13th, 2005. Around this time, the Southwest ridge appraisal wells were drilled in the field by Unocal (GC0825 #1; GC0826 #4). Results from these wells extended the known hydrocarbon column 800 feet (245 m) deeper on the west side of the structure. An updip sidetrack found another wet compartment (GC0826 #4_ST02) adding further complexity to the subsurface picture. Additional problems were encountered when a planned deep test well (GC0826 #2) encountered several 20 foot (6 m) thick intervals of mobile tar in 100 feet (30 m) of section that flowed up the wellbore and proved to be impassable (Romo et al., 2007). A second deep test at a different location (GC0826 #3) effectively twinned the discovery well without problems and confirmed an oil-water contact (OWC) in the early Miocene reservoir in the east compartment at 20,640 feet (6,291 m) true vertical depth subsea (TVDSS). This well also proved an accumulation of high-viscosity hydrocarbons in low permeability Paleogene-aged sands.
Wide azimuth towed streamer acquisition
The game changing technological development for imaging the Mad Dog Field was wide-azimuth towed streamer acquisition (WATS) (for details see Michell et al., 2007). This method of shooting the same source lines repeatedly as the receiver boats sail farther and farther away allowed the collection of longer offset data, which increases the chance of recording reflections that have taken complicated or unusual paths through the overburden. This made it uniquely wellsuited for deployment in subsalt areas having complex salt geometries such as the Mad Dog salt body. The resulting image (Fig. 7C) and corresponding map (Fig. 8) revealed a step-change in the reflectivity over the crest of the structure and in the number of faults that can be mapped. This increased information revealed for the first time the size of the potential prize and allowed more challenging wells to be planned.
Mad Dog West appraisal
Pressure transient analysis of data from the A5 producer (GC0782 #9_ST01) was interpreted to show that the well is was connected to a large volume of fluid, in a long, narrow corridor. With the previous poor quality subsalt image, the team was not able to place a second producer on the west side of the field confidently; however, the map generated from the WATS seismic data (Fig. 8) revealed a large untested fault block. The GC0781 #11 pilot hole was drilled to test an expected OWC just below the lowest known oil in the Southwest ridge appraisal wells. This target depth was reached, but the sands were still found to be full of oil. The team was then able to chase the OWC downdip, using real-time biostratigraphy, real-time azimuthal density and gamma ray information, and other logging-while-drilling (LWD) data to geo-steer a high-angle well, subparallel to bedding. The OWC was finally tagged at 23,330 feet (7,111 m) TVDSS and the stripped-down borehole apparatus was cemented in the hole. The well was sidetracked up dip to land the producer at the crest of the fault block, within 275 feet (84 m) of net measured depth (MD), early Miocene reservoir sand.
Mad Dog South
The deeper OWC on the west side tagged by GC0781 #11 greatly increased the hydrocarbon resource estimate beyond the designed capacity of the Mad Dog spar. Preliminary work began to evaluate options for extracting the additional hydrocarbons in a timely fashion, including facility de-bottlenecking and the implementation of a subsea tieback development. The team concluded that, in order to optimize recovery, the geographic and volumetric extent of the resources needed to be fully appraised. Against this backdrop, the Mad Dog South well was planned. The major concern with the south flank of the field was the 2005 wet penetration updip (GC0826 #4_ST02), which was thought to reduce the probability of any hydrocarbon accumulation in the south flank and inhibited further appraisal. By 2009, however, improvements in seismic processing were able to extract more information out of the data. The application of Vertical Transverse Isotropy (VTI) by CGGVeritas in processing the data revealed an intricate web of faults over the crest of the south flank (Bowling et al., 2010). By mapping these faults (seen in Fig. 8, based on a later 2010 iteration using tilted transverse isotropy (TTI)) the team could interpret that the updip wet penetration (GC0826 #4_ST02) was contained within a separate fault block. This explained the elevated pressures and salinities in the block, suggesting that it was isolated from charge and from the primary aquifer, and opened up a large compartment on the south flank of the field as an appraisal target.
The Mad Dog South appraisal well was targeted to test a similar column height to that observed in the east, ~ 1240 feet (380 m), which would suggest an OWC midway between the contacts seen on the east and west flanks of the structure. The well (GC0826 #5) was drilled by BHP Billiton Petroleum in May 2009, as BP temporarily ceded operatorship of the GC0826 block to BHP Billiton Petroleum, which had a rig slot available to drill the well. The well exceeded predrill upside expectations and discovered 280 feet (85 m) of net MD pay in early Miocene sands that were full to base. The team implemented the upside contingency plan and sidetracked south (GC0826 #5_ST01), down-dip as far as mechanically possible to look for an OWC that would constrain the hydrocarbon footprint to the south. The rig lost communication with the BHA at 23,464 feet (7,151 m) TVDSS. At that depth, the reservoir was still full to base, extending the lowest known oil ~130 feet (40 m) deeper than the west flank OWC. This result showed a deeper contact and larger column height in the south than elsewhere in the field and led to another large increase in the oil in-place resource estimate to more than 7 times the initial estimates.
Mad Dog Phase 2 development
The Mad Dog west and south appraisal results showed that the in-place Mad Dog resources covered a geographic footprint that extended beyond the drilling capabilities of the original rig and that the hydrocarbons could not be produced through the existing Mad Dog spar within the design life of the facility. The Mad Dog Phase 2 project was initiated to ensure the development of the full-field resource base in the safest and most efficient way possible. After a concept selection process, an all-subsea development was recommended, which is planned to be tied back to a second spar host facility having 130,000 BOPD capacity. The size of the south and west hydrocarbon footprints relative to their respective aquifers was highlighted by preliminary uncertainty studies as a potential problem for project viability. The decision was made to implement a water-flood project having flank and mid-dip injectors to increase recovery. The water-flood project is also planned to include BP's LoSal ™ enhanced oil recovery technology to maximize recovery from the field. The development will be operated by BP having a 60.5% WI, in partnership with BHP Billiton Petroleum (23.9 % WI) and Chevron (15.6% WI). The project is expected to be sanctioned by BP, BHP Billiton, and Chevron in 2013, and pre-drill activity will commence in 2014. The new facility will be called “Big Dog Spar.”
Mad Dog North
The most recent drilling activity in the field took place in 2011, when BP temporarily ceded operatorship of Block 738 to BHP Billiton Petroleum in order to test the northern flank of the field. The Mad Dog North appraisal well (GC0738 #1) found 166 feet (50.6 m) of net pay. A bypass well (GC0738 #1_BP01) was drilled around 70 feet (21 m) from the original well and successfully collected a conventional core throughout the entire reservoir section for the first time in the history of the field, 13 years after the field discovery. This core more than doubled the amount of whole core collected from the Mad Dog field, and information derived from it will help to refine the design parameters for the Phase 2 development program and the placement of remaining wells to be drilled from the Mad Dog spar “A.” The north drilling program also included north- and east-directed sidetracks to delimit the hydrocarbon footprint and OWC in the northern part of the field. Two of these well bores had to be bypassed after encountering mobile tar at middle Miocene stratigraphic levels. The discovery of mid-Miocene tar may eventually lead to a revision of the pre-existing tar model (Romo et al., 2007).
Mad Dog rig replacement
Drilling operations on the Mad Dog spar were terminated on September 9th, 2008, as the facility was evacuated ahead of the approaching Hurricane Ike. When the platform was re-occupied workers quickly discovered that the rig had been toppled by the hurricane. It was later found to be located on the seafloor a short distance from the platform, where it had landed without causing any further damage to the facility or its surrounding infrastructure. A rig replacement program was initiated and a new rig was installed in March, 2012. Spar drilling activities are anticipated to commence in late 2013/ early 2014.
The Mad Dog early Miocene reservoir sands were deposited in an unconfined deep-water basin setting in the southern Green Canyon protraction area. These sands are part of a basin floor fan complex that encompasses BP's Atlantis Field to the east and BHP Billiton Petroleum's Shenzi Field to the north. The fan complex extends laterally for 60 by 50 miles (95 km by 80 km), constrained by regional stratigraphic and paleontologic correlations (Fig. 9). At least 5 individual fans make up the early Miocene fan complex, four of which comprise the main reservoir sands in the Mad Dog Field (Fig. 5). Facies observed in core, rock properties across the field, and reservoir pressure data indicate that the Mad Dog field is situated in a mid-outer fan position and is dominated by a series of compensationally stacked submarine lobes. Mad Dog system sands are thought to be sourced from at least one major canyon to the west-northwest (Fig. 9) based upon regional seismic mapping and log correlation.
Mad Dog contains three early Miocene reservoir sands, the DD, EE, and FF, which are interpreted to represent at least four individual fans. The DD and EE appear to be individual fans, while the FF is subdivided into an upper and a lower unit, which are thought to represent two amalgamated fans. Each of the individual fans can be correlated over a distance greater than 30 miles (50 km) (Fig. 9). Overall fan dimensions compare favorably with basin floor fans in the Tanqua Karoo basin, South Africa (Bouma and Wickens, 1994). The fan complex at Mad Dog is typically around 450 feet (137 m) thick, with individual DD, EE, and FF average thicknesses of 67, 52, and 123 feet (20, 16, and 37.5 m) respectively. The sands are separated by 50 to 100 feet (15–30 m) thick sections of muddy silt and occasional marine shale layers containing Glossifungites-rich surfaces and a few thin, discontinuous sands. The EE sands, FF sands, and the shale packages between the DD/EE and the EE/FF exhibit little variation in stratigraphic thickness across the field. The DD sand is more variable, as unfaulted sections range from 30 to 100 feet (10 to 30 meters). The sand bodies are high (80%-90%) net-to-gross (N:G),and very fine- to fine-grained. There is little variation in N:G within the three sands throughout the field. Strong lateral continuity of the sand bodies throughout the southern Green Canyon region, and the high N:G properties, support the interpretation that the early Miocene sands were deposited in an unconfined basin fan system (sensu Reading and Richards, 1994).
Examination of conventional core from the early Miocene sands reveals repeated intervals of deposition from conventional turbidity currents that have been interpreted to indicate that Mad Dog occupies a lobe-dominated, mid-outer fan position (sensu Johnson et al., 2001). The cores are dominated by massive, amalgamated, structureless sandstones and massive, dewatered sandstones. In a vertical stacking sense, muddy debrites between sand units are more pervasive at the bases and tops of each of the EE and FF sand packages, reflecting initiation and retreat phases (sensu Gardner et al., 2008). Structured sandstones, which display features such as traction and ripple-laminated units, are less common and are generally thinner than the dewatered sandstones, which contain dish structures and consolidation lamina. The cores lack any major slump features, supporting the interpretation of deposition on a stable basin floor. Low volume fractions of discrete erosional features, such as scours, are observed in core, reflecting high-energy deposition within an axial region or an unconfined-to-semi-confined submarine depositional setting. Vertical stacking successions interpreted in individual Mad Dog well logs are interpreted to reflect the relative positions of submarine lobes within a larger fan (e.g., Prelat et al., 2009).
Alternative depositional models have been considered for well correlation, including a more extensively channelized system. These channelized models predict unconfined stacked lobes that have significant thickness and reservoir quality variation. This is not consistent with the well data correlations and production history matching, both of which suggest that the reservoir sands are sheet-like over tens of kilometers; dimensions that are consistent with lobes in unconfined systems (Fig. 10) (Karoo basin, South Africa: Prelat et al., 2010).
Over 25 well bores have penetrated the Mad Dog early Miocene reservoir interval. Well data collection has included LWD and wireline logs, static pressures, and four conventional cores. The three reservoir zones are texturally and mineralogically similar. On average, the sands are fine-grained and moderately well sorted. However, a wide range of grain sizes and sorting exists, from very fine- to medium-grained and poorly sorted to well sorted. The dominant framework grain is quartz and there are varying amounts of feldspar and rock fragments. Ductile components consist primarily of detrital clay, carbonaceous material, and mud clasts. There is very little evidence of authigenic material in the core.
A depositional package scheme has been used to describe the reservoir, based on observations from the cores. These packages are identified based using lithology, texture, sedimentary structures, and vertical succession. Each package is interpreted to have distinct flow properties and form the basis for modeling reservoir flow. The cored wells have been calibrated to logs and used to construct the field petrophysical model. The petrophysical model is used to assign the depositional package scheme to the sands in the un-cored wells using a variety of log responses. Six depositional packages (P1 to P6) have been described in the early Miocene reservoir section by Badley Ashton (Fig. 11). The majority of the reservoir sands consist of P2, P3, and P4. Only one instance of P1, a conglomerate, has been identified in the four cores.
Each depositional package contains a distinct set of characteristics. The P2 package consists of amalgamated, massive sands that may include dewatering structures. Bed thicknesses are on the order of several feet (a few meters). This package tends to have the coarsest grain size and best sorting (Fig. 11).
The P3 package represents amalgamated event beds that have massive sandstones at the base with preserved finer grained bed caps. Structures grade from massive to cross-bedded and contain additional carbonaceous material. Dominant grain size and sorting is similar to P2 in the massive part of the package. The cross-bedded section is finer in grain size and poorer sorting results from an increase in carbonaceous material.
The P4 package is a sand-prone heterolithic interval, in which the sands tend to be laminar to ripple cross-bedded and may contain bed caps. There is very limited core across this package and it tends to represent a minor percentage of the pay sands as a whole.
The P5 and P6 packages are found between the DD, EE, and FF sands. They are mud prone heterolithic units that may contain thin, structured very fine-grained sand and silt beds that are moderately to poorly sorted. The depositional environment across Mad Dog ranges from axial to off-axis to fringe. The axial deposits have a high proportion of P2 and minor amounts of P3. Off-axis deposits contain lesser amounts of P2 and higher proportion of P3.
Reservoir quality is texturally controlled. Grain size is the dominant controlling factor. Clay content and sorting have secondary impact. There are only minimal authigenic clay and cement components present. Porosity loss is caused by mechanical compaction; chemical processes are minimal. Porosities range from 20% to 27% and average 24% for all sands. Air permeability averages range from several hundred millidarcies to greater than one darcy. Typically, the EE has the highest permeability sand, with close to one darcy average air permeability. The DD and FF sands have averages on the order of 200 to 300 millidarcies. Water saturation averages range from 14% to 42%, having the highest values in the FF.
The Mad Dog structure is a salt-cored anticline having a four-way dip closure. The structure was active from the Eocene through the early Pleistocene. At the early Miocene reservoir level, there is over 15,000 feet (4,500 meters) of structural relief from the crest of the anticline to the deepest parts of the basin to the northwest. Several fault trends dissect the structure and divide the reservoir into compartments (Fig. 12). These compartments contain large hydrocarbon columns that are trapped at their updip and lateral limits by sealing faults with over 150 feet (~45 m) of throw, and generally open to the aquifer at their down dip margins.
The evolution of the Mad Dog structure was controlled by salt withdrawal, far-field compression, downdip translation, and local salt inflation. The region is underlain by a layer of Jurassic-aged salt (Peel et al., 1995; Hall, 2002) which acted as a fluid over geological time, moving in response to pressure differences driven by sedimentary loading and other factors (e.g., Hudec et al., 2009). Thickness variations in the rocks (which may be Cretaceous or Jurassic in age) directly overlying the salt indicated that differential movement commenced with deposition (Fig. 13). By the Eocene, a basic precursor of the Mad Dog structure was manifest in the area as thinner sediments draped over a pair of gentle anticlines surrounded by thicker sediments in synclines on either side. This early formed structure might have resulted from a pulse of contractional deformation (Grando and McClay, 2004) but was buried and remained dormant through the early and middle Miocene, (Fig 13(i)), as seen in the generally isopachous strata deposited throughout that time. Fold amplification began during middle Miocene (ii) deposition and a pronounced asymmetry developed, as the southern basin subsided while salt inflated beneath the crest and northern flank of the structure. The major faults over the field began to form during this phase of deformation to accommodate the extension from bending in the outer arc of the fold. These faults primarily display normal offset, although rare thrust faults seen below the neutral surface deep in the fold core also probably began to move during this phase. A small structural closure in the early Miocene stratigraphy was in place by middle Miocene (ii) time, and therefore any hydrocarbons generated after that time could be trapped by the growing structure. By the late Pliocene, the Mad Dog structure was a broad, south-facing monocline, which had extensive crestal faulting.
Around the Plio-Pleistocene boundary the allochthonous salt canopy was emplaced rapidly across the area (Hudec, 2008). The transfer of vast amounts of salt from beneath the Mad Dog structure to on top of it may have been triggered by increased Pleistocene deposition (Fig. 14) on the basin floor (Fig. 13) as the melt waters from retreating Laurentide ice sheets carved canyons in the shelf and deposited large volumes of sandy mud far out into the Gulf of Mexico basin (Galloway et al., 2000). The evacuation of salt from beneath the Mad Dog structure caused the progressive collapse of the north flank and a final phase of motion on the crestal normal faults. The deep salt layer eventually welded out in places, which halted the growth of the anticline and froze the structure into the shape seen today. Continued Pleistocene sediment deposition and salt advance (right side Fig. 13) buried the structure and led to the development of the Sigsbee Escarpment bathymetric feature (Jackson et al., 2008).
Faults, fault systems and fault seal
The majority of the faults in the Mad Dog area appear to be extensional, related to outer arc bending during the development of the anticline. Some of the largest faults in the field, however, steepen with depth and appear to show evidence for early motion and later reactivation. This early phase of motion may be related to the basinward translation of the entire region on top of the salt detachment during deformation in the Mesozoic. These first-formed faults may have accommodated strike-slip or tearing motion as parts of the system moved at different speeds. As sediments appear to be thinnest over the Mad Dog anticline, this would represent a natural weak point in the system, and the large graben faults may have been among the first to form. The overall Mad Dog structure has a slightly sigmoidal shape (Fig. 3) consistent with mild right-lateral strike-slip deformation.
The working hypothesis for the subsurface team is that any observed compartments and baffles are most likely caused by faults. The relatively uniform stratigraphy drilled to-date and the unconfined basin floor fan depositional environment for the early Miocene sands suggests that lateral stratigraphic heterogeneities are unlikely to be present at an inter-well scale. Analysis of fault seal capacity supports this assumption. Around 200 feet (60 m) of fault throw is needed to seal the EE/ FF sands, while less than 100 feet (30 m) of throw is needed to seal the DD sands (Fig. 15, left side). Segments of faults with less than ~100 feet (30 m) of throw are predicted to act as baffles, but not as barriers to the movement of hydrocarbons through the reservoir. The early Miocene reservoir section contains sufficient sand that it should not seal when self-juxtaposed. These values have been calculated using the Mad Dog Deep 2 (GC0826 #3) well results and Badley's TrapTester and Triangle software, using a shale gouge ratio (SGR) cut off value of 25% (e.g., Bretan et al., 2003).
The calibration of fault seal capacity across the field can be used to visualize potential sealing and non-sealing faults on a reservoir map (Fig. 15, right side) which aids future wellbore planning and reserve calculation. The 2009 TTI processing of the WATS seismic data provides enough resolution to map faults with offset greater than 100 feet (30 m) confidently. Faults with offset between 100 feet (30 m) and 30 feet (9 m) can be mapped with less confidence, and faults with offset of less than 30 feet (9 m) cannot be mapped and are considered to be below seismic resolution (Fig. 16). This resolution means that all faults that seal the entire early Miocene reservoir sands should be visible on the seismic data. Segments of faults that only seal the DD, but not the EE/FF may be mapped with low confidence on the data, while faults that baffle production may not be seen at all. The major compartment-bounding faults in the field are mapped to have greater than ~200 feet (60 m) of displacement along the sections where they are interpreted to seal the reservoir section. This compares favorably with the amount of missing section recognized in well bores that penetrate these faults.
The resolution limitations of the 2009 TTI seismic reprocessing require that any products derived from the data must be evaluated for kinematic, geometric, and geomechanical soundness. An example of this evaluation process is shown in Figure 16. It is generally accepted that fault populations follow scaling relationships (e.g., Dawers and Anders, 1995), and published compilations of normal fault measurements suggest that they tend to have length-to-throw ratios between 10 and 100 (e.g., Evenick, 2009). These simple scaling rules-of-thumb have been used to assess the quality of the seismically interpreted faults across the Mad Dog field. The results showed that the Mad Dog fault pattern generally fell within the boundaries expected from the literature. Most faults seem to display a length-to-throw ratio within the 20 to 50 range and the fault set also shows tighter clustering than the fault set mapped on the 2007 VTI data, reflecting an improvement in our ability to map faults on the new data. Faults having ratios that fall outside the 10 to 100 ranges have been reviewed and generally there are satisfactory, unique explanations; for example, the fault is not fully contained within the dataset. One observation from this work is that the largest southeast ridge faults (yellow diamonds, Fig. 16) appear to be long relative to the amounts of throw mapped on them. This observation is in spite of the fact that this fault set is predominantly outboard of the shallow salt, in the best-imaged area of the field. This supports the interpretation that there may be an amount of strike-slip displacement resolved on these faults, further increasing their length relative to the amount of vertical displacement mapped across them.
A final observation from Figure 16 is that the number of mapped faults decreases rapidly in the seismic uncertainty domain. It is widely accepted that fault systems have a certain scale-invariance (e.g., Yielding et al., 1992); therefore, if a fault population contains a certain number of big faults, it will have correspondingly more small faults. This matches operational experience in Mad Dog, where a well bore that crosses one seismically visible fault will typically have around three times that number of faults in a dipmeter interpretation. Additionally, production history matching indicates that wells in the east compartment communicate with each other less freely than expected, suggesting there are baffles of reduced permeability in the intervening rock volume (Fig. 17). In addition, two of the four whole cores of early Miocene rocks taken in the field have contained small faults. An example of this can be seen in photographs of the core from the early development GC0782 #1_ST03 well that intersected a small fault having an estimated 20 feet (6 m) of missing section (Fig. 18). Structural logging of the core indicates a zone of increased frequency of deformation features that extended 13 feet (4 m) above and 6 feet (2 m) below the main fault zone.
Petrophysical testing shows that these deformation features can have a reduction in permeability of over five orders of magnitude compared to the host rock. The asymmetry in this deformation zone might be related to lithological variation, as 6 feet (2 m) below the fault, the host rock changes from clean sandstone to thinly bedded sandstones and shales. The scientific literature suggests that fault damage zone thickness scales with displacement on the fault. Childs et al. (2009) suggest that this relationship may be around 1:1 for faults of this size. Extrapolating these results throughout the Mad Dog field reveals there could be large volumes of rock in deformation haloes around faults that will have reduced permeability, leading to decreased sweep efficiency and a lower recovery factor.
There are eight main faults in the structure that have been interpreted as sealing (Fig. 12 inset) and therefore delimit the major compartments that explain the production, pressure and fluid results of the wells drilled to date. There are four major compartments with separate OWCs. faults are represented by red lines on Figure 17, in contrast with the model boundaries and white lines, which represent seismically visible faults.
The south compartment contains lowest known oil at 23,464 feet TVDSS (7,150 m), which was observed in the south sidetrack (GC0826 #5 ST01). Faults 2 and 3 are thought to seal because the Mad Dog South well (GC0826 #5) tested a virgin pressure fault block. Fault 2 is interpreted to isolate the compartment from the ~6 years of production from the GC782 #4_ST02 (A3) and GC0782 #9_ST01 (A5) wells in the southwest ridge, while fault 3 is interpreted to isolate it from the ~6 years of depletion in the east compartment. Fault 3 is also thought to maintain the 3,500 feet (~1,080 m) difference in OWC between the south and east compartments. Faults 2, 5, 7, and 8 are thought to seal to explain the wet penetration at the crest of the Southwest ridge (GC0826 # 4_ST02). The high pressures and salinities recorded by this well are interpreted to show it landed in an isolated compartment that is not connected to the primary aquifer.
The east compartment is interpreted to have an OWC at 20,640 feet TVDSS (6,290 m), based on a penetration in the GC0826 #3 well. The oil-filled east compartment is separated from the wet penetrations to the north (e.g., GC0783 #1) by Fault 4 (Fig. 12). It is separated from the wet penetrations in the graben (GC0782 #1_ST02) by the eastern graben bounding fault (Fault 6 on Fig. 12). Fault 3 is believed to seal, which explains the >3,500 feet (1067 m) difference in OWC between the east compartment and the south compartment. The east compartment contains four actively producing wells. History matching the production and pressures from these wells suggests they exhibit good connectivity with minor baffling. These baffles are believed to be subseismic faults much like the fault captured in the GC0782 #1_ST03 core. These
The west compartment is interpreted to have an OWC at 23,330 feet (7,111 m), based on a penetration in the extended reach GC781 #11 downdip pilot hole. Fault 1 is thought to seal because the A7 well (GC0782 #11_ST01) did not measure any depletion from the ~3 years of production from the A3 and A5 wells, suggesting they produce from an adjacent compartment. Fault 5 is believed to seal to form the up dip limit of the hydrocarbon-filled west compartment.
Southwest ridge compartment
The southwest ridge compartment is thought to have lowest known oil at 22,000 feet (6706 m) TVDSS, assuming it is located in the same compartment as the GC825 #1_ST01 well. This compartment is known to be separate from the west and south compartments because the depletion from production by the A3 and A5 wells was not seen when the west and south compartments were drilled. The OWC in this block is interpreted to be 23,330 feet TVDSS (7,111 meters), based on analysis of pressure data. This depth approximately corresponds with the downdip mapped extent of the block, and the block is therefore considered fully charged.
Overall performance from the early Miocene wells has been excellent. Peak production rates from individual wells range from 15,000 to 27,000 BOPD, depending on the permeability and thickness of the reservoir interval, viscosity of the hydrocarbons and near wellbore skin effects. The wells are completed with dual frac packs (DD sand and EE/FF sand), which have initial skins ranging from slightly negative to as high as 10.
The Mad Dog early Miocene reservoir is remarkably consistent, having similar reservoir characteristics, sand thicknesses, and fluid properties from well-to-well and within the production interval. Differences in well performance are primarily due to permeability, completion efficiency, proximity to water, and faulting in the reservoir. Some degradation in productivity has been observed in wells after they have begun to produce water.
The wells in the east compartment appear to have outstanding aquifer support, having an aquifer to hydrocarbon ratio of approximately 30. The producers communicate well with each other, and reservoir pressures remain fairly stable despite high off-take. All of the east compartment wells are currently producing water, with between 5-40% water-cut; however, the water production in several wells is lower than originally estimated. In particular the GC0782#5 (A2 producer) that is farthest downdip and closest to the OWC is producing much less water than predicted. This may indicate that it is separated from the aquifer by a subseismic fault.
Various surveillance data have been collected from wells in the east compartment. Production logs (Fig. 19) are interpreted to show that all reservoir sands are contributing to flow but high permeability streaks may be present, which will impact water movement through the reservoir. Pressure measurements in development wells drilled into depleted compartments also have provided valuable insights. Figure 20 shows the normalized original reservoir pressures of the east compartment exploration, appraisal, and development wells along with the depleted pressure measurements taken in the A6 and A8 producers. It is apparent from these measurements that the wells in the east compartment are in pressure communication but have a small differential pressure between sand intervals pre-production. These differential pressures have grown with production, suggesting that the shales between the sands are causing vertical compartmentalization on a production time scale. Differential depletion within sands is interpreted to show the impact that the differing porosity and permeability properties of each depositional package has on flow. By combining the pressure data and water production history, it is apparent that subseismic geologic features exist that impact reservoir connectivity on a small-scale. However, the east compartment wells have exhibited some connectivity to each other and to the aquifer. As a result, the subseismic features do not appear to have significantly degraded the overall east compartment recovery
The production wells in the west and southwest ridge compartments have also provided very encouraging results to-date. Material balance calculations and pressure transient analysis suggests the highly productive A5 and A7 wells are producing from large, well-connected fault blocks. The A5 in particular has provided sufficient dynamic data to conclude that the reservoir is well-connected, even in an area of the field that is pervasively faulted and poorly imaged seismically. The west compartment appears to be connected to a moderately sized aquifer and has an aquifer to hydrocarbon volume ratio of 8. The southwest ridge compartment is connected to little or no aquifer.
The lessons learned from the early production history at Mad Dog have provided a great foundation on which to build the design of the Phase 2 project. The dynamic performance has been fundamental to the overall appraisal of the Mad Dog Field guiding the interpretation of static description, reservoir characterization, and reservoir connectivity.
Mad Dog is a giant, deep water, subsalt field that should produce hydrocarbons in the Gulf of Mexico for the co-owners BP, BHP Billiton Petroleum, and Chevron for the next forty years. Since the field discovery in 1998 the in place resources have increased eight fold to an estimated 4 billion barrels. The early Miocene reservoirs have been deposited by turbidites in an unconfined, structurally quiet, deep-water basin fan setting. This has resulted in a series of laterally-extensive, well-connected, porous, permeable, and sand-rich reservoirs.
Major structural development began in the middle Miocene and continued to the Plio-Pleistocene. Downdip translation of sediments and movement of the underlying salt created a giant anticlinal structure having four-way dip closure. Large, bending-related normal faults divided the reservoir into a series of compartments, many of which were later charged with hydrocarbons. Production data showed that these compartments are separated on a development time scale, but most likely were connected over geologic time.
Within each compartment, reservoirs deplete as expected from the porosity and permeability structure, although high-permeability streaks and subseismic faults create challenges for predicting reservoir performance and water breakthrough.
The Mad Dog co-owners have greatly benefitted from the phased-development approach taken on the project. This has reduced subsurface uncertainty by collecting valuable dynamic information while producing the hydrocarbons identified by the discovery and appraisal program. The development of the field has also benefitted from being the laboratory for new seismic acquisition and processing technologies. This investment in subsalt imaging has proved to be a game changer, allowing the subsurface team to appraise this giant field more fully, de-risk the Phase 2 development significantly, and begin to unleash the full potential of the Mad Dog.
The authors wish to thank all the co-workers, past and present, who have contributed to the success of the Mad Dog project and advanced our understanding of the subsurface structure. In particular we single out Glen Anderson, Jerry Bowling, Joaquin Naar, and Zeke Snow for improving this paper and freely sharing their ideas, words, and graphics. We thank our current team leaders, Sneha Chanchani (BP), Patrick Massey (BHP Billiton Petroleum), and Brian Grant (Chevron) for allowing us the time and space to work on this paper and for encouraging us to share the Mad Dog story. We thank the managers in BP GOM Production and GPO and our co-owners BHP Billiton Petroleum and Chevron for permission to publish and for their continuing belief and investment in the Mad Dog project. This paper has benefitted from helpful reviews by Zeke Snow, Ed Lisle, and Jim Brenneke.
LoSal™ is a trademark of BP p.l.c.
Figures & Tables
New Understanding of the Petroleum Systems of Continental Margins of the World
- Atlantic Ocean
- chemically precipitated rocks
- clastic rocks
- deep-water environment
- geophysical methods
- geophysical profiles
- geophysical surveys
- Green Canyon
- Gulf of Mexico
- North Atlantic
- oil and gas fields
- oil-water interface
- reservoir rocks
- sedimentary rocks
- seismic methods
- seismic profiles
- Sigsbee Escarpment
- subsalt strata
- well logs
- Mad Dog Field