Overpressure at the Macondo Well and its impact on the Deepwater Horizon blowout

At the Macondo well, the overpressure (fluid pressure greater than hydrostatic) in the main reservoir is nearly identical to that within a stratigraphically equivalent sandstone at the Galapagos development 21 miles (34 km) to the south; we interpret that the reservoirs share a permeable, laterally extensive, and hydraulically connected aquifer. At Macondo, pore pressure approximately parallels the overburden stress to a depth of 17,640 ft (5,377 m) subsea and thereafter decreases abruptly by 1,200 psi (8.3 MPa) over 370 ft (113 m) as the main sandstone reservoir is approached. In contrast, at Galapagos, pore pressure increases with the overburden stress for the entire well depth. The pore pressure regression at Macondo was responsible for a reduction in the least principal stress. This, in combination with the extreme pore pressures within overlying strata, drastically narrowed the range of safe operational borehole pressures. These geologic phenomena produced challenging conditions for drilling, prevented successful temporary abandonment of the well, and contributed to the well’s failure.

is annotated A-A′ with a white dashed line. The green dashed line denotes the M56 reservoir shape from BP's exploration plan 15 , but is artificially truncated N-S. The structural map of the M56 reservoir is interpreted from a 3-D seismic volume that is zero-phase, narrow-azimuth, tilted transversely isotropic, and pre-stack reverse-time-migrated in depth. Dark pink indicates truncation of the M56 by salt stocks. The M56 steeply shallows against SE diapir forming a potential fluid leak point. The narrow-azimuth survey does not image bedding well beneath salt (light pink). The hydrostatic pressure, u h , assumes a constant fluid density of 1.024 g/cm 3 (seawater) from the sea surface. The overburden stress, σ v , is calculated by integrating the density of the sediment below the seafloor (see Methods). Direct measurements of pore pressure are shown with symbols (triangles, squares, circles; see Methods for discussion). The mudstone pressure, u ms , interpreted from the (sonic) velocity log is shown with the blue line and the annular pressure-while-drilling (APD) measured near the drill bit is shown with the black line. The depths associated with the M57 and M56 sandstones are highlighted in yellow. (b) Lithology penetrated by the Macondo well based on cuttings. rock compaction is a function of effective stress 12,13 and our approach is described in the Methods section. The mudstone pore pressure profile is quite similar to the measured sandstone pressures. From 11,650 to 17,640 ft (3,551 to 5,377 m), u ms increases subparallel to the overburden. Below 17,640 ft (5,377 m), the mudstone pore pressure estimate fully captures the magnitude of the pore pressure regression measured in the M56 reservoir.

Seismic Interpretation and Stratigraphic Correlation
We map the spatial distribution of the top of the M56 across a 20 by 29 mile (32 by 47 km) area using a 3-D seismic volume (Fig. 1b). We tie the top of the M56 reservoir from projected log data to a reflection in the seismic data, and then track this event across the seismic volume. The top M56 surface ranges from 15,500 to 26,500 ft (4,700 to 8,100 m) resulting in over 11,000 ft (3,350 m) of relief within our study area (Fig. 1b). The map shows one structural high to the north penetrated by the Macondo well, and a second to the south targeted by the Galapagos development. Our mapped surface closely correlates with BP's independent analysis of the depth of the M56.
The sandstone itself could not be resolved with these seismic data. However, significant sandstones correlate with the M56 surface at every well penetration shown in Fig. 1b. In this region, the transport of sand by turbidity flows in the Middle Miocene was NW to SE 14 . The geologic model from BP's exploration plan defines the M56 reservoir as an amalgamated, low-relief channel-levee complex that trends NW-SE and has an average thickness of 25-43 ft (7-13 m) 15 . Modern analogs of elongate, continuous, sand-prone channel-levee complexes exceed 30 miles 16 . This characterization of the M56 sandstone (Fig. 1b, dashed green line) is consistent with subsequent reservoir simulation that supports a long narrow aquifer 17 .

Aquifer Pressure
We compare the aquifer pressure at Macondo with the aquifer pressure at the Galapagos development (Fig. 1). The aquifer pressure is the water-phase pressure in the sandstone and it removes the effect of hydrocarbon buoyancy 18 . The aquifer overpressure, u a *, is a single number that is independent of depth within a permeable hydraulically connected sandstone 19,20 . At Macondo, we calculate u a * to equal 3,386 psi (23.35 MPa) whereas at the Galapagos wells, u a * is 3,433 psi (23.67 MPa) (see Methods). The difference in u a * between the Macondo and Galapagos locations is 47 psi (0.32 MPa) which is less than 1.5% of the total pressure. We interpret that the nearly identical aquifer pressure records hydraulic connectivity through a shared aquifer.

Pressure and Stress Profile through the M56 at Macondo and Galapagos
We compare the pore pressure and velocity profiles across the M56 reservoir at both Macondo and 562-1, one of the Galapagos wells (Fig. 3). The depth below sea surface of the M56 and its pore pressure are essentially identical at the two locations. At 562-1, the pore pressures above and below the M56 (Fig. 3d, symbols) record a continuous and gradual increase in pore pressure with depth subparallel to the overburden. In contrast, the pore pressure at Macondo (Fig. 3c) is much lower within the M56 than above it. The mudstone velocities at both wells (Fig. 3b,e, black lines) increase where the sandstone pressures decrease, reflecting increased compaction due to increased effective stress. At Macondo, there is a sharp increase in mudstone velocity across the M56 interval. The average mudstone velocity (Fig. 3b, thick black line) is 9,500 ft/s (2,900 m/s) across the M57 interval (17,250-17,640 ft or 5,258-5,377 m), but average mudstone velocity increases to 11,000 ft/s (3,350 m/s) across the M56 interval (17,640-18,250 ft or 5,377-5563 m). Although not shown, resistivity and density also increase in this interval, reflecting the increased compaction. In contrast, at 562-1, the velocities show a continuous and gradual increase with depth (Fig. 3e, thick black line). Likewise, our mudstone pressure estimation (Fig. 3d, blue line) is nearly continuous across the M56 at 562-1 in contrast to the pore pressure regression at Macondo.

Basin Hydrodynamics
We integrate the observations at the Galapagos and Macondo wells with the map of the M56 surface to present a conceptual model of overpressure across the region (Fig. 4a). We have documented nearly constant aquifer overpressure, u a *, in the M56 at both locations and we assume u a * remains constant between them (Fig. 4, green). From the seafloor down, the mudstone overpressure increases linearly, subparallel with the lithostatic stress as is observed at both Macondo and Galapagos (Figs 2 and 3). At Galapagos, the mudstone pressure is approximately equal to the sandstone pressure at the M56 (Fig. 3d). In contrast, to capture the pore pressure regression at Macondo, there is a reversal in the mudstone pressure trend as the M56 is approached (Fig. 3c); this results in a return to cooler colors (Fig. 4). Beneath the M56, mudstone overpressure again increases. In the conceptual model, contours are connected between wells by assuming a linear decrease in the mudstone overpressure gradient from A to A′. Contouring adjacent to the M56 assumes the pore pressure regression, if present, is approximately the same distance from the M56 as is observed at Macondo (Fig. 4a, dashed black line).
This overpressure is also expressed in a plot of overpressure vs. depth below seafloor (Fig. 4b). In this view, the constant overpressure of the reservoir at the depths mapped is illustrated with a vertical solid black line. The overpressure in the bounding mudstone away from the reservoir is shown with white lines that represent both Macondo and Galapagos. The M56 pressure is lower than the bounding mudstone pressure at depths below seafloor greater than present at Galapagos (11,650 ft or 3,550 m).
The overpressure cross-section is a fluid potential map: water flows orthogonally to the overpressure contours within material of isotropic permeability. Flow within the mudstone is illustrated by black arrows. In areas where there is a pore pressure regression (Fig. 4a, area between the dashed line and the M56), flow is focused toward the M56. Elsewhere flow is upward: pore pressure gradually dissipates as fluids flow to the seafloor. We interpret that the Galapagos and Macondo reservoirs are hydraulically connected because they have nearly identical aquifer overpressures. In fact, the aquifer pressure at Galapagos is interpreted to be 47 psi (0.32 MPa) greater than at Macondo. In a 2-D view, this implies flow from Galapagos towards Macondo. Although the pressure difference is small, it can drive a lateral flow rate of 200 mm/year given the 300 mD permeability that is estimated for these sandstones.
www.nature.com/scientificreports www.nature.com/scientificreports/ It is well recognized that in many basins, regionally connected high-permeability aquifers at a nearly constant overpressure are encased in low-permeability overpressured mudstone such as is illustrated here in the M56 [21][22][23][24] . A key question is, what controls the aquifer overpressure in these systems? One common interpretation is that there is a leak point where the aquifer pressure equals the least principal stress. At this leak point,  www.nature.com/scientificreports www.nature.com/scientificreports/ the pore pressure bleeds off through fractures and the aquifer pressure is fixed to the least principal stress 19,20 . We mapped the M56 reservoir up to 9,500 ft (2,900 m) below the seafloor against the salt diapir 3 miles SE of Macondo (Fig. 1b). At this depth, the aquifer pressure converges to within 1000 psi (6.9 MPa) of the overburden stress ( Fig. 4b. 'v'): we interpret that the leak point is at or near this location.

Implications of the Pore Pressure Regression
The pore pressure regression hindered the drilling and temporary abandonment of the Macondo well. To illustrate this, we express the downhole pressures and stresses with an equivalent mud weight (EMW) or density plot (see Methods section) (Fig. 5a,b). Within the exposed borehole, a single mud weight is used to maintain the borehole pressure (1) below the fracture pressure to avoid the loss of drilling fluid (mud) through fractures into the formation and (2) above the pore pressure to prevent flow from the formation into the borehole. The difference between the equivalent mud weight necessary to cause fractures anywhere in the exposed borehole and the EMW that equals the formation pressure anywhere in the exposed borehole is the drilling window. During operations in the deepest well segment at Macondo, the formation was exposed below the base of the 9 7/8″ liner (Fig. 5d). Along this segment, the drilling window was extremely narrow (Fig. 5b,c, orange rectangle): the left bound of this window is constrained by the pore pressure in the M57 of 14.20 ppg (1.702 g/cm 3 ) EMW (Fig. 5b,c, red circle) and the right bound is constrained by the fracture pressure (the least principal stress) within the M56 sand (14.3-14.4 ppg (1.714-1.726 g/cm 3 ) EMW) (Fig. 5b,c, red triangle).
This narrow drilling window created challenging drilling conditions. Gas flowed into the well from the M57 (Fig. 5b,c, open triangle), indicating that borehole pressures had dropped below the pore pressure. Furthermore, on three occasions, mud was lost into the formation (Fig. 5b, brown triangles), indicating that borehole pressures had exceeded the fracture pressure. In fact, these events constrain the drilling window. The two mud-loss events into the M56 document a lower fracture pressure within this interval than in the upper half of the well segment (Fig. 5b, brown square and uppermost triangle). This drop in fracture pressure (least principal stress) is most likely a result of the reduced pore pressure, but could also be due to different mechanical properties in sands relative to mudstones 25 . The lack of sufficient drilling window meant that BP was forced to terminate drilling without fulfilling all of its objectives, which included drilling to 19,560 ft (5,962 m).
The narrow drilling window impacted the approach used to cement the production casing in place. To maintain the pressures along the cement column within the drilling window (Fig. 5c, gray lines within orange rectangle), BP and Halliburton used 16.74 ppg (2.006 g/cm 3 ) cement foamed with nitrogen to reduce its downhole density to 14.5 ppg (1.738 g/cm 3 ) to keep dynamic borehole pressures below 14.583 ppg EMW (1.747 g/cm 3 ) 26,27 . The particular foam cement mixture was shown to be unstable during testing prior to and after the blowout 28,29 . At trial, it was accepted by both parties that this cement failed, and it was BP's position that failure of the cement was the primary cause of the failure to seal the well 4,29,30 .
The Macondo well penetrates a complicated hydrogeologic system. A sedimentary section of near lithostatic pressures overlies the lower pressured M56 sand, the exploration target (Fig. 2). The M56 is a regional aquifer whose  www.nature.com/scientificreports www.nature.com/scientificreports/ pressure is likely controlled by a leak point located where the M56 shallows and its pressure converges on the least principal stress (Figs 1, 4b 'v'). To drill and produce this hydrocarbon target required a delicate balance to keep the borehole pressure above the pore pressure present and below the fracture pressure. The technical challenges associated with drilling and cementing this complicated hydrodynamic system contributed to the ultimate blowout of the Macondo well.

Methods
Macondo Pore Pressure and Stress Profile. The overburden stress is calculated by integrating the weight of the water column and the weight of the overlying sediment. We combine density log data from nearby wells in portions of the Macondo well where no density data were acquired. Logs are corrected to account for borehole washout and for the presence of hydrocarbons. Where no density data are available, a velocity-to-density transform is used 31 . If neither density nor velocity data are present, an exponential interpolation between density above and below the interval is used 12 .
Industry routinely measures pore pressure and takes fluid samples from relatively permeable formations with wireline tools (e.g. Modular Formation Dynamics Tester TM , MDT) and directly from the drill string (Geotap TM ). At the Macondo well, BP recorded 21 pressures in four sandstones at the base of the well between 17,600 and 18,150 ft (5,364 and 5,532 m) (Fig. 2a, circles). 70 MDT pressures were recorded in nine sandstones between 8,900 and 12,500 ft (2,700 and 3,800 m) (Fig. 2a, squares) at the Texaco 252-1 well, located 1.27 miles (2.04 km) SW of the Macondo well. These MDT measurements are corrected to the Macondo well location assuming continuous stratigraphy parallel to the seafloor 32 .
We also constrain pore pressure from fluid influxes into the borehole (kicks) and elevated gas levels detected in the incoming drilling mud. Kicks and high gas occur when pore pressure exceeds hydraulic pressure from the drilling fluid in the exposed borehole. Six such events occurred during drilling operations (Figs 2, 3 and 5, open triangles). Using drilling information prior-to, during, and after an event, we estimate the location and pore pressure.
Drilling information includes the location of sandstones, length of exposed borehole, gas content of the incoming mud, surface mud weight, equivalent static density, equivalent circulating density, and shut-in drill pipe pressure. The equivalent mud weight is another way of expressing pressure using the average density of the drilling fluid from the drill floor to a location in the borehole. The equivalent static density is the downhole pressure expressed as an equivalent mud weight when the mud pumps are off and thus, there is no circulation. The equivalent circulating density is the downhole pressure expressed as equivalent mud weight as while the drilling fluids circulate. The circulating density is greater than the equivalent static density because of friction caused by fluid circulation. www.nature.com/scientificreports www.nature.com/scientificreports/ The fracture pressure is the borehole pressure necessary to hydraulically fracture the formation. It is commonly close to the regional least principal stress but can be affected by stress perturbations due to the borehole geometry and the cohesive strength of the rock. The fracture pressure is constrained at four locations below the 9 7/8″ liner (Fig. 5). The downhole static and dynamic drilling pressures leading up to, during, and after each lost mud event are used to bracket the fracture pressure interpretations (Fig. 5, brown triangles). We define the upper bound of the fracture pressure with the equivalent circulating density when the losses began and the lower bound from the highest static or dynamic pressure at which the well is stable before or after the loss event (see ref. 32 for detailed explanation). It is generally accepted that the in-situ stress of mudstone is higher than that of sandstone 25 , so the loss location is assumed to occur in the sandstone nearest to the bit at the time of the loss event. Fracture pressure is also constrained with the 9 7/8″ formation integrity test, FIT (Fig. 5, brown square). After drilling out of the cemented liner shoe, pressure on the exposed formation was increased to above overburden stress without experiencing fluid loss. This test result provides further evidence that the subsequent losses occurred deeper, in the M56 reservoir interval.
Mudstone pore pressure. Rapid deposition of this low permeability material is the primary source of overpressure in the Gulf of Mexico 33 . It is not practical to directly measure the pressure within these low permeability mudstones. Instead, mudstone pore pressure is commonly estimated from the compaction state (porosity) of the rock, which is typically measured by resistivity, density, or velocity 34,35 . In this approach, a correlation is established between one of these petrophysical proxies and the vertical effective stress, σ′ v . Once the correlation is established, then σ′ v is determined given the observed property (e.g. velocity, density, resistivity). Once σ′ v is determined, pore pressure, u, is easily determined if the overburden stress, In deepwater Gulf of Mexico Neogene sediments, pore pressure is not accurately described by a single compaction curve. This is because deeper, hotter, and older mudstones have undergone more compaction than shallower mudstones at the same effective stress. Clay diagenesis is thought to be the primary cause of this behavior and the smectite-to-illite transformation (S/I) is considered the most significant [36][37][38] . More illitic material has a lower porosity at a given effective stress than a more smectitic material 39,40 . We follow ref. 39 and assume: The left side of Eq. 1 is the total porosity, φ, less the pore volume that is filled by clay-bound water, φ m . The molecular structure of smectite has an easily hydratable interlayer, whereas illite does not 41 ; thus the clay-bound water in the illite is less than that in the smectite (φ m , i < φ m , s ). The right side of Eq. 1 is a well-established trend for mudstone compaction (e.g. refs 13,35 ) and here it describes intergranular porosity loss with effective stress. It is not well known whether φ 0 or B vary with the degree of the S/I transformation, so we assume that they are constant (ref. 39 ) We calibrate the model by determining the effective stress within mudstones adjacent to where pressure has been measured in sandstones. We assume that the overpressure, u*, in the mudstone equals u* measured in the nearby sandstone (e.g. ref. 21 ), and use the mudstone pressure and overburden to calculate effective stress (u = σ v − σ′ v ). Next, we determine the mudstone porosity at each location from the velocity log after 42 : where v ma is matrix velocity, v is the velocity log measurement, and x is an empirically derived acoustic formation factor exponent. We assume x = 2.19 and v ma = 14,909 ft/s (4,545 m/s) following precedent for Gulf of Mexico Neogene sediments 21,35,42 . The shallow locations with cooler in-situ temperatures have a higher porosity for a given effective stress than the deeper and warmer locations (Fig. 6). This contrast is most apparent at a vertical effective stresses equal to 1,500 psi (10 MPa) where the porosity, φ, in the shallow section is 9 porosity units greater (Fig. 6, green symbols) than the deeper section (Fig. 6, red symbols). We interpret that the deeper sediments have lost clay-bound water φ m as the smectite in the mudstone was converted to illite with burial. We assume that porosity loss from clay-bound water release during the S/I transformation is linearly proportional to temperature, and that transformation begins at 70 °C and plateaus at 110 °C. This approximates the main phase of S/I transformation [43][44][45] without additional constraints on depositional history and chemical composition 46 . We follow Lahann 39 and assume φ m = 0.12 for smectitic mudstone and φ m = 0.03 for illitic mudstone. Based on these assumptions, the clay-bound water porosity is: where T is temperature, and T s and T i are the smectite (70 °C) and illite (110 °C) transformation boundary temperatures. We combine Eqs 2 and 3, and solve for φ − φ m for all the φ vs. σ′ v points in Fig. 6. We then use least-squares regression to constrain Eq. 1 and find φ 0 = 0.22 and B = 2.9E −4 psi −1 (Fig. 6, black line). Given B and φ 0 , Eq. 1 is then used to estimate mudstone pressure along the borehole (Fig. 2a, blue line) with φ m calculated from Eq. 2. To calculate mudstone velocity, we picked mudstones along the borehole at 30-40 ft (9-12 m) intervals and applied a 5-pick moving average to the corresponding compressional sonic log measurements. For each mudstone pick, we calculate φ from mudstone velocity (Eq. 2) and φ m from temperature (Eq. 3). φ and φ m are entered into Eq. 1, solving for σ′ v and then u.
www.nature.com/scientificreports www.nature.com/scientificreports/ We apply this method (calibrated at Macondo) to estimate the mudstone pressure at 562-1 (Fig. 3). The close match between the estimated mudstone pressures and the measured sandstone pressures, independent of local calibration, supports the accuracy of our method within this region. Effective stresses at 562-1 are roughly 500-1,300 psi (3-9 MPa) higher than at Macondo (outside of the pressure regression). Mudstone sonic porosities are similar in both wells, but the temperature gradients are different. The Macondo well has an average temperature gradient of 28.4 °C/km versus 26.1 °C/km at 562-1. The lower temperature gradient and deeper water at 562-1 results in M56 temperatures that are nearly 20 °C lower than M56 temperatures at Macondo. The lower temperature indicates that the mudstone at 562-1 is more smectitic than the mudstone at Macondo for a given depth, so the sonic porosities transform to higher σ ′ v (Fig. 6).
Aquifer Pressure. We (Figs 1, 7). These wells are chosen because the pressure measurements were made before production at either location; thus, the measurements are interpreted to record the in-situ pressures unaffected by production or the Macondo release ( Fig. 1, red circles and yellow stars). Many of the measurements were made within hydrocarbon-bearing sections. To determine the aquifer overpressure in such cases, the buoyant effect of the hydrocarbon column must be removed (e.g. ref. 18 ). Specifically, the hydrocarbon pressure is projected down to the hydrocarbon-water contact (HWC) using the MDT-derived hydrocarbon density (Fig. 7). For each well at Macondo and Galapagos, we constrain the HWC, hydrocarbon-phase density, and water-phase density with log, MDT and seismic data. We then calculate aquifer overpressure at Macondo and Galapagos, taking into account pore-water density (u a * = u − ρ pw gz SS ). At Macondo, we interpret that the 4-way closure of the M56 structure (Fig. 1b) was filled to its spill point. We interpret a structural crest at 17,720 ft (5401 m), a saddle at 18,375 (5601 m), and thus a column height of 655 ft (200 m) by depth-correcting BP's predrill interpretation 15 . BP interpreted that the seismic amplitudes supported this filled-to-spill interpretation for the HWC 15 . We calculate the aquifer overpressure, u a *, to equal 3,386 psi (23.35 MPa) using a hydrocarbon gradient of 0.24 psi/ft (5.43 MPa/km) and a pore-water gradient of 0.465 psi/ ft (10.52 MPa/km). It is possible that the structure was not filled to spill thus the HWC is shallower. LLOG-253-1 (Fig. 1, northernmost blue dot) provides the deepest hydrocarbon-bearing penetration of the M56 in the Macondo structure at 18,150 ft (5,532 m), which yields an upper bound to the aquifer overpressure of 3,436 psi (23.69 MPa) The three Galapagos development wells (519-1, 519-2, and 562-1) (Fig. 1) constrain the aquifer pressure at this location to a single value (Fig. 7). At 519-1, two vertically stacked sandstone lobes comprise the M56. Each lobe shows a distinct HWC, but both share a u a * of 3,436 psi (23.69 MPa). 519-2 encountered only water in the M56, which yields u a * of 3,430 psi (23.65 MPa). We use these 519-2 MDT measurements to estimate the M56 pore water density of 0.465 psi/ft (10.52 MPa/km). 562-1 encountered hydrocarbon in the M56 and did not penetrate a HWC. An aquifer pressure calculation that assumes the HWC is just below the sandstone yields a u a * of 3,433 psi (23.67 MPa), which is nearly identical to those observed in the 519-1 and 519-2 wells. We use the average, 3,433 psi (23.67 MPa), to describe the aquifer overpressure at the Galapagos development. www.nature.com/scientificreports www.nature.com/scientificreports/ Temperature Profiles. We determined the temperature profiles at Macondo and 562-1 using temperatures recorded during MDT pore fluid sampling (Fig. 8, open symbols). Temperatures between 113.3 and 113.7 °C were recorded at three MDT sample points in the Macondo well between 13,008 and 13,064 ft (3,965 and 3,982 m) below seafloor (Fig. 8, rectangles). At 562-1, four MDT sample points record temperatures between 93.5 and 98.4 °C for depths between 11,633 and 12,316 ft (3,545 and 3,754 m) below seafloor (Fig. 8, diamonds). BP's temperature model for Macondo (Fig. 8, upper black line) 8 is 3.8 °C higher than the average of the recorded temperatures in the M56 (Fig. 8, rectangle error bars). We assume this difference reflects a correction for borehole cooling. At Macondo, MDT measurements were acquired three days after drilling was completed, which is comparable to the four day gap at 562-1. Therefore, we apply the same 3.8 °C correction to the measurements at 562-1 (Fig. 8, diamond error bars). Our temperature model for 562-1 assumes a linear decrease from the corrected reservoir measurements to the seafloor (Fig. 8, lower black line). Seafloor water temperatures in deepwater Gulf of Mexico approach 4 °C for the water depths observed at Macondo and 562-1.

Data Availability
The data that support the findings of this study are available in this thesis 32 and by request from the corresponding author F.W.M.P. Trial related documents are available at http://www.mdl2179trialdocs.com. The seismic data are not publicly available due to license restrictions. BP's temperature model is used at Macondo; 562-1 temperatures are modeled using a linear projection to the seafloor. Color scheme and dotted lines show the temperature-derived S/I transition zones.