seismic rock physics of steam injection in bituminous-oil...

Path: K:/SEG-HEAVYOILS-10-0501/Application/SEG-HEAVYOILS-10-0501-ch06.3d

Date: 10th September 2010 Time: 15:30 User ID: muralir

Chapter 6

Seismic Rock Physics of Steam Injection inBituminous-oil Reservoirs

Evan Bianco,1 Sam Kaplan,1 and Douglas Schmitt1

IntroductionThis case study explores rock physical properties of

heavy-oil reservoirs subject to the steam-assisted gravitydrainage (SAGD) thermal-enhanced recovery process(Butler and Stephens, 1981; Butler, 1998). Previously pub-lished measurements (e.g., Wang et al., 1990; Eastwood,1993) of the temperature-dependent properties of heavy-oilsaturated sands are extended by fluid substitutional model-ing and wireline data to assess the effects of pore fluid com-position, pressure, and temperature changes on the seismicvelocities of unconsolidated sands. Rock physics modelingis applied to a typical shallow McMurray formation reser-voir (135–160 m depth) encountered within the bituminousAthabasca Oil Sands deposit in Western Canada to constructa rock-physics-based velocity model of the SAGD process.Although the injected steam pressure and temperature con-trol the fluid bulk moduli within the pore space, the effectivestress-dependent elastic frame moduli are the most poorlyknown yet most important factors governing the changes ofseismic properties during this recovery operation. The resultsof the fluid substitution are used to construct a 2D syntheticseismic section to establish seismic attributes for analysisand interpretation of the physical SAGD process. The find-ings of this modeling promote a more complete descriptionof 11 high-resolution, time-lapse, 2D seismic profiles col-lected over some of the earliest steam zones.

The SAGD process has been adopted as the recoverymethod of choice for producing bitumen from the Atha-basca Oil Sands in Western Canada and it has changedrelatively little since the first test installation and experi-ment at the underground test facility in the early 1990s.The invention of Geophysics Research, Department ofPhysics, this thermal-enhanced oil recovery and horizontaldrilling technology was born out of the challenges

associated with producing extremely dense and viscous oil[American Petroleum Institute (API) density <10�] fromshallow siliciclastic reservoirs. Steam carries a significantportion of its energy as latent heat, and it is much moreefficient at transferring heat to the reservoir than merelycirculating hot water. Engineering models of this thermalprocess assume that steam chamber growth is symmetricabout the well pairs, but because of lithologic hetereogene-ities and steam baffles, this is most certainly not the case(Figure 1). Before seismic profiling can be optimized as atool for tracking the movement of steam in the reservoir, itis important to understand the behavior of oil-sands mate-rial when subjected to elevated temperatures, pore pressure,and fluid saturation conditions inflicted by SAGD.

Figure 2 shows a scanning electron micrograph (SEM)image of typical McMurray oil sand. Oil sand, by defini-tion, lacks or has very little cementation. As such, its mod-uli (bulk and shear) are largely dependent on grain-to-graincontacts. These contacts are held in place by confiningpressure, and any reduction in effective pressure will resultin a reduction in effective moduli. The micrographs displaya subtle interlocked texture characterized by relatively highincidences of long and interpenetrative grain contacts. Thissuggests that the shallow oil sands were once much deeperthan they are today, although lithification did not occur.Furthermore, the bitumen has undergone biodegradationand is highly viscous; it may actually support the sandgrains in much of this material and act as partial cementwhen strained at seismic frequencies.

Dipole sonic logs and density logs can provide localmeasurements of effective elastic properties. Effectiverock properties (i.e., effective P-velocity, VP, effective S-velocity VS, and effective density, q) are extracted fromwireline data to establish reasonable elastic parameters(jeff, leff, q) for input to Gassmann’s equation for fluid

1University of Alberta, Institute for Geophysical Research, Department of Physics, Edmonton, Alberta, CanadaThis paper appeared in the September 2008 issue of THE LEADING EDGE and has been edited for inclusion in this volume.

107



substitution (Figure 3). Upon injecting steam into the res-ervoir, a depleted oil zone expands gradually and replacesthe initial bitumen saturation with a combination ofwater, steam, and residual oil at elevated temperaturesand pressures. Also, oil production rates are increased ifthe steam is injected at high pressures because the uncon-solidated oil sand is deformed although there is currentlylittle understanding of what actually takes place within asteam zone. Permeability has been reported to increase as

a result of geomechanical changes (e.g., Wong, 2003; Liand Chalaturnyk, 2006), and the portion of the reservoirthat is touched by steam is completely physically alteredfrom its original state. In that sense, unconsolidated reser-voirs cannot be treated with simple fluid substitution cal-culation and the pressure-dependent effects of the framebulk modulus must be taken into account.

The fluid is also complicated. Ternary diagrams aidin studying the effective P-velocity response of this

Figure 1. Schematic representation of the plumbing sys-tem for oil drainage into horizontal well injector/producerpairs in a typical SAGD process. Steam is pumped into thetop well (red arrows), and heated oil drains along the edgesof a steam chamber under gravity toward a producing wellall along the length of the wellbore (green arrows).

Figure 2. SEM image of (a)uncleaned oil-sands material,and (b) oil sands material withorganic components removed(cleaned). The reflective andresinous material in cracksand pores of (a) is bitumen,and trace amounts of clay canbe seen in (a) and (b).

Figure 3. Scatterplot of (a) VP versus VS, (b) bulk modulus (j) versus shear modulus (l), and (c) incompressibility versusshear modulus (l) computed from dipole-sonic log measurements through the McMurray formation and the overlying mud-stones of the Clearwater formation. The color scale for each panel is cast to the gamma-ray reading (API) for each depth.

108 Heavy Oils: Reservoir Characterization and Production Monitoring



three-component fluid ensemble (oil, water, steam) forvarious temperature and pore-pressure scenarios (Fig-ure 4). The result is a more accurate description of the insitu elastic parameters within an idealized steam chamber,because the frame has likely been completely altered byincreased pore pressure and temperature. Ternary dia-grams also present a visual representation of the stability

of a given effective property of interest. In this reservoir,we assume that a considerable amount of oil remains irre-ducibly trapped within the steam zone; the initial oil satu-ration of 88% falls to 62%, corresponding to a modestrecovery factor of 30% within the depleted zone. If weassume that recovery is 60%, then the resultant change ineffective P-velocity of this new material is only a fewpercent higher. Compared to an initial velocity of 2150m/s and a final velocity of 1506 m/s (a 30% decrease invelocities), the error associated with such a saturationrange is negligible. The ternary diagram can be used toevaluate a property of interest for all possible saturationcombinations of a three-component mixture.

Integrating Rock Physics andReservoir Parameters for Improved

Synthetic ModelingA background or baseline acoustic velocity model was

created from a composite of logs from the general area. This“high-resolution” velocity model (Figure 5) was impreg-nated with the idealized velocity anomaly shown in themiddle panel of Figure 6.

To explain the propagation of seismic energy through anoil-depleted steam chamber surrounded by cold, untouched,virgin reservoir, a finite-difference algorithm was used to cal-culate the wavefield generated through the acoustic velocitymodel. Shots were collected every 2 m along the profile, andreceivers were placed every 2 m on either side of the shot.The poststack image was created using conventional normalmoveout (NMO) correction and stacking with a resultingcommon midpoint (CMP) spacing of 1 m and is shown inthe bottommost panel of Figure 6.

The top of the steam zone is marked by a trough event(Figure 6), as expected, because of a strong negative reflec-tion coefficient at the top. The base of the anomaly ismarked by a strong peak event produced by an increasedreflection coefficient at the bottom of the reservoir. Thesteam zone yields a large increase in amplitude and a sig-nificant time delay that may serve as a proxy for steamchamber thickness. Furthermore, this example also showsthat scattering plays a significant role in the seismicresponse with diffraction hyperbolas, complicated rever-berations, and multiple reflections from within the steamzone (Figure 6) apparent. The perturbation in the wavefieldis localized about the steam zone, and internal multiplespersist beneath its thickest part (approximately 108–118 malong the profile in Figure 6) for several cycles beneath thereservoir. Furthermore, the top and base of the steam zoneare seen to produce independent sets of remnant diffrac-tions that remain after NMO correction and stacking havebeen applied. The two sets of diffractions actually have

Figure 4. Ternary diagrams can be used to plot mixtureproperties of three-component mixtures. The black diamondin (a) indicates a pore fluid containing 20% steam, 30%water, and 50% oil. The effective P-wave velocity has beencalculated for all permutations of oil þ water þ steam sat-uration scenarios. Typical reservoir temperatures and pres-sures were used to estimate the change in P-wave velocityof oil sands from 2150 m/s before steam injection to 1506m/s after steam injection. The pressure-dependent decreasein frame bulk modulus (Kdry) is computed using a modifiedHertz–Mindlin formulation extrapolated from unconsolidatedspherical grains (c). Curves in (c) represent typical porositiesfor unconsolidated materials.

Chapter 6: Seismic Rock Physics of Steam Injection in Bituminous-oil Reservoirs 109



different curve shapes; the top set is faster (shallow hyper-bolas), and the bottom set is slower (steeper hyperbolas). Itis possible that the bottom set of diffractions are time-retarded propagation within the low-velocity steam zone,whereas the top set is not retarded and is merely scatteringfrom the outside surface of the steam zone. These observa-tions promote the need for 2D migration. Fomel et al.(2007) have demonstrated the ability to perform poststackmigration on data using diffraction information aloneand there may be some promise to using their method orother standard migration procedures to actually extractquantitative time-lapse information from such features. Theabsence of diffractions and internal multiple events leads toa simpler and ideal image of the steam chamber. For com-parison, an ideal image of the reservoir and steam zone isshown in Figure 7, where the noisy events of 2D wavepropagation are not incorporated. Here, under ideal condi-tions, the exact shape of the steam zone can be traced outand the geometry and volume of the depleted reservoir canbe accurately determined. The same is not true for the seis-mic section generated by finite difference. There is ofcourse more detail, and the physics is 2D (instead of theless complete 1D convolutional model), but the final imageis plagued with higher ambiguity. Here, the brightestamplitudes do not extend across the whole width of thesteam zone. Direct mapping of the traveltime delay oramplitude anomaly in this case would lead the interpreterto conclude that the steam zone is actually smaller than itreally is. The edge effects and resolution limits of small-scale velocity anomalies such as the one presented heremust be modeled on a case-by-case basis, and simple Fres-nel zone arguments will not be sufficient in determining themaximum horizontal resolution (Schmitt, 1999). Recently,Zhang et al. (2007) have presented an extensive study thatshows how important adequate resolution is to characteriz-ing such reservoirs.

Figure 5. High-resolution velocity modelused as input for finite-difference forwardmodeling. The background horizontal layersare derived from closely spaced wells at theunderground test facility where observationwells are spaced 40 m apart. This extremelyhigh-density wireline sampling allows for anintricate and high-resolution cross sectionthrough a steam zone.

Figure 6. The top panel shows a temperature cross section,and the middle panel shows the computed change in P-wavevelocity from a developed steam chamber. This velocityanomaly has been inserted into a background cross-sectionalvelocity model for the input to the finite-difference algo-rithm used for generating synthetic seismic data (shown inthe bottom panel). These are poststack seismic data thathave undergone conventional processing, velocity analysis,NMO correction, and stacking.




Comparison with Real SeismicData and 2D Time-lapse Imaging

Over a 5-year period, a series of 11 high-resolution 2Dlines were collected by the University of Alberta at a SAGDsite in northern Alberta. Because the steam zones were rela-tively shallow, an exceptionally small 1-m CMP spacingwith offsets ranging from 48 to 142 m (48 channels were

available and spaced 2 m apart) was used. Care was takenin the repositioning of the sources and the receivers duringthese tests, and the final sections exhibit good repeatability.Figure 8 shows a volume CMP position by traveltime bycalendar date visualization of these time-lapse data; the“brightest” amplitudes have been highlighted as green“isosurfaces.” This image shows two interesting results thatare consistent with the numerical seismic experiment: (1)

Figure 7. Comparison betweenan “ideal” seismic profile (left)using simple 1D convolution, anda more realistic seismic profileover a steam zone (right) shownin Figure 7 that was acquiredusing a finite-difference algorithm.The image on the right was highlysampled, conventionally pro-cessed, and stacked.

Figure 8. Three-dimensional representation of repeated 2D seismic (time-lapse) data collected over three steaming horizontalwell pairs at the underground testing facility. Position and two-way traveltime are plotted on the x- and y-axes respectively,and the volume of data is given a “depth” perspective by stacking the repeated sections along the z-axis (in ascending calen-dar date). The “brightest” amplitudes (positive and negative) have been rendered as semitransparent “isosurfaces.” These iso-surfaces are thought to be indicators of the lateral extent of the steam chambers. The reverberations are proportional to themagnitude of steam in the reservoir and coincide with the modeled reverberations in Figure 4. The approximate location ofthe well pairs are indicated by the black lines; however, their size and vertical separation are not to scale.

Chapter 6: Seismic Rock Physics of Steam Injection in Bituminous-oil Reservoirs 111



The amplitude increase caused by the steam is very largeand detectable, and (2) the steam anomalies are similar tothe reverberation and scattering symptoms as modeled.However, the steam anomalies are not symmetric about thewell pairs (in particular, the leftmost steam chamber isentirely asymmetric). Furthermore, there appears to be asharp lateral dropoff in the amplitudes between the steamzone, and this could be attributed to the interference of dif-fraction energy off the top of the steam zones.

ConclusionsThe oil sands of Western Canada are truly unique

materials and represent an immense resource that is only inthe early stages of exploitation. To progress the under-standing of rock physics in these settings, theoretical rockphysics relationships such as Biot–Gassmann theory mustbe modified to account for the unconsolidated form andhigh viscosity of the oils. Additionally, there is much thatremains unknown about the viscoelastic behavior of oilsand and the effect that it has on seismic wave propagation.Ideally, our synthetic full-field finite-difference experimentshould be adapted to treat shear-wave and attenuationinformation. Such modeling would provide a more realisticdepiction of wave propagation; however, there are still toomany uncertainties about attenuation and viscosity charac-teristics at seismic frequencies to promote full viscoelasticmodeling at the same scale of spatial resolution.

The results from this analysis suggest that the feasi-bility of seismic monitoring does not only depend on thethermal and mechanical related changes associated withfluid substitution, but also on the scale of the steamanomaly itself. Thorough rock physics modeling can aidin the long-term survey design for monitoring reservoirdepletion. Parameters such as spatial sampling, optimalfold, repeat time intervals, and source type, to name onlya few, can all be evaluated before first steam. Futurework is required to correlate these seismic measurementsto the thermocouple profiles from observation wells (e.g.,Birrell, 2003), after which time the full extent and utilityof such highly sampled data can be ascertained.

AcknowledgmentsThis research was funded by National Sciences and

Engineering Research Council and earlier contracts fromAlberta Oil Sands Technology and Research Authority.

ReferencesBirrell, G., 2003, Heat transfer ahead of a SAGD steam

chamber: a study of thermocouple data from phase Bof the underground test facility (Dover project): Jour-nal of Canadian Petroleum Technology, 42, 40–47.

Butler, R., 1998, SAGD comes of age!: Journal of Cana-dian Petroleum Technology, 37, 9–12.

Butler, R. M., and D. J. Stephens, 1981, The gravity drainageof steam-heated heavy oil to parallel horizontal wells:Journal of Canadian Petroleum Technology, 20, 90–96.

Eastwood, J., 1993, Temperature-dependent propagation ofP- and S-waves in Cold Lake oil sands: comparison oftheory and experiment: Geophysics, 58, 863–872.

Fomel, S., E. Landa, and M. T. Taner, 2007, Poststackvelocity analysis by separation and imaging of seis-mic diffractions: Geophysics, 72, no. 6, U89–U94.

Li, P., and R. J. Chalaturnyk, 2006, Permeability varia-tions associated with shearing and isotropic unload-ing during the SAGD process: Journal of CanadianPetroleum Technology, 45, 54–61.

Schmitt, D. R., 1999, Seismic attributes for monitoring ofa shallow heated heavy oil reservoir: a case study:Geophysics, 64, 368–377.

Wang, Z., and A. Nur, 1990, Effect of temperature on wavevelocities in sands and sandstones with heavy hydro-carbons: SPE Reservoir Engineering, 3, 158–164.

Wong, R. C. K., 2003, A model for strain-induced per-meability anisotropy in deformable granular media:Canadian Geotechnical Journal, 40, 95–106.

Zhang, W. M., S. Youn, and Q. Doan, 2007, Understand-ing reservoir architectures and steam-chamber growthat Christina Lake, Alberta, by using 4D seismic andcrosswell seismic imaging: SPE Reservoir Evaluation& Engineering, 10, 446–452.


seismic rock physics of steam injection in bituminous-oil...

Documents