Rock Physics Constraints for Oil Sands Seismic Interpretation

Quick Nav

• Define the constraint before interpreting the seismic
• Set the interpretation objective and the decision boundary
• Assemble a traceable rock physics data package
• Condition sonic, density, and lithology logs before modeling
• Choose the mineral and frame model for oil sands conditions
• Model bitumen, water, and gas effects without overextending Gassmann
• Generate elastic scenarios that match real reservoir questions
• Forward model the expected seismic response
• Calibrate, validate, and log discrepancies
• Report uncertainty without overstating mapped classes
• Citations
• Copy-forward worked example

Define the Constraint Before Interpreting the Seismic

A rock physics constraint is a calibrated relationship between reservoir properties, elastic properties, and the seismic response used to limit plausible interpretations.

That definition matters in oil sands work because the seismic answer is rarely unique. Porosity, shale content, water saturation, bitumen properties, gas, temperature, and pressure can overlap in P-impedance and Vp/Vs space. A clean bitumen sand at one temperature can move toward the response of a different lithology when viscosity, gas, or shale changes.

The constraint does not make seismic prove bitumen saturation, permeability, or economic pay. It narrows the interpretation. It says which geological explanations remain physically consistent with the logs, core, fluids, pressure, temperature, and seismic bandwidth in hand.

Bottom Line: Treat rock physics as a boundary-setting method, not as a substitute for core, pressure, production, or reservoir engineering evidence.

Working depth and temperature frame

For baseline modeling, keep the depth and thermal context explicit. The reference intervals here sit between about 180 and 520 m TVD, with baseline temperatures from roughly 8 to 65 C. Those ranges shape fluid behavior, effective pressure, and the chance that elastic clusters will separate cleanly.

Set the Interpretation Objective and the Decision Boundary

Name the decision before building the model. The workflow changes depending on whether the task is reservoir presence, top McMurray picking, net sand discrimination, shale barrier mapping, gas flagging, steam-chamber monitoring, or bitumen-quality screening.

A useful objective has a decision boundary. “Map better sand” is too loose. “Separate clean bitumen-bearing sand from shale baffles within a target McMurray interval” is workable because the expected elastic contrast can be tested.

Direct elastic constraints versus inference

• More direct: P-impedance and Vp/Vs changes tied to velocity, density, porosity, clay content, and fluid effects.
• Partly inferential: net sand, shale continuity, fluid class, gas risk, or heated interval state.
• Not directly solved by seismic alone: permeability, bitumen grade, completion quality, or economic pay.

In the reference setup, target attributes are limited to P-impedance and Vp/Vs within a 5-45 Hz bandwidth. That keeps the exercise honest. If the seismic cannot carry the frequency content needed to see the contrast, the map should not pretend otherwise.

Assemble a Traceable Rock Physics Data Package

Start with provenance, not equations. A model that cannot be traced back to a well name, run date, tool type, correction history, and depth reference will create arguments later, usually when the map has already influenced a drilling or monitoring decision.

The minimum well-side package includes compressional sonic, shear sonic where available, bulk density, neutron porosity, gamma ray, resistivity, caliper, core porosity, core grain density, lithology descriptions, fluid data, pressure, and temperature. Compressional and shear sonic plus core grain density anchor the volumetric balance before any substitution.

Seismic-side inputs should include migrated gathers or angle stacks, the processing report, wavelet estimate, time-depth relationship, horizon framework, and inversion products if they already exist.

Provenance fields to lock down

• Well name, license identifier, and datum
• Run date and tool type; included wells in the reference package span 2017-2021
• Edited intervals, environmental corrections, and bad-hole flags
• Core-analysis vendor, depth reference, and unit system
• Pressure data; the reference range is about 2.8-4.1 MPa

A University of Calgary graduate seminar should be able to rerun the package from the metadata alone. That is a useful standard for conference-ready interpretation: not ornate, just auditable.

Condition Sonic, Density, and Lithology Logs Before Modeling

The hypothesis is simple: most false impedance stories start as small log problems. The method is also simple, though not quick. Depth-match logs to core, flag bad-hole intervals with caliper and density correction, remove obvious sonic cycle skips, and preserve all edits in an audit curve.

In the reference conditioning pass, cycle skips were removed over roughly 0.8-2.4 m intervals, and audit curves retained edits at about 0.15 m sampling. That sampling matters because later upscaling can smooth the evidence of the correction while leaving the elastic consequence intact.

Prioritize density and sonic QC before mineral modeling. A density artifact at bed scale can become a convincing P-impedance contrast after upscaling, especially where thin McMurray facies stack tightly.

Important: Do not replace poor log values just because the curve looks untidy. Create modeled replacement curves only where justified by clean local intervals, core calibration, or a documented model assumption.

Replacement curve labels

• Empirical: derived from local clean-log trends.
• Core-calibrated: tied to measured core porosity, grain density, or lithology descriptions.
• Model-derived: generated from a stated rock physics model with listed inputs.

Choose the Mineral and Frame Model for Oil Sands Conditions

Build the solid matrix from measured or locally justified quartz-clay proportions. Do not assume a generic clean quartz sand across the McMurray interval just because the reservoir target is sand-rich.

Prior work by Han, Nur, and Morgan remains useful because it ties sandstone velocity behavior to porosity and clay content. The gap is oil sands texture. Unconsolidated grains, bitumen, and weak frames shift the response away from many conventional sandstone expectations.

The proposed approach is to test a small set of frame models against the conditioned logs and core flags: soft-sand, contact-based, friable-sand, or empirical velocity-porosity-clay trends. For friable sand cases, local coordination number adjustments are not optional bookkeeping; the reference range is 6-9, with average effective pressure near 3.2 MPa.

Field Note: Models trained on conventional reservoirs often misclassify bitumen viscosity effects outside the 10-40 C range. The frame model and fluid model need to be checked together, not in separate notebooks.

Model Bitumen, Water, and Gas Effects Without Overextending Gassmann

The fluid protocol runs in three steps. First, assign density and bulk modulus for formation water, bitumen, and gas. Next, compute effective pore-fluid properties for defined saturation cases. Then test the elastic response under stated saturation and temperature scenarios.

Heavy oil and bitumen deserve extra scrutiny. Standard Gassmann substitution assumes low-frequency elastic behavior and connected pore fluids. In bitumen systems, frequency, viscosity, and temperature can move the answer enough to change the interpretation class.

Eastwood’s Cold Lake oil sands work is a useful anchor for temperature-dependent P- and S-wave behavior. The qualification matters: field calibration is required for other deposits, production states, and seismic frequencies. In the reference fluid tests, the bitumen viscosity proxy spans roughly 150000-450000 cP at 15 C, and the frequency band spans about 8-60 Hz.

Fluid test checklist

• State the assumed temperature for every modeled fluid case.
• Keep water, bitumen, and gas properties in the same unit system.
• Separate saturation scenarios from lithology scenarios.
• Flag any case where Gassmann assumptions are doubtful because of viscosity, frequency, or disconnected pore fluids.

Generate Elastic Scenarios That Match Real Reservoir Questions

A scenario library beats a single deterministic curve. One curve invites over-reading. Discrete scenarios force the interpreter to tie elastic behavior back to lithology flags, fluid assumptions, and the decision boundary.

Recommended scenario classes are clean bitumen sand, shaley bitumen sand, wet sand, lean sand, shale baffle, gas-affected interval, and heated or partially depleted interval where relevant. In the reference package, scenarios were built as three classes per well over roughly 12-28 m intervals, then upscaled to a 4 ms sample interval.

For each scenario, calculate or compile Vp, Vs, density, P-impedance, S-impedance, Vp/Vs, and selected AVO attributes. The goal is not to populate every possible case. The goal is to cover the cases that can actually change the interpretation decision.

Scenario table logic

• One row per lithology-fluid-temperature case.
• One column for each elastic attribute.
• One field for the core or log flag that justified the class.
• One field for whether the class should be mapped, flagged, or only discussed qualitatively.

Forward Model the Expected Seismic Response

Data presentation comes first here: the reference forward modeling uses a 28 Hz peak-frequency wavelet and adds noise at roughly 8-12 percent of peak amplitude. Those two choices are enough to turn a clean log-domain separation into a marginal seismic-domain separation.

The interpretation follows. Upscale the logs to seismic scale, build synthetic reflectivity, convolve with the representative wavelet, compare near-to-far response, and test whether P-impedance and Vp/Vs separation survives bandwidth, noise, phase uncertainty, wavelet uncertainty, and tuning.

Use angle-dependent modeling when AVO or elastic inversion sits inside the interpretation objective. If the modeled contrast disappears after convolution, do not rescue it with a color bar. Move the class into a lower-confidence category or drop it from the map.

Calibrate, Validate, and Log Discrepancies

Calibration should assign every mismatch to a likely source before the interpretation is accepted. Common bins include time-depth error, phase error, anisotropy, missing shear information, wrong mineral mix, unresolved thin beds, or a fluid assumption that does not match local temperature.

The reference validation setup uses blind wells spaced roughly 1.2-3.8 km apart, with phase match tracked to within about 12 degrees. That does not certify the model. It tells the interpreter whether the synthetic, inversion attribute, and geological pick live in the same practical space.

Discrepancy log fields

• Well and interval
• Observed mismatch in P-impedance, Vp/Vs, phase, or horizon timing
• Assigned source of discrepancy
• Action taken: revise, downgrade, retain, or exclude
• Reason for the action

Report Uncertainty Without Overstating Mapped Classes

Report elastic ranges from documented wells, not from wishful interpolation. In the reference deliverables, P10-P90 ranges come from 4-7 wells. That range format gives readers a disciplined envelope without implying that every voxel has been independently verified.

Shale barrier mapping needs gamma-ray calibration above roughly 85 API. Without that calibration, a low-quality shale map can look precise while carrying the wrong lithology basis.

Deliverables should separate mapped classes from interpretive flags. A mapped class has log, core, model, and seismic support. A flag marks a possible condition such as gas effect or heated response that needs additional evidence before it drives a reservoir decision.

Bottom Line: The final product should show what the seismic can constrain, what geology must infer, and what remains a flag rather than a class.

Citations

• Han, D.-H., Nur, A., and Morgan, D., sandstone velocity work on porosity and clay effects, used here as the research anchor for velocity-porosity-clay trends.
• Eastwood, J., Cold Lake oil sands work on temperature-dependent P- and S-wave behavior, used here as the research anchor for heavy-oil and bitumen temperature effects.

Copy-Forward Worked Example

Use this template on one McMurray line with two wells.

1. Set the decision: discriminate clean bitumen sand from shale baffle across a target interval 14-31 m thick.
2. Limit the seismic attributes: use P-impedance and Vp/Vs only, and keep the test inside the 5-45 Hz interpretation bandwidth.
3. Build the well package: load compressional sonic, shear sonic where available, density, gamma ray, resistivity, caliper, core porosity, core grain density, lithology, pressure, and temperature for both wells.
4. Condition the logs: depth-match to core, flag bad hole, remove cycle skips, and store every edit in an audit curve at about 0.15 m sampling.
5. Create three scenarios per well: clean bitumen sand, shaley bitumen sand, and shale baffle. For each one, calculate Vp, Vs, density, P-impedance, S-impedance, and Vp/Vs after applying the local quartz-clay matrix and the selected friable-sand frame settings.
6. Forward model the line: upscale to 4 ms, build reflectivity, convolve with a 28 Hz wavelet, add noise at roughly 8-12 percent of peak amplitude, and tie the synthetics over an 80 ms window.

Then make the map from only the scenario separations that survive the synthetic tie. Label clean bitumen sand as the mapped class, shale baffle as the barrier class, and any unresolved overlap as an interpretation flag. The reader can copy those six steps into a project notebook and replace the wells, interval tops, and local mineral proportions with the next dataset.