Validating Dynamic Reservoir Simulation Using Rock Physics Modeling in SubsurfaceAI

When Equinor and its partners released the full Volve field dataset to the public, they provided something unique: a complete, real-world North Sea oil field, from seismic and wells all the way through static and dynamic reservoir models. Equinor+1

Volve is a decommissioned oil field in the central part of the North Sea, about 5 km north of Sleipner Øst, in 80 m water depth. It produced oil from Middle Jurassic Hugin Formation sandstones between roughly 2700 and 3100 m depth from 2008 to 2016. Energistics+1 In 2018, Equinor released around 40,000 files of subsurface and operating data—static and dynamic reservoir models, seismic, wells, logs, production, drilling data, and reports—under an open data licence. Equinor+1

This makes Volve an ideal real-world testbed for the forward rock physics workflow described in Part 2 of this series: using Rock Physics Modeling (RPM) and Rock Physicist Templates (RPTs) in SubsurfaceAI to generate synthetic seismic from static and dynamic reservoir models—and then comparing these synthetics with the real Volve seismic.

In this case study, we show how a seismic interpreter can perform the entire workflow inside SubsurfaceAI, with no coding required, using the Volve dataset as a concrete example.

Case Study Objective

Our goal is to answer a very practical question:

Given the Volve dynamic reservoir simulation and a rock physics model calibrated from Volve wells, does the synthetic seismic generated from the simulation resemble the real seismic?

If the answer is “yes,” we gain confidence that:

• The reservoir model is consistent with both wells and seismic
• The dynamic behavior (saturation and pressure changes) in the simulation is plausible

If the answer is “no,” the mismatch becomes diagnostic feedback for improving the reservoir model, rock physics assumptions, or both.

Data Used From Volve

From the Equinor Volve open data, we use: Equinor+2Offshore Magazine+2

• Well logs (P-sonic, S-sonic, density, GR, porosity logs) from key Volve wells
• Static and dynamic reservoir models, including:
  • Porosity grids
  • Oil/water saturation grids as a function of production time
  • Pressure grids for each simulation time step (Eclipse format)
• 3D seismic data covering the Volve field
• Production time steps corresponding to different stages of field depletion

All of this content is loaded into a single SubsurfaceAI project, where interpreters can navigate wells, seismic, and models side by side.

Overview of the Workflow

The workflow mirrors the eight-step process described in Part 2, distilled into six major stages for this case study:

1. Rock Physics Modeling (RPM) with Volve well data, saved as Rock Physicist Templates (RPTs)
2. Import Eclipse simulation into SubsurfaceAI, including static and dynamic grids
3. Compute Vp, Vs, and density from porosity, saturation, and pressure grids using the RPTs
4. Map elastic grids to a regular seismic grid and define 3D volume geometry
5. Generate synthetic 3D seismic volumes for each simulation time step using a wavelet or wavelet volume
6. Analyze and compare saturation and seismic amplitudes between real and synthetic data

All of this is performed inside SubsurfaceAI, using point-and-click workflows—no external tools, and no scripts.

Step 1 – Rock Physics Modeling (RPM) Using Volve Well Data

We begin with the Volve wells that intersect the Hugin Formation reservoir. These wells provide:

• P-sonic (DTp), S-sonic (DTs), and density logs
• Gamma ray and derived Vshale
• Porosity from density/neutron or core
• Fluid information (oil/water contacts, test data where available)

1.1 Select Wells and Intervals

Within SubsurfaceAI, the interpreter:

• Loads the Volve wells into the project
• Defines the Hugin reservoir interval based on tops/picks
• Selects representative wells that sample key facies (e.g., clean sand, shaly sand, non-reservoir intervals)

1.2 Build the Rock Physics Model in RPM

Using SubsurfaceAI’s RPM module, the interpreter chooses suitable rock physics formulations. For Volve’s Hugin sandstone, a typical approach might be:

• Soft-sand or stiff-sand models for the clean reservoir sands
• A Han-type sand–shale model for mixed lithologies
• Gassmann fluid substitution to model oil vs. brine conditions
• Empirical trends (e.g., Castagna, Han) to guide Vp-Vs-porosity relationships

The key point: all of these models are available and configurable through the SubsurfaceAI RPM interface—no equations to implement, no scripts required.

1.3 Calibrate to Volve Well Data

The interpreter then:

• Crossplots Vp, Vs, and density vs. porosity and Vshale
• Adjusts mineral and frame parameters until the model curves match the Volve log trends
• Uses known oil/water intervals to constrain fluid substitution behavior
• Validates that modeled elastic responses at the well level are consistent with measured logs

Once satisfied, the interpreter:

1.4 Saves the Result as Rock Physicist Template (RPT)

The calibrated Hugin reservoir model is saved as an RPT. This template encapsulates:

• Model choices and parameters
• Fluid properties
• Lithology and porosity relationships
• Saturation behavior

This RPT will later be applied automatically across the entire Eclipse grid at all simulation times.

Step 2 – Import Eclipse Simulation (Static & Dynamic Models)

Equinor’s Volve dataset includes static and dynamic reservoir models in Eclipse format, including:

• Static properties: porosity, facies or Vshale, initial saturation
• Dynamic properties: saturation and pressure grids for many production time steps

SubsurfaceAI’s model import tools:

• Read Eclipse grid geometry and property arrays
• Preserve corner-point geometry and grid dimensions
• Associate each time step with a production date

Once imported, we have a time-series of 3D reservoir property grids ready for rock physics modeling.

Step 3 – Compute Vp, Vs, and Density From Porosity, Saturation, and Pressure

Now we apply the RPT to the Volve grids.

For each simulation time step, SubsurfaceAI:

1. Reads the reservoir properties from the Eclipse model:
   • Porosity
   • Oil/water saturation
   • Pressure
   • Vshale or facies code, if available
2. Uses the RPT to convert properties → elastic parameters:
   • For each grid cell, the appropriate rock physics model is invoked
   • Porosity, Vshale, and saturation drive Vp, Vs, and density
   • Pressure is used to adjust frame moduli where relevant
3. Generates full-field 3D volumes of:
   • Vp (P-wave velocity)
   • Vs (S-wave velocity)
   • Density
4. Repeats this for each production time in the dynamic simulation.

The result: a time-series of elastic models consistent with:

• Volve well data (via the calibrated RPT)
• The Eclipse simulation’s evolution of saturation and pressure

All of this is done with SubsurfaceAI’s GUI—no coding, no external tools.

Step 4 – Define the Seismic Grid and 3D Volume Geometry

To generate synthetic 3D seismic volumes, we need to ensure that the elastic grids are mapped onto a regular seismic volume with:

• Inline and crossline spacing
• Vertical sample rate (in time or depth)
• Spatial extent that matches the real Volve seismic

SubsurfaceAI provides:

• Tools to align the model extent with the seismic survey
• Options to interpolate or resample Vp, Vs, and density onto the chosen grid
• Controls for vertical sampling (e.g., 2 ms or 4 ms)

The outcome is a set of seismic-ready elastic volumes:

• Vp(x, y, z, t)
• Vs(x, y, z, t)
• Density(x, y, z, t)

for each simulation time t.

Step 5 – Generate Synthetic 4D Seismic Volumes

Now we convert the elastic models into synthetic seismic data.

5.1 Define or Select a Wavelet / Wavelet Volume

We have two options:

• Extract a wavelet from the real Volve seismic using well ties within SubsurfaceAI
• Use a previously derived wavelet volume (angle- or space-dependent) stored in the project

The key is to ensure that:

• The wavelet is representative of the Volve seismic bandwidth and phase
• The same wavelet (or wavelet volume) is used for both base and monitor synthetic volumes, to isolate the effect of reservoir changes

5.2 Forward Model Synthetic Seismic

For each simulation time step:

1. Compute reflection coefficients from Vp and density (and Vs for pre-stack modeling)
2. Convolve with the chosen wavelet or wavelet volume
3. Generate either:
   • Pre-stack synthetic gathers (angle/offset dependent)
   • Partial stack volumes (near, mid, far)
   • Full stack 3D volumes

All of this is computed inside SubsurfaceAI, and the outputs are standard seismic volumes that can be:

• Viewed in the 3D/seismic viewer
• Sliced inlines/crosslines/time slices
• Compared directly with the real Volve seismic cubes

Now we have a synthetic 3D seismic volume for each production time in the Eclipse simulation.

Step 6 – Compare Saturation and Seismic Amplitudes (Real vs Synthetic)

This is the most important step from an interpretation and reservoir-management perspective.

6.1 Visual Comparisons

Interpreters can:

• Display real and synthetic inlines/crosslines side by side
• Blend volumes with transparency to see differences
• Step through time to visualize how synthetic seismic responds to saturation/pressure changes
• Overlay well trajectories and production data for context

For 4D analysis:

• Compare synthetic base vs synthetic monitor
• Compare real base vs real monitor
• Compare synthetic 4D difference vs real 4D difference

6.2 Quantitative Comparisons

Using SubsurfaceAI’s Seismic Volume Calculator and cross-plot tools, interpreters can:

• Compute difference volumes (real − synthetic) for each time step
• Extract amplitude vs time vs saturation crossplots
• Map seismic difference against changes in oil saturation from the Volve Eclipse model
• Identify areas where simulated saturation changes should produce a clear 4D signal—and check whether they do in the real data

Because the dynamic Volve models include time-dependent saturation and pressure grids, we can test hypotheses such as:

• Do simulated waterflood fronts produce the seismic softening or hardening we see?
• Are mapped amplitude anomalies in the real 4D data consistent with modelled saturation changes?
• Is there a mismatch between modeled pressure effects and observed seismic attributes?

Some mismatches will point to:

• Incorrect dynamic reservoir behavior (over- or under-estimation of sweep efficiency)
• Mis-positioned faults or flow barriers in the model
• Over-simplified facies or net-to-gross distributions
• Rock physics assumptions that need refinement (e.g., saturation model, frame stiffness)

Because everything—rock physics, reservoir models, synthetics, real seismic—is hosted in one SubsurfaceAI project, interpreters can iterate quickly.

What This Case Study Shows

Using Volve as a real-world example, this case study demonstrates that:

1. RPM + RPTs enable interpreters to transform reservoir properties into elastic properties, without writing a single line of code.
2. SubsurfaceAI can ingest Eclipse static and dynamic models from the Volve dataset, generating a time-series of elastic models that evolve as saturation and pressure change. Offshore Magazine
3. Synthetic seismic volumes can be generated inside SubsurfaceAI for each simulation time step, using wavelets tied to the real Volve seismic.
4. Real and synthetic seismic volumes can be compared visually and quantitatively, allowing interpreters and reservoir engineers to test whether the Volve model is consistent with seismic responses over time.
5. The entire process is interpreter-friendly, repeatable, and fully integrated, unlike traditional rock physics modeling workflows that require specialist tools and custom coding.

Why Volve Is a Perfect Testbed

Because Equinor disclosed all subsurface and operating data from Volve, including static and dynamic reservoir models and seismic, it has become one of the most widely used open datasets in our industry. Equinor+1

For SubsurfaceAI, Volve is an ideal demonstration case:

• It is a real, decommissioned field in the central North Sea, with a clear production history and well-documented data.
• The dataset includes exactly the elements needed for our Part 2 workflow: wells, seismic, static models, dynamic simulations.
• The open nature of the data lets users replicate and extend this case study themselves.

From Case Study to Daily Practice

While this case study is based on an open dataset, the real power lies in applying the same workflow to your own fields:

• Calibrate RPTs from your wells
• Import your simulation models
• Generate synthetics over time
• Validate and improve your reservoir models using seismic

SubsurfaceAI makes this workflow realistic for everyday seismic interpreters, not just rock physics specialists. It turns rock physics from an occasional, expert-driven exercise into a standard part of the seismic–reservoir integration loop.

In the next parts of this series, we’ll show how the same RPTs used here for forward modeling can also be used to:

• Convert inversion-derived Vp, Vs, and density into reservoir properties (Part 3)
• Use machine learning to automate and scale rock physics workflows (Part 4)
• Perform interactive “What-If” scenario analysis for thickness, net-to-gross, and fluid changes (Part 5)

All with the same philosophy in developing subsurfaceAI:

Designed and coded by geoscientists, for geoscientists. No coding required.