Predicting CPT Measurements with Seismic

CPT prediction using seismic data is becoming more widely adopted, driven by the increasing number of sites being prepared to host floating offshore wind farms. The challenge with floating wind is the significantly greater number of seabed contacts required to anchor a single turbine—typically around four anchor points—compared to the single contact point used for fixed-bottom turbines. Each seabed contact requires prior knowledge of sub-seabed soil and rock conditions. However, the resulting 4x increase in geotechnical sampling costs can make development financially unviable for many floating wind sites.

How does CPT-prediction use seismic data?

This is where the industry sees seismic inversion as a potential solution. The concept behind pre-stack seismic inversion is to generate models of subsurface elastic properties, from which rock or soil types can be inferred. These models can then be calibrated against known downhole measurements—such as cone penetration tests (CPTs), P- and S-wave logs (PS logs), and borehole data. By inverting amplitudes within seismic data using established geophysical principles, we can derive elastic properties, which in turn inform predictions of geotechnical parameters. In offshore wind, where data underpins engineering design, we are producing synthetic or predicted CPTs that provide properties such as sleeve friction, cone resistance, bulk modulus, shear modulus, and density.

When the seismic wavefield intersects subsurface interfaces (e.g., boundaries between geological units), wavefield partitioning occurs. This is where P-wave energy is partially converted into both P- and S-waves. The degree of this partitioning depends on the elastic properties of the materials and the angle of incidence. Since pre-stack seismic data varies with angle of incidence, we can measure the amplitude variation across offsets (the distance between source and receiver), which tells us how much energy is converted to shear waves. This phenomenon is called Amplitude Variation with Offset (AVO).

Why don’t Final Stacks give enough information?

Seismic inversion can be performed on post-stack datasets, but these results should not be relied upon for final engineering design. It could be argued that using a stack for seismic inversion can be up to 100 times less reliable than using pre-stack data, here’s why.

Difference between stacks and gathers

When acquiring seismic data, a vessel moves over the same midpoint using multiple offset traces. The configuration depends on the number and spacing of hydrophones on the cable. For each midpoint, the traces from different offsets are collected into what is called a Common Midpoint (CMP) gather. Combining all this information produces a stacked trace.

Why stacks can hide issues

Stacking is a powerful noise-attenuation process. Signal-to-noise improvement is proportional to the square root of the number of traces in a gather. You can demonstrate this by adding random noise to a gather—no matter how severe the noise, stacking can suppress it to the point of invisibility in the final image. While the stacked image may appear clean, the underlying gathers could be contaminated with random noise that distorts amplitudes—critical for seismic inversion and CPT prediction.

Why we prefer reprocessing from gathers

When reprocessing from gathers, geophysicists can use the distinct patterns of signal and noise to differentiate and enhance useful energy. In the stack domain, these patterns are lost. You're left only with geological patterns, making it harder to separate signal from noise. Gathers provide more information and greater flexibility for processing, around 100 times more… for example, many shallow seismic surveys use a 96-channel streamer and stacking all the information received by these channels into 1 trace means you essentially have only 1% of the information remaining to play with.

Now we have established why the preference should always be pre-stack data, simply requesting gathers from your seismic provider isn’t enough. Here’s why…

Processing tailored for CPT Prediction

Seismic surveys should be designed with the end goal in mind. If the goal is to generate predicted CPTs via elastic inversion, high-quality pre-stack gathers are essential and should be explicitly requested from your seismic contractor.

That said, while many acquisition contractors can deliver pre-stack data if requested, few possess the in-house expertise to carry out high-quality, AVO-compliant pre-stack processing. For seismic data to be reliable for inversion and CPT prediction, the processing must preserve true amplitudes and avoid damaging the AVO signal. Using standard stack-oriented processing flows won’t suffice. The sequence must be re-designed from raw data with AVO-compliant steps, addressing both noise and acquisition artefacts.

The key processing steps that make a difference include:

• Sea state statics

• Deghosting

• Designature

• Demultiple

Deghosting and designature relate to wavelet fidelity in seismic data, especially in Ultra High-Resolution (UHRS) datasets. At RockWave, we apply full pre-stack deghosting using inversion-based techniques—not relying on stacking to clean up the data. This ensures accuracy in wavelet shapes and resolution.

For designature, we have deep expertise in managing complex wavefields and directional variations, which are common in 2D and 3D UHRS/EHRS surveys. Most acquisition contractors lack reliable near-field hydrophone data, which makes this step even more important. By designing designature operators that honour directional wavefield variations, we achieve a more uniform wavefield and ensure AVO compliance across all angles and azimuths.

Demultiple is another challenge—especially in rough sea conditions. While stacking helps suppress multiples, it’s not sufficient for AVO-compliant gathers. We use advanced convolution-based methods like SRME, along with iterative, adaptive matching techniques, to compensate for non-ideal sea surfaces.

Statics correction is also essential, particularly in UHRS data. We resolve static errors in a surface-consistent way, breaking them down into source and receiver components to align the data to common datums and preserve AVO signal integrity.

We also perform amplitude-preserving pre-stack migration as standard. This returns full offset gathers across the full angle range, suitable for inversion.

Additionally, we integrate specialist AVO QC analyses from the beginning of processing through to the final deliverables. These quantitative checks go beyond what’s required for stack-only products.

With RockWave’s approach, clients receive a complete pre-stack solution. While most providers focus on producing a final stacked image, that may only be 80% of what’s needed. If your goal is inversion-ready gathers for CPT prediction, 80% isn’t enough. Work with a processing specialist who can deliver a 95%+ solution—reducing engineering risk and helping developers avoid costly overdesign due to uncertain subsurface data.

What Developers need to push their seismic acquisition contractors to deliver

We’ve already discussed the importance of high standards in seismic processing. But as mentioned earlier, if developers want to use seismic data to generate predicted CPTs at any location across a site (as long as that location has been sampled by 2D or 3D seismic), then the seismic dataset must also be acquired with that end-use in mind.

In fact, all seismic surveys should be designed with their ultimate purpose in mind.

Here’s what developers should be asking of their seismic acquisition contractors:

1.     Reliable source signatures from near-field hydrophones (NFH):

Capturing clean and usable source signatures from NFH remains a challenge. As highlighted by Benedict Robbins of SUT OSIG (LinkedIn post), some of the key issues include:

• Overdriven amplitudes due to proximity to the source

• Electrical noise and interference

• Hydrophone movement relative to the source

• A lack of stable, purpose-built acquisition equipment

The industry as a whole should be pushing to overcome these barriers. In the meantime, RockWave has consistently succeeded in generating its own designature operators, as discussed earlier, to compensate for limitations in NFH data.

2.     Sensible towing depths for both source and hydrophone streamers:

It's vital to have good control over source and receiver depths—or at the very least, a thorough understanding of their movement during acquisition. This includes accounting for variations due to sea state at the time of data collection.

3.     Sufficient streamer lengths and offsets:

This matters for two key reasons:

#1: You need adequate offset to record a broad range of angles. These angles are essential for AVO analysis and seismic inversion.

#2: Inversion requires a low-frequency model to support absolute property prediction. Ultra-High-Resolution Seismic (UHRS) data is often lacking in low frequencies, which limits inversion results to being relative (i.e., comparing contrasts between layers) unless supplemented. For absolute inversion—where predictions can be tied to real-world geotechnical measurements—you need a reliable low-frequency model.

In UHRS, this low-frequency content typically comes from the velocity model or, if possible, from a Q model, provided that Q (attenuation) can be reliably extracted across the dataset.