You are Home   »   News   »   View Article

How to use gravity with seismic - Geosoft

Friday, November 20, 2015

Geosoft has developed an effective modelling strategy for using gravity data together with seismic in subsalt interpretation.

There are many big oil and gas reservoirs beneath salt (including offshore Brazil), but oil and gas companies are often challenged when it comes to imaging the subsalt using seismic data only, because of the strong velocity contrast at the salt-sediment interface.

Gaud Pouliquen, technical analyst with Geosoft, explained how you can use gravity data together with seismic data to get a better understanding of the subsurface.

As a starting point you can use seismic data to understand the rock above the salt, and use gravity to work out how thick the salt layer is (since salt has a lower density to the rock around it), and then use that together with your seismic data to understand the rock beneath.

To produce a model of the subsurface from the seismic data, you need to build a velocity model, in order to convert or migrate the seismic data from time to depth. Subsalt imaging often requires several iterations of migration and interpretation to produce a reliable velocity model and ultimately a subsurface model. To reduce the number of iterations in the velocity model building process, a 3D velocity model can be easily converted to a 3D density distribution, and gravity data can then be used to determine the base of the salt-sediment interface.

Traditionally, inverting gravity data to recover an interface or layer between two distinct density domains is done through a layered Earth approach where each domain is defined by relief surfaces. Within these layers, density is defined using a constant value or a density-depth function (e.g., to represent compaction). This type of approach has been used to invert on the base of salt. Although it is a fast method, it lacks the flexibility needed to account for complex geometry such as salt bodies (best represented by triangulated surfaces) and 3D density distributions.


An alternate approach to layered modelling is 'voxel modelling' where you split the subsurface up into millions of tiny cubes (the word 'voxel' is a join-up of 'volume' and 'pixel').

'It gives us far more flexibility in terms of representing complex geometry,' she said.

However the voxel based approach also has challenges to overcome: it is more demanding in terms of processing time, is limited by non-uniqueness (i.e., many different models can fit the data) and voxel-based inversions tend to produce smooth transitions between domains rather than a sharp interface.

Hybrid approach

Geosoft has been working on a method which gets the best of both worlds (hybrid), called 'Voxel Assisted Layered Earth Modelling' or VALEM.

VALEM introduces more flexibility to build the subsurface model. It allows for three different types of data structures within a single model: layers, voxels and 3D triangulated surfaces. The latter can describe complex salt bodies, while densities are described by a full 3D density voxel derived from the seismic model.

With the voxel based approach, it is also easier to add in other constraints, based on what you know about the geology, and therefore to address the challenge of non-uniqueness. And in order to sharpen the transition between the salt and the sediment densities during the inversion, VALEM uses an Iterative Reweighting Inversion (IRI) focusing technique. At the end of the gravity inversion you end up with a relief surface of the base of salt.

Base of salt inversion

'Working out the position of base of salt is an important part of the seismic interpretation workflow,' Gaud said.

VALEM can be used to fine tune your velocity model or discriminate between different models, or you can let VALEM do the first attempt at working out the base of salt.

In that case the VALEM inversion is fed with a residualised density model where you can remove everything from the model that you know (e.g., what is above the top of salt). The inversion area is then constrained within the salt boundaries (i.e., you assume that you know the extent of the top of salt from the seismic).

Testing it

The process was tested out using a synthetic salt model created by an SEG (Society of Exploration Geophysicists) research committee in 1996, designed to be a typical US Gulf coast salt structure.

The process is to start with an initial velocity model which contains a top of salt but is "flooded" with sediment below the top salt.

Then you use this starting model to calculate what the gravity would look like if the model were correct, compare it to the observed gravity (i.e., the gravity anomaly generated by the true salt model), and calculate the difference or misfit between the two.

You can send that difference to VALEM and invert for a base of salt that will minimize the misfit.

Then you can check how gravity computed from the model compares to actual gravity data and validate the recovered base of salt.

You might be slightly overestimating or underestimating the salt thickness in the deepest part of the salt body but there is essentially a very good match between the true model and the model recovered by VALEM.

Looking at real world data, OMV tested VALEM using seismic and gravity data from offshore Africa, with water depth of 500 - 1200m.

'The seismic imaging was struggling with the salt and a high density in the crust', she said. 'Using this process, they recovered a plausible base of salt, and sent it back to the seismic team for validation.'

'For OMV it was a way to add value to the data at pretty low cost,' said Gaud. 'I should probably remind you that gravity field data is pretty cheap compared to seismic.'

'We've been testing VALEM with FTG (full tensor gravity) data. It's not something we commercialize yet, but something to introduce in the future, alongside magnetic data inversion for sub-salt,' she said.

'Geosoft offers a service to run the computer processing in the cloud, via Microsoft Azure service,' she said. 'So you don't need your own high performance computer to run it.'

Associated Companies
» Geosoft
comments powered by Disqus


To attend our free events, receive our newsletter, and receive the free colour Digital Energy Journal.


Latest Edition Jul-Aug 2021
Jul 2021

Download latest and back issues


Learn more about supporting Digital Energy Journal