You are Home   »   News   »   View Article

Multi-physics interpretation, integration and predictive analytics at NEOS

Friday, July 1, 2016

NEOS, a company based in California’s Silicon Valley, has developed its techniques and workflows to build models of the subsurface utilising many different types of survey data, including seismic, electromagnetic, gravity, hyperspectral, magnetic, surface geology, well data and radiometric.

The company uses each dataset to image the same geology but with different physical properties, reducing exploration risk and cross constraining the datasets.

NEOS interprets and integrates the datasets; following this, predictive analytics algorithms are run to gain valuable insights, such as where the subsurface is most similar to a producing well or where there is analogous geology to a specific regional play.

Investors include Microsoft founder Bill Gates, and Silicon Valley venture capital firm Kleiner, Perkins, Caufield and Byers. Much of NEOS' work has been in North and South America, including a large study in Argentina. In addition a project onshore and near-shore Lebanon was recently completed.

Although non-seismic measurements make up a substantial proportion of NEOS' workflows, seismic data often provides the framework for building interpretations of the subsurface, providing vertical constraints to the other datasets. To improve the company's seismic capabilities, it recently purchased the onshore seismic data processing and imaging group from ION Geophysical - now called the NEOS Seismic Imaging Group (SIG) - whose advanced techniques such as AVO analysis, depth imaging and azimuthal velocity anisotropy analysis add significant value to an integrated earth model as well as being able to offer conventional seismic processing to clients.


NEOS believes an integrated analysis of many data sources can individually and collectively improve insights into the subsurface. A typical project would include the following geological and geophysical datasets;

-Electromagnetic data provides information about the resistivity of rocks, giving an understanding of lithology and hydrocarbon charge.

-Radiometric data detects radioactivity of the surface and near-surface, which can be useful for understanding shallow fractures, the total organic carbon (TOC) of shale units at the surface as well as some indirect hydrocarbon indicators.

-Gravity data measures lateral density contrast in the subsurface, which is used for structural analysis.

-Magnetic data is also used for structural analysis, identifying volcanics in the sedimentary section, as well as assessing basement heterogeneity.

-Hyperspectral sensors, by measuring absorbed and reflected light (both visible and invisible) at hundreds of points across an energy spectrum, can identify both direct and indirect indicators of hydrocarbons.

Seismic datasets, well information and surface geological mapping are also integrated into their workflow.

'The benefit of using all of these different technologies - is looking at the same thing but through different physical properties,' said James Dodson, business development director at NEOS. 'Each are individually valid, but when you bring them together you reduce uncertainty.'

'We have some datasets more applicable to shallower interpretations, some at reservoir level, some at depth.'

Once the company has manually interpreted and integrated all available data, creating a 3D earth model, the next step is to bring in predictive analytics, or statistical modelling, including some data mining techniques, to spot trends, correlations and patterns.

'This workflow is highly repeatable, completely scalable, from 1 million km2 to 100s km2,' he said. 'We've done work in some very remote regions with not a lot of data available.

We can also use huge datasets, such as with seismic data onshore US. We can integrate data of different vintages and resolutions.'

DJ Basin, Colorado

NEOS was asked to help develop subsurface understanding of the DJ Basin in Colorado, where there is a large amount of seismic data and production information from many wells.

Wells had been drilled on the basis that they were close to an already producing well, he said. But the results were variable, with wells producing between 25 and 500 bopd.

The oil company 'asked us to give them a reason to drill somewhere that's a bit more intelligent, and potentially find new areas for leasing,' Mr Dodson said.

Several airborne datasets were acquired, interpreted, integrated, and assessed to determine which data seemed most important in predicting a well's production, and how to therefore predict where the best and worst performing wells could be located for future drilling.

The output was a generated map that indicated areas to avoid and areas that could be recommended for drilling. The project team also identified a new field in the south of the project area that the oil company leased and successfully drilled, confirming the team's predictions.

Onshore & near-shore Lebanon

NEOS, in partnership with the Lebanese Petroleum Administration (LPA), has completed a project in Lebanon to better understand the regional prospectivity, including the onshore northern half of the country and the near-shore along the Mediterranean coastline, and to high-grade acreage throughout the survey area to support future leasing, drilling and G&G investments.

The project builds upon heightened exploration interest in the Eastern Mediterranean region as a whole and is aimed at identifying geological features that extend into the project area from offshore structures, Syria's onshore petroleum systems as well as several other regional plays.

For the study, approximately 6,000km2 of new data, including gravity, magnetic, hyperspectral, magnetotelluric (MT), and radiometric data was acquired. NEOS acquired, processed, interpreted and integrated the data in just 7.5 months, which included four airborne and one ground-based dataset.

The first stage of the study was to interpret the datasets individually.

The gravity, magnetic and electromagnetic data was used to get an idea of the subsurface structure.

Magnetic data was used to identify volcanics within the sedimentary section as well as basement heterogeneity.

The surface was mapped using hyperspectral and radiometric data. The hyperspectral data analysis showed that there are many oil seeps onshore Lebanon; this being exciting because it is direct evidence of a working hydrocarbon system.

The second step of the study was to manually integrate the data. Ultimately, the aim is to create a 3D model where 'everything complements each other and cross correlates,' he said.

There was a small amount of borehole data available, which could be used to show interval thickness and rock properties.

2D models were created using surface geology, strike and dip information, regional knowledge and rock property information. These models were used to forward calculate the gravity and magnetic signal, which was compared to the acquired data, with the models being altered until the forward calculated and acquired signals matched.

The hyperspectral data was integrated into the model, showing where seeps appeared at surface fault locations, within specific rock units or above subsurface structures, such as anticlines.

The third step in the process was to apply the predictive analytics techniques. The project team used geological knowledge and geophysical attributes from fields in nearby Syria, identifying regions in the subsurface of Lebanon with similar geological and geophysical properties.

'In Lebanon we found analogues to onshore Syrian fields' he said. 'We also have analogues to offshore Triassic deep gas. We have four different play types.'

Neuquén Basin, Argentina

In Argentina's Neuquén Basin, NEOS' aim was to 'high-grade' the acreage of a supermajor, i.e. try to work out which parts of the land were most likely to provide oil.

The work included estimating thermal maturation and TOC of the shales, as well as interval thickness.

NEOS acquired magnetic, gravity and hyperspectral data, and did a lot of geochemical work, analysing seeps on the ground. There was some seismic and well data available.

Magnetic data was interpreted and used to help create a geothermal model of the basin (its temperature gradient).

All of the results were integrated to create a 3D earth model; following this, a toolbox of predictive analytics algorithms were applied.

The project team was able to create a map of the Vaca Muerta shale, predicting where oil, condensate and dry gas production would be high.

The supermajor has subsequently drilled in the project area, with one well drilled in a manually high graded area, and another drilled in an area high graded by the predictive analytics methods.

'Both wells encountered oil, one of them is the best producing well they have in the basin,' Mr Dodson said. 'The other is not far behind. From our perspective, this a really good result.'

Associated Companies
» NEOS GeoSolutions
comments powered by Disqus


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


Clustering Considerations in the Machine Learning Workflow – Examples with Exploration Data
Philip Lesslar
from Precision DM


Latest Edition Apr-May 2020
Apr 2020

Download latest and back issues


Learn more about supporting Digital Energy Journal