PAST,

PRESENT AND

FUTURE

TRENDS OF SEISMIC ANALYSIS40 YEARS OF RESEARCH AND DEVELOPMENT, IMPLEMENTATION AND

EXPLOITATION OF SEISMIC ANALYSIS – AND WHERE ARE WE?

40 Years of Seismic Stratigraphy

The application of seismic data to stratigraphy and depositional

systems analysis has been widespread at least since the

publication of “AAPG Memoir 26” from 1977, 38 years ago.

The memoir was based on a Research Symposium on Seismic

Stratigraphy presented at the “1975 national convention of the

American Association of Petroleum Geologists”.

It has gone 40 years since this convention, 40 years of progress

and time for research on the topic.

Charles E. Payton, Special Editor of the Memoir 26 wrote:

“ Seismic stratigraphy is one of the fastest growing geoscience

disciplines”

Integration of Seismic Stratigraphy

and Seismic Geomorphology

Geologic interpretation of Seismic data can be done either as an analysis:

• of cross-section views, or seismic stratigraphy.

This has been the traditional approach extracting geologic insights from seismic data, especially on 2-D seismic

data, but not limited to such.

or

• of plan-view images, or seismic geomorphology.

This approach necessarily involves 3D seismic data and is the basis for the analysis of the geological significance

of identified landforms.

or

• combine seismic stratigraphy and geomorphology

Clearly, integrating insight derived from stratigraphic and the geomorphologic analyses provide the most robust

geologic interpretations.

Future Trends in 3-D Seismic

Analysis: The Integration of

Seismic Stratigraphy and

Seismic Geomorphology By

Henry W. Posamentier(2004)

11 years ago Henry W. Posamentier wrote:

“Only relatively recently has the emphasis shifted to 3-D seismic, with sometimes astonishing results. In some instances entire depositional systems with discrete depositional elements can be directly imaged, resulting in highly accurate predictions of lithofacies relationships in time and space. Such direct imaging of geology has resulted in refinement of depositional models, especially within the context of sequence stratigraphy.”

The word, Sequence Stratigraphy was used

in the context of Seismic Stratigraphy.

3-D Yields Strat Geologic Insights

Points from the Geophysical Corner column in AAPG Explorer, February, 2004 entitled “3-D

Yields Strat Geologic Insights”

• The power of visualization

• Value of integrating seismic stratigraphy and seismic geomorphology

• Seismic geomorphological analysis of stratigraphic unit reveals geological well-known

geometries, allowing for geological interpretation of geological processes

• Seismic stratigraphy enable us to identify locations of geological seismic sequences of

interest to be integrated with the geomorphological analysis results.

• Combined, it is possible to reveal geological seismic defined systems, seismic facies and

potential geological environments.

This was said already in 2004, based on understanding already established 40 years ago.

So what has happened since?

What has happened within

Seismic Stratigraphy and Geomorphology?

• The power of visualization has become dramatically better due to computer and graphics

power

• integrating seismic stratigraphy and seismic geomorphology is better understood and

empirical data gives us a waste amount of examples to use to compare with in new

datasets.

• Seismic geomorphological and stratigraphic analysis of stratigraphic unit is made more

accurate and better defined with the massive amounts of seismic attributes available to

define an seismic image suitable for specific geological features.

• Seismic imaging techniques and processing capabilities have made it possible for us to have

higher resolution on seismic and also enable us to perform integrated rock physics products

which can approach the lithofacies domain of the data too.

• Seismic and geological modeling domain approaches each other, as resolution aspects

getting narrower and narrower, albeit not yet there.

• Computing power combined with better mathematical algorithms to utilize computers within

the learning domain, provide us with a potential for the future

What will happen within

Seismic Stratigraphy and Geomorphology?

The value of integrating seismic stratigraphy and seismic geomorphology will become

increasingly more valuable utilizing learning machine technologies and

methodologies.

Challenges in acquisition and processing will still have challenges to obtain quality

and resolution required to obtain a 1:1 relationship between seismic data and

geological models.

Handling of the big Data has already become a challenge on how to obtain

analytical value of already acquired data and new data to come in more forms and

shapes than before. How to integrate all this data to obtain optimal results from our

seismic stratigraphic and geomorphology analysis to obtain accurate and higher

chance of success identification of hydrocarbon traps, which is the ultimate goals

within the oil and gas industry.

What will happen within

Seismic Stratigraphy and Geomorphology?

The value of integrating seismic stratigraphy and seismic geomorphology will become

increasingly more valuable utilizing learning machine technologies and methodologies.

Challenges in acquisition and processing will still have challenges to obtain quality and

resolution required to obtain a 1:1 relationship between seismic data and geological models.

Handling of the big Data has already become a challenge on how to obtain analytical value

of already acquired data and new data to come in more forms and shapes than before

(attributes, derivatives, pre- and post-stack data and so forth). How to integrate all this data

to obtain optimal results from our seismic stratigraphic and geomorphology analysis to obtain

accurate and higher chance of success identification of hydrocarbon traps, which is the

ultimate goals within the oil and gas industry.

Integration with rock physics techniques and methods to approach next level, seismic

reservoir description will be a challenge, but algorithms and understanding provide us with

hopes for better results in medium to short term here.

Market Prognosis support

Big Data Challenge

From “Marine Seismic Equipment & Acquisition Markets 2015-2025” report by

Visiongain

Marine seismic acquisition submarkets

which will thrive from 2015-2025• Multi-Client Seismic Acquisition

• Proprietary Seismic Acquisition

• 3D Seismic Acquisition (3D, 4D, WAZ)

• Ocean Bottom Seismic (OBS) Acquisition ( 3C, 4C, PRM)

• 2D Seismic Acquisition

Big Data, for better or worse: 90% of world's data generated over

last two years

Over a Third of All Software Purchased Goes Unused, Reveals New

Research from 1E

How much Seismic Data is available, globally, regionwise or within

your company alone?

MUCH, AND MORE TO COME, FROM VARIOUS SOURCES AND TYPES

Big Data is an issue for

Seismic Analysis

Big data is undeniably important and is already responsible for great gains in efficiency and

knowledge. Big data is not a miracle worker, but it is changing our way we work.

How does big data differs from regular data, other than there just being a lot more of it?

Regular data doesn't magically become big data just because you've got 100 million data

points instead of a thousand.

Aside from sheer quantity, there are three defining characteristics of big data.

• Mega sourcing: taking data from huge numbers of distributed sources. Aggregate data

from many sources, satellites and third-party data sources (well, attributes, products,

service companies, IoT, NoT etc.).

• Automation: the ability to analyze data as fast as it can be collected means that the

results can be put in play automatically, without anyone having to examine the data

manually. The sheer quantity of data is becoming too great for humans to analyze even

with the benefit of extra time.

• Feedback: Big data does not measure static or pristine systems; it puts its results back into

these systems and changes their behavior. (This, naturally, makes the effects of big data

that much more complicated and dependent.)

Big Data in Seismic AnalysisAutomatization

We choose to only talk about Automatization aspect in this talk.

We are to talk what we can do about automation of data within the Seismic Analysis and

performance of Seismic Stratigraphy and geomorphology.

It is apparent that there is and will be a need to have the ability to analyze data as fast as it

can be collected and processed. There will be a need for the machines and our codes to

be able to automatically, without anyone having to examine the data manually.

The sheer quantity of data has already become and will become too great for us to

analyze even with the benefit of extra time.

This is the entry point for learning machines – or the concept of Deep learning.

Deep learning in

Seismic Analysis Challenges

The Challenge in Seismic Data Interpretation is multifold:

• 3D replaces 2D domain seismic interpretation in a larger degree than before.

• More data types creates data and attribute overload, not possible to manually screen

properly with high degree of confidence and reliability.

• Multiple surveys over same areas require governance and comparison/ calibration

which is time-consuming and full of potential traps.

• 4D domain seismic interpretation introduces a full suite of new parameters to take into

consideration.

• Amount of attributes and lack of clarity in their inter-dependencies and importance to

describe the geology or reservoirs have become overwhelming.

Deep learning in

Seismic Analysis Attributes

Seismic attributes are any measurable property of seismic data. In turn, these attributes are

input to self-organizing-map (SOM) training. Efforts distilling numerous seismic attributes into

volumes that are easily evaluated for their geologic significance and improved seismic

interpretation. Commonly used categories of seismic attributes include instantaneous,

geometric, amplitude accentuating, amplitude-variation with offset, spectral

decomposition, and inversion.

Principal component analysis (PCA), a linear quantitative technique, has proven to be an

approach for use in understanding which seismic attributes or combination of seismic

attributes has interpretive significance. The PCA reduces a large set of seismic attributes to

indicate variations in the data, which often relate to geologic features of interest. PCA, as

a tool used in an interpretation workflow, can help to determine meaningful seismic

attributes

Deep learning in

Seismic Analysis Classification

Unsupervised classification of Seismic Attributes

Classification without supervision of patterns into groups is formally called clustering.

Depending on the application area these patterns are called data lists, observations or

vectors.

For exploration geophysicists, these patterns are usually associated with seismic attributes,

seismic waveforms or seismic facies.

The main objective here is to show how one of the most popular clustering algorithms -

Kohonen Self-Organizing Maps (KSOM), can be applied to enhance seismic interpretation

analysis associated with one and two-dimensional color maps.

Deep learning in

Seismic Analysis Self-Organize

The KSOM (Kohonen, 2001) clustering is one of the most commonly used tools for non-

supervised seismic facies analysis, with KSOM providing ordered clusters that can be

mapped to a gradational color bar (Coléou et al., 2003).

KSOM is closely related to vector quantization methods (Haykin, 1999).

We assume input variables, i.e., the seismic attributes, can be represented by vectors in the space ℜn, aj = [aj1, aj2, ..., ajN], j = 1, 2, ..., J; where N is the number of seismic attributes

and J is the number of seismic traces when KSOM is applied to surface attributes or is the

number of voxels (Matos et al., 2005) when KSOM is applied to volumetric attributes.

The objective of the algorithm is to organize the dataset of input seismic attributes, into a

geometric structure called the KSOM.

Deep learning in

Seismic Analysis Train Data

The number of prototype vectors in the map determines both its effectiveness and

generalization capacity. During the training, KSOM forms an elastic net that adapts to the

"cloud" formed by the input seismic attribute data.

Data that are close to each other in the input space will also be close to each other in the

output map. Since KSOM can be interpreted as a reduced version of the input n-

dimensional data ruled by a lower dimensional grid that attempts to preserve the original

topological structure and since seismic data measures the changes in geology.

KSOM approximates the topological relation of the underlying geology.

Deep learning in

Seismic Analysis KSOM

Deep learning in

Seismic Analysis Convolutional Neural Network

The rapid rise of computer vision technology and the increasing number of companies developing

image recognition platforms are enormous.

Until recently, computer vision technology has been used primarily for detecting and recognizing faces

in photos. While facial recognition remains a popular use of this technology, there has been a rapid rise

in the use of computer vision for automatic photo tagging and classification.

This increase is largely due to recent advances in artificial intelligence (AI), specifically the use of

convolutional neural networks (CNNs) to improve computer vision methods.

So far, this technology has not won any major terrain within the Oil and Gas Industry.

Deep learning in

Seismic Analysis Pattern recognition

Stratigraphic interpretation of seismic data is a time consuming and highly subjective methodology where

the result is highly dependent upon the operators skills, training and mostly experience to recognize

depositional environments and their associated geometrical attitude and occurrence.

Combine this with varying quality of the data foundation, seismic data quality and type, there are many

ways this could go wrong.

The task at hand is to identify geometric patterns in the data, generate image captions/ descriptions

Deep learning in

Seismic Analysis Learn patterns

Why not use computer vision algorithms to analyze digitized images of seismic data (original or

attribute versions, does not matter). The algorithms could be trained to detect and understand

visual similarities in seismic facies pattern and automatically classify these based on style,

occurrence etc.

Utilize Convolutional Neural Networks (CNN) that are able to learn complex visual concepts using

massive amounts of data,, could save time and efforts, but not only that; create a more objective

analysis of the data.

The use of machine learning and image processing algorithms to analyze, recognize and

understand visual content could prove to be a ground breaking way to analyze large amount of

data, both in Supervised Neural Networks (SNN), but also as Unsupervised Neural Networks (UNN),

like the CNN.

The computer gets trained to find patterns within the data with the use of deep learning-based

computer vision technology to analyze, recognize and understand the content of an image.

Deep learning in

Seismic Analysis Observe both plane and cross section

Seismic Stratigraphic image learning Seismic Geomorphology image learning

Deep learning in

Seismic Analysis Models in plane and cross section

Seismic Stratigraphic image models Seismic Geomorphology image models

Algorithms already established in geological modeling

software.

Require some guidance with a low frequency surface

model in data to mimic dips and curvatures in stratigraphic

response of data

Train the data to recognize geometrical patterns and

utilization of “iPhoto” and “Facebook” technology and

methodology to interact with the training.

Deep learning in

Seismic Analysis Tag with “facies” recognition

Yes

No

Yes

No

Type a NameType a Name

Yes

No

Type a Name

Type a Name

Type a Name

You give input to the unsupervised training of your data. It will automatically identify similar ones

and/or give you a choice of places it finds similar, and you choose to tell its right or wrong.