ADAGEO-SUMMIT 2021 GEO-DS Experts Panel

https://adageo.github.io/summit-2021/program/

Data Science-based Geophysics: connections, challenges & research opportunities

Panel – Teasing questions

The following teasing questions will guide your 15-20 minutes panel statement. Of course, according to your expertise, you can address other issues you want to bring into the discussion. The questions have been organised according to three topics:

  • Dealing and exploiting data
  • Expectations for a possible data centred geophysics practice
  • R&D perspectives: towards revolutionary path-breakings

We state whether each topic is more addressed to geophysicists than to data scientists and computer scientists. Again this classification is just indicative.

After your panel intervention, we will open the floor to discussion among panellists and with the audience. The panel chairperson will bring other “surprise” questions, and you can also propose other questions.

Dealing and exploiting data

In a Larry Lines paper “Addressing Milo’s challenges with 25 years of seismic advances” (https://library.seg.org/doi/10.1190/1.2112389), he goes through the geophysics challenges announced by Milo Backus, SEG President, in the 1980s, and he refers to the following statement:

“Proliferation of technical information. [There is so much.] It seems to me there is an ongoing need to try to figure out how we can better synthesise all of this information. It’s overwhelming for most, and so the tendency is – I can’t go there because even if the answer I seek is available, I’ll be forever trying to figure out how to get it quickly and easily. So Library Sciences, IT and how we can continue to deliver pertinent, effective and meaningful information to the people that need it in a timely fashion is going to be a big and ongoing challenge.“

It seems that the boom of democratisation to access data collections has touched Geophysics. However, from data scientists and computer scientist perspectives, it appears that there is a small beginning in the use of Geophysicists of the considerable information content available in exploration geophysics data and results to illuminate, for example, geologic processes and geologic history.

  1. Geophysicists perspective: What is the role of data, what has the community done, e.g., standardisation, open data, experiments reproducibility? What blocking aspects prevent an aggressive automatisation of data exploration, maintenance, sharing (openly or through business models?)
  2. Data science/bases perspectives:
    1. Do scientific data, besides disorder and volume, maybe velocity, open particular challenges for the domain?
    2. Are there questions, analysis, or knowledge discovery that are challenging to implement? Are these querying exercises, or do these exploitation calls for mathematical models to be used or proposed to look for “new insight/foresight”
    3. Do requirements refer “only” to the construction, integration and maintenance of databases/warehouses/lake with advanced but still well-known techniques?

Expectations for a possible data centred geophysics practice

The book Quo Vadimus: Geophysics for the Next Generation was published by the American Geophysical Union as part of the Geophysical Monograph Series, Volume 60. It is a  monograph that contains the contributions to the Union symposium, held at the XIX General Assembly, Vancouver, Canada, in August 1987and is known as the Quo Vadimus symposium. It shows a survey about the open problems in Geophysics. One principal purpose of the seminar was to identify outstanding issues of geodesy and geophysics and thereby stimulate efforts to solve them over the next few decades. Most were already related to data or problems needing to process data.

Problemvotes
Less secrecy, more sharing / Free the data20
Improving seismic resolution17
Acceptance of error — live with it12
How to get more oil & gas from old fields9
Water — need to develop national and global understanding & policy9
How to resolve damage from resource extraction vs environmental protection7
Lack of science (E&P) acceptance in society7
Gas hydrates development R&D (×4)6
  1. What/where are the blocking aspects that have prevented scientists from proposing significant advances in these open issues since the 80’s’? Does data have a role in some blocking aspects? Which are the requirements, and why haven’t they been addressed? Is it that database and computer science solutions (algorithms, data processing) have not been mature enough to respond to these requirements? Where is the gap?
  2. Data science: Regarding “less secrecy, more sharing/free data,” which seems to be a critical challenge, there is an associated statement saying the following:

“The simplest unsolved problem comes from data acquisition where surface conditions vary locally, either wiping out reflections or changing in depth through multiples or mode converted to noise. Acquisition parameters are set to get the average solution best. Acquiring more data with more geophones with the mega channel cableless systems helps to improve the signal to noise and is a start, but comes at a cost. Interpolation has helped this issue tremendously from my experience.“

2.1 Can data acquisition be stated as a data science/databases querying/analytics problem that could help in the decision-making process to design and calibrate data collection settings that could maximise the possibility of acquiring data with given quality?

2.2 Which would be the data analytics or models challenges would quantitatively study and model the “variation of surface conditions.” 

Which type of models could help set data collection protocols and even generate synthetic scenarios to test in silico before trying in situ solutions?

3. Another requirement stated in the book Quo Vadimus: Geophysics for the Next Generation says:

“We are forever seeking better resolution (horizontal and vertical) from our seismic data so we can better image the seafloor and its subsurface and draw the remotely sensed data closer to reality. There are clearly limitations in physics that we approach, but tremendous gains have been made in acquisition and processing technologies in the last decade that takes us close to these physical limitations.“

I believe that this regards the “simulation” of synthetic scenarios that can help, predict and respond to what-if questions in virtual environments, what we wall today “digital twin” virtual spaces to reason economically safely about problems:

3.1 Geophysics: Does this exist in Oil and Energy companies? What would be the characteristics of such simulations? Which are the data that must be combined and integrated consistently to have such solutions? How would geophysicists validate the credibility of this kind of setting?

3.2 Geophysics: is it possible today to list the geophysics current and open problems that need data integration and say which collections need to be integrated to address them? To which extent are these data collections profiled concerning the format, the content of data, metrics, etc. The sense of this question is based on an interview with  Eric von Lunen in March 2017. 

 Eric von Lunen is an experienced geophysicist who is known for his reservoir characterization expertise for conventional, as well as low porosity/perm unconventional plays. He has worked at Seismograph Service Corp, Schlumberger and UT-BEG and has done extensive consulting and contracting for geophysical application research for major and large petroleum independents such as Exxon-Mobil; Aramco; Hunt Oil; Marathon; and YPF-Maxus amongst others. His focus on value-oriented geophysical method applications has been particularly noteworthy. (https://csegrecorder.com/interviews/view/interview-wth-eric-von-lunen)

He answered the following question (https://csegrecorder.com/interviews/view/interview-wth-eric-von-lunen):

As you have championed the benefits of integrated shale resource characterisation, I would like to ask you how much value you think multicomponent seismic data can bring to such an exercise?

3.3 Data science/databases perspective: This challenge goes beyond data analytics, querying, and discovery and touches on a real interaction between data and models. What would be the role of mathematical models in the generation of such simulation? Are they complex enough that alternatives from heuristics like those proposed in artificial intelligence can have a crucial role here? Where are we concerning state-of-the-art, and to which extent could we in the short term have solid solutions?

R&D perspectives: towards revolutionary path-breakings

  1. What directions in which some of the R and D worldwide are focused on the geophysics industry? What is your impression about (a) the essential developments that people can expect in geophysics shortly (b) anything path-breaking that we can expect that would revolutionise things?
  2. What are the directions in which some of the R and D worldwide focus on data science, databases, and computing science regarding energy, geology, and experimental sciences? What is your impression about (a) the essential developments that people can expect in cross-disciplinary domains soon (b) anything path-breaking that we can expect that would revolutionise things?

References