Prompt Like a Pro: Get More from EarthDoc AI

The Conventional EarthDoc vs. EarthDoc AI

• Conventional EarthDoc: This is essentially an SQL-driven database with a web interface. It has been around for many years; people are accustomed to it and have learned to live with its limitations. It simply finds data, if available in the database, and retrieves it to your screen as a list of links. There is no explanation of relevance, no intelligent background information, and nothing to help you digest the answer you receive. You must do all the analysis, draw the conclusions, and sort the information yourself.
• EarthDoc AI: This new system, on the other hand, does not retrieve information for you per se. Instead, it constructs an answer dynamically, providing a description and a conclusion in the chat pane alongside the documents it based its response on. This is what we call 'grounding'. 

The 'Smart Colleague' Analogy 

While the 'old' EarthDoc is a search tool that produces a list of links, keep the following analogy in mind for the new EarthDoc AI: it is more like a highly experienced, intelligent colleague.

You can ask this colleague anything regarding technical geoscience as it appears at EAGE conferences and in EAGE publications. It will answer you dynamically, usually under a minute, pretty impressive for a synthesised, referenced answer, we think. This allows you to have a genuine conversation. Treat the first answer from EarthDoc AI purely as a starting point. This colleague is so experienced that he not only knows off the top of his head decades of EAGE conference papers and First Break articles (more than 70,000!), but he also understands natural language text, not just English, but other major languages and their grammar, such as Spanish, Dutch, French, Chinese, and Ukrainian etc. It is capable of understanding logic, problem-solving, tone of writing, patterns, and statistics. 

Embracing the Paradigm Shift

Once you understand the difference between the two systems, you will realise that you must ask your questions differently. Gone are the days of creating queries consisting of just a few keywords, as you would in Google or the conventional EarthDoc.

• Be Specific: Simply asking "seismic" or "helium gas" is too vague to start a conversation. You undoubtedly have a specific intention behind those topics; articulate it and let EarthDoc AI know exactly what you are after.
• Provide Context: A student requires a different explanation than a seasoned geophysicist. Start the conversation with a brief explanation of what you want and what your role is.
• Challenge the AI: Feed EarthDoc AI your actual question, challenge it to think alongside you, and explicitly instruct the application to ask you questions or make suggestions if you wish to dive deeper into the topic.

Let’s give it a try. Imagine the following scenario:

The User: Dr. Aris Thorne, a senior geophysicist at an energy company.

The Subject: Niche but Critical Issue. When dry supercritical CO₂ is injected into a saline aquifer, it evaporates the formation water, leaving behind salt (halite) near the wellbore. This changes the porosity and the acoustic properties of the rock.

The Goal: Aris wants to deepen his understanding of this particular critical issue.

How Dr. Aris Thorne Interacts with EarthDoc AI

Instead of just asking for a generic word like CCUS, Aris leverages EarthDoc AI, a geoscience-focused assistant grounded in the EAGE repository. Knowing EarthDoc AI has access to decades of applied geophysics articles. Aris wants the relationship explained between the chemical reaction and the resulting geophysical signal.

Dr. Aris Thorne:

I am an experienced geophysicist and want to deepen my knowledge about CCUS and specifically about how does CO₂-induced salt precipitation in saline aquifers affect near-wellbore acoustic impedance, and what are the limitations this creates for quantitative 4D seismic interpretation? Please explain the relationship between the chemical reaction and the resulting geophysical signal as detailed as possible.

Answer coming from EarthDoc AI:

Objective:

This review synthesises the geophysical and geochemical impacts of CO₂-induced salt precipitation in saline aquifers, specifically detailing how these chemical reactions affect near-wellbore acoustic impedance and limit quantitative 4D seismic interpretation.

Key Concepts & Methods

CO₂ injection into high-salinity aquifers disrupts the thermodynamic equilibrium between the formation brine and minerals. Acidification triggers geochemical fluid-rock interactions, which can accelerate mineral dissolution or precipitation. Near the wellbore, continuous injection of dry CO₂ vaporises water from the surrounding brine, leading to formation dry-out and the precipitation of halite. Predictive workflows integrate geomechanical and compositional reservoir flow simulators with petro-elastic models (PEM) based on Gassmann fluid substitution equations to capture these coupled processes.

Findings & Comparisons

The precipitation of halite within the pore space alters both hydraulic and elastic rock properties. Salt precipitation causes pore clogging, which significantly reduces injectivity near the wellbore. From a geophysical perspective, the crystallization of salt within the rock frame modifies the rock stiffness and shear properties. Chemically driven reservoir alterations, including salt precipitation and mineral dissolution, can induce seismic velocity variations of comparable magnitude to geomechanically induced deformation. While fluid substitution of saline water by supercritical CO₂ typically causes a softening effect; decreasing P-wave velocity and reducing acoustic impedance by up to 15%; the precipitation of solid halite acts as a cement or pore-filler, resulting in a localized hardening effect that increases rock stiffness.

Practical Implications & Limitations

The main limitations for quantitative time-lapse monitoring are:

• Signal Ambiguity: The acoustic softening from CO₂ saturation is counteracted by the hardening signature of salt precipitation, complicating quantitative saturation mapping.
• Saturated Sensitivity: Acoustic impedance displays low sensitivity to changes in gas saturation at high saturation thresholds, typically exceeding 20% 6.
• Interference of Non-Isothermal Effects: Temperature changes alter the thermophysical properties of the CO₂ and the evaporation rate, significantly influencing the spatial distribution of salt crystallization.
• Resolution Limits: Small-scale chemical alterations and thin CO₂ accumulations often fall below the seismic tuning thickness, restricting accurate localized quantification.

Short Conclusion

CO₂-induced halite precipitation near the wellbore increases rock stiffness and opposes the acoustic impedance reductions driven by fluid substitution. This chemical interaction introduces significant ambiguity into 4D seismic datasets, limiting the accuracy of quantitative fluid saturation and pressure profiling.

This generated explanation is followed by a list of the grounding documents. Seven in total. They are not shown here.

Now, Aris could dive deeper into the topic and challenge EarthDoc AI. We won't do that here; I think you understand the concept by now. The example above is authentic output from EarthDoc AI, intended to clarify the value of context when writing a good prompt, rather than providing a geophysically correct answer per se. I defer to the domain experts for this.

I encourage you to replicate the example, or, even better, come up with your own specific analytical challenges. The EarthDoc AI approach might save you quite some time.

Peter Verweij is Chief Information Officer at the European Association of Geoscientists & Engineers (EAGE).