Monday, 30 March, 2026/Edition 104

Lithium, a battery-grade material, is a key ingredient for global electrification. Most of the world’s lithium is mined from China and Australia in the form of spodumene from rocks; however, oilfield-produced brine is also an important nonconventional source of lithium.

Each day, about 250 million barrels of water are produced from oil and gas fields globally. Although the produced waters vary chemically, some of them do contain considerable lithium content that can be extracted before the waters are reinjected or processed for other uses.

Let’s take a look at some new studies and data on lithium extracted from oilfield-produced brine.

Rasoul Sorkhabi

Editor, Core Elements

Bakken Source for Oilfield Brines in the Williston Basin

Bakken Formation core and stratigraphy. Wikipedia Commons/USGS

Kyle Henderson and colleagues published a study in Chemical Geology that links lithium in oilfield brines to shale diagenesis in the Williston Basin’s Bakken and Three Forks formations.

What they did:

• The researchers collected 255 drill core samples from the three members of the Bakken Formation and Three Forks Formation, both of Late Devonian age.

• They analyzed the core samples for major and trace element concentrations using ICP equipment and for Rock-Eval pyrolysis data.

• The researchers collected 23 brine samples from the Bakken and Three Forks pay zones.

  • They also analyzed the brine samples for their major cation and trace elements using the ICP equipment.

What they found:

• Brines from the Bakken samples have a median lithium concentration of 50.1 milligrams per liter (parts per million), which is about 250 times greater than that of normal seawater (about 0.2 parts per million).

• Lithium concentrations in the Bakken shales decrease from about 90 parts per million along the basin margins to about 70 parts per million in the thermally mature interior, corresponding to a loss of approximately 20 percent of the bulk lithium.

• Monte Carlo mass-balance calculations using formation thickness, porosity, water saturation, and shale lithium content showed that diagenetic lithium loss from the shales can reproduce Bakken brine lithium concentrations.

• Monte Carlo simulations could not match the lithium concentrations in the Three Forks Formation brine, which require additional sources for the brine.

Why it matters: What controls the enrichment of lithium in brines is a big question that impacts lithium exploration. Several factors include:

1. Interaction of fluids with silicate rocks

2. Dissolution of lithium-bearing evaporites

3. Diagenetic modification

This study provides data for the third mechanism, at least for the Bakken formation and brines.

• Studies of this kind are important for understanding the lithium mobility and source-sink framework in oilfield brines and sedimentary basins.

Too much subsurface data with not enough clarity?

Join Spotfire leaders Drew Scherer and Alessandro Chimera for a free webinar presentation, as they walk through real upstream use cases to demonstrate how modern analytics and AI are helping geoscientists and engineers turn complex data into faster, more confident decisions.

U.S. Geological Survey, Lithium Brine Data, and Machine Learning Studies

Image source: USGS

In 2022, the U.S. Geological Survey included lithium as one of the critical minerals. Lithium’s position has remained the same in the 2025 USGS list.

Here are two new USGS studies on brine lithium resources.

U.S. Geological Survey National Produced Waters Geochemical Database Viewer Version 3.0

• USGS has published a new version of its geochemical database of U.S.-produced waters.

• The database contains 113,135 samples for injection, geothermal, coal, shale, and sedimentary well plays.

• The geochemical data can be queried for various elements, including lithium.

• The database also includes identification and location information, well descriptions, dates, rock properties, physical properties of the water, organic chemistry, and more.

Go deeper: Read this USGS publication by Blondes and colleagues.

Machine Learning Applied to Lithium Brines

Emil Attanasai and colleagues from USGS have published a paper in Data Science in Science that applied machine learning (ML) methods to predict lithium concentrations in U.S. oil and gas fields.

Data Sources:

• USGS National Produced Waters Geochemical Database (PWGD) Version 3.0

• Pennsylvania Department of Environmental Protection (PDEP) Oil and Gas Well Waste Reports

Researchers applied four ML algorithms:

1. Random Forest (RF)

2. Gradient Boosting Trees (GBT)

3. Extreme Boosting (XGBoost)

4. Deep Neural Networks (DNN)

These four algorithms are commonly applied to classification and regression problems.

What they did:

• The researchers investigated whether ML algorithms applied to a suite of geochemical concentration data (Ba, Br, Cl, K, Mg, Sr) may be used to predict the lithium concentration of an unknown sample.

• They then applied an estimated economic lithium grade cutoff of 150 parts per million and assessed the utility of ML to predict whether a produced water sample would fall above or below the grade cutoff.

Shale formations in the study:

• The Marcellus Shale: 168 samples with a lithium median value of 59.6 parts per million (USGS Database)

• The Smackover Formation: 160 samples with a lithium median value of 97.5 parts per million (USGS Database)

• The Utica Shale: 46 samples with a lithium median value of 53 parts per million (USGS Database)

• The Marcellus Shale, North-central eastern Pennsylvania fields, with a lithium median value of 210 parts per million (PDEP Database)

• The Marcellus Shale, Southwestern Pennsylvania fields, with a lithium median value of 120 parts per million (PDEP Database)

What they found: With appropriate tuning, the ML approaches can be applied:

1.  In a classification sense, to accurately assign produced water samples to their formation or the producing area of origin

2. In a predictive sense, to estimate the concentration of lithium in produced water from related constituents

AAPG’s Most-Cited Paper

Soloplayer13892/Shutterstock.com

Last week, I asked: What is the most cited paper published among all AAPG publications?

My research shows that the winner with 10,891 citations (as of March 26, 2026) is: “Classification of carbonate rocks according to depositional texture” by Robert Dunham in the AAPG Memoir No. 1 in 1962.

Next week, we will look at the longest paper ever published in AAPG Bulletin! Any guesses? Write me at editorial@aapg.org

Want to help AAPG grow?

Consider supporting AAPG's free resources, like this one, by donating today.