Sands of time: AI helps uncover the history of Earth’s most common grains

Stanford researchers have created an artificial intelligence tool called SandAI, designed to uncover the history of quartz sand grains dating back hundreds of millions of years.

Using SandAI, scientists can accurately determine whether the sand grains were shaped and deposited by forces like wind, rivers, waves, or glaciers. This breakthrough offers a new way to explore the geological history of sand with impressive precision.

“Working on sedimentary deposits that haven’t been disturbed or deformed feels about as close as you can get to being in a time machine – you’re seeing exactly what was on the surface of Earth, even hundreds of millions of years ago.”

“SandAI adds another layer of detail to the information we can pull from them,” said Michael Hasson, lead author of a new study demonstrating the tool.

Analyzing sand more effectively

The SandAI tool offers a novel approach for geological and archaeological studies, especially for periods or environments where other evidence, like fossils, is scarce. Its advanced microtextural analysis provides a clearer window into the past, revealing how sand grains were shaped by natural forces such as wind, rivers, waves, or glaciers.

SandAI’s machine learning-based method makes the traditionally subjective and time-consuming process of microtextural analysis more quantitative and consistent, analyzing individual sand grains for a comprehensive assessment.

“Instead of a human going through and deciding what one texture is versus another for sand grains, we are using machine learning to make microtextural analysis more objective and rigorous,” said Mathieu Lapôtre, the senior author of the paper. “Our tool is opening doors for microtextural analysis applications that were not available before.”

In addition to historical studies, SandAI can assist in modern forensic investigations, particularly against illegal sand mining. Sand is the most widely used resource after water, critical for construction materials like concrete and mortar, which require specific types of angular sand for proper strength.

90% accuracy achieved

To develop SandAI, the Stanford team utilized a neural network designed to mimic the learning process of the human brain. In this system, correct predictions strengthen connections between artificial neurons, allowing the model to learn from its mistakes.

Collaborating with experts worldwide, researcher Yonaton Hasson gathered hundreds of scanning electron microscope images of sand grains from various environments, including rivers (fluvial), windblown deserts (eolian), glaciers, and beaches.

Hasson explained the goal in the press release : “We wanted this method to work across geological time, but also across all of the geography that we have on Earth.” This required including examples of different conditions, such as both wet and dry windblown dunes, to ensure SandAI could accurately analyze a broad range of sand grains.

The neural network trained itself by analyzing these images, identifying subtle features that might escape human researchers. SandAI iteratively refined its predictions, eventually achieving an impressive 90% accuracy.

The researchers then tested the model with new samples from well-characterized environments, ranging from the present to about 200 million years ago during the Jurassic period. SandAI successfully traced the transport histories of these grains, demonstrating its robustness.

The team has made SandAI accessible online for public use and plans to enhance the tool based on feedback from users. They are excited to see how it will be applied across various scientific and forensic fields.

The study has been published in the Proceedings of the National Academy of Sciences .