AI Model Reconstructs Earth's Biodiversity History
Researchers developed a machine-learning tool to map 300 million years of marine species evolution and extinction.
Updated on Oct. 20, 2026 in Environmental
Live Poll
Do you trust AI-based models to provide accurate insights into historical or scientific data?
Scientists have introduced a machine-learning model called OneStrata that analyzes fossil records to create a high-resolution timeline of Earth's biodiversity. The model reveals that three of the five major mass extinctions likely occurred through gradual decline.
Why it matters
Previous models struggled to process massive datasets, but this computational breakthrough allows for the analysis of thousands of species over millions of years. This tool provides a clearer understanding of how life has changed over deep geological time.
The OneStrata model utilizes the OneStratigraphy dataset to achieve a 26,000-year resolution for estimating species appearance and extinction. This follows a 2020 algorithmic method that successfully reconstructed timelines for 11,000 marine species.
The players
Shuzhong Shen
He is a researcher based in Nanjing, China, who led the presentation of the new biodiversity modeling research.
Geological Society of America
This is a global professional society dedicated to the geosciences that hosts major annual meetings for scientific exchange.
The details
Researchers trained the OneStrata model on statistical patterns within fossil deposits found at tens of thousands of sites worldwide. By learning these patterns, the system accurately predicts when specific fossils were likely deposited, overcoming the computational limits that previously hindered large-scale high-resolution modeling.
Timeline
2020: Researchers utilized an initial algorithmic method on a supercomputer to analyze fossil data.
October 12, 2026: Shuzhong Shen presented the research findings at a Geological Society of America meeting in Denver.
The Big Picture
This research provides a new methodology for interpreting the five major mass extinctions, fundamentally shifting the understanding of whether these events were abrupt or gradual. It bridges traditional paleontology with modern data science to resolve long-standing debates about extinction mechanics.
While this study focuses on ancient geological history, the underlying machine-learning framework could eventually be adapted to analyze modern ecological data. This could offer scientists better tools to track species loss and habitat changes in the current era.
The takeaway
This study demonstrates how advanced computational tools can unlock secrets from the deep past by identifying patterns invisible to human researchers. These findings encourage a broader appreciation for how data-driven modeling can refine our understanding of Earth's long-term history.
Further reading
For more on the latest research in this field, visit our Environmental section.
Source note: This article includes information reported by Nature.
Live Poll
Do you trust AI-based models to provide accurate insights into historical or scientific data?
