AI tool aggregates microplastics across studies

As microplastics have multiplied, so have the studies quantifying them. And across all those studies, plastics researchers use thousands of different categories for materials, shapes, and size ranges, which means those mounting data can’t be easily combined to show a larger picture.

In terms of data compatibility, “it’s been the Wild West for a long time,” says Win Cowger, research director at Moore Institute for Plastic Pollution Research. “We need a tool to wrangle everything together.”

Cowger joined forces with environmental science graduate student Hannah Hapich and hydrologist Andrew Gray at the University of California, Riverside, to create such a tool (Environ. Sci. Technol. 2024, DOI: 10.1021/acs.est.4c02406). Hapich saw how the growth in natural language processing (NLP) technology, computing power, and open-source artificial intelligence software could automate data harmonization in a way that wasn’t possible even a year ago.

The team’s AI-powered model sorts material and shape categories into hierarchical nested bins. It incorporates more than 50 reference and conversion databases to convert information on microplastics’ length, material, and shape to their volume, mass, and density. Hapich has posted the open-source app and dubbed it MaTCH (Microplastics and Trash Cleaning and Harmonization).

The software, its creators found, can harmonize otherwise incomparable datasets (such as river and drinking water studies) with differing formats, measurements, and particle descriptions. Its NLP technology assigns unfamiliar categories to correct bins 71–94% of the time—which is high for NLP models, Hapich says. “The tool exponentially speeds up the process.”

The tool outputs sunburst plots that organize data hierarchically, which make clear how the tool categorized and sorted the input datasets. “The whole idea is to maximize comparability without flattening the data and destroying specificity,” Hapich says.

California recently became the first government in the world to require testing for microplastics in drinking water, and Cowger wants to make it standard practice to run drinking water analyses through MaTCH. “It really just makes the data extremely useful and rapidly comparable. This is how everybody should do it,” he says.