We can't find the internet
Attempting to reconnect
Something went wrong!
Hang in there while we get back on track
A Paint Library of Plastic Particles (PLoPP): a spectral library, visual key, and machine learning model for paint microplastic identification
Summary
Peeling paint from buildings, cars, and roads is a major but understudied source of microplastic pollution, and scientists have lacked good tools to identify these paint particles once they end up in the environment. Researchers built a new reference library and identification guide for paint microplastics, tested with real-world samples, to help scientists more accurately track this pollution source. Better detection tools are an important first step toward understanding how much paint-based microplastic is out there and what risks it might pose to ecosystems and, potentially, to people who breathe, drink, or eat these particles.
The degradation of paint is thought to be a major pathway of microplastics to the environment. However, pollution from paint microplastics is challenging to assess due to a lack of paint-specific spectral libraries and visual keys to aid in the differentiation of paint microplastics from non-paint microplastics. Here we begin to fill this gap is filled by creating a visual key for paint microplastic identification and a spectral library, the Paint Library of Plastic Particles (PLoPP), using attenuated total reflectance Fourier-transform infrared micro-spectroscopy (µATR-FTIR). PLoPP includes 263 spectra from 90 paints used in seven sectors (architectural, automotive, consumer, general industrial, marine, road markings, and wood) spanning 15 colors, five appearances (glitter, gloss, matte, pearl, and semi-gloss), and at least 25 polymers, though primarily polyurethane, polyurethane acrylics, and polyvinyl chloride. To assess the accuracy of the library in identifying paint microplastics from other microplastics, a spectral analysis pipeline using a machine learning model was developed to differentiate between pristine paint and non-paint microplastic samples with an overall accuracy of 92%. Using environmental particles from Plymouth, UK, and spectra from Charleston, SC, we verified the utility of these tools. The machine learning model had an accuracy of 55% when differentiating between environmental paint and non-paint microplastics, while the visual key had an average accuracy of 92% for particles. The correlation-based FTIR library searches in OMNIC TM software using PLoPP correctly classified 86% of the environmental particle spectra as paint/non-paint. This research demonstrates the utility of a paint-specific spectral library and visual key to better identify paint microplastics. Further work should identify whether paints can be characterized by sector.