Related Experiment Video
Updated: Mar 27, 2026
Sampling and Identification of Microplastics in Groundwater
Published on: November 7, 2025
A pipeline for meso- and microplastic identification in aquatic systems using machine learning and hyperspectral
Petros Chatzitoulousis1, Stephanie Oswald2, Nikolaos Ploskas3
1Institute for Molecules and Materials, Analytical Chemistry, Radboud University, P.O. Box 9010, 6500GL, Nijmegen, the Netherlands; Biomedical Engineering Laboratory, School of Electrical & Computer Engineering, National Technical University of Athens, 15773, Athens, Greece.
This study presents a new method using Hyperspectral Imaging in the Near-Infrared range (HSI-NIR) and a machine learning model to identify microplastics. The approach accurately detects common polymers like polyethylene and polypropylene in environmental samples.
Area of Science:
- Environmental Science
- Analytical Chemistry
- Data Science
Background:
- Microplastic pollution is a significant global environmental and health concern.
- Current identification methods are often destructive, slow, and costly.
- Hyperspectral Imaging in the Near-Infrared range (HSI-NIR) offers a non-destructive alternative but faces data analysis challenges.
Purpose of the Study:
- To develop and validate an efficient analytical pipeline for microplastic identification using HSI-NIR.
- To apply a machine learning model for accurate, pixel-wise classification of microplastic particles.
- To assess microplastic distribution and polymer types in diverse environmental samples.
Main Methods:
- An analytical pipeline combining HSI-NIR with optimized preprocessing techniques was developed.
- A Multi-Layer Perceptron (MLP) machine learning model was employed for pixel-wise microplastic classification.
- The method was tested on samples from Lanzarote Island, the Wadden Sea, and the Waal and Rhine rivers.
Main Results:
- The MLP model demonstrated high accuracy in identifying polyethylene (PE), polypropylene (PP), polyethylene terephthalate (PET), and polystyrene (PS).
- The MLP model outperformed other common classification algorithms like Support Vector Machines and Random Forests.
- PE and PP were the most prevalent polymers found across all sampling sites.
Conclusions:
- The developed HSI-NIR and MLP pipeline provides an effective, non-destructive method for microplastic identification and quantification.
- This approach can significantly aid in assessing the environmental impact and distribution of microplastic pollution.
- Further research can build upon this methodology for broader environmental monitoring applications.
