Related Experiment Video
Updated: Sep 16, 2025
05:31
Sampling, Sorting, and Characterizing Microplastics in Aquatic Environments with High Suspended Sediment Loads and Large Floating Debris
Published on: July 28, 2018
16.2K
Introduce multivariate two-dimensional information to establish a data-driven volume estimation model for complex
Zhan-Ao Zhang1, Rong Zhang1, Rui Gan1
1State Key Laboratory of Water Pollution Control and Green Resource Recycling, School of the Environment, Nanjing University, Nanjing, Jiangsu 210023, China.
The Science of the Total Environment
|July 5, 2025
Summary
This study introduces a new machine learning method to accurately estimate microplastic fiber volume, overcoming limitations of traditional geometric approaches for better environmental risk assessment.
Area of Science:
- Environmental Science
- Polymer Science
- Data Science
Background:
- Microplastic (MP) fibers are widespread environmental contaminants with potential toxicity.
- Accurate quantification of MP fiber volume is essential for assessing ecological risks and environmental impact.
- Traditional geometric methods struggle to accurately measure the volume of complex-shaped MPs due to limitations in capturing curvature and detailed dimensions.
Purpose of the Study:
- To develop and validate a novel machine learning framework for accurate microplastic fiber volume estimation.
- To improve upon the accuracy of traditional geometric methods in quantifying MP fiber volume.
- To identify key features influencing MP fiber volume for better environmental monitoring.
Main Methods:
- A machine learning model was developed using image recognition and shape descriptors to estimate MP fiber volume.
- The model incorporated 2D features like area, aspect ratio, circularity, and solidity.
- Performance was evaluated against geometric models using real-world MP samples, with accuracy and Mean Absolute Percentage Error (MAPE) as key metrics.
Main Results:
- The machine learning framework achieved 89.43% accuracy and a MAPE of 10.58% ± 4.30% in external testing.
- The ML approach demonstrated superior performance compared to traditional geometric models, overcoming their inherent limitations.
- Interpretability analysis identified area, aspect ratio, circularity, and solidity as significant features for MP volume estimation.
Conclusions:
- Machine learning offers a powerful and accurate tool for estimating microplastic fiber volume.
- This novel approach provides environmental scientists and policymakers with a better method for MP pollution assessment.
- Accurate volume estimation is critical for understanding MP flux and mitigating ecological risks.
