Rapid Mass Conversion for Environmental Microplastics of Diverse Shapes

Area of Science:

• Environmental Science
• Analytical Chemistry
• Marine Biology

Background:

• Rivers are major pathways for microplastic pollution entering oceans globally.
• Discrepancies exist between microplastic number and mass concentration measurements.
• Accurate conversion models are needed to reconcile these measurement paradigms.

Purpose of the Study:

• To develop efficient and accurate models for converting microplastic number concentration to mass concentration.
• To improve the assessment of seaward transport flux of riverine microplastics.

Main Methods:

• Established a comprehensive environmental microplastic dataset.
• Utilized a deep neural residual network (ResNet50) for microplastic shape classification (fiber, fragment, pellet) with 92.67% accuracy.
• Incorporated circularity (C) and thickness information into 2D microplastic images for enhanced analysis.

Main Results:

• Developed accurate microplastic mass conversion models with low absolute estimation errors (0.2-7.1%) across different shapes and circularity values.
• Models demonstrated significantly faster processing times (saving ~2 hours per 100 microplastics) compared to previous methods.
• Validated models in environmental samples, showing 7-fold lower estimation errors than existing empirical models.

Conclusions:

• The developed models provide a robust and efficient method for microplastic mass estimation.
• This advancement aids in more accurate quantification and understanding of microplastic pollution dynamics in aquatic environments.
• The integration of shape and thickness data improves the precision of microplastic mass flux assessments.