Related Experiment Video
Updated: Aug 8, 2026
08:27
Sampling and Identification of Microplastics in Groundwater
Published on: November 7, 2025
Field readiness of Raman-AI pipelines for microplastic analysis across environmental matrices
R Muhammad Nur Nasyrah1, Nurul Muthmainnah Ramlan2, Isnaeni Isnaeni3
1Department of Physics, Hasanuddin University, Makassar 90245, Indonesia; Research Center for Photonics, National Research, and Innovation Agency, Serpong 15314, Indonesia.
Journal of Hazardous Materials
|August 6, 2026
Summary
Raman spectroscopy with AI/ML for microplastic analysis is not yet field-ready due to significant reporting gaps and limited validation across different conditions. Further development is needed for reliable real-world application.
Area of Science:
- Environmental Science
- Analytical Chemistry
- Data Science
Background:
- Microplastic monitoring requires robust analytical workflows applicable beyond laboratory settings.
- Raman spectroscopy combined with Artificial Intelligence/Machine Learning (AI/ML) shows promise for polymer identification.
- Current high benchmark accuracies do not guarantee workflow reliability across diverse matrices, instruments, and preprocessing methods.
Purpose of the Study:
- To critically evaluate the evidence supporting the field-readiness of Raman-AI microplastic analysis workflows.
- To identify essential reporting elements and validation strategies for reliable field deployment.
- To develop a framework for assessing the readiness of Raman-AI workflows for real-world microplastic monitoring.
Main Methods:
- A systematic Scopus search identified 61 relevant studies published between 2023-2025.
- Studies were coded based on sample matrix, Raman modality, acquisition metadata, preprocessing techniques, AI/ML model family, analytical task, and validation design.
- A readiness-oriented framework was developed to evaluate the complete Raman-AI workflow, not just classifier performance.
Main Results:
- The corpus is dominated by classification/identification tasks (68.9%) using Micro-Raman (59.0%) in water or controlled matrices.
- Significant reporting gaps exist concerning spectral resolution, normalization, cosmic-ray removal, leakage control, uncertainty analysis, interpretability, and data/code availability.
- Quantification/regression and imaging/mapping applications remain limited.
Conclusions:
- Raman-AI microplastic analysis workflows require substantial improvements in reporting and validation before being considered field-ready.
- Key areas for development include open benchmarks with code, grouped and external validation, uncertainty quantification, interpretability, traceable preprocessing, and cross-matrix/cross-instrument testing.
- Field-readiness claims for current Raman-AI workflows should be treated as conditional pending these advancements.
