Weathering-independent differentiation of microplastic polymers by reflectance IR spectrometry and pattern recognition. (August 2022)

• Record Type: Journal Article
• Authors: Ferreiro, Borja; Andrade, Jose M.; Paz-Quintáns, Carlota; Fernández-González, Verónica; López-Mahía, Purificación; Muniategui, Soledad
• Is Part Of: Marine pollution bulletin. Volume 181(2022)
• Publication Date: 2022-08
• Subjects: Microplastics -- Infrared spectrometry -- Reflectance -- Weathering -- Pattern recognition -- Variable selection; Marine pollution -- Periodicals; Marine Biology -- Periodicals; Water Pollution -- Periodicals; Mer -- Pollution -- Périodiques; Publications périodiques; Pollution des mers; Lutte antipollution; Electronic journals; 363.7394
• DOI: 10.1016/j.marpolbul.2022.113897
• Languages: English
• ISSNs: 0025-326X

Abstract

Abstract: The presence and effects of microplastics in the environment is being continuously studied, so the need for a reliable approach to ascertain the polymer/s constituting them has increased. To characterize them, infrared (IR) spectrometry is commonly applied, either reflectance or attenuated total reflectance (ATR). A common problem when considering field samples is their weathering and biofouling, which modify their spectra. Hence, relying on spectral matching between the unknown spectrum and spectral databases is largely defective. In this paper, the use of IR spectra combined with pattern recognition techniques (principal components analysis, classification and regression trees and support vector classification) is explored first time to identify a collection of typical polymers regardless of their ageing. Results show that it is possible to identify them using a reduced suite of spectral wavenumbers with coherent chemical meaning. The models were validated using two datasets containing artificially weathered polymers and field samples. Graphical abstract: Unlabelled Image Highlights: Weathered polymers can be identified using a reduced suite of IR wavenumbers. In most cases 10 variables were selected out of 3400 spectral wavenumbers. Models were validated using a weathering series and field samples. The best model used SVM and ATR wavenumbers from the most relevant loadings.