Microplastic detection in terrestrial systems using a classification on optical values and surface characteristics

There is a growing concern that the steady increase in plastic production is leading to a substantial contamination of our environment with microplastic particles. While aquatic ecosystems are more and more studied, there is still a substantial lack in knowledge regrading terrestrial (mainly soil) system. This knowledge gap is partly related to the challenges to detect and analyses microplastic particles in soils. Firstly, it is difficult to extract microplastic from a matrix of organic and inorganic particles of similar size. Secondly, the well-established spectroscopic methods to detect microplastic in water samples are sensitive to organic material and are moreover very time consuming. Eliminating very stable organic particles (e.g. lignin) from soil samples without affecting the microplastic to be measured is hardly possible. Hence, a robust analytical approach is needed to tackle the microplastic detection in soils. In this study, we combine a density separation scheme, a 3D Laser Scanning Confocal Microscope (Keyence VK-X1000, Japan) and a machine learning algorithm to classify and analyses microplastic particles in soil samples. For the analysis a silty loam (16% sand, 59% silt, 25% clay, 1.3% organic carbon) and a loamy sand (72% sand, 18% silt, 10% clay, 0.9% organic carbon) were spiked with different concentrations of high density Polyethylene (HDPE), low density Polyethylene (LDPE), Polystyrene (PS) and Polybutylene adipate terephthalate/Ploy lactic acid (PBAT/PLA) microplastic (HDPE 50 - 100 and 250 - 300 µm, LDPE <50 and 200 - 800 µm, PS <100 µm, PBAT/PLA < 2 mm). The classification with a machine learning algorithm is an essential data processing step to distinguishes between plastic, mineral as well as organic particles left after density separation. In case microplastic adopts the soil color, a combination of optical information and surface characteristics are used for a successful classification. Overall, the 3D Laser Scanning Confocal Microscopy in combination with a machine learning algorithm is a promising tool to detect, quantify and analyses microplastic in soils.