Tlanezi Lucero Martinez Reyes belegte von 2020 bis 2022 den Studiengang "Water Resources Engineering and Management". Ihre herausragende Masterarbeit mit dem Titel "Innovative method for detection and identification of microplastic particles in sediment bed, based on fluorescent tagging with Nile Red and Methylene Blue - Case study Großer Brombach lake" wurde von Maria Ponce und Dr. Stefan Haun vom Institut für Wasser- und Umwelsystemmodellierung betreut.
Die Masterarbeit befasst sich mit einem fluoreszenz-basiertem Verfahren zur Bestimmung von Mikroplastik. Im Hintergrund steht die Aufgabe, die Belastung der aquatischen Umwelt durch Mikroplastik, insbesondere von Sedimenten in Talsperren, mit einer automatisierbaren Methode zu bestimmen. Die Arbeit hat damit eine grundlegende Bedeutung, da das Thema Mikroplastik in der aquatischen Umwelt ein ernstzunehmendes Problem darstellt.
Zusammenfassung (in English)
Worldwide, microplastic particles (MPs) have generated increasing concern due to their ubiquity in the environment, accumulation rates in the environment, and their potential to harm human and animal health. However, monitoring MPs in the environment is a time-consuming and expensive task, therefore, the need for the development of cost- and time-effective methods is imperative to advance microplastic research. The developed work proposes an innovative approach to MPs identification that combines a staining protocol capable to reduce the fluorescence intensity of OM by combining MB 0.6%+NR dyes with the advantages of a supervised machine learning algorithm called Random Forest classifier that allows an efficient categorization of data. The proposed approach used Red-Green-Blue (RGB) values, extracted from images obtained from NR-tagged particles under fluorescence microscopy filters (Rhodb, nilRe, DAPI, FITC) to create a dataset. The data set was used to train and validate the Plastic Identifier Model (PIM). The created model has an accuracy of 0.94, AUC 0.99, and Kappa 0.93 which indicates an excellent classification performance of the constructed model. The applicability of the PIM was demonstrated by successfully applying the model to identify 90% of artificially prepared samples, those samples contained OM and MPs particles previously digested. Furthermore, the proposed methodology was applied to analyze eight natural samples from the GBS reservoir, and it was found the presence of fibers such as PS, nylon and PET monofilament, PA particles, and also PVC polymers on sediment bed samples. Nevertheless, in the case of natural samples, a confirmatory method such as FTIR, or RAMAN is required to corroborate the PIM predictions of particles identified by this method for the GBS samples. Without a backup method to confirm the PIM results, it was not possible to demonstrate the applicability of PIM to natural samples.
The generated method that combines the MB+NR staining protocol and the PIM is an innovative, semi-automated, and cost-efficient method for identifying particles that has the potential to become an economic method for MPs detection and identification of natural samples.