Introduction
Microplastic contamination of soil and water is a global environmental issue that has not yet been thoroughly addressed. There are no standardized methods of data collection and evaluation, as well as comprehensive sample databases. In recent studies, infrared scanning has been used to identify plastic particles in soil without extraction and was proved to be an effective way of solving the sample collection problem. In the research by Ng et al., a convolutional neural network (CNN) was suggested as a method of measuring and classifying the levels of contamination. This proposal aims to continue the research of microplastics contamination using infrared scanning and the CNN model by creating an extensive database of samples and expanding the spectral library.
Identified Problem and Previous Research
Microplastic contamination is one of the most serious current environmental concerns. Microplastics are plastic pieces less than 5 millimeters in size used to manufacture consumer products, such as synthetic fabrics, toothpaste, and skincare items. Being washed down the drain, they are not removed by wastewater treatment and get into the soil and water. There are concerns that microplastics produce a negative effect on the environment as they absorb toxic chemicals that build up over time and can affect human health.
The current research on microplastics pollution focuses on the development of the methods of microplastic identification and extraction, the study of the main sources of pollution, and its effect on the environment and human health. The presence of microplastics in water sources has been studied in more detail, while its impact on soil remains largely unexplored. No standardized methods have yet been developed for identification and measurement. Is it still not established how much microplastics can exist within the soil and what can be regarded as the maximum concentration limit. The inconsistency of sampling methods results in various sampling units, such as abundance per surface area, abundance to depth, volume in cubic meters, and weight ratios. Different extraction techniques are used that affect the measurements of the collected plastic contaminants. Along with the development of detection, identification, and quantification methods, the current research objectives include the analysis of ecological risks, impact on health, pollution characteristics, and remediation strategies.
The most recent strategy for microplastics identification in the soil is infrared screening. The research conducted by Ng et al. used visible-near-infrared (vis-NIR) spectroscopy and a convolutional neural network (CNN) model to detect microplastics and measure their concentration. Vis-NIR scanning was proved to be feasible in predicting microplastics in the soil, while the CNN model allowed the researchers to classify various degrees of contamination based on concentration. The methods turned out to be efficient for the identification and quantification of microplastics and have the potential for further development and implementation. The proposed research is based on the results of the previous studies and aims to expand the spectra library to include more plastic types and collect a larger database of samples. It is intended to contribute to environmental protection initiatives and can become an important step towards the development of a strategy of addressing the problem of microplastics contamination.
Hypothesis
The expansion of the spectral library for infrared scanning and the collection of an extensive sample database will help to effectively classify types of plastic and the quantity of soil contamination using the CNN model.
Aims
- Expansion of the spectral library for infrared scanning, particularly accounting for the various colors and plastic polymers.
- Collection of an extensive sample database to ensure more accurate results of further research.
- Determining the effectiveness of the CNN model for contamination measurement.
Research Plan
The proposal suggests using infrared spectroscopy for soil microplastic contamination screening in different spectra to put together a database for further research and analysis. The primary suggested model of screening is vis-NIR spectroscopy. Each plastic polymers exhibit unique infrared spectral signatures, and the Vis-NIR spectra can be used to predict various physical and chemical properties of soil samples and quantify microplastics in soil. A Vis-NIR spectrometer measures the amount of light that is reflected from a surface within the wavelength range of 350 to 2500nm, giving a reflected percentage for each wavelength. Each sample should be scanned several times at different spots to ensure that its heterogeneity is captured. The scanning results are correlated with the chemical structure of the sample and used to predict the composition of new sample sets.
The vis-NIR scanning provides several benefits compared to other methods of sample analysis. It circumvents the need for microplastic extraction and allows researchers to measure bulk soil samples, which helps to avoid extensive sample preparation. By reducing the number of required manipulations, makes overall analytical time shorter and diminishes biases caused by human handling. Overall, it helps to solve one of the main problems in microplastics detection related to sample preparation difficulties.
To evaluate the level of microplastic contamination, a screening model instead of a regression model is proposed because of the physical characteristics of plastics and the minimal surface contact area with the Vis-NIR probe. For this research, a CNN model was chosen that provides an alternative to classical neural image analysis based on deep learning methods. It is a tool that extracts features directly from the image and, then, takes actions to obtain the final result and make a classification. The CNN model was determined to be effective in predicting various soil properties using spectral data, with the model trained from scratch performing better than the transfer learning model. However, due to a relatively small sample size used in the previous research, the hypothesis of the superiority of the first type of model is yet to be proved. The collection of a more extensive sample database and the expansion of the spectral library within the proposed research can help either verify or reject this assumption. It will also help improve the CNN model’s accuracy and establish it as a reliable method for evaluating the level of contamination.
The development of an extensive spectral library is a challenging task, and careful preparation and planning are required. The library should be expanded to include plastic particles representing a diversity of polymer types, colors, and morphologies. They should be collected across a range of metrics, geographies, and time and include samples containing a range of additives and pigments. It will help to enhance data quality, accuracy, and consistency and avoid color-related noise by taking into account different colors and dyes. A comprehensive library should be built progressively and include samples obtained by different research groups in different regions. It is necessary to collect samples from various types of soils and areas to achieve better representation.
The proposed research is expected to provide both qualitative and quantitative contributions to the studies of microplastics contamination. The quantitative contribution includes the expansion of the sample database, and the quality of further research will be facilitated by the development of a comprehensive spectral library and using the CNN model for evaluation. The research is intended to become an important step towards achieving standardization in the identification and measurement of soil contamination by microplastics. Monitoring the pollution level within the soil is vital for assessing soil condition and health and understanding the risks related to exposure to microplastics for both the environment and human health.
Conclusion
The proposed research is based on the already tried-out methods of measuring and evaluating the level of soil contamination by microplastics. It intends to use infrared scanning for identification and the CNN model for the classification of microplastics in the soil and focus on the expansion of the spectral library and sample database. The database should be built to include samples collected across a range of metrics, areas, and time, and include plastic particles representing a diversity of polymer types and colors. The study also aims to prove that the CNN model is a reliable classification tool that can be used in further research. Both qualitative and quantitative contributions are expected that will facilitate environmental management.
References
Corradini, F. et al., ‘Predicting Soil Microplastic Concentration Using Vis-NIR Spectroscopy’, Science of the Total Environment, vol. 650, no. 1, 2019, pp. 922-932, Web.
He, D. et al., ‘Microplastics in Soils: Analytical Methods, Pollution Characteristics, and Ecological Risks’, TrAC Trends in Analytical Chemistry, vol. 109, 2018, pp. 163-172, Web.
Ng, W., Minasny, B., and McBratney, A., ‘Convolutional Neural Network for Soil Microplastic Contamination Screening Using Infrared Spectroscopy’, Science of the Total Environment, vol. 702, 2019, p. 134723, Web.
Tox Town, Microplastics [website], Web.