The development of methods to automatically determine the chemical nature of microplastics by FTIR-ATR spectra is an important challenge. A machine learning method, named k-nearest neighbors classification, has been applied on spectra of microplastics collected during Tara Expedition in the Mediterranean Sea (2014). To realize these tests, a learning database composed of 969 microplastic spectra has been created. Results show that the machine learning process is very efficient to identify spectra of classical polymers such as poly(ethylene), but also that the learning database must be enhanced with less common microplastic spectra. Finally, this method has been applied on more than 4000 spectra of unidentified microplastics. The verification protocol showed less than 10% difference in the results between the proposed automated method and a human expertise, 75% of which can be very easily corrected. Highlights ► A machine learning algorithm was developed to determine the chemical nature of microplastics. ► This method allows a fast and reliable automated identification even when several thousand of FTIR spectra have to be studied.► This method is the first part of a software dedicated to the study of microplastics: POSEIDON.

The study of microplastic pollution involves multidisciplinary analyses on a large number of microplastics. Therefore, providing an overview of plastic pollution is time consuming and, despite high throughput analyses, remains a major challenge. The objective of this study is to propose a protocol to determine how many microplastics must be analyzed to give a representative view of the particle size distribution and chemical nature, and calculate the associated margin error. Based on microplastic data from Tara Mediterranean campaign, this approach is explained through different examples. In this particular case, the results show that only 3% of the collected microplastics need to be analyzed to give a precise view on the scale of the North West Mediterranean Basin (error <5%), and 17.7% to give an overview manta per manta (error <10%). This approach could be an important practical contribution to microplastic studies.