Raman Spectroscopy and Machine Learning for Agricultural Applications: Chemometric Assessment of Spectroscopic Signatures of Plants as the Essential Step Toward Digital Farming.

A growing body of evidence suggests that Raman spectroscopy (RS) can be used for diagnostics of plant biotic and abiotic stresses. RS can be also utilized for identification of plant species and their varieties, as well as assessment of the nutritional content and commercial values of seeds. The power of RS in such cases to a large extent depends on chemometric analyses of spectra. In this work, we critically discuss three major approaches that can be used for advanced analyses of spectroscopic data: summary statistics, statistical testing and chemometric classification. On the example of Raman spectra collected from roses, we demonstrate the outcomes and the potential of all three types of spectral analyses. We anticipate that our findings will help to design the most optimal spectral processing and preprocessing that is required to achieved the desired results. We also expect that reported collection of results will be useful to all researchers who work on spectroscopic analyses of plant specimens.