The global coffee industry is a booming one with a rich history that spans the globe. Due to the sheer size of this industry and the multitude of factors affecting growth and processing, there is a high degree of complexity in the resulting coffee quality, which is intrinsically linked to its volatile composition. There is no standardisation of what constitutes light, medium or dark roast in the industry and no rapid, non-destructive methods of predicting the coffee volatiles as an indication of quality. The current study aimed to establish a rapid Near Infrared Spectroscopy (NIR) method of detecting coffee volatiles non-destructively coupled with a machine learning approach to assess the effect of different roasting degrees on coffee bean volatile composition. Washed arabica coffee beans from Ethiopia and Congo were roasted in triplicate roast batches to industry-validated light, medium and dark degrees. These were analysed with HS-SPME-GC/MS and NIR in parallel. The PLS-R model was constructed to predict volatiles from the NIR data, enabling rapid indication of the roast degree. This work successfully predicted 32 compounds (such as furans, ketones, phenols, pyrroles, esters, and pyridines) from the NIR data with R<sup>2 </sup>values above 0.8 and low error. Ultimately, this approach allowed for the successful rapid, non-destructive prediction of key roast degree discriminatory marker volatiles and paves the way for the development of fast, hand-held NIR equipment methodology to predict coffee volatiles in the industry to discriminate based on roast degree and beyond.

Potential of near infrared spectroscopy and machine learning to predict volatile changes during coffee roasting

Where thousands of chemistry professionals meet to share ideas and advance scientific and technical knowledge. Spring 2023 abstracts, posters, and presentations will launch at the end of March.

ACS Spring 2023

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.