Method instantly evaluates the origin and authenticity of special Brazilian canephora coffees

Group that brings together researchers from Unicamp, Embrapa and Federal Technological University of Paraná obtained more than 90% accuracy in discriminating between samples from different producers and Brazilian states

21.06.2023 | 15:46 (UTC -3)
Karina Ninni, FAPESP Agency
A group that brings together researchers from Unicamp, Embrapa and the Federal Technological University of Paraná obtained more than 90% accuracy in discriminating between samples from different producers and Brazilian states; Photo: Michel Rocha Baqueta/Unicamp
A group that brings together researchers from Unicamp, Embrapa and the Federal Technological University of Paraná obtained more than 90% accuracy in discriminating between samples from different producers and Brazilian states; Photo: Michel Rocha Baqueta/Unicamp

Common coffee consumed in Brazil is generally a mixture of Arabica beans (Coffea arabica) with the so-called robusta – a generalized designation for varieties of the canephora species (Coffea canephora). But, despite being traditionally used in blends and historically described as a low quality coffee compared to arabica, canephora is on the rise in the specialty coffee scene, the result of a new industry perception about the species.

“Brazilian canephora coffees are standing out and reaching a sensorial quality comparable to that of Arabica. Producers, consumers and processors are interested in this market. Special Brazilian canephora coffees produced in two different regions have already received geographical indication registration”, reveals doctoral student in food science, Michel Rocha Baqueta, from the Faculty of Food Engineering at the State University of Campinas (FEA-Unicamp).

Under the guidance of Unicamp professor Juliana Azevedo Lima Pallone and with the collaboration of researchers Patrícia Valderrama, from the Federal Technological University of Paraná, and Enrique Anastácio Alves, from Embrapa in Rondônia, Baqueta developed a method to instantly assess the origin of Brazilian coffees from canephora species from the main producing states, making it possible to distinguish its botanical varieties (conilon and robusta) and the canephora coffee cultivars from the Amazon.

The method developed by the group uses infrared spectroscopy and chemometrics tools (the application of statistical or mathematical methods to interpret data of chemical origin, using machine learning algorithms, based on artificial intelligence) to discriminate between the canephoras cultivated in Brazil. The work generated several articles, two already published in the journals Journal of Food Composition and Analysis and Analyst.

The research was supported by FAPESP through two projects (19/21062-0 e 22/03268-3).

Accuracy and advantages

According to scientists, existing traditional chemical methods, which could be applied for the same type of evaluation, are more laborious and complex, as they involve chemical analyzes that take three to four days to generate results based on chemical compounds present in the product.

“Products considered special and with geographical indication have greater commercial value. This is because the geographical indication indicates not only the origin, but the expected characteristics. In general, specialty coffees are classified based on their sensory characteristics. However, legitimizing the attribution of geographical indication or origin of the product by applying sensory tests is not easy, since these tests provide a global impression of the product, determining its aromas and flavors, but not its origin. A sensory test may indicate that that coffee has the very particular sensory characteristics of coffee from a certain place. But it will be difficult to attest to its origin with a high level of accuracy”, explains Baqueta.

According to the doctoral student, there is a bottleneck in the certification of this type of product in the country and the method developed by the group helps to overcome the difficulties associated with legitimizing specialty coffees, with emphasis on those with geographical indication, offering advantages: it can be done in seconds, with roasted coffee, already ground, without sample preparation and waste generation.

Brazil is the second largest producer of canephora in the world, with emphasis on the states of Espírito Santo (Southeast region) and Rondônia (North region), followed by Bahia (Northeast region). Two varieties are cultivated in the main production sites: Robusta and Conilon. In Rondônia, coffees are produced by indigenous peoples, in addition to conventional producers, and are called Amazon Robusta. Recently, Robustas produced in Rondônia and conilons from Espírito Santo received registration as a geographical indication.

“We had results greater than 90% with the classification model in discriminating between Robustas cultivated by indigenous people and those cultivated by non-indigenous people, and also between samples from Espírito Santo and Bahia and also in differentiating samples of canephoras and arabicas. In the case of discriminating between canephoras with indication and without indication [with and without certification], the accuracy was 100%.”

Chemical fingerprint

The first step in applying the technique depends on obtaining spectra in the near-infrared region of samples of known origin. For this, a spectrometer is used that performs measurements in the spectrum range corresponding to the near-infrared region (800 to 2.500 nanometers). The equipment emits electromagnetic radiation onto the sample and a detection system records the signal. This information is then processed by software and generates spectra relating to the interactions of electromagnetic radiation with the sample, which depends on its chemical composition.

“It is a chemical fingerprint of the sample. This type of evaluation had never been used to discriminate Brazilian special canephora coffees. When we started, we had doubts about the possibility of really determining the origin of the camphors with this method, as coffee after roasting becomes very complex, the chemical composition changes completely. But we did it.”

Pallone clarifies that infrared spectroscopy associated with chemometrics tools can be applied to different objectives in food analysis. “The big challenge is to verify whether these spectra generated by the equipment and these chemometrics tools would be suitable for specific purposes, like ours. It is necessary to test a series of factors: the best way to obtain the spectra, check if there is a better region of the spectrum to use in the analysis. Our purpose was to answer the following question: if we collect the spectra of samples of special canephora coffees from different origins, would we be able to obtain a classification model that allows us to distinguish between them? Therefore, we collected the spectra of all the samples and tried to evaluate the best way to work with the data obtained and the best chemometrics tool to apply to the results and answer the question.”

According to the researcher, the equipment generates spectra based on chemical bonds of the structures that make up the sample. “We imagined that coffees from different origins and different locations could generate different spectrums. We received a very complex result from the equipment, containing spectra from a large number of samples, from different origins. To work with this data, it is necessary to use machine learning algorithms for chemical data, chemometrics, which allow the use of complex spectral data obtained and the association with the origin of specialty coffees. Thus, it is possible to obtain models that indicate the classification of samples, according to the producing region. The models generated are tested and validated and it is possible to know their percentage of success in the classifications.”


Pallone recalls that the samples of special coffees from Rondônia and Espírito Santo were obtained thanks to a partnership with Embrapa Rondônia, which sent samples from producers in the Indigenous Lands of Rondônia and also from non-indigenous people who produce special Robusta in the State, in addition to copies of the Holy Spirit.

“A representative set consisting of 527 samples was analyzed, including Amazon Robusta from indigenous and non-indigenous producers, conilon from Espírito Santo and Bahia, special Arabica and low-quality canephora. We also had samples of unspecified canephora coffees. We tested the model with recommended and non-recommended coffees and saw that it worked. We also compared high-quality canephora with high-quality arabica, to find out whether special Robusta are comparable to arabica in chemical terms.”

The researchers explain that Robusta coffee produced exclusively in Rondônia (Amazon Robusta) has a designation of origin (Matas de Rondônia). Pallone recalls that there was knowledge about the variety of samples from Rondônia and, therefore, it was possible to discriminate, in addition to origin, the botanical varieties cultivated in the Amazon.

Baqueta reports that Embrapa Rondônia is developing a national sensory analysis protocol for canephora in Brazil. “There is an international protocol, but the reality in Brazil is different, there is a lot of diversity here. We need a protocol for the country. Arabica already has a national protocol, although there is an international one as well.”

According to him, one of the legacies of the work recently published in Journal of Food Composition and Analysis was the development of a database. “We collected more than 1.500 spectra. We are able to evaluate the authenticity and origin of any canephora sample that comes from a region included in our database”, he says.


Scientists explain that there are several types of equipment (spectrometers). “The benchtop ones, traditionally used in laboratories, generate spectra with better resolution and cover an area with which more chemical compounds can be detected. These should be in the US$100 range. But there is a portable version of this equipment, the price of which starts at US$1. There are some in the $15, $20 range. The resolution is a little lower, but we have already done some tests and seen that the level of accuracy of portables is very close to that of bench equipment. This is in one of the articles we recently published in the newspaper Analyst”, reveals Baqueta.

The article Brazilian Canephora coffee evaluation using NIR spectroscopy and discriminant chemometric techniques be accessed here.

The study Discrimination of Robusta Amazônico coffee farmed by indigenous and non-indigenous people in Amazon: comparing benchtop and portable NIR using ComDim and duplex is available here.

LS Tractor February