Study site and evaluated material

This study was conducted at the Robusta coffee germplasm bank of the EETP, Los Ríos Province, Ecuador (1º 4’ 33’’ S, 79º 29’ 15’’ W, altitude 80 m; average temperature 24 ºC and relative humidity 82 %). The subjects of the study were 25 elite trees, selected in a study for their agronomic characteristics of commercial interest, such as plant height, total number of productive branches, length of productive branches, total number and distance between nodes (^{Plaza et al., 2015}).

Production data and determination of physical and chemical properties

During the harvest period (April to October) 2012, coffee cherries were collected every 15 d from the 25 trees. After each collection, the fruits were dry-processed, exposed to sunlight in a glass canopy. The fruits were distributed in layers 3 to 5 cm thick, turned three times a day until moisture was 10 to 12 % (“café bola seca”) and hulled.

Physical and chemical variables

In each sample, moisture, density and grain size were determined. Physical defects were evaluated according to the norm INEN-ISO 10470:2012 (^{INEN, 2012}). Real loss of mass and sensorial impact were determined to estimate the value of the defects. Soluble solids (SS) were evaluated in 10 g ground coffee in 200 mL of water, which was placed in a heater-shaker for 15 min until boiling and the liquid extract was centrifuged 10 min. A sample of 10 mL of the supernatant was dried 3 h in an oven, it was kept in a desiccator for 5 to 10 min, weighed, and yield of soluble “dry extract” coffee was obtained. Percentages of caffeine, glucose, fructose and sucrose were determined by HPLC.

Organoleptic variables

In a cylindrical roaster, 100 g green coffee was roasted at 180 to 220 ºC for 8 to 12 min until it turned dark brown. The coffee beans were then reduced to small particles (500 µm) with and electric mill.

A tasting panel of technical workers of INIAP and of the firm Solubles Instantáneos S.A. (SICA) was formed. With each sample, five cups of coffee were prepared with 8.25 g roasted ground coffee in 150 mL boiling water. The tasters scored the sensorial parameters: fragrance-aroma, taste, aftertaste, salt-acidity balance, bitter-sweet balance, sensation in mouth, uniformity, balance, purity, and general score on a scale of 1 to 10; 1 was the lowest score.

Statistics

The principal components (PCA) and cluster analyses were performed with the production, physical, chemical and organoleptic variables using the software Infostat version 12.0 (^{Di Rienzo et al., 2014}). Cluster analysis was based on Euclidian distance and, with the cluster hierarchy method, a similarity dendrogram was obtained.

Yield

The individuals with the highest average yields between 2010 and 2012 were COF004-p9, COF004-p15, COF003-p7 and COF003-p18 (Table 1), and nine individuals produced annual averages of coffee cherries of less than 10 kg per plant. Yield above this value is a criterion for selecting high-producing clones in Ecuador (^{Enríquez et al., 2014}). The high standard errors indicate wide production variability due to the biannual pattern (high alternating with low annual production) in the first years of production (^{Cilas et al., 2011}). Production of C. canephora cultivars can be high in one place and different in others (^{Montagnon et al., 2000}). Also, in allogamous species, such as Theobroma cacao, it is essential to evaluate productive variables in each specific region in function of soil conditions and climatic variables of at least four years.

Physical variables of coffee beans

Density of the coffee beans was 742 (individual NP2024-p7) to 801 g L^-1 (individual NP3018-p19) and the coefficient of variation among individuals was 1.97 (Table 1). All the individuals were characterized as “high density” (>650 g L^-1) (^{Duicela et al., 2004}). The conversion rate from coffee cherry to green coffee was similar among the individuals (R² = 0.21) (Table 1), with an average of 4.61 (Figure 1), indicating that high-yielding cultivars also produce a larger quantity of green coffee (Figure 2). Individual COF003-p8 had the lowest conversion rate (4.11), a valuable commercial characteristic. Bean size varied from 5.62 to 7.44 mm, and 11 individuals had beans 5 to 6 mm, while another 11 had 6 and 7 mm beans. Three individuals (COF001-p2, COF003-p7 and COF004-p15) had values above 7 mm required by the industry (^{Jiménez and Soto, 1995}). In our study, the individual with large beans did not have a lower conversion rate from coffee cherry to green coffee, indicating that there is no direct relationship.

Figure 1 Relationship between coffee cherry yield and conversion rate of coffee cherry to Green coffee in C. canephora individuals. Each value is the conversion average. EET-Pichilingue, 2012. 

Figure 2 Relationship between coffee cherry and Green coffee of 25 C. canephora individuals. EET-Pichilingue, 2012. 

Average SS was 25.09 % (Table 1), and individual NP3018-p19 had the highest percentage (26.66 %), and these values were higher than those reported by ^{Castaño et al. (2000)}. According to norm INEN-ISO 10470:2012, individuals COF003-p5, COF003-p2, COF001-p2, COF001-p10, COF004-p7, COF004-p9, NP3018-p8 and NP3018-p9 had physical defects with a score of less than 3, showing little insect damage, deformities or extraneous material (stones, sticks, clods). The small amount of insect damage is considered intrinsic to each individual and could be used to select them as resistant to the coffee borer (Hypothenemos hampei). Moreover, these genotypes are of better quality because they have few perforated beans.

Chemical characteristics

Percentage of caffeine varied from 1.65 to 2.21 % (Table 2), which are similar to those obtained by ^{Koshiro et al. (2006)} in C. canephora. Caffeine concentrations of this species (1.9 %) are higher than in C. arabica (1 %). The values found in our study coincide with ^{Mazzafera et al. (1991)}, whose study, however, did not show a direct relationship between percentage of caffeine and organoleptic score. Individuals COF001-p2, COF003-p2, COF003-p5, COF003-p19, COF004-p7, COF004-p9, NP2044-p6, and NP2044-p17 were outstanding for their higher values of caffeine.

The percentages of sucrose were several times higher (4.43 % average) than those of glucose and fructose (0.29 and 0.13 %) (Table 2), coinciding with that documented in C. canephora (^{Rogers et al., 1999}) and with the content of sucrose in C. canephora from Hawaii (^{Koshiro et al., 2006}), although values for fructose and glucose were higher than for sucrose.

Percentages of sugars in C. canephora clone coffee beans changes with provenance and parent material. In our study, all the C. canephora individuals were from the Ecuadorian Amazon region and show trends similar to other regions of Ecuador, which have been documented by ^{Rogers et al. (1999)}, who confirmed that Ecuadorian C. canephora has higher concentrations of sucrose than of fructose and glucose. The values for sucrose in our study were lower than those of C. arabica reported by ^{Rogers et al. (1999)}. Lower sucrose content has a negative effect on cup quality of C. canephora (^{Koshiro et al., 2006}). However, individuals COF001-p1 and NP2044-p17 had higher concentrations of sucrose and, thus they were of high quality.

Organoleptic evaluation

Fourteen individuals had scores above 80 points (Table 3) and, according to ^{ICO (2010)} norms, they are fine coffees. Individuals COF005-p6 and COF003-p18 obtained higher total scores, and all of the characteristics had averages equal to higher than 8 (Figure 3). Moreover, both individuals were outstanding for their high yield. The type of diagram of these individuals was the result of the evaluations of the panel of tasters.

Figure 3 Diagrams of cup quality profiles of two individuals with the highest total scores in the organoleptic evaluation. EET-Pichilingue (2012). 

The correlation between coffee cherry yield and defects in green coffee was significant and positive (0.55). Between chemical and sensorial variables, the correlations were not high; the highest correlations were between sensorial variables (Table 4), which showed similar quality evaluations by the members of the panel.

Identification of groups based on multi-variate analysis

The principal components analysis explained variability in physical, chemical and organoleptic traits. Eleven principal components (PC) explained 99 % of the variation. PC1 and PC2 explained 23 and 20 % of the total variation (Table 5). The PCA showed that the variables that most contributed to discrimination of the individuals in PC1 were percentage of moisture, average bean size, percentage of caffeine, percentage of fructose and percentage of sucrose. In PC2 the important variables were coffee cherry yield, conversion coffee cherry/green coffee, bean size and defects.

The dual graph of principal components shows separation of individuals with high yield, abundant defects and larger beans (positive quadrant of PC2) separated from those placed in the positive quadrant of PC1 and PC2, which included those with high production and few defects (Figure 4).

Figure 4 Principal components analysis of physical, chemical and sensorial variables of 25 Coffea canephora genotypes. EET-Pichilingue, 2012. 

The cluster analysis revealed five groups (Figure 5). In the first group (G1) were COF004-p15 and COF003-p7, with higher yield, high percentage of defects and bean size larger than 7mm. The second group (G2) was COF003-p8 with high yield and few defects. The third group (G3) included NP2044-p17, NP2044-p6, NP2044-p16 and COF001-p2, which had low organoleptic scores. The fourth group (G4) was COF003p-18 with high organoleptic score and high green coffee yield. There were 17 individuals in group (G5) with medium size beans and few defects. Grouping the individuals with similar characteristics can contribute to selection of materials for industrial use. The first group had the two highest-yielding individuals, but they also had the largest quantity of defects. This last variable should be evaluated to differentiate defects caused by insects from those caused mechanically.

Figure 5 Cluster analysis of 25 Ecuadorian Robusta coffee (Coffea canephora) individuals based on Euclidian distance of physical, chemical and organoleptic characteristics. EET-Pichilingue, 2012. 

Implications for selection programs

The results of this study give added value to previous selections, which were based exclusively on yield. The data on quality identified two individuals that will contribute to improving quality and yield of C. canephora in Ecuador. Studies that combine productivity and quality should be conducted and validated in other regions.