MATERIALS AND METHODS

Study site: The experiment was conducted at Oromiya Estate in Jimma District, at Agaro Agricultural Research Sub Center, Ethiopia from 2004 to 200. It is located at 45 km in the southwest of Jimma town at an altitude of 1630 m.a.s.l. It is situated at 7°50’35”N latitude and 36°35’30”E longitude. The mean annual rainfall of the area is 1616 mm with average maximum and minimum temperatures of 28.4 and 12.4°C, respectively (Elias, 2005).

Experimental material, design and management: The study was carried out during the main season in 2009/10 on batch II forty-nine Limu coffee trial of six-year old including two standard checks (744 and F-59) (Table 1). The coffee germplasm accessions were collected in 2003 from the potential and representing areas in the Limu-Kossa Wereda of Jimma Zone and were planted in the field in August 2004. The experiment was laid out in a 7x7 simple lattice design with 2 replication and seven accessions per incomplete block. Six trees per accessions were planted in 2x2 m spacing. All the management practices such as shading, weeding and fertilization were uniformly applied to all plots as per the recommendation.

Experimental procedure and data collection: Ripe Red coffee cherries were handpicked. Before pulping fully ripened and healthy berries were separated from foreign materials and unripe green cherries and green bean coffee samples were prepared during 2009/2010 cropping season. A total of 98 samples were prepared from forty nine accessions (six trees per accessions bulked together). The samples were prepared from six trees per accession per replication at peak harvest period. The samples were carefully prepared using wet processing method (pulping, fermentation, and drying) and prepared for sensorial analysis using the following procedures.

Table 1:	Geographical origin of coffee accessions used in the study

Pulping: Fully ripened beans of berries were separated from the skin and pulp by using a pulping machine that squeezes the berries between fixed and moving surfaces.

Fermentation and drying: The beans were then stored in a fermentation tank for 48 h till first washing was made. Then, samples were stored for 24 h for final washing. The samples were placed on mesh wire under sun for drying. During drying, the moisture content of the bean wasmeasured by moisture tester H-E50 to maintain the moisture level at 10-12% for all samples uniformly. About 300 g of green coffee bean samples were prepared per accession per replication separately for each accession for organoleptic quality characteristics analysis. To attain homogenous bean size and healthy beans for organoleptic quality analysis, samples were screened through a mesh sieve 15 (5.95 mm). Then, samples on and above screen 15 were used for analysis. The sensorial quality analyses were carried out at the Jimma Agricultural Research Center by a group of well trained cup tasters/panelists of the center.

Roasting and grinding: The roaster machine, probatBRZ6, was first heated at about 160-200°C. About 150-200 g of green coffee bean sample prepared per accession per replication was used for roasting. When the roast starts to crackle (burst open), the gas were turned down. When the coffee was considered medium roast (7 min on average) it was tipped out into the cooling try. Cold air was blown through the coffee to produce rapid cooling-off. When the roast was cool (4 min on average), it was blown to remove the loose silver skins before grinding. Variability in roasting was controlled by measuring weight loss. The weight loss found was between 8 and 10% and these matches from medium to dark roasting degree reported by Agwanda et al. (2003). About 12 g medium seized ground coffee was prepared using Mahlkoing electrical grinder with middle adjustment.

Brewing: Soon after grinding, coffee powder weighing about 8 g was placed in a cup with a capacity of 180 mL. Then, boiling water was poured on to the ground coffee up to about half way in the cup. Soon after, volatile aromatic quality and intensity parameters were recorded by sniffing. Then, the contents of the cup were stirred to ensure an infusion of all coffee grounds. The cup was then filled to the brim with boiled water. The brew was made ready for panelists within 8 min.

Cup tasting: This was carried out once the beverage cooled to around 60°C (drinkable temperature). Two cups per sample were prepared for tasting session. The genotypes and the replicates were arranged at random. Aroma (aromatic quality and intensity), flavor, acidity, bitterness and astringency were scored using scales ranging from 0 to 5. Typical flavor was assessed as an after taste aromatic quality that could vary from winy to flowery (winy, fragrant, floral, citrus, moka, spicy and others). There was also an overall standard for liquor quality based on the above attributes that ranged from 0 to 5 scales. Mean of each variable by the panel was used for statistical analysis (ISO, 1992).

Data analysis: The data were statistically analyzed using lattice analysis of variance format design by using the statistical procedures described by Gomez and Gomez (1984). All statistical and data processing were performed Using XLSTAT (XLSTAT, 2008), Computer program, SAS (SAS, 2002) version 9.2 software and Genes version 7.0 soft ware. For characters having significant mean differences, the difference between treatment means was compared using LSD at 5% probability level.

The phenotypic variances and coefficients of variations were estimated as per Singh and Chaudhry (1985). Heritability in the broad sense for all organolyptic quality attributes was computed using the formula suggested by Singh and Chaudhry (1985). The genetic advance expected under selection, assuming selection intensity of the superior 5% of the genotypes was estimated in accordance with the methods illustrated by Allard (1960).

Phenotypic and genotypic correlations were computed following the method described by Singh and Chaudhry (1985).

Clustering of the 49 accessions for organolyptic quality attributes was performed using the proc cluster procedure of SAS version 9.2 by employing the method of average linkage clustering strategy of observations. The number of clusters was determined by following the approach suggested by Copper and Milligan (1988) by looking into three statistics namely, pseudo F, pseudo t² and the Cubic Clustering Criteria (CCC). The number of cluster was decided where the CCC and pseudo F statistics combined with a small value of the pseudot² statistics and large pseudo t² statistics for the next cluster fusion.

Genetic divergence between clusters was determined using the generalized Mahalanobis’s D² statistics. Mahalanobis (1936) developed this method to determine divergence prevailing among population in terms of generalized group distance-deleted (Sharma, 1998)-deleted. Testing the significance of D² values obtained for a pair of clusters were taken as the calculated value of chi-square (χ²) and tested against the tabulated value of χ² for P degrees of freedom (P is the number of organolyptic quality attributes) (Singh and Chaudhry, 1985) at appropriate probability level, that was considered.

RESULTS AND DISCUSSION

The mean square values for the studied traits ranged between 0.045 and 0.248 (Table 2). The Analyses of Variance (ANOVA) revealed the presence of significant differences among the Limmu coffee accessions for the organoleptic quality attributes (acidity, flavor and overall standard). However, ANOVA showed non-significant differences among the accessions for aromatic intensity, astringency, body, and bitterness (Table 2).The significant variations among coffee accessions for the quality traits indicated the existence of variability to have an effective selection for improvements. Similarly, different studies indicated the existence of high genetic diversity of arabica coffee for quality specifically in major coffee growing areas in Ethiopia (Yigzaw, 2005; Getu, 2009) and in Kenya (Walyaro, 1983). Similarly, Alsemaan et al. (2011) reported the existence of genetic diversity within Rosa damascena cultivated in Syria and recommended Almarahl and Bab Alnayrab accessions to be used to broaden the production of rose oil. In addition, Parthiban et al. (2011) found variability in Jatropha for some character studied. Gichimu and Omondi (2010b) studied on morphological characterization of five newly developed lines of Arabica coffee and two commercial cultivars in Kenya and they reported non significant difference for internodes length. However, Dar and Sharma (2011) obtained highly significant difference among the genotypes for all the quality traits studied in tomatoes.

In this study, we recorded genetic variance values ranging 0.00 to 0.09 and phenotypic variances between 0.06-0.15 (Table 3). Genetic variance of quality attributes was observed less as compared to environmental variance except aromatic quality (Table 3). The results revealed that bitterness, astringency and body variation was mainly influenced by environment. According to Verma and Agarwal (1982), heritability values greater than 50% are considered as high, whereas values less than 20% are low and values between 20 and 50% as medium.

Table 2:	Mean squares of genotypes for organoleptic quality traits
*,**significant at p<0.05, 0.24 and p<0.01, 0.34, respectively

Table 3:	Variability parameters for organoleptic quality trait in coffee accessions
σ2g: Genotypic variance, σ2P: Phenotypic variance, H2: Heritability, GCV: Genotypic coefficients of variation, PCV: Phenotypic coefficients of variation, GA: Genetic advance, GA%: Genetic advance as percent of mean

Accordingly, higher estimates of broad sense heritability were observed for aromatic quality (64.74%) and medium broad sense heritability values were recorded for acidity (43.64%), over all standards (40.77%) and flavor (39.70%). However, aromatic intensity, astringency, bitterness and body were grouped under low heritability range. Variations of genotypes for bitterness, astringency and body were due to environmental factors and thus broad sense heritability was estimated to be small. Thus, the study depicted that bitterness, astringency and body are the appropriate parameters to control variation due to roasting problem during liquoring. Therefore, these attributes are less important to evaluate organoleptic performance of genotypes.

Phenotypic coefficients of variation values ranging from 8.37 to 37.50% and genotypic coefficient of variation values ranging from 0.00 to 9.51% were obtained Table 3). The organoleptic cup quality attributes were influenced by both environment and genotype. Genotypes respond differently for quality attributes and their magnitude is measured by Phenotypic Coefficient of Variation (PCV) and Genotypic Coefficient of Variation (GCV) values greater than 20% are regarded as high, whereas values less than 10% are considered to be low and values between 10 and 20% to be medium (Deshmukh et al., 1986).

The organoleptic cup quality attributes were influenced by both environment and genotype. Genotypes respond differently for quality attributes and their magnitude is measured by Phenotypic Coefficient of Variation (PCV) and Genotypic Coefficient of Variation (GCV) values greater than 20% are regarded as high, whereas values less than 10% are considered to be low and values between 10 and 20% to be medium (Deshmukh et al., 1986).

High Phenotypic Coefficient of Variation (PCV) was recorded for astringency and bitterness. Medium phenotypic coefficient of variation was obtained for aromatic quality, acidity, flavor and over all standard. In contrast, low values of phenotypic coefficients of variation also recorded for aromatic intensity and body. All quality attributes categorized under low GCV range. Environmental variation had large effect on the expressions of astringency, bitterness and body where difference between Genotypic Coefficients of Variation (GCV) and Phenotypic Coefficients of Variation (PCV) was relatively large. Thus, selection on phenotypic basis of these attributes may not be effective for the genetic improvement. However, relatively, small difference was obtained for other attributes indicating both environmental and genetic variation contributed towards expressions of the traits and selections on phenotypic basis for this trait may improves the population for coffee quality.

High heritability, genotypic coefficient of variation coupled with high genetic advance as percent of mean for aromatic quality attribute implies the potential for the coffee genotype improvement through selection.

Although the phenotypic coefficients of variation were greater than genotypic coefficients of variation for quality attributes aromatic intensity, aromatic quality, acidity, flavor and over all standard a narrow gap was found between PCV and GCV values, indicating little influence of environment in the expression of quality trait. The present study is in agreement with the study by Getu (2009).

The phenotypic coefficient of variability were although greater for quality attributes aromatic intensity, aromatic quality, acidity, flavor and over all standard than respected genotypic ones but a narrow gap was found indicating little influence of environment in the expression of quality trait. The present study is in agreement with the study by Getu (2009).

We obtained phenotypic association values ranging between -0.51 and 66 (Table 4). Significant positive phenotypic association was observed among aromatic quality, aromatic intensity, over all standard, body, acidity, and flavor (Table 4). Attributes which correlated strongly and positively with flavor were considered as good cup quality attributes. Flavor revealed negative association with bitterness. Therefore, aromatic quality, aromatic intensity, acidity, body and over all standard were described as good cup quality attributes, whereas bitterness were considered as poor cup quality attributes. Body was positively and significantly associated with good cup quality attributes such as aromatic intensity, aromatic quality, flavor and overall standards. Expression of characters of crop plants is correlated due to genotypic and/or environmental factors. Similarly, Yigzaw (2005) indicated positive association among good cup quality attributes. Moreover, Agwanda et al. (1999) identified flavor as an all round organoleptic attribute to be considered during selection to develop superior coffee genotypes which is in support of present finding. In addition, direct observable phenotypic association of characters resulted from genotypic and/or environmental correlations (Falconer and Mackay, 1996). Therefore, these organoleptic quality attributes contributed to good final cup quality. Thus, considering good cup quality attributes like flavor during selections would improve future target population for quality. Similar to present finding, Umamaheswari and Mohanan (2011) identified internodal length, vine length, number of inflorescences per plant and leaf area as lead characters in Vanilla planifolia Andrews and they recommended that due weightage can be given to these characters in improvement of this species.

The results of genotypic association for organoleptic quality attributes were given in Table 4 and we recorded the values ranging between 0.68 and 0.99. In this study, positive and significant genotypic association was observed among aromatic intensity, aromatic quality, acidity, astringency, bitterness, body, flavor and overall standard. Likewise, Yigzaw (2005) and Getu et al. (2009) indicated similar positive association among good cup quality attributes in Arabica coffee genotypes.

Table 4:	Senso-phenotypic (above diagonal) and genotypic (below diagonal) correlation
*, **significant at p<0.05, 0.24 and p<0.01, 0.34, respectively. AI: Aromatic intensity, AQ: Aromatic quality, AC: Acidity, AS: Astringency, BI: Bitterness, BO: Body, FL: Flavor, OVS: Over all standard

Agwanda et al. (1999) indicated that flavor has got strong genetic association with preference and therefore it is the best selection criteria for the genetic improvement of Arabica coffee liquor quality. Moreover, flavor used for discriminating among varieties because it was found as an all round beverage quality attribute which in corporate other aromatic attributes and well associated with good cup quality attributes like acidity and body. Nikhila et al. (2008) reported that length of primary branches, number of primary branches, intermodal length and bush spread are the character that should be given premium importance while carrying out crop improvement programmes in robusta coffee.

Cluster distributions of 49 coffee genotypes using organoleptic quality traits is given in Table 5. The number of accessions classified in each cluster was 31, 17 and 1, in cluster I, II and II, respectively. Cluster I composed of 31 (63.27%), Cluster II 17 (34.69%) accessions and Cluster III 1 (2.04%), indicating how closely related different genotypes were grouped together. The first group comprised 31 coffee accessions that were characterized by low organoleptic quality. The second cluster comprised 17 that were characterized by medium in organoleptic quality and the third remaining cluster comprised of one coffee accession which was relatively superior in organoleptic quality. The present study was in agreement with previous findings that reported by Dessalegn et al. (2008) and on forty-two Ethiopian collections of Arabica coffee genotypes.

Mean organoleptic quality attributes of clusters for eight organoleptic quality attributes in 49 coffee accessions was given in Table 6 and it ranged 0.25 to 4.58. Cluster III and I was found relatively different because distant and divergent groups constituted these clusters and thereby characterized by extreme values of organoleptic quality attributes. High mean values of aromatic intensity, aromatic quality, acidity, body and flavor and overalls standard characterized cluster III while the lowest mean values of these attributes characterized in turn cluster I. Mean of clusters therefore, indicated the presence of two quite different groups (I and III) consisted of contrasting sensorial quality attributes. Cluster II was average for good cup quality attributes except astringency. The released coffee varieties (744 and F-59) were used as check to identify Limu coffee accession that perform similar to the check. The two checks were grouped in clusterI, indicating more than half percent of Limu accessions had similar quality attributes with the standard checks. The present finding corroborates with that of Yigzaw (2005) and Dessalegn et al. (2008).

Table 5:	Cluster distribution of the 49-coffee genotypes for organoleptic quality traits
*Represents the check varieties

Table 6:	Mean organoleptic quality attributes in three clusters
*,** represent the lowest and highest mean values, respectively. AI: Aromatic intensity, AQ: Aromatic quality, AC: Acidity, AS: Astringency, BI: Bitterness, BO: Body, FL: Flavor, OVS: Over all standard

Squared mahalanobis distance between clusters for organoleptic quality attributes indicated that most of the clusters were highly significantly distant from each other (Table 7). Significant distance analysis and association of genotypes to specific cluster group indicated the presence of genetically distant materials and the association of genotypes with specific cluster group of distinct characteristics. Cluster I showed the maximum and significant genetic distance (38.82) from Cluster I. The result also indicated that cluster III had maximum genetic distance from cluster I. Cluster III the inter cluster distances between clusters II and III, I and III in that order were found to be significant. This confirmed the presence of genetically diverse coffee materials for the quality characters as pointed out by Dessalegn et al. (2008). The significant inter cluster distances indicated that there is a high opportunity for obtaining transgressive segregates and maximizing hetrosis by crossing accessions belonging to different clusters as there is a higher probability that unrelated genotypes would contribute unique desirable alleles at different loci (Peeters and Martinelli, 1989). Similarly, Rohman et al. (2004) also reported that high divergence between the clusters for 25 sorghum varieties which were grouped in to four clusters.

Similarly, Souza and Sorrells (1991) pointed out that categorizing germplasm accessions in to morphologically similar, more particularly genetically similar groups is useful for selecting parents for crossing. Iftekharuddaula et al. (2002) reported that clustering pattern was not influenced by the geographic origin rather it was influenced by the pedigree of the breeding lines in rice which is again in support of present study. In contrast to present study, Falconer (1981) reported that genetic diversity has probably arisen through diversity in origin (geographical separation), ancestral relationship, gene frequencies and morphology. Zhuang et al. (2011) also reported that the clustering analysis done on Persian Wheat (Triticum turgidum ssp. carthlicum) accessions using EST-SSR Markers suggested that most of the accessions with adjacent geographic origins had the tendency to cluster together which is also in contrast to our finding. Souza and Sorrells (1991) pointed out that categorizing germplasm accessions in to morphologically similar, more particularly genetically similar groups is useful for selecting parents for crossing. Falconer (1981) reported that genetic diversity has probably arisen through diversity in origin (geographical separation), ancestral relationship, gene frequencies and morphology. These workers indicated that plants differing in either one or more of these factors would differ by a significant number of genes. Iftekharuddaula et al. (2002) reported that clustering pattern was not influenced by the geographic origin rather it was influenced by the pedigree of the breeding lines in rice. On the other hand, Zhuang et al. (2011) reported that the clustering analysis done on Persian Wheat (Triticum turgidum ssp. carthlicum) accessions using EST-SSR Markers suggested that most of the accessions with adjacent geographic origins had the tendency to cluster together. Three principal components (PC) PC1, PC2 and PC3 with eigen values of 4.23, 1.42 and 1.01, respectively, explained 81.37% of the total variance (Table 8). The first two principal components PC1 and PC2, with percent variability of 52.87 and 17.77%, respectively, explained 70.64% of the total variance. All the organoleptic quality attributes had contribution to genotype classification. However, some of the characters had relative values closer to unity in the first Principal Component (PC1) and thus contributed more to the classification (Chahal and Gosal, 2002).

Table 7:	Mahalanobis distance (D2) of the three clusters for organoleptic quality trait
*D2 significant at p = 0.05 (χ2) = 14.07, **D2 = Significant at p = 0.01 (χ2 ) =18.48

Table 8:	Eigenvalues, total variance, cumulative variance, and eigenvector for quality traits in Limu coffee germplasm accessions

Aromatic quality, acidity, flavor and over all standard had higher score as compared to others and contributed the highest variability in PC1. Therefore, these quality attributes were the cases of diversity among the coffee genotypes.

CONCLUSION

There has been no compressive and systematic quality analysis carried out in Limu coffee germplasm collections, though some promising selections were identified based on yield performances and disease resistances. The present study was undertaken using batch II involving 49-Limu coffee accessions with the prime aim to characterize them based on organoleptic quality traits and thereby determine the extent of genetic diversity. The findings showed significant differences among the coffee accessions for most quality characters, indicating the existence of organolepitc variability for future improvements. Moreover, there was high phenotypic coefficient of variation for astringency and bitterness. Medium phenotypic coefficient of variation was obtained for aromatic quality, acidity, flavor and over all standard. This is in contrast to the low phenotypic coefficients of variation recorded for aromatic intensity and body. It must be known that Limu coffee is of a restricted environmental importance. In an attempt to identify and develop suitable coffee varieties, it is necessary to include additional accessions and evaluate them over seasons based on disease resistance, yield performance, quality standards and other biochemical characteristics.

ACKNOWLEDGMENTS

We would like to acknowledge Jimma University College of Agriculture and Veterinary Medicine for the source of budget for the research and we are also thankful to the Jimma Agricultural Research Centers (JARC) management and coffee breeding for providing study materials, land and laboratory facilities. Our heartfelt gratitude also goes to the JARC coffee quality panelists: Mr. Getu Bekele, Mr. Niguse Mekonen, Mr. Abrar Sualeh, Mr. Alemseged Yilma, Mr. Jafar Dawid, Mr. Ashenafi Ayano, Mrs. Zenebech Teshome, Mrs. Fantaye Mekonen, and Mrs. Foziya Kemal for the laboratory sensorial quality analysis.