MATERIALS AND METHODS

Experimental site: The experiment was conducted during 2009-2010 at two locations representing mid to high agro-ecologies i.e., Tepi National Research Center (TNRC) (7°3' N, 35°E, 1200 masl) and Bahir Dar (12°31' N, 37°3' E, 1890 masl) in Southwest and Northwest of Ethiopia, respectively (Hailemichael and Tesfaye, 2008; Adam, 2008). The soil type is Dystric Nitosol at Tepi and sandy at Bahir Dar. Bahirdar receives an annual average rainfall of 1545 mm. The mean monthly minimum and maximum temperatures of the sites are 12.2 and 25.8°C., respectively. Tepi receives the annual average rainfall of 1688 mm and has mean maximum and minimum temperatures of 29.5 and 15.4°C, respectively.

Experimental materials: Thirty six diverse ginger accessions representing the germplasms of the crop in the country were used for the study. The planting materials were kindly provided by TNSRC.

Table 1:	Locally collected and introduced ginger accessions that were used in the experiment

Of these, twenty seven were local collections; while the other nine were introduced from abroad (Brazil, China, Rafinfua, Surinam and Mauritius). Details of the accessions that were used in the experiment are given in Table 1.

Experimental design and management: The experiment was laid out in 6x6 Lattice Design with two replications. The spacing was 30x15 cm between rows and plants, respectively, 50 cm between plots and 1 m between replications in plot size of 1.2x1.5 m with 4 rows per plot. Three to five centimeter long one-year-old rhizomes having at least one active bud were used as planting materials for all accessions. Fertilizer was not applied at TNSRC; as there were non-significant results from fertilizer trials (Hailemichael et al., 2008). At Bahir Dar, a recommended rate of fertilizer i.e., 100 and 200 kg ha^-1 DAP and urea, respectively, was applied at planting. Other agronomic practices, such as planting, weeding and harvesting were carried out according to the recommendations of Agricultural and Rural Development Office (MoARD, 2007a) of the site.

Data collection: For data collection, five plants were randomly selected from each inner two rows. The parameters recorded included date of planting, number of plants/plot (nppp), days taken to maturity (dm), fresh rhizome yield (q ha^-1), dried rhizome yield (dry in q ha^-1), oleoresin content (oc in w/w%), volatile oil content (voc in v/w%) and fiber content (fb) (1 = having less fiber and 5 = having higher fiber content), leaf area (cm²) which was calculated using lengthxwidth multiplied by adjustment factor (K = 1.426) (Anteneh et al., 2008). Oleoresin content (w/w%), volatile oil content (v/w%) and fiber content were determined as described by Toure and Zhang (2007).

Data analysis: Analysis of variance for each character was computed using the standard statistical procedure of Gomez and Gomez (1984), using statistical software SAS 9.1 (SAS, 2002). Means were separated using LSD value at p = 0.05 level of significance. Test of homogeneity of variances of each character for the tow locations were performed by Cochran’s test (Cochran and Cox, 1957) and the test showed heterogeneity of the two locations for almost all characters considered. The experiment was conducted using lattice design with two replications. However, Randomized Complete Block Design (RCBD) was used for the genetic variability analysis, because RCBD was more efficient than lattice for estimating the variance components of the different characters considered.

Genotypic (σ²g) and phenotypic (σ²p) variances were computed using the procedures of Burton and de Vane (1953). The genetic advance was computed as per the method reported by Johnson et al. (1955). Non-significant characters of the two sites were not taken in calculations of genetic variations.

Cluster analysis for genetic divergence among the genotypes was estimated by D² (Mahalanobis, 1936). Principal component analysis was performed by using correlation matrix by employing procedure printcomp corr of SAS version 9.2 (SAS, 2008) in order to examine the relationships among the quantitative characters that are correlated among each others by converting into uncorrelated characters called principal components.

RESULTS

Analysis of variance, range and mean values: Separate analysis of variance (Table 2) for individual location was made for each parameter. Number of plants per plot, days taken to maturity, fresh rhizome yield, dry rhizome yield, oleoresin, volatile oil and fiber content showed highly significant differences at both locations. Other characters, like number of leaves per plant, leaf length, leaf width, leaf area, plant height, number of fingers per rhizome, rhizome length and rhizome width were non-significant between accessions at Tepi and highly significant at Bahir Dar. Number of tillers per hill showed non-significant difference between accessions at both locations.

At Tepi, fresh rhizome yield ranged from 60.00 to 366.7 kg ha^-1. More than 33.3% of the accessions gave above the grand mean fresh rhizome yield i.e., 181.10 q ha^-1 (Table 3). At Bahir Dar, fresh rhizome yield ranged from 50.5 to 228.6 kg ha^-1. At Bahir Dar, though the range was as wide as that of Tepi, the grand mean was lower (10661 kg ha^-1). Fifty percent of the accessions gave above the grand mean. At Tepi, dry rhizome yield ranged from 11.30 to 69.9 kg ha^-1 and at Bahir Dar, the same character ranged from 9.3 to 53.2 kg ha^-1.

Table 2:	Analysis of variance for 16 characters of Ginger accessions grown at Tepi and Bahir Dar during 2009-2010 cropping season
**and * indicate significant differences at 1 and 5%, respectively, ns: Not significant

Table 3:	Estimates of ranges and standard deviation for char acres of ginger accessions grown at Tepi and Bahir Dar during 2009-2010 main cropping season
NLPP: No. of leaves per plant, NPPP: No. of plants per plot, LL: Leaf length, LW: Leaf width, LA: Leaf area, NFPR: No. of fingers per rhizome, RL: Rhizome length, RW: Rhizome width, DM: Days taken to maturity, FRY: Fresh rhizome weight, DRY: Dry rhizome weight, OC: Oleoresin content, VOC: Volatile oil content and FB: Fiber content

Table 4:	Estimates of variances, coefficients of variations, heritability in broad sense (H%) and genetic advance as percent of mean (GAM) of ginger accessions grown at Tepi

Table 5:	Estimates of variances, coefficients of variations, heritability in broad sense (H%) and genetic advance as percent of mean (GAM) of ginger accessions grown at Bahir Dar

At Tepi, oleoresin content ranged from 2.7 to 6.0 w/w while at Bahir Dar it ranged from 2.4 to 5.0 w/w. At Tepi, volatile oil content ranged from 0.6 to 2.0 v/w; while at Bahir Dar, volatile oil content ranged from 0.5 to 1.6 v/w. The fiber content ranged from 1.0 to 2.0 at both sites. At Tepi days taken to maturity ranged from 251 to 285 and at Bahir Dar 241 to 273.

Genotypic and phenotypic variation: Characters such as number of plants per plot, fresh rhizome yield, fiber content, volatile oil content and dry rhizome yield produced high GCV and PCV at Tepi (Table 4). However, days taken to maturity showed the least GCV and PCV values.

At Bahir Dar number of plants per plot, fresh rhizome yield, number of fingers per rhizome and dry rhizome yield produced high GCV and PCV (Table 5). Here also days taken to maturity showed the least GCV and PCV values.

Heritability in the broad sense: Broad sense heritability was estimated at Tepi for all agronomic traits and it ranged from 49.17 to 98.90% (Table 4). The values of heritability were high for all characters measured at Tepi.

At Bahir Dar heritability values were ranged between 62.80 and 99.90% (Table 5). The values of heritability were high for all characters measured at this particular location.

Table 6:	Eigen values, total variance, cumulative variance and eigen vectors for 16 characters of 36 ginger accessions grown at Tepi during 2009-2010 main cropping season
NLPP: No. of leaves per plant, NTPH: Number of tillers per hill, NPPP: No. of plants per plot, LL: Leaf length, LW: Leaf width, LA: Leaf area, NFPR: No. of fingers per rhizome, RL: Rhizome length, RW: Rhizome width, DM: Days taken to maturity, FRY: Fresh rhizome weight, OC: Oleoresin content and VOC: Volatile oil content

Estimates of genetic advance: Genetic advance was relatively high for dry rhizome yield (48.760%) and fresh rhizome weight (70.427%) and this was coupled with high heritability and high GCV at Tepi location (Table 4).

Similarly, genetic advance was relatively high for dry rhizome yield (48.760%) and fresh rhizome weight (70.427%) and this was coupled with high heritability and high GCV at Bahir Dar (Table 5).

Principal component analysis: At Tepi, the first six principal components explained 77.7% of the total variation (Table 6). The first PC which accounted for 33.6% of the total variation, was the average effect of traits, such as leaf length (0.626), leaf area (0.612), plant height (0.659), rhizome width (0.668), dry rhizome yield (0.999) and days to maturity (0.983). This perhaps emphasizes the significance of these characters for the overall genetic diversity in PC1. The second PC which accounted for 11.4%, of the variability among accessions was the average effect of traits such as rhizome width (0.233), fiber content (0.241) and fresh rhizome yield (0.995).

Similarly at Bahir Dar, the five principal components with eigenvalues greater than one explained for 67.7% of the total variations among accessions (Table 7). The first PC explained 23.3% of the variation among accessions was highly influenced by fresh rhizome yield. The second PC that explained 15% of the variation among accessions was highly influenced by days taken to maturity (-0.993).

Cluster analysis: At Tepi, the germplasm were grouped into 7 cluster (Table 8) whereas at Bahir Dar the germplasm were classified into 11 clusters (Table 9) based on the morphological and quality traits associated with them.

Table 7:	Eigen values, total variance, cumulative variance and eigen vectors for 16 characters of 36 ginger accessions grown at Bahir Dar during 2009-2010 main cropping season.
NLPP: No. of leaves per plant, NTPH: No. of tillers per hill, NPPP: No. of plants per plot, LL: Leaf length, LW: Leaf width, LA: Leaf area, NFPR: No. of fingers per rhizome, RL: Rhizome length, RW: Rhizome width, DM: Days taken to maturity, FRY: Fresh rhizome weight, OC: Oleoresin content and VOC: Volatile oil content

Table 8:	Clustering patterns of 36 ginger accessions grown at Tepi during 2009-2010 main cropping season.

Table 9:	Clustering patterns of 36 ginger accessions grown at Bahir Dar during 2009-2010 main cropping season

Table 10:	Generalized square distance (D2) among the clusters of accessions grown at Tepi during the growing season of 2009-2010 main cropping season
* and ** significance at X2 of p<0.01% and p<0.05% are 30.58 and 25.00

Table 11:	Generalized square distance (D2) among the clusters of accessions grown at Bahir Dar during the growing season of 2009-2010 main cropping season
* and ** significance at X2 of p<0.01% and p<0.05% are 30.58 and 25.00

Genetic distances: Based on Mahalanobis’s statistics, highly significant inter-cluster distances were obtained at Tepi (Table 10). At Tepi Cluster II showed the maximum and significant distance (1644.00) from cluster VII whereas cluster VI revealed the maximum genetic distance (1145.00) from cluster XI. These distances indicated that the germplasm in the above clusters are significantly divergent from each other.

At Bahir Dar, there was a significance difference between all clusters (Table 11). The maximum distance was found between C I and C X (D² = 1284.0). The second most divergent distance was between C VI and C XI (D² = 1145.0). The third most divergent distance was C VII and C X (D² = 1022.0). The genetic divergence distance between the rest of the clusters was also highly significant.

DISCUSSION

The existence of significant difference among ginger germplasm accession for most of the characters studied at both Tepi and Bahirdar locations indicates the existence of sufficient genetic variability among the studied accessions. This finding is in-line with results of Chandra et al. (1999). Parthiban et al. (2011) observed significant variability among plant height, basal diameter, sturdiness quotient, number of primary and secondary branches and seed yield in jatropha hybrids. However, this is not in line with the findings of Dar and Sharma (2011) who reported significant variation between quantitative character among 60 tomatoes genotypes. Further, Gichimu and Omondi (2010) reported that morphological characterization of five newly developed lines of arabica coffee as compared to commercial cultivars in Kenya and they found non significant difference for internodes length.

At Tepi, fresh rhizome yield ranged from 366.7 to 60.0 kg ha^-1. At Bahir Dar, fresh rhizome yield ranged from 228.6 to 50.0 kg ha^-1. At Bahir Dar, though the range was as wide as that of Tepi the grand mean was lower (10661 kg ha^-1). This may be attributed to long growing period of the crop and lower rain shower, especially late in the season at Baihr Dar. The average fresh rhizome yield in all Indian accessions varied from 7000.0 to 10000.0 kg ha^-1; despite up to 40000.0 kg ha^-1 report from experimental plots (Pruthi, 1998). Weiss (2002) reported average fresh rhizome yield in India up to 150.0 t ha^-1 from selected cultivars of ginger. Similar work done under Ethiopian condition confirmed that the average fresh rhizome yield varied between 11600.0 to 21440 kg ha^-1 (Hailemichael and Tilahun, 2004). Jansen (1981) also reported rhizome yield of up to 38000.0 kgq ha^-1 in Ethiopia. At Tepi, dry rhizome yield ranged from 69.9 to 13.3 kg ha^-1. Similarly, at Bahir Dar dry rhizome yield ranged from 6990 to 1130 kg ha^-1, indicating that there was a considerable difference between accessions for the character. The existence of variation in yield of accessions may be considered during varietal development selection process.

At Tepi, the maximum oleoresin content was obtained from Ging 141/73; while at Bahir Dar it was recorded from Ging 84/00. At Tepi, the highest volatile oil content was obtained from Gin 141/73; while at Bahir Dar, it was obtained from Ging 74/86. An experiment conducted in India using 14 popular ginger cultivars showed that the oleoresin content varied from 3.6 to 10.4% (Pruthi, 1998). In a similar experiment conducted in India using 24 genotypes found that the highest oleoresin content was 5.31% (Tiwari, 2003b). This study demonstrates a high level of intraspecific differences in oil composition due to environmental factors and, more particularly, soil characteristics.

In this study, wide range in days to maturity was recorded. Such a wide range in days taken to maturity of accessions offers a great opportunity for the development of ginger varieties suitable for various agro-ecological zones of Ethiopia; experiencing diversities in precipitation and length of growing period. The accessions differ significantly for most of the characters and relatively wide range of mean values for most of the characters indicated the existence of variation among the tested accessions.

Genetic variability is a prerequisite for response to selection and knowledge of the extent and nature of phenotypic variability is, then, one of the basic needs for the breeders for further improvement of the crop (Adam, 2006). The amount of genotypic and phenotypic variability that exists in a species is essential in developing better varieties and in initiating a breeding program. The estimates of components of variance, phenotypic and genotypic variances, heritability in the broad sense, genetic advance for Tepi and Bahir Dar are presented in Table 7 and 8, respectively. Deshmukh et al. (1986) classified Genotypic Coefficients of Variation (GCV) and Phenotypic Coefficients of Variation (PCV) estimates as high (>20%), medium (10-20%) and low (<10).

Selection based on traits with high PCV and GCV is effective and the phenotypic expression of such characters is a good indication of the genotypic potential. Progress from selection depends on genetic variability existing in the population and selection is more effective when the genetic variation is higher than environmental variation (Allard, 1960). According to Moll and Stuber (1974) genetic effects are dependant on non-genetic effects such as physiological factors.

Generally, the differences between GCV and PCV for almost all the corresponding characters were small, indicating thereby these characters were less influenced by the environment. Similar findings were reported by Jalata et al. (2011) in field pea. However, this finding contradicts with the findings of Singh et al. (2011) who reported the magnitude differences to be medium to low in PCV and GCV for most of the characters studied in field pea. For all characters studied, PCV value was greater than GCV at both sites similar with the findings of Tiwari (2003b).

Broad sense heritability was high for all characters measured. Dabholkar (1992) categorized heritability estimates as low (5-10%), medium (10-30%) and high (>30%). Garedew (2006) also found high heritability with high phenotypic and genotypic coefficient of variation in tuber weight/hill and tuber number/hill for (common name) Plectranthus edulis (Vatke) Agnew. Chattopadhyay et al. (2004) reported maximum heritability (88.5%) for weight of secondary rhizome followed by plant height (83.3%), total fresh rhizome yield (79.9%) and number of secondary finger (72.1%) in ten turmeric cultivars studied in India. In addition to this, Abera (1998) reported high heritability (79.06%) of marketable root yield in medium maturing sweet potato cultivars. The value of heritability is affected by the breeding material, the character under consideration and environmental conditions to which the genetic materials are exposed (Dabholkar, 1992). In this study, relatively high heritability values were obtained may be due to favorable growing conditions and adaptability of the ginger germplasms used.

Since heritability estimates show only the effectiveness with which selection of genotypes can be based on phenotype, their utilities increased when used only with estimates of genetic advance (Johnson et al., 1955; Allard, 1960). Besides, Genetic advance was relatively high for dry rhizome yield (48.760) and fresh rhizome weight (70.427) and this was coupled with high heritability and high GCV at both locations (Table 4, 5). This high heritability coupled with GCV estimates also gives a reliable estimate of the amount of genetic advance through selection which indicates the effectiveness with which selection of genotype can be based on phenotypic performance (Johnson et al., 1955).

At Tepi the maximum variation showed by the first principal component was based on traits leaf length, leaf area), plant height, rhizome width, dry rhizome yield and days to maturity. Similar finding was reported by Kebede and Bellachew (2008) for leaf area. The largest value for fresh rhizome yield in the first PC at Bahir Dar displayed that this trait was mainly responsible for explaining 23.32% of the total variance. This emphasizes the importance to these traits for assessment genetic diversity in ginger.

The 36 ginger grermplasm accessions were grouped into 7 clusters (Table 8). Six are real clusters and one is solitary. This solitary is also considered as cluster, even if it consists of one accession. This indicates the tested ginger germplasm were highly divergent. Cluster I contained one accession (Ging 28/79) from Mauritius and the rest from Ethiopia. Cluster II was characterized by accessions from Ethiopia (Gamogofa and Kafa-Sheko). The third cluster was characterized by accessions from Surinam, Ethiopia and Kenya. Cluster IV was characterized by accessions from Ranifua, Ethiopia and Australia. Clusters V and VII were characterized by accessions from Ethiopia only. The introduced germplasm were concentrated in four clusters. The germplasm from Gamo Gofa district (Ethiopia) were distributed in five different clusters. Similarly, accessions from Kafa district (Ethiopia) were distributed in three clusters. Generally the germplasm from Ethiopia were concentrated in 10 clusters, even if most of them are concentrated in cluster I. This indicated that there was more genetic variability in Ethiopian accessions.

At Bahir Dar cluster I consisted of nine accessions from Mauritius, Brazil and Ethiopia. Clusters II and VIII were mainly characterized by the accessions from Ethiopia and Australia, whereas Cluster III was characterized by accessions from Rafintua and Ethiopia. Clusters IV, VI, VII, IX and X were characterized by accessions from Ethiopia. The last cluster (cluster XI) contained accession from Australia only. Similarly, accessions from Kafa district (Ethiopia) were distributed six clusters. Generally the germplasm collected from different major ginger growing zones of Ethiopia were scattered over 10 clusters out 11 clusters at Tepi and all over clusters at Bahir Dar, displaying possibility of exchange of planting materials among farmers from different zones. Further, this result also suggested that the accessions from Ethiopia were more variable.

In addition, we recorded overlapping of clustering patterns in germplasm in the majority of origin. This could be explained as lack of differentiation among origins, probably arising partly due to gene flow among regions (Ayana and Bekele, 1999). In general, it is observed that accessions from Ethiopia are relatively more variable in their clustering patterns, compared to those from abroad. This indicates that due emphasis must be given to Ethiopian germplasm to improve most of the attributes in the crop. Nevertheless, the ginger accession found in Ethiopia have to be classified accurately by using molecular markers , qualitative traits, etc.

The study has revealed highly significant inter-cluster genetic distances at both locations (Table 10, 11). At Tepi Cluster II showed the maximum and significant distance (1644.00) from cluster VII whereas cluster VI revealed the maximum genetic distance (1145.00) from cluster XI at Bahir Dar). These distances indicated that the germplasm in the above clusters are significantly divergent from each other. Since, the magnitude of heterosis largely depends upon the degree of genetic diversity among the parental lines, the progenies of those accessions belonging to the pairs of distant clusters could be useful in a hybridization program for obtaining a wide spectrum of variation among the segregants (Peeters and Martinelli, 1989). Thus, it may be concluded that the germplasm accessions from cluster II and VII at Tepi and VI and XI at Bahir Dar could offer relatively better parental lines that, when intercrossed, could produce hybrids with high heterotic effect, even though other clusters with significant genetic distances are also good sources of parental lines.

Dissimilarities recorded in this study revealed a wide variation in land races of ginger. Greater diversity may be obtained if a more collections from geographically distant areas was included in the study. Nevertheless, the information on the genetic relationships will be vital to utilize available germplasm for crop improvement.

ACKNOWLEDGMENT

We appreciate the support of the Ethiopian Ministry of Education in funding the project.