Coffee belongs to the family Rubiaceae and genus Coffea, which is further subdivided into Mascarocoffea, Eucoffea, and Paracoffea sections (Coste 1992; Vieira 2008). The commercially significant species of coffee (Coffea canephora Pierre ex A. Froehner and Coffea arabica L.) belong to the subsection Paracoffea and they have originated from sub-tropical montane areas of central and eastern Africa (Nair 2010). Arabica coffee is the more significant of the two species in terms of production (70%), value and area planted (I.C.O 2020). Arabica coffee is an allotetraploid with a chromosome constitution of 2n = 4x = 44, and is known for its high quality, intense aroma, lower caffeine content, and a less bitter taste, all of which make it more commercially viable and profitable compared to robusta coffee (Lashermes and Anthony 2007). The genetic base for coffee production is therefore very small because there are more than 1000 species in the genus coffea and yet only two species are of commercial significance. Although the other species have no known commercial value yet, they represent a significant gene pool and source of genetic variation that is important for breeding traits such as drought, pest and disease resistance (Rhiney et al. 2021).

Arabica coffee thrives in specific niches as it is sensitive to extremes of temperature, preferring temperatures bet-ween 15 and 25℃, and high and well distributed rainfalls (> 1800 mm/annum) for healthy growth and productivity (Nair 2010). It is produced across a range of altitudes although, generally, it does very well between 1200-1800 m above sea level in the equatorial and tropical zones (Coste 1992). Initial efforts in coffee breeding focused on increasing coffee yields for profitable production which narrowed and limited selection to most productive lines, with efforts towards disease and drought tolerance coming later on (Eskes and Leroy 2009). The ability of plant to adapt to biotic and abiotic stress factors, such as the effects of climate change, depends on the intraspecific trait diversity that has evolved in the species over time (Warschefsky et al. 2014). This is already very limited for coffee because producers have, over time, tended to rely on specific but limited lines, with genetic and phenotypic plasticity producing distinct phenotypes in response to environmental variation even from same parentage (Eskes and Leroy 2009; Ferreira et al. 2020). It is therefore imperative to assess and ascertain genetic diversity in these coffee germplasms accessions over time, especially where there are risks of genetic erosion, random crossings and mislabelling of genotypes.

The most critical step in a plant breeding programme is the understanding of available genetic diversity. As such, many studies have been carried out to analyse genetic diver-sity in coffee across many producing countries such as DR Congo (Vanden Abeele et al. 2021), Ethiopia (Alemayehu 2017; Benti et al. 2021), Kenya (Gichimu et al. 2012; Omingo et al. 2017), Uganda (Kiwuka et al. 2021) and Indonesia (Ramadiana et al. 2018). these studies provide important information on the distribution and variation of a series of desirable (and undesirable) traits in coffee that are needed to develop a new cultivar or conserve existing cultivars on the basis of their value. Variability analysis have been carried out in coffee using morphological markers (Akpertey et al. 2019), biochemical markers (Herrera and Lambot 2017) and molecular markers (Jingade et al. 2019) mainly in Northern Africa, Central Africa and other continents. There is paucity of data on genetic diversity or characterization on coffee genotypes produced in Southern Africa (Angola, Malawi, Tanzania, Zambia and Zimbabwe). This is despite coffee being an important economic activity in the region, supporting many smallholder farmers and other stakeholders in the coffee value chain. For example, in Zimbabwe, coffee is the second most profitable crop after tobacco and produced by around 5000 smallholder farmers, many of which rely on coffee as their main source of livelihood (Chemura et al. 2016).

Preliminary studies have shown that common coffee genotypes are produced across Southern Africa in Malawi, Tanzania, Zambia and Zimbabwe (Murphy et al. 2008). Other studies have shown that the response of these genotypes to environmental conditions differ between commercial genotypes in the region (Baker et al. 2001; Chemura et al. 2014). With many studies indicating increases in production challenges due to climate change and variability (Kutywayo et al. 2013; Chemura 2014), it is important that research and resources be channeled towards the improvement of coffee to mitigate against these challenges. One of the sustainable ways of mitigating against these challenges is coffee breeding to widen its genetic base (Herrera and Lambot 2017; Akpertey et al. 2019). Coffee germplasm with a broad genetic base can offer a more durable solution to adapt to climate variability, withstand biotic pressures and ensuring profitability at farm level.

Generating information on the nature and magnitude of variability, heritability and expected genetic gains in coffee populations is critical for the development of a successful coffee breeding program. The success of any crop breeding program depends on the variability present in a population as well as the extent to which the variability can be inherited to other generations, which sets the limit of progress that can be achieved (Falconer and Mackay 1996; Teressa et al. 2010; Ferreira et al. 2020). Heritability is a measure of the phenotypic variance which is attributed to genetic cause.

Information on heritability influences the choice of selection procedures most useful to improve traits, predict gain from selection and to determine the importance of genetic effects (Falconer and Mackay 1996; Laghari et al. 2010). Traits with high heritability estimates are easily inherited and fixed using simple selection methods, while traits with low heritability estimates are slow to fix. Studies have also proven that heritability estimates alone, without genetic advance estimates cannot give adequate information regarding the predicted gain from selection (Najeeb et al. 2009; Ogunniyan and Olakojo 2014). Genetic advance gives information on expected gain in a character under a given selection pressure. High heritability estimates toge-ther with high genetic advance indicates the presence of additive genes, which signifies the most suitable conditions for selection (Ogunniyan and Olakojo 2014).

The use of morphological traits in genetic variability studies is the first step in studying genetic relationships among individuals. Different researchers across the globe reported different results on coffee variation, relationships, heritability estimates and genetic advance for different morphological traits (Beksisa and Ayano 2016; Akpertey et al. 2019; Alemayehu 2019, Merga et al. 2021). These studies have not been carried out in Zimbabwe, where most of the coffee cultivars are common in Southern Africa as well. This study was therefore carried out to understand the coffee diversity, heritability estimates and relationships among the coffee germplasm found in Zimbabwe.

Study site and materials

The study was conducted at Coffee Research Institute (CoRI), situated 8km South West of Chipinge town. Its geographic position is (20⁰21¹S and 32⁰37¹E) and lying at an altitude of between 1060 and 1290 metres above sea level (masl). In terms of the agro ecological classification used in Zimbabwe, CoRI is in Natural Region 1 where mean maximum and minimum temperatures are 28 and 14℃ respectively. Annual rainfall varies from 800-1300 mm. A large portion of the soils at CoRI are derived from Umkondo quartzite which form the oxisols group and another portion is made of sandstones that are leached and strongly weathered (Fao 2006).

Twelve genotypes coded CR1 to CR12 were used in the study. The genotypes were a collection of coffee material in Zimbabwe and Southern African countries. These geno-types were conserved in the coffee gene bank at Coffee Research Institute, and represent useful genetic resources cultivated by farmers as well.

Experimental design and data collection

The trial was superimposed in the coffee gene bank planted in November 2009. The twelve genotypes were laid out in a randomized complete block design (RCBD) in three blocks. Each block had twelve plots, with each plot having 15 coffee trees. In row spacing was 2.5 m × 2 m. All coffee management practices from planting were carried out according to the standard coffee handbook (Logan and Biscoe 1987). Data were collected from five central coffee plants from each plot. Data collection was collected on quantitative traits. The selected bushes were marked by a plastic yellow paper for ease of identification. Mature coffee was harvested after every two weeks from the first harvest and weighed using a balance. The total yields were bulked in each replication for each genotype. Data on percentage of ripened fruits at each harvest were also collected. A count on the number of primary branches per plant, bearing branches on the five central coffee trees per plot was taken. Leaf area was considered very important in this study. A random selection of five middle leaves were recorded on marked primary branches and an average of five middle leaves was taken for each replicate. Leaf length was measured using a 30 cm ruler along the midrib, from the area where the leaf attaches to the stem to the end where the midrib ends. Leaves from which leaf length was taken were also measured for leaf width using a 30 cm ruler across the longest part of the leaf. Data on seed length was collected by measuring the longest part of the seed using Vernier calipers. This was done after removing coffee pulp to remain with coffee seed with parchment only. Seed width (mm) was considered to be the length across the mid part of the seed using Vernier callipers.

Data analysis

Data collected on quantitative characters were subjected to variability analysis in R (R Core Team 2022). Descrip-tive statistics, phenotypic variance, genotypic variance, environmental variance, Genotypic coefficient of variation (GCV), phenotypic coefficient of variation (PCV) were estimated using the formulas proposed by Deshmukh et al. (1986).

Phenotypic variance (δ²p)

[1]δ2p=δ2g+δ2e

Genotypic Variance (δ²g)

[2]δ2g=MSg−MSe/r

Genotypic coefficient of Variation (GCV)

[3]GCV=δ2g*100x¯

Phenotypic coefficient of Variance (PCV)

[4]PCV=δ2gx¯*100

Where, δ²g = Genotypic variance, δ²p= Phenotypic variance, δ²e = Environmental Variance and x¯
= Grand mean, MSg = Mean Square of genotypes, MSe = Mean square error.

Estimation of Heritability and genetic advance were also calculated.

Broad sense heritability (H) values were computed based on formulae by Falconer and Mackay (1996) as follows:

[5]H=δ2g/δ2p*100

The heritability estimate was categorized as described by Robinson et al. (1949) as follows:

0-30% = low, 30-60% = medium and > 60% = high.

Genetic advance under selection (GA)

GA was estimated using the formulae:

[6]GA=KHδ2p

Genetic advance as percentage of mean (GAM)

[7]GAM=GAx¯*100

Where Where, δ²g = Genotypic variance, δ²p = Phenotypic variance and x¯
= grand mean, H = heritability in the broad sense, GA is expected genetic advance, k = Selection differential at 5% selection intensity (k = 2.063).

Principal Component Analysis (PCA) was carried out for quantitative traits to determine the contribution of different characters to total variation. Correlation analysis was done to determine the relationships among morpho-logical traits measured in this study. Data on quantitative traits was further subjected to cluster analysis. The clus-tering was done using Unweighted Pair-Group Method with Arithmetic Average (UPGMA), showing clusters of different coffee genotypes.

Genetic variability, heritability and genetic advance estimates

There were highly significant differences (P < 0.05) in coffee yield, plant height, stem girth, primary branches, leaf properties and number of branches per plant. There was no significant variation (P > 0.05) in seed traits and percentage of bearing branches. Estimates of genotypic variance, phenotypic variance, genotypic coefficient of variance and phenotypic coefficient of variance are presented in Table 1. High values of PCV were obtained in leaf area and percentage ripening which were 25.1% and 33.8% respectively. The other traits showed a moderate PCV which ranged between 10.5-17.8%. Moderate PCV and GCV values were obtained for percentage ripening, coffee yield, plant height and leaf traits. Low PCV and GCV values were obtained from stem girth and seed width. Low GCV were obtained from stem girth, percentage of bearing branches and seed traits with values ranging from 4.9% to 8.5% (Table 1).

Table 1 . Mean performance of twelve coffee genotypes under field conditions.

Traits	Mean	Min	Max	%CV	GV	PV	GCV%	PCV%	H2(%)	GA	GAM	P
Yield (kg)	1726.0	943.8	2141.6	8.0	75510	94514	15.9	17.8	79.9	506	29.3	***
PH (cm)	84.3	53.9	111.6	5.1	199.1	217.8	16.7	17.5	91.4	27.8	33.0	***
SG (cm)	25.8	21.8	31.6	6.0	4.8	7.2	8.5	10.4	67.1	3.7	14.3	***
BP	71.6	51.4	93.2	12.7	33.3	116.6	8.0	15.1	28.5	6.3	8.9	*
BB	58.1	42.1	76.2	13.1	33.6	91.6	9.97	16.5	36.7	7.2	12.4	*
PBB	81.4	62.2	94.3	9.1	18.1	73.2	5.2	10.5	24.8	4.4	5.4	ns
PR	57.2	25.8	91.7	26.9	137.3	374.7	20.5	33.8	36.6	14.6	25.5	*
LL (mm)	147.8	109	181	7.9	218.3	355.1	10.0	12.8	61.5	23.9	16.1	***
LW (mm)	59.3	40.2	85.1	10.5	4.4	82.7	11.2	15.3	53.6	10.0	16.9	**
LA (mm2)	54.5	26.2	78.2	15.3	118.3	187.9	19.9	25.1	63.0	17.8	32.6	**
SW (mm)	6.6	5	9	12.3	0.11	0.78	4.9	13.3	13.8	0.25	3.78	ns
SL/W	1.7	1.3	2.3	14.1	0.001	0.06	2.25	14.3	0.03	0.01	0.74	ns

Significantly different at P = 0.05*, P < 0.01**, P < 0.001***.

BB: bearing branches, LL: leaf length, LW: leaf width, LA: leaf area, SW: seed width, PR: percentage ripening, PH: plant height, SG: stem girth, BP: branches per plant, PBB: percent bearing branches.

Broad sense heritability estimates for the quantitative traits ranged from 0.03% for seed LW ratio up to 91.4% for plant height (Table 1). High heritability (> 60%) estimates were observed for plant height (91.4%), coffee yield (79.9%), stem girth (67.1%), leaf length (61.5%) and leaf area (63%). Low heritability estimates (0-30%) were ob-tained for seed traits, number of branches per plant and percentage of bearing branches per plant. Values of the expected genetic advance (GA) from selecting 5% of the genotypes revealed a gain of 506 kg for coffee yield, 33.0 cm for plant height, 14.3 mm for stem girth, 17.8 mm² for leaf area and 10.0 mm leaf width. High GAM was obtained for coffee yield (29.3%), plant height (33.0%), leaf area (32.6%) and percentage ripening (25.6%). Low GAM values were obtained for seed traits and the number of branches per plant.

Principal component analysis (PCA) for quantitative morphological traits

The relative importance of each of the quantitative traits was analysed using Principal Component Analysis (PCA). Table 2 is showing the principal components, loading vectors of each variable, and the cumulative variance of each PC. The relative discriminating power of the PCA was high for PC1, PC2, PC3 and PC4 having eigen values of 6.39, 2.21, 1.31 and 1.17 respectively (Table 2). The first principal component (PC1) explained 37.5% of the entire variability. Variation in the first PC was mainly contributed by yield (‒0.315), plant height (‒0.3176), leaf area (‒0.356), stem girth (‒0.273) and leaf length (‒0.32). PC2 accounted for 18.1% of the total variation, being dominated by seed length (0.44), leaf width (0.33), percentage of bearing branches (0.38) and leaf area (0.31). The third and fourth PC contributed 9.8% and 8.9% respectively, with the majority of the contributions coming from coffee yield, seed width, stem girth, percentage of bearing branches and percentage ripening.

Table 2 . Contribution of individual traits and cumulative variances for the different principal components of coffee.

Trait	PC1	PC2	PC3	PC4
BP	‒0.296	‒0.28562	0.085396	0.019427
LL	‒0.32457	0.282923	0.179216	‒0.29275
LW	‒0.29113	0.335714	0.117145	‒0.02443
LA	‒0.35602	0.317559	0.19405	‒0.12491
SL	‒0.05803	0.443363	‒0.0528	0.027801
SW	‒0.29726	‒0.02328	‒0.50394	‒0.35799
SLWR	0.247689	0.288651	0.432006	0.422165
BB	‒0.3707	0.022949	‒0.11226	0.26648
PBB	‒0.1502	0.389093	‒0.27474	0.366745
PR	0.089786	0.09694	‒0.45947	0.355048
PH	‒0.31768	‒0.27816	0.048067	0.256147
SG	‒0.27265	‒0.20153	0.40399	0.031524
yield	‒0.31522	‒0.25513	‒0.0334	0.438627
Eigenvalues	2.2085	1.5357	1.12636	1.07784
Variance	0.3752	0.1814	0.09759	0.08937
Cumulative
Variance	0.3752	0.5566	0.65418	0.74355

BP: branches per plant, BB: bearing branches, PB: percentage of bearing branches, LL: leaf length, LW: leaf width, LA: leaf area, SL: seed length, SW: seed width, FB: fruits/branch, PR: percentage ripening, PH: plant height, SG: stem girth.

Correlations between coffee morphological traits

Significant Pearson’s correlation coefficients were ob-tained for the different quantitative traits measured in this study. There were significant associations between yield and branches per plant (r = 0.685), seed width (0.621), seed length to width ratio (‒0.702), bearing branches (0.777), plant height (0.888) and stem girth (0.640). Significant correlations were also obtained between plant height and stem girth (0.606), bearing branches (0.69) and branches per plant (0.59). Seed width was also positively correlated with plant height (Table 3).

Table 3 . Correlation between morphological traits in coffee.

	BP	LL	LW	LA	SW	SLWR	BB	height	girth	yield
BP
LL	0.202
LW	0.298	0.685*
LA	0.344	0.924***	0.884***
SL	‒0.122	0.433	0.408	0.446
SW	0.591*	0.469	0.496	0.556
SLWR	‒0.605*	‒0.273	‒0.261	‒0.314	‒0.861***
BB	0.794**	0.394	0.528	0.574	0.557	‒0.489
PR	‒0.310	‒0.197	0.064	‒0.139	‒0.194	0.223	‒0.185
PH	0.590*	0.338	0.226	0.411	0.681*	‒0.732**	0.691*
SG	0.712**	0.548	0.326	0.537	0.379	‒0.542	0.649*	0.606*
yield	0.685*	0.302	0.337	0.419	0.621*	‒0.702*	0.777**	0.888***	0.640*

BP: branches per plant, LL: leaf length, LW: leaf width, LA: leaf area, SL: seed length, SW: seed width, SLWR: seed length to width ratio, BB: bearing branches, PR: percentage ripening, PH: plant height, SG: stem girth.

Clustering of coffee genotypes

The relationships between the coffee genotypes were evaluated using cluster analysis based on 13 quantitative traits. The results of the clustering revealed that coffee genotypes have at least 70% similarity. Considering a similarity level of 85% (shown by the red line in Fig. 1), four groups can be formed. The first group comprised of three genotypes CR01, CR12 and CR03. This group had the highest proportion of ripened fruits at each harvest. Also, the group had the second highest number of branches, yield, stem girth and height, on top of every other trait. The second cluster had one genotype CR08. The third cluster had six genotypes which were mainly distinguished by average trait performance across all the measured traits. The fourth cluster had two genotypes; CR06 and CR07 which were mainly distinguished by excellent performance across the majority of traits such as highest yielders, tallest plants and highest proportion of bearing branches (Fig. 1).

Figure 1. Cluster dendrogram showing similarity of coffee genotypes based on quantitative traits measured at Coffee Research Institute under field conditions. The solid red line shows the cut-off point used to determine the number of clusters.

The genetic distances between clusters for the 12 coffee genotypes basing on quantitative traits is given in Table 4. The minimum inter cluster distance (4.6) was recorded between cluster 1 and 2, while the highest interclass distance (7.7) was between clusters cluster 2 and cluster 4. Intraclass variation was high for cluster 3 followed by cluster 1 (Table 4).

Table 4 . Inter-class and intraclass (*) distances for the coffee quantitative traits.

	Cluster 1	Cluster 2	Cluster 3	Cluster 4
Cluster 1	3.239581*
Cluster 2	4.577103	0*
Cluster 3	4.624531	6.481865	4.230744*
Cluster 4	5.278984	7.746616	5.575188	3.156091*

The aim of this study was to evaluate genetic variation, heritability and relationships of coffee genotypes grown in Zimbabwe using morphological traits. The study also sought to identify the most important traits which can be used to distinguish between coffee genotypes and their associations in order to enhance opportunities for the utilization of coffee germplasm in production and breeding programs.

The efficient exploitation of coffee germplasm lies in proper understanding of genetic variation and the relation-ships with respect to key growth and yield parameters (Reuben and Misangu 2003). The study showed significant variation in morphological traits such as coffee yield, stem girth, plant height, leaf traits and primary branches. The existence of such variation provides evidence that selection is possible. The use of morphological markers in detecting variation in coffee was vindicated. Similar studies ob-served a wide range of morphological variation among coffee genotypes in traits such as coffee yield, plant height, stem girth, number of branches, fruit traits and leaf characteristics (Beksisa and Ayano 2016; Akpertey et al. 2019; Merga et al. 2021).

Significant differences in leaf length, leaf width and leaf area in coffee varieties means that the coffee varieties are genetically diverse in leaf characteristics. Leaf area is an important indicator of radiation interception, water balance, energy conversion and yield. It is a reliable estimate of the results of interventions such as fertiliser, moisture, genetics and the interaction of these factors. Genotypes with large leaves utilize more photosynthetically active radiation required for photosynthesis. Significant variation in leaf traits agrees with a study by Cesar et al. (2010). On the other hand, Hue (2005) reported insignificant variation in coffee leaf traits. Therefore, the observed morphological diversity provides possibilities for the identification and selection of diverse germplasm to develop new coffee varieties.

Results of this study revealed small range of values between GCV and PCV values for yield, plant height and stem girth. According to Weldemichael et al. (2017), the extent of environmental effect on any trait is indicated by the magnitude of the difference between phenotypic and genotypic coefficient of variation. The small difference between GCV and PCV values for plant height and coffee yield indicates minimal influence of the environment on expression. This therefore implies that selection of coffee basing on yield and plant height will be effective to improve coffee. On the other hand, huge differences existed between GCV and PCV values for branches per plant, seed width, seed length to width ratio and percentage ripening. Selection of coffee using these traits would thus not bring any significant genetic gain.

Estimates of broad sense heritability were classified as low (< 30%), moderate (30-60% and high (> 60%), according to Robinson et al. (1949). According to Najeeb et al. (2009), heritability estimations together with genetic advance estimates of the mean gives very good estimates of predicted gain under selection than heritability alone. Basing on this classification, coffee yield and plant height had high heritability and genetic advance estimates of the mean (GAM). These high estimates for both heritability and GAM imply that selection of these traits can be very effective as these traits have limited influence of the environment in their expression. According to Beksisa and Ayano (2016), when heritability and GAM are high, any selection technique will result in a significant genetic gain provided the environment is maintained at a constant. The high heritability estimates in this study indicates that coffee genotype played important roles in determining its pheno-type, implying the existence of both additive and dominant gene effects in inheritance. In addition, because coffee is high value crop with long gestation period, selection for plant height at early growth stages may provide oppor-tunities for increasing the efficiency and speed required in coffee breeding.

Moderate heritability estimates combined with mode-rate to low GAM were obtained for stem girth, percentage ripening, and leaf width. This implies that these traits are influenced by moderate environmental effects, and are likely governed by both additive and non-additive types of gene action (Yirga et al. 2020). It is therefore complicated and most unlikely to improve coffee using simple selection methods for these traits since they exhibit low genetic gain. The use of molecular markers, progeny testing, the use of pedigree and genotype trait associations may provide addi-tional information on the relevance of these traits in coffee selection and breeding. The results of this study are in agreement with previous work. For instance, Weldemichael et al. (2017) reported high heritability and GAM estimates for coffee yield and plant height while high heritability estimates and low GAM were observed for stem girth, nodes per primary and average internode length. Similar results on stem diameter were also reported by Beksisa and Ayano (2016). In related studies, high heritability and GAM estimates were also recorded for coffee yield and plant height (Alemayehu 2019). This information is very valuable for the Southern African community in the context of coffee improvement.

The relative contribution of the quantitative traits was assessed using the Principal Component Analysis. Accord-ing to Iezzoni and Pritts (1991), PCs with eigenvalues greater than 1 are the most important in explaining variance in the data. The results revealed that four PCs had important contributions to variation as they have eigenvalues greater than 1. The four PCs contributed a total of 74.4%. Traits such as coffee yield, plant height, branching traits and leaf width contributed much in PC1 indicating that these traits are important in discriminating between coffee genotypes studied. The contribution of the four PCs was comparable to work done by Gessese et al. (2015), Gichimu and Omondi (2010) where the first two PCs obtained a total of 60% variation and Akpertey et al. (2019) who obtained variation of 71.3% in the first three components. Generally, traits with large absolute values close to a unity influence clustering more than those with lower absolute values closer to zero. This therefore implies that coffee yield, plant height, number of bearing branches and leaf area played a major role in showing variation among genotypes in this study. These traits should therefore be considered in selecting parents for hybridization programs in Southern African communities.

Knowledge of association among various quantitative traits in coffee is important in the design of breeding strategies, as it provides opportunities for indirect and simultaneous selection of traits. The correlation table in our analysis have revealed some important associations among traits. The positive relationships between coffee yield and the number of branches per plant, seed width, number of bearing branches per plant, plant height and stem girth is an indication that selection for yield can be done indirectly using these traits. An increase in one of these traits would result in a corresponding increase in coffee yield. These results agree with Atinafu and Mohammed (2017) who observed that coffee bean yield was positive and signi-ficantly correlated with hundred green bean weight, stem diameter and percent of bearing primary branches. In another study, Akpertey et al. (2019) observed a positive significant relationship between coffee yield components and number of branches. Correlated traits offer oppor-tunities for indirect selection of traits which are difficult to select. In addition, selection of coffee at vegetative growth stage may offer opportunities for early selection and simul-taneous improvement of traits. This will be a significant milestone in quickening breeding since coffee is a peren-nial crop with a long gestation period.

Cluster analysis was used to classify coffee genotypes based on their morphological traits. The analysis clearly classified coffee into four groups. The groupings revealed that coffee genotypes have at least 70% similarities. High levels of similarity were expected as the genotypes evalu-ated in this study are Arabica coffee, a self-pollinating specie. Selfing promotes homozygosity, and a reduction in variation among plant species (Hue 2005; Wright et al. 2008). The four groups formed were mainly separated by traits such as coffee yield, plant height, number of bearing branches and leaf area. The fourth cluster had genotypes, which produced the best coffee in terms of height, branches and yield. The genotypes can be of use in selections that targets high yield and vegetative traits; tallness and branch proliferation. Three genotypes CR12, CR01 and CR03 grouped together, and were mainly distinguished by early maturity, evidenced by the highest proportion of ripened fruits at harvest. While coffee is harvested manually in Zimbabwe, and other Southern African countries, the selection of genotypes with uniform maturity may offer opportunities for machine harvesting in the future (De Sousa Filho et al. 2015).

It is very important that efforts be directed towards widening genetic constitution of coffee genes in Zimbabwe. Broad genetic base is generally associated with a reduction in negative inbreeding effects as well as the potential to adapt to any environmental challenges (Frankham 2005; Wernberg et al. 2018). In other parts of the world, genetic diversity studies were conducted using morphological and agronomic traits. For instance, Gichimu and Omondi (2010) observed that coffee has a narrow genetic base with variation of about 25% considering coffee yield, canopy size, leaf shapes and seed traits. In that study, the narrow genetic base was attributed to continuous selection in coffee by farmers. From a breeding standpoint, it is very important to consider genes in other species of coffee to widen the genetic base. In order to have diverse coffee genotypes, it is important to introduce more coffee from other countries. Coffee breeding programs need to focus on yield improvement, with emphasis also being given to the correlated traits as they also give an indication of yield performance of a variety.

The study also determined the inter and intra cluster distances among coffee. Cluster distances are important determinants of genetic diversity when selecting parents to use in crop improvement programs. Cluster analysis in this study produced four clusters which differed within and between the clusters. Highest interclass distance between clusters 2 and 4 gives opportunities for the production of transgressive segregates and heterosis maximisation (Rahim et al. 2010; Tounekti et al. 2017). On the other hand, small intercluster and within cluster differences provides opportunities for the production of pure lines. Unique genes can only come from these divergent cluster families. The existence of both very divergent and uniform genotypes among coffee germplasm was also observed by Muvunyi et al. (2017).

The overall aims of this study were to report genetic variability, heritability and relationships among morpho-logical traits in common coffee genotypes in Southern Africa, targeting important traits to select for breeding. Significant differences, high heritability estimates, positive significant correlations among major traits such as plant height, coffee yield, number of bearing primaries and stem girth were obtained. Selection of these traits was found to be possible using phenotypic traits, while other low heritable traits could be selected using molecular markers and direct selection procedures. The study revealed a low genetic diversity for quantitative traits, with possibility of producing transgressive segregates from the most divergent families. The need to broaden genetic diversity in Zimbabwe through introductions, selection and hybridiza-tions for further breeding programs was discussed.

My acknowledgements goes to the Coffee Research Institute agronomy and plant pathology staff for helping in data collection. Special mention also goes to Dr Abel Chemura for his valuable contribution in this manuscript.  

The authors declare that they have no conflict of interest.