Genetic Variability of Coffee ( Coffea arabica L.) Germplasm in Biennial Bearing and Its Influence on Selection Efficiency

Arabica cofee is a perennial cash crop and highly afected by biennial bearing which disturbs farmers’ annual income and world’s cofee industries. Developing nonbiennial bearing variety is prominent in addition to applying feld management practices. Tis study was conducted from 2012 to 2020 in southwestern Ethiopia at Tepi and Gera to test the extent of genetic variability among Arabic cofee germplasm in biennial bearing and understand the infuences of bienniality on advanced selection. Te pooled analysis of variance revealed handiness of genetic variability in yield and biennial bearing. Te moderate genotypic coefcient of variation (10–20%), heritability (30–50%), and high genetic advance as percentage of the mean ( > 20%) were manifested in yield and biennial bearing. Response to selection and selection efciency were negatively afected by biennial bearing. Early selection excluded 30–40% of the top high yielders from advanced selection. Selection at four harvesting seasons revealed 90% and more selection efciency. Tus, one has to be conscious of the alternate bearing nature of lines during advanced selection. Both T43/11 and T51/11 were among the top high yielders and showed low biennial bearing at Gera and Tepi. T33/11, T49/11, T55/11, and T61/ 11 showed very low biennial bearing at both locations. Tese are promising lines and could be recommended for further biennial bearing improvement breeding programs.


Introduction
Cofee is a perennial cash crop that belongs to the family Rubiaceae and the genus Cofea which consists of 124 species [1]. Among these species, Cofea arabica L. and Cofea canephora P. are the dominant species in the world cofee market. Arabica cofee contributes 65% of worlds' cofee production, and 35% contributed from Canephora production. Te former is tetraploid (4x � 2n) and selfpollinated, and the latter is diploid (2x � 2n) and selfincompatible species.
Among the headache in perennial crop production, bienniality is an alarming issue. Many fruit trees and other horticultural crops such as mango, apple, pear, apricot, and avocado are highly afected by bienniality [2][3][4]. Also, Arabica cofee is one of the horticultural crops that are afected by biennial bearing. Tus, the Cofea arabica L. yield is fuctuating or higher one year and lower the next [5,6]; this of course afects farmers' annual incomes. Te fuctuation of cofee yield results in food security problem, especially in developing countries such as Ethiopia. Te producers' of and on harvesting and supply afect the total cofee production and demand in the world market. Tis has become a bottleneck for world's cofee industries today.
Te biennial bearing nature in yield performance afects selection efciency in cofee; this is due to its negative efect on yield accumulated over years [2,7]. Moreover, bienniality causes heterogeneity growth variable and temporal correlation pattern over multiple harvesting seasons; this fuctuation in cofee yield makes the process of selecting the best performing progenies/lines difcult. Finally, this may lead to wrong conclusions for any gauge of variance components and selection of promising lines that exhibit biennial bearing nature.
Despite enormous agronomic management practices applied to alleviate biennial bearing in Arabica cofee, reducing it to the required level became still difcult for cost users to steadily generate the annual income from cofee production. Tus, developing regular bearing variety which can regulate the undesirable traits of heavy and light bearing is more economical. It is well known that regular bearing/ nonbiennial cultivars are preferred to biennial bearing cultivars [8]. Tis cultivar helps farmers get stable annual income, stable world cofee industries, and reduce the gap between supply and demand in the world cofee markets [9].
Cofee species respond diferently to bienniality efects; the production of Conilon robusta is relatively less afected by biennial production change [10]. However, relative to robusta types, Arabica is highly afected by bienniality [10]. Te diference between these two species implies that bienniality in cofee can not only be controlled by genes but also by environmental factors and feld management practices. Variability in biennial bearing is reported for some fruit trees; the biennial bearing of stone fruits such as mango, olive, and plum varies from cultivar to cultivar [11]. Also, variability reported among apple and pear genotypes is in biennial nature [2,12]. Additionally, Guitton et al. [13] confrmed that the gene related to hormone is more responsible for biennial bearing than fower related gene in apple. In the feld, variability is observed among Ethiopian Arabica cofee germplasm in biennial bearing. However, so far, no well-planned and designed implementation has been conducted for studying genetic variability among these cofee germplasm in biennial bearing and its efect on advanced selection. Tus, this study was designed with the main objectives to evaluate the response of Arabica cofee to biennial bearing and identify the biennial bearing efects on selection efciency.  (Table 2). Augmented design was used at both testing sites. Six cofee trees were planted per plot with the spacing of 2 m × 2 m between plants and rows. Te feld managements of the experiments such as shade and fertilizer application were applied according to Endale et al. [14].

Methods and Data
Recoded. Red cherry of cofee yield data was selectively picked and recorded per plot in gram. All cofee berries are not changed to red cherry at the same time. After red cherry is completely harvested, the leftover dry and green cofee fruits were collected separately and recorded in gram which were later changed to red cherry using 2.26 and 1.04 conversion factors, respectively [15]. Te mean of red cherry was computed by dividing the total amount of red cherry in gram per plot for the total number of bearing cofee trees per plot. Ten, the mean of red cherry was converted to clean cofee yield in Qha −1 , multiplying red cherry by 0.00417 (conversion factor). Finally, the yield data in Qha −1 were converted to kg·ha −1 which is the SI unit for weight.

Data Analysis.
Five and four years of clean cofee yield data were analyzed for Tepi and Gera locations, respectively. All collected data were subjected to R-software (version 4.3) for statistical analysis. Data uniformity was tested using the Shapiro-Wilk test; earlier, combined data analysis homogeneity variance was tested using the F-max method.
Te alternate or biennial bearing of cofee germplasm has been characterized with several descriptive statistics [16] and the mean relative diference index as Hoblyn et al. [17]. I � 1/n − 1 (|y t − y t−1 |/y t + y t−1 + |y t+1 − y t−2 |/y t+1 + y t−2 + . . . + y t+1 − y t /y t+1 + y t ), where y t is the t th observed yield in an ordered series of size n, |y t − y t−1 | is the absolute value of the diference in yield between two successive years t and t − 1, and y t + y t−1 is the sum of the yields over these two years and then standardized over the total number of years in the time series, n, minus one. I varies between 0 and 1, with I � 0 representing no alternate bearing behavior and I � 1 corresponding to strict alternate bearing behavior.
Relative percentage of biennial bearing (RP): it was calculated according to Morettini [18] and Singh [19] as follows. RP � difference between two successive crops x100/sum of the two successive crops. It is obvious that the index RP can vary between 0, in the case of a regular bearing pattern, and 100, in the case of a pronounced biennial bearing pattern.
For per year phenotypic analysis, the following linear model is utilized: Y ijk � µ + g i + b k + ε ik , where y ik is the phenotypic value for the genotype i and the block k, µ is the population mean, b k is the efect of the kth block, g i is the random efect of the ith genotype, and ε ik is the random efect of residual.   International Journal of Agronomy

Components of Variance.
Error (σ 2 e) and genotypic (σ 2 g) and phenotypic (σ 2 p) variance were computed by the following formula suggested by Hallauer et al. [20] and Singh and Chaudhary [21]. h 2 � s 2 g /sp 2 , where s 2 g is genotype, sp 2 is phenotypic variance, and h 2 is broad sense heritability. Te following random model is used to estimate variance components and response to phenotypic selection of pooled analysis: where y ijk is the phenotypic value for the genotype i, year j, and block k, µ is the population mean, h j is the random efect of year, g i is the random genotypic efect, b k is the efect of kth block, gh ij is the interaction random efect between genotypes and years, and e ijk is the random efect of residuals.
is the interaction between g and y, s 2 e is the experimental error variance, and r and y are the number of replicates and years, respectively. Response to selection: RS � ih σp, where i is the selection intensity (at 5%), h is the square root of heritability (√h 2 ), and σp is the phenotypic standard deviation.

Variability in Biennial Bearing and Yield.
In most years or harvesting seasons including over year mean, a nonsignifcant diference was observed among accessions in yield and biennial bearing at both locations (Table 3). Tis is due to a high mean square of error against which the whole accession mean square was tested. However, a highly signifcant (P < 0.01) diference in yield performance was observed among cofee accessions in the 2020 harvesting season at Tepi and signifcant diference (P < 0.05) in yield performance at Gera in the same harvesting season; this may be due to a diference in yield potential expression of cofee accessions as the evaluation years extend. Likewise, variability among Arabica cofee accessions in clean yield was reported by many investigators [15,[22][23][24][25]. At Tepi, a highly signifcant diference in biennial bearing was observed among testing materials during early harvesting seasons (2017 and 2018). Tis resulted from the bienniality range recorded from 0 (nonbiennial) to 38.9 in these consecutive years (Table 4). In agreement, at early stage, signifcant variability was detected among pistachio in alternate bearing [24]. Despite nearly null to complete range in biennial bearing being revealed among accessions including standard checks, a nonsignifcant diference was observed at Gera (Tables 3 and 5), which resulted from a high mean square of error (high environmental contribution) against which bienniality of cofee was tested.
At Tepi, a high genotypic coefcient of variance (GCV > 20%) manifested in yield across harvesting seasons except in 2016 and 2018, whereas from over year mean of yield, moderate GCV (10-20%) was recorded. However, high PCV (>20) was observed across all seasons. High and moderate genetic gains as percentage of the mean (GAM > 20) and (GAM 10-20%) were observed in these harvesting seasons at this location. High heritability (62.9-88.71%) was recorded for yield in 2017, 2019, and 2020, but moderate to low was observed in the rest seasons. Moderate GCV (10-20%) was recorded at Gera except in 2020 (which was 26.69%), but except in 2017, high PCV (>20%) was observed in all seasons including over year mean   International Journal of Agronomy of yield. In line with this, fuctuation of genetic parameters' results was observed across harvesting seasons in cofee yield [25,26]. From the over year mean of yield, moderate GCV (10-20%), GAM, and heritability were clearly pointed out at both locations (except GAM (>20) at Gera) which elucidated the existence of moderate genetic variability among accessions in yield performance. Concurring, moderate genetic variability among cofee accessions was found in yield [22,27]. Additionally, gene involvement for controlling biennial bearing had been detected [10,13]. By selecting the top 5% high yielders' genotypes from the accessions in 2020 harvesting season, it is possible to improve clean cofee yield by 504.84 kgha −1 and 909.16 kgha −1 at Tepi and Gera, respectively (Table 3). High GCV, GAM, and H (33.76% and 24.70%, 61.60% and 40.71%, and 79.22% and 64.65% for I2 and I3, respectively) were recorded from over two and three years for biennial bearing; moderate GCV (16.86 and 16.84%), H (46.71 and 47.75%), and high GAM (23.62 and 23.92%) were noted from over four (I4) and fve (I5) harvesting seasons, respectively, at Tepi (Table 3). Also, moderate GCV, H, and high GAM (19.89 and 18.02%, 36.09 and 37.22%, and 24.50 and 22.54% for I3 and I4, respectively) were observed at Gera which indicated that the handiness of genetic variability among cofee accessions in biennial bearing. Te present results confrmed with the fndings of Esmailpour [28] and Todd et al. [5] who reported variability among pistachio sp. in alternate bearing. From the two locations' genetic parameters results, it was elucidated that the possibility of alternate bearing improvement via selection and/or hybridization. From the population of the top 5% low biennial bearing genotypes, it is possible to mitigate bienniality to the average values of 0.06 and 0.08 at Tepi and Gera, respectively.

Alternate Bearing of Arabica Cofee across Harvesting
Seasons. Although the ANOVA of biennial bearing showed nonsignifcant diferences in some seasons, the bienniality range between pair of harvesting seasons was very large (Table 5). For the pair consecutive of harvesting seasons (2017 and 2018, 2018 and 2019, and 2019 and 2020), the relative percentage of bienniality ranges from 0.07, 0.04, and 0.39 to 100%, receptively. From theses ranges, it was clearly observed that the existence of almost nonbienniality (0.07 and 0.04 which was recorded by T13 and T34/11, respectively) to complete bienniality (100% which was recorded for T71, T37, T32, T83, and T29/11) among accessions was seen. Also, the bienniality range from 0 to 39.90% was observed between 2017 and 2018 harvesting seasons at Tepi (Table 4). For the others pair of consecutive harvesting seasons less than one (<1) to 90.68%, the range of bienniality was recorded at Tepi. Tis implies that the presence of genetic variability among Arabica cofee germplasms in biennial bearing which is a desirable trait for solving alternate bearing problem of commercially released varieties.
Te lowest relative bienniality between 2017 and 2018 was recorded for T13/11 genotypes at both locations (Tables 4 and 5). Accessions such as T19, T28, T35, T52, and T66/11 showed zero in biennial bearing between 2017 and 2018 harvesting seasons. Te relative biennial bearing for some accessions may vary from seasons to seasons and location to location; this indicates that the selection of cofee genotypes for less biennial to nonbienniality using relative biennial bearing of consecutive seasons should be supported by biennial bearing intensity. In agreement, variability of bienniality across seasons was reported for pistachio [24]. During nonbiennial bearing selection, one has to be conscious of the extraneous factors such as environment and weather condition of the harvesting seasons in addition to genetic factor.

Biennial Bearing Intensity (I) and Response to Selection (R).
Biennial bearing intensity showed an increase during early harvesting seasons (from 2016-2017) at Tepi (Figure 1 (A)) but showed a decrease at Gera (Figure 1 (B)). At early, over two and three harvesting seasons (2016-2017 and 2016-2018, respectively) response to selection an upsurge and decrease, respectively, such as biennial bearing intensity at Tepi. From over three harvesting seasons to four (2016-2019), the overall bienniality was increased exponentially; also, a relatively increment of response to selection was observed. From over four to fve seasons, the alternative bearing intensity and the overall mean performance almost remained constant, but response to selection was increased. Tis may be because sometimes cofee genotypes bear extremely high yield during on years relative to of years which may result in high biennial bearing and high phenotypic performance in yield which may contribute for increments of response to selection. Tus, biennial bearing and response to selection may be increasing or decreasing together or may  International Journal of Agronomy 5 show opposite relation. Concurring, Merga et al. [15] reported that bienniality nature of Arabica cofee could afect the genetic gain and response to selection. At Gera, biennial bearing and response to selection showed contrasting relation (Figure 1 (B)). In addition, at early harvesting seasons, biennial bearing decreased. However, response selection showed upswing. Over two to three harvesting seasons, bienniality was upraised, whereas the reverse was true for response to selection. Response to selection exponentially increased, while biennial bearing exponentially decreased at over three to four harvesting seasons. Te overall mean of yield across all harvesting seasons was increasing at this location. Te relationship between response to selection and biennial bearing at Gera was logically expected as bienniality under normal state negatively infuenced the overall yield performance which may afect response to selection of the top high yielder genotypes. Te negative impact of bienniality on genetic gain or response to selection was confrmed in Arabica cofee [15].

Biennial Bearing Intensity and Selection Efciency.
Cofee accessions showed better performance having low biennial bearing found in rank of the top high yielders when compared with accessions with high biennial bearing    International Journal of Agronomy (Tables 6 and 7). Tis is due to their inherency in consistent high yielding performance across years as compared to those exhibiting high biennial bearing. Cofee genotypes such as T60/11, T43/11, and T80/11 revealed 0.1, 0.22, and 0.25 biennial bearing intensity (I), respectively, at Gera (Table 6). Also, at Tepi, T41/11, T42/11, and T43/11 showed low biennial bearing intensity (0.31, 0.35, and 0.24, respectively) ( Table 7). Tese genotypes were ranked the top 15 (at Gera) and the top 10 high yielders (at Tepi) in all over year yield performance from over two (2YRS) to four (at Gera) and fve (at Tepi) harvesting seasons. For such cofee genotypes, early selection (before or at three harvesting seasons) may be as efcient as late (four or late harvesting seasons) selection. T41/11 showed high relative percentage of biennial bearing across overall harvesting (72.44% from over two seasons and 84.25% from over three seasons) and biennial intensity (0.54) at Gera (Table 6). Tus, it ranked below the top 15 high yielders in yield performance by over two and three (3YRS) mean of yield; thus, early selection may exclude such genotypes from the next breeding program. Likewise, biennial bearing intensity ranges from 0.19 to 0.93 and 0.14 to 0.89 were recorded for pecan and pistachio, respectively [24,29]. Some cofee genotypes such as T70/11, T78/11, and T85/11 showed high relative percentage of bienniality (RP) and biennial bearing intensity (0.5, 0.58, and 0.42, respectively) at Gera (Table 6); also, T37/11, T38/11, and T48/ 11 recorded high RP in pair of consecutive seasons and biennial intensity (0.4, 0.49, and 0.47, respectively) at Tepi (Table 7). But from population, they ranked the top 15 high yielder in over two, three, and four harvesting seasons (at Gera) and fve harvesting seasons (at Tepi) (Tables 6 and 7). Additionally, genotype T78/11 had recorded the highest biennial bearing intensity (0.58) and high RP (which were 82.94 and 81.46% in RP1 and RP2, respectively); however, it ranked 1 st , 1 st , and 2 nd by the mean of 2YRS, 3YRS, and 4YRS, respectively. Tis resulted from the extremely high yield recorded by this genotype during on years relative to of year and other genotypes that compensate its low yield in of seasons. Tus, the mean of yield over seasons maintains superiority of genotypes, while a high diference in yield between on and of seasons causes high alternate bearing intensity. Tus, while promising line selection, cofee breeders should be conscious of the biennial bearing nature of cofee genotypes.
High relative percentage of biennial bearing was observed early in RP1 and RP2 at Gera and in RP1 at Tepi (Tables 6 and 7). Depending on the top 15 high yielders by 4YRS mean yield performance, 60 and 66.7% of the top 15 were selected by 2YRS and 3YRS, respectively, at Gera (Table 6), whereas at Tepi, 70, 80, and 90% of the top 10 high yielders were selected by 2YRS, 3YRS, and 4YRS when compared with 5YRS of the top 10 high yielders (Table 7). Tus, the two locations' results elucidated that selection is ideal and more efcient at fourth harvesting seasons; selection of early four harvesting seasons may ignore some promising line from further evaluation.

Advanced Selection of Top High Yielder and Top Less
Alternate Bearing. Te top 10 and 15 high yielders were selected from accessions using over fve and four harvesting seasons at Tepi and Gera, respectively, regardless of their biennial nature (Tables 8 and 9). Cofee accessions were less performed in yield at Tepi when compared with Gera. All accession yield performance at Tepi was less than checks. Te highest yielder accession at Tepi was T37/11 which had recorded 1838.07 kgha −1 . From the top 10 in yield performance, T21/11, T43/11, and T51/11 showed relatively low biennial bearing. Te check, Desu showed very low biennial bearing (I � 0.09) such as T33/11, T61/11, and T62/11 which were the top 10 in low alternate bearing at Tepi. Also, this check was a high yielder (recorded 2336 kgha −1 ) in addition to its low in biennial bearing. Tis material is prominent in the future cofee improvement for alternate bearing.

Conclusion
Variability was observed among cofee accessions in yield performance and biennial bearing in some harvesting seasons. Highly signifcant (p < 0.01) and signifcant (p < 0.05) variability was shown among the entire testing materials at Tepi and Gera, respectively, in the 2020 harvesting season. Among cofee accessions, a highly signifcant diference was indicated in biennial bearing at early stage at Tepi. Te over season pooled analysis of yield and alternate bearing intensity revealed the existence of moderate genetic variability among cofee accessions. Te moderate genetic coefcient of variance (GCV) (17.57 and 14.48%), heritability (H) (24.46 and 48.28%), and high genetic advance as percentage of the mean (GAM) (17.81 and 20.74) were recorded in yield at Tepi and Gera, respectively. For bienniality, moderate GCV (18.02 and 16.57%), H (37.22 and 47.75%), and high GAM (22.54 and 23.92%) were observed at Gera and Tepi, respectively.
Alternate bearing could afect response to selection which leads to less selection efciency. Early selection, using two and three harvesting seasons, excludes 30-40% and 33.3-20% high yielders from advanced selection. Selection at four harvesting seasons revealed 90% and more selection efciency which is the appropriate time for promising line selection.
Over year yield performance of cofee accessions at Tepi was less than checks. However, at Gera, most of the top 15 selected ones were high yielder than standard checks. T43/11 and T51/11 were the top high yielders and showed low biennial bearing at Gera and Tepi. T33/11, T49/11, T55/11, and T61/11 showed very low biennial bearing at both locations. Tus, these accessions need to be taken into consideration during the bienniality improvement breeding program. Generally, one has to be conscious of the biennial bearing nature of Arabica cofee during advanced selection.

Data Availability
Te data used to support the fndings of this study are available from the corresponding author upon reasonable request.

Conflicts of Interest
Te authors declare that they have no conficts of interest.