Agronomic characterization of high-yielding irrigated barley accessions in the Cerrado

The objective of this work was to estimate the genetic, phenotypic, and environmental parameters of 113 barley accessions (Hordeum vulgare), previously selected based on high yield, in an irrigated production system in the Cerrado (Brazilian savanna), for use in breeding programs in Brazil. The experiment was conducted in 2013 at two sites in the Federal District, using a randomized complete block design with three replicates. The evaluated traits were: grain yield, kernel plumpness, thousand grain weight, plant height, lodging, and days to heading. Significant differences were observed between environment and accessions for all studied traits, as well as genotype x environment interaction. The MCU 3832 PI 402310, MCU 3484 PI 401962, CI 08053 Custer, MCU 3634 PI 402112, and MCU 3816 PI 402294 accessions stood out regarding grain yield and kernel plumpness. It is possible to obtain high selection gains with high heritability values. However, indirect selection is not recommended in the experimental conditions due to the low genotypic and phenotypic correlations obtained. It is also possible to select barley accessions with high yield and early heading date simultaneously, an important factor in the genotype selection process for future experiments. The accessions from Colombia are the most adapted to the Cerrado conditions.


Introduction
Barley is the fourth most grown cereal crop in the world, after corn, wheat, and rice. It is a plant with great adaptability, sown in several regions of the globe and at different latitudes. In Brazil, it is mainly sown in the South as a winter crop (Agostinetto et al., 2015). The country produces 317 thousand tons in 85 thousand hectares (Faostat, 2016), which only meets 43% of the demands of the Brazilian malt-production industry. For this reason, to supply the national beer industry, 350 thousand tons of barley are still imported

Materials and Methods
A total of 113 hulled beer barley (Hordeum vulgare L.) acessions of two-row an six-row were evaluated, and the six-row cultivar BRS 180 was used as the control (Tables 1 and 2). Theses accessions were previously selected for grain yield in a study performed by Monteiro (2012) at Embrapa Cerrados, agronomically characterizing 435 accessions of barley from the germplasm bank of Embrapa Recursos Genéticos e Biotecnologia. The experiments were conducted from May to September 2013 in two sites in Distrito Federal, Brazil, under center pivot irrigation system. Site 1 is located in the experimental field of Embrapa Cerrados (CPAC) (15º35'57"S, 47º42'38"W, at an altitude of 1,007 m), whereas site 2 is located at the experimental field of Embrapa Produtos e Mercado (SPM) (15°54'53"S, 48°02'14"W, at an altitude of 1,254 m). The soil of both sites is classified as a Latossolo Vermelho distrófico típico argiloso (Rhodic Hapludox). A randomized complete block design with three replicates was used. The plots consisted of six 5-m long rows, spaced 20 cm apart, with a useful area of 4.8 m 2 for each plot, with a density of 300 plants per square meter. According to the soil analysis, 16 kg ha -1 N, 120 kg ha -1 P 2 O 5 , and 64 kg ha -1 K 2 O were applied to the sowing groove, and 40 kg ha -1 N were applied by the time the fifth fully expanded leaf arose (Amabile et al., 2007).
Six traits were evaluated: estimated grain yield (kg ha -1 ); kernel plumpness (>2.5 mm) in percentage, according to Brasil (1996); thousand grain weight (TSW) in grams (Regras…, 2009); plant height in centimeters, as the average of three plants per plot; degree of plant lodging (LOD), with data transformed into arc sine X 0.5 ×100 -1 , in which X is the value in percentage of lodging; and days to heading (DH), which is the period, in days, from emergence to when 50% of the spikes inside the useful area of the plot were visible. The assessments of plant height, degrees of lodging, and heading were performed in the experimental fields of CPAC and SPM. The evaluations of grain yield, kernel plumpness, and TSW were carried out at the seed laboratory of Embrapa Cerrados.
The collected data were subjected to analysis of variance (Anova), and the averages were grouped by the Scott-Knott test, at 1% probability. Phenotypic variance (σ 2 f ), genotypic quadratic component (σ 2 g ), heritability (h 2 ), and the coefficients of environmental (EVC), genetic (GVC), and relative (CVr) variance were estimated for each trait, using the Genes software (Universidade Federal de Viçosa, Viçosa, MG, Brazil). For the Anova, individual and group analyses were also performed for each trait.
The phenotypic, genotypic, and environmental correlations were measured based on the estimates of  the respective variances and covariances between the traits, two by two, according to Kempthorne (1966), also using the Genes software (Universidade Federal de Viçosa, Viçosa, MG, Brazil).

Results and Discussion
The joint Anova showed the existence of significant differences among the accessions for all evaluated traits (Table 3), highlighting the variability among barley accessions, which is desired in pre-breeding works.
Significant effects of environment and genotype x environment interaction were also observed for the studied traits. This can be largely attributed to the difference in altitude between the experimental fields, of approximately 200 m, which directly affects temperature and favors environments of milder climates. This difference contributes to the increase in the genotype x environment interaction and, consequently, to the greater phenotypic variability of the quantitative traits, as reported by Kaczmarek et al. (1999).
According to Resende & Duarte (2007), regarding selective accuracy, in order to achieve values of 90% or more, considered ideal by Steel & Torrie (1980), F-values for the cultivar must be equal to or higher than 5.26 for a safe statistical interference. In the present study, regarding genotype x environment interaction, the F-value was only lower than the ones proposed by Resende & Duarte (2007) for lodging (4.7), showing high selection accuracy (Table 3).
Another genetic parameter that indicates experimental precision is the EVC; the lower this value is, the more precise is the experiment. However, the analysis of the EVC value must be considered according to the particularities of each trait and crop evaluated (Costa et al., 2002). In the present study, the values of the EVC varied from 1.39% for DH to 6.93% for LOD, which are considered low, indicating a good environment control (Table 3).
Genotypic variance is associated with the genetic differences among the specimens. Higher values for this component are an indication of higher genetic variability, which enables the identification of superior genotypes (Vencovsky, 1987). It was observed that the estimate of genotypic variance was the main component of phenotypic variance among the following traits: yield, kernel plumpness, TSW, height, LOD, and DH (Table 3).
The CVr is represented by the ratio between the GVC and the EVC. According to Vencovsky (1987), values above the unit for the CVr indicate good chances of gains with the application of selection among populations, suggesting that the trait can be handled with higher probability of genetic gains in breeding in experiments with two or three replicates.
The values of the GVC were higher than those of the EVC for all traits. The traits with higher CVr were: DH, yield, and LOD, with 6.23, 4.55, and 3.88%, respectively (Table 3). The value of the CVr for LOD contrasts with that obtained by Amabile et al. (2015) in irrigated conditions in the Cerrado, which can be explained by the fact that these authors studied a lower number of cultivars, which are considered elite and, therefore, probably presented a lower CVr value for LOD (0.54%); however, as in the present study (Table 3), Amabile et al. (2015) also found high values for DH (5.33%) and yield (2.72%).
The percentage of genetic variability that is transmitted from the parents to the descendants is obtained through h 2 (Lush, 1949). The high yields aspired by the breeder can be reached by the selection of traits that contribute to grain yield and that have both high h 2 and genetic gain. Consequently, the components of variance and h 2 with the genetic parameters are important for the definition of strategies aiming at the higher efficiency of the breeding program. The broadsense h 2 ranged from 99.57% for LOD to 97.91% for TSW (Table 3). The following values of high magnitude were also found for the same traits: above 84%, in 18 accessions gathered in Ethiopia (Addisu & Shumet, 2015); and from 62 to 84%, in F1 and F2 populations from ten Indian cultivars in a study performed during winter in India (Yadav et al., 2015).
The genotypic, phenotypic, and residual correlation coefficients between the evaluated traits in the present study were, in general, considered of low magnitude when compared with those found by Sayd (2014) Amabile et al. (2015), of 0.6 and 0.8, respectively. However, higher relationships, of 0.4, were observed between PK and TSW (Table 4), which is in alignment with Amabile et al. (2015), who reported a strong correlation of 0.6 between PK and TSW, which facilitates the evaluation process of the traits and also indirect selection. The genotypic and phenotypic correlation coefficients between height and LOD, in turn, were inexpressive, allowing the selection of tall plants with an adequate architecture that minimizes lodging, according to Ceccarelli et al. (1992).
In the comparison between averages for estimated grain yield at the CPAC site, the Colombian genotype MCU 3484 PI 401962 stood out positively with 6,508 kg ha -1 , similarly to the Brazilian control cultivar BRS 180, with 6,142 kg ha -1 (Table 1). At the SPM site (Table 2), the control cultivar BRS 180, with 7,358 kg ha -1 , was overcome statistically by the four following genotypes, presenting values considered of high magnitude: the Afghan cultivar H HOR 2325/58 PI 329126, with 9,187 kg ha -1 ; the Colombian genotypes MCU 3816 PI 402294 and MCU 3832 PI 402310, with 7,724 and 7,695 kg ha -1 , respectively; and the North American cultivar CI 08053 Custer, with 7,559 kg ha -1 (Table 2). At both sites, phenotypes from the American continent presented higher yield potential when compared with the genotypes from the other continents. Moreover, as expected, the six-row genotypes showed higher values for yield.
Values of 9,000 kg ha -1 were also achieved in experiments at Embrapa Cerrados with lineages from International Maize and Wheat Improvement Center (CIMMYT) introduced into the Cerrado, but with no malt quality . Sanchez et al. (2015) obtained lower values -5,848, 5,449, and 4,767 kg ha -1 , respectively, in the Cerrado -for the BRS Sampa, Manduri, and BRS 195 cultivars, recommended for irrigated areas. Values lower than the ones observed in the present study were also reported by Amabile et al. (2013), who found average of 5,908 kg ha -1 for the BRS Savanna cultivar, with values ranging from 4,726 to 8,659 kg ha -1 in trials performed in different locations and years in the Cerrado. However, there is still the possibility of grain yield gains using genotypes from the germplasm bank that are being introduced as cultivars and used in breeding blocks.
For kernel plumpness, the SPM site presented 51 genotypes with values above 80%, the lower limit adopted by the breeding program, whereas the CPAC site presented only 13 genotypes. At the CPAC site, the following genotypes stood out positively: six-row cultivars from Morocco -CI 09958 (87.6%) and CI 09959 (84.6%) -and from Iran -CI 09962 (91.6%) and CI 09961 (85.3%); besides two-row cultivars from Germany -CI 11493 Frugherste Stankas (82%) and Carina PI 371632 (84%) -and from the Czech Republic -CI 11367 Branisovicky (81%) ( Table 4). At the SPM site, 37 genotypes were similar to the control BRS 180 Table 3. Analysis of variance and genetic parameters for grain yield, kernel plumpness (PK), thousand seed weight (TSW), plant height, degree of plant lodging (LOD), and days to heading (DH), evaluated in 113 barley (Hordeum vulgare) accessions at two sites. (1) DF ( (Table 2). These results highlight the varying responses of the genotypes to the different evaluated environments. However, the Moroccan cultivar CI 09958 stood out at both sites, with values of 87.6% for CPAC and of 93% for SPM, showing high potential and higher stability for this trait in comparison with the other accessions. High values for kernel plumpness, above 80%, were also reported by: Amabile et al. (2013), who studied 39 elite two-row and six-row genotypes in irrigated conditions in the Cerrado; and Smith et al. (2013), who assessed 7 cultivars of high quality adapted the central north region of the United States, during six years, at four different sites.

Sources of variation
The average for TSW of the accessions at the CPAC site (40.2 g) was lower than the one at SPM (46.4 g), showing the difference between the two environments (Tables 1 and 2, respectively). However, the Iranian cultivar CI 09962 had high values at both sites: 57 g at CPAC, which was statistically similar to the weight of the Moroccan cultivar CI 09958, of 58 g; and 62.6 g at SPM, which does not differ statistically from the weights of the Ethiopian cultivars CI 12918 and CI 14041, of 60.3 and 58.6 g, respectively. These values show the positive correlation between the traits PK and TSW (Table 2), which is benefic to the breeding program. Among the accessions, only CI 12918 presented yield above 6,000 kg ha -1 in at least one site, emphasizing the low correlation between TSW and yield. Two other two-row cultivars stood out positively at both sites: CI 02109 White Smyrna from the United States and CI 09882 Gondar from England.
In the literature, Smith et al. (2013) observed TSW ranging from 56 to 59 g in seven beer cultivars recently released, values close to the ones of the accessions analyzed in the present study. Mirosavljević et al. (2015) reported an average of 40.8 g for 19 genotypes of two-row winter beer barley during three years of assessment in Serbia.
For plant height, the tallest accessions, all above 94 cm, were: CI 07255 from the United States and CI 06147 from Egypt (Table 1) in the CPAC site; and CI 10016 Raspa from Colombia and CI 07255 from the United States in the SPM site ( Table 2). The Ethiopian cultivar E 3/416 PI 356495 showed the lowest height at both environments, 57.6 cm at CPAC and 69 cm at SPM. The small size of this genotype makes it interesting to explore, since breeding programs have predicted, among their goals, the selection of short plants to avoid lodging (Amabile et al., 2015). The height of the accessions ranged from 57.6 cm for E 3/416 PI 356495 to 99 cm for CI 07255 at the CPAC site, and from 69 cm for E 3/416 PI 356495 to 103.6 cm for CI 07255 at the SPM site, with averages of 79.5 and 86.4 cm, respectively.
The Canadian genotypes presented heights above 84 cm at both sites and were characterized as tall. In contrast, the genotypes from South America showed great variation in height, varying from short, to medium, to tall. The Asian, European, and African genotypes showed smaller variations for this trait and were characterized as medium, except the CI 06147 cultivar from Egypt.
The behavior of the genotypes regarding lodging was considerably more intense at the SPM site than at CPAC. This behavior is due to the higher altitude of the first environment, which creates a more humid microclimate, favoring higher growth and, consequently, plant lodging. Of the 113 accessions evaluated, 14 did not lodge at the CPAC site and 8 did not lodge at the SPM site. Still, at the SPM site, 87 Table 4. Genotypic (r g ), phenotypic (r p ), and environmental (r e ) correlation coefficients between grain yield, kernel plumpness (PK), thousand seed weight (TSW), plant height, degree of plant lodging (LOD), and days to heading (DH) in 113 barley (Hordeum vulgare) accessions. The values for lodging obtained in the present study were higher than the averages of 28% reported by Mirosavljević et al. (2015), in trials performed with winter barley during three years, and of 25%, by Smith et al. (2013), for seven cultivars released in the United States. The Colombian genotypes MCU 3592 PI 402070, MCU 3576 PI 402054, and MCU 3452 PI 401930, as well as the control cultivar BRS 180, stood out positively and did not present lodging at both environments. No grouping tendency was observed when considering countries of origin and level of lodging.
Lodging, therefore, is a non-limiting trait, which can be solved with the use of growth reducers. However, the search for high-yielding genotypes with a more resistant stem allows the producer to reduce production costs and diminishes the risk of losses at harvest due to lodging.
Regarding DH, the most precocious genotypes were all six-row, originating from Colombia: MCU 3571 PI 402049, CI 13711 (both 49.3 days), and MCU 3469 PI 401947 (49.6 days) at the SPM site; and MCU 3484 PI 401962 (50.3 days) and MCU 3827 PI 402305 (51 days) at the CPAC site. The genotypes with higher yields presented heading cycles of around 50 to 60 days, similar to the ones obtained by the cultivar recommended for the Cerrado by Amabile et al. (2013). Values of around 55 days were also found for seven cultivars released in the United States in trials performed during six years by Smith et al. (2013).
The average of the genotypes was of 55 days at the SPM site and of 61 days at the CPAC site, with a difference of 6 days from one environment to the other (Tables 1 and 2). These averages are considered low when compared with the one obtained by Yadav et al. (2015) of 75 days for precocious Indian genotypes, in trials performed in India. It should be noted that more precocious genotypes are interesting for use in irrigated systems in Brazil, considering the advantage of the early release of areas for sowing of summer crops in the country. 2. High selection gains may be obtained with high hereditability values, but indirect selection is not recommended in the experimental conditions because of the low genotypic and phenotypic correlations observed.

Conclusions
3. It is possible to select barley accessions with high yield and precocious heading cycle simultaneously, a decisive factor in the process of genotype selection for future trials.
4. The accessions from Colombia are the most adapted to the Cerrado conditions.