Introduction

Compared to other cereals, sorghum is more resistant to environmental stressors and its production is more economic than other grains (^{Awika and Rooney, 2004}). It is the fifth most significant cereal in the world (^{Li et al, 2010}). Sorghum is commonly adapted to drought-prone regions with severe salinity and regions with relatively low production inputs (^{Li et al, 2010}). Therefore, it is widely used for feeding animals, especially over marginal lands (^{Barile et al, 2007}). The parts of the plant that remain after grain harvest are usually not attractive for animals, but parts of stay-green plant species can be used to make silage for easy consumption and digestion. Such silage can both constitute a feed source for animals and allow recycling of waste materials.

In vitro gas production techniques and chemical composition analyses are commonly used to identify potential nutritional values of feeds. These methods are fast, easy and cost-effective techniques (^{Kaplan et al., 2014}) and are also used to determine methane production. Methane is created through rumen fermentation and contributes significantly to global warming (^{Lin et al., 2013}).

GT biplot analysis allows visual assessment of data and is commonly used in economic, medical, genetic and agronomic studies. In GT biplot analysis, several attributes of the genotypes can be presented graphically, and thus, relations among several genotypes and attributes can be visually assessed and compared (^{Yan et al, 2001}). Therefore, researchers commonly prefer the GT biplot method due to ease of interpretation and assessment (^{Yan, 2014}).

The present study was conducted to investigate yield, chemical composition, fermentation, gas and methane production of stay-green sorghum genotypes for silage production to estimate metabolic energy and organic matter digestibility of these genotypes and ultimately to compare the genotypes with regard to investigated traits.

Material and methods

Among 274 local populations commonly used in sorghum breeding, 41 sorghum genotypes identified as stay-green after harvest and a standard cultivar (Sugargraze) were used as the plant material of the present study. Sorghum genotypes were sown in a lattice experimental design (6×7). As fertilizer, 180 kg ha⁻¹ N and 80 kg ha⁻¹ P₂O₅ were used. All the phosphorus and half of the nitrogen were applied at sowing, and the remaining half of the nitrogen was applied when the plants reached a height of 50 cm. Cultural practices were performed for weed, disease and pest control.

Results

Relationships between yield and chemical composition of sorghum genotypes

Biplot analysis was used to compare yield and chemical composition of sorghum genotypes and to identify good-genotype groups with regard to genotypes or chemical characteristics. By using reciprocal relationships among the investigated chemical traits, separation power can be obtained from the vector image of GT biplot chemical composition. Traits with long vectors have a high genotype separation capacity. In this case, green herbage yield, ADF, NDF, ME and methane gas production are identified as significant traits for separation of genotypes (Figure 1). Biplot vector images provide information about the relationships among investigated traits. Significant negative relationships were observed among gas production, ADF and NDF, as well as among herbage yield, crude protein, organic acids, pH and crude ash. Significant positive relationships were observed among ADF, NDF and dry matter, as well as among methane, metabolic energy, organic matter digestibility and gas production and among acetic, butyric, propionic acids, pH, crude ash and crude protein (Figure 1).

Figure 1 GT biplot based on yield and chemical composition-focused scaling for yield and chemical composition. 

GT biplot polygons also indicate which genotypes are prominent with which traits (GTI: genotype trait interaction) (Figure 2). In this case, G22 OMD, G23, G3, G24 and G14 genotypes were prominent with lactic acid; G32, G20, G15, G4, G16, G38, G29, G18 and G1 genotypes with crude protein, methane %, pH, butyric acid, acetic acid and crude ash; G11, G30, G41, G26, G33, G36, G9, G40 and G39 genotypes with ADF, NDF and dry matter; and G7 and G6 genotypes with green herbage (Figure 2; Table 1).

Figure 2 Polygon views of the GT biplot based on symmetrical scaling for the which-won-what pattern for the genotypes and yield and chemical composition. Details of the genotypes are presented in Tables 1. 

With GT biplot polygons of investigated traits, it is possible to identify the ideal genotypes. Considering all the traits, it was observed that genotype G20 was the most prominent, followed by G32, G15, G18, G16, G4, G28, G38 and G23. The Sugargraze cultivar, which was used as a standard cultivar, was far behind the genotypes in comparison (Figure 3).

Figure 3 GT biplot based on genotype-focused scaling for comparison of the genotypes with the ideal genotype. 

Discussion

^{Bolsen et al. (1996)} reported decreased NDF and ADF content of silages because carbohydrate sources increased the number of some anaerobic bacteria such as lactic acid bacteria in silage ambient and accelerated NDF, ADF and crude cellulose degradation of silages. Such a case indicated the availability of stay-green plant parts for silage.

^{Ball et al. (1996)} indicated that variations in dry matter and protein contents mostly resulted from genetic differences and that these parameters commonly varied based on leaf, cob and shoot ratios, ripening period, temperature and fertilization practices. Increasing ADF and NDF ratios are observed with the progress of the ripening period (^{Kaplan et al., 2014}). Increasing ADF and NDF ratios complicate the digestion and reduce crude protein content, gas production, metabolic energy and digestible organic matter quantities (^{Kaplan et al., 2014}).

Low pH levels indicate the existence of an acidified ambient with the fermentation of soluble sugars (^{Islam et al., 2012}). Dissolved carbohydrates (i.e., grain) increase lactic acid production, and increased lactic acid quantities reduce the pH levels of the ambient (^{Bates, 2009}). High pH levels of the present silages (the desired level is between pH 3.86-4.02) indicate low lactic acid accumulation in this study (^{McDonald et al., 1991}). With the progress of harvest, increased ADF and NDF ratios decrease crude protein and the WSC ratio of the silages (^{Hargreaves et al., 2009}) and negatively affect pH levels.

With decreasing grain ratio (i.e., starch), energy also decreases (^{Mould et al., 1983}), and silage quality is impaired (^{Cox et al, 1993}). The more fermentable carbohydrates are present, the more gas will be produced (^{Blümmel and Orskov, 1993}). For metabolic energy calculations, the principles specified in ^{Menke et al. (1979)} were used in this study. Gas production and crude protein ratio are used in this calculation. Cutting panicles of sorghum will reduce the carbohydrate ratio and gas production and consequently the metabolic energy.

Dry matter content of sorghum is desired to be 24.60% at minimum for quality silage (^{Carmi et al., 2006}). In the present study, a few genotypes had a dry matter ratio lower than this specified value.

^{Islam et al. (2012)} reported increasing crude protein contents with increasing water-soluble carbohydrate (WSC) contents. The majority of WSC is fermented by silage microorganisms for lactic acid, ethanol and VFA. Lactic acid is the very last main product, reducing pH levels for better preservation of silage (^{Miron et al., 2006}). There was a linear relationship among crude protein, lactic acid and pH of the present study.

^{Lopez et al. (2010)} classified the anti-methanogenic potential of the feeds based on gas production through fermentation as low (>11% and ≤14%), medium (>6% and <11%) and high (>0% and<6%). Based on this classification, the anti-methanogenic potential of stay-green sorghum genotypes was classified as low (14.15-21.89% CH₄).

Although biplot analysis is recommended to test a certain trait under different environmental conditions, it can be used just as effectively for all genotypes through user inputs of dual data such as genotype characteristics (^{Yan and Kang, 2003}), and genotypes can be screened for desired characteristics (^{Yan and Tinker, 2006}). In this study, stay-green sorghum genotypes were screened through yield and nutritional characteristics. With GT biplot analysis, it is possible to identify the best genotype with regard to these investigated traits (^{Yan and Kang, 2003}). The ideal genotype is either the one within the central circle or the closest one to the central square (^{Kaya et al., 2006}). In present study, the genotype G20 was the closest to the central circle and was consequently identified as the ideal genotype.

Biplot analysis can be used to visualize the interrelationships among the traits because the cosine of the angle between the vectors of any two traits approximates the correlation coefficient between them (^{Yan and Kang, 2003}). Based on this application, ADF and NDF were positively correlated with ME and OMD (^{Kaplan et al, 2014}), and crude protein was positively correlated with pH (^{Islam et al, 2012}).

The polygon is commonly formed by connecting the markers of the genotypes that are farther away from the biplot origin to gather all other genotypes in the polygon (^{Kaya et al, 2006}).

The main conclusions are as follows. Although some silage samples had low silage and nutritional characteristics, others yielded values close to or even higher than full sorghum silage. The genotypes with high green herbage yield, crude protein content, ME, OMD and lactic acid content, as well as low ADF, NDF, gas production and pH, can be used for silage both after grain production and grain harvest. The other sorghum lines with lower quality values than full sorghum silage can be used in ruminant nutrition during periods of deficit for forage supplies. Further breeding research on stay-green genotypes may provide significant contributions to plant and livestock production activities.