Assessment of Genetic Diversity in Various Yield Traits to Determine a High Yielding New Type of Upland Rice

— This experiment's objective was to estimate the genetic diversity of F5 populations on various yield components for developing high-yielding upland rice. The experiment was carried out from April to August 2018 at Semarang Village, Bengkulu, Indonesia. Plant materials used in the experiment were 160 accessions of F4, obtained from a pedigree selection of crossing between two local landraces (Sriwijaya and Bugis) and two introduced accessions (IR-148 and IR-7858-1). The experiment was arranged in Augmented Design with a spaced planting system (20 x 20 cm). The variance between populations was determined by Principal Component Analysis (PCA) with XLSTAT version 9.0. Broad sense genetic diversity was found for the number of traits, such as number of panicles, the total number of grains per panicle, number of empty grains per panicle, number of filled grain per panicle, percentage of empty grain, the weight of 1000 grains, the weight of grain per hill, which potentially improved high yielding. These genotypes were categorized into three groups. Group I showed superior traits for the total number of grains per panicle, the number of filled grain per panicle, weight of 1000 grains, and weight of grain per hill found in genotypes BKL4-B-2, BKL3-B-3, BKL4-B-3, BKL3-B-2. Group II had superior traits for panicle length, the number of empty grains per panicle, and the percentage of empty grain were BKL1-RS*1-3, BKL1-RS*1-1, BKL3-RS*1-1, BKL3-RS*1-3, BKL1-RS*1-2, BKL2-RS*1-2, and BKL2-RS*1-1. Groups III was superior to the number of panicle traits found in genotypes BKL2-B-2, BKL3-B-1, BKL1-B-1, BKL1-B-2, and BKL2-B-1.


I. INTRODUCTION
The contribution of upland rice to the national rice production in Indonesia has been remarkably insignificant. Its productivity is only 2.5 ton.ha -1 , much lower than that of lowland rice, reaching 5.15 ton.ha -1 [1]. However, as the land for growing lowland rice has become limiting, increasing national rice production must be done by growing upland rice at the upland soil, because of which improves the performance of upland rice has always been of paramount importance.
Many factors contribute to the low productivity of upland rice, such as an unfavorable environment, improper crop management, and lack of high-yielding seeds [2], [3]. One way to cope with those problems is by developing a highyielding new type of upland rice adaptive to such environmental conditions. Assembling a high-yielding and drought-tolerant rice variety may be done by combining all the excellent traits of two parental lines, like high yield and tolerance to drought. The parental lines come from local landraces or introduced accessions, having desirable-superior traits [4], [5]. An ideal new type of high-yielding rice paddy is expected to increase the potential yield, which is in line with the Indonesian government's program so-called food selfsufficient program.
Assembling a new variety of upland rice may start by crossing two or more parental lines to incorporate the expected superior traits within a new population. The population is then selected, evaluated, tested for its environmental adaptability to get a promising line ready to be released as a new high-yielding variety. Selection is an effective method to obtain the essential traits, having a high chance of succeeding. If a specific trait has high genetic diversity, it must have a high intergroup diversity so that a selection process will be easier to get the expected traits. Thus, it is essential to have information on the genetic diversity of a population to get an expected new variety. The trait variability of plants significantly determines the yield potency and improves the efficiency of using the genetic materials in the breeding program for increasing yield [6], [7], [8].
Trait superiority must be prioritized during the new variety program's assembly so that each promising variety owns a specific trait different from the existing variety. A specific trait identity must be identified thoroughly to prevent duplication and ensure the newly released variety's identity. A selection process will be effective, provided that the targeted traits have a high value of heritability. Heritability is essential to determine the choice of a selection method and how generation selection of the expected trait should be made [9].
Genetic advance reflects how effective the selection process was taking place. A selection process will be effective provided that the value of genetic advance is high and supported by a high value of heritability. Furthermore, heritability value also determines the selection progress, in which the higher the heritability values, the higher the selection progress will be, and the faster a high-yielding variety will be obtained. Genetic diversity plays an essential aspect in which breeders always work [10], [8], [11], [12].
Principal component analysis might be used to determine trait diversity and to identify the general trait, while biplot visualization can determine the specific trait of a particular genotype ( [13], [14]). This experiment's objective was to identify the agronomic characters and estimate the genetic diversity of F5 populations on various yield component varieties and trait quality for developing high-yielding new types of upland rice.

II. MATERIALS AND METHODS
The experiment was carried out from April to August 2018 at Semarang Village, Bengkulu City, Bengkulu Province, Indonesia. Plant materials used in the experiment were 160 accessions of F4 Generations, obtained from a pedigree selection of crossing between two local landraces (Sriwijaya and Bugis) and two introduced accessions (IR-148 and IR-7858-1).
The experiment was arranged in Augmented Design with a spaced planting system (20 cm x 20 cm). Each accession was planted in six rows with one seed per hole using a head-torow system (about 800 population). At ten days after planting (DAP) the crops were fertilized with 150 kg. ha -1 Urea, 100 kg. ha -1 SP36, and 100 kg. ha -1 KCl. At 30 DAP the crops were fertilized with 100 kg. ha -1 Urea, 100 kg. ha -1 SP36, and 100 kg. ha -1 KCl. Pests, diseases, and weeds were managed intensively. Variables measured included length of panicle, number of panicles per hill, number of total grains per panicle, number of filled grain per panicle, percentage of empty grain per panicle, the weight of 1000 grains, and weight of filled grain per hill.

A. Data Analysis
Data were analyzed with Microsoft Excel Statistical Program combined with Minitab 15. Principal Component Analysis (PCA) determined the variance between populations, correlation matrix, and biplot with XLSTAT version 9.0.
The following Formula could determine the value of genetic gain (GG): GG = S.h2bs GG (%) = Gx100% X where S = selection differential, G= selection advance, G (%) = genetic gain in percentage, h2bs = broad sense heritability, X = average of initial population. Selection was done for selected individual from single trait at 10% of intensity.

A. Genetic Diversity of F5 Population
The estimated value of plant genetic trait showed that the coefficient of genetic diversity (CGD) ranged between 7.63% and 62.46%. The lowest CGD value (7.63%) was found in the panicle length trait, while the highest value was found in the percentage of emptied grain per panicle (62.46%). The absolute value of CGD (0 -62.46%) was then used to decide the relative value of CGD, in which 62.46% was assumed to be the 100% relative value. Therefore, the criteria for absolute value transformed to low (0.0% < x < 15.62%), medium-low (15.62% < x < 31.23%), medium-high (31.23% < x < 46.85%), and high (46.85% < x < 62.46%).
Traits having a low and medium-low of CGD values were categorized as traits with narrow genetic diversity. In contrast, traits having medium-high and high CGD values were categorized as traits with broad variability ( [17], [18]). Based on those criteria, we found that the panicle's length is a trait with narrow genetic variability while the number of panicles, total grain per panicle, and empty number grain per panicle as traits with medium broad genetic variability.
Furthermore, the number of filled grain per panicle, percentage of empty grain per panicle, weight of 1000 grains, and weight of grain per hill were categorized as traits with broad genetic variability. In other words, there was one trait that has low genetic variability and seven traits having broad genetic variability, having the chance to improve the genetic performance of the rice through the following traits: number of panicles, number of total grains per panicle, number of empty grains per panicle, number of filled grain per panicle, percentage of empty grain, the weight of 1000 grain, and weight of grain per hill. A broad genetic variability means that the selection processes of the trait run effectively and be able to improve the genetic traits for the following generation ( [19], [8]). The selection process could be done in a more convenient way for traits with broad genetic variability that can be used for crop improvement. The estimated value of trait heritability was 0.78 for panicle length and 1.0 for the weight of 1000 grain (Table 1.). Based on [15] criteria, the estimated heritability values of all traits were high. However, the high genetic advance was found only for the total number of grains per panicle, the number of filled grain per panicle, weight of 1000 grain, weight of grain per hill. Traits that have high heritability determine the effectivity of the selection process and speed up the advance of crop improvement. High heritability values suggested that genetic factors are more dominant than the environmental factors, and that selection process could be done in the early generation [20], [21], [7].

B. Lines Performance of F5 Population
The panicle length produced in this experiment ranged from 24.22 to 31.04 cm, with 27.91 cm on average (Table 2.). The lines obtained from the hybridization produced lines with the shortest panicle (24.22 cm) and the longest panicle (31.04 cm). Some lines, however, showed shorter panicle than their parental lines. The length of the panicle is strongly correlated to the number of grains per panicle. The number of panicles ranged from 3.00 to 27.00, with an average of 12.46. To improve the potential yield of 10% higher than New Plant Type (NPT), some traits are needed, such as 330 panicles per m 2 and 150 grain per panicle, 22 ton per hectare of biomass (with 14% of moisture content), and 50% harvesting index [22], [23]. A strategy has been developed for getting a new type of rice with specific panicle traits and 150 grains per panicle [23]. The number of panicles per m2, percentage of filled grain, total biomass, and harvesting index are needed for developing the new type of rice [24].
The lines obtained from the crossing demonstrated the number of grains per panicle up to 141.00-307.42, with 214.35 on average, the highest number of the filled grain of 268.75 with 166.43 in average, and the low percentage of empty grain (22.01%), as shown in Table 2. Poor grain filling is caused by the low level of apical dominance at the panicle, the poor arrangement of grain on the panicle, and the limiting activity of phloem in transporting assimilates [24]. Furthermore, the inefficient partition of assimilates causes the new rice paddy's poor grain filling [25].
The NPT rice in Indonesia must have the reasonable number of the tiller (12-18 tiller) [26]. However, all of them must be productive, with the number of grains per panicle up to 150-250, percentage of the filled grain of 85-95%, the weight of 1000 grain about 25-26 g, short but vigorous stem (80-90 cm), and early season (110-120 days). With those traits, the newly bred rice paddy is expected to yield 9-13 tons per hectare. The population of the F5 line demonstrated a high number of grain (307.42, 214.35 on average), a low percentage of empty grain (22.01%), and a high percentage of fille grain (268.67, 166.43 on average) per panicle. In short, these lines were the potential for producing high yield because they demonstrated the high number of panicles per hill, 12.46 on average (Table 2.). The previous study recommended avoiding an extreme trait when obtaining newly breed rice paddy, such as a high number of grain (200-250) per panicle leading to the poor grain filling, because of which IRRI set a standard for 150 grain per panicle to ensure a high percentage of filled grains [24]. Besides, the weight of 1000 seeds of the new lines produced in this experiment was about 26.84-55.52 g, with an average of 40.74 g. Ideally, a new variety must weigh 26.84-55.52 g for 1000 seeds [27]. Therefore, all newly bred lines resulting in this experiment were considered to have a high potential yield based on this standard. To increase the yield of newly bred rice, one needs to have parental lines having long panicles and a high number of grain per panicle ( [28], [24]). Also, recommended by [27] that an ideal new variety has 180-240 grain per panicle, and 85% of the grain is fully filled.

C. Correlation and Analysis of Principal Component between Yield Component and Yield
The correlation values could determine the relationship between the yield component and yield. The correlation between characters was presented in Table 3. Panicle length was negatively correlated to and significantly different from the number of panicles (-0.59) and weight of grain per hill (-0.49), and positively correlated with and significantly different from the number of empty grains per panicle (0.64) and percentage of emptied grain (0.57) Number of panicles, number of filled grain, and weight of 1000 grains positively correlated to and significantly different from grain per hill weight. Consequently, those three traits could be used as selection criteria for increasing grain yield. The weight of 1000 grain is correlated to the size and degree of fullness of the grain. The high percentage of filled grain and the enormous size of grain lead to the high weight of 1000 grain ( Table 3). The increase of the number of filled grain per panicle leads to the increase of the number of grains per hill significantly.
In contrast, increasing the number of empty grains per panicle reduces grain yield per hill. The high number of filled grain per panicle, combined with a high percentage of filled grain per panicle, will significantly increase grain yield per hill. The grain filling in compact-panicle rice becomes poor subject to an expression of the recessive allele for high ethylene production, but the allele is amenable for suppression by the corresponding dominant allele [29].
The results of principal trait analysis showed that some traits significantly affected the variability of the tested lines. The results significantly reduced total morphological traits evaluated to three main components (KU-1, KU-2, and KU-3), which represented the eigenvalue of >1 and the variability of line performance up to 88.15% (Fig 1). Moreover, the trait with a higher coefficient value of the main component contributes to the main component [30].
Eigenvalue contributed significantly to total diversity. The first principal component (KU-1) with an eigenvalue of 3.96 contributed to 49.47% of total diversity. The second principal component (KU-2) with an eigenvalue of 2.33 contributed to 78.53% of total diversity among the tested lines, and the third component (KU-3) with an eigenvalue of 0.77 contributed to 88.15% of the total diversity of the lines tested.

D. Morphological Characters Main Component
Analysis of the identifier vector shows which trait contributes maximally to the diversity of the tested lines (Table 4), as the trait with the highest point of identifier vector and positive value. For KU-1, the traits contributing to the diversity of tested lines included length of panicle, number of filled grain per panicle, number of empty grains per panicle, percentage of empty grain, the weight of 1000 grain, and weight of grain per hill. Besides, it was found in the number of total grains per panicle for KU-2 and the number of panicles for KU-3. The contribution of diversity explaining the indicators used to observe the relationship between yield potency and yield was 100%. However, by conducting biplot analysis, which reduced all the indicators into two-side dimensions, the information that could be explained was only about 78.53%, meaning that the information that could be explained by biplot analysis was >70%. It meant that biplot analysis has already represented enough information on the relationship among those eight indicators.   Table 3.
Analysis of the main coordinate was carried out to show each tested line's relative position, as presented in Fig 2. The results demonstrated that the following traits, the total number of grains per panicle, filled grain per panicle, the weight of 1000 grain, weight of grain per hill, had a greater chance to yield a higher population average than the other traits. To improve the yield of new type rice, it needed parental lines with superior grain per panicle length and length of the panicle [24], [28]. One of the ideal traits for new rice is that the number of grain must be 180-240, with the percentage of filled grain about 85% [27]. A number of filled grain were used as the indicator trait to develop a new type of rice [23], [31]. Besides, an ideal new type of rice must have 150 number of filled grain [23] or 160 number of grain [31]. From this discussion, it may be suggested that the traits of the total number of grain per panicle, filled grain per panicle, the weight of 1000 grain, weight of grain per hill have to be considered while selection for high yield as they expressed positive and significant correlation with grain yield. A positive inter-correlation was also noticed between these traits. Hence, a balance should be maintained while selecting for these traits. It will bring up improvements in the yielding potential and also the traits themselves.

IV. CONCLUSION
Broad sense genetic diversity was found for the number of traits, such as number of panicles, the total number of grain per panicle, number of empty grain per panicle, number of filled grain per panicle, percentage of empty grain, the weight of 1000 grains, the weight of grain per hill, which potentially improved high yielding. These genotypes were categorized into three groups, each of which had its own characteristics. Group I showed superior traits for the total number of grain per panicle, the number of filled grain per panicle, weight of 1000 grains, and weight of grain per hill found in genotypes BKL4-B-2, BKL3-B-3, BKL4-B-3, BKL3-B-2. Group II had superior traits for panicle length, the number of empty grain per panicle, and the percentage of empty grain were BKL1-RS*1-3, BKL1-RS*1-1, BKL3-RS*1-1, BKL3-RS*1-3, BKL1-RS*1-2, BKL2-RS*1-2, and BKL2-RS*1-1. Groups III was superior to the number of panicle traits found in genotypes BKL2-B-2, BKL3-B-1, BKL1-B-1, BKL1-B-2, and BKL2-B-1.