Principal Component Analysis for Yield and Yield Attributed Traits in Lowland Rice (Oryza sativa L.) Genotypes

Principal component analysis was utilized to determine the variation and to estimate the relative contribution of various characters for total variability. The experiment was laid out using randomized block design with three replications during 2017/2018 main cropping season at Fogera, Ethiopia. The first four principal components axes accounted for 85.3% cumulative variance of the total variability for seventeen agronomic characters. PC1, PC2, PC3 and PC4 explained 44.15%, 19.31%, 14.913% and 6.97% of variation from the total variation, respectively. Thus, maximum variation was found in first PC; therefore, selection for characters under PC1 would be desirable. The variability in PC1 was accounted by flag leaf length, panicle length, days to heading and days to 50% flowering, while PC 2 was accounted by harvest index. For future breeding program that employ hybridization, parental material selection should be carried out considering principal components influence to breeders’ interest.

protein (FAO, 2002). In Ethiopia, rice offers a variety of uses. It is used in the preparation of local foods (injera, dabbo, genffo, kinchie,shorba) and local beverages (tella and katikalla/Areki) either alone or mixed with other cereal grains (Heluf and Mulugeta, 2006). However, the average rice productivity in Ethiopia is estimated at 2.81 t ha -1 (CSA, 2017), which is much lower than the world's average of 4.6t ha -1 (FAO, 2015). Despite the fact that rice has been recognized by Ethiopian government as "the new millennium crop of Ethiopia" to attain food security, lack of improved varieties, lack of recommended crop management, lack of pre and postharvest management coupled with biotic and abiotic stresses limit the production and productivity of the crop in the country (Tesfaye et al., 2005;MoARD, 2010;EIAR/ FRG II, 2011). Among these problems, lack of improved varieties for different agro ecologies of the country is the most serious (MoARD, 2009;EIAR/ FRG II, 2011). In many countries, rice is a long established crop and cultivars have been selected that are well adapted to local conditions and the local market. It is estimated that about 120,000 varieties of rice exist in the world (Sassaki and Moore, 1997). But in Ethiopia which has diverse agro-ecologies, there are no more than elven lowland rice varieties in the whole country.
Farmers of South Gondar, especially those in Libokemikem, Fogera and Dera districts, largely produce lowland rice under rain-fed condition. Due to swampy nature of the study area, crop production was limited before rice adoption. Fogera and surrounding districts are swampy areas which are ideal for lowland rice cultivation. However, one of the major constraints in the area is the absence of high yielding improved lowland rice varieties resistant to diseases and to terminal water deficit (terminal moisture stress). Hence, as rice is a potential crop in study area, increasing its productivity per unit area and its total production will enable farmers get encouraging returns and improves its role in achieving food self-sufficiency. To increase the productivity of rice in the country, research has been conducted mainly at Fogera National Rice Research and Training Center (FNRRTC). The center introduced a bulk of genotypes from International Rice Research Institute (IRRI) and African Rice Center (WARDA), which are sources of variability for future rice improvement in Ethiopia.
The success of plant breeding research depends on the availability of genetic variation. However, full information is lacking on the genetic variability and character association of grain yield and yield related traits available within recently introduced low land rice genotypes in the study area. Genetic improvement mainly depends on the amount of genetic variability present in the population which is a universal property of all species in nature (Dutta and Burua, 2013). Variability in genotypes for yield and yield component traits forms the basic factor to be considered while making selection (Haydar et al., 2007). The character yield reflects the performance of all plant components and might be considered as the final result of many other traits. i.e. every plant contains an inherent physiological production capacity that operates on energy required for normal plant performance. Not all genotypes have the same inherent physiological capacity to yield (Welsh, 1981).
The knowledge of diversity and genetic distance among groups of genotypes helps to identify parental lines for hybridization programs. Therefore, keeping in view these urgent needs, the present investigation has been undertaken to know the level of genetic divergence among twenty seven rain-fed lowland rice genotypes with three check varieties for yield and yield contributing traits through principal component analysis.

MATERIALS AND METHODS Experimental Site Description
The experiment was conducted in the North-Western part of Ethiopia at Fogera National Rice Research and Training Center (FNRRTC) during the rainy season (June-December) of 2017/18. FNRRTC is located in Amhara Regional state, in the North-Western part of Ethiopia, 607 km far from Addis Ababa. The experimental site is found at Woreta and located 11 0 58' N latitude, 37° 41' E longitude and at an elevation of 1810m above sea level. Based on ten years' average meteorological data, the annual rainfall, and mean annual minimum, maximum and average air temperatures are 1300mm, 11.5°C, 27.9°C and 18.3°C, respectively. The soil type is black Vertisol with pH of 5.90 (Dejen, 2020). The main water source for rice production in the study area is rain-fall water. Irrigation water from rivers Rib and Gumara was also used in the off season for production of vegetables as the second crop after rice. www.iiste.org ISSN 2224-3216 (Paper) ISSN 2225-0948 (Online) Vol.11, No.16, 2021

Experimental Materials
Thirty genotypes consisting of 27 lowland rice genotypes introduced in 2015 from African Rice Center (Formerly called WARDA), two released lowland varieties (Ediget and Hibir) and one locally available genotype (X-Jigna) , obtained from FNRRTC, were used for this study (Table 1).

Experimental Design
The experiment was laid out in randomized complete block design with three replications. Each plot had six rows each 4m long, with a spacing of 25 cm between rows and 15cm between plants. The plot size was 4 x 1.5m = 6 m 2 . Net plot size was 4 rows x 4m =4m 2 . The distance between plots and replications was 0.3 m and 1 m, respectively. Three healthy and uniform sized seeds were drilled per hill on date 28 June 2017 and thinning was conducted after germination to ensure single plant per hill.
Fertilizer in the forms of N and P2O5 was applied at a rate of 69/23 Kg/ha, Urea and NPS, respectively. All the NPS was applied at sowing. Urea was applied as split three times, 1/3 at sowing, 1/3 at tillering and the remaining at panicle initiation stage. All other agronomic practices were applied as recommended for rice production in the study area. The first four principal components (PC's) accounted for 85.34% and the first and the second PC's accounted for 44.15% and 19.31% (total 63.46%) of the variance, respectively. Component loading of the first four principal components is given in Table 2.To aid visualization of the overall variability in the tested genotypes, the first two components scores (PC's) is plotted (Figure 2). Out of the total variation, PC1 and PC4 explained the largest (44.15%) and smallest (6.97%) variation, respectively, while PC2 and PC3 accounted for 19.31% and 14.913% of the total variation, respectively. In each PC indicated maximum variation was found in first PC, therefore, selection for characters under PC1 may be desirable.
Most yield and component characteristics contributed positively to PC1 except tillers per plant, panicles per plant, thousand-grain weight and harvest index. However, the highest variability was related with days to heading, days to 50% flowering, panicle length and flag leaf length. The second principal components (PC2) had positive loading for 10 traits out of the 17 studied traits, but the most contributed trait for the variation was dominated by harvest index. The third principal component (PC3) had positive loading for 11 traits out of the total studied traits, but dominated by panicles per plant, tillers per plant, grain yield and harvest index. The long vectors indicate that, they have a large contribution to the total variation (Yan and Kang, 2002).  Characters with lower absolute values, closer to zero, have less influence on clustering than largest absolute values within the principal components. Therefore, the above-mentioned characters, which load high positively or negatively, contributed more to the diversity and they were the ones that most differentiated the principal components. The graph of PC biplot classified the lowland rice genotypes into different group based on the most influential quantitative traits.
As a result clearly showed a breeder can easily pinpoint distances between the genotypes and make decisions based on the principal component simultaneously. In the bi plot, genotypes close to each other are similar while the ones found near the origin are distinct and the ones further out are extremes. PCA analysis classifies the genotypes into groups over the four quadrants based on the concentrations of these seventeen traits. The genotypes were distributed throughout the quadrant demonstrating large genetic variability in these traits.
The ones on the upper left quadrant were related in their harvest index and thousand grain weight while, the upper right quadrant contained genotypes related in their grain yield, panicle weight, plant height, filled grain per panicle, panicle length, flag leaf length and flag leaf width. The selected five rice genotypes (G8, G14, G26, G29 and G30) with grain yield above 6000 kg ha -1 were located in this quadrant, which clearly indicates the positive association of these traits for grain yield of rice genotypes (Fig3).
The right bottom quadrant contain genotypes related in their biomass yield, days to heading, days to flowering, days to maturity and unfilled grains per panicle plant whereas, the left bottom genotypes related in their number of tillers per plant and number of panicles per plant.
The distance between the locations of any two genotypes on the biplot is directly proportional to the degree of similarity/difference between them as per the traits considered (Shegro et al., 2013). The bi plot showed that genotypes G3, G20, G22 and G29 were the most divergent from the major group. These extreme genotypes are favorable for breeding programs due to their morphological and biochemical differences from the rest which makes unique. Genotypes which nearly overlapped in the principal component axes had similar quantitative characters. Thus, (G14 and G26) at cluster III; (G15 and G24) at cluster I in the upper right quadrant and (G10 and G12) at cluster II in the left bottom quadrant showed similar relationship in the principal component axis (Fig2).

SUMMARY and CONCLUSION
The present study was an attempt to know the principal components for yield and yield contributed traits in lowland rice genotypes for future utilization in the breeding program. To generate this information a total of 27 lowland rice genotypes with two standard checks and one locally available genotype were evaluated using randomized complete block design with three replications during the 2017/18 main cropping season at FNRRTC. The analysis of variance showed highly significant differences among the tested genotypes for all 17 studied traits, which indicates that there is a considerable genetic variability in the tested rice genotypes.
The maximum cluster distance was found between cluster two and three while the minimum was found between cluster one and three. The first four principal components explained 85.34% of the total variation. PC1, PC2, PC3 and PC4 accounted for 44.15%, 19.31%, 14.913% and 6.97% of the total variation, respectively. Based on the present investigation results, it can be concluded that there is adequate genetic variability for most of quantitative characters evaluated, that the genotypes with high grain yield should be selected from different clusters and crossed so as to improve grain yield. The study also identified the best performing genotype for further evaluation and/or recommended for release for possible commercialization.
For future breeding programs that employ hybridization, parental material selection should be carried out between clusters rather than within clusters. It is recommended to repeat the study at more seasons and locations with more number of genotypes to predict genotypic performance across seasons and locations which helps to validate the obtained current results. Moreover, the future rice research should be supplemented by molecular characterization to further confirm the outcome of current study findings.