Path analysis for agronomic traits and yield of salt tolerant rice cultivars under moderate salinity condition in Central Vietnam 

The experiment was conducted on the moderate saline soil (EC = 6.35 dS m^-1) with 10 salt-tolerant rice varieties in winter–spring season from January to April, 2017 in Thua Thien Hue province, Central Vietnam. Soil properties of the experimental site were shown in Table 1.
 

        
The experiment was arranged as a randomized complete block design (RCBD) with three replications. The area of   each plot was 10 m² (5 m × 2 m). The rice was planted in row of 10 cm × 20 cm spacing with one plant per cluster. Fertilizer was applied following the recommended dosage used by local farmers such as 200 kg lime ha^-1, 100 kg N ha^-1, 60 kg P₂O₅ ha^-1 and 60 kg K₂O ha^-1. Basic application was 100% P₂O₅ and 30% N. The remaining fertilizer was applied twice for top dressing. The first top dressing application was at 15 days after planting with 50% K₂O and 40% N and the second was at 40 days after planting with 50% K₂O and 30% N.
        
Thirty plants were randomly selected from each treatment to record agronomic traits. One week before harvesting, plant height was determined by measuring the distance from the soil surface to the end of the longest panicle. At harvest, these 30 plants were harvested, their roots removed, and then the number of panicles per plant was counted.
        
Thirty panicles from the thirty selected plants were measured their panicle length, and then bagged individually and dried until the grain’s water content reduced to 11-12%. After drying, number of total grains per panicle, number of filled grains per panicle and number of unfilled grains per panicle, weight of filled grains weight per panicle were measured. To determine the above ground dry biomass of each plant, shoots were oven-dried at 70°C for 1 week and then weighed. The fertilization rate (%) was calculated by dividing the total number of filled grains per panicle by the total number of grains per panicle (Zeng et al., 2002).
        
Data were analyzed using MS Excel 2007 and Statistics for Window 10 (Tallahassee, FL, USA). The average value, standard deviation (SD), standardize of the data, and the indirect coefficient (iC) were calculated using MS Excel 2010. To compare the differences between the agro-biological traits of rice varieties, one-way ANOVA and Tukey’s test at α= 0.1 were applied. Pearson’s correlation and the direct effect coefficient (dC) were processed by Statistix for Window 10. Residual effects coefficients (rC) were calculated based on the formula Res =          , in which R² is the coefficient of determination.

Correlation of agronomic traits and individual yield
 
Expression of agronomic traits and yield per plant of 10 experimental salt-tolerant rice cultivars were shown in Table 2. Due to the influence of salinity and low soil fertility of the experimental soil, the growth, development, and yield of the 10 selected salt-tolerant rice cultivars were low average values. Panicles per plant, panicle length, panicle weight, grains per panicle, filled grains per panicle, fertilization rate, 1000 grains weight, dry biomass and yield per plant were 8.4 panicles, 22.4 cm, 2.4 g, 113.2 grains, 92.7 grains, 82.0%, 26.5 g, 34.5% and 16.6 g, respectively.
 

        
The correlation coefficient (r) between the agronomic traits of the salt-tolerant rice varieties were obtained in Table 3. Individual yield had a significant positive correlation coefficient with the studied traits as did plant height (r = 0.3624), panicles per plant (r = 0.7019**), panicle weight (r = 0.4530*) and dry biomass per plant (r = 0.7837***). Plant height had a significant positive correlation coefficient with panicle length (r = 0.3664*), panicle weight (r = 0.4548*), number of grains per panicle (r = 0.4502*), number of filled grains per panicle (r = 0.4062*) and dry biomass per plant (r = 0.5057**). Thus, as the height of plant increases the above traits will increase. Similar to height of plant and yield, panicles per plant, panicle length, panicle weight, and number of grains per panicle had a significant positive correlation with dry biomass per plant with r = 0.5479**, 0.4198*, 0.4057* and 0.4207*, respectively.
 

       
Because the correlation between agronomic traits and yield was greatly influenced by environmental factors and research materials (varieties, fertilizers, etc.), the results of the correlation coefficients were not uniform among crops if the environmental conditions and varieties were different in usage (Ramakrishnan et al., 2006; Rasheed et al., 2002). Oad et al., (2002) concluded that individual yield positively correlated with the weight of 1000 grains, number of effective branches per plant and length of panicle. Khan et al., (2009) suggested that individual yield was positively related to panicle length and the number of grains per panicle. Akinwale et al., (2011) concluded that individual yield was positively correlated with the number of branches per plant, panicle weight, and grains per panicle. Sürek and Beser (2003) showed a positive correlation between individual yield with dry biomass per plant and the number of filled grains per panicle. Therefore, it can be concluded that studying the correlation between yield and its components, we should analyze and conclude for each detailed research and materials condition.
 
Direct and indirect effects of agronomic traits on individual yield
 
Table 4 showed that panicle weight had the highest positive direct coefficient on individual yield with dC = 0.8294, followed by panicles per plant (dC = 0.5524), number of grains per panicle (dC = 0.4355), fertilization rate (dC = 0.2561), and dry biomass per plant (dC = 0.2516). Although the plant height has a positive correlation with yield (r = 0.3624*), it has a very low direct coefficient (dC = 0.1330). This parameter had an indirect effect on yield through panicle weight (iC = 0.3772) and grains per panicle (iC = 0.1960). Thus, the direct influencing of plant height on individual yield of salt-tolerant rice cultivars in this area was very small (dC = 0.1330). The number of panicles per plant was highly positively correlated with yield (r = 0.7019***) and the coefficient of direct effect (dC = 0.5525) which were much larger than other indirect effect coefficients. This result indicated that the number of panicles per plant strongly and directly influences the yield of salt-tolerant rice cultivars. Panicle weight was highly positively correlated with yield (r = 0.4530*) and a high coefficient of direct effect (dC = 0.8294) whichs were much larger than other indirect effect coefficients. Similar to the number of panicles per plant, yield was strongly and directly governed by panicle weight.
 

        
In addition, panicle weight had a high indirect effect coefficient on yield via the number of grains per panicle  (iC= 0.3344). Although the number of grains per panicle insignificantly correlated for yield (r = 0.2085^ns), it had a high direct (dC = 0.4355) and high indirect effect coefficients on yield through trait of panicle weight (iC = 0.6369). Thus, the number of grains per panicle also played an important role in rice yield. Moreover, the dry biomass per plant had a high positive correlation coefficient with individual yield (r =0.7837***). However, it had a low direct effect coefficient (dC = 0.2516), indirectly affecting the yield through panicle weight (iC = 0.3365) and panicles per plant (iC = 0.3027). Thus, the dry biomass per plant did not directly influence the final rice yield.
       
Analysis results of r, dC, and iC in s 3 and Table 4 shown that the panicles per plant and panicle weight directly influence the final rice yield. Total dry biomass per plant and plant height do not directly affect yield, but indirectly affected via traits of panicle weight and grains per panicle. Similar to the correlation between yield and its components, the direct (dC) and indirect (iC) coefficients were also governed by the different factors of environment and research materials. For example, yield was influenced directly by the number of branches per plant and number of days of flowering (Amirthadevarathinam, 1983), panicle length (Arvind et al., 2011), number of panicles per plant, number of grains per panicle, weight of 1000 grains (Yang, 1986), number of filled grains per panicle and plant height (Bhadru et al., 2012), number of effective branches per plant, panicle length and flowering time (Ibrahim et al., 1990), plant height and branch numbers per plant (Kumar, 1992), number panicles per plant and number of spikelet per panicle (Lin and Wu, 1981), number of effective branches per plant, number of grains per panicle and weight of 1000 grains (Ram, 1992), number of grains per panicle and number of effective branches per plant (Sundaram and Palanisamy, 1994) and dry biomass per plant, harvesting index and weight of 1000 grains (Süreket_al1998).

The properties of experimental saline soil in Central Vietnam indicated that its fertility was low in general. This means that these kinds of saline soil in this area were considered not to be suitable for rice cultivation. The growth and yield of rice was low, greatly limited by salinity and other chemical and physical properties. In all the agronomic traits studied, individual yields had a significantly positive correlation coefficient (r) with the studied traits of plant height (0.3624*), panicles per plant (0.7019***), panicle weight (0.4530**) and dry biomass per plant (0.7837***) under moderate soil salinity (6.35 dS m^-1), soil pH from moderate to slightly acid (5.6 to 6.6) and low soil fertility conditions. Among studied agronomic traits, three traits such as panicles per plant, panicle weight and number of grains per panicle had the highest direct effect coefficients on rice yield with dC = 0.5524, 0.8294 and 0.4355, respectively. Therefore, these three agronomic traits should be used for selecting good salt-tolerant rice cultivars for the moderate salinity levels of soil in the central region of Vietnam. However, the relatively high residual effects (rC = 0.3703) imply that in addition to the above agronomic traits and yield were also significantly affected by other unstudied agronomic traits or other land and environmental factors. Thus, these factors should be more intensely studied.

The publication of this article is funded by Hue University.