Evaluation of yield-attributing parameters in Aus rice for enhancing productivity

Abstract

This research aimed to assess the agronomic performance of the progeny (F₃ and F₄ generations) of 48 newly developed Aus rice lines, using a randomized-complete-block-design under rainfed conditions. We found a wide range of variations in yield and yield-contributing traits among the studied genotypes. High board sense heritability percentages were found for sterility percentage (99.50 and 97.20), thousand-grain-weight (88.10 and 90.20 g), plant-height (84.90 and 86.90 cm) and day-to-maturity (84.50 and 97.60 d) in both F₃ and F₄ generations, respectively. However, the highest genetic advance as mean percentage was observed for sterility (48.00 and 50.60), effective tillers number per hill (ET) (44.70 and 47.10), total tillers number per hill (TT) (43.00 and 45.40) and filled-grains per panicle (41.00 and 43.20) respectively. Notably, the correlation study also identified the traits, TT (r = 0.31 and 0.45), ET (r = 0.30 and 0.44), straw yield (r = 0.57 and 0.39) and harvest index (r = 0.63 and 0.67) as effective for improving grain yield in both F₃ and F₄ generations, respectively. We identified higher grain yield per hill (g) and shorter to moderate crop growth duration (days) in several distinct accessions, including R1-49-7-1-1, R3-26-4-3-1, R1-6-2-3-1, R1-13-1-1-1, R1-50-1-1-1, R3-49-4-3-1, R1-47-7-3-1, R2-26-6-2-2, R3-30-1-2-1 and R1-44-1-2-1, among the 48 genotypes in both the F₃ and F₄ generations. A further location-specific agronomic study is recommended to assess the drought tolerance of these promising genotypes. This will further assess their suitability as potential breeding materials when developing rice varieties adapted to grow under fluctuating rainfalls conditions.

Introduction

Rice (Oryza sativa L.) stands as the predominant staple food crop, providing sustenance for over half of the global population. It serves as a major source of calories and nutrition for peri-urban populations in both Asia and Africa (Dewan et al., 2023). Rice holds a predominant position in Bangladesh, thriving in the riverine agro-based climate that is conducive to its cultivation and is consumed routinely by approximately 160 million (M) people (Khushi et al., 2020). Being a primary staple food grain in Bangladesh, rice is cultivated on nearly 11.5 M hectares (ha) of land, accounting for approximately 71% of the total cropped area (BBS, 2018). The total rice production is dominated by Aus, Aman and Boro rice growing seasons in Bangladesh. The respective area coverage for each Aus, Aman and Boro rice is 1.14, 5.88 and 4.75 M ha, resulting in the production of 2.76, 14.00 and 19.56 M tonnes (t) with an average grain yield of 2.51, 2.55 and 4.08 t/ha, respectively, in the fiscal year 2018–19 (BBS, 2021). However, with the global population increasing at a rate of 2 M people/ year, there is a need to increase rice production to meet the growing demand. This challenge is further exacerbated by the adversities of climate change, such as: drought, high and low temperatures, heavy precipitation at a time, floods and salinity intrusion, which pose a significant stress on crop production (Ghosh et al., 2021a, 2021b, 2022; Kupdhoni et al., 2023).

Rice is one of the most susceptible crops to reduced yield from water deficits due to its several morpho-physiological and biochemical mechanisms (Ji et al., 2012). Among the three types of rice in Bangladesh, Aman rice is an irrigation-independent rainfed monsoon crop, while Boro rice is completely irrigated. However, Aus rice is a rainfed or limited irrigated crop among the rice-growing seasons in Bangladesh (Mainuddin and Kirby, 2015). Aus rice is grown in the very hot summer season and the yield of Aus rice is low compared to Aman and Boro rice (BRRI, 2011). Increasing the yield of Aus rice not only contributes to the total productivity of rice, but also reduces the use of groundwater (UNB, 2016). Hence, Aus rice could be a substitute for groundwater-dependent Boro rice cultivation (BRRI, 2011). Besides, the government of Bangladesh has launched an incentive programme for small and marginal farmers in an attempt to rejuvenate Aus rice cultivation (BRRI, 2011). But the available varieties of Aus rice cannot meet the farmer's demand. So, there is a necessity to develop a high-yielding Aus rice variety with a short life span to meet the increasing demand for rice by leveraging the advantage of its lower irrigation requirement and reducing the dependency on groundwater for irrigation in rainfed ecosystems (Hossain et al., 2024).

Genetic traits are primarily controlled by polygenes, and their expression is significantly influenced by environmental factors (Islam et al., 2016). To unveil the genetic control of any attribute, it is crucial to assess variability components, including phenotypic and genotypic variations, along with the determination of broad-sense heritability and genetic advance. These key parameters play a vital role in understanding the gene action that governs the desired characteristics (Bekele et al., 2013; Khan et al., 2023). To comprehend the genetic variability of yield-contributing traits using a multivariate approach, it is essential to explore the relationships among these traits and their contribution to the grain yield of rice. Many researchers have effectively classified and estimated diversity in various rice varieties using agro-morphological features through multivariate analyses (Ravikumar et al., 2015). The variation in genetic divergence in agronomic traits is crucial for identifying desirable genetic material with anticipated characteristics in breeding programmes and thereby expanding the gene pool (Ahamed et al., 2021). The genetically heritable portion, known as heritability, assists plant breeders in making informed selections and evaluating the extent of genetic improvement (Tuhina-Khatun et al., 2015). Heritability, combined with genetic advances, can more accurately predict genetic gain under selection than heritability alone (Shrestha et al., 2021a). Cluster analysis using Euclidean distance is a useful statistical method for assessing the genetic diversity of germplasm concerning traits collectively (Shrestha et al., 2021b). Moreover, exploring the correlation between yield and yield-attributing characteristics facilitates improvements in yield by revealing the extent of the relationship between yield and its contributing components (Aditya and Bhartiya, 2013). So, enhancing the productivity of Aus rice requires a thorough understanding of the yield-attributing traits, such as grain yield, days to maturity and other agronomic traits. By identifying and evaluating the genotypes which possess significant variations and superiority of yield-attributing parameters compared to others, we can enhance the overall productivity of Aus rice. Therefore, these investigations were conducted to assess the extent of variability into the agronomic traits of newly developed 48 Aus rice accessions in F₃ and F₄ generations and to identify promising genotypes for future breeding under rainfed conditions.

Materials and methods

Plant materials and experimental sites

Two experiments were carried out in the research field of the Agronomy Department of Bangabandhu Sheikh Mujibur Rahman Agricultural University over a 2-year period, between 2019 and 2020, during the Aus rice growing seasons (March–August). The experimental sites were situated at 24°05′N latitude and 90°16′E longitude, positioned in the middle of the Madhupur tract, within agro-ecological zone 28. The research site experiences a subtropical climate marked by three distinct seasons, namely summer, wet and winter, characterized by warm to hot temperatures, heavy rainfall and increased humidity during summer, and cooler temperatures compared to summer during winter. The metrological data during the crop growing period are presented in online Supplementary Table S1. The soil is characterized by heavy clays and is silt-loam in texture. It is acidic in nature, with suboptimal physico-chemical properties. A recommended dose of fertilizer was applied in accordance with the recommended fertilizer guidelines for both experiments to ensure normal crop growth under field conditions (Guide, 2015; Sultana et al., 2015). The 48 Aus rice accessions originated from nine distinct parents, namely Dhalasaitta, Laksmilota, Kataktara, Narica-ABSS, BRRI dhan43, BRRI dhan55, BR7, Nipponbare and Parija, through hybridization; carried out by the Department of Agronomy and Genetics & Plant Breeding at BSMRAU. The newly developed 48 Aus rice accessions were used in this experiment for evaluation based on agronomic traits in both F₃ (online Supplementary Table S2) and F₄ generations (online Supplementary Table S3).

Experimental design and agronomic management practices

The experiments were conducted following randomized-complete-block-design comprising three replications. Each replication in both experiments consisted of 25 individual plants for each accession. Two rice seeds were manually sown per hill, maintaining a spacing of 20 cm × 15 cm in both instances. In the F₃ generation, rice seeds were sown on 30 March 2019, while in the F₄ generation, sowing took place on 4 April 2020. Light irrigation was provided to ensure the even germination of the sowing seeds of Aus rice accessions. Necessary intercultural operations, including gap filling, weeding, irrigation, drainage and plant protection actions, were uniformly performed to ensure the normal growth and development of the Aus rice in both experiments. The rice harvest took place when 80% of the fertile spikelets on the panicle attained a golden yellow colour.

Data collection

The data on yield and yield-contributing agronomic traits, as described in Table 1, were collected from five randomly selected plants within each investigated Aus rice accession in both experiments. The genotypic variance (σ ²_g) and phenotypic variance (σ ²_p) were calculated using below equations, respectively (Ullah et al., 2016; Makky et al., 2019)):

(1)$$\sigma ^2_{\rm g} = \displaystyle{{{\rm M}{\rm S}_{\rm g}-{\rm M}{\rm S}_{\rm e}} \over r}$$

(2)$$\sigma ^2_{\rm p} = \sigma ^2_{\rm g} + \sigma ^2_{\rm e} $$

where MS_g, genotype mean square; MS_e, error mean square; σ ²_e, environmental variance; r, the number of replications, the methodology described by Johnson et al. (1955).

Table 1. Name of the studied traits, along with their acronyms, measurement units and data collection procedures

The phenotypic coefficient of variation (PCV) percentage and genotypic coefficient of variation (GCV) percentage were calculated using following formulas, respectively:

(3)$${\rm PCV}\;( {\rm \% } ) = \displaystyle{{\sigma ^2_{\rm p} } \over {\bar{X}}} \times 100$$

(4)$${\rm GCV}\;( {\rm \% } ) = \displaystyle{{\sigma ^2_{\rm g} } \over {\bar{X}}} \times 100$$

where $\bar{X}$ denotes mean of the population, the methodology described by Burton and DeVane (1953). The PCV and GCV values were classified as 0–10, 10–20 and ≥20% indicated low, moderate and high, respectively.

Broad-sense heritability (H ²) percentage was calculated using below equation, the methods described by Johnson et al. (1955):

(5)$$H^2\;( \% ) = \displaystyle{{\sigma ^2_{\rm g} } \over {\sigma ^2_{\rm p} }} \times 100$$

The heritability percentage was classified as 0–30, 30–60 and ≥60% respectively, as low, moderate and high, respectively, according to Robinson et al. (1949).

Genetic advance (GA) and genetic advance as mean percentage (GAM) were calculated using below formulas, respectively:

(6)$${\rm GA} = K \times H^2 \times \sqrt {\sigma ^2_{\rm p}}$$

(7)$${\rm GAM} = K \times \displaystyle{{\sqrt {\sigma ^2_{\rm p} } } \over {\bar{X}}} \times H^2 \times 100$$

where K is a constant representing the selection intensity. When K is 5%, the corresponding value is 2.06 according to Assefa et al. (2000). GAM was classified as 0–10, 10–20 and >20% indicated low, moderate and high, respectively, according to Johnson et al. (1955).

Statistical analysis

The descriptive statistics, including the range, mean, standard error of the mean, skewness, coefficient of variation and Fisher's least significant difference of each studied characteristic were conducted using statistical software SPSS version 25 (Islam et al., 2019). The variance of components, such as σ ²_g, σ ²_p, σ ²_e, PCV, GCV, H ², GA and GAM, was assessed using the ‘variability’ package in R version 3.6.3. The packages ‘corrplot’, ‘pheatmap’ and ‘factoextra’ were used to generate correlation matrix, heatmap, hierarchical cluster and K-means clustering, respectively, in R statistical software version 3.6.3 (https://www.rproject.org/, accessed on 17 March 2023) (Islam, 2022).

Results

Agro-morphological variability of Aus rice accessions

In the current study, we estimated the range, mean, skewness, coefficient of variation (CV) and least significant difference at 5% level of significance (LSD_0.05) of the studied agronomic traits such as day-to-maturity (DM), plant-height (PH), total tillers number per hill (TT), effective tillers number per hill (ET), panicle length (PL), filled-grains per panicle (FG), sterility percentage (St), thousand-grain-weight (TW), grain-yield (GY) and harvest-index (HI) of Aus rice accessions in the F₃ and F₄ generations from the 48 Aus rice accessions (Table 2). The investigated genotypes exhibited average values for DM (99.10 d), PH (119.50 cm), TT (16.10), ET (14.90), PL (26.50 cm), FG (92.40), St (33.60%), TW (28.60 g), GY (23.20 g) and HI (42.10%) in the F₃ generation of Aus rice accessions. Meanwhile, the F₄ generation showed mean values for DM (99.50 d), PH (125.00 cm), TT (14.20), ET (12.50), PL (25.70 cm), FG (108.00), St (27.60%), TW (26.20 g), GY (28.60 g) and HI (51.10%) (Table 2). The studied traits exhibited both positive and negative skewness according to their distribution pattern. In the F₃ generation, all the traits skewed positively except for PL and HI, whereas in the F₄ generation, traits such as DM, PH, PL, FG, St and TW exhibited positive skewness, while TT, ET, GY and HI skewed negatively (Table 2). In the F₃ generation, the highest CV value was found in St (49.50) followed by GY (44.30), TT (32.80), FG (31.80), ET (30.90) and HI (22.80). Conversely, in the F₄ generation, the highest CV value was observed in St (47.70) followed by ET (21.70), TT (20.90), FG (19.90) and GY (18.70). These high CV values expressed greater phenotypic variations, while other traits exhibited lower variations.

Table 2. Descriptive statistics of the studied traits in the F₃ and F₄ generations of Aus rice accessions

CV, coefficient of variations; DM, day-to-maturity; ET, effective tillers number per hill; FG, filled-grains per panicle; GY, grain-yield; HI, harvest-index; PH, plant-height; PL, panicle length; SEM, standard error of mean; St, sterility %; TT, total tillers number per hill; TW, thousand-grain-weight.

Values were derived from 48 rice accessions in the F₃ and F₄ generations from three distinct biological replicates (n = 3), each replicate consisting of five plants.

Phenotypic and genotypic coefficients of variation

The assessment of the genetic variability including genotypic variance (σ ²_g), phenotypic variance (σ ²_p), PCV percentage, GCV percentage, H ² percentage, GA and GAM is vital for the improvement of new genotypes. In the present study, the σ ²_p for TT, ET, PL and TW was slightly superior to the σ ²_g in both the F₃ and F₄ generations (Table 3). There were noticeable deviations between σ ²_p and σ ²_g for the traits of DM, PH, FG, St, GY and SY. There was greater variation in values of GCV and PCV for the studied parameters in both the generations and most of the traits had higher PCV values than GCV. There was minimal differentiation between GCV and PCV in most of the studied traits, except for TT, ET, FG and GY in both experiments (Table 3). In the present study, the traits including TT, ET, FG, St, GY and SY exhibited higher PCV values (>20%), whereas PH and TW, DM and PL showed moderate (10–20%) and lower values (10%) respectively. Similarly, the level of GCV was high (>20%) for TT, St and FG, and moderate (10–20%) for PH, ET, TW, GY and SY, in both the F₃ and F₄ generations (Table 3).

Table 3. Estimation of genotypic variance (σ ²_g), phenotypic variance (σ ²_p), GCV, PCV, broad-sense heritability (H ²) percentage, genetic advance (GA) and genetic advance as mean percentage (GAM) for the studied traits in the F₃ and F₄ generations of Aus rice accessions

DM, day-to-maturity; ET, effective tillers number per hill; FG, filled-grains per panicle; GY, grain-yield; PH, plant-height; PL, panicle length; St, sterility %; SY, straw yield; TT, total tillers number per hill; TW, thousand-grain-weight.

Heritability and genetic advance of Aus rice accessions

Heritability and genetic advancement form a powerful duo in the geneticist's toolkit. Heritability elucidates the historical genetic backdrop, elucidating the intrinsic nature of trait variability, while genetic advance provides a roadmap for future progress, guiding breeders in their efforts to systematically enhance the genetic potential of a population through selective breeding strategies. In our results, the H ² varied from 39.30 (ET) to 99.50 (St) in the F₃ generation, while it ranged from 40.30 (ET) to 97.60 (DM) in the F₄ generation (Table 3). A high range of H ² implies that the genetic component has a substantial impact on the trait's expression, making it a promising candidate for selection and improvement in breeding programmes. The studied traits showed notably high (H ²) in both the F₃ and F₄ generations for St (99.50 and 97.20), TW (88.10 and 90.20), PH (84.90 and 86.90) and DM (84.50 and 97.60), respectively. In contrast, the traits exhibited high GAM in St (48.00 and 50.60), ET (44.70 and 47.10), TT (43.00 and 45.40) and FG (41.00 and 43.20) in both the F₃ and F₄ generations, respectively (Table 3).

Correlation coefficient of the studied agronomic traits

The Pearson's correlation coefficient is a frequently used approach that provides information about the nature and extent of relationships, thus aiding in the selection of traits for improvement (Singh et al., 2021b). The correlation coefficient analysis showed both positive and negative relationships between the studied agronomical traits in the F₃ and F₄ generations (Fig. 1). In the F₃ generation, strong-significant positive correlations were observed between TT and ET (r = 0.97), and moderate positive associations between GY and HI (r = 0.63), GY and SY (r = 0.57), ET and SY (r = 0.59), TT and SY (r = 0.64) whereas low positive correlations were established between PH and PL (r = 0.42), DM and St (r = 0.36), DM and GY (r = 0.33), TT and GY (r = 0.31), DM and HI (r = 0.31), ET and GY (r = 0.30) and PH and TW (r = 0.30). Meanwhile, moderate negative associations were observed between FG and TW (r = −0.53) as well as negative relationships between FG and St (r = −0.42), ET and TW (r = −0.31) and TT and TW (r = −0.31) (Fig. 1a). On the other hand, the correlation among the agronomic traits of F₄ generation of Aus rice accessions varied from strong positive to low negative. Strong positive correlations were observed between ET and TT (r = 0.97), and moderate positive correlations were found in GY and HI (r = 0.67) whereas a low positive association was found between PH and PL (r = 0.46), TT and GY (r = 0.45), ET and GY (r = 0.44), TW and SY (r = 0.41), GY and SY (r = 0.39), TT and SY (r = 0.35), ET and SY (r = 0.30) and DM and FG (r = 0.30). Conversely, moderately negative associations were observed among ET and FG (r = −0.62), TT and FG (r = −0.61) and FG and St (r = −0.58) in the F₄ generation (Fig. 1b).

Figure 1. Illustration of the Pearson's correlation among the investigated agronomic traits in Aus rice accessions. The correlation coefficient (r-value) was calculated from the mean values of rice accessions in the F₃ (Fig. 3a) and F₄ (Fig. 3b) generations. The colour intensity signifies the degree of correlation, with +1 indicating a strong positive correlation (dark purple) and −1 indicating a strong negative correlation (dark red) between the two traits. DM, day-to-maturity; ET, effective tillers number per hill; FG, filled-grains per panicle; GY, grain-yield; HI, harvest-index; PH, plant-height; PL, panicle length; St, sterility %; SY, straw yield; TT, total tillers number per hill; TW, thousand-grain-weight.

Cluster analysis revealed interactions between the studied traits and accessions

In the present study, the heatmap with hierarchical cluster analysis (HCA) generated two main row and column clusters. Subsequently, investigated rice accessions were sorted into four row clusters horizontally, and the important characters were grouped into five column clusters vertically using the mean values of all investigated traits of rice accessions in both the generations (Fig. 2). The HCA revealed that clusters 1 and 4 had the highest number of accessions (16), followed by cluster 2 (13) and cluster 3, which had only three accessions in the F₃ generation (Fig. 2a). The colour intensities of red to blue colour clearly illustrate the values of the yield and yield-attributing traits, such as GY, TT, ET, FG and SY, were found to be maximum in the accessions of cluster 1, followed by cluster 4. On the other hand, cluster 3 accessions had the shortest maturation duration, followed by cluster 2. The accessions of cluster 4 exhibited moderate position for both yield and maturity duration in the F₃ generation (Fig. 2a). The traits SY, TT and ET were assigned to group 1, while TW, PH and PL were in group 2. Group 3 included two characters, DM and St, while group 4 consisted of only FG. The traits GY and HI were connected to group 5. The performance gradient (highest to lowest) of individual accessions against the trait was indicated by colour intensities from red to blue and vice versa. The investigated accessions, namely, R1-6-2-3-1, R1-9-10-1-1, R1-13-1-1-1, R1-31-6-2-1, R1-49-7-1-1, R1-50-1-1-1, R2-26-6-2-2, R3-26-4-3-1 and R3-32-10-1-1 produced the highest grain yield, whereas R3-30-1-1-1, R3-30-1-2-1, R2-56-5-2-1, R3-11-9-1-1, R3-30-1-3-1, R1-44-1-2-1, R2-31-1-2-1 and R2-54-3-3-1 had the lowest maturity duration. There were also some potential accessions, such as R1-49-7-1-1, R1-6-2-3-1, R1-50-1-1-1 and R3-32-10-1-1, with moderate growth duration and higher grain production (Fig. 2a). On the other hand, the heatmap clustering of the F₄ generation showed that 12 investigated rice accessions were organized into cluster 1, while clusters 2, 3 and 4 had 8, 15 and 13 accessions, respectively (Fig. 2b). The studied traits, like GY, TT, ET and SY were found in cluster 4, but FG was higher in cluster 1. However, cluster 2 had the shortest crop maturity period, followed by cluster 4. The column cluster of important agronomic traits revealed that group 1 was linked with St, TT and ET; group 2 with PH and PL; group 3 with TW and SY; group 4 with GY and HI, and finally, group 5 allied with DM and FG (Fig. 2b). The Aus rice accessions namely R3-49-4-3-1, R1-47-7-3-1, R2-26-6-2-2, R3-30-1-2-1, R1-44-1-2-1, R3-30-1-3-1, R3-49-4-1-1, R2-31-1-2-2, R1-6-2-3-1, R1-50-1-1-1 and R1-6-2-1-1 showed maximum grain yield, whereas R2-56-5-2-1, R3-11-9-1-1, R3-30-1-1-1, R3-30-1-2-1, R3-30-1-3-1, R2-54-3-3-1, R3-49-4-1-1, R3-49-4-3-1, R2-30-9-3-1, R2-12-8-3-1 and R1-56-3-3-1 accessions revealed the minimum maturity duration among the accessions (Fig. 2b).

Figure 2. Heatmap with hierarchical clustering generated by Euclidean distance using Ward's method based on yield and yield-related agronomic traits of F₃ (a) and F₄ (b) generations of Aus rice accessions. In the double dendrogram, each row (X-axis) depicts an accession, and each column (Y-axis) represents a trait. The colours (red to blue) and their intensities (3 to −3) were adjusted based on the genotypes–trait relationship. The colour spectrum illustrates that the values greater than the mean are categorized as golden, and vice versa for black. DM, day-to-maturity; ET, effective tillers number per hill; FG, filled-grains per panicle; GY, grain-yield; HI, harvest-index; PH, plant-height; PL, panicle length; St, sterility %; SY, straw yield; TT, total tillers number per hill; TW, thousand-grain-weight.

K-means cluster analysis was performed with a specific emphasis on the traits of GY and DM across the F₃ and F₄ generations, to confirm the association of accessions with the traits GY and DM (Fig. 3). In the F₃ generation, the genotypes were arranged into clusters 1, 2, 3, 4 and 5 comprising 9, 6, 5, 14 and 14 genotypes, respectively (Fig. 3a). Among the cluster groups, cluster 2 accessions, including R1-6-2-3-1, R1-13-1-1-1, R1-31-6-2-1, R1-49-7-1-1, R1-50-1-1-1 and R3-26-4-3-1, showed the highest GY (42.20 g) and also had the highest TT (20.20), ET (18.80), FG (114.30), SY (45.50 g) and HI (48.70%) followed by cluster 4 in the F₃ generation (online Supplementary Table S4). Furthermore, in the F₄ generation, clusters 1, 2, 3, 4 and 5 consisted of 12, 14, 12, 3 and 7 investigated accessions, respectively. The rice accessions of cluster 2 were more tightly associated with grain yield, followed by cluster 3, whereas cluster 5 had the lowest maturity duration, followed by cluster 2 (Fig. 3b). The mean value of the cluster 2 accessions like R1-6-2-1-1, R1-6-2-3-1, R1-48-3-2-1, R1-26-6-2-2, R1-44-1-2-1, R1-47-7-3-1, R1-50-1-1-1, R2-31-1-2-2, R3-30-1-1-1, R3-30-1-2-1, R3-30-1-3-1, R3-32-10-1-1, R3-49-4-1-1 and R3-49-4-3-1 exhibited the highest GY (34.50 g) with the second lowest DM (97.70 d) among the cluster groups. They also had the highest TT (15.30), PL (26.30 cm), FG (113.80), TW (27.00 g), SY (30.00 g) and HI (53.90%) (online Supplementary Table S4). Interestingly, the accessions, namely R1-6-2-3-1, R1-13-1-1-1, R1-31-6-2-1, R1-49-7-1-1, R1-50-1-1-1 and R3-26-4-3-1 in the F₃ generation, and the accessions, R1-47-7-3-1, R2-26-6-2-2, R3-30-1-2-1, R1-44-1-2-1, R3-30-1-3-1, R2-31-1-2-2, R1-6-2-3-1, R1-50-1-1-1, R1-6-2-1-1, R3-30-1-1-1, R3-30-1-2-1, R3-30-1-3-1, R3-49-4-1-1 and R3-49-4-3-1 in the F₄ generation, exhibited the highest yield and yield-attributing traits and moderate growth duration according to K-means clustering. The mean values of grain yield and yield-related traits for each individual rice accession with genotypic code is presented in online Supplementary Tables S2 and S3 for both the F₃ and F₄ generations, respectively.

Figure 3. K-means clustering of Aus rice accessions in F₃ (a) and F₄ (b) generations performed in relation to grain-yield and days-to-maturity. The genotypes were grouped into five distinct clusters, with different clusters represented by various coloured polygons. Genotypes were identified using their genotypic code, and the corresponding names were provided in online Supplementary Table S2 for F₃ and online Supplementary Table S3 for F₄ generations.

Discussion

Our investigation revealed substantial variation in growth, yield and yield-attributing traits among the newly developed Aus rice accessions in both the F₃ and F₄ generations. The growth, yield and yield-contributing characters including DM, PH, TT, ET, PL, FG, St, TW, GY, SY and HI of the investigated Aus rice accessions differed significantly due to their genetic variation and opened the opportunity for selection to enhance desirable traits in both the generations (Table 2). In the realm of descriptive statistics, skewness serves as a valuable tool for researchers, aiding in the assessment of the shape and characteristics inherent in a dataset. Our dataset suggests that most traits, exhibiting either positive or negative skewness, are likely influenced by complementary and duplicate gene interactions. This highlights the potential for effective selection in varietal improvement. Patel et al. (2020) noted that complementary and duplicate interactions are present when the skewness value is greater or smaller than zero. The higher values of CV found in St, ET, TT, FG and GY indicate more phenotypic variation, and breeders can use selection to improve crops by managing the available genetic variability of such traits. Moreover, the positive and negative skewness observed in agronomical traits in this study suggested complementary gene interactions in succeeding segregating Aus rice populations. These results are consistent with the findings of Abdala et al. (2016) and Kiani (2012), who observed significant differentiations between genotypes for the studied traits and phenotypic variations.

Genetic variability among genotypes is a crucial criterion in breeding programmes, enabling the selection of the desired types for hybridization (Nihad et al., 2021). The assessments of variability, such as σ ²_g, σ ²_p, GCV, PCV, H ², GA and GAM were computed for F₃ and F₄ Aus rice populations, encompassing investigated agronomical traits in this study (Table 3). Genotype selection based only on σ ²_p may be misleading as it splits into genotypic as well as environmental effects, respectively (Zaid et al., 2022). Moreover, it is necessary to estimate σ ²_g along with σ ²_p to extend the opportunity for selection (Hasan-Ud-Daula and Sarker, 2020). In this study, the traits PH, FG and St had high σ ²_g, indicating these features had more genetic attribution and a wider chance for selection. Additionally, the variance component σ ²_p was observed to be higher than σ ²_g, suggesting that environmental factors have a more significant influence on the expression of these traits compared to genetic factors. These imply a broad spectrum of genomic diversity in these traits among the population, implying the potential for increased rice production by choosing individuals exhibiting these traits. This observation aligns with the results reported by Dhurai et al. (2016), who found minor differences between σ ²_p and σ ²_g for PH, FG, St and HI in rice. The PCV and GCV can be used to assess and compare the nature and range of variability existing for the studied traits (Gupta et al., 2022). High PCV and GCV values indicate a significant extent of genetic diversity, suggesting that the traits could be selected in the varietal improvement procedure (Sharma et al., 2021). The results of our study indicated that there were high to moderate GCV and PCV values in the traits such as St, TT, FG, GY and SY in both the generations. The results closely agree with the previous research of Faysal et al. (2022), who reported moderate-to-high GCV and PCV for traits such as PH, FG, St, GY, flag leaf length and number of secondary branches per panicle in Aman rice. Moreover, the assessed values of PCV were consistently higher than GCV for all of the investigated traits, indicating the influence of environmental variables on the expression of these traits, which aligns with the results of Singh et al. (2021a).

The H ² and GAM are vital for any breeding programme as they guide parent selection based on the measurable characters, and also indicate the selection response, considering the existing genetic variability and heritability for specific traits (Burton and DeVane, 1953). In our study, the investigated traits displayed high-to-moderate heritability. Notably, the high H ², linked with high GAM was observed for four traits, namely, PH, FG, St and TW in both the F₃ and F₄ generations. Proposing the additive for the traits that suggests the selection process is effective in advancing the genetic potential for specific traits within the population in the rice breeding programmes. These findings underscore the potential of selecting for these specific traits as a substantial contribution to the enhancement of Aus rice accessions. Similar findings were reported by Hossain and Haque (2016), who observed that the FG, St, GY and secondary branches number per panicle in rice, Hasan-Ud-Daula and Sarker (2020) and Faysal et al. (2022) for yield and yield-attributing traits in rice.

The relationship level between the investigated traits determines the efficiency of trait selection for breeding programmes. Thus, the analysis of the correlation among the studied traits is imperative for considering the possibility of selecting two or more traits (Sharma et al., 2021). A positive relationship between two desired traits facilitates concurrent improvement in both traits in a plant breeding programme (Islam et al., 2018). In our present study, a positive and significant correlation was found between GY and HI, SY in the F₃ generation (Fig. 1a) and GY and TT, HI, ET and SY in the F₄ generation (Fig. 1b). However, the trait DM had a significant positive correlation between GY, HI and St in the F₃ generation (Fig. 1a) and also only with FG in the F₄ generation (Fig. 1b). These results align with the findings of Fentie et al. (2021), who also observed a positive association between GY and FG, BY, HI, PL, panicle weight and flag leaf length of rice. Selection for these traits may be essential for improving crop varieties as they have an indirect effect on yield improvement of rice. These results agree with the research findings of Sarkar et al. (2014), who observed significant positive and negative relationships among the yield-related agronomic traits in rice.

Cluster analyses assist breeders in choosing parents for plant breeding programmes by elucidating trait–trait relationships among the examined genotypes (Huqe et al., 2021; Ata-Ul-Karim et al., 2022). In our study, HCA was employed to categorize distinct clusters based on the 11 investigated agronomic traits and 48 Aus rice accessions in the F₃ (Fig. 2a) and F₄ generations (Fig. 2b). Furthermore, we carried out K-means clustering with an emphasis on GY and DM. The rice accessions, R1-6-2-3-1, R1-13-1-1-1, R1-31-6-2-1, R1-49-7-1-1, R1-50-1-1-1 and R3-26-4-3-1 in the F₃ generation, and the accessions R1-47-7-3-1, R2-26-6-2-2, R3-30-1-2-1, R1-44-1-2-1, R3-30-1-3-1, R2-31-1-2-2, R1-6-2-3-1, R1-50-1-1-1, R1-6-2-1-1, R3-30-1-1-1, R3-30-1-2-1, R3-30-1-3-1, R3-49-4-1-1 and R3-49-4-3-1 in the F₄ generation, were identified as top performers based on higher GY and shorter DM through HCA and K-means clustering (Figs. 2 and 3). It is indicating that these rice accessions have the highest genetic potential in the breeding programme. These findings align with Kaysar et al. (2022), who identified promising cultivars such as BRRI dhan29, BRRI dhan58, Binadhan-10, Hira-2 and Tejgold under subtropical conditions through hierarchical clustering based on the studied agronomic traits.

Conclusion

The yield and yield-related agronomical traits of 48 rice accessions exhibited significant genetic variability in both the F₃ and F₄ generations. Variance components, including PCV and GCV, provided insights on the extent of phenotypic variation in the elite rice accessions. The extended value of H ² and GAM in TT, ET, FG, St, TW and GY indicated them as the most important traits and thus deserve greater attention in breeding programmes. Certain Aus rice accessions demonstrated higher grain yield (GY) and shorter to moderate days-to-maturity (DM). Specifically, the F₃ generation included accessions R1-49-7-1-1, R3-26-4-3-1, R1-6-2-3-1, R1-13-1-1-1 and R1-50-1-1-1, while the F₄ generation included accessions R3-49-4-3-1, R1-47-7-3-1, R2-26-6-2-2, R3-30-1-2-1 and R1-44-1-2-1. These promising Aus lines could be useful in future breeding programmes under rainfed conditions. This study recommends evaluating the location-specific performance of chosen genotypes for higher yield stability where water is scarce.