Direct-seeded rice as a promising approach for sustainable rice cultivation: multi-trait evaluation and selection of adapted genotypes

1 Introduction

Rice (Oryza sativa L.) is a prominent food crop and serves as a staple dietary food for nearly half of the global population, providing essential calories and nutrients to approximately 3.5 billion people (Fukagawa and Ziska, 2019; Kumari et al., 2020; Mohidem et al., 2022). A growing population, along with food demand and urbanization, is expected to increase the global rice production from approximately 510 million tonnes in 2023 to 550 million tonnes by 2030 and nearly 590 million tonnes by 2040 (Yuan et al., 2021). Asia remains the leading producer and consumer of rice, with India being the second-largest producer. In India, rice is cultivated on about 51.42 million ha, producing 149.07 million tonnes, with an average yield of 2,899 kg ha^-1 (Indiastat, 2025). Total irrigated area under irrigated rice is estimated at nearly 6.0 million hectares, and lowland rice cultivation is estimated at nearly 13.0 million hectares (ICAR–Central Rice Research Institute, 2025). Traditionally, rice is grown with puddled transplanted systems where nursery-grown seedlings are transplanted into flooded fields 15–40 days after germination. Although this technique yields well and keeps weeds free, it is highly labor-intensive and water-demanding, requiring nearly 5000 L of water to produce 1 kg of rice, with agriculture set to experience a 10-15% decrease in water supply (Kaur and Singh, 2017). Due to these major constraints, Puducherry rice production declined from 3,133 to 2761 kg ha^-1 (CEIC Data, 2026). Therefore, the development and adaptation of water-efficient crop establishment methods have become essential for sustainable rice production.

Direct-Seeded Rice (DSR) has emerged as a water-conserving alternative strategy to traditional transplanting because of its lower water and labor requirements (Dhillon and Mangat, 2018). The three crop establishment methods, namely dry direct-seeded rice (DDSR), wet direct-seeded rice (WDSR) and transplanted rice (TPR), create contrasting soil and crop-growing environments. DDSR is established by sowing dry seed directly into unpuddled soil, whereas WDSR involves sowing pre-germinated seed in puddled and wet soil, while TPR involves transplanting nursery-raised seedlings into puddled soil. Consequently, these establishment methods differ in soil moisture dynamics, aeration and water regime which can influence crop growth and grain yield. DDSR is generally associated with relatively drier and more aerobic soil conditions during crop establishment, whereas WDSR and TPR are established under wetter and less-aerated conditions following puddling. Such differences in soil moisture and aeration, together with differences in crop establishment conditions, can influence crop growth, resource-use efficiency and grain yield (Jin et al., 2023). In recent years, DSR has gained considerable popularity among South and Southeast Asian rice-growing farmers because it reduces labor requirements, irrigation water use and production costs compared with transplanted rice (TPR) (Saravanane, 2020; Pooja et al., 2021; Karthickraja et al., 2025). It is a viable option for shifting from transplanted rice to direct-seeded rice (Kumar and Ladha, 2011). DSR can also attain a grain yield as compared to transplanted rice (Reddy et al., 2025). However, its success largely depends on adopting suitable agronomic practices, including optimal sowing time, seed rate, choice of the variety, weed management practices and irrigation management (Chaudhary et al., 2023). Furthermore, mechanized DSR enhance operational efficiency and improves benefit-cost ratios rather than TPR (Devkota et al., 2019; Sandhu et al., 2021). Since DSR is grown under aerobic or intermittently flooded conditions, it results in considerably lower methane emissions and reduced global warming when compared to TPR (Alam et al., 2020). In addition, DSR preserve soil structure, improves soil physical and biological properties and contributes to good organic carbon accumulation, enhancing sustainable crop production and improving the performance of succeeding crops (Yadav et al., 2017; Alam et al., 2020; Kumar et al., 2022). Therefore, the present study was conducted to evaluate the performance and stability of rice genotypes across three contrasting crop establishment methods (DDSR, WDSR and TPR) and to identify suitable genotypes and the most favorable crop establishment method for rice production under water-deficit conditions.

2 Materials and methods

2.1 Experimental details

The experiment was carried out in Pandit Jawaharlal Nehru College of Agriculture and Research Institute, Karaikal (PAJANCOA & RI), during the zaid season (March-June, 2025). The study was laid out in a strip plot design with three replications. Three crop establishment methods were assigned to the horizontal plots, viz., Dry direct-seeded rice (DDSR), Wet direct-seeded rice (WDSR) and Transplanted rice (TPR). Thirty rice genotypes were allocated to the vertical plots under each establishment method. The experimental materials comprised 17 released varieties, four landraces and nine advanced breeding lines (Table 1). The study aimed to evaluate the best environment among three establishment methods and also to identify good genotypes for each environment. Each experimental plot measured 3.0 m × 1.0 m with five rows maintained at a spacing of 20 × 15 cm.

Table 1. List of genotypes used in this study.

2.2 Crop establishment and intercultural operations

The three crop establishment methods, namely dry direct-seeded rice (DDSR), wet direct-seeded rice (WDSR) and transplanted rice (TPR), were established following the recommended practices for rice cultivation under the region. The total experimental area of 0.2 ha was equally allocated among the three crop establishment methods. The intercultural operations followed under each crop establishment method are presented in Table 2 (Anonymous, 2020).

Table 2. Intercultural operations and crop management practices followed under different crop establishment methods in rice.

2.3 Trait measurement

Five representative plants from each genotype in every replication were randomly tagged for recording observations. Data were recorded for 22 parameters categorized into phenological, physiological, yield-related and grain-quality traits viz., normalized difference vegetation index (NDVI), chlorophyll content index (CCI), canopy temperature depression (CTD), chlorophyll fluorescence (Fv/Fm), days to 50 per cent flowering (DFF), number of productive tillers per plant (NPTP), days to maturity (DTM), plant height (PHT), root length (RL), shoot fresh weight (SFW), root fresh weight (RFW), shoot dry weight (SDW), root dry weight (RDW), panicle length (PL), number of spikelets (NOS), panicle fertility percent (PFP), single plant yield (SPY), thousand grain weight (TGW), plot yield (PY), milled rice recovery (MRR), head rice recovery (HRR) and starch content (STC).

2.4 Statistical analysis

All statistical analyses were performed using R statistical software (version 4.5.2; R Core Team, 2025). Analysis of variance (ANOVA) was conducted to determine the significance of genotype, crop establishment method and their interaction. Phenotypic correlation analysis, hierarchical clustering and the multi-trait genotype-ideotype distance index (MGIDI) and box-and-whisker plot analysis were performed to evaluate trait relationships, classify genotypes, identify superior genotypes based on multiple traits and compare trait distribution across crop establishment methods.

3 Results

A detailed analysis of various traits, including phenological, physiological, yield-related and grain-quality traits under three contrasting environments was evaluated. The mean performance and descriptive statistics were provided in Table 3, showing the distribution of the trait values for the three environments and indicating changes in central tendency, variability and dispersion.

Table 3. Mean performance for twenty-two characters observed in thirty rice genotypes under three different treatments.

^awithin genotypes (G), ^bwithin treatment (T), ^cG × T.

3.1 Analysis of strip plot variance

ANOVA (Table 4) revealed significant differences between genotypes as well as treatments in all the 22 traits measured under DDSR, WDSR and TPR, demonstrating the existence of a large amount of genetic variability in the material and different establishment methods under investigation. The genotype x Environment interaction was significant in most of the characters except thousand grain weight (TGW), milling rice recovery (MRR) and head rice recovery (HRR).

Table 4. Strip plot analysis of variance for the observed traits.

*significance at 10% level, significance at 5% level, *significance at 1% level; ^awithin genotypes (G), ^bwithin treatment (T), ^cG x T, ns, non significant.

3.2 Phenotypic correlations and yield associations over three environments

3.2.1 Phenotypic correlation for DDSR

Phenotypic correlation for various morphological and yield traits was estimated to determine the relationship between the establishment methods and the single plant yield. Under DDSR, SPY had high positive correlations with shoot fresh weight (r = 0.36*), shoot dry weight (r = 0.36), number of spikelets per panicle (r = 0.51), panicle fertility percentage (r = 0.52) and thousand grain weight (r = 0.31) (Figure 1*).

3.2.2 Phenotypic correlation for WDSR

In WDSR, single plant yield (SPY) had significant positive relationships with shoot fresh weight (r = 0.34*), shoot dry weight (r = 0.34), number of productive tillers per plant (r = 0.30), number of spikelets per panicle (r = 0.53), panicle fertility percentage (r = 0.29) and thousand grain weight (r = 0.24) (Figure 2*).

3.2.3 Phenotypic correlation for TPR

In TPR, significant positive correlations were observed between SPY and NDVI (r = 0.240), shoot fresh weight (r = 0.40), shoot dry weight (r = 0.40), number of productive tillers per plant (r = 0.28), number of spikelets per panicle (r = 0.61) and percentage of panicle fertility (r = 0.41) in TPR (Figure 3**).

3.3 Hierarchical cluster analysis over three environments

3.3.1 Cluster analysis for DDR

All 30 genotypes for various traits under the DDSR condition were grouped into five clusters (Figure 4). Among the five clusters, Cluster I was the largest, comprising 10 genotypes. Cluster II included 7 genotypes, followed by Cluster IV with 5 genotypes. Clusters III and V were the smallest, with 4 genotypes each.

3.3.2 Cluster analysis for WDSR

Cluster analysis of the WDSR condition was grouped into five clusters for 30 genotypes (Figure 5). Among the five clusters, cluster I was the largest, comprising 12 genotypes. Clusters II and V contained 6 genotypes each, followed by Cluster IV with 5 genotypes. Cluster III was the smallest, consisting of only 2 genotypes.

3.3.3 Cluster analysis for TPR

Under the TPR condition, 30 genotypes were grouped into five clusters (Figure 6). Among the five clusters, Clusters II and V were the largest, each comprising 8 genotypes. Cluster III consisted of 7 genotypes, while Cluster I had 4 genotypes. Cluster IV was the smallest, containing only 3 genotypes.

3.4 Multi-trait genotype-ideotype distance index for DDSR, WDSR and TPR

MGIDI analysis for three different environments retained eight principal components, explaining 78.94%, 78.10% and 78.94% accounts total phenotypic variation DDSR, WDSR and TPR, respectively (Tables 5–7), indicating the retained factor shows adequate variability among evaluated traits. 78.94%, 76.83% and 76.83%, demonstrating its effectiveness in simultaneously improving multiple agronomic, physiological, root, yield and grain quality traits for DDSR, WDSR and TPR, respectively (Tables 8–10). Based on MGIDI index genotypes like MTU 2716, I.W. Ponni, PKM 4 and KR 21003, CO 51, ADT 58, KKL(R) 2 and TKM 9 for DDSR (Figure 7; Table 11). In WDSR MTU 2716, I.W. Ponni, ADT 48, IR 50, CO 51, ADT 54 and KKL(R) 2 (Figure 8; Table 11). For TPR, BPT 5204, IR 20, I.M. Samba Mashuri, MTU 2716, KR 21003, IR 50, KKL(R) 2 and CO 51 (Figure 9; Table 11). These genotypes are identified as the best performers for their respective environment. Strength and weakness analysis suggested each genotype possessed different desirable attributes for different factor groups, indicating the presence of a complementary trait combination useful for variety development.

Table 5. Factor loadings under dry direct-seeded conditions for twenty-two traits.

Table 6. Factor loadings under wet direct-seeded conditions for twenty-two traits.

Table 7. Factor loadings under transplanted conditions for twenty-two traits.

Table 8. Selection gain (%) with desired gains under dry direct-seeded conditions for twenty-two traits.

Table 9. Selection gain (%) with desired gains under wet direct-seeded conditions for twenty-two traits.

Table 10. Selection gain (%) with desired gains under transplanted conditions for twenty-two traits.

Table 11. MGIDI index for 30 genotypes under three treatments.

3.5 Box-and-whisker plot for plot yield of DDSR, WDSR and TPR

The box-and-whisker plot revealed clear variation in plot mean yield of thirty genotypes among three different crop establishment methods (Figure 10). E2P (WDSR) recorded the highest yield, followed by E3P (TPR) and E1P (DDSR) showed a low yield. The wider spread of yield distribution in E2P and E3P indicates greater variability among genotypes and shows that E2P provides the most favorable environment for potential genotypic expression.

4 Discussion

Significant effects of genotype, crop establishment method and their interaction indicate that rice performance was strongly influenced by both genetic constitution and the production environment. The differential responses of genotypes across environments suggest that the expression of agronomic, physiological and grain quality traits is environment-dependent, This finding is consistent with Khare et al. (2024), who reported significant effects of genotype, environment and genotype × environment interaction for yield and yield-associated traits in rice. Similarly, Bocianowski et al. (2024) and Popoola et al. (2024) reported significant genotype × environment interactions and differential genotypic performance across environments, highlighting the importance of multi-environment evaluation for identifying stable and widely adapted genotypes. In the present study, the significant genotype × crop establishment method interaction indicates that genotype performance varied across DDSR, WDSR and TPR, confirming that the suitability of a genotype was dependent on the crop establishment environment.

Correlation analysis revealed single plant yield exhibited significant positive associations with shoot fresh weight, shoot dry weight, number of spikelets per panicle, panicle fertility percentage and thousand-grain weight under DDSR, these positive associations indicate that biomass accumulation, spikelet formation, panicle fertility and grain weight contributed to variation in single plant yield under DDSR. Greater biomass production may enhance assimilate availability during reproductive development, while higher spikelet number and panicle fertility can increase sink capacity and grain yield (Liang et al., 2023). Zhou et al. (2019) suggested that increased dry matter accumulation improved grain yield, highlighting the importance of harvest index for achieving higher productivity.

Under WDSR, single plant yield showed significant positive associations with shoot fresh weight, shoot dry weight, number of productive tillers per plant, number of spikelets per panicle, panicle fertility percentage and thousand-grain weight. These relationships suggest that both vegetative biomass production and reproductive attributes were positively associated with yield variation under wet direct-seeded conditions. Similar positive associations of grain yield with productive tillers, spikelet fertility, panicle traits and grain weight have been reported in recent rice studies (Chattar et al., 2024; Kumar et al., 2025).

Single plant yield is positively associated with Normalized differential vegetative index, shoot fresh weight, shoot dry weight, number of productive tillers per plant, number of spikelets per panicle and percentage of panicle fertility in TPR. Greater biomass production and improved growth characteristics significantly improve the grain yield and productive tillers under transplanted conditions (Liu et al., 2018; Huang et al., 2020).

Amegan et al. (2020) reported clustering based on a dissimilarity matrix approach for identifying genetically diverse genotypes that can be utilized in hybridization programs to generate genetic variability. Hierarchical clustering grouped the 30 rice genotypes into five clusters under each crop establishment method, although the distribution of genotypes among clusters varied across DDSR, WDSR and TPR. This variation suggests that environmental conditions associated with crop establishment influenced the expression of phenotypic traits used for clustering. Rudresh et al. (2021) reported similarly that genotypes under different environments, like DDSR, WDSR and TPR, belong to different clusters and exhibit greater genetic diversity between genotypes.

MGIDI effectively integrates multiple correlated traits into a single selection approach used for the identification of genotypes with balanced agronomic performance, thereby reducing the effect of undesirable trait combinations. The strengths and weaknesses analysis further demonstrated that various selected genotypes contributed to different factor groups, facilitating the identification of complementary genotypes for future breeding programs. In MGIDI analysis, genotypes located near the cut-off point may also possess desirable attributes and be considered in future breeding programs (Olivoto and Nardino, 2021). Similar applications of the MGIDI index for efficient multi-trait genotype selection have also been demonstrated by Gabriel et al. (2019) and Debsharma et al. (2023).

The box-and-whisker analysis showed variation in plot yield among the three crop establishment methods, with WDSR exhibiting a higher plot yield than DDSR and TPR. The relatively higher performance of WDSR may be attributed to favorable crop establishment conditions associated with wet seeding in puddled soil, which can promote uniform crop establishment and subsequent yield development. Similar studies have reported that wet-seeded rice can achieve grain yields comparable to or, under suitable management conditions, higher than those of transplanted rice (Kaur et al., 2022).

5 Conclusion

The present study demonstrated significant genotype × crop establishment method interactions for most traits, confirming that genotypic performance was influenced by the establishment environment. Phenotypic correlation analysis identified biomass- and yield-component traits, particularly shoot fresh weight, shoot dry weight, number of spikelets per panicle and panicle fertility percentage, as important contributors to single plant yield across establishment methods. Hierarchical clustering revealed substantial phenotypic diversity among the 30 genotypes, with genotype grouping varying across DDSR, WDSR and TPR. Among the three establishment methods, WDSR recorded the highest plot yield and was identified as the most favorable environment for yield expression. MGIDI-based multi-trait selection identified MTU 2716, I.W. Ponni, ADT 48, IR 50, CO 51, ADT 54 and KKL(R) 2 as superior genotypes under WDSR. Overall, WDSR, particularly with the identified superior genotypes, appears to be a promising crop establishment option for maintaining high productivity under the conditions evaluated in the present study. However, the experiment was conducted at a single location during one season and direct measurements of water use, water-use efficiency and economic returns were not included. Therefore, multi-location and multi-season evaluation, together with assessment of water-use efficiency and economic viability, is required to validate the stability, adaptability and practical potential of the identified genotypes and establishment method.

Generative AI statement

The author(s) declared that generative AI was not used in the creation of this manuscript.

Any alternative text (alt text) provided alongside figures in this article has been generated by Frontiers with the support of artificial intelligence and reasonable efforts have been made to ensure accuracy, including review by the authors wherever possible. If you identify any issues, please contact us.

Publisher’s note

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.

Footnotes

Edited by: Xiaowen Wang, Jiangsu University, China

Reviewed by: Sanjoy K. Debsharma, Texas A&M AgriLife Research and Extension Center at Beaumont, United States

Fikri Adriansyah, Sriwijaya University, Indonesia