Sorghum (Sorghum bicolor L. Moench) is a versatile cereal crop cultivated for grains, forage and many industrial products. It ranks fifth in importance after wheat, rice, maize, and barley (Ritter et al., 2007; Motlhaodi et al., 2014). Sorghum is cultivated for food grains in most of the arid and semi-arid regions of Asia and Africa. Whereas, in the western countries it is cultivated for green forage and feed (Almas et al., 2015; Getachew et al., 2016; Aruna and Cheruku, 2019; Mundia et al., 2019). In most of the resource poor countries sorghum stalk is fed to animals during the lean period of the year. Sorghum produces excellent forage, the forage quality viz. digestibility is on-par with maize. Based on the number of cuts two types of forage sorghums are grown, single cut and multi-cut. Sorghum is also found better for silage and pellet making which can be stored for a long time and also it improves digestibility. The sorghum dry stalk after harvesting panicles is stacked and is fed to animals. It is the most efficient use of dual type of sorghum which is grown for grains and dry fodder (Aruna and Cheruku, 2019).

North-Eastern India is a non-conventional area for forage sorghum. Rice is predominantly grown in the region. The other vegetation in the region is less amenable for dairy and meat industry. The milk and animal products demands are linear to the fodder demand in the region. The year-round fodder demand can be met by cultivating a crop like sorghum which is high yielding, drought tolerant, wide adaptability and also photo-insensitivity types available (Wolabu and Tadege, 2016). Looking into the prospects of forage sorghum in North-Eastern (NE) India, breeding program was initiated. The limited availability of genetically diverse resources tailored for forage sorghum development has hindered progress in developing varieties and hybrids. No high-yielding forage sorghum varieties or hybrids have been tested for suitability in NE India, including Assam (Bora et al., 2020). The development of stable, high-yielding genotypes and hybrids are necessary to meet the fodder demand in these regions. To start with breeding material from ICAR- Indian Institute of Millets Research (ICAR-IIMR) was used. The breeding material included male sterile lines and restorer lines. This marks the initiation of forage sorghum breeding in NE India. The success of sorghum hybrid breeding relies on the complementary interaction of the selected parental lines (A, B and R). These lines play a crucial role in the early stages of sorghum hybrid breeding programs, serving as the foundation for developing improved and high-performing cultivars and hybrids. Understanding the stability and adaptability of each parent line allows breeders to optimize combinations that exhibit heterosis, leading to hybrids with superior performance compared to their parents (Kannababu et al., 2017).The sorghum B-lines were specifically bred for fodder purposes, and their adaptability to a wide range of environments will allow breeders to effectively exploit these selected lines in the development of forage sorghum hybrids. The future objectives aim for the implementation of shuttle breeding with stable lines on the Indian subcontinent, where forage sorghum is currently in its infancy. Dominance and overdominance variance influence hybrid performance. It is more important to select stable and high-yielding lines as parents for crosses than to use the whole set of A-lines and test the hybrids that are made in multi-environmental trials (METs).

Forage yield is a complex trait that is influenced by many factors, including the genotype of the plant, the environment in which it is grown, and management practices. Genotype-by-environment interaction (GEI) can make it difficult to identify superior genotypes and can lead to reduced genetic gain in breeding programs. This results in reduced genetic gain and slows down the breeding process (Quintero et al., 2018). Prior to introduction of the crop in non-conventional areas it must be tested for its performance and stability. Also, it is important to prioritise multi-environment trials (METs) in order to examine the stability and pattern of GEI across environments for effective crop improvement programmes (Yan and Kang, 2002; Vaezi et al., 2019). It is crucial to evaluate the performance of selected genotypes based on their adaptability to the local environment. The stability of traits and high heritability contribute to the selection of breeding material for improvement. Stability is one of the important criteria in assessing suitability of genotypes to a region. Several analytical models and techniques, including analysis of variance (ANOVA), regression analysis viz. Finlay and Wilkinson (1963) and Eberhart and Russel (1966); non-parametric methods like, principal component analysis (PCA), additive main effects and multiplicative interaction (AMMI), and genotype and genotype plus environment (GGE) bi-plots, were developed in order to understand the unpredictable effects of genotype, environment, and their interaction (Kendal, 2016; Singamsetti et al., 2023). In light of India’s complex climate, breeding climate-resistant and region-specific high-yielding hybrids has emerged as a top priority for sorghum breeders. The increase in yield may be attributed to the selection of grain yield along with favourable expression of secondary traits. Identifying high-performing genotypes or treatments across multiple traits has been a difficult task. Optimising genotype selection based on grain yield and other relevant agronomic traits could enhance efficiency. Incorporating multiple trait data sets without encountering multi-collinearity in the selection process has posed a challenge for breeders. Therefore multi trait based stability evaluation methods are played crucial role in precise selection of genotypes. The Smith-Hazel index, a commonly used selection index for multiple traits, is not recommended for use in METs due to biased index coefficients and multi-collinearity issues (Smith, 1936; Hazel, 1943; Singamsetti et al., 2023). Breeders have suggested several improved novel multi-trait based stability evaluation methods, including genotype - ideotype distance index (MGIDI), multi-trait stability index (MTSI), multi-trait stability and mean performance (MTMPS) and multi-trait index based on factor analysis and genotype-ideotype distance (FAI-BLUP Index), to address these issues (Rocha et al., 2018; Olivoto et al., 2019b; Zuffo et al., 2020; Olivoto and Nardino, 2021). These methods are devised to aid breeders in maintaining sustainable progress in primary traits, such as forage yield, while preserving genetic advancements in secondary traits. These genotypes enable plant breeders to suggest stable and high-yielding, reliable forage genotypes for specific regions.

In the present study, breeding material consisted of B- lines and R lines were evaluated for their performance and stability. The data was analysed for variability, heritability, inter-trait associations, and the GEI effect. Considering the four multi-trait models, a comparative study is also made to find the most suitable multi-trait stability model and stable genotypes.

In this present investigation, the performance of 95 Sorghum forage genotypes were evaluated for 14 agronomic traits in two consecutive planting years (2020–2021) under the climatic conditions of Jorhat of Assam and Indian institute of millet research (IIMR) of Rajendra Nagar, Hyderabad. To our knowledge, this is the first report of employing WAASB and multi – trait stability model in understanding Genotype × Environment interactions and stability analysis in Sorghum forage hybrids. Developing high-yielding fodder hybrids is one strategy that could be used to address the massive global shortage of green and dry fodder. A significant hindrance to the progress of forage sorghum hybrid creation has been the scarcity of female parents (male-sterile), maintainer, and restorer lines that have been particularly designated or customized for the purpose of forage sorghum hybrid development. Identifying promising parents and subsequently producing a commercial hybrid from them is a formidable undertaking (Patel et al., 2019). The evaluations of stability and adaptability of these lines are in accordance with the principles of sustainable agriculture. They aim to foster the growth of crop varieties that are less susceptible to failure, contribute to long-term environmental and economic sustainability, and assist in tackling the difficulties posed by fluctuating and evolving environmental conditions (Rao et al., 2011). The work is primarily focused on minimizing the duration of breeding, specifically addressing the environment, lowering resource usage, and supplying the appropriate genotypes for hybrid development.

Genotype-by-environment interactions (GEI) can significantly impede selection efficiency and hinder the development of adapted varieties, particularly for complex traits like forage yield (Rao et al., 2011). Accuracy of prediction models aid in interpretation of multi-environment trials (Gauch and Zobel, 1988). GEI studies with linear mixed models are more efficient in predicting genotypic responses compared to fixed effect models like AMMI family models (van Eeuwijk et al., 2016). Mixed models can be used to estimate various biometrical parameters, including broad-sense variance, heritability estimates, and genetic correlation among traits (Olivoto et al., 2019a). Olivoto et al. (2019b) proposes to use mean performance and stability of multiple desirable agronomic traits to improve varietal recommendations. The ideotype is postulated to possess a maximum WAASBY score of 100 for all the considered traits. The genotype with the lowest multi trait based selection index is selected as it is closer to the ideotype. In this present investigation, simultaneous selection for forage yield and early maturity has been carried out through four important multi trait based stability evaluation methods including MGIDI, MTSI, MTMPS and FAI – BLUP and a comparative study in the selection of ideal genotypes had done. Among the four methods, MGIDI had given more weightage in this current study.

Evaluating the agronomic performance of germplasm is crucial for selecting stably performing genotypes across locations with desirable traits. Such selected genotypes can be constituted to from a core-subset for trait improvement. In the present study lines with high value for multiple traits were identified as superior viz. G44 (309B), G69 (352B), G17 (CSV33MF), G45 (373B), G90 (424B) and G63 (442B) showed desirable mean performance for multiple traits along with high forage yield. The genotypes, G63, G38, G62, G64, G32, G75 and G44 were the desirable genotypes with early maturity with high yield and these genotypes could be included in development of high yielding varieties breeding programmes.

The results of pooled ANOVA showed that there is significant genotype, environment, and genotype x environment interactions effects for most of the traits suggested that the genotypes were suitable for estimating the GEI and stability parameters. Similarly, Patel et al. (2019) found significant differences in genotypes and G x E for days to flowering, plant height, number of leaves per plant, stem diameter, and yield of green and dry fodder per plant in both the individual and pooled environments. Likewise, these kinds of results were also reported by Chala et al. (2019) and Nagesh Kumar et al. (2021).

The BLUP model, which incorporates genotypic and interaction effects (GEI) as random, is superior to other models in predicting random effects and genotype mean values with greater precision. Genetic variability is crucial for enhancing agronomic traits in plant breeding selection procedures. Heritability knowledge aids plant breeders in selecting an appropriate method for enhancing a trait, estimating the selection’s benefit, and assessing the genetic effects’ significance. The study found statistically significant genotypic and GEI effects for all recorded traits (p ≤ 0.001). This showed that genotype mean performances differed among growing conditions, which could increase genotype diversity and provide sufficient variation for easy selection (Yue et al., 2022). So, a higher heritability of a trait indicates a higher effectiveness of selection with high selection accuracy. The high selection accuracy of these traits illustrated the model’s reliability in the selection of better genotypes. The low values of R² _ge were observed in almost all the traits indicating considerable residual variance in the G x E interaction component, unlike the AMMI ANOVA, which explained most of it via the first two IPCAs and the GEI variance contribute less towards the phenotypic variance component (Olivoto et al., 2019a; Yue et al., 2022).

The high genotypic correlation among environments (r_ge) was recorded in most of the traits indicated a consistent trend across environments and easier identification of stable and superior genotypes while it was low for other remaining traits, suggesting that selecting stable and superior genotypes for these attributes is challenging and requires detailed, reliable information. BLUP based genetic parameters were comparatively smaller than mean based genetic parameters, with the same trend. Similar results were reported by Sousa et al. (2019) in cowpea for immature seed production, Koundinya et al. (2021) in 25 cassava genotypes and Yue et al. (2022) in 28 maize genotypes. The four factors (FA1 to FA4) with eigenvalues greater than one accounted for a significant proportion of the total variation. Therefore, it was possible to maintain strong explanatory strength while reducing the dimensionality of data. The maximum value of communality indicating that a significant portion of the variance of each trait might be explained by these factors (FA). All the studied traits were grouped in 4 factors based on their communality value. Similar kinds of results were also reported in barley for identification of salt-tolerant barley genotypes using multiple-traits index and yield performance at the early growth and maturity stages by Pour-Aboughadareh et al. (2021); Yue et al. (2022) in maize hybrids; Olivoto et al. (2019b) in oat; Olivoto and Nardino (2021) in wheat and Debsharma et al. (2023) in rice.

Yield is a complex trait that arises from the interplay of various yield-contributing factors that exhibit either positive or negative correlations with yield as well as with each other. The correlation coefficient is useful in identifying beneficial indicators of high yield by evaluating the relative influence of various characters on yield and among themselves during selection. Characters that had notable positive or negative correlations with green fodder yield were prioritised during the selection process for achieving a high green forage yield per plant. Arvinth et al. (2021) also reported similar results, where Green fodder yield per plant have positive and significant association with number of leaves per plant, stem girth, leaf length of blade, leaf width of blade and dry fodder yield per plant. These results were also supported by several breeders (Santhiya et al., 2021; Thant et al., 2021; Patil et al., 2022 and Chauhan and Pandey, 2021).

Selected genotypes were used to calculate selection differentials. Genotype G15, with an MGIDI score of 4.44, exhibited the cut point at the final red circle, accounting for selection intensity. Genotype G90, located in close proximity to the circle, may possess significant traits. Therefore, further investigation is required for the genotypes at the cut point. Positive selection differences were noted in adaptive traits except for leaf-to-stem ratio, stem girth, and days to 50% flowering. This suggests that these selected genotypes were more stable, as per the WAASBY index for these traits. A negative selection differential (%) and genetic gain under selection (%) were advantageous due to the lower desired value of the trait, such as days to 50% flowering. Genotypes selected for dry forage yield per plant, inter-nodal length, green forage yield per plant, and plant height based on the WAASBY index exhibited high heritability and genetic gain, indicating their stability and desirability for these traits. The aforementioned traits could be improved by directly selecting and incorporating these genotypes into the breeding programme. Olivoto and Nardino (2021) reported similar results in wheat, while Yue et al. (2022) found comparable outcomes in maize hybrids.

The strength and weakness of traits could be judged by the factor analysis viz., the contribution of each component to the MGIDI. Traits that are closer to the ideotype tend to have lower proportions, explained by the factor, which is located near the exterior edge of the figure. All genotypes, except G45 and G38, exhibited weak contribution and stability and were closer to the ideotype for various plant characteristics such as number of leaves, leaf width, leaf area index, stem girth, dry forage yield, and green forage yield. Genotypes G69, G66, G33, and G45 exhibited greater stability in terms of plant height, leaf length, number of nodes per plant, and inter-nodal length. G31 and G14 exhibited greater stability with respect to panicle length. All genotypes exhibited stability and proximity to the ideotype for days to 50% flowering. So, strength and weakness plot is a potential graphical tool to identify and select genotype based on their strength and weakness for the respective traits can be improved. Similar kinds of results were reported by Olivoto et al. (2021) in strawberry, Benakanahalli et al. (2021) in Guar, Mamun et al. (2022) in EMS induced rice mutant and Debsharma et al. (2023) in rice.

The application of a multi-trait-based stability method would enable the selection of genotypes based on their stability across multiple traits and mean performance. The method relies on genotype-ideotype distance and factor analysis scores. The ideotype exhibits the maximum WAASBY values across all the observed traits (Olivoto et al., 2019b). Multi-trait-based stability methods’ lower values indicate stable genotypes across multiple traits.

The genotypes G36 and G89 were consistently observed in all models and were considered highly desirable and reliable. Consequently, these genotypes were incorporated into the ideotype-based breeding programmes. The two genotypes exhibited ideal mean performances based on MGIDI selection and stability as per the MTSI selection criteria. Genotype G90 was found to be close to the cut point in the MGIDI and was detected in the models MTSI and MTMPS. On the other hand, genotype G89 was detected in all models and exhibited a high green forage yield per plant. These genotypes were considered ideal for both forage yield and ideotype design in the ideotype breeding pipeline. Yue et al. (2022) reported comparable findings in Maize hybrids, while Olivoto and Nardino (2021) and Benakanahalli et al. (2021) observed similar outcomes in Wheat and Guar, respectively. When MET data is used for genetic analysis and genomic selection, noise is produced and phenotypic values are misinterpreted, leading to wrong inferences. Multi-trait stability analysis could speed up the process of curation of phenotypic data by identifying stable genotypes in a varied panel across contexts. The correlation between the stability of parents and hybrids has not been demonstrated. However, quality parameters are more susceptible to being impacted by multi-environment factors compared to parameters relating to forage production.

The current study based on four techniques to identify genotypes that consistently perform well across many traits. may introduce bias towards specific traits or combinations of traits. This could lead to the exclusion of potentially valuable genotypes that perform well in certain environments or for specific traits not emphasized in the selection process. Since the current study was with limited number of environments, a better picture could have been possible with the use of more diverse environmental conditions. A long-term assessment of genotype performance and stability over multiple years may provide more robust insights into their suitability for climate-resilient fodder sorghum production. The performance of the genotypes is determined by both the direct measurement of the characteristic and the influence of associated traits. The influence of the associated characteristics on the economic characteristic helps in choosing genotypes in populations that remain consistent across different environments, which also serves as an indirect indicator of heritability across different locations. Further research is warranted to validate the effectiveness of the identified genotypes, especially under field conditions and across a wider range of environments. Additionally, exploring additional traits relevant to climate resilience and abiotic stress tolerance could enhance breeding efforts for developing climate-resilient forage sorghum genotypes.

The use of the BLUP based multi –trait stability techniques in multi-environment trials facilitated an accurate depiction of genotype-environment interaction (GEI). Four distinct methods for evaluating genotype ranking, namely MGIDI, MTSI, MTMPS, and FAI-BLUP, have been employed. The multi-trait-based stability evaluation methods facilitated the selection of stable genotypes, with positive selection differentials for traits that wanted to increase and negative selection differentials for one trait that wanted to decrease. This means that breeders and agronomists that aim at simultaneous selection for mean performance and stability while considering various attributes might find these strategies effective. It provides a unique selection process, including weights for each trait that are easy to interpret, and considers the correlation structure among the traits, which will make the index more applicable and provide better treatment recommendations. The study aims to enhance specific traits in specific environments. Overall, this study opens the door to the use of these tools to analyse plant multivariate data, standing out as a powerful tool to develop better recommendation strategies in sorghum forage. The results showed that genotypes G36 (302B) and G89 (348B) were found to be common across all four evaluation methods based on all the studied traits. The aforementioned genotypes were found to be the most reliable, high-yielding, and early-maturing and could be suggested for variety development and ideotype breeding programmes to ensure the food and nutritional security of an ever increasing population.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

PB, RN and AS designed the research and performed the experiments. BB, AS and VR contributed experiment materials to perform field experiment and provide critical support during laboratory based analysis. RS, JB and PM provide support during data analysis. PB, HV, LG, NB and NS wrote the manuscript with help from all the other authors. All authors contributed to the article and approved the submitted version.