Sorghum [Sorghum bicolor (L.) Moench] ranks fifth among the most important cereal crops in the world, after maize, rice, wheat, and barley (faostat, 2020). In 2020, global sorghum production was 29.8 million metric tons (MMT) with 1.5 tonnes per hectare (t ha⁻¹) of average productivity (Faostat, 2020). It is a food security crop for more than half a billion people in developing countries, mainly in arid and semi-arid regions where moisture stress is a major constraint (Ejeta, 2005). Sorghum is a multipurpose crop, which is used for food, feed, and fuel (Waniska et al., 2016; Stamenković et al., 2020). Due to its gluten-free nature, high starch and protein levels, and high content of condensed health-beneficial compounds, its grain has high nutritional value (Waniska et al., 2016; Dykes, 2019).

Despite its importance, the productivity of sorghum is highly limited by many factors but drought remains the major abiotic constraint. About 80% of sorghum production in the world is under dryland conditions (Assefa et al., 2010) and in Sub-Saharan Africa, sorghum is mainly cultivated in drought-prone areas that cover nearly 60% of the total area (Hadebe et al., 2017). In Ethiopia, 66% of the areas where sorghum is predominantly cultivated are prone to frequent drought (Geremew et al., 2004). Several studies have reported the impact of drought stress on sorghum. Drought stress affects the growth and development of sorghum, which ultimately leads to a substantial reduction in grain yield (Bobade et al., 2019; Queiroz et al., 2019; Abreha et al., 2022). For instance, Assefa et al. (2010) reported that drought stress at the vegetative and reproductive stages reduced the sorghum yield by more than 36% and 55%, respectively.

The northeastern part of Africa, most likely modern Ethiopia and Sudan, is considered to be the center of origin and domestication of cultivated sorghum (Doggett, 1970; Dillon et al., 2007). The presence of high genetic diversity in the Ethiopian sorghum gene pool has been reported by several authors (Adugna et al., 2013; Adugna, 2014; Enyew et al., 2022). The original sources of the most widely used stay-green genotypes B35 and SC56 are Ethiopia and Sudan, respectively. The likelihood of the diverse sorghum gene pool that exists in this region being a potential source of more novel drought tolerance and stay-green genotypes is high. In line with this, Disasa et al. (2016) reported the presence of high genetic diversity in the Ethiopian sweet sorghum germplasm that has not yet been utilized in sorghum breeding programs. Crop landraces and wild relatives are potential sources of desirable genes for crop improvement (Kyratzis et al., 2019; Ochieng et al., 2020). In this regard, the Ethiopian sorghum gene pool could serve as a potential source of resistant/tolerant genotypes against biotic and abiotic stresses.

Multiple drought-tolerant genotypes, such as 00MN7645, QL41, B35, BTx642, SC-56, and E-36-1 have been identified and used in sorghum breeding programs (Tao et al., 2000; Xu et al., 2000; Kebede et al., 2001; Haussmann et al., 2002; Sukumaran et al., 2016). However, it is still necessary to find additional sources of drought and stay-green genotypes to avoid dependence on only a few sources, which is the case in the current sorghum breeding programs across the world. Drought tolerance is a complex trait, controlled by many genes. In addition, various environmental factors affect drought severity, making sorghum breeding for drought tolerance challenging (Tuinstra et al., 1996; Abreha et al., 2022). Nevertheless, selecting drought-tolerant sorghum genotypes is crucial for improving the production of sorghum through the application of modern breeding methods.

Selecting sorghum landraces that have drought tolerance-related traits such as stay-green and high chlorophyll content is a crucial step in sorghum breeding programs. This is because improving such traits, which have moderate to high heritability (Harris et al., 2007; Mutava et al., 2011; Ochieng et al., 2020) could lead to an increased level of drought tolerance of the crop (Harris et al., 2007; Mutava et al., 2011). Drought stress often causes a reduction in chlorophyll content of sorghum plants and promote leaf senescence (Thomas and Ougham, 2014; Hou et al., 2021), which results in reduced grain yields (Borrell et al., 2000; Djanaguiraman et al., 2020). High chlorophyll content has also been associated with improved stay-green in sorghum and reduces post-flowering drought-induced senescence (Harris et al., 2007). Similarly, the growth of panicle exertion and peduncle length were lower in the sorghum plants grown under stress conditions as compared to the control (Maluk, 2018). Furthermore, the panicle exertion was reported to be affected by water stress occurring after floral initiation in sorghum and rice (Tsuda, 1986; Praba et al., 2009). The panicle exertion was also reported to increase harvest index and improve yield in wheat (Liang et al., 2013). Therefore, the use of drought related traits such as stay-green, chlorophyll content, number of leaf per plant, leaf area, panicle exertion and peduncle length as selection criteria to identify drought tolerant genotypes could lead to increased grain yield in sorghum in drought prone areas. However, not much known about the genotype by environment interactions and the stability of drought tolerance related traits in sorghum

The analyses of genotype stability across different environments have been carried out by using different statistical models as reviewed by Pour-Aboughadareh et al. (2022). Additive main effects and multiplicative interaction (AMMI) and genotype main effects plus genotype by environment interaction (GGE) biplot models are the most effective and commonly used models for the analyses of stability, adaptability and ranking of genotypes and for selecting suitable mega environments. In AMMI, genotype type stability is determined through the analysis of AMMI stability value (ASV) (Purchase et al., 2000), sums of the absolute value of the IPCA scores (SIPC), averages of the squared eigenvector values (EV) (Sneller et al., 1997), weighted average of absolute scores (WAAS) (Olivoto et al., 2019), and absolute value of the relative contribution of IPCAs to the interaction (ZA) (Zali et al., 2012). Low values of these parameters suggest high stability of the genotypes across environments. Most sorghum production areas in Ethiopia are in arid or semi-arid regions with high rainfall variability and low soil water storage. Consequently, the total annual rainfall and its distribution throughout the sorghum-growing season play a critical role in sorghum grain yield. The availability of sufficient moisture in the soil at certain growth stages is particularly crucial for obtaining a good grain yield (Assefa et al., 2010). Evaluation of moisture availability during the growing season of sorghum (and other crops) in Ethiopia for a long period led to Melkassa, Mehoni, and Mieso being designated moisture stress sites (dry lowland) suitable for testing sorghum’s tolerance to moisture stress. Even though these sites receive annual rainfall that may appear to be sufficient for growing crops without much moisture stress, they are considered moisture-stressed due to factors such as inefficient distribution of rain during the crop-growing season, a relatively low moisture-holding capacity of the soils, and occasional high temperatures. Accordingly, these sites have been used as suitable testing sites for moisture stress-related research on sorghum (Wagaw et al., 2020; Wondimu et al., 2020; Teressa et al., 2021) including for the present study.

The objectives of the present study were to determine the drought stress response and stability of the Ethiopian sorghum landraces through the determination of genotype by environment interactions of different traits. This was done through the application of AMMI and GGE biplot models to identify novel sources of drought tolerance for use in sorghum breeding programs.

This study revealed that there exists substantial phenotypic variation within the Ethiopian sorghum gene pool for the drought-tolerance related traits studied across three locations. The observed phenotypic variation in this study and previous studies on Ethiopian sorghum (Girma et al., 2019; Enyew et al., 2021) suggests that tolerant genotypes to abiotic stresses can be obtained through screening its diverse gene pool. The observed high broad-sense heritability for most of the traits analyzed suggests a major role of genetics in the phenotypic variation of the traits, which could be used for improving drought-tolerance related traits through selection. In line with this study, a high broad-sense heritability for drought-tolerance related traits, such as chlorophyll content and stay-green were reported in previous studies in sorghum (Kapanigowda et al., 2013; Naoura et al., 2019; Ochieng et al., 2020). However, low heritability of chlorophyll content in this crop has also been reported (Badran (2020). This difference may be due to the variation in the genotype, environment and genotype by environment interaction effects.

The significant positive correlation observed among the chlorophyll content and grain yield is in agreement with previous studies in sorghum (Ochieng et al., 2020; Abreha et al., 2022) and other crops (Kamal et al., 2019). Similarly, chlorophyll content showed a highly significant positive correlation with stay-green (SG). This is due to the ability to maintain high chlorophyll content during drought-stress conditions, thereby delaying the process of leaf senescence. Xu et al. (2000) reported the correlation of leaf chlorophyll content and SG in sorghum during post-flowering drought suggesting the use of these traits for screening drought tolerance and yield in sorghum. Sorghum genotypes that maintain high chlorophyll content under drought stress are more likely to produce higher grain yields than non-SG types. This indicates that the stay-green genotypes identified in the current study are highly valuable genetic resources for breeding sorghum for drought tolerance. The highly significant positive correlation observed between GLNF and grain yield in this and previous reports on sorghum (Reddy et al., 2014) indicate that GLNF, being easy to measure, could serve as one of the criteria for selecting genotypes that can be used in sorghum breeding programs that aim at developing drought tolerant cultivars. PAE showed significant positive correlation with CHLF, CHLM and SG. Similarly, Maluk (2018) reported that PAE positive association with stay-green trait. This may be due to the genotypes which have stay-green trait may maintain high leaf water potential which allows the plant to maintain the extension of the PAE and PDL under drought stress conditions.

The present study confirmed that genotype, environment and G×E interaction have significant effects on the drought-tolerance related traits (CHLM, CHLF, SG, GLNF, FLA, PE and PDL). This highlights the significance of multi-environment field trials for genotype evaluations aimed at the identification of stable genotypes with a high mean performance for the studied traits. The phenotypic variation due to genotype was high for SG, CHLM, CHLF and GLNF, indicating that genotypes played a larger role in phenotypic variation than environments. The effects of environment and G × E interaction were higher for FLA, PE and PDL, indicating that the effects of genotypes on the phenotypic variation in these traits were lower.

High genetic variation exists in the Ethiopian sorghum gene pool (Amare et al., 2015; Girma et al., 2019; Enyew et al., 2021), which has been used as a source for resistance to biotic (Wu et al., 2006) and abiotic traits such as stay-green (Borrell et al., 2000). Genotype B35, derived from a cross between two Ethiopian durra sorghum genotypes (BT×642), is among the well-known drought-tolerant genotypes and has been used as a major source of stay-green genes in sorghum breeding programs (Evans et al., 2013; Abreha et al., 2022). Whereas, Melkam, Argiti, and ESH4 are drought tolerant and high-yielding improved varieties being widely cultivated in Ethiopia. In the present study, genotypes G48, G65, and G66 for CHLF, G59 for CHLM and G22 and G66 for SG had higher mean performance than B35 as well as the improved varieties (Melkam, Argiti, and ESH4). Interestingly, these genotypes with a high mean performance for CHLF, CHLM and SG had high mean grain yield and other farmer preferred traits, such as plant height as reported in our previous work (Enyew et al., 2021). Hence, this study suggests the potential to discover sorghum genotypes with a stronger drought-tolerance within the Ethiopian sorghum gene pool. Further study that includes these desirable genotypes is required to gain deeper insight into the genetic control mechanisms of these drought-tolerance related traits. These results further highlight the potential of germplasm maintained in gene banks in contributing new genetic sources required for crop improvement.

In AMMI analysis, low values of the stability parameters, ASV, EV, SIPC, WAAS and ZA indicate high stability of genotypes and low G×E interaction (Sneller et al., 1997; Purchase et al., 2000; Zali et al., 2012; Pour-Aboughadareh et al., 2022). In this study, G244 was identified as the most stable genotype across the three locations based on all AMMI stability parameters, followed by G222 and G201, which were identified based on their ASV, EV and SIPC values for chlorophyll content at the flowering stage. For chlorophyll content at the maturity stage, G114, G302, and G240 showed high stability according to their EV, SIPC, WAAS and ZA values while ASV suggests that G81, G53 and G255 were the most stable. Among the top three stable genotypes identified by all AMMI stability parameters, G53 and G114 had the highest CHLF value while G201 and G222 had the highest CHLM value. All other genotypes had CHLF and CHLM values lower than the corresponding mean values. However, stable genotypes could have low mean performance, and so they should not be the only factors to select germplasm for breeding. In the present study, the top ranking genotypes which had both high mean performance and stability were selected based on their GSI, which is calculated based on ASV and genotype mean ranking (Farshadfar, 2008; Mahmodi et al., 2011). Accordingly, G222, G53 and G239 for CHLF and G53, G308 and G54 for CHLM were identified as stable genotypes across environments with high mean performance. These genotypes were originally collected from Tigray and Oromia regions, suggesting the significance of these areas as potential sources for drought tolerant genotypes. Among the stable genotypes with high mean performance, G54 and G239 had high grain yield, as shown in our previous work (Enyew et al., 2021). Therefore, these genotypes could be prioritized for use in sorghum breeding programs to enhance the desirable characteristics of these drought-tolerance related traits with the ultimate goal of developing drought tolerant cultivars. This GSI method has been effectively applied to select the top ranking and stable genotypes in sorghum and other crops (Mohammadi et al., 2010; Nduwumuremyi et al., 2017; Donkor et al., 2020; Enyew et al., 2021).

The best performing genotypes could be selected through interpreting the G×E interaction, MGE clustering, and Which-Won-Where GGE biplot that show a particular adaptation of genotypes (Gauch, 1992; Yan et al., 2000; Yan and Tinker, 2006; Rakshit et al., 2012). In the GGE biplot, the vertex genotypes located away from the biplot origin are more reactive to environmental changes and considered as specifically adapted genotypes (Yan and Kang, 2002). These genotypes could be least or top performing in some or in all tested environments (Yan and Kang, 2002). According to the ‘Which-Won-Where’ GGE biplot, the testing environments were clustered into two MGEs in this study with different top performing genotypes for CHLF and CHLM. For both CHLF and CHLM, MGE1 was represented by MH and MK environments, and contained G48 and G65 for CHLF, and G316 and G59 for CHLM as top performing genotypes. Whereas, MGE2 comprised only environment MS where G117 and G175 for CHLF and G48 for CHLF had the top mean values. The results suggest the genotypes have different levels of adaptation to different MGEs, signifying the role of G×E interactions (Mohammadi et al., 2010). Several authors have reported the identification of top performing genotypes adapted to a specific MGE (Rakshit et al., 2012; Aruna et al., 2016; Yihunie and Gesesse, 2018; Vaezi et al., 2019).

GGE ranking biplot are commonly used to select top ranking and stable genotypes across environments through based on AEC (Yan, 2001). In this study, the AEC in ranking biplot showed genotypes G48, G66, G65 and G175 as the top ranking in CHLF and G59, G321, G48, G114 and G18 for CHLM. However, due to the G×E interaction effect, the top ranking genotypes such as G59 for CHLF and G59 and G48 for CHLM were less stable. One of the major applications of the ranking biplot is to detect genotypes with both high mean performance and high stability. In this study, GGE ranking biplot identified G48, G66 and G175 for CHLF and G118, G114 and G321 for CHLM as high performing stable genotypes. The identified stable sorghum genotypes that had the highest mean performance with respect to their chlorophyll content could be used as potential novel sources of drought tolerance for sorghum improvement. It indicates that selection of best genotypes through ranking biplot analysis is a valuable approach for identifying high performing stable genotypes as shown in previous studies (Aruna et al., 2016; Oliveira et al., 2020).

Phenotypic evaluation is vital for efficient selection of crop germplasm for conservation and use in plant breeding programs. In this regard, the landrace accessions remain a vital source of novel alleles for traits of interest. Ethiopian sorghum landrace accessions possess significant phenotypic variation for all drought-tolerance related traits (SG, CHLM, CHLF, GLNF, FLA, PE, and PD) targeted in the present study. This suggests the high possibility of identifying novel drought-tolerant sorghum genotypes within the Ethiopian sorghum gene pool, which consists of a large collection of sorghum germplasm in the national gene bank, which has not yet been evaluated for drought tolerance. High broad-sense heritability observed for the studied traits suggests a strong genetic control, which is a favorable opportunity for improving the traits through crossbreeding and selection. The analysis of variance confirmed that genotype, environment, and G×E interaction have significant effects on the studied traits. Hence, multi-environment field trials should be conducted for reliable identification of top performing and stable genotypes in terms of drought tolerance. Overall, G222, G53, G239, G48, G66, and G175 for CHLF, and G53, G308, G54, G118, G114, and G321 for CHLM were identified as stable genotypes with high mean performance, and could be prioritized for use in sorghum breeding programs. The landrace accessions identified as having a better mean performance than the improved varieties with respect to their chlorophyll content could be potential sources of novel alleles for drought tolerance. Hence, they should be further investigated through methods such as genome-wide association studies.

The original contributions presented in the study are included in the article/ Supplementary Material . Further inquiries can be directed to the corresponding author.

All authors conceived and designed the experiment. ME conducted the experiment, analyzed the data and wrote the draft manuscript. MG, TF, AC, CH, KT and AS reviewed the manuscript. All authors contributed to the article and approved the submitted version.