Introduction
Proso millet (Panicum miliaceum L.), is one of the minor millets grown in India. It originated in Manchuria (Northern China) before 10,000 BC and used to be an important cereal crop in China and Eurasia prior to the introduction of wheat and barley in these regions (Rajasekaran et al., Reference Rajasekaran, Francis, Mani, Ganesan, Farooq and Kadambot Siddique2023). It is grown in Arabia, China, India, Egypt, Japan and Western Europe (Vetriventhan et al., Reference Vetriventhan, Azevedo, Upadhyaya and Naresh2019). The cultivated forms of P. miliaceum are divided into five subspecies, miliaceum, patentissium, contractum, compactum and ovatum, based on the size and compactness of the inflorescence (Vetriventhan and Upadhyaya, Reference Vetriventhan and Upadhyaya2018). In India, it is cultivated in Bihar, Tamil Nadu, Madhya Pradesh, Maharashtra, Eastern Uttar Pradesh, Andhra Pradesh and Karnataka. In India, proso millet is cultivated in about 0.041 million ha with a production of 0.022 million tonnes and yield 531 kg/ha as compared to 0.77 million ha area, 1.19 million tonnes production and 1535 kg/ha yield across the world according to the estimates of 2015–16 (Bhat et al., Reference Bhat, Tonapi, Rao, Singode, Santra and Johnson2018). Proso millet is one of the most preferred crops after finger millet in the Indo-Gangetic plains, attributed to its high and easily digestible protein content. Therefore, it is important to expand the area of proso millet grown in the states Bihar, Jharkhand, Uttar Pradesh and Bengal.
Proso millet is a self-pollinated, allotetraploid with the chromosome number 2 n = 4x = 36 and a genome size of 923 Mb (Zou et al., Reference Zou, Li, Miki, Li, Tang, Xiao, Rajput, Deng, Peng, Jia and Huang2019). Proso millet is preferred by growers as it is a short duration crop, maturing in only 60–90 days, has extremely low water requirement and high-water use efficiency. Due to its early maturity and low water requirement, it evades drought and extends better anticipation for rigorous cultivation in dryland areas (Rajasekaran et al., Reference Rajasekaran, Francis, Mani, Ganesan, Farooq and Kadambot Siddique2023). It is generally grown during the kharif season under rainfed conditions. It exhibits special traits, like early maturity, that make it a promising crop for crop rotation. Proso millet exhibits a reduction in the height of its main stem, as temperatures surpass 30°C, enabling enhanced resilience against drought. Being a C4 plant with a low transpiration ratio, it has efficient carbon fixation even in challenging conditions such as limited carbon dioxide (CO2) and nitrogen availability, high temperatures and drought (Habiyaremye et al., Reference Habiyaremye, Matanguihan, Guedes, Ganjyal, GM, Kidwell and Murphy2017). It therefore becomes a crop of importance that has potential for cultivation as a summer catch crop in the Indo-Gangetic plains.
Proso millet is a highly nutritious grain with a low glycaemic-index, a rich source of minerals (zinc, iron and calcium), vitamins (vitamin B-complex, niacin and folic acid), starch, essential amino acids (cysteine and methionine) and phenolic compounds such as beta-glucan and antioxidants (Amadou et al., Reference Amadou, Gounga and Le2013). Despite all the benefits, proso millet is a neglected crop. As breeding strategies are not well developed in this strictly self-pollinated crop, there is less availability of commercial high yielding varieties; therefore, farmers are still using the long-preserved landraces. Crop improvement methods focus on yield performance of the breeding lines and selected germ plasm in multi-environment and yield stability. However, the yield performance and stability are dependent on the genotype by environmental interactions (GEI). Improving yield stability across environments will imperatively enhance the grain yield. Cluster analysis, principal component analysis (PCA) and biplots are multivariate tools for effectively determining GEI. Genotype plus Genotype by environment (GGE) biplots are widely used to determine the GEI patterns, access genotype rankings and identify stable genotypes across environments which will determine their adaptability in different environments (Khan et al., Reference Khan, Rafii, Ramlee, Jusoh and Mamun2021).
The aims of the study were to identify which proso millet varieties are suitable for growing in the summer period (February–May) and also those which are the most stable and high yielding in both summer and rainy seasons. To achieve this, the research focused on evaluating the GEI interactions of local landraces and proso millet germ plasm collected from various ecologies of Bihar, Jharkhand and Uttar Pradesh states lying in the Indo-Gangetic plains of India. The genotypes were evaluated for their adaptation, stability and performance by measuring yield-related traits, in both summer and kharif (rainy season).
Materials and methods
Genetic material and experimental site
The genetic material evaluated were 60 proso millet genotypes (Supplementary Table 1) which are landraces and local germ plasm collected from various districts of eastern Uttar Pradesh, Bihar and eastern and central parts of Jharkhand, all of which lies in the Indo-Gangetic plains (Supplementary Fig. 1). Proso millet has been a preferred crop of this region since ancient times along with finger millet, and it finds mention in some of the ancient Indian texts and manuscripts. The farmers have preserved the indigenously grown proso millet landraces, which are a source of their food and livelihood. The genotypes were planted in augmented block design having 6 blocks with 10 genotypes per block and 3 checks (national checks were TNAU-164, TNAU-202 and the local check was BR-7) randomised in each block. They were planted with spacing of row to row = 25 cm, plant to plant = 10 cm and block to block = 1 m. The study was conducted at the small millets’ research farm of Dr. Rajendra Prasad Central Agricultural University, Pusa, Bihar, India, during the two summer seasons (summer 2022, summer 2023) and one kharif season (kharif 2022). The comparative weather data for past 5 years in the region are given in Supplementary Figure 1. Proper agricultural operations like hand weeding (2 weeding during summer as well as kharif), irrigation (2 irrigations during summer and no irrigation during rainy season) and fertiliser applications (N:P:K::40:20:20).
Which-won-where biplot showing best performing genotype in environment for (A) grain yield, (B) panicle length, (C) number of productive tillers and (D) 1000 grain weight. The polygon is constructed by joining the farthest genotypes. The perpendicular lines separate the polygon into sectors. The genotypes at the vertex are the best performing genotype in the environment falling in that sector.
Phenotypic observations
Quantitative traits recorded were days to 50% flowering (DTF), days to maturity (DTM), panicle length (PL), fodder yield per hectare (FY/ha), grain yield per hectare (GY/ha), plant height (PH), number of basal tillers per plant, number of productive tillers per plant (NPT), 1000 grain weight (TGW), flag leaf blade length, flag leaf blade width and peduncle length. Qualitative traits recorded were plant growth habit, plant pigmentation at leaf sheath, inflorescence shape, culm branching, panicle compactness, lodging, seed shattering, grain colour and grain shape. Five randomly selected plants from each row, representing a specific genotype were chosen for taking the observations.
Statistical analysis by analysis of variance and GGE biplot
The method given by Federer (Reference Federer1956) and expanded by Federer and Raghavarao (Reference Federer and Raghavarao1975) was used in this study for calculating analysis of variance for augmented designs under heterogenous conditions. The total variance that could be assigned to different sources was partitioned as mentioned in Supplementary Table 2.
The effectiveness of GGE biplot have demonstrated by many researchers as a tool for graphically addressing problems related to genotype (G) evaluation in multi-environment trials (Yan and Kang, Reference Yan and Kang2003; Yan and Tinker, Reference Yan and Tinker2005). In this method, the most stable environment is placed close to the dot of ideal and average environment and in the concentric area of ideal environment dot. In case of GGE biplot for varieties, the desirable varieties are placed near to ideal variety and in the concentric variety of ideal variety dot. Grain yield, TGW, PL and NPT, collected across all the environments, were subjected to GGE biplot construction as per the model given by Yan et al. (Reference Yan, Hunt, Sheng and Szlavnics2000), and Yan and Kang (Reference Yan and Kang2003), based on singular value partitioning (SVP) of the first two principal components as follows:
Yij = μ + άi + βj + Фij + ∑ij
where Yij is performance (Y) of ith genotype in the jth environment, μ = grand mean, άi = genotype row main effect, βj = environment main effect, Фij = individual GEI and ∑ij = error.
All the variables in this study focus on environment centred data model and the data subjected to SVP. The resultant principal components 1 and 2 from the SVP were used to construct the GGE biplot, and the other smaller PCA was regarded as residual and not explained.
Results
Mean performance of genotypes for different traits across pooled environment
The average performance of 60 proso millet genotypes along with three checks over three environments showed significant variation for grain and stover yield, productive tillers, PL, DTF, DTM, PH and TGW (Table 1 and Supplementary table 3). Disease and pest incidence was negligible in proso millet, some shoot fly incidence was observed at the vegetative stage but it did not damage the crop and the crop overcame the incidence sturdily.
Analysis of variance for quantitative traits in proso millet across three environments
* ns P > 0.05; P ≤ 0.05; **P ≤ 0.01.
Note: DTF = days to 50% flowering. DTM = days to maturity, PL = panicle length (cm), FY/ha = fodder yield per hectare (kg/ha), GY/ha = grain yield per hectare (kg/ha), PH = plant height (cm), NBT = number of basal tillers per plant, NPT = number of productive tillers per plant, TGW = 1000 grain weight (g), FLL = flag leaf blade length (cm), FLW = flag leaf blade width (cm), PDL = peduncle length (cm).
Correlation coefficient analysis for quantitative traits
The traits FY/ha, PH and TGW were significant and positively correlated with GY/ha. DTF and DTM were significant and negatively correlated with GY/ha (Table 2).
Phenotypic correlation coefficient among different quantitative traits based on the mean data of three seasons
* Significant at 5%, **Significant at 1%, NS = non-significant.
Stability of proso millet genotypes based on GGE biplot
The GGE biplot illustrating the best performing genotypes for grain yield, PL and NPT is shown in Fig. 1. The best performing genotypes for environmental stability are shown in Figs. 2 and 3.
GGE biplot showing mean vs stability of 63 proso millet genotypes over three environments for (A) grain yield, (B) panicle length, (C) number of productive tillers and (D) 1000 grain weight. The average environmental axis (AEA) is shown by dotted blue line, AEA abscissa passes through the biplot origin, depicts the average environment and points towards higher mean values. The AEA ordinate which is represented by perpendicular lines to the AEC passing through the biplot origin depicts the stability of the genotypes. The concentric circles include the best performing genotypes.
GGE biplot showing relationship between environments of 63 proso millet genotypes for (A) grain yield, (B) panicle length, (C) number of productive tillers and (D) 1000 grain weight. The average environment axis is shown by red dotted line, the abscissa of which points towards the ideal environment shown by blue dot and the average environment is shown by the red circle. The red lines represent the environmentvectors.
Grain yield
The polygon view of GGE biplot evaluated the performance of 63 genotypes in three different seasons (Fig. 1A) shows that the genotypes were grouped into three sectors (I, II and III). All three seasons were located in sector III. V20, V7, V26, V28, V45, C1, C2 and C3 were located on the vertices of the polygon. Of these, C1(BR-7), C2(TNAU-164), C3(TNAU-202) and V45(RAU PM-45) had the highest yield and so were the best performers in all the three seasons. There were outstanding performers in other sectors such as V20, V7, V26 in sector I and V28 in sector II, but no environment was included in these groups. Some other genotypes fell into these groups, but these genotypes could not be recommended for any of the seasons taken into consideration. Moreover, 99% (PC1 = 97.4%, PC2 = 1.6%) of the total GGE variation revealed the model’s ability for predicting the stability of proso millet genotypes across the environments. The best performing genotypes present inside the concentric circles (Fig. 2A) were BR-7, TNAU-202, TNAU-164 and other genotypes are far from them. The ranking stability to grain yield was BR-7 > TNAU-202 > TNAU-164 > V22 (RAU PM-22) > V15 (RAU PM-15) > V2 (RAU PM-2) > V57 (RAU PM-57) = V60 (RAU PM-60) > V54 (RAU PM-54) > V23 (RAU PM-23) > V10 (RAU PM-10) > V38 (RAU PM-38). Though no genotype outperformed the check varieties, RAU PM-22 exhibited a comparable performance to check varieties and hence this genotype along with the checks can be recommended for multilocation testing for summer seasons. Summer 2022 was located very close to the centre of concentric ring indicating that it is an ideal environment, followed by summer 2023 and kharif 2022 within the concentric ring has features similar to an ideal environment (Fig. 3A).
Panicle length
The best performing genotype in environment for PL are shown in Fig. 1B where the genotypes were grouped into eight sectors (I, II, III, IV, V, VI, VII and VIII). Season 1 and season 3 was located in sector V and season 2 was located in sector III. The genotypes located in the vertices of polygon performed better compared to others with respect to that particular environment belonging to the sector. V20, V27, V53, V11, V54, V39, V40, V32, V35, V33, V41, V10, V28, V26, V55 were located on the vertices of the polygon. Among them, V53 (RAU PM-53) was best performer in season 2 and in season 1 and season 3, V32 (RAU PM-32), V39 (RAU PM-39), V40 (RAU PM-40) were the best performers. There were outstanding performers in other sectors such as V20 in sector I, V27 in sector II, V11 and V54 in sector IV, V35 in sector VI, V41 and V33 in sector VII, V10, V28, V26, V55 in sector VIII, but no environment was included in these groups so these genotypes could not be recommended for any of the seasons taken into consideration. The total GGE variation was 84.4% (PC1 = 55.6%, PC2 = 28.8%). The best performing genotype is shown (Fig. 2B) in the concentric circle and the graph showed V54 (RAU PM-54) with maximum value for PL and have V39 (RAU PM-39), V11 (RAU PM-11), V13 (RAU PM-13), V52 (RAU PM-52), V40 (RAU PM-40) and other genotypes located in a circular pattern away from the centre. The ranking of genotypes for PL was RAU PM-54 > RAU PM-39 > RAU PM-11 > RAU PM-13 > RAU PM-52 > RAU PM-40. The evaluation for environment for PL clearly depicts that the summer 2022 was located within the concentric ring indicating that it has features similar to an ideal environment (Fig. 3B).
Number of productive tillers per plant
The performance of 63 genotypes in three different seasons evaluated through GGE biplot (Fig. 1C) showed that the genotypes were grouped into 5 sectors (I, II, III, IV, V). All three seasons were located in sector III. The genotypes located in the vertices of polygon performed better compared to others with respect to that particular environment belonging to the sector. V1, V24, V6, V4, V17, V35, V45, V23 genotypes were located on the vertices of the polygon. Among them, V17 (RAU PM-17) and V35 (RAU PM-35) had the highest number of tillers in all the three seasons. The total GGE variation was 95.3% (PC1 = 81.5%, PC2 = 13.8%) The best performing genotype was RAU PM-17 with maximum NPT (Fig. 2C). In terms of stability to productive tillers per plant, the ranking of genotypes was: V17 (RAU PM-17) > V35 (RAU PM-35) > C3 (TNAU-202) > V27 (RAU PM-27) > V46 (RAU PM-46) > V51 (RAU PM-51) > V49 (RAU PM-49) > V10 (RAU PM-10). In all seasons, RAU PM-17 and RAU PM-35 outperformed the check varieties and hence these genotypes can be used for breeding programmes in future. The evaluation for environment for productive tillers per plant depicts that summer 2022 was located within the concentric ring indicating that it has features similar to an ideal environment (Fig. 3C).
1000 grain weight
GGE biplot for TGW (Fig. 1D) shows that the genotypes were grouped into 4 sectors (I, II, III, IV). All three seasons were located in sector II. In the study, V24, V1, V22, V17, V37, V41 genotypes were located on the vertices of the polygon, among them V17 (RAU PM-17) won in all the three seasons. There were outstanding performers in other sectors such as V24 and V1 in sector I, V37 in sector III and V41 in sector IV, but no environment was included in these groups. The generated biplot in the study revealed. The total GGE variation was 92.4% (PC1 = 78.5%, PC2 = 13.9%). The best performing genotype with highest value for TGW was V17 (RAU PM-17) (Fig. 2D). The ranking of genotypes can be done as V35 (RAU PM-35) > V15 (RAU PM-15) > V31 (RAU PM-31) > V57 (RAU PM-57) > V17 (RAU PM-17) > V4 (RAU PM-4) > V27 (RAU PM-27) > V10 (RAU PM-10) > V22 (RAU PM-22) in terms of stability to TGW. V35 and V17 showed stable performance along with high TGW in all three seasons and hence this genotype can be used for selection for TGW. The evaluation for environment for TGW clearly depicts that the summer 2022 and summer 2023 were located within the concentric ring indicating that it has features similar to an ideal environment (Fig. 3D).
Discussion
The aims of our study were to identify the highest yielding proso millet genotypes that had greatest stability in both summer and kharif (rainy) seasons. In India, proso millet is generally grown only as a kharif season crop and all released varieties are bred with this intention. However, our study demonstrated that proso millet can be introduced as a potential summer crop not only because it yields better in summer as compared to kharif but also because it is short duration crop with minimum irrigation requirement and can easily fit into the Indian cropping system. It can easily be grown as a summer catch crop as it can be sown after the harvest of the rabi (winter) crop and can be harvested before the kharif (rainy) crop. In this way, the growers will reap a remunerative crop in the cropping gap period and will not have to leave their field fallow. Summer 2022 and 2023 were the most ideal mega-environments for proso millet crop, and further multilocation trials can be constituted for summer season in the Indo-Gangetic plains (so proso millet can be encouraged as a summer season crop).
BR7, TNAU 202 and TNAU 164 had the widest adaptability as they were stable across the three environments summer as well as kharif and their yield were comparable across the environments. These released varieties can be directly released for summer season after multilocation trials in the Indo-Gangetic plains. RAU PM 17, RAU PM 22, RAU PM 35 and RAU PM 54 were the landraces identified, which showed specific adaptation/stability and high yield performance in summer so they can be utilised for further breeding programmes like varietal development by using them in crossing programmes as parents to develop high yielding varieties for summer or by selection to improve their yield in summer, and they can also be included in the multilocation trials constituted for summer season.
The mean performance of yield and yield attributing traits was comparable in summer and kharif (rainy) seasons, which implies that there is no yield compromise and proso millet can perform equally well when grown in summer season. Correlation studies showed that yield, PH and grain weight had significant positive correlation with grain yield while DTF and DTM had significant negative correlation with grain yield in all environments, which follows a similar trend as observed in major cereal crops. It implies that early flowering and maturity in proso millet in drier and warmer summer season will lead to higher grain yield which can be attributed to the fact that the plant will utilise the energy in grain filling which would have otherwise been expended in maintaining the vegetative growth. However, Mishra et al. (Reference Mishra, Tiwari, Sapre, Nema, Sahu, Tantwai and Tiwari2022) reported that the DTF in barnyard millet was positively correlated with grain yield/plant; barnyard millet has greater PH than proso millet, so later the flowering in barnyard will lead to greater biomass accumulation which in-turn would have translated into grain yield. PH normally shows a negative correlation to grain yield in C3 crops; however, in this study, PH shows a positive correlation with grain yield. This can be explained as, since proso millet is a C4 crop, it can fix CO2 more efficiently into photosynthates compared to C3 plants so more the biomass, more will be the C-fixation that can be efficiently partitioned to the sink.
As summer season has been identified as the ideal environment for proso millet cultivation in the Indo-Gangetic plains, further studies for multilocation testing of the identified varieties and the landraces need to be conducted in the different agro-ecologies of the region. The identified varieties and landraces can be utilised for further breeding programme like varietal development by using them in crossing programmes as parents to develop high yielding varieties for summer or by selection to improve their yield in summer. Stability studies also need to be done pan-India to identify the locations where proso millet can be introduced as a summer crop and consequently develop varieties specific to summer season for the identified ecology. The outcome of the study can be utilised by the All India Coordinated Research Project in the crop improvement programmes and development of millet-based cropping system.
Conclusion
The stability of 63 proso millet genotypes was evaluated in three seasons in the Indo-Gangetic plains. RAU PM- 17, RAU PM-22, RAU PM-35, RAU PM-54, BR-7, TNAU-202 and TNAU-164 showed higher grain yields and other yield attributing characters in summer 2022 and summer 2023 and were found to be most stable and wining genotypes for summer season as compared to other genotypes. Summer was found to be the most stable and ideal environment for proso millet performance based on yield, TGW, NPT and PL. Proso millet can be effectively introduced as a summer crop in this region for better performance, enhancing production, fitting the existing cropping system, summer adaptability, pre-breeding programmes and breeding for high yielding proso millet cultivars.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1479262125100221.
Acknowledgements
The authors acknowledge Dr. Rajendra Prasad Central Agricultural University, Pusa, Bihar, India, for providing facilities and funding to conduct this research work.
Competing interests
The authors declare none.
