Introduction

Today, the human population is confronted with two major paradoxes, the persistence of hunger and the narrowing of genetic diversity, inviting immediate attention to ensure a world without hunger. Monoculture and the use of farming practices that focus on high yields through various inputs are detrimental not only to human health but also to the environment and biodiversity. The massive issue of feeding approximately 9.7 billion people by 20501 requires an astonishing increase in food production and an unprecedented degree of strain on a limited range of crops. In search of an affordable solution to these challenges underutilized/potential crops including buckwheat have come to light as these crops are nutritionally rich, resilient to environmental stresses, and provide various health benefits, making them suitable candidates for sustainable agriculture and diversifying our diets.

Buckwheat has gained immense attraction in recent years due to its high nutritional value, environmental sustainability and adaptability to diverse agro-climatic conditions. Buckwheat is a pseudocereal that belongs to the family Polygonaceae and genus Fagopyrum. Although many species of buckwheat exist, only nine species are recognized as being economically important2 with the two most widely cultivated being common buckwheat (F. esculentum) and tartary buckwheat (F. tartaricum). Wild species are found mainly in China, India, Nepal, Bhutan, Pakistan and other South Asian countries. The origin of common and tartary buckwheat is considered to be southwestern China near eastern Tibet3,4. This area is also considered as the distribution and diversity center of buckwheat. From here it spread to southern and northern China, Korea, Japan and Central Asia5. During the 9th–12th century, it diffused to Russia and in the 13th–15th century it further spread to northern and central Europe6. Today, it is grown in Russia, China, Ukraine, France, Kazakhstan, Poland, Japan, Korea, India, Nepal, Bhutan and some other countries. During the year 2023, buckwheat was grown on 2.18 million ha area with a production of 2.2 million tons and a productivity of 1007.5 kg/ha7. As a pseudocereal, buckwheat is rich in proteins, fiber and many other essential amino acids8, making it a critical food source, especially where traditional cereal crops may not thrive well. Being gluten free9, it has becomes a crucial food source for people who have gluten allergy. One of the key advantages of buckwheat is that it requires fewer inputs such as fertilizers10, pesticides11 and water compared to other cereal crops, making it a promising candidate for areas facing climate changes and for promoting sustainable agriculture. The rich genetic diversity of buckwheat can be leveraged to boost biodiversity and enhance the quality of other crop species12.

As the demand for nutritional security under a changing global scenario rises, understanding the growth behavior and adaptation of buckwheat under varying environmental conditions becomes imperative for increasing its production and utilization. Genotype × environment (G × E) analysis is a crucial aspect of understanding the performance and adaptability of genotypes. G × E interaction studies helps in identifying the genotypes that perform consistently well across the diverse environmental conditions, enabling the development of cultivars that are both high yielding and resilient to climatic variations13.

Additive Main Effects and Multiplicative Interaction (AMMI) and Genotype plus Genotype-by-Environment Interaction (GGE) biplots are statistical tools used to analyze G × E interactions. AMMI combines the overall performance of each genotype and each environment along with how each genotype responds to different environments to provide sound understanding of how genotypes perform across various environments14. It combines analysis of variance with principal component analysis to separate main effects from interaction effects, providing insights into the performance stability of genotypes15. On the other hand, GGE biplots are a graphical tool used to analyze the performance and stability of genotypes across multiple environments16 and understanding environmental influences.

This study provides significant scientific value by identifying high yielding and stable buckwheat genotypes suitable across diverse environments. The results underscore the critical role of genotype × environment interactions in directing selection for productivity in underutilized crops. This research enhances the usefulness of DUS testing by linking trait stability with environmental response and encourages incorporation of advanced statistical analytics into buckwheat improvement programs. This study was conducted to evaluate the stability and adaptability of 20 buckwheat genotypes across two distinct environments namely, Shimla and Sangla in Himachal Pradesh, India over three consecutive years. These locations having unique agro-climatic conditions provide valuable data on how buckwheat genotypes respond to specific environmental factors. The findings from these trials could not only help local farmers to identify the high yielding varieties for their region but also provide valuable information for improving global buckwheat production. This research is especially significant for regions with increasing buckwheat cultivation in Asia, Europe and North America where climatic conditions are identical.

Materials and methods

Plant material

The genetic material used in the experiment comprised of two species namely common buckwheat (Fagopyrum esculentum) and tartary buckwheat (Fagopyrum tartaricum). A total of twenty genotypes of buckwheat were used out of which thirteen belonged to tartary buckwheat and seven to common buckwheat. The experimental material comprised of released varieties, indigenous and exogenous collections obtained from Indian National Gene Bank. The genetic material along with their genotypic code, source and species is shown in Table 1.

Table 1 List of genotypes along with code, source and species.
Full size table

Environments

The crop was grown consecutively for 3 years from 2022 to 2024 at two distinct experimental locations i.e., Shimla (31° 05’ 56” N, 77° 09’ 35” E) and Sangla (31° 25′ 56″ N, 78° 15’ 4” E). The experiment was conducted in the research farm of ICAR-National Bureau of Plant Genetic Resources, Regional Station, Shimla. At Sangla, it was organized at Chaudhary Sarwan Kumar Himachal Pradesh Krishi Vishvavidyalaya, Mountain Agriculture Research and Extension Centre which is situated at a height of 2680 m AMSL. The soil in Shimla is clay loam, slightly acidic to neutral pH, with relatively high organic carbon content. The soil in Sangla is sandy loam with pH ranging from 5.5 to 6. Organic carbon content is relatively less than Shimla. Nitrogen and phosphorus content in Sangla is comparatively low than Shimla. The environments (a combination of seasons and locations) showed considerable degree of variation in various climatic parameters. Table 2 depicts the different environment along with the environmental codes.

Table 2 List of environments along with environment code.
Full size table

Experimental design and observations

The experiment was conducted in Randomized Complete Block Design (RCBD) with three replications in each environment. Plant to plant spacing was maintained at 20 cm, row to row distance was 45 cm and row length was 2 m. Recommended agronomic practices were followed for raising the experimental crop. Ploughing was done to prepare the land for sowing and organic manure was added. Hand weeding was performed at regular intervals. The application of insecticides and fungicides was conducted based on necessity. In Shimla, the crop was sown on 03/06/2022, 16/06/2023 and 07/06/2024. In Sangla, it was sown on 17/05/2022, 03/06/2023 and 14/06/2024. Harvesting was done by handpicking mature seeds.

For this study, twelve quantitative characteristics were recorded as described by Mahajan et al.17. The quantitative traits investigated included days to 50% flowering, leaf blade length (cm), leaf blade width (cm), petiole length (cm), number of primary branches, number of inflorescences per plant, inflorescence cyme length (cm), plant height (cm), days to 80% maturity, number of seeds per inflorescence, seed yield per plant (g) and 1000 seed weight (g).

Statistical analyses

The recorded data was subjected to rigorous statistical analysis using ‘R’18, package ‘metan’ (Multi environmental trial analysis)19. An individual analysis of variance (ANOVA) was done to assess each environment while a pooled ANOVA was performed across all environments to access the significance of genotypes, environment and G × E interactions. Subsequently, AMMI analysis20 was conducted to partition the G × E interaction into principal component analysis. The AMMI model for G ×E interaction21 is expressed as:

$${\text{Y}}_{{{\text{ij}}}} = {\text{ }}\mu {\text{ }} + {\text{ g}}_{{\text{i}}} + {\text{ e}}_{{\text{j}}} + {\text{ }}\Sigma \lambda _{{\text{n}}} \alpha _{{{\text{in}}}} \gamma _{{{\text{jn}}}} + {\text{ }}\theta _{{{\text{ij}}}}$$

Where, Yij is the mean yield of ith genotype in the jth environment, µ is the general mean, gi is the ith genotypic effect, ej is the jth location effect, λn is the eigen value of the Principal Component Axis n, αin and γjn are the ith genotype, jth environment principal component analysis (PCA) scores for the PCA axis n, θij is the residual, n is the number of PCA axis retained in the model. Also, AMMI stability indexes were computed to identify the stable genotypes across the environments. AMMI stability Index (ASI) quantifies the result based on first two PCs22 and is calculated as:

$$ASI\:=\sqrt{\left[{\left({\varvec{P}\varvec{C}1}_{\varvec{s}\varvec{c}\varvec{o}\varvec{r}\varvec{e}}\right)}^{2}\times\:{\left({\varvec{P}\varvec{C}1}_{\varvec{\%}\:\varvec{e}\varvec{x}\varvec{p}\varvec{l}\varvec{a}\varvec{i}\varvec{n}\varvec{e}\varvec{d}\varvec{m}\varvec{e}\varvec{a}\varvec{n}\varvec{s}\varvec{u}\varvec{m}\varvec{o}\varvec{f}\varvec{s}\varvec{q}\varvec{u}\varvec{a}\varvec{r}\varvec{e}}\right)}^{2}\right]+\left[{\left({\varvec{P}\varvec{C}2}_{\varvec{s}\varvec{c}\varvec{o}\varvec{r}\varvec{e}}\right)}^{2}\times\:{\left({\varvec{P}\varvec{C}2}_{\varvec{\%}\:\varvec{e}\varvec{x}\varvec{p}\varvec{l}\varvec{a}\varvec{i}\varvec{n}\varvec{e}\varvec{d}\varvec{m}\varvec{e}\varvec{a}\varvec{n}\varvec{s}\varvec{u}\varvec{m}\varvec{o}\varvec{f}\varvec{s}\varvec{q}\varvec{u}\varvec{a}\varvec{r}\varvec{e}}\right)}^{2}\right]}$$

Where,\(\:{PC1}_{score}\)is first principal component score of interaction effect, \(\:{PC1}_{\%\:explainedmeansumofsquare}\) is percentage sum of squares explained by first principal component interaction effect, \(\:{PC2}_{score}\) is second principal component score of interaction effect, \(\:{PC2}_{\%\:explainedmeansumofsquare}\) is percentage sum of squares explained by second principal component interaction effect. AMMI stability value (ASV) is a measure to quantify and rank genotypes in terms of yield stability23. It is calculated as:

$$ASV\:\sqrt{[\frac{{PC1}_{sumofsquares}}{{PC2}_{sumofsquares}}{\left(PC1\right)}^{2}+{\left(PC2\right)}^{2}}]$$

The weighted average of absolute scores (WAAS) is calculated by taking into account each interaction Principal Components from the singular value decomposition of the G × E interaction impact matrix by a linear mixed-effect model24, as follows:

$$WAAS_{i} = \frac{{\sum\nolimits_{{k = 1}}^{p} {\left| {PC_{{ik}} \times EP_{k} } \right|} }}{{\sum\nolimits_{{k = 1}}^{p} {EP_{k} } }}$$

where \(\:WAA{S}_{i}\) is the weighted average of absolute scores of the ith genotype, \(\:P{C}_{ik}\) is the score of the ith genotype in the kth interaction Principal Component, and \(\:E{P}_{k}\) is the explained variance of the kth PC for k = 1,2,.,p, considering p = min(g − 1;e − 1). The AMMI and GGE biplots are graphical representations that illustrate the G × E interaction and genotype ranking according to stability and mean. GGE biplot analysis25 was performed for seed yield per plant to visualize the interaction patterns and identify the superior genotypes. This helped in determining the best performing genotypes in specific environments and provided insights into the discriminative ability and representativeness of testing environments.

Results

Individual and pooled ANOVA

In all the six environments, ANOVA across each environment (Table 3) indicated significant variation among the genotypes for all the characters. Pooled ANOVA across the environments (Table 4) demonstrated that the environments and G × E interaction and nested G × E interaction were highly significant for all the characters while genotype was found significant for all the characters except petiole length and number of primary branches. This indicated that sufficient variation was found in the environment. However, characters showing higher coefficient of variation indicated greater influence of environmental factors suggesting sufficient genetic diversity for selection in breeding programs.

Table 3 Individual analysis of variance across each environment.
Full size table
Table 4 Pooled analysis of variance.
Full size table

Pooled ANOVA using AMMI model

AMMI analysis provided a deeper insight into genotypic stability. Pooled analysis using the AMMI model (Table 5) showed that mean sum of squares due to environment, genotype and G × E interaction was significant for all the characters under consideration. G × E interaction was further decomposed into five principal components (PCs). PC1, PC2 and PC3 were found to be significant for all the characters and accounted for a substantial proportion of the total variation due to G × E interaction, explaining over 80% of the variation in all the characters. For seed yield per plant, genotype explained 42% of the total variation, whereas G × E interaction and environment explained 35.1% and 22.2% respectively. The first two PCs for this character explained more than 70% variation due to G × E interaction. Residuals accounted for the remaining unexplained variation which was found low suggesting most of the variations were explained by the model.

Table 5 Pooled analysis using AMMI model.
Full size table

AMMI based stability indexes

The AMMI model provided stability rankings for twenty genotypes for seed yield per plant based on three stability indexes (Table 6): AMMI Stability Index (ASI), AMMI Stability Value (ASV) and Weighted Average of Absolute Scores (WAAS). The high yielding genotype, IC 109,729 (3.80 g per plant) was ranked 1st in WAAS and 2nd in ASI and ASY for stability in seed yield per plant. Best stability scores was found in Sangla B-129 which was ranked 1st in ASI and ASV, and 3rd in WAAS while also having a competitive yield (2.86 g per plant). Entry IC 202,226 was also found stable ranking 3rd in ASI and ASV, and 2nd in WAAS despite having lower yield. This table also indicated the stable characters for each genotype. Seed yield per plant was found stable in Sangla B-129, IC 109,729 and IC 202,226.Days to 80% maturity was stable in Shimla B-2, IC 14,889, IC 17,371 and IC 26,594 whereas days to 50% flowering were found stable in Sangla B-118, Himpriya, PRB 1 and IC 202,226. Two genotypes i.e. IC 26,594 and IC 258,233 were found stable for maximum number of characters (four). Sangla B-129 and IC 412,722 were found stable for three characters.

Table 6 AMMI based stability indexes.
Full size table

AMMI biplots for seed yield per plant

AMMI I biplot (Fig. 1) revealed that Shimla B-1, Sangla B-118, Sangla B-129 and IC 109,729 had a higher yield than the mean and had positive interaction PC score whereas Shimla B-3, Sangla B-1, Sangla B-5, Sangla B-214, Himpriya, IC 26,594, IC 258,233 had negative PC scores with high yield. Shimla B-2, VL-7, PRB-1, IC 17,371, IC 202,226, IC 274,426, IC 412,722, EC 323,730 recorded lower yield than mean and had positive PC score while IC 14,889 had lower yield and negative PC score. Sangla B-129 and IC 258,233 had interaction PC scores near to zero and high mean indicating that they are stable. Sangla, 2022, Shimla, 2023, Sangla, 2023 and Sangla, 2024 showed high seed yield per plant, meanwhile Shimla, 2022 and Shimla, 2024 showed low seed yield per plant. Genotype EC 323,730 showed specific adaptation to Shimla, 2024 while Himpriya showed specific adaptation to Shimla, 2023. AMMI II biplot (Fig. 2) revealed that Sangla, 2022, Sangla, 2023 and Sangla, 2024 exerted less interaction forces than the rest. Sangla B-129, IC 109,729, IC 202,226, Shimla B-2 and IC 258,233 were closer to the origin and therefore, were less sensitive to the environment. Genotype IC 14,889 was found suitable for Sangla, 2024 and Sangla, 2022 whereas Sangla B-214 was found suitable for Sangla, 2023 and Shimla, 2023. Sangla B-129 was also suitable for Shimla, 2023. IC 109,729 and VL-7 were found suitable for Shimla, 2024. IC 202,226, Shimla B-2, IC 17,371 and IC 258,233 were found suitable for Shimla, 2022. The environments can be further partitioned into four mega environment groups. Environments Sangla, 2022, Sangla, 2024 and Sangla, 2023 form one group while the remaining three environments can be classified into separate groups. Environments showing positive correlation with each other were Sangla, 2022, Sangla, 2024, Sangla, 2023, and Shimla, 2023. Environments showing no correlations were Shimla, 2024 with Sangla, 2024, Sangla, 2022 and Shimla, 2022. Negative correlation was shown by Shimla, 2022 with Shimla, 2023, Sangla, 2023, Sangla, 2022, and Sangla, 2024. Meanwhile Shimla, 2024 showed negative correlation with Sangla, 2023 and Shimla, 2023.

Fig. 1
Fig. 1
Full size image

AMMI 1 biplot for seed yield per plant (g). The key to genotype and environment code is given in Tables 1 and 2, respectively.

Fig. 2
Fig. 2
Full size image

AMMI 2 biplot for seed yield per plant (g).

GGE biplots for seed yield per plant

Which-won-where view of the GGE biplot

The biplot in Fig. 3 identified the best performing genotypes in specific environments by forming a polygon. The genotypes present on the vertices had the longest vectors and are among the most responsive genotypes. Sangla B-118 performed best in Shimla, 2022. Himpriya excelled in Shimla, 2024. Genotype IC 109,729 outperformed other genotypes in Sangla, 2022, Sangla, 2023 and Sangla, 2024.

Fig. 3
Fig. 3
Full size image

Which-won-where view of the GGE biplot for seed yield per plant (g).

Discriminativeness vs. representativeness

The biplot (Fig. 4) helped in analyzing the ability of each environment to differentiate the genotypes. Additionally, it also indicated how well an environment represents the overall testing conditions. Shimla, 2022, Sangla, 2023 and Sangla, 2024 had longer vectors and thus can effectively distinguish among the genotypes. Whereas, Shimla, 2024 recorded shorter vector length and exhibited less discriminativeness.

Fig. 4
Fig. 4
Full size image

Discriminativeness vs. representativeness for seed yield per plant (g).

Genotype–environment interaction biplot

Figure 5 depicted the interaction between genotypes and environments, with PC1 explaining 62.86% of the variation and PC2 explaining 14.79% accounting for a total of 77.65% of the variation in seed yield per plant. Sangla B-129, IC 109,729, Sangla B-214 and IC 258,233 were closer to the origin and reported more stable seed yield per plant across environments. Sangla B-118 has been assessed to be closely associated with Shimla, 2022, showing specific adaptation. The spread of environment vectors suggested varying growing conditions, with Shimla, 2022 being distinct from other environments.

Fig. 5
Fig. 5
Full size image

Genotype by genotype–environment interaction biplot for seed yield per plant (g).

Mean vs. stability

In this biplot (Fig. 6) horizontal axis represents mean performance while vertical axis represents stability. Genotypes IC 109,729, Sangla B-118, Sangla B-5, Sangla B-1 and IC 26,594 positioned further to the right exhibited high mean seed yield per plant. Genotypes that are closer to the horizontal axis viz., IC 109,729 and Sangla B-129 are more stable.

Fig. 6
Fig. 6
Full size image

Mean vs. stability for seed yield per plant (g).

Relationship among environments

Figure 7 represents the biplot that assessed the correlation between different environments. Sangla, 2022, Sangla, 2023, Sangla, 2024 and Shimla, 2024 grouped closely together suggested that environmental conditions are identical, whereas Shimla, 2022 is separate from others indicating unique environmental conditions. Shimla, 2022, Sangla, 2023 and Sangla, 2024 have the longest vectors, indicating a strong ability to differentiate genotypes. Sangla, 2022, Shimla, 2023, Sangla, 2023, Shimla, 2024 and Sangla, 2024 are all showing acute angles with each other and thus these environments had positive correlation with each other. Meanwhile, Shimla, 2022 is showing an obtuse angle with Shimla, 2024 thus indicating negative correlation.

Fig. 7
Fig. 7
Full size image

Relationship among environments for seed yield per plant (g).

Ranking environments

In this biplot (Fig. 8) the ideal environment is represented at the center of concentric lines, which acts as a measure to assess how far each environment is from the ideal one. Shimla, 2023 and Sangla, 2023 are the closest to the ideal environment, and therefore, are most desirable of all environments. Whereas, Shimla, 2022 and Shimla, 2024 being the farthest from the ideal center are least desirable test environments.

Fig. 8
Fig. 8
Full size image

Ranking environments for seed yield per plant (g).

Ranking genotypes

Figure 9 helps in ranking the genotypes based on their adaption across the environments. It compared all the genotypes with the ideal genotype defined as having the highest yield in all environments. IC 109,729 and Sangla B-1, being positioned closer to the ideal genotype, were found to be outstanding as compared to other genotypes with high mean yield and stability.

Fig. 9
Fig. 9
Full size image

Ranking genotypes for seed yield per plant (g).

Discussion

The significant G × E interaction observed underscores the complexity of genotype evaluation in diverse environments. The findings from this study confirm the significant G × E interactions on the yield and stability of buckwheat genotypes. Pooled ANOVA revealed that there was considerable variation present in the environment and genotypes. The significant effects of environment on performance of genotypes highlight the importance of multi-environmental trials to identify the stable and high yielding buckwheat genotypes. Similar findings have also been reported by previous studies26,27,28 accentuating the role of environmental effects.

AMMI analysis further provided deeper insight into G × E interactions through partitioning the G × E interactions into principal components. It revealed that PC1 captured the majority of interaction variations. IC 109,729 and Sangla B-129 expressed remarkable stability with lower interaction scores. Similar results have been reported by previous studies29,30,31. Liu et al.32 used the AMMI model to analyze the buckwheat genotypes and reported that genotypes with lower interaction scores were most stable. Furthermore, AMMI biplots helps in understanding the genotype stability and adaptability across different environments33. Genotypes such as IC 109,729, Sangla B-118 and Shimla B-1 having high yield with moderate PC1 scores could be promising candidates for breeding and selection. GGE biplots collectively provide a comprehensive understanding of genotypic performance, stability and G × E interactions34. IC 109,729 and Sangla B-129 exhibited broad adaptability while others performed well in specific environment like Sangla B-118 in Shimla during the year 2022. Overall, IC109729 and Sangla B-129 were identified as most stable and high yielding genotypes. They demonstrated low sensitivity to environmental changes, making them suitable for large scale adoption. These findings align with previous studies on stability analysis of pseudocereals like quinoa35 and grain amaranth36 emphasized the importance of multi-environmental trials.

Environments like Shimla, 2022, Sangla, 2023 and Sangla, 2024 exerted strong interaction forces on the genotypes, whereas Shimla, 2023 and Sangla, 2023 can be considered closer to the ideal environment. Thus, Shimla, 2023 and Sangla, 2023 were found to be discriminative and representative environments that can effectively differentiate genotypic performance. Similar observations have been reported in quinoa37 and grain amaranth38,39. These findings reinforce the significance of G × E interactions in genotype selection and breeding. In case of locations, Sangla outperformed Shimla in terms of seed yield per plant (Fig. 10). The influence of altitude, temperature and precipitation likely contributed to the variation40. Variability is slightly lower in Sangla location indicating more consistent performance of buckwheat genotypes. The ability to pinpoint the suitable testing environment enhances the efficiency of selection, ensuring that only most stable genotypes get promoted for further testing. Based on these findings, future research should focus on testing genotypes under abiotic stress conditions such as drought, flood, heat and cold. Modern breeding approaches such as marker assisted selection41, genome editing via CRISPR/Cas9 technology42 and genome-wide allele frequency fingerprinting (GWAFF)43 can be utilized to complement the phenotypic data. Additionally, expanding the research to more diverse agro-climatic conditions would confirm the stability of genotypes across an even broader range of environments for recommending to commercial farming.

Fig. 10
Fig. 10
Full size image

Average of seed yield per plant (g) over three years at two locations.

Conclusions

The AMMI and GGE biplot analysis of twenty buckwheat genotypes across two locations over three years provided crucial insights into how performance and stability are influenced by genetic, environment and G × E interaction effects. Among the tested genotypes, IC 109,729 and Sangla B-129 exhibited relatively higher yield and stability, making them suitable candidates for large scale adoption, selection and breeding. The implications of this study go beyond merely identifying the superior genotypes. Moreover, Shimla, 2023 and Sangla, 2023 were identified as highly discriminative and representative environments making them ideal for future testing. These findings highlight the importance of conducting multi-environmental trials in breeding programs to ensure the development of buckwheat varieties that are resistant to climate change. By identifying the genotypes that perform well across locations and years, breeders can make informed decisions to develop varieties that are productive and stable under various environments. This is even more significant in the current context of climate change where stability is becoming as important as the yield. This study further lays the foundation for the development of improved buckwheat cultivars with superior agronomic and nutritional characteristics adapted to the similar environments around the globe.