Introduction

Rice (Oryza sativa L.) is an important member of the family Poaceae, along with wheat (Triticum aestivum L.) and maize (Zea mays L.) and; it is among one of the three crops on which human beings depend for their dietary nutrition. Rice is currently the most significant crop and a staple food in east Asia and many other regions worldwide1,2. Rice is cultivated in over 100 countries and the area under rice cultivation is about 164 million hectares with a production of 510 million metric tons3. Rice consumption has been slightly increased in the world during the last couple of years as global intake of rice was around 502 million metric tons during 2020–21, whereas in 2008–2009 it was around 437 million metric tons4. Per capita consumption of rice is 45 kg per annum5,6.

Rice is the second most important staple food, and Pakistan earn almost 2.5 billion US$ every year7,8. The rice share in agriculture value addition is 1.9 percent and its contribution in the total GDP of Pakistan is 0.4 percent. During the year 2020–2021, the area under rice cultivation was 3.5 million hectares, and the total rice production was 9.3 million tons for both coarse and Basmati rice9. Whereas, during the year 2023 there was a decline in area of production by 15 percent and yield loses by 21 percent due to high input prices and flood damages10. Production of rice is expected to increase every year to ensure food security as the world’s population is increasing and the rice demand is also increasing11. During 2050, world population is expected to be more than 10 billion and more food would be required to meet the food security challenges Click or tap here to enter text12. Rice crop faces number of challenges that hinder its production i.e., flood, drought, diseases and insect’s infestation, heat stress during pollination, cost of production, and availability of quality seed13.

Pakistan imports almost five thousand tons of hybrid rice seed every year from China and other sources; and almost 30 thousand tons of hybrid seed is produced locally by private companies. Seed production of locally adopted open-pollinated varieties is very less in the country to meet the farmers’ demand14. Due to unavailability of improved varieties and quality seed, poor farmers use leftover seed repeatedly that is a major cause of low yield. Under these circumstances, the development of improved rice breeding materials and their subsequent testing over numbers of locations across the region / country becomes prerequisite to recognize rice genotypes having higher yield potential, stable performance, and stress resilience value.

Rice production enhancement on a sustainable basis by using Green Super Rice (GSR) concept was proposed for rice breeding and production for rapid accelerating the development in rice functional genomics. As a result of collective efforts of Scientists, many rice cultivars have been developed with enhanced multiple biotic & abiotic stress tolerance, water & nutrient used efficiency and now these cultivars have been released in many countries in Asia and Africa15. Green Super Rice has characteristics of higher nutrient uptake efficiency, judicious use of water, resistance to economically important diseases (BLB, BLS etc.), insect pests and abiotic stresses specially drought, higher yield potential and good end use quality16. Stable performance of cultivars in terms of paddy yield and adaptability over a broad range of ecologies is a pre-requisite. Green Super Rice germplasm was acquired from China and IRRI in Pakistan through Pakistan Agricultural Research Council’s coordination system and evaluated at country wide Multilocation trials during the year 2021 by Rice Program, Crop sciences Institute, National Agricultural Research Center, Islamabad, Pakistan for assessing stability and adaptability of best- performing genotypes over a broad range of agro-ecologies.

Stability and adaptability analysis are important concepts in the study of rice genetics and breeding. Stability analysis is used to identify rice varieties that perform consistently better across different environments or growing conditions. This is important because rice is grown in diverse agro-ecological zones and climatic conditions. Stability analysis is conducted using statistical methods like regression analysis, analysis of variance (ANOVA) and AMMI (additive main effects and multiplicative interaction) model. AMMI model has shown to be quite helpful in analyzing multi-location yield trials in rice17,18. Multilocation yield trials (MLYTs) are widely used to assess the yield performance of rice germplasm in multiple environments. In MLYTs, the genotype into environmental interaction (GEI) is a main source of variation, which can be due to both genetic and non-genetic factors. The model is a type of analysis of variance (ANOVA) that separates the sources of variation in the data into additive and multiplicative components. The additive component of variance is due to the presence of main effects of the genotypes and environments and multiplicative component represents the interaction of genotypes and environments. The genetic variance can be estimated from the additive component of the model. Another commonly used method is the genotype plus genotype-by-environment interaction (GGE) biplot analysis. It is a graphical tool that visualizes the patterns of GEI and identifies the genotypes that perform well across environments19,20.

There are various statistical tools and packages that are used to perform stability analysis and G x E interaction estimation. The METAN R package is a commonly used tool for analyzing data from multilocation yield trials (MLYTs)21. The package provides functions for analyzing genotype by environment interaction (GEI) and for estimating the stability and adaptability of genotypes across environments. The METAN R package uses a mixed-effects model to estimate the genetic and environmental effects on the performance of genotypes in MLYTs. The model accounts for both the mean and the variance of the genotypes across environments. The package also provides functions for visualizing the patterns of GEI and to identifying stable yield performing genotypes21. Several studies have been conducted in MLYTs in rice and data was subjected to METAN R package for stability analysis and GEI analysis22,23. Present study aimed to scale out suitable genotypes that have stable yield performance in a broad range of rice-growing ecologies in Pakistan; that have the potential to become a variety for general cultivation by farmers.

Material and methods

Rationale of the study

Chinese scientists launched the Green Super Rice (GSR) project in 200524 in response to the rising concerns about scarce resources, environmental pollution, and ecological destruction. The primary objective of research was to grow new rice strains with ecologically favorable traits such as insect and disease resistance, responsive to fertilizers effectively and drought tolerant25. Now, the GSR idea includes more than just creating new types with eco-friendly characteristics, which exemplifies the larger philosophy of crop breeding technology that emphasizes resource conservation and environmental friendliness. It also entails implementing risk-free, secure, effective, and high-yield crop management techniques15,16. The GSR project has been designed with five aims for GSR breeding1,2: creation of whole-genome selection platforms, combining environmentally favorable genes, involves developing novel germplasms with greater tolerance to biotic pressures as well as improved resistance to a variety of abiotic challenges, producing new sustainable GSR cultivars which includes hybrid and inbred types, increase grain yields and grain quality, these features are combined and lastly, the creation of field management strategies for higher production tailored for GSR. Therefore, considering these major focuses of GSR project, we designed this study with following objectives to have the tentatively available GSR genotypes for specific adaptability mechanisms in different ecological areas of Pakistan: 1) Evaluation and screening of drought-resistant and water-saving genotypes with better yield and quality under water deficit conditions; 2) Assessment of genotypes with better nutrient-use efficiencies with better growth, yield, and quality under reduced fertilizer application; 3) Evaluation of genotypes with better resistance to insects and pests under conditions of reduced application of pesticides; and 4) screening to the genotypes with more tolerance to abiotic stress such as salinity, heat, cold etc. with stable grain yield and quality.

Plant materials and study sites description

In the current study, screening of the adaptability mechanisms of eighteen GSR experimental genotypes with different growth durations comparing with two checks were conducted at twelve ecological locations across three provinces of Pakistan (Table 1; Fig. 1). The experiment was conducted using a randomized complete block design (RCBD) with three replications in 2021. Table 1 presents the ecological location, geography and environmental conditions which prevailed during the experimental season whereas Table 2 is represents the details of genotypes selected in this experiment with accession numbers, parentage, and source of germplasm. The rice genotypes and check cultivars selected for this experiment are mentioned in Table 2.

Table 1 Geographical details of locations for trials evaluation during the year 2021.
Full size table
Figure 1
figure 1

Map of experimental sites across Pakistan with legends indicating the locations of the Institutes where experiment was conducted. (ArcGIS-10.8 Software was used to generate the map1).

Full size image
Table 2 Genotypes names, their codes, gene bank-accession numbers, source of germplasm and source of germplasm.
Full size table

Agronomic and cultural practices

Considering the climatic conditions of each study site, during the first week of June, 2021, a 100-g pre-treated seed of GSR genotypes with two genotypes of checks was sown for nursery through dry method of nursery sowing on 2 ft2 plot size24. The plots were labelled with genotype names and codes. Depending on the prevailing climatic conditions, a 30 days-old nursery was shifted to paddy transplantation fields at the respective study location, where the transplantation was done manually. Transplantation of all selected genotypes was done between the 1st to 10th of July through a common straight-row technique comprising three replications at each study site. The plot size of the transplanted seedlings was maintained at 2.0 m × 1.0 m with five rows containing eight seedlings per row. The line-to-line and plant-to-plant spacing was maintained at 20 cm within the plot. To measure the variation in the genetic components, the grain yield and yield-related attributes were estimated at the physiological maturity stage during which five randomly selected seedlings were taken in each replicated plot. Plant height and panicle length measurements were carried out with standard designated methods26. The data for the number of productive tillers, grains per panicle, 1000-grain weight and grain yield was calculated with standard agronomic procedures27. To fulfil the crop nitrogen (N) requirements, synthetic Urea fertilizer was applied in three equal splits viz-10, 40, and 65 days of seedling transplantation. Whereas phosphorus (P) and potash (K) required dosages were applied in the form of diammonium phosphate (DAP), and muriate of potash (MOP) after 10 days of transplantation. All the N, P and K fertilizer management was done at the recommended dosage of 150, 90, and 60 kg/ha, respectively, which marginally varied based on the locality. Weeds were removed manually two times through hand-weeding as well as through recommended pesticides28.

Statistical tools

Analysis of variance

The morphological data for all the experimental genotypes at selected localities was subjected to the combined analysis of variance (ANOVA) through R software (4.1.3 version) (https://cran.r-project.org/bin/windows/base/old/4.1.3/). Additionally, ANOVA-based data was subjected for further evaluation of the effects of genotypes (G), environments (E), and replications (R) and determining the interaction magnitude of the G × E. ArcGIS-10.8 Software was used to generate the map29.

Genetic components

The mean sum of squares of all the genotypes was used to determine the genetic and environmental effects for different traits30. For each specific trait, the average of the sample data was taken to have replication mean values31,32,33. The average data among all traits was analyzed statistically34,35 as well as biometrically36,37.

  • Genotypic and phenotypic variance

$${\sigma }^{2}g=\frac{{\text{GMS}}-{\text{EMS}}}{{\text{r}}}$$

where GMS refers to the genotype mean square, EMS designates the error mean square, and r denotes the replications.

$${\sigma }^{2}p={\sigma }^{2}g+{\sigma }^{2}e$$

Here \({\sigma }^{2}p\) represents phenotypic variance, \({\sigma }^{2}g\) is denots the genotypic variance, and \({\sigma }^{2}e\) designats towards the environmental variance.

  • Environmental variance

$${\sigma }^{2}e=\frac{{\text{EMS}}}{{\text{r}}}$$

Here, \({\sigma }^{2}e\) denotes environmental variance, EMS designates towards error mean square, and r represents the replication number.

  • H2

$${h}_{B}^{2}=\frac{{\sigma }^{2}g}{{\sigma }^{2}p}$$

where \({h}_{B}^{2}\) denots the broad-sense heritability, whereas \({\sigma }^{2}g\) represents genotypic variance and \({\sigma }^{2}p\) is the phenotypic variance.

Calculation of stability statistics

Stability prediction among GSR genotypes and assessment of the yield components across different locations, the univariate and multivariate stability analyses were conducted because of the possible occurrence of substantial variations among different environments.

Univariate stability analysis

Univariate stability analysis was performed for the above-mentioned yield and yield-related components of all genotypes by using AMMI Stability Value (ASV)38,39 and AMMI Stability Index (ASI)40, Shukla’s stability variance (σ2)41 and Wricke’s ecovalence (Wi2)42.

AMMI stability value (ASV)

$$ASV=\sqrt{ {\left(\frac{{SS}_{IPCA1}}{{SS}_{IPCA2}} \left(IPCA1\right)\right)}^{2}+{\left(\left(IPCA2\right)\right)}^{2}}$$

\({{\text{SS}}}_{{\text{IPCA}}1}\) and \({{\text{SS}}}_{{\text{IPCA}}1}\) denote the sum of squares while IPCA1 and IPCA2 represent the scores of genotypes in the first and second principal component interactions, respectively39.

  • AMMI stability index (ASI)

AMMI-model based AMMI Stability Index (ASI) was designed by40 to calculate the stability among genotypes by the following equation:

$$ASI=\sqrt{ \left[{(IPCA1 \times {\theta }_{1}^{2})}^{2}+{(IPCA2 \times {\theta }_{2}^{2})}^{2}\right]}$$

IPCA1 and IPCA2 represent the values of the first and second principal component interactions, whereas, and \({\uptheta }_{1}^{2}\) and \({\uptheta }_{2}^{2}\) show the percentage sum of square descripted by these interactions.

  • Regression coefficient (bi)

Regression coefficient was estimated according to Eberhart and Russell43. Here bi is equal to 1. The genotypes not equal to 1 are sensitive to environmental change.

  • Wricke’s ecovalence

The overall participation of respective genotype to the sum of squares of G × E is evaluated by the ecovalence parameter analysis42 as given following:

$${W}_{i}^{2}={\sum ({X}_{ij}-\overline{{X }_{i.}}-\overline{{X }_{.j}}-\overline{{X }_{..}})}^{2}$$

In this context, \({X}_{ij}\) represents the average of genotype “i” in environment “j”, additionally, \(\overline{{X }_{i.}}\) refers to the mean grain yield of genotype “i”, \({X}_{.j}\) signifies the average yield of the jth environment, and \({X}_{..}\) corresponds to the overall grand mean.

  • Shukla’s stability variance

The concept of Shukla's stability variance was introduced by41. It aims to evaluate the variability of genotypes across diverse climatic conditions and is defined using the following equation:

$${\sigma }^{2}=\left[\frac{p}{(p-2)(q-1)}\right]{W}_{i}^{2}-\frac{\sum {W}_{i}^{2}}{(p-1)(p-2)(q-1)}$$

where p and q represent the numbers of genotypes and environments, respectively, and \({W}_{i}^{2}\) is the Wricke’s ecovalence of the ith genotype.

  • Weighted average of absolute score (WAASBY)

To quantify the genotypic stability, we used the function WAASB from METAN R package21 to compute the WAASB index44, as follows:

$${WAASB}_{i}=\frac{\sum_{k=1}^{p}\left|{IPCA}_{ik}\times {EP}_{k}\right|}{\sum_{k=1}^{p}{EP}_{k}}$$

where \({WAASB}_{i}\) represents the weighted average of absolute scores for the ith genotype; while \({IPCA}_{ik}\) denotes the score of the ith genotype in the kth interaction principal component axis (IPCA); and \({EP}_{k}\) signifies the amount of variance explained by the kth IPCA. Genotypes with a lower WAASB value are considered to exhibit greater stability, while environments with higher WAASB values indicate higher genotypic variance.

Multivariate stability analysis

Multivariate stability analysis through AMMI45 and GGE biplot46 were conducted to evaluate the best genotypes respective to testing environments comprising the high stability and performance, thereby better understanding of the G × E.

AMMI model

In the present study, multivariate stability was evaluated using the Additive Main Effect and Multiplicative Interaction (AMMI) model to assess genotype vs. environment interaction and predict the stability of GSR genotypes. The AMMI model combines pooled ANOVA to assess additive main effects, followed by employing singular value decomposition (SVD) on the total error matrix to compute the interaction principal components (IPCs). We determined the AMMI model in R by utilizing the metan library (Olivoto & Lúcio, 2020). AMMI model. To compute the AMMI model, the following equation was used45:

$${Y}_{ij}=\mu +{\alpha }_{i}+{\beta }_{j}+{\sum }_{k=1}^{n}{\uplambda }_{k\gamma ik}{\delta }_{jk}+{\varepsilon }_{ij}$$

Here, \({Y}_{ij}\) denotes the mean values of performance of the ith genotype in the respective jth environmental conditions, \({\alpha }_{i}\) represents the fixed effect of the respective genotype, \({\beta }_{j}\) shows the environmental effect, n is the total number of IPCA hold in AMMI model, λk is individual value for IPC axis k, γik represents the ith GSR genotype value of eigenvector for IPC axis k, \({\delta }_{jk}\) shows the eigenvector value for jth environment for IPC aaxis k, and lastly \({\varepsilon }_{ij}\) denotes the mean residual.

GGE biplot analysis

A study analysis was conducted for the sustainable phenotypic reliability of the different climatic conditions of the biplot graphic after rating the adaptable GSR genotypes47. The biplot graphs designed in this study represent the variables which impact the genotype, where the ordinate is the IPCA1. The genotype with IPCA1 falling near to zero would be considered an ideal and stable genotype, where low stability represents low productivity of the respective genotype (Gauch & Zobel, 1996; Kempton, 1984.

To compute different parametric and non-parametric stability statistics, we also used the web-based STABILITYSOFT tool48. The AMMI stability was used in conjunction with the simultaneous selection indices (SSI) technique, which is based on average yield and stability49. The average of collected data was analyzed graphically (Yan & Tinker, 2006 through R 4.1.3 software (https://www.r-project.org/) and GGE Biplot GUI tools (https://CRAN.R-project.org/package=GGEBiplotGUI). Currently, there could be several techniques to deliver the key findings: firstly, polygon biplot graphics to analyze how each genotype responded to specific environments across various experimental sites, revealing the GGE pattern. Secondly, the study focused on identifying the most suitable genotype(s) based on both mean performance and stability. Moreover, graphical displays of concentric circles with vectors of entries were used to unveil associations between different environmental factors and genotypes, with a particular emphasis on identifying stable genotypes.

Ethical approval

The plant collection and use was in accordance with all relevant guidelines.

Results and discussion

Results

Mean performance and combined analysis of variance

The mean performance of genotypes at 12 locations is mentioned in Table 3. Combined analysis of GSR genotypes (Table 4) revealed significant differences for mean squares (p < 0.05) for traits under study like Plant height (PH), number of tillers per plant (NT), panicle length (PL), grains per panicle (GPP), thousand grain weight (TGW) and paddy yield in Kg per hectare (PY. Kg ha−1). The mean squares for genotypes (G), environments (E) and genotypes x environmental interaction (G X E) for almost all the traits depicted significant differences (p < 0.05). Combined analysis of variance showed highly significant (p > 0.01) differences for G X E interaction of genotypes for all the parameters under study, which means genotypes responded differently in some environments and need to be further tested for stability and adaptability analysis.

Table 3 Mean performance of GSR lines over 12 locations countrywide, pooled average yield and percent increase or decrease than local check cultivars.
Full size table
Table 4 Combines ANOVA of significant yield attributing trains of GSR lines on the basis of mean squares.
Full size table

Estimation of genotypic and phenotypic components of variances

Analysis of genotypic and phenotypic variances for metric traits of GSR lines is represented in Table 5. Phenotypic variance has been partitioned into genotypic variance and environmental variance. In our study, all the traits except NT revealed higher genotypic variance than environmental variance. Broad sense heritability was estimated in the range of 44.36% to 98.60%. Maximum heritability value was recorded for PH (98.60%) while NT showed minimum value for heritability (44.36%). All other traits under study revealed higher values for heritability (> 70%). Higher genotypic variance and higher values for broad sense heritability depicts that parameters under study have additive genes and more stable characters for selection in the genotypes for variety development.

Table 5 Estimation of genetic components of GSR rice for yield and yield-related attributes.
Full size table

Univariate stability statistics

Estimation of univariate parametric stability statistics

The results of parametric stability statistics of yield and yield-related traits of 20 GSR lines are presented in Table 6. The parametric stability statistics like AMMI stability value (ASV), AMMI stability index (ASI), stability value based on regression coefficient (bi), Shukla’s stability variance (σ2i), Wricke’s ecovalence (Wi2) and weighted average of absolute scores (WAAS) were studied to assess the stable genotypes based on paddy yield. These parametric stability statistics were used based on concepts that genotypes that have a stability value near to zero (0) are more stable. Parametric stability statistics values were assessed for 20 genotypes over 12 locations country wide in four provinces (Punjab, Sind, Khyber Pakhtunkhwa, Baluchistan) and Islamabad, Federal Capital, in Pakistan. Based on ASV, ASI, bi, Wi2, σ2i and WAAS statistics, the genotypes G1, G4, G5, G8, G11 and G12 revealed lowest values for parametric statistics and considered more stable genotypes on the bases of paddy yield than rest of other genotypes. Minimum ASV value revealed by genotypes i.e., G12 (2.26) and maximum value by G19 (13.6). The minimum value for ASI was depicted by G12 (8.09) followed by G11 (11.7) and maximum value by G19 (48.8). The lowest value for bi was revealed by G19 (0.54) and highest value for bi was recorded in G2 (1.38). Genotypes, G1, G4, G5, G8, G11 and G12 revealed lowest values for Wi2 (< 4 × 106), σ2i (< 4 × 105) and WAAS (4.3 to 10.8).

Table 6 Parametric stability statistics of yield and yield related traits of GSR lines evaluated at 12 locations across Pakistan.
Full size table

Multivariate stability statistics

AMMI analysis of variance

The additive main effects and multiplicative interaction (AMMI) model was used to assess the genotype × environmental interaction (G × E) of 20 GSR lines over 12 locations for paddy yield that revealed significant variation (p < 0.05) for genotypes and non-signification for environment and highly significant for G × E interaction (Table 7). The variation proportion of PC1 and PC2 from interaction revealed 67.2% variability for paddy yield.

Table 7 AMMI analysis of variance for grain yield of 20 GSR lines cultivated across the country at 12 locations.
Full size table
Mean verses stability analysis of GGE (genotype + genotype × environment) biplot

The stability pattern of GSR lines was assessed based on their mean performance across various environments. The mean vs. stability graph is generated by the intersection of vertical and horizontal ordinate lines (Figs. 2, 3). Every genotype has a single arrow that has direction towards a mean performance for trait under study. In the present study mean vs. stability analysis of GGE biplot revealed 68.5% variability for genotypes on the bases of paddy yield over multilocation. The genotypes, G2, G5, G11 and G18 revealed higher paddy yield in environments, E3, E5, E7, E8, E10, E11 and E12. The genotypes depicted less arrowhead lines to these environments and considered as a best performing genotype in almost all locations.

Figure 2
figure 2

The GGE biplot design of genotype × environment interaction of 20 GSR lines and 2 check cultivars planted at 12 environments during the year 2021 for paddy yield. Note: METAN R package was used to generate figures.

Full size image
Figure 3
figure 3

‘Mean versus stability’ GGE biplot of 20 GSR lines including checks planted at 12 environments during the year 2021 for paddy yield.

Full size image

‘Which-won-where’ GGE biplot

The GGE biplot is presented in polygon view in Fig. 4. It determined the best performing genotypes for paddy yield in a group of locations (Environments). The ‘which-won-where’ GGE biplot revealed 68.62% variability in first two PCs for paddy yield. The GGE biplot explained according to, as described by50. The genotypes present near the vertex of polygon with no environment nearby are considered less stable and genotypes present on the vertex of polygon where one or more environments are prevalent are considered best performing genotypes. The genotypes present inside the polygon are less responsive to tested environment and are considered the best performing genotypes over a broad range of environments. In Fig. 4 the polygon is divided into five sectors representing 12 environments. Sector-I contains E1, E4, E7 and E12 environments. Sector-II contains E3, E8, E9, E10, and E11 environments while Sector-III includes only one environment E5. Sector-IV is represented by E6 while Sector-V contains E2 and E4 environments. The genotype, G2, G13 and G7 were best performing genotypes in E6. Genotypes G9 and G16 were best-performing genotypes in tested environments, E2 and E4. The genotypes (G6, G14, G19, G17, G18, G15 and G20) are poorly performing genotypes because they are lying in an area with no representation of any environment. The genotypes (G1, G4, G5, G8, G10 and G14) that are less responsive to any environment and exhibited stable performance for yield in all the environments.

Figure 4
figure 4

‘Which − won − where’ view of GGE biplot 20 GSR lines including checks under 12 environments during the year 2021 for paddy yield.

Full size image

‘Discriminativeness vs. representativeness’ pattern of stability

The GGE biplot polygon is divided into four quadrants (Fig. 5). Genotypes, G11, G15 and G20 that fall in the upper right quadrant of the polygon have high discriminativeness and high representativeness and are considered the most desirable genotypes. Genotypes, G4, G5, G6, G18, G19 and G17 that fall in the lower right quadrant have high ‘discriminativeness and low representativeness’ and are suitable for specific environments or management practices. In this case G17 is the most suitable for these genotypes. Genotypes (G2, G3, G7, G8, G10, G13 and G14) in the upper left quadrant have low Discriminativeness and high representativeness, are best suited for average or low-yielding environments. In this case environments, E5 (D.I Khan, KPK) and E6 (Usta Muhammad, Baluchistan) were best suited environments for yield performance. Finally, genotypes located in the lower left quadrant have low discriminativeness and low representativeness and should be discarded.

Figure 5
figure 5

‘Discriminativeness vs. representativeness’ view of GGE biplot of 18 GSR lines and 2 check cultivars grown under 12 environments during the year 2021 for paddy yield.

Full size image

IPCA and WAASB/GY ratio-based stability heat-map

Interaction Principal Components (IPCA) and weighted average of absolute Scores (WAASB) heat map (Fig. 6) is a graphical representation that uses color-coded cells to display data in a matrix format. In the context of stability analysis and the estimation of the WAASB (Weighted Average of Absolute Scores on Principal Component Axes) index, a heatmap can be used to show the genotype ranking based on the number of principal component (PC) axes utilized. Axis Labels: The horizontal axis of the heatmap represents the number of principal component axes used for estimating the WAASB index. It typically starts from a lower value and increases incrementally. The vertical axis represents the genotypes being analyzed. Color Coding: Each cell in the heatmap is color-coded to represent the genotype’s ranking based on the WAASB index. The colors may range from a gradient of a single color (e.g., lighter to darker shades of blue) or a spectrum of multiple colors (e.g., a rainbow palette). The color intensity or hue can indicate the relative ranking, with darker or more vibrant colors representing higher rankings and lighter or paler colors representing lower rankings. Genotype Ranking: The heatmap provides a visual representation of how different genotypes are ranked based on their WAASB index scores. The rows (vertical axis) correspond to individual genotypes, and the cells’ colors reflect their relative rankings across different numbers of PC axes. Interpretation: Observing the heatmap, one can identify patterns or trends in genotype rankings. For example, genotypes that consistently rank high across different PC axes will display cells with darker colors throughout the heatmap. Conversely, genotypes that consistently rank low will have cells with lighter colors. Inconsistencies or variations in rankings may be represented by cells with mixed colors or transitions between lighter and darker shades. Optimal Number of PC Axes: By examining the heatmap, researchers can assess the influence of the number of PC axes on genotype rankings. They can observe if there is a specific number of PC axes where genotypes consistently achieve higher rankings. Heat map revealed stable genotypes i.e., G8, G5, G4, G12, G11, G2, G3 and G1 based on IPCAs and WAASB.GY ratio.

Figure 6
figure 6

a Shows the genotype ranking depending on the number of principal component axis used for estimating the WAASB index. Four clusters of genotypes are shown by label colors (red) unproductive and unstable genotypes; (blue) productive, but unstable genotypes; (black) stable, but unproductive genotypes; and (green), productive and stable genotypes. b Shows the genotype ranking depending on the WAASB/GY ratio. The ranks obtained with a ratio of 100/0 consider exclusively the stability for the genotype ranking. On the other hand, a ratio of 0/100 considers exclusively the productivity for the genotype ranking.

Full size image

Ranking environments

The environment E2 (KSK, Lahore) is the best environment for paddy yield performance for all the genotypes tested (Fig. 7). The tested environment for paddy yield was revealed based on nearness to the concentric centers. In this biplot analysis, genotypes are considered as random samples of testing genotypes. The pattern of environment ranking for GGE biplot for paddy yield is E2 > E4 ≈ E1 > E1 > E3 > E5 ≈ E8 ≈ E ≈ E2 ≈ E4 ≈ E5 ≈ E9 ≈ E10 ≈ E11≈ E12 > E7 > E6.

Figure 7
figure 7

The GGE biplot ‘Ranking Environments’ profile to rank environments for the ideal location of 20 GSR lines including checks grown under 12 environments during the year 2021 for paddy yield.

Full size image

Ranking of genotypes

The G + G × E biplot for genotype ranking revealed ideal genotype as compared to other genotypes under study (Fig. 8). The arrowhead indicates the ideo-types that performed well in all tested locations. The genotypes near to the concentric arrow are considered the best performing genotypes. In this biplot analysis locations are considered as random samples of testing environment. The pattern of genotypes ranking for GGE biplot for paddy yield is G9 ≈ G2 ≈ G14 ≈ G3 > G1 > G13 > G12 > G8 > G11 > G7 ≈ G10 ≈ G4 > G6 > G5 > G11 > G20 > G19 > G17 > G18 > G15.

Figure 8
figure 8

The GGE biplot ‘Ranking genotypes’ pattern to rank genotypes to find out the ideal genotype of 20 GSR lines including checks evaluated under 12 environments during the year 2021for grain yield.

Full size image

Discussion

Introduction of GSR is a success story in Pakistan after its yield performance we revealed at local climatic condition as compared to locally adopted check cultivars. Our findings for GSR paddy yield performance are in comparison with the results of previous studies by51. For variety development, GSR germplasm was tested over broad range of environmental conditions country wide in Pakistan. Multilocation yield trials are the major activity of breeding program to assess the genotypes or candidate line for stable performance and adaptability over the broad range of environmental conditions. Selection of genotype for multi-environment based on prediction values in comparison with observed values is crucial44. There are three main factors that can increase the accuracy of prediction of suitable genotypes in multi-environments (ME). The first one is the use of experimental design which includes the plot size, field management and area of experiment. The second one is the increased number of replications to minimize the experimental error. Third one is the statistical analysis that can predict accuracy in comparison with observed values for selection of suitable ideotype52.

From breeding point of view, the estimation of genotypic variance, phenotypic variance, environmental variance, interaction of G × E by analysis of variance and heritability estimates are very important in terms of gene expression controlling complex traits like grain yield over a broad range of environmental conditions. Based on information about variance contribution by genotypes and environment, an ideal genotype can be selected with less influence by environment. Only heritable variations transfer from generation to generation for a particular trait in a genotype that is essential for selection of suitable genotype53. A genotype with high broad sense heritability and genotypic variance is more stable as compared to genotypes with more influenced by environmental variance54,55,56. In the present study, analysis of variance revealed significant differences for all the traits studies (PH, PL, GPP, TGW and YLD) except NT, which was non-significant. Significant G x E interaction depicted that multi-environment (ME) played important role for the expression of genotypes for paddy yield and these similar finding was reported by57. Genotypic variance and broad sense heritability estimates were recorded in higher magnitude for all the traits except NT and similar findings were reported by54,55. However, careful selection of genotypes should be made while selection of genotypes for NT because this trait is governed by environmental variance.

From the agronomic point of view of AMMI biplot analysis, GEI biplot, ‘which-won-where’ biplots and WAASB facilitates selection of genotypes performed well in a particular environment44. In our study, we evaluated genotypes at multi-location, and data was subjected to both univariate and multivariate statistics. Univariate statistics such as ASI, ASV, bi, Wi2, σ2i and WASSB is used to identify the most stable genotypes based on certain parametric statistics. Based on ASV, ASI, bi, Wi2, σ2i and WAAS, the genotypes G1, G4, G5, G8, G11 and G12 revealed the lowest values for all the stability statistics. These genotypes are considered the most stable genotypes based on these statistics and results obtained in this study are in accordance with the early findings58. Univariate stability statistics have some limitations as compared to multivariate statistics such as AMMI analysis and GGE biplot analysis58.

Multivariate stability statistics are often preferred due to the complex nature of rice production systems and the interdependencies among various factors affecting stability. Rice production and stability are influenced by multiple variables, such as climate, soil conditions, agronomic practices, and management strategies. Multivariate stability statistics can consider the joint effect of these factors, providing a more holistic view of stability compared to analyzing individual factors in isolation. Among multivariate methods, the AMMI analysis is mostly used for G x E interaction54. Non-parametric statistics such as the biplot mostly point out suitable genotypes in a particular location based on their performance. In our study, AMMI analysis of variance revealed significant GGE interaction and the first two Interaction Principal Component Axis (IPC1 and IPC2) were significant and depicted 39% and 67% variance for paddy yield, respectively (Table 6). This result indicates interaction of the environment had an important contribution to the performance of a genotype53.

Performance of genotypes for yield, agronomic and other botanical traits is influenced by multiple environmental factors such as photoperiod, temperature, humidity, darnel changes, soil types, rain, competitions with local biotic factors (weeds, diseases, insects, soil microbiomes interactions with plants, acidity/alkalinity of soil fertility status of soil, time of sowing etc. Keeping in view, various multivariate statistics i.e. ‘Mean vs. Stability’ analysis of GGE biplot, ‘Which-won-where’ GGE biplot, ‘Discriminativeness vs. representativeness’ pattern of stability, ‘ranking environments’, ‘ranking genotypes’ were used to identify higher yielder and stable genotypes in particular environments By using ‘Mean vs. Stability’ analysis of GGE biplot, the genotypes, G2, G5, G11 and G18 revealed higher paddy yield in environments, E3 (NIAB, Faisalabad), E5 (ARI. D.I Khan, KPK), E7 (RRS, Bahawalnagar, Punjab), E8 (RRI, Dohkri, Sind), E10 (Shaikhupura, Punjab), E11(Gujranwala, Punjab) and E12 (Shikarpur, Sind). According to the ‘Which-won-where’ GGE biplot, 12 environments were grouped into five sectors i.e., Sector-I (E1, E4, E7 and E12), Sector-II (E3, E8, E9, E10, and E11), Sector-III (E-5), Sector-IV (E6) and Sector-V (E2 and E4). These sectors are based on similar environmental conditions and the performance of genotypes in these environments. The genotypes, G2, G13 and G7 were the best performing genotypes in Sector-IV which contains only a single environment, E6 (Usta Muhammad, Baluchistan). These genotypes can be recommended in Usta Muhammad, Baluchistan area for cultivation after further testing and evaluations for yield, this environment has highly fertile soil and high temperature conditions. The genotypes G9 and G16 were best performing in Sector-V (SSRI, Pindi Bhattian, and KSK, Lahore, Punjab); these genotypes are best performing in alkaline soils because one of the environments i.e. E4 has highly saline soil conditions. The GGE biplot polygon is utilized to characterize the discriminativeness and representativeness of genotypes in multi-environment. The polygon characterized a graphical interface between genotypes and environments, where each point in the polygon corresponds to a genotype and represents its performance across different environments. The x-axis of the polygon represents discriminativeness which is a measure of the genotype's ability to differentiate among the tested environments. A genotype with high discriminativeness can be used to select the best performing genotypes across different environments. The y-axis of the polygon represents representativeness which is a measure of the genotype's stability across different environments. A genotype with high representativeness performs consistently in all environments, even if it does not necessarily have the highest yield. In case of the ‘Discriminativeness vs. representativeness’ pattern of stability, the genotypes, G2, G3, G7, G8, G10, G11, G13 and G14, G15, G17 and G20 that fall in the upper right quadrant of the polygon have high discriminativeness and high representativeness and are considered the most desirable genotypes. In this pattern stability study, we identify two environments, E5 (D.I Khan, KPK) and E6 (Usta Muhammad, Baluchistan) that were best-suited environments for yield performance under hot weather conditions. The genotypes that performed best in these two environments are likely to heat and drought resistant genotypes and should be recommended for cultivation under drought and high-temperature environments after farther testing for yield and agronomic traits.

According to the ranking environment, environment E2 (KSK, Lahore) is best environment for paddy yield performance for all the genotypes tested and ranking genotypes, the genotypes G1, G2, G8, G11, G12, G13 and G14 were identified as best performing genotypes in case of paddy yield. Heatmap (Fig. 7) also revealed stable genotypes i.e., G8, G5, G4, G12, G11, G2, G3 and G1 based on IPCAs and WAASB.GY ratio. The genotypes can be utilized for stable and high yielding variety development process and can be used in the future breeding program for high-yielding, salt, and drought tolerant variety development.

Conclusion

Paddy yield in rice is a complex polygenic trait, governed by several factors. Environmental interaction of genotypes for the expression of genes related to yield is an important phenomenon. Testing of genotypes in a wide range of environmental conditions is a prerequisite before the commercial release of cultivar by plant breeders. To identify the most stable genotypes in a multi-environment, we evaluated 20 GSR lines including checks by using univariate and multivariate statistical approaches. Based on ASV, ASI, bi, Wi2, σ2i and WAAS statistics, the genotypes G1, G4, G5, G8, G11 and G12 revealed lowest values for parametric statistics and observed more stable genotypes. To further validate the results, data was subjected to multivariate statistics and AMMI analysis. This revealed significant differences among genotypes and G x E interaction. GGE biplot depicted 67% variability among genotypes in the first two PCs. Based on GGE biplot analyses, ‘IPCA and WAASB/GY’ ratio-based stability Heat-map and ranking of genotypes, the genotypes namely G1, G2, G3, G5, G8, G10, G11 and G13 were observed better performing and stable. It is concluded from our study that these genotypes could be recommended for commercial cultivar development after further testing in national yield trials.