Agronomic Performance and Yield Stability of Elite White Guinea Yam (Dioscorea rotundata) Genotypes Grown in Multiple Environments in Nigeria

Abstract

Yam (Dioscorea spp.) is a main staple tuber crop in Nigeria and the West African region. Its performance is determined by genotypes and also the environment of growth. This study assessed the agronomic performance and yield stability of elite white yam (Dioscorea rotundata) genotypes across diverse Nigerian environments. A total of 25 genotypes were evaluated at three locations in two consecutive growing seasons, 2022 and 2023, for fresh tuber yield, disease resistance, and tuber quality traits. The genotype’s performance and stability for the measured traits were assessed using various analytical tools such as additive main effects and multiplicative interaction (AMMI) and multi-trait stability index (MTSI). The AMMI analysis revealed significant differences among the genotypes and across the environments for all traits (p < 0.001, p < 0.01). The PCA revealed that the first two principal components (PC1 and PC2) explained a substantial portion of the total variation (49.84%). The MTSI identified four clones: G18, G19, G24, and G16 as promising candidates for improved yam production in Nigeria with high and stable performance for the multiple traits.

1. Introduction

Root and tuber crops (RTB) have been playing increasingly important roles in food security and in meeting global dietary requirements [1]. Among the RTB’s, yam (Dioscorea spp.) holds unparalleled significance as both a staple food and a cash crop in Africa’s sub-humid tropics. West Africa accounted for 96% of the world yam production figure of 88.3 Mt [2]. Current estimates put the Nigeria production figure at 61.2 MT representing 72% of West African production [2].

Yam is a multi-species crop consisting of approximately 600 species. Within the diverse array of yam species, White Guinea yam (Dioscorea rotundata Poir) reigns supreme, flourishing extensively in West Africa’s tropical regions, and supporting the livelihoods of over 400 million people, and its cultivation extends from Cameroon to Sierra Leone [3]. Yam is primarily grown as subsistence crop, heavily reliant on natural environmental conditions. However, its demand far exceeds supply due to low productivity, which is constrained by several biotic and environmental factors. With the global population projected to reach 9.6 billion by 2050, enhancing the productivity of vital food crops like yam, while maintaining environmental sustainability, is vital for ensuring food security [4].

In Nigeria, institutions like the International Institute of Tropical Agriculture (IITA) prioritize yam breeding endeavors aimed at developing novel varieties that meet end-users’ preferences, encompassing both farmers and consumers. This focus revolves around critical traits such as consistent tuber quality, high yield, less oxidative tuber flesh browning tendency, increased dry matter content, and resistance against yam mosaic virus and anthracnose disease [2,5,6].

Breeding is a potent tool for sustainable methods to enhance yam cultivation and ensure food security. However, progress in trait improvement relies heavily on comprehensive genetic information and the understanding of trait behavior across diverse environments. The exploration of yam quantitative traits [7,8,9] has consistently highlighted the significant impact of genotype-by-environment interaction (GEI). GEI refers to the differential responses of genotypes across growth environments, and dissecting such patterns in breeding trial data is crucial because it helps identify genotypes that are superior and consistent in performance across multiple environments.

Research across diverse environments has demonstrated that both genetic and non-genetic factors contribute to phenotypic variation. The proportion of total variation due to the environment, genotype, and their interaction varies depending on the specific trait being studied [8,9,10]. However, GEI studies are complex due to the large datasets involving multiple genotypes tested across various environments. Making data such analysis and interpretation challenging due to the multiple traits involved in breeding trials. This complexity underscores the need for a multi-trait stability index, which can capture and evaluate the stability of multiple traits simultaneously. This approach is crucial when selecting genotypes based on more than one trait. Several studies [7,8,11] have shed light on this crucial interaction, demonstrating its role in determining yam genotype performance across different environmental contexts. These findings emphasize the importance of considering GEI in yam breeding and advocate for the use of multivariate analysis techniques to address its complexities effectively [12]. As breeding strategies evolve, understanding heritable variations and genetic correlations among economically significant traits becomes imperative. This knowledge forms the foundation for selecting breeding plans, guiding the development of desired progenies within breeding populations.

This study aims to estimate variance components and the heritability of agro-morphological traits and analyze the genotype–environment interaction patterns of elite white Guinea yam genotypes grown across multiple environments in Nigeria. We employed the Additive Main Effects and Multiplicative Interaction and Multi-Trait Stability analysis models due to their ability to handle complex GEI interactions, provide clear visualization, and support multi-trait breeding objectives.

2. Materials and Methods

2.1. Plant Materials and Trial Establishment

Twenty-five yam genotypes from the International Institute of Tropical Agriculture (IITA) Ibadan, Nigeria, yam breeding program were used in this study. The genotypes were chosen based on their attributes, which include dry matter content, fresh tuber yield, and disease resistance. Details of the genotype taxonomy, testing locations, and climatic data are presented in Supplementary Tables S1–S3. The experiments were carried out at Ikenne, located at 6°84′ N and 3°69′ E; Abuja located at 9°13′ N and 7°23′ E and Ibadan located at 7° 49′ N and 3° 90′ E in Nigeria for two consecutive years, 2022 and 2023. The field was ridged in 1 m spacing at a soil depth of 40 cm. Seed tubers of size 150 gm from each genotype in the trial were planted at 1 m spacing using the normal hoe at a depth of 10 cm on top of the ridge. The planting was carried out in an alpha lattice design of three plants per plot at a spacing of 1 m × 1 m in two replicates. The trial was established in April for Ibadan and Ikenne and in May for Abuja due to their variations in climatic factors. The trial was staked at 2 m high and proper weeding was performed manually with hoes to ensure that the field was kept clean and free of weeds. NPK 15-15-15 (15% Nitrogen, 15% Phosphorus, 15% potassium) fertilizer was applied at the rate of 200 kg ha⁻¹ at 8 weeks after planting and the best pesticide with the active ingredient Cypermethrin (EC, cypermethrin 10%, m/m, 90% agricultural emulsifier and solvent) was applied at 0.6 L ha⁻¹

Agronomic and yield traits (

Table 1. Agronomic traits, description, and time recorded for this study at each month after planting (MAP).

S/N	Traits	Full Names	Description	Time Recorded
1	AUDPCYAD	Area under disease progression curve yam anthracnose disease	The rating of symptoms caused by anthracnose over a period of 2-5MAP and converted to area under disease progression curve	Over the period of 2-5MAP
2	AUDPCYMV	Area under disease progression curve yam mosaic virus	The rating of symptoms caused by virus over a period of 2-5MAP and converted to area under disease progression curve	Over the period of 2-5MAP
3	PLNV	Plant Vigor	How vigorous the plants appear at 3MAP	3 MAP
4	INTOX30	Intensity of tuber oxidation 30 min	Visual tuber oxidation was accessed at harvest	At harvest
5	INTOX180	Intensity of tuber oxidation 180 min	Visual tuber oxidation was accessed at harvest	At harvest
6	TTY	Fresh tuber yield	Yield was estimated per plot using the formula total tuber weight divided by the effective plot multiplied by ten	At harvest
7	DM	Dry matter content	Percentage of dry matter content of tuber	At harvest
8	ATW	Average tuber weight	Average weight of tuber per plot was accessed at harvest	At harvest

All trait information was accessed using the Standard Operating Protocol for Yam Genotype Performance Evaluation Trial [13].

) were collected at the vegetative stage of the plant and at harvest using the “Standard Operating Protocol for Yam Genotype Performance Evaluation Trial” [13]. All agronomic management practices were carried out equally and properly per the earlier recommendations.

The area under the disease progression curve (AUDPC), is a valuable quantitative summary of disease severity for YMV and YAD over time and, was estimated using the trapezoidal method [14] This method represents the time variable and calculates the average disease severity between each pair of adjacent time points:

$$AUDPC=\sum _{i=1}^{n-1}\left({\displaystyle \frac{y_i+y_{i+1}}{2}}\right)\left(t_{i+1}-t_i\right)$$

where n is the number of assessments made, y_i is the anthracnose or virus severity score on date i, and t is the time in months between assessments y_i and y_i+1.

The rate of pathogen reaction to yam mosaic virus (YMV) and yam anthracnose disease (YAD) (severity scores) were recorded monthly from two to six months after planting following the scale as shown in (Figure 1) below.

DMC (%) = (dry tuber weight (g)/fresh tuber weight (g)) × 100

TTY (t ha⁻¹) = (total tuber weight harvested (kg)/effective plot (m²)) × 10

$$\mathrm{ATW}({\displaystyle \frac{\mathrm{kg}}{\mathrm{tuber}}})=\left(\mathrm{total}\mathrm{tuber}\mathrm{weight}\mathrm{harvested}\mathrm{per}\mathrm{plot}(\mathrm{kg})/\mathrm{total}\mathrm{tuber}\mathrm{numbers}\mathrm{harvested}\right)$$

2.2. Statistical Analysis

All statistical analyses were performed using various packages in R Studio [15]. The function “find outliers” from the “metan” package Version 1.18.0 [16] was used to ensure there were no outliers and that the data followed a normal distribution. Adjusted means (LS means) for each trait across environments were calculated using the “emmeans” package version 1.10.4.900001. [17]. Analysis of variance (ANOVA) was carried out using the alpha lattice design model below [18].

Y_ijkl = μ + Env_l + Rep(Env)_lj + Gen_i + (Gen:Env)_il + e_ijkl

where Y_ijkl is the observation response, $\mu$ represents the overall mean, Envi is the fixed effect of the lth environment, Rep(Env)_lj is the effect of the jth replication within the lth environment, Gen_i is the fixed effect of the ith genotype, (Gen:Env)_il is the interaction effect between the ith genotypes and lth environment, and e_ijkl represents the residual error term.

The adjusted trait mean of the ith genotype in the l^th environment from the above model was used for principal component and correlation analysis. Principal component analysis (PCA) was conducted using the “Factoextra package” Version 1.0.7 [19] to generate the Eigenvalues, the percentage of the variation accumulated by the PCA, and the load coefficient values between the original characters and respective PCA. Pearson correlation analysis was performed using the “metan” package in R software version 4.1.1. The “corr_coef” function from the package was used to plot the graph.

Genetic parameters such as broad-sense heritability and phenotypic coefficients were estimated using the function “get model data” in the metan package in R. Plot-based broad-sense heritability across all environments for all traits was calculated using the following formula:

$${\mathrm{H}}^2={\mathsf{\sigma }}_{\mathrm{g}}^2∕\left({\mathsf{\sigma }}_{\mathrm{g}}^2+{\mathsf{\sigma }}_{\mathrm{g}\mathrm{x}\mathrm{e}/\mathrm{l}}^2+{\mathsf{\sigma }}_{\mathrm{e}}^2/lxk\right)$$

Phenotypic coefficient of variation.

$$\mathrm{C}\mathrm{V}\mathrm{p}={\displaystyle \frac{{\sqrt{\mathsf{\sigma }}}^2\mathrm{p}}{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}}{\mathrm{x}}_{100}$$

Genotypic coefficient of variation.

$$\mathrm{C}\mathrm{V}\mathrm{g}={\displaystyle \frac{{\sqrt{\mathsf{\sigma }}}^{2}g}{\mathrm{g}\mathrm{r}\mathrm{a}\mathrm{n}\mathrm{d}\mathrm{m}\mathrm{e}\mathrm{a}\mathrm{n}}}{\mathrm{x}}_{100}$$

σ²_g is the genotypic variance_, σ²_e is the residual variance, and σ²_gxe is the genotype-by-environment interaction variance, $\mathrm{l}$ is the number of environments, k is the number of replicates. Heritability was categorized as low (<30%), moderate (31–60%), and high (>60%) according to Robinson et al. [20].

Genotype by environment interaction (GEI) analysis was performed using the Additive Main Effects and Multiplicative Interaction (AMMI) model [21]

$${\mathrm{y}}_{\mathrm{i}\mathrm{j}}=\mathsf{\mu }+\mathsf{\alpha }\mathrm{i}+\mathrm{T}\mathrm{j}+\sum _{\mathrm{k}=1}^{\mathrm{p}}{\mathsf{\lambda }}_{\mathrm{k}}{\mathrm{a}}_{\mathrm{i}\mathrm{k}}{\mathrm{t}}_{\mathrm{j}\mathrm{k}}+{\mathrm{P}}_{\mathrm{i}\mathrm{j}}+{\mathrm{e}}_{\mathrm{i}\mathrm{j}}$$

where y_ij is the observed mean yield, λ_k is the singular value for the k-th interaction principal component axis (IPCA); a_ik is the i-th element of the k-th eigenvector; t_jk is the j^th element of the k^th Eigenvector. A residual e_ij remains, p IPCA are used, where p ≤ min (g − 1; e − 1). This approach depends on the analysis of variance (ANOVA) for estimating genotype and environment main effect, principal component analysis (PCA) for decomposing GEI structure into Interactive Principal Component Axes (IPCAs), and biplot for graphical presentations. AMMI provides a suitable approach to separating the genotypic effect from the genotype-by-environment effect with cultivar ranking in mega-environment [22]. At the same time, GGE is ideal for grouping sites and cultivars without cultivar rank change [23]. Superior cultivars and test environments were selected.

The multi-trait stability index was determined using the formula used by [24] to select stable genotypes for the studied traits. The genotypes with the lowest MTSI were considered superior [25].

$$MTSIi={\left({\sum }_{j=1}^f{\left(f_{ij}-f_j\right)}^2\right)}^{0\cdot 5}$$

MTSIi = multi-trait stability index for the i-th genotype, $f_{ij}$= j-th score for the i-th genotype, $f_j$ = j-th score of ideal genotypes.

3. Results

3.1. Variation in Quantitative Traits across Environments for 25 Yam Genotypes

Figure 2 displays the variation of quantitative traits assessed for 25 yam genotypes across different environments with each plot representing the distribution of adjusted means for a specific trait in multiple environments. The box represents the interquartile range, which is the range between the first quartile (Q1, 25th percentile) and the third quartile (Q3, 75th percentile) and 50% of the data lies in between. The wider boxes indicate more variation in adjusted means among genotypes within that environment and the narrower boxes indicate less variation and more consistency among the genotypes. The points outside the whiskers (blue dots) are adjusted means, which are significantly different from the others. The median (the horizontal line) in each environment varies in terms of their traits, which implies that the environment influences the performance of the genotypes for each trait. ATW, PLNV, AUDPCYAD, AUDPCYMV, and Oxi180 have higher variability among others. Traits such as PLNV, DMC, ATW, and TTY have their means (red diamond) higher than the median, which indicates that few genotypes have higher performance. PLNV has the smallest boxplot, compared to Oxi30, which has the largest variability.

3.2. Analysis of Variance on Quantitative Traits

Table 2. Mean squares and heritability estimate of 25 genotypes evaluated in 6 environments during two-year growing season.

Source	df	AUDPCYMV	AUDPCYAD	PLNV	ATW	TTY	Oxi30	Oxi180	DMC
Env	5	77,615 ***	55,359 ***	2.691 ***	2.533 ***	601.1 ***	99.30 ***	83.65 ***	109.8
Rep (Env)	6	474	903	0.563	0.605	43.9	3.09	7.16	38.1
Genotypes	24	2096 ***	3418 ***	0.251	1.268 ***	123.9 ***	6.83 ***	10.20 ***	114.2 *
Geno × Env	120	1610 ***	2046 ***	0.262	0.767 ***	73.7 ***	3.84 *	6.26 **	68
residual	144	907	1054	0.195	0.426	37.3	2.78	4.07	61.5

*, **, *** = significant at 0.05, 0.01, and 0.001 probability levels, respectively; Env = environment; Geno = genotypes; AUDPCYMV = area under disease progression curve yam mosaic virus; AUDPCYAD = area under disease progression curve yam anthracnose disease; PLNV = plant vigor; ATW = average tuber weight; TTY = total tuber yield; DMC = dry matter content; Oxi30 and Oxi180 = oxidation index at 30 and 180 min, respectively.

shows the combined analysis of variance for the measured agronomic traits across six environments. The results showed highly significant (p < 0.05 or p < 0.001) differences among genotypes for all the traits measured except for plant vigor (PLNV).

Genotype X environment interaction (GEI) was significant (p < 0.05, p < 0.01, and p < 0.001) for all measured traits except for dry matter content and plant vigor.

3.3. Genotypic Coefficients, Phenotypic Coefficients, and Broad-Sense Heritability

This study showed that the phenotypic coefficient of variation is higher than the genotypic coefficient of variation (

Table 3. Genetic parameters estimate in the selected white yam genotypes.

Genetic Parameters	AUDPCYMV	AUDPCYAD	ATW	TTY	PLNV	Oxi30	Oxi180	DMC
GV	2.18	4.9	4.93	5.07	0.00	3.33	2.94	5.21
PV	52.20	55.10	48.80	54.70	56.30	62.70	62.40	18.00
H2	23.00	40.00	39.00	41.00	0.00	44.00	39.00	41.00
CVg	0.88	0.95	153.13	14.47	0.00	81.10	47.37	6.67
CVp	4.28	3.19	481.82	47.53	295.33	351.95	218.23	12.38
Mean	168.65	232.86	1.45	15.56	2.54	2.25	3.62	34.24

GV = genotypic variance; PV = phenotypic variance; H2 broad sense heritability; CVg = Genotypic coefficient of variance; CVP = phenotypic coefficient of variance; PLNV: plant vigor; AUDPCYAD: area under disease progression curve yam anthracnose disease; AUDPCYMV: area under disease progression curve yam mosaic virus, ATW: average tuber weight; TTY: total tuber yield; OXi30 and OXi180: intensity of tuber oxidation at 30 and 180 min; and DMC: dry matter content.

). The genotypic coefficient of variation ranged from 0 for plant vigor to 153.13 for average tuber weight. On the other hand, phenotypic coefficient of variation (CVp) values ranged from 3.19 to 481.82 in all the traits. Low broad sense heritability (H²) (0–30%) was observed for plant vigor and AUDPCYMV, while moderate broad sense heritability was observed for all other traits (Table 3).

3.4. Traits Importance and Contribution

The principal component analysis displayed the contribution of the quantitative traits in the white yam genotypes (

Table 4. Principal component analysis and variables contribution on each factor for quantitative traits.

Variables	PC1	PC2	PC3	PC4
AUDPCYMV	−0.29	0.42	−0.08	−0.20
AUDPCYAD	−0.17	0.28	−0.01	−0.63
ATW	0.38	0.19	−0.10	−0.28
TTY	0.56	0.30	0.09	0.07
PLNV	0.01	0.18	−0.71	−0.22
Oxi30	0.23	−0.54	−0.14	−0.28
Oxi180	0.25	−0.46	−0.06	−0.39
DMC (%)	−0.06	0.05	0.66	−0.44
eigenvalue	2.35	2.14	1.08	1.05
variance (%)	26.08	23.76	12.02	11.69
cumulative (%)	26.08	49.84	61.87	73.55

PLNV: plant vigor; AUDPCYAD: area under disease progression curve yam anthracnose disease; AUDPCYMV: area under disease progression curve yam mosaic virus; ATW: average tuber weight; TTY: total tuber yield; OXi30 and OXi180: intensity of tuber oxidation at 30 and 180 min; DMC: dry matter content; and PC = principal component. The bold denotes higher contributing factors in each PCs as explained in the text.

). The first four principal components of the quantitative traits accounted for 73.55% of the total variation. The first principal component (PC1) had average tuber weight (ATW) and fresh tuber yield (TTY) which contributed most to the total variation. PC2 had an area under the disease progression curve for yam mosaic virus (AUDPCYMV), with TTY and oxidation intensity at 30 and 180 min contributing most to the total variation. PC3 had plant vigor and dry matter content, contributing most to the total variation. PC4 had oxidation intensity at 180 min, dry matter content, and area under the disease progression curve, with yam anthracnose disease contributing most to the total variation.

3.5. Phenotypic Correlation Coefficient between the Quantitative Traits Measured

Pearson correlation coefficients among the agronomic traits for the 25 white yam genotypes evaluated across six environments are displayed in Figure 3. There were significant (p < 0.001) positive correlations between ATW and TTY (0.44) and AUDPCYMV and AUDPCYAD (0.35). Such correlations were negative but not significant between AUDPCYAD with oxidation intensity at 30 min (−0.25) and AUDPCYMV with oxidation intensity at 180 min (−0.35) and Oxi30 (−0.49), and significant (p < 0.05) between AUDPCYMV and TTY (−0.12). Plant vigor (PLNV) showed a positive but negligible correlation with AUDPCYMV (0.14) and a significantly negative correlation (p < 0.01) with oxidation intensity at 180 min after tuber cut.

3.6. Additive Main Effect and Multiplicative Interaction of Agronomic Traits

The mean squares of AMMI analysis of variance are presented in Supplementary Table S4. The AMMI analysis revealed significant variation in the main effects of the environment (p < 0.001) for all traits except for dry matter content. The effect of genotype was significant at (p < 0.001) for all observed traits except PLNV. Genotype X environment interaction (GEI) was decomposed into two principal components, and PC1 mean squares were highly significant (p < 0.001; p < 0.01) for all traits. PC2 mean squares were significant (p < 0.001; p < 0.01; p < 0.05) for all the traits.

3.7. GxE Interaction Pattern

Figure 4 shows the AMMI1 biplots where genotypes and environments are depicted as points on a plane with the first PC on the y-axis. The horizontal line within the plot shows the interaction effect and the vertical axis represents the main effect. The vertical axis showed interaction between genotypes and the environments while the horizontal showed the main effect of the genotype and environment. The superior genotypes for fresh tuber yield are the genotypes on the right two quadrants (top and bottom right) of the biplot. The environments were distributed from the low-yielding environments in the left quadrants to the high-yielding ones in the right quadrants. IK23, AB23, IK22, and IB23 were identified as high-yielding environments (Figure 4a). The superior genotypes identified for dry matter were the genotypes on the right two quadrants (top and bottom right) of the biplot. The environments were distributed from the low dry matter content environments on the left quadrants to the high dry matter content on the right quadrants. IK23, AB23, AB22, and IB23 were identified as high dry matter content environments (Figure 4b).

AMMI2 biplots of white Guinea yam genotypes are illustrated in Figure 5. The AMMI 2 revealed the importance of PCA2 scores along with the first PCA in explaining the complexity of GEI involving significant multi-environments and identifying the adaptation of genotypes. The environmental vectors (solid green lines) are the strength and pattern of interaction between each environment and the genotypes. The longer lines are stronger interaction with the genotypes which are more discriminating than the shorter lines which is of less interaction. The dashed lines which connect genotypes shows the spatial relationship between the genotypes and how they relate to each other in-terms of principal components (PC1 and PC2). The PC2 values were 25.9% and 12.7% for fresh tuber yield, and dry matter content. Genotypes and environments near the origin (center of the plot) have average performance and stability because environmental changes less influence their performance. In contrast, those far from the origin and close to each other indicated strong specific interaction. For fresh tuber yield, G22 performed well in AB 23, G7 and G6 performed well in IK23, G1 and G14 performed well in AB22, and G17 performed well in IB23 environments. For dry matter content, G14 performed well in IB23, G16 and G19 performed well in IK23, and G25 performed well in AB22 and AB23 environments.

Figure 6a shows the mean performance and stability of 25 genotypes for total tuber yield across six environments. The arrowed average tester coordination (ATC) lines from the biplot show the experiment’s lowest vs. highest and stable vs. unstable genotypes [26]. A deviation from the ATC lines indicates stability; therefore, G16, G10, and G4 were among the most stable genotypes. G17 and G7 were among the least stable genotypes for total tuber yield because they strongly deviated from the ATC lines. Genotypes near the arrow have the highest mean performance for total tuber yield. Figure 6b shows the mean performance and stability of 25 genotypes for dry matter content across six environments where deviation from the ATC lines indicates stability; therefore, G3 and G15 were among the most stable genotypes. Genotype G14 was the least stable for dry matter content because it strongly deviated from the ATC lines. Genotypes close to the arrow are the genotypes with the highest mean performance for dry matter content.

3.8. Multi-Trait Selection for Agronomic Traits

The traits under study were selected through a multi-trait stability index (MTSI). The selection differential was negative for all the traits. The selection differential for the WAASB (Weighted Average of Absolute Scores) index was 1.09%, 9.08%, 19.2%, 19.3%, 24.3%, 26.2%, 26.4%, and 39% for DMC, AUDPCYAD, TTY, AUDPCYMV, ATW, PLNV, Oxi180, and Oxi30, respectively (

Table 5. Selection differential of the WAASB index for the observed traits in white yam genotypes *.

Variable	Factor	FA1	FA2	FA3	FA4	Xo	Xs	SD	SD%	Communality
Oxi30	FA 3	−0.35	0.21	−0.69	−0.14	0.22	0.14	−0.09	−39.00	0.67
Oxi180	FA 3	0.21	−0.20	−0.76	0.14	0.30	0.22	−0.08	−26.40	0.68
PLNV	FA 4	0.10	0.06	−0.41	−0.76	0.12	0.09	−0.03	−26.20	0.76
ATW	FA 4	−0.17	−0.20	0.28	−0.75	0.22	0.16	−0.05	−24.30	0.72
AUDPCYMV	FA 2	−0.17	−0.89	−0.19	−0.03	1.29	1.04	−0.25	−19.30	0.85
TTY	FA 1	−0.99	−0.02	0.03	0.01	0.66	0.53	−0.13	−19.20	0.97
AUDPCYAD	FA 2	−0.04	−0.92	0.07	−0.05	1.48	1.35	−0.14	−9.08	0.86
DMC	FA 2	−0.10	0.31	−0.13	0.25	0.21	0.21	0.00	−1.09	0.18
Average										0.7368696

* (Xo) = original mean, (Xs) = mean of the selected genotypes SD = selection differential, % = percentage. The bold denotes higher contributing factors in each PCs as explained in the text.

). The four selected genotypes with the lowest MTSI values (Supplementary Table S5) were G18, G19, G24, and G16 (Figure 7a). Figure 7b provides a visual representation of the strengths and weaknesses of the selected genotypes based on the contribution of each factor in the MTSI index. Genotypes G24, G18, and G19, which had the lowest value in the first factor, were close to the ideal genotype for fresh tuber yield (Table 5). Genotypes G24 and G18 had the lowest share of the second factor indicating their proximity to the ideal genotype in terms of AUDPCYMV and AUDPCYAD. Genotypes G16, G18, and G19 had the lowest share of the third factor, indicating their proximity to the ideal genotype in terms of Oxi30 and Oxi180. Genotypes G19 and G16 had the lowest share of the fourth factor, indicating their proximity to the ideal genotype in terms of PLNV and ATW. The selection differential percentage varied from −39% for oxidation intensity at 30 min to −1.09% for dry matter content. The mean of the selected genotypes (Xs) was generally lower as compared to the original mean (Xo) for all genotypes in the population, which makes the selection differential (SD) negative ranging from −0.25 to 0.

4. Discussion

This study assessed the genetic potential of 25 genotypes of white Guinea yam (Dioscorea rotundata) for its agronomic attributes at six environments (3 locations by 2 years). The analysis of these traits revealed a significant genotype-by-environment interaction (GEI) showing the complexity of breeding efforts. The boxplot analysis illustrated variability in the adjusted means among genotypes within each environment, with high variability indicating genotypes with extreme performance values. The varying median values across environments for each trait indicated that environmental conditions influence genotypic performance. These findings highlight the importance of considering environmental influence when assessing genotypic performance for quantitative traits. Traits with high variability may require environment-specific strategies for selection and breeding, whereas traits with low variability suggest more stable performance across different conditions. Similar findings were also reported in yam and rice [27,28,29].

Genetic variance is a critical component in selecting superior genotypes and enhancing desirable traits within a population. Our study identified significant genotype effects across most traits with the exception of plant vigor, suggesting opportunities for genetic improvement through hybridization and selective breeding strategies. This observation aligns with findings in Cassava by Cynthia et al. [30], reinforcing the potential for targeted breeding strategies to enhance genetic gains. However, traits such as tuber yield, YMV resistance, and tuber flesh oxidation showed significant phenotypic variation, with the phenotypic coefficient of variation (PCV) generally exceeding the genotypic coefficient of variation (GCV). This pattern was consistent with earlier studies [11,31], indicating substantial environmental influence on these traits.

The significant differences detected among genotypes for the measured traits highlight the genetic variability within the studied population. A key finding of this analysis is the notable genotype-by-environment interaction (GEI) for most traits, as supported by Adjei et al. [7]. This interaction shows that genotype performance is influenced by both genetic makeup and environmental conditions. A strong GEI for traits like tuber yield and yam mosaic virus can complicate efforts to identify superior genotypes for diverse environments [32], as performance may not be consistent across different conditions [33].

The low broad-sense heritability (0–30%) observed for plant vigor and yam mosaic virus (AUDPCYMV) indicates that genetic differences account for only a small portion of the phenotypic variation in these traits. This suggests that achieving improvements might require more intensive selection or consideration of additional factors. Conversely, the moderate broad-sense heritability for traits, such as total tuber yield, dry matter content, and area under the disease progression curve for yam anthracnose disease implies a stronger genetic influence. These findings differ from previous studies that reported high heritability for tuber yield and yam mosaic disease [34], as well as moderate to high heritability in various yam traits [8,35], adding complexity to interpreting genetic influences on yam traits.

Principal component analysis (PCA) reveals the underlying factors driving variation in quantitative traits among genotypes, consistent with other studies [8]. These insights can guide breeders in selecting parental candidates for targeted breeding to enhance specific traits [35]. Phenotypic traits have been crucial in understanding yam diversity and differentiation [36,37,38], informing more effective breeding strategies.

The correlations among the studied traits showed the strength of their linear associations. This implies that certain traits can be reliable proxy selection parameters for evaluating different genotypes. This correlation analysis offers valuable insights into the interplay among diverse traits. It can significantly enhance the precision of selecting and breeding yam genotypes tailored to exhibit specific attributes.

The AMMI1 and AMMI2 biplots provided a comprehensive view of the genotype-environment interactions for tuber yield and dry matter content. Location x year contributions were considered to be the environment in this study, which is consistent with [39]. The AMMI model is a crucial tool for evaluating the significance of genotype-by-environment (G × E) interactions in agronomic traits across multi-environment trials. This model allows for a detailed understanding of the relationships between genotypes and the environments in which they are tested. The AMMI model 1 biplot is widely used for identifying high potential yield and stability [24]. In the AMMI1 biplot, genotypes positioned on the right exhibit higher yields, reflecting superior performance. This aligns with the model’s utility of identifying high-yielding and stable genotypes [24,40,41].

The IPCA1 and IPCA2 scores are essential for understanding the function of genotype-by-environment interaction and the adaptability of genotypes in the test environments [42]. Different genotypes of yam possess distinct varietal characteristics and preferences for different environmental conditions [43,44] where they can exhibit their full potential which is the basis for the evaluation of genotypes across different environments so as to identify the genotypes that are suitable for a particular environment [44]. Genotypes like G17 showed low mean performance and stability, indicating specific environmental adaptations, while G21 demonstrated high performance and stability across environments. This information assists researchers in making informed decisions about genotype adaptability and potential applications.

The Multi-Trait Stability Index (MTSI) identified G18, G19, G24, and G16 as the best-performing genotypes. Genotypes with lower MTSI values are considered better in performance and stability. The analysis revealed that traits such as total tuber yield contributed significantly to genotype selection, while traits like AUDPCYMV and AUDPCYAD were less influential. MTSI has been effectively used in previous studies to evaluate yield and agronomic traits across various crops [45,46,47,48] and has also been applied to bush-yam [31] and cassava [49]. Multi-trait selection methods, including FAI-BLUP, are useful for identifying desirable genotypes for breeding programs [12].

5. Conclusions

This study explored the genetic potential of twenty-five genotypes of white Guinea yam (Dioscorea rotundata) across six environments The findings revealed significant genotype and environment interactions for the studied traits, with genetic factors playing a major role in trait variability. Each genotype exhibited unique agronomic characteristics, offering promising avenues for targeted selection and breeding programs to enhance specific traits of interest. The application of the Multi-Trait Stability Index validated previous studies, demonstrating its efficacy in evaluating yield and other agronomic traits across different genotypes. These findings offer valuable insights into yam genetics and agronomy, providing a useful basis for refining breeding strategies focused on improving specific traits relevant to the environments studied.