In this study, a total of 285 Double haploid lines, were produced through anther culture of the (F₂ − F₃) generations derived from interspecific crosses among Korean Tongil-type (Semi-dwarfism, high-yielding potential with medium amylose and short grain) and Japonica (Earliness, low-medium amylose, soft and sticky endosperm) with African varieties and land-races (high biomass, disease resistance and low yield) were established at (Tissue culture laboratory Africa-Rice St. Louis). Of these 10 promising Double Haploid lines were selected from on station and on-farm advanced yield trials based on their high yielding potential, submergence tolerance and early maturity along with the local cultivar Umgar as a check were used. The pedigree, source, origin and ecology of the studied genotypes were presented in Table 1.

A field experiments were conducted at three agricultural research corporation stations under different environmental conditions to maximize the influence of the interaction of phenotypic values with environments. The three locations, namely Kosti, Medani and Maatug in 2019 to 2020 rainy seasons, making a total of six environments, are shown in Table 2.

The genotypes were randomly assigned to plots of 3 × 6 m size and consisted of 10 rows spaced 30 cm a part. The seeds were drilled on July 1^sd to 5^th, at all the five environments. Sowing was done by hand drilling on a seed rate of 60 kg/ha and then thinned after seedling establishment to about 3 - 5 cm intra row spaced plants. Fertilizers were applied at the rate of 86 kg N ha⁻¹ and 43 kg P₂O₅ ha⁻¹ in the form of urea and triple super phosphate (TSP), respectively. P₂O₅ was applied as a basal dose during final land preparation. Urea was top dressed in two

equal split doses one at 21 days after sowing and the other before panicle initiation. Irrigation at every 3 - 7 days intervals was kept under the saturated submerged condition at first four weeks and drainage was conducted whenever heavy rains resulted in pounding at this stage then whole plots were flooded gradually until reaching the peak (10 to 20 cm) at reproductive and ripening stages. In all environments, a uniform package of practices and necessary precautions were taken to maintain the crop in its best condition.

Hand weeding was performed three times per season across the six environments following the local recommendations. Insect damage and disease incidence were not recorded during the cropping season at all environments.

Data on quantitative traits at appropriate growth stages of rice plants were collected according to the standard evaluation system (SES) for rice, [17] . Data on days to flowering (DF) (days from sowing time when the 50% of the plants reached flowering), days to maturity (DM) (days from sowing time when 50% of the panicles reached full maturity), number of panicle/m², 1000 grain weight (g) and grain yield (t/ha) were collected on a plot basis. At harvest, ten randomly selected plants in the middle of the row of each plot were used to collect data on plant height (cm), panicle length (cm), number of tillers/plant, number of filled grains/panicle, number of unfilled grains/panicle.

The trials were conducted using a randomized complete block design (RCBD) with three replications.

Analysis of variance (ANOVA) for individual experiments was carried out to assess the genotypic effects and their interaction for all studied characters using the general linear model (GLM) procedure. A combined analysis on mean grain yield and its components was conducted after testing the homogeneity of variances. The GE interaction of 11 rice genotypes over the six environments were assessed by the AMMI model as proposed by [18] , using the statistical analysis program GenStat 12^th edition [19] .

The AMMI model is:

Y g = μ   + ∝ g +   β e + ∑ n   = 1 N λ n γ g δ n + p g

where Y g = yield of the genotype (g) in the environment(e); μ = grand mean; ∝ g = genotype mean deviation; β e = environment n = mean deviation; N = No. of IPCAs (Interaction Principal Component Axis) retained in the model; λ n = singular value for IPCA axis n; γ g = genotype eigenvector values for IPCA axis n; δ n = environment eigenvector values for IPCA axis n and p g = the residuals.

The combined analysis of variance for yield and other agronomic characteristics is presented in Table 3. The main effect of genotypes mean squares was significant for number of panicle per m², number of filled grain per panicle and number of unfilled grain per panicle (0.05) and highly significant for other agronomic traits (0.01), indicating that genotypes differed in their genetic potential for

*, ** significant at 0.05 and at 0.01 probability level, respectively.

these traits. Main effects of season × genotype mean squares were highly significant for all studied traits, indicating varietal differential seasonal response and thus agree with the results found by [13] . Mean square for genotype × location was highly significant (p < 0.01) for grain yield and its most components indicating that some genotypes performed differently in different locations.

Mean grain yields ranged between 5.77 tha⁻¹ for k150147 to 4.12 tha⁻¹ for UMGAR (Table 4). k150147, k150870, k150032, k150912, and k150307 had significantly higher grain yield than the local check Umgar over the six environments. They produced a grain yield increase of 40%, 38%, 34%, 34% and 33% over Umgar, respectively (Table 5).

Grain yield stability and wide adaptation of rice varieties are becoming increasingly important in mitigating the adverse effects of climate change, which is becoming more prominent over the last years. The analysis of variance as per AMMI model revealed that there was significant contribution for variation by main effects (genotypes and environments) and interaction effects for yield (Table 6). Significant mean sum of squares due to genotypes indicate that there existed genotypic differences and significance of environment explains that environmental effects differ across different locations and test locations were diverse. In addition, G × E interaction effects signify that genotypes behave differently across different environments. Grain yield showed that 43.2% of the total sum of squares was attributable to genotypes, 44.2% to environments and 12.6% to G × E effects (Table 6). A large sum of squares for genotypes (43.2) near to environments (44.2) indicated that the genotypes were diverse, with large

*, ** Significant 0.05 and at 0.01 level of probability, respectively.

*, **, *** indicate highly significant at 0.05, 0.01 and 0.001 respectively. GEI = Genotypes by Environments Interaction. IPCA = Interaction Principal Component Axis.

differences among environmental means causing most of the variation in grain yield, also genotypic has a strong influence on grain yield. Present results are in harmony with results of [20] , who observed that mean sum of square due to genotypic main effect was high for grain yield when rice genotypes were evaluated in different growing seasons. These results suggest that variations in environment means are mainly due to location differences than seasonal variations. In the other hand our study on lines with [21] , he evaluated fifteen upland rice varieties in two locations and they found that environmental mean variations were higher than genotypic mean variations for yield and hence the tested locations were diverse. The magnitude of the genotypes sum of squares was also higher than that for G × E interaction, indicating that there were substantial differences in genotypic response across environments, which is in agreement with previous reports of [22] .

The multiplicative variance of the treatment sum of squares due to G × E interaction was further partitioned into two principal components IPCA1 and IPCA2. The two principal components were highly significant for grain yield. The percent contribution of each of the IPCAs to the sum of squares of the G × E interaction for grain yield is tabulated in Table 6. [23] proposed that two IPCAs for AMMI model were sufficient for a predictive model. Therefore, the interaction of the eleven rice genotypes evaluated over the six environments were best predicted by the first two principal component axes, which explained 90.33% of the total interaction effects of which the contribution of PC1 was 49.61% and that of PC2 was 34.02%. The rest of the variations were taken as residual effect. Therefore, IPCA2 vs IPCA1 biplots were generated to illustrate the genotype and environment effects simultaneously (Figure 1).

The PC 1 biplot indicated that the environments E3, E1, E2 and E5 were high yielding locations with high additive genotypic main effects, while the yields in the other environments were below the environmental mean. The scatter plot of the genotypes in this biplot indicated that genotype G7, G1, G2, G6 and G4 were the highest yielding genotypes. The AMMI biplot indicated that the genotypes G6, G4, G3, G7 and G5 were located close to the origin and proved highly stable, although their mean yields were on the high side and should therefore be recommended. Similar results regarding the stability of genotypes due to first two IPCAs values were recorded by [22] Vaezi et al., (2017) and [24] Baraki et al., (2020). The genotypes G11, G10 and G9 located far away from the origin, were unstable and expressed higher positive or negative GxE interactions. The AMMI- based stability parameter, the AMMI stability value (ASV), was calculated based on the first two PCAs to produce a balanced measurement between them and can be useful in situations where the two first IPCs explain a considerable part of the GxE interactions (Table 5).

According to ASV, genotypes k150912, k150870, k150307 and k150147 were identified as high yielding and stable for having lower ASV values, whereas genotypes k150874, k150223 and k15451G5 were identified as being more stable and low-yielding genotypes based on the genotype stability index, which takes both the overall mean yield and ASV into consideration (Table 5). Similar results were reported by [22] , who studied AMMI stability parameters to identify stably performing barley lines. According to the AMMI-based stability results, the four genotypes k150912, k150870, k150307 and k150147 are considered as prospective DH genotypes having higher grain yield, stability and wide adaptability and deserve to be released for irrigated system condition (aerobic rice) commercial production in the central clay plains of the Sudan.

Flowering and maturity: Genotypes differed significantly in flowering and maturity times, 50% flowering and maturity ranged between 76 and 91 days and 104 and 113 days after sowing, respectively, (Table 7). Generally, genotypes were harvested between 98 to 120 days after planting or 21 days after 50% flowering. Plant height: Ranged from 69.7 to 95.5 cm for KF150659 and Umgar, respectively. According to IRRI, aerobic rice plant height is classified as semi-dwarf if it is less than 110 cm. In the present study, most of the tested genotypes are classified under semi-dwarf class (Table 7). This indicated that the tested genotypes had inherent variability in stature to develop lodging resistant rice and will have higher response to nitrogen fertilization similar result was reported by [25] .

(DF) = days to 50% flowering, (DF) = days to maturity, (PH) = plant height, (PL) = panicle Length, (NT) = number of tillers per plant, (NP) = number of panicles per meter square, (FG) = number of filled grains/panicle, (UG) = number of unfilled grain/panicle, (TGW) = 1000 grain weight.

Panicle length: Significantly different among genotypes at each location (Table 7). The grand mean was 21.8 cm, genotype KF150659 having the shortest (19.4 cm) and KF150912 has having the longest panicle length (24.0 cm). The number of productive tillers per plant: Significantly varied with the genotype at each locations and all combined location (Table 7). It ranged from 3 for KF150659 and KF150426 and 6 for KF150912. These numbers are relatively low compared to other findings. For instance, [26] reported number of productive tillers in the range of 4 - 7 tillers per plant. The low number of productive tillers could be due to soil type and fertility and also could be due to water stress especially waterlogging. Number of panicles/m²: The highest number of panicles/m² of 373, 366 and 363 were exhibited by genotypes KF150870, KF150912 and KF150147, respectively, while, the lowest of 291 and 300 were exhibited by the genotypes KF150659 and KF150874, respectively (Table 7). For high grain yield in rice, the plant should have large number of panicles per m² with good grain filling and panicle size. Number of filled grains/panicle: The highest number of filled grains per panicle was recorded for KF150032 (101), KF150307 (95), KF150912 (93), KF150870 (92) and KF150147 (92). While the lowest for KF150426 (80) (Table 7). This trait offers a prime scope for the development and selection of high yielding upland rice genotypes. Number of unfilled grains per panicle: In order to increase yield, it is important to reduce spikelet sterility or increase spikelet fertility. Genotypes differed significantly in number of unfilled grains which ranged between 13 and 22 (Table 7). This could be due to drought and high temperature during flowering time. The 1000-grain weight: There is significantly different among genotypes (Table 7). Umgar recorded the heaviest 1000 grain weight (25.9 g), while KF150451 recorded the lightest (21.5 g) over the six environments. According to the standard measurement of the international Rice Research Institute (IRRI), the highest yielding varieties range between 25 to 28 g for 1000 grain weight, which indicates variability among the tested genotypes in their efficiency of partitioning, assimilates into grain yield.