Evaluation of yield and yield components of some groundnut genotypes under rainfed condition in Mali using biplot analysis

Sixteen groundnut genotypes were evaluated under three different environmental conditions of Mali during 2015/2016 rainy season. The environments were Tioribougou, Djijan and Samanko affiliated at International Crops Research Institute for the Semi-Arid Tropics (ICRISAT). The experiments were conducted in a 2×8 alpha lattice with three replications. Within the environment, the main effect of a genotype was significant. Genotypes by environmental interaction were also significant. Results showed that three genotypes ICGX-IS 13005F2-B1-287-1 (grain yield 2,197.1 kg/ha), ICGX-IS 13005F2-B1205-1 (grain yield 1,922.3 kg/ha) and ICGX-IS 13012F2-B1-29-1 (grain yield 2,106.0 kg/ha) were found to be stable across environments. The genotypes with high pod yield for each specific environment were ICGX-IS 13005F2-B1-287-1 with 2197.1 kg/ha in Tioribougou and ICGX-IS 13005F2-B1-252-1 with 2382.8 kg/ha in Samanko.


INTRODUCTION
Groundnuts are main food product for thousands of small scale farmers in Mali. FAOSTAT (2015) data on groundnut production showed that, for the country, the increase in groundnut production resulted from the expansion of the area cultivated and increase in the number of farmers involved in the production rather than the increase of yield productivity per se for more than 50 years. Other studies indicated that severe drought occurred twenty times during 55 years in Mali (Ndjeunga et al., 2003). Moreover, SAS Institute 2009 argued that in West Africa, groundnut productivity follows similar trends with the rainfall pattern, which is mostly dominated by the occurrence of recurrent drought spells. The groundnut production is rain-fed in Mali and most of the time; the growing phase terminates under a period of drought. This exposes the crop to end-of season drought causing yield loss ranging from 44 to 85% in Mali (Sanogo et al. unpublished). Though, breeding for drought tolerance should be an important objective for breeders in the Sahelian region, Ravi et al. (2011) proposed the development of varieties that are better adapted to waterlimited conditions. One of the most important steps in a breeding programme is the assessment of new breeding materials in target environments. Assessing the performance of breeding lines is crucial because it involves both breeders and farmers and allows critical judgments from both sides before the acceptance of the promising lines by farmers. Testing of potential new genotypes in diverse environments can reveal wide adaptation, or yield stability of cultivar, and good mean performance (Simmonds, 1991). According to Kang et al. (2006), genotype-by-environment interaction (GEI) is expected in trials that involve cultivars with diverse genetic backgrounds in diverse test sites. Estimates of significant GEI effects is an indication that genotypes respond differently under different growing conditions, suggesting that environments have effect on the genotypes' performance (Worku et al., 2001). In Sub-Saharan Africa, groundnut breeders generally select genotypes with high pod yield to specific environments. GEI is common under drought situations (Bänziger et al., 2004), and drought events are unpredictable and occur randomly. In a GEI study across Niger and India, Hamidou et al. (2012) observed a high GEI between the two countries suggesting the need for environmentspecific selection in groundnut. The present study was designed to assess the new developed drought tolerant groundnut lines along with their four parental lines and the two landraces and to identify promising genotypes across three different locations in Mali. The specific objectives were to: (i) determine the importance of genotype x environment interaction in the different genotypes, (ii) identify varieties with stable performance or specific adaptation to be recommended to farmers in Mali for cultivation.

Genetic material
Sixteen genotypes comprising 10 F 2:4 drought tolerant groundnut and six controls (4 parental lines and 2 landraces) were evaluated in this study (Table 1).

Field location and experimental conditions
The ten superior elites genotypes and their four parents were evaluated at Tioribougou (580 m above sea level masl, latitude 13°22´ N, longitude 7°69´ W), Djijan (320 m asl latitude I3°05' N, longitude 9° 28' W) and Samanko ICRISAT Research Station (331 m, latitude 12°54'N, longitude 8°40'W), during the rainy season 2015/2016. The average annual rainfall at Tioribougou, Djijan and Samanko affiliated with ICRISAT are 700, 900 and 800 mm, respectively. Currently, two local controls: 47-10 and Tigaba which are the most popular in Mali were included. All experiments were laid out as 8×2 alpha lattice designs with 3 replications. Each plot consisted of a row with a length of 4 m with 60 cm within-row spacing and 15 cm between-row spacing. Prior to planting, basal fertilizer was applied as 100 kg ha -1 Simple Super phosphate. Standard cultivation practices, including hand planting, hand weeding. Data were analyzed in analysis of variance (ANOVA) across locations using PROC GLM in SAS (SAS Institute, 2009). Replications, Environments' and incomplete blocks were considered as random effects while genotypes were considered as fixed effects. PROC CORR was used to perform correlation analysis. The GGE biplot analysis (Yan and Tinker, 2006) was performed using GenStat software Edition 15 for grain yield means adjusted for block effects to obtain information on the genotypes in the three experimental environments. It was also applied to estimate genotypic mean performance across environments for stability analysis. GGE biplot analysis allows the disintegration of the G x E interactions. The first two principal components (PC1 and PC2) were constructed in GGE biplot that were derived from subjecting environment-centered grain yield means for each location to singular value disintegration. The data were not transformed ('Transform=0'), but standardized ('Scale=1), and were environment-cantered (Centering=2').

Mean squares from ANOVA for yield and drought traits in 16 groundnut lines
Results of the analyses of variance combined across environments and for each environment are presented in Table 2. ANOVA revealed highly significant (P<0.01) effects of genotypes for 50%flowering, hundred seed weight, harvest index, shelling percentage whereas genotypes effects were significant (P<0.5) for pod and grain yield (Table 2). Environment effects were highly significant (P<0.01) for 50% flowering, hundred seed weight, harvest index, shelling percentage, pod yield and grain yield (Table 2).

Environmental mean performance and coefficient of variation of traits in 16 genotypes
By comparing the three environments for their mean performance across the 16 genotypes, higher values were recorded in Djijan compared to Tioribougou and Samanko for most of the traits except for 50% DF, pod yield and grain yield (Table 3). Fifty percent date to flowering of plants ranged from 25 days in Tioribougou and Samanko to 26 days in Kita. Biomass production varied from 290 g in Djijan to 172.20 g in Samanko.  Pedigree  Source  Status  G1  ICGX-IS 13005F2-B1-205-1  ICGV 91317/ICGV 87378  ICRISAT  F 4  G2  ICGX-IS 13005F2-B1-252-1  ICGV 91317/ICGV 87378  ICRISAT  F 4  G3 ICGX
Harvest index ranged between 79.30% in Tioribougou and 67.70% in Samanko. Hundred seed weight ranged from 29.50 g in Samanko to 40.90 g in Djijan. Shelling percentage ranged between 73.90% in Samanko and 74.20% in Djijan and Tioribougou. The overall mean of groundnut pod yield was variable between 1,868 kg/ha in Samanko and 3,406.74 kg/ha in Djijan. Grain yield ranged from 1,980.50 kg/ha in Samanko to 2,365 kg/ha in Tioribougou. Coefficient of variation obtained ranged from 2.32% for SP in Tioribougou to 28.08% for GY in Djijan. The low yielding environment (Samanko) exhibited lower coefficient of variation for several characters than the two higher yielding environments (Djijan and Tioribougou) ( Table 3).

GGE biplot analysis
Figures 1 to 3 provide graphical results of the GGE biplot analysis. The GGE biplot of the grain yield of the 96 genotypes revealed that PC1 explained 61% of the total variation while PC2 explained 28%. The total variation for grain yield of the genotypes across the three environments account for 89% with PC and PC2 ( Figure  1). The GGE biplot was based on genotype-metric  ).

DISCUSSION
The first objective of this study was to evaluate genotypes to random drought. Drought is, though not predictable but occurs frequently in Mali. However, rainfall patterns during the period of the experiment from planting to harvesting across the three environments were favorable for plant growth. As a result, there was no drought spell affecting the crop during the reproductive phase. Hence the true potentials of the newly developed lines under drought are yet to be determined. Therefore the testing of the lines under controlled drought conditions is necessary to confirm the presence of drought tolerance alleles in the newly developed lines. In the present study, correlations were calculated in each environment separately and also based on the mean values obtained from the pooled analysis. The 50% DF and hundred seed weight were correlated with grain yield in one environment but not in other environment indicating that environment might have some influence. Therefore, selection of these traits would need attention for further characterization and improvement. This suggests that these traits are not stable across environments and are therefore unreliable as secondary traits for genetic improvement of grain yield. Pod yield, shelling percentage, harvest index and biomass production exhibited a positive and significant (<0.5) phenotypic correlation with grain yield in all the environments or at least two environments indicating selection of these traits will simultaneously improve the yield. Similar results were obtained by Sharma and Varshney (1995), Johnson et al. (2008) and Ravi Kumar et al. (2012) who reported a significant and positive correlation of grain yield with pod yield. Wunna et al. (2009) detected strong correlation between grain yield and shelling percent, biomass production. The results from the current study were not in agreement with other findings such as those reported by Uddin et al. (1995) who found negative correlation between 100-seed weight and grain yield while Sumathi and Ramananthan (1995) and Sah et al. (2000) showed positive correlation between 100-seed weight and grain yield in groundnut. The correlation between grain yield and several characters in groundnut were reported by a number of investigators. For example, Sharma and Varshney (1995), Johnson et al. (2008) and Ravi Kumar et al. (2012) reported significant and positive association of grain yield with pod yield. Wunna et al. (2009) detected strong correlation between grain yield and shelling percent, biomass production. The results of the current

vs Controls
*** ns ns * ** ns * ns = non-significant, *P<0.05, **P<0.01, ***p<0.0001, respectively. 50% DF= Day to 50% flowering (days), SCMRf = SPAD meter reading at 60 DAS, SCMRz = SPAD meter reading at 80DAS, SLAf = Specific leaf area (cm 2 g -1 ) at 60 DAS, SLAz = Specific leaf area (cm 2 g -1 ) at 80 DAS, HSW = hundred seed weight (g), SP = Shelling percentage (%), PY= pod yield (kg ha -1 )and GY = grain yield (kg ha -1 ). study have some implications in breeding for agronomic and yield components traits in groundnut in Mali. Grain yield associated negatively with 50% DF suggests delayed crop growth may reduce grain yield. This is probably because of the possibility that the crop may be exposed to several prevailing abiotic stresses like drought and thereby reducing yield (Nagabhushanam, 1981). This scenario should be avoided in the major groundnut cultivation areas in Mali such as Kolokani by selecting drought tolerant materials or short duration varieties. The correlation between grain yield and other yield components indicated that possible superior segregants could be selected in combinations of traits such as days to 50% flowering, hundred seed weight and shelling percentage. The multi-environmental trial analyses conducted in three locations Djijan, Tioribougou and Samanko revealed that genotypes, environments and their interactions were significant for almost all traits which agreed with a number of previous results such as Nath and Alam (2002), Bucheyeki et al. (2008), Hamidou et al. (2012 and Makinde et al. (2013) working on groundnut. Both of genotypic and environmental differences played important roles in the expression of the different traits with significant GEI reflected in combined analysis. The genotypes revealed a considerable variability for agronomic traits and yield components that could be potentially useful for the improvement of drought related traits as well as yield components. The combined analysis across the three locations for the F 4 progenies compared to the checks revealed improvements made in some traits of interest. This may be due to new genetic combinations produced in the hybridization process. It is important to note that the ten elites genotypes were selected as the top 10 F 4 high performance among 90 drought tolerant genotypes in the breeding process. Each of the ten F 4 genotypes had a good yield compared to controls and the farmers' best control (Tigaba) in the three experimental condition except ICGX-IS 13012F2-576-1. Four genotypes (ICGX-IS 13005F2-205-1, ICGX-IS 13005F2-252-1, ICGX-IS 13005F2-91-1 and ICGX-IS 13012F2-29-1) were identified with significantly higher pod yield compared to the best control (ICIAR 19BT). The best F 4 genotype (ICGX-IS 13005F2-252-1) with 3,278.80 kg/ha represented 24.5% higher yield than the best control ICIAR 19BT. Farmers normally adopt varieties that yield more than their locally adapted cultivars; and meet the preferred traits which differ from one community to another (Gowda et al., 2000). The selection criteria of variety largely depend on the importance of the crop in the farming system and their uses (Abebe et al., 2005). Genotypes out yielding with more than 20% over landraces open the possibility of introduction or replacing with the landrace Tigaba, a late maturing variety. Tigaba cannot be cultivated in low falling periods or drought years. The variety 47-10, the most widely grown variety in Mali, produces a lot of pods with few seeds resulting in low yield. In addition to that, the variety is suspected to have high aflatoxin levels which is damaging to humans and cattle when consumed. These new genotypes should be evaluated for other farmers and end-users before adoption. Each of the ten F 4 genotypes had a good yield performed well compared to mean controls and the farmers' best check (Tigaba) except ICGX-IS 13012F2-576-1. The genotype-by-environment interactions effects on pod yield were significant. Genotype-by-environment interaction effects on pod yield have been reported in a number of previous studies but the most recent results from Bucheyeki et al. (2008), Jogloy et al. (2009), Zhang et al. (2011), Balota et al. (2012 and Makinde et al. (2013) were consistent with the present results. The presence of genotype by environment interactions resulted in the yield performance of genotypes and ranking of genotypes varying from one environment to another, which suggests that environment specific varieties could be selected. Genotypes that has significantly a better yield than the controls were identified specifically for each environment including ICGX-IS 13005F2-B1-287-1 (G3) for Tioribougou, ICGX-IS 13005F2-B1-205-1 (G1), ICGX-IS 13005F2-B1-252-1 (G2) for Samanko and ICGX-IS 13012F2-B1-29-1 (G6) for Djijan. In the three experimental environments, genotype ICGX-IS 13012F2-B1-29-1 (G6) whereas ICGX-IS 13005F2-B1-205-1 (G1) and ICGX-IS 13005F2-B1-287-1 (G3) had high yield and high stability.

Conclusion
The genotype by environment interaction influenced drought and traits related to yield or yield components. Some environments were better than others in the expression of the quantitative traits. The difference in ranking genotypes based on yield performance across the three environments confirmed the presence of the genotype by environment interactions. The genotype by environment interaction analysis using GGE biplot method was powerful to visually analyze the yield performance of genotypes in the three locations. Genotypes that performed significantly better than the checks were identified specifically for each environment including ICGX-IS 13005F2-B1-287-1 (G3) for Tioribougou and ICGX-IS 13005F2-B1-252-1 (G2) for Samanko. Genotypes showing little crossing-over interactions and great potentials for high genetic gains were ICGX-IS 13012F2-B1-29-1 (G6), ICGX-IS 13005F2-B1-205-1 (G1) and ICGX-IS 13005F2-B1-287-1 (G3). Six new drought tolerant genotypes had a good yield of up to 18% over the six controls and three of the same genotypes showed yield superiority of 36% six checks. Increase in yield of the new drought tolerant genotypes over the controls showed substantial achievement in breeding for tolerance to drought. The new genotypes will contribute to enhancing the production of groundnut in Mali especially in these areas experiencing cycles of drought. Despite these substantial results, further generations and evaluations will enable better homogeneity (fixation of alleles) of lines and clear varietal recommendations to farmers.