ADDITIVE MAIN EFFECTS AND MULTIPLICATIVE INTERACTIONS ANALYSIS OF YIELD PERFORMANCES IN COWPEA GENOTYPES UNDER UGANDAN ENVIRONMENTS

Agbahoungba Symphorien 1,2 , Karungi Jeninah 1 , Talwana Herbert 1 , Badji Arfang 1 , Kumi Frank 1 , Mwila Natasha 1 , Edema Richard 1 , Gibson Paul 1 and Rubaihayo Patrick 1 . 1. Department of Agricultural Production, College of Agricultural and Environmental Sciences, Makerere University, P.O. Box 7062, Kampala, Uganda. 2. Laboratory of Applied Ecology, Faculty of Agronomic Sciences, University of Abomey-Calavi, 01 PO Box 526, Cotonou, Benin. ...................................................................................................................... Manuscript Info Abstract ......................... ........................................................................ Manuscript History


ISSN: 2320-5407
Int. J. Adv. Res. 5 (6), 349-360 350 Among the African biggest cowpea producers, Uganda is ranked 8 th with production of about 84,000 metric tons (Ddamulira et al., 2015). This volume of production demonstrates the importance of cowpea cultivation as a component of Ugandan farming system and with cultivation expanding beyond northern and eastern regions traditionally known for the crop in recent time (Karungi et al., 2000a;Ronner et al., 2012). In Uganda, cowpea is ranked 4 th legume after beans, groundnuts, and soybean (Ronner et al., 2012). Cowpea is mostly grown in the drier eastern and northern parts of Uganda (Dungu et al., 2015) because of its tolerance to drought and adaptation to warm weather, hence enabling it to produce significant yield where other legumes like beans fail to grow (Bisikwa et al., 2014).
Farmers-traditional cultivars are known to be well adapted to the low input conditions, but generally poor in yield and highly susceptible to the major diseases and pests. These production constraints are the main target of cowpea breeding program both at the national and regional levels in sub-Sahara Africa. Although past research efforts have brought some improvement into farmer's yield, available statistics still indicate significant instability in yield across locations and years (FAO, 2013) and yield at farm-gate is far below optimum.
Yield has been described as a complex phenotypic trait in plants because of being a final aggregate product of many interwoven physiological and development traits controlled by different arrays of genes. Understanding interrelationship between yield and environments (Nwofia, 2012) is vital to achieving high and stable yield. In addition, unregulated seed distribution system and research centers resource limitations have led to poor release and distribution of 'improved' varieties with doubt on yield stability. This problem could be addressed through a decentralized system where improved lines from research and seed centers are subjected to post-varietal-release evaluation to ascertain genetic stability, in this case yield.
The venture to develop high yielding genotypes on par with traditional cultivars is still in progress. The developed cultivars adapted to a wide range of environments, is the eventual goal of plant breeders. Hence, pattern of response of genotypes is studied by the plant breeders by testing genotypes in different environments to study genotype x environment (GxE) interaction. To estimate the level of interaction of genotypes to environments and to eliminate as much as possible the unexplainable and extraneous variability contained in the data, several statistically techniques have been developed to describe GxE and measure the stability of genotypes.
Since GxE interaction is naturally multivariate, the Additive Main effects and Multiplicative Interaction (AMMI) offers an appropriate first statistical analysis of yield trial that may have a GxE interaction (Zobel et al., 1988). The objectives of this study were to assess the extent of GxE interaction and to select the stable genotypes of cowpea genotypes in Ugandan environments over seasons.  Table 1. The cowpea cultivars used in this study were obtained from cowpea collection at MUARIK (Table 2). Eight breeding lines from International Institute of Tropical Agriculture (IITA), 16 breeding lines from Uganda, and 48 Ugandan landraces were used.

Experimental Design:-
The genotypes were planted in an alpha lattice design (8 blocks x 9 genotypes per block) with two replications in three locations and two seasons. The first planting was done in September 2015 and the second in April 2016. Three seeds were planted per hole and the seedlings were thinned to two plants per stand 10 days after sprouting. Each plot consisted of 4 rows of 5m long and 0.75m apart with an intra-rows space of 0.25m. Regular weeding of the fields till maturity were done with hand hoe.
The cultivars were given protection against aphids during the vegetative stage by spraying with the insecticide chlorpyrifos (as Ascoris 48 EC) applied at the rate 2.5 g (a.i.) ha -1 once at 15 days after planting. They were also given protection against podding stage pests, by spraying with λ-cyhalothrin (as Karate 2.5 EC) applied at the rate 2.5 g (a.i.) ha -1 using a CP-15 knapsack sprayer at 50 % podding (Abudulai et al., 2006).

351
At maturity, the whole plot was harvested and the total dried grain weight (g) was taken using electronic weighing scale. Harvesting was done twice and the yield (kg.ha -1 ) was estimated from the total dried grain weight per plot.

Statistical Analysis:-
A combined analysis of variance was done on the yield data across locations and seasons using linear mixed model (REML) procedure in GenStat 12.0 software (Payne et al., 2009). The model described by Smith et al. (2005) was used as follow: Where, is the observed value for the i th genotype from j th location, k th season, m th block nested within the l th replication; μ is the general mean effect; is the i th genotype effect (considered as fixed); is the j th location effect (considered as fixed); s k is the k th season effect (considered as random); r l is the l th replication effect (considered as random); is the effect of m th replicated nested within the l th replication (considered as random); is the k th season and j th location interaction effect (considered as random); is the interaction effect of k th season and i th genotype (considered as random); is the interaction effect of j th location and i th genotype (considered as random); is the effect of the three-way interaction between k th season, j th location and i th genotype (considered as random); and is the experimental error considered as random.
Yield means were separated using Least Significant Difference (LSD) at 5% level.
To establish the adaptability and the stability of the genotypes to the different environments, an additive main effects and the multiplicative interaction (AMMI) analysis was performed. AMMI is a unified approach that fits the additive effects of genotypes and the environments by the usual analysis of variance and then describes the nonadditive parts by principal component analysis fitted to the AMMI model according to the following equation: ̃ Where = yield of genotype g in environment e for replication r, =grand mean; =mean deviation of the genotype g (genotype mean minus grand mean); and =mean deviation of environment mean; =the eigenvalue of the principal component (IPCA) axis n; ̃ = the genotype g eigenvector value for IPCA axis n; = the environment e eigenvector value for IPCA axis n; = the residual; and = the random error (Zobel et al., 1988).

Results:-Cowpea Grain Yield as Influenced by Genotypes, Locations and Years (Seasons):-
The results from the analyses of variance are presented in Table 3. Genotypes, locations, and seasons significantly (P<0.001) affected the grain yield in cowpea. In terms of interactions, genotypes significantly (P<0.001) interacted with locations and with seasons for cowpea grain yield. The three-way interaction (Y x L x G) effects were also highly significant (P<0.001). The factors explained revealed that cowpea grain yield was mostly affected by locations (72.06%), genotypes (5.73%) and the interaction year x location x genotype (2.75%).

Additive Main Effects and Multiplicative Interactions Analysis of cowpea grain yield:-
The AMMI analysis of variance for cowpea grain yield (kgha -1 ) of the 72 genotypes tested in six environments showed that 69.16% of the total sum of squares was attributable to environmental effects; only 5.36% to genotypic effects and 12.74% to G x E interactions effects ( Table 6).

Discussion:-Cowpea Grain Yield as Influenced by Genotypes, Locations and Years (Seasons):-
Evaluation of cultivars in contrasting environments and across years is an essential step in determining their desirability and cultivars with average response across the environments that have a wide scope of adaptation. Expression of wide genetic variability recorded in this study offers opportunity for quality improvement that would allow selection of individuals with better attributes for cowpea grain yield. Data on wide genetic variability in cowpea for grain yields are well documented (Idahosa et al., 2010;Manggoel et al., 2012;Nwosu et al., 2013). The analysis of variance across environments showed highly significant (P<0.001) genotypic effects for the grains yield. Furthermore, the mean sum of squares of environments and genotype x environment interaction were significant indicating broad range of diversity existed among the genotypes across the tested environments (Anandan et al., 2009). Such statistical interaction resulted from the changes in the relative ranking of the genotypes or changes in the magnitudes of differences between genotypes from on environment to another (Tarakanovas and Ruzgas, 2006). The significant L x G effects (P<0.001) demonstrated that genotypes responded differently to the variation in the environmental conditions of location and indicated the necessity of testing cowpea varieties at multiple locations. The significant Y x L x G effects (P<0.001) showed that cowpea yield largely depends on climatic conditions, in particular on the seasonal variation of temperature and the total precipitation in the experimental years. Similar results have been reported by Ddamulira et al. (2015) while evaluating the genotype by environments interaction effects on Brazilian cowpea yield in Uganda.
The best performing cultivars with regard to yield across locations and years was MU9 (854.68 kgha -1 ). The high yield obtained at Arua 2016A was explained by the rainfall pattern that occurred in this area compared to two other environments where the experiments were conducted. Arua has a bimodal rainfall pattern with much longer first season rains. The rain was received during cowpea germination, vegetative and reproductive stage, yet sufficient soil moisture during the reproductive stage is known to enhance grain filling which result into increased grain yield as reported by Faisal and Abdel (2010) and Agoyi et al. 2017). On the other hand, Serere has been reported to be an ideal environment for cowpea production in Uganda with its sandy loamy soil suitable for proper and healthy cowpea growth because it does not restrict root development, has good aeration and drainage (Ecocrop, 2009; Directorate Agricultural Information Services, 2011). But in this study, the lowest cowpea grain yields were recorded in Serere 2015B and could be explained by the severe diseases infestation experienced in that year especially, scab and rust diseases. These disease infestations could be attributed to the fact that the trial of 2015B was set on the previous year site, so there may have been disease build up in the soil prior to planting. Different trends were reported earlier on 29 cowpea genotypes under diverse Ugandan environments by Ddamulira et al. (2015), who observed high cowpea yield in Namulonge and Serere. This shows the difficulties encountered by breeders in selecting new genotypes for release; these difficulties arise mainly from the masking effects of variable environments (Goncalves et al., 2003;Tarakanovas and Ruzgas, 2006). Thus, it is important to study adaptation patterns of genotypes response and their stability in multi-location trial.

Additive Main Effects and Multiplicative Interactions Analysis of Cowpea Grain Yield:-
In the current study, the contribution of the environment to the total variation was higher than the effect of the genotypes and genotype by environment (G × E) interaction. The environments were diverse and caused the greatest variation in grain yield. The AMMI analysis for the grain yield indicated that G x E interaction effects was highly significant (P < 0.01) with a sum of square 2.3 times larger than that for genotypes, which determined sustainable differences in genotypic responses across environments. Similar results were reported by Ddamulira et al. (2015) while evaluating the grain yield and protein content of Brazilian cowpea genotypes under diverse Ugandan 358 environments. In this study, the G x E interaction effects were partitioned in the two first principal component axes (IPCA1 and IPCA2). The two first IPCA explained 92.41% of the interaction sums square (Table 6). This implied that the interaction of cowpea genotypes with the six environments was predicted by the two first component of genotypes and environment which is in agreement with the findings by (Gauch and Zobel, 1996) who recommended that the most accurate model for AMMI can be predicted using the first two IPCAs. However, this contradicts the findings by Asio et al. (2005) while evaluating the local and improved cowpea genotypes in Uganda. These results indicate that the number of terms to be included in an AMMI model cannot be specified prior without first trying AMMI predictive assessment as reported by Kaya et al. (2002). In general, factors like type of crop, diversity of the germplasm and range of environmental conditions will affect the degree of complexity of the best predictive model (Crossa et al., 1990;Ddamulira et al., 2015).
The Interaction Principal Component Axes (IPCA) scores of a genotype in the AMMI analysis indicate the stability of a genotype across environments. The closer the IPCA 2 score are to zero, the more stable the genotypes are across their testing environments. Considering only the IPCA 2 scores it became clear that the genotypes WC30, NE 45, NE 31, NE 51 were the most stable genotypes, they were well adapted to high yielding environments that are more favorable. These genotypes have good potential for genetic improvement of Ugandan cowpea germplasm. The cultivars NE21, NE32, WC55, and WC17 posed close to zero of IPCA1 showed that they are more stable but with lower yield than WC 30, NE 45, NE 31 and NE 51. The cultivars NE48, NE5 and WC 26 had a yield significantly over grand mean grain yield and had an IPCA 1 score greater than the other cultivars but were less stable and may be characterized by specific adaptation in favorable environments. The biplot also showed the yield of a variety at individual sites. For instance, the cultivars NE30, NE67, NE48, WC48, WC26 and NE5 were best for high-yielding environments E2 and E5 (Arua 2015B and 2016A). The specific adaptability to certain environments possibly explained the highest variation in their grain yield. It is presumed that although in certain environments, NE30, NE67, NE48, WC48, WC26 and NE5 yielded highly, in other environments like Serere, the same genotypes might be less adapted due to limited ability to mobilize growth resources which reduces on their ability to produce high dry matter and grain yield as reported by Ddamulira et al. (2015). With respect to the test environments, E1 (MUARIK 2015B) was most discriminating as indicated by the longest distance between its marker and the origin. Thus, MUARIK could be recommended as best environment for cowpea genotypes evaluation. In addition, the length of a genotype vectors reflects the amount of interaction for that genotype. Thus according to Fig.1B, most of the GEI is due to the fact that the genotype MU9 has grain yield beyond average and large IPCA 2 score value in the trial. A similar result on the genotype MU9 was reported by Asio et al. (2005) in Uganda. As a result, this genotype is most suitable for poor environments.

Conclusion:-
This study was conducted to understand the yield performance of cowpea genotypes under diverse environments in Uganda. The grain yield varied based on the genotypes, environments and their interactions. Although genotype MU9 had the highest yield, it was only adapted to specific environments and could be used in those specific areas. Hence, genotypes WC 30, NE 45, NE 31, NE 51 were high yielding, stable and adapted to the environments tested, and should be recommended for genetic improvement of cowpea germplasm in Uganda. In terms of environments, the best grain yield was obtained from Arua, which implied that this environment was favorable for growing cowpea lines in Uganda. The genotype x environment interaction also affected grain yield which implied that, the grain yield of cowpea differed based on different environmental factors (soil types, temperature and rainfall).