Chili (Capsicum frutescens L.) is an important vegetable and spice crop cultivated throughout Ethiopia especially in South, Central and South West part of the country. The small fruited chili locally called ‘Mitmita’ are either prepared as crushed fresh or powder of dry fruits used for special local sauce preparation, to eat raw meat or eaten with local bread/injera for its unique pungency which adds value in local food preparations. Thus, high quality is very important in chilies that are used in different food preparations; for its high pungency, bright red color, high oleoresin concentration and good seeds in the fruit are the main characters on which quality is based and priced for used as spice in the local market. Apart from other environmental factors the production and quality of chili is governed by the inherent genetic potential of the genotypes. In Ethiopia, chili has become almost an essential ingredient of the daily diet of the rich and the poor societies. It is an important commercial product supplied to the local market and exported to different countries. In its major area of production, with hot pepper has a huge potential for improving the income and livelihood of thousands of smallholder farmers and can plays a vital role for food security in Ethiopia. In some parts of the country where pepper and chilies are dominantly grown, sales from these crops contribute 50-60% of the household income, as the green fresh fruits fetches good price and sold at Ethiopian Birr $80-100 per kg in the retail market. Therefore, it is important to produce suitable, high yielding genotype with acceptable agronomic and quality traits. Therefore, varietal evaluation at different locations is very essential for developing either widely or specifically adapted genotype. Therefore, this study was under taken to develop high yielding genotype with desirable fruit qualities for local and export market [1-6].

Five chili genotypes (PBC-586, PBC-401, PBC-067, PBC-142A and Local check) were evaluated under main season with supplemental irrigation in the off season at Melkassa and Debreziet Agricultural Research Centers (MARC and DZT) and Wonji commercial farm for two years (2013-2014) (Table 1) and a trial within a year was considered as one environment and these totally made six environments.

Table 1: Altitude, rainfall, soil type and temperature of experimental sites.

Seedlings were raised on seedbed and transplanted to the field 50 days after sowing. Randomized complete block design (RCBD) with 3 replications were employed, using a plot size of 2.8 m × 3 m at 70 cm inter-rows and 30 cm intra-row spacing. Fertilizers, at a rate of 200 kg/ha DAP applied at planting. Split application of 50 kg/ha Urea at transplanting and the remaining 50 kg/ha side dressed one and half months after transplanting. To control different leaf disease and insect pests, Ridomil MZ 63.5% and Bylaton at the rate of (3.5 kg/ha) and Karate 5% (2.5 l/ha) was applied respectively. All other cultural practices such as weeding, and cultivation were applied as needed.

Data on plant height, days to 50% flowering and fruit, fruit numbers/plant and fruit length, diameter and weight, yield (total and marketable), were recorded. The combined analysis of variance over years and location was computed using SAS/STAT Software 9.2 computer program and mean separation was done using Least Significance Difference (LSD). Analysis of variance, environmental index and linear regression coefficient (bi) measured over environmental index were computed as suggested [7]. Subsequently, the genotype by environment interaction (GEI) was analyzed using the AMMI statistical model by applying principal component analysis to the ANOVA interaction effects.

Pooled Analysis of Variance

The combined analysis of variance of five chili genotypes over six growing environments indicated significant variations of genotypes in fruit yield at (P<0.05) due to their inherent difference in performances. Thus, from the perspective of mean yield, PBC-586 and PBC-142A were the top yielding genotypes, followed by PBC-401. Moreover, MS of the environment source in the ANOVA was significant that showed environments were also diverse, for causing the variation in fruit yield. The GEI mean squares were also significant, indicating genotypes had tremendous variability with respect to their sensitivity to changes in environments.

The overall vegetative performance and field establishment of promising genotypes was satisfactory at all locations, whereas, at 50% flowering date, genotype PBC-067 was found late flowering type, which was directly reflected on fruiting and maturity period.

The overall total yield performances of genotypes in both years showed significant difference, though the performance of all genotypes was higher in the second years of the experiment (Table 2). The overall marketable yield performance was the highest from PBC-586 and PBC-142 (28 q/ha and 27 q/ha) (Table 3) respectively. And these was significantly different from other genotypes. In general, the overall total and marketable yield potential and fruit characters of PBC-586 and PBC-142A were very satisfactory under all testing sites of the experiment. Though the local check gave lowest fruit length, the pod diameter was statistically at bar with PBC-586. However; there was statistically significant different among all the genotypes. However, genotypes PBC-586 and PBC-142 gave attractive pod characters in color, length and diameter than all the rest.

Table 2: Total dry pod yield (q/ha) of chili genotypes at Melkassa and Debrezeit Agricultural Research Centers and Wonji commercial farm in, two years of 2013 and 2014.

Table 3: Overall combined performance of chili genotypes at Melkassa, Debrezeit Agricultural Research Centers and Wonji commercial farm in, 2013 and 2014.

Yield Stability analyses

The mean fruit yield per hectare across the six environments ranged from 21 q ha^-1 for (local check) to 33 q ha^-1 (PBC-586). The analysis of variance (ANOVA) of fruit yield of the five genotypes showed that 25% of the total SS of the model was attributable to environmental effects and the rest to genotype effects (45%) and Genotype × Environment interaction (5%) (Table 4). The significant MS of the environment source in the AMMI analysis of variance indicated that the environments were diverse, causing most of the variation in with large differences among environmental means in fruit yield, which is in harmony with the findings. Moreover, the genotype sources of variation in the AMMI analysis was significant indicating genotypes had highly variability in their performance in environments. However, GEI sum of squares were about 9 times lower than that of genotype SS, indicating the GEI effects was less strong in causing the variability in yield over locations than the inherent variability of the genotypes. This further indicated the greater opportunity towards selection for favorable traits in multi-environmental trials [8,9].

Table 4: Combined analysis of variance (ANOVA) according to the AMMI 2 models for the six environments.

Moreover, using of the conventional ANOVA stability model, from the criteria of higher mean yield and lowest Coefficient of Variation (CVi) jointly as stability parameters, indicated, PBC-401 had the lowest CV (11%), with significantly lower yield values (ranked 3rd) than of the first two high yielding genotypes (i.e., PBC-586 and PBC-142A. Hence, balancing lower CV value with higher mean yield as the criteria for decision, PBC-586 and PBC-142A were found the most stable variety, as they had the 1st and 2nd highest yield score with the lowest the 3rd and 2nd CV value in that order [10]. This may enable PBC-586, PBC-142A and PBC-401 attractive with respect to yield stability across environments (Table 5)

Table 5: Mean performance of five genotype for total fruit yield over different environments. Grand mean yield=26.8; R-squared=0.85; CV=15%.

The Principal Component Analysis (PCA) of AMMI model

The complete AMMI model contained 85% of the total SS, and the residual was 15%. The AMMI model clearly demonstrated the existence of a significant genotype × environment interaction, which is partitioned in to the first and second IPCA axes. The IPCA scores of the genotypes in the AMMI analysis are indication of the stability of a genotype across environments. The closer the IPCA scores are to zero, the more stable the genotypes are across the testing environments. The greater the IPCA scores, negative or positive (as it is a relative value), the more specifically adapted a genotype is to certain environments [11-13].

Thus, when the IPCA scores of a genotype are interpreted the first IPCA axis of the interaction captured 40% and 45% of the interaction degrees of freedom and sum of squares, respectively. On the other hand, the second IPCA axis captured 30% and 29% of the interaction degrees of freedom and sum of squares, respectively. Moreover, the third IPCA axis explained about 30% of the interaction degree of freedom and 25% interaction SS and was not-significant, which is considered as a residual or noise; not interpretable; and thus discarded. Partitioning of the interaction sum of squares by AMMI was quite effective as the MS for the first two IPCA axes is over three times the MS of the residual [14,15]. The AMMI analysis, based on the rank of genotypes on the IPCA 1 and IPCA 2 scores, indicated that, PBC-536 ranked 1st in IPCA 1 score but ranked 2nd in its IPCA 2 score, PBC-401 ranked 2nd in IPCA 1 score and 3rd in IPCA 2 score and PBC-142A ranked 3rd and 4th in IPCA 1 and IPCA 2 respectively (Table 6).

Table 6: IPCA 1 and IPCA 2 scores for five genotypes evaluated at six environments.

However, considering both IPCA scores the most stable genotypes was PBC-586 with above average yield. Furthermore, PBC-142A and local check were found better adapted to both high and low yielding environments as they had an IPCA value of the different sign and magnitude. The chili genotypes, PBC-067 and local check was specifically adapted to the low yielding environments had a below average yield (Table 6).

There is a good variation in the different environments ranged from the lower yielding environments of MARC-2013 and DZT-2014 to the high yielding environment of MARC-2014, Wonji-2013 and Wonji-2014. Furthermore, the AMMI analysis of the environment with the IPCA 1 and IPCA 2 scores, showed MARC-2014, 1st in IPCA 1 and 2nd in IPCA 2 score and DZT-2014, 3rd and 1st in its IPCA 1 and IPCA 2 scores respectively. From the interaction point of view, location MARC-2014 was the most discriminating environment, favoring respective genotype with similar IPCA signs and values. On the other hand, location ‘(Wonji-2013 and DZT-2014) were the least discriminating ones, as they had a lower negative IPCA scores, which may imply that genotypic differences at this site would be highly consistent with that averaged over sites (Table 7). These lower PCA scores of both locations ‘Wonji-2013 and DZT 2014 may also imply that the relative ranking of genotypes was stable at these sites. To this end, genotypes with lower negative PCA scores of PBC-067, Local check and PBC-142A were favored in these locations. In general, the candidate genotypes, PBC-586 and PBC-401were found the most stable genotypes with the lowest IPCA scores.

Table 7: The IPCA 1 and IPCA 2 scores for the six sites, sorted on environmental mean yield, used in the study.

Analysis of environmental index

Environmental index, of the two years in three locations (six environment) on yield performance of chili genotypes indicated that the highest mean yield was recorded at MARC-2014 during the second year and Wonji-2013 during the first year (Y2 and Y1) and were considered the best yielding environment with high positive environmental index of 5.7 and 2.2 respectively, followed by Wonji-2014 (Y2) which had also positive environmental index (0.77). However, the environmental index of the three locations, DZT-2013, and MARC-2013 (Y1,) and DZT-2014 (Y2) were negative.

In general, the highest mean yield of 32.5 and 29 q ha^-1 with high positive environmental index of 5.7 and 2.2 was recorded from Melkassa-2014 and Wonji-2013, respectively (Table 5), which was greater than the overall average mean yields of (26.8 q/ha) of six environments. However, the yield level was lower than the grand mean 26.8 q/ha and the environmental index was negative in the rest of the environments. The overall yield potential of Wonji at both years, Wonji-2013 and Wonji-2014 was 29 q/ha and 27 q/ha respectively were greater than the overall average mean yields of (26.8 q/ha) of six environments. However, at Debrezeit in the two years of the experiment, there was negative environmental index with below average yield, indicating as it is the poorest yielding environment for chili production. Further, this study clearly indicated that chili performance was better in low land sandy loam soils of MARC and Wonji areas than high land heavy soils of Debrezeit.

Analysis of regression coefficient

Considering regression value (bi) approximating unity; when associated with high mean yield, genotypes have general adaptability but when associated with low mean yield, genotypes are poorly adapted to all environments. As per the overall analysis the genotypes had considerably less variation around the grand mean yield of 26.8 q/ha. When, a regression value (bi), is greater than unity indicates genotypes are highly sensitivity to change in environment and called (below average stability) and are specifically of adaptable to high yielding environments. However, when a regression value (bi) is less than unity provides a measure of greater resistance to environmental change (above average stability) and are specifically adaptable to low yielding environments [7].

From the, linear regression analysis of a single genotype on the average yield of all genotypes in each environment resulted in the regression coefficient ranged from 0.77 to 1.2 for fruit yield. This large variation in regression coefficient explains different responses of genotypes to environmental changes. All chili genotypes showed significantly different regression coefficient (bi) from unity which indicated that the linear response of the genotypes to environmental index is different in different environment. Though, the regression coefficient of all genotypes was positive, the two top yielding genotypes (PBC-586 and PBC-142A), had a regression coefficient above >1, which indicated as these are specifically adapted to high yielding environments. Therefore, in this study PBC-586 and PBC-142A exhibited superior mean fruit yield than the grand mean (26.8 q ha^-1), with regression coefficient values of (bi=1.2 and 1.1) close to unity indicating their average stability suitable for all environments.

Chili is the potential vegetable crop produced under rain fed and irrigated conditions of the rift valley areas of Ethiopia. The overall yield performance is better in these hot areas with good supplementary irrigation for early plant establishment and subsequent high yield potential. Genotypes also showed different in days to flowering, crop establishment and fast growth and development and high in marketable and potential yield. Considering, the overall average performance, PBC-586 and PBC-142 found to be better performing especially in fruit quality marketable and total fruit yield than the check and the rest of the genotypes at all test locations. Considering wide range adaptability as a breeding strategy, the best genotypes that combine high yield and stable performance across range of environments is the best strategy. Thus, along with this analysis, from all the genotypes, PBC-586, BC-142A and PBC-401, were found stable, had comparably lower yield fluctuation in different environments which was found appealing for recommendation and release for wider environments. In addition, their attractive pod length, diameter, color and uniformity, have got high local consumer preference besides good yield potential. Therefore, PBC-586 and PBC-142 was released with local name called 'Melka-Dera and Melka-Oli' for PBC-586 and PBC-142A respectively. Moreover, the potential of chili production is widely recognized as important cash crops by small farmers in different climatic regions of the Rift Valley areas. Since chili can be produced in the short rainy season can consistently augment hot pepper production in the off season under irrigated condition so that chili and pepper products can be available throughout the year for in the local in the market for consumers.

We are very much grateful to Ethiopian Institute of Agricultural Research (EIAR), for the financial support and cooperation to undertake this research activity and MARC, DZT, Wonji-Elfora commercial farm for managing the field trial.