African Crop Science Journal African Crop Science Society ISSN: 1021-9730 EISSN: 2072-6589 Vol. 9, Num. 2, 2001, pp. 359-367

African Crop Science Journal, Vol. 9. No. 2, pp. 359-367

Genotype x Row spacing and Environment interaction of Cowpea in Semi-arid Zones

Ndiaga Cisse
N. Cissé, ISRA / CNRA BP53 Bambey, Sénégal

(Received 13 December, 1999; accepted 17 December, 2000)

Code Number: CS01017

INTRODUCTION

Close spacing between and within rows significantly increase biological and grain yields in cowpea (Vigna unguiculata L. Walp), (Ezedinma, 1974). Jallow and Fergusson (1985) reported a linear seed yield response to density between 40 and 250 thousand plants ha^-1 (TPH). However Nangju et al. (1975), concluded that cowpea cultivars with different plant morphologies would require different optimum densities to express their full seed yield potential. Kwapata and Hall (1990) did not find cultivar responsiveness to density to be associated with difference in plant morphology; but they found that cowpea seed yield was significantly greater at 400 TPH than at 100 TPH in irrigated conditions.

Jallow and Fergusson (1985), and Kwapata and Hall (1990) reported a significant cultivar x density interaction. Row spacing x genotype interaction caused by changes in ranking of genotypes for seed yield could require twice the resources to identify superior genotypes for use in either row spacing (Hugie and Orf, 1989). If a trait that does not interact with row spacing but is related to seed yield could be identified, the resources needed to test lines could be reduced. Harvest index (HI) which is defined as the ratio of seed yield to total shoot dry-matter (Donald, 1968) has been proposed as an indirect selection criterion to improve cowpea grain yield ( Hall et al., 1997). In grain legumes, considerable foliage is lost by the time of final seed harvest. Actual HI uses as the denominator, the total shoot dry-weight produced during the entire growing period, by collecting all the foliage lost from the plants due to senescence. In apparent HI, the shoot dry-matter present at the time of harvest is used as the denominator. For soybean [Glycine max (L.) Merr.] Schapaugh and Wilcox (1980) concluded that apparent HI reflected actual HI, and that either method could be used to compare partitioning efficiencies. For cowpea, structural HI, which uses the shoot dry-matter without leaves and pedicels as the denominator gave similar information as apparent HI concerning genotype differences of photosynthate (Kwapata and Hall, 1990). Imrie and Butler (1983) found HI a useful indicator of cowpea seed yield. Ranking of cowpea genotypes for HI was consistent across a broad range of plant densities, indicating that selection for HI may be effective with widely spaced plants in the F₂ generation (Kwapata and Hall, 1990). However, Kahn and Stofella (1985) did not find harvest index and forage yield closely related to cowpea seed yield on a single factor basis. But a two-variable model using both these factors accounted for a high percentage (98%) of the variability of cowpea seed yield.

Most cowpea cultivar development and testing are conducted at plant densities lower than the corresponding recommended row spacing. Potential problems of breeding and evaluating cowpea genotypes at wide row spacing for subsequent release and production in higher plant densities have not been studied in semi-arid environments. Thus, it is important to estimate the magnitude of the genotype x row spacing interaction for seed yield and its effects on the testing and selection procedures used in the development of cowpea cultivar. Therefore, the objectives of this study were to: (i) examine the effects of two row spacing on grain yield and harvest index of 10 cowpea cultivars; (ii) determine the genetic relationships of yield and HI of cowpea, at 50 x 50 and 50 x 25 cm spacing, and at two different locations; and (iii) examine the consequences of genotype x row spacing interaction for seed yield on selection practices used to identify superior lines for use in wide and narrow spacing.

MATERIALS AND METHODS

The experiments were conducted in 1988 and 1989 growing seasons at the agronomic research stations of Bambey (14 ° 42 ' N, 16 ° 28 ' W) and Louga (15 ° 36 ' N, 16 ° 13'W) in Senegal. The soil at Bambey is a deep, slightly leached, tropical ferruginous with 7% clay; while at Louga the soil is sandy with a low percentage of clay (3%) and a low field capacity (8%). The soil pHs are about 5.5 at Louga and 7 at Bambey.

Ten cowpea genotypes were included in this study of which three (Ndiambour, 58-57, IS86-239) of them have a prostrate and indeterminate plant morphology; six (Mougne, Tvx3236, IS86-247, IS86-275, IS86-279, IS86-283) are semi-erect and indeterminate; and the tenth, (CB5) is erect and determinate. The five lines designated IS86- are experimental, from the cross 58-57 x IT81D-1137 (Cissé et al., 1995). The rest are released varieties in Senegal.

Fields were chosen where pearl millet had been grown the previous year and were ploughed at the end of the dry season. Fertilisers were applied in both locations at the rate of 9, 30, and 15 units of NPK, which is a low level presently recommended by the extension service for cowpea as a general practice. Chemical control of insects was implemented in all plots at the two locations. Hairy caterpillar (Amsacta moloneyi) and cowpea aphid (Aphis craccivora) were controlled with endosulfan (800 g ha^-1), and flower thrips (Megalurothrips sjostedti) with deltamethrine (15 g ha^-1). Annual precipitations were respectively 639 and 805.5 mm at Bambey in 1988 and 1989, and 437 and 470 mm at Louga. Compared with pan evaporation, rainfall would have supported a 70 and 90-day crop i 1988 and 1989 at Bambey, 60 and 70-day crop at Louga, respectively (Thaiw et al., 1993). Mid-season drought was observed in both years at Louga, during late vegetative stage to early flowering.

A randomised complete block design with split-plot arrangement and four replications was used in each location. Main plots were genotypes and subplots were within row spacing (WRS). Each main plot had two subplots; within each main plot the subplots were arranged at random. Each subplot contained six rows 5 m long with 50 cm between rows. The two WRS were 50 cm and 25 cm, planted at two seeds/hill to realise 80 and 160 thousand plants ha^-1, respectively. These plant densities correspond to the recommended row spacing for prostrate indeterminate and erect determinate genotypes (Séne and Ndiaye, 1974).

At physiological maturity (95 % of the pods yellow), pods were harvested on 4 m of each subplot centre four rows, for yield estimation. On the remaining 1 m of the centre four rows, pods were harvested and the plants were cut close to ground level. Only the shoot matter present at harvest was collected, so harvest index refers here to apparent HI. Pods and shoots were oven-dried at 70° C for 5 days before they were weighted. Harvest index (HI) was estimated with shoot dry-matter ha^-1 and seed yield ha^-1 obtained on the 1 m sample of each subplot. The formula HI = seed yield ha^-1 / shoot dry-matter ha^-1 was used. The 1 m sample was used to estimate HI, separately from subplot yield to avoid collinearity in the genetic correlations.

Statistical analyses. Locations and years were considered random factors, while genotypes and WRS were fixed effects. A separate analysis of variance was performed for each location over years, in addition to a combined analysis over locations and years. Spearman’s rank correlation coefficient was estimated to test the change in ranking of genotypes over within row spacing and within and between locations (Steel et al., 1997). Procedures from the Microcomputer Statistical program (MSTAT-C, 1991) were used for these analyses. Genetic correlation coefficients for individual traits were used to determine consistency in expression of yield and harvest index at the two planting densities, and the relationship of these two traits. The formula given by Falconer (1989) was used to calculate the genetic correlation coefficients as;

r_g = σ²_xy / (σ²_Gx. σ²_Gy)^1/2

where:
σ²_xy = Genetic covariance between 50 cm and 25 cm within row spacings.
(σ²_Gx. σ²_Gy)^1/2 = Geometric mean of genetic variances at 50 cm (x) and 25 cm (y) within row spacings.

Standard errors of the correlation were calculated after Mode and Robinson (1959). Significance at the α = 0.05 and 0.01 levels for genetic correlation was declared if the coefficient exceeded its standard error by two and three times, respectively (Mode and Robinson, 1959).

RESULTS AND DISCUSSION

The combined analyses of variances over locations and years showed significant genotype effects on yield and HI (Table 1). The effect of within row spacing was not significant for the two traits. The average grain yields were similar for 50 and 25 cm within row spacing. The difference in grain yields between seeding rate means of the same genotype was significant only for IS86-275, a semi-erect indeterminate line, with an advantage of 229 kg ha^-1 (+15 %) at 25 cm within row spacing. The second largest yield advantage (166 kg ha^-1, + 19 %) was also at 25 cm within row spacing and was from CB5, the erect-determinate genotype but it was not significant (Table 2).

Averaged over genotypes, HI values were 46.2 % at 50 cm and 42.1 % at 25 cm within row spacing. The differences in HI values between seeding rate means of the same genotype were significantly larger at 50 cm than at 25 cm within row spacing for five of the ten lines (Table 2). Lower plant populations tended to increase harvest index in cowpea as in soybeans (Weber et al., 1966; Wilcox, 1974; Ismail and Hall, 2000).

The significant genotypes x years, genotypes x locations, and within row spacing x year x location interactions for yield and HI suggested that relative responses of genotypes and responses due to within row spacing changed across years and locations. Hence, environmental conditions might influence the outcome of selection if yield or HI is used as a selection criterion. Grain yields for individual genotypes were much higher at Bambey than at Louga. The higher yields at Bambey were associated with wetter conditions, shoot biomass being positively correlated with seasonal rainfall (Thiaw et al., 1993; Jost and Cothren, 2000). The higher soil fertility at Bambey was an additional factor contributing to yield. The variance component of the genotype x year interaction effects was not significant for harvest index and for grain yield; it was (17473.7), which was about one-fourth the magnitude of the genotype x location (66212) interaction variance. Thus, the effects of the genotype x year interaction were relatively less important than the genotype x location interaction effects. Hence, the data for each location were combined over years and analysed separately.

In the over year analyses within location, significant varietal differences for grain yield were observed at Bambey and Louga (Table 3). Genotypic differences for HI were identified at Louga but not at Bambey. Significant genotypes x year interactions were also observed for both traits at Bambey and Louga, suggesting that at individual sites, environmental conditions might also influence the outcome of selection for yield and HI. Different results were, however, obtained by Kwapata and Hall (1990) in cowpea and Egli et al. (1985) for soybeans who found harvest index to be a relatively stable parameter across years.

Within row spacing did not have a significant effect on both yield and HI at the two locations. The average yields of the two within row spacing were similar at Bambey (1767.5 and 1753.3 kg ha^-1). The low plant density (50 cm within row spacing) obtained 10 % more yield than the high density (25 cm within row spacing) in 1988, but the latter was better performing by 12 % in 1989. This rank change explained the significant year x within row spacing at Bambey. However, for individual genotypes, only CB5 with an erect and determinate plant type, had a yield advantage at 25 cm within row spacing in both years, with an average of 18 %; while the prostrate and semi erect genotypes performed equally at the two densities. At Louga, the high plant density (25 cm within row spacing) had 15 and 10 % greater yield than the low density (50 cm within row spacing) in both years, respectively, resulting in a 12 % average advantage, significant only at the 10 % probability level. For 9 of the 10 genotypes, grain yields were larger at 25 cm within row spacing, and the difference between seeding rate means of the same line was significant only for IS86-275 (Table 4). Therefore, it is likely that the main advantage of high plant densities is that it offsets some forms of environmental stress (Ball et al., 2000). If a cultivar is not particularly stressed such as in Bambey, the response to high plant densities may be small. Narrow spacing may be viewed as a management practice to stabilise yield potential over varying environments, which is in agreement with Marshall and Ohm (1987).

The interaction of primary interest is the genotype x within row spacing, which was not significant for yield and HI in the combined analysis over years and locations, and in the analysis over years within location, indicating that genotypes did not respond in different ways to within row spacing. Therefore, selection at low densities would be effective in identifying superior genotypes, which are productive at wide or narrow within row spacing. However, the significant year x within row spacing and year x genotype x within row spacing interactions for both traits, indicate that the response to high plant density varies with the environment, and also with the effect that the environment has on specific cultivars. The effectiveness of selection would then apply to relatively stable genotypes, which can be tested for more than a year. Menéndez and Hall (1996) did observe low heritabilities for HI and indicated that this trait might be more effectively selected for, in advanced rather than in early generation. These results are in contradiction with the statement that selection for HI may be effective with spaced plants in the F₂, because of consistent ranking of cowpea genotypes for this trait, across a broad range of plant densities (Hall et al., 1997).

The significant genotype x plant densities interaction for grain yield reported by Jallow and Fergusson (1985) and by Kwapata and Hall (1990) were based on one year data at one location, hence the genotype x within row spacing x environment components were included in the genotype x within row spacing effect. In this study, within each year the genotype x within row spacing interaction (not shown) were significant in all four analyses for HI and in three out of four, for grain yield. Thus, more than a year testing is necessary for accurate evaluation of genotype x plant density interaction in cowpea.

Spearman correlations, which are correlation of order of ranking, were used to compare the ranking of the genotypes in the two within row spacing. A spearman correlation of 1.0 would mean that the genotypes ranked exactly the same in both plant densities and that the effects due to difference in the direction of response were nonexistent. There were large positive and highly significant Spearman correlations for yield (0.98), and HI (0.94) between the two within row spacings in the combined analysis, suggesting that rankings of the genotypes did not change significantly over the two plant densities. The rank correlation coefficients for grain yield and HI at Bambey (0.87; 0.93) and at Louga (0.85; 0.88) also suggest that rankings of the genotypes for these traits were the same over the two within row spacing (Table 4). The rank correlation coefficients between the genotype means at Bambey and those at Louga for yield (0.05) and harvest index (0.53) were significantly different from 1, further indicating the presence of significant genotype x environment interaction.

Within both locations, the genetic correlation coefficients between yield and HI were highly significant (Table 5), indicating that attaining high grain yield can be achieved by selecting for optimal partitioning of carbohydrate to grain, as estimated by HI (Kwapata and Hall, 1990). But the genetic correlation coefficient between yield at Bambey and yield at Louga were low. Nonsignificant relationship between HI at Louga and Bambey were also observed. Thus, performance at these two locations is conditioned by substantially different sets of alleles (Atlin and Frey, 1990). It is possible also that the same sets of alleles controlling performance may be acting differently at the two locations, as indicated by a more recent study with QTLs for yield in two-rowed barley (Hordeum vulgare L.), that allelic effects may change magnitude or sign depending on environment (Tinker et al., 1996). The five IS86 - lines gave higher yield at Bambey, where they were selected from preliminary yield trials, compared to one of their parents, 58-57. This latter line, with substantial adaptation to low-productivity environment (Thiaw et al., 1993), was the best performing genotype at Louga. Only a portion of the gain from selection at Bambey was expressed at Louga. Therefore, preliminary selection in large populations, need to be done in both environments, since the goal is to preserve an elite fraction for further testing. The highly significant genetic correlation between yield and harvest index within locations and between yield at Bambey and harvest index at Louga, suggest that genotypes which are most productive in both environments can be identified. Selection for high yield would be achieved in the more favourable conditions (Bambey) and for high HI in the low-productivity environment (Louga).

The results of this study indicate that within row spacing did not have a significant effect either on both yield and harvest index. Genotypes responded similarly to WRS, suggesting that selection at low densities would be effective in identifying superior genotypes, which are productive at wide and narrow within row spacing. However, the effectiveness of selection would apply to relatively stable genotypes that can be tested for more than a year. Preliminary testing of large populations in high-productivity environment only, would not be sufficient for maximising gain from selection; but genotypes which are the most productive in both environments can be identified through concomitant selection in high- and low-productivity conditions for yield and harvest index respectively.

ACKNOWLEDGEMENT

This research was supported by the Bean / Cowpea CRSP, USAID Grant # Dan-1310-G-SS-6008-00, and the Government of Senegal.

REFERENCES

1. Atlin, G.N. and Frey, K.J. 1990. Selecting oat lines for yield in low-productivity environment. Crop Science 30:556-561.
2. Ball, R.A., Purcell, L.C. and Vories, E.d. 2000. Optimizing soybean plant population for a short-season production in the Southern USA. Crop Science 40:757-764.
3. Cissé, N., Ndiaye, M.,Thiaw, S. and Hall, A.E. 1995. Registration of ‘Mouride’ cowpea. Crop Science 35:1215-1216.
4. Donald, C.M., 1968. The breeding of plant ideotypes. Euphytica 17:385-403.
5. Egli, D.B., Guffy, R.D. and Legget, J.E. 1985. Partitioning of assimilate between vegetative and reproductive growth in soybean. Agronomy Journal 77:917-922.
6. Ezedinma, F.O.C. 1974. Effect of close spacing on cowpea (Vigna unguiculata) in Southern Nigeria. Experimental Agriculture 10:289-298.
7. Falconer, D.S. 1989. Introduction to quantitative genetics. Third edition. John Wiley & Sons, New York. 438 pp.
8. Hall, A.E., Singh, B.B. and Ehlers, J.D. 1997. Cowpea breeding. Plant Breeding Reviews. Jules Janick (Ed.). John Wiley & Sons, Inc. 15:215-274.
9. Hugie, W.V. and Orf, J.H. 1989. Genotypic interaction of early maturity soybean with row spacing. Crop Science 29:1447-1451.
10. Imrie, B.C. and Butler, K.L. 1983. Joint contribution of individual plant attributes to
11. seed yield of cowpea (Vigna unguiculata L.) in small plots. Field Crops Research 6:161-170.
12. Ismail, A.M. and Hall, A.E. 2000. Semidwarf and standard-height cowpea responses to row spacing in different environments. Crop Science 40:1618-1623.
13. Jallow, A.T. and Fergusson, T.U. 1985. Effects of planting densities and cultivar on seed yield of cowpeas [Vigna unguiculata (L.) Walp.] in Trinidad. Tropical Agriculture (Trinidad) 62:121-124.
14. Jost, P.H. and Cothren, J.T. 2000. Growth and yield comparisons of cotton planted in conventional and ultra-narrow row spacings. Crop Science 40:430-435.
15. Kahn, B.A. and Soffella, P.J. 1985. Yield components of cowpea grown in two environments. Crop Science 25:179-182.
16. Kwapata, M.B. and Hall, A.E. 1990. Determinants of cowpea (Vigna unguiculata) seed yield at extremely high plant density. Field Crops Research 24:23-32.
17. Marshall, G.C. and Ohm H.W. 1987. Yield response of 16 winter wheat cultivars to row spacing and seeding rate. Agronomy Journal 79:1027-1030.
18. Menéndez, C.M. and Hall, A.E. 1996. Heritability of carbon isotope discrimination and correlation with harvest index in cowpea. Crop Science 36:233-238.
19. Mode, C.J. and Robinson, H.F. 1959. Pleiotropism and the genetic variance and covariance. Biometrics 15:518-537.
20. MSTAT-C. 1991. A Microcomputer Statistical Program. Michigan State University.
21. Nangju, D., Little, T.M. and Anjorin-Ohu, A. 1975. Effect of planting density and spatial arrangement on seed yield of cowpea [Vigna unguiculata (L) Walp.]. Journal of the American Society of Horticultural Science 100:467-470.
22. Schapaugh, W.T. Jr. and Wilcox, J.R. 1980. Relationships between harvest indices and other plant characteristics in soybeans. Crop Science 20:529-533.
23. Séne, D. and Ndiaye, S.M. 1974. L’amélioration du niébé (Vigna unguiculata) au CNRA de Bambey: de 1959 à 1973. L’agronomie Tropicale 8:772-802.
24. Steel, R.G.D., Torrie, J.H. 1997. Principles and Procedures of Statistics. A Biometrical Approach. 2nd ed. McGraw-Hill Book Co., New York. 633 pp.
25. Thiaw, S., Hall, A.E. and Parker, D.R. 1993. Varietal intercroppping and the yields and stability of cowpea production in semiarid Senegal. Field Crops Research 33:217-233.
26. Tinker, N.A., Mather, D.E., Rossnagel, B.G., Kasha, Kleinhofs, K.J., Hayes, P.M., Falk, D.E., Ferguson, T., Shugar, L.P., Legge, W.G., Irvine, R.B., Choo, T.M., Briggs, K.G., Ullrich, S.E., Franskowiak, J.D., Blake, T.K., Graf, R.J., Dofing, S.M., Shaghai Maroof, M.A., Scoles, D.J., Hoffman, D., Dahleen, L.S., Kilian, A., Chen, F., Biyashev R.M., Kudrna, D.A. and Steffenson, B.J. 1996. Regions of the genome that affect agronomic performance in two-row barley. Crop Science 36:1053-1062.
27. Weber, C.R., Shibles, R.M. and Byth, D.E. 1966. Effect of plant population and row spacing on soybean development and production. Agronomy Journal 58:99-102.
28. Wilcox, J.R. 1974. Response of three soybean strains to equidistant spacings. Agronomy Journal 66:409-412.