African Crop Science Journal
African Crop Science Society
ISSN: 1021-9730
African Crop Science Journal
Vol.5. No.2, pp. 119-125, 1997

Yield evaluation and stability analysis in newly selected 'KSA' cotton cultivars in western Kenya


National Fibre Research Centre, Kibos, P.O. Box 1490, Kisumu, Kenya

(Received 4 May, 1994; accepted 9 April, 1997)

Code Number: CS97018
Sizes of Files:
    Text: 26.1K
    Graphics: No associated graphics files


Performance for seed cotton and lint yields, lint percentage, and days to 50% boll opening of nine cultivars and strains of cotton (Gossypium hirsutum L.) was evaluated across 21 environments in Western Kenya during 1981 through 1986. The objectives of the study were to assess improvement in yield by selection and examine adaptability and stability for seed cotton yield. Two introduced, REBA B50 and HARNA(73)44, one obsolete, BPA68, and two commercial cultivars, UKA59/240 and BPA75, three strains, KSA20, KSA153 and KSA112, selected from UKA59/240, and one multiline of the strains KSA81M were chosen for evaluation at six locations over the years. Genotypic and environmental effects were highly significant (P<0.01) for all the traits studied. Genotype X environment interaction effects were highly significant for seed cotton and lint yields. Seed cotton yields ranged from 1389 to 1559 kg ha^-1, with the strains and multiline demonstrating superiority over the commercial variety they were selected from in seed cotton and lint yields. Stability analysis for seed cotton yield indicated linearity for a considerable portion of genotype X environment interaction effects as evidenced by high significance of the linear component of the interaction. Estimates of parameters describing stability found the multiline to be most stable, and widely adaptable to cotton growing areas of western Kenya.

Key Words: Adaptability, Earliness of maturity, Genotype X environment, Gossypium hirsutum L., lint percentage, lint yield, multiline, selection, seed cotton yield, stability


La performance du coton a graines et des produits a peluche, le pourcentage de la peluche, et les jours marques par 50% d'ouverture des capsules de neuf cultivars et des varietes de coton (Gossypium hirsutum L.) ont ete evalue dans 21 milieux differents de l'ouest du Kenya de 1981 a 1986. L'objectif de cette etude etait, d'une part, d'evaluer l'amelioration la production du coton a graines par la selection et, d'autre part, d'examiner son adaptabilite et sa stabilite. Deux varietes introduites, REBA B50 et HARNA(73)44, une ancienne variete BPA68, deux cultivars commerciaux UKA59/240 et BPA75, trois varietes KSA20, KSA153 et KSA112, selectionnees a partir de UKA59/240, et beaucoup de varietes KSA81M ont ete choisies pour cette evaluation dans six endroits differents sur plusieurs annees. Les facteurs du genotype et du milieu ont revele des effets hautement significatifs (P<0.01) pour tous les caracteres etudies. Les effets de l'interaction du milieu sur le genotype x se sont averes hautement significatifs pour le coton a graines et les produits a base de peluche. La production du coton a base de peluche. La production du coton a graines, y compis celle des varietes et de plusieurs lignees, etait comprise entre 1389 et 1559 kg ha^-1. De telles valeurs demontrent leur superiorite par rapport a la variete commerciale a partir de lonquelle elles ont ete selectionnees. L'analyse de la stabilite de la production du coton a graines indique le caractere lineaire d'une bonne partie des effets du milieu du genotype x. Ceci est mis en evidence par l'importance considerable de la composante lineaire de l'interaction. Les estimations des parametres descriptifs de la stabilite ont prouve que la multilignee etait beoucoup plus stable, et hautement adaptable au coton cultive a l'ouest du Kenya.

Mots Cles: Adaptabilite, maturite precoce, milieu du genotype x, Gossypium hirsutum L., pourcentage de la peluche, production de la peluche, multilignee, selection, production du coton a graines, stabilite


The cotton cultivar UKA59/240 has been grown in several parts of Kenya since early 1970's. Seeds have been recycled continuously between farmers and ginneries without new seed issues. This was believed to have resulted in loss of purity and deterioration through mixing with other cultivar seeds and outcrossing. It was, therefore, considered necessary to carry out selection.

Selection for yield in cotton cultivars, especially when criterion for indirect selection is highly heritable, has been reported to be successful. Manning (1963) and Miller and Rawlings (1967) improved lint yield through selection. Manning (1963) attributed improvements obtained by selecting within a cotton variety which had been inbred for several generations to the amphidiploid nature of the cotton plant. Scholl and Miller (1976) predicted response of yield to selection for lint percentage to be as great as direct selection for yield.

Miller et al. (1959) and Bilbro and Ray (1976) recommended yield evaluation as a necessary step in development of improved cotton cultivars. According to Miller et al. (1959), cotton is grown over a range of soil types, variable fertility levels, climatic, cultural, disease and insect conditions. They noted these factors to vary from location to location in the same year, and even year to year in same location, which are then reflected in yields of cotton cultivars. Similar variations are found in cotton growing areas in Kenya.

Several procedures have been proposed for evaluating stability of cultivars. Lin et al. (1986) have reviewed nine stability statistics. Liu and Sun (1993) evaluated 17 statistics recommended for description of cultivar stability, and favoured use of Eberhart and Russell's (1966) regression model in yield stability analysis of cultivars. Paroda and Hayes (1971) emphasised that linear regression could be regarded as a measure of response of particular cultivars, whereas the deviation around the regression line (S2di) is the most suitable measure of stability. Becker (1981) found the regression coefficient, bi, to be closely correlated with the environmental variance.

In the present study, lint percentage was used as a criterion of selection for higher yield in the commercial cotton cultivar UKA59/240. The objectives were to evaluate improvement in cotton yield through selection and to determine cultivar seed cotton yield stability. The commercial cultivars were used as standard checks in their respective ecological zones.


The base population for selection was a commercial cultivar, UKA59/240, selected from UKA67 which originated from selections of a cross between 'Mwanza local' and 'Albar 51' at Ukiriguru, Tanzania. Albar 51 was a selection from 'Nigerian Allen' for resistance to Xanthomonas campestris p.v. malvacearum (Arnold, 1970). UKA59/240 was introduced in Kenya and became widely grown in central, eastern, coastal, and parts of western Kenya.

Nine-kg seed samples of UKA59/240 were obtained from each of the four cotton growing regions. The samples were mixed to form the base population. During the main season of 1978 this base population was planted at Kibos. At crop maturity 1,000 randomly selected plants were labelled with a progeny number. Seed cotton from each progeny was harvested into appropriately labelled bags, and weighed. Ginning was done separately for each progeny and lint weighed for computation of lint percentage. A selection pressure of 5% was exerted to provide 50 plants with highest lint percentages, ranging from 34.3 to 37.7%.

The progenies selected were grown in progeny rows. The best eight strains in lint percentage were retained and assigned 'KSA' (Kibos Selection Albar) prefix, followed by the original progeny serial number. Equal weights of seed from the eight strains were mixed to synthesise a multiline, 'KSA81M', which has since been registered and described (Opondo et al., 1993).

Evaluation in replicated experiments included KSA81M, the strains KSA20, KSA153, and KSA112, commercial cultivars UKA59/240 and BPA75. An obsolete commercial cultivar, BPA68, and two introduced cultivars, HARNA(73)44 and REBA B50 were also included for comparison purposes due to their fibre qualities and yield.

The nine cultivars and strains were planted at four sites, Kibos, Alupe, Siaya, and Homa Bay, over six years (1981-1986), except Siaya (four years) and Homa Bay (five years). An 'environment' refers to any single year-location combination. The altitude, temperature, rainfall, and some landscape and soil characteristics at the experimental sites are shown in Table 1.

TABLE 1. Altitude, mean temperature and rainfall, and some landscape and soil characteristics at the experimental sites

Location  Altitude  Temperature   Rainfall  Terrain/Soil characteristics 
            (m)    (C) min./max.    (mm)
Kibos      1137     15.4/28.9       1230    Plain/black heavy clay
Alupe      1250     15.9/27.8       1450    Gentle slope/brown clay loam
Siaya      1270     15.8/27.5       1450    Gentle slope/red clay soil
Homa Bay   1145     17.9/33.6       1200    Steep slope/black heavy clay

At each environment entries were grown in a randomised complete block design with five replicates. The plot size was two 10-m rows spaced 1 m apart. Seeds were sown in hills 0.30 m apart and thinned to two plants per hill, 2 to 3 weeks after planting. Inorganic fertilizers were applied at recommended rates for each location, with chemical pest control and hand weed-control.

Days after planting when at least 50% of the bolls had split open (days to 50% boll opening), seed cotton and lint yields were recorded on a plot basis. Data on lint percentage were recorded for each environment on a single sample of a genotype over-all replications. The analysis of variance for this trait had to be calculated on single values per environment. Therefore only environments, cultivars, and environment X cultivar effects were the sources of variation estimated. Bilbro and Ray (1976) employed a similar approach using mean cotton lint yields.

Analyses of variance for individual environments were performed. Bartlett's t-test, as described by Snedecor and Cochran (1967), was used to assess homogeneity of variances prior to combined analysis. In the combined analysis, genotypes (cultivars and strains) were considered fixed effects and environments assumed random. Mean squares for genotype X environment were used to test significance of genotypic and environmental effects, while genotype X environment interaction effects were tested using pooled error (McIntosh, 1983). Stability analysis for seed cotton yield was performed following Finlay and Wilkinson (1963), and Eberhart and Russell (1966).


Mean yields for all entries were moderate for rain-grown cotton (Table 2). Environmental mean seed cotton yield ranged from 1380 to 1570 kg ha^-1, with mean yields for other environments reasonably spaced between these limits. Significance (P<0.01) was indicated for genotypic and environmental effects in the traits (Table 3). Significant (P<0.01) genotype X environment interaction effects were also indicated for seed cotton and lint yields.

TABLE 2. Means for four traits of new cotton strains, a multiline and five cultivars in 21 environments in western Kenya

Genotype     Seed cotton     Lint       Lint    Days to 50%
              (kg ha^-1)  (kg ha^-1)     (%)     maturity
KSA20         1480ab        521cde      35.7bcd    142a
KSA153        1529a         550abc      36.0abc    141ab
KSA112        1559a         559ab       36.2ab     141ab
KSA81M        1514ab        533bcd      35.6cd     139cd
BPA75         1482ab        515de       35.2de     139cd
UKA59/240     1389c         474f        34.3g      141ab
BPA68         1428bc        492ef       34.5fg     139cd
HARNA(73)44   1510ab        526cd       34.9ef     139cd
REBA B50      1553a         566a        36.3a      137e
S.E.          29.1          9.4          0.21       0.7
CV (%)        18.10        18.35         2.84       4.74
LSD (P<0.05)   80            26          0.57       1.8
Means within a column followed by same letters are not significantly different at P < 0.05 (Duncan's Multiple Range Test)

All the new 'KSA' strains, including KSA81M, yielded significantly (P < 0.05) higher seed cotton and lint, and lint percentage than the source cultivar, UKA59/240. An introduction, REBA B50, gave the highest yields and lint percentage. This cultivar had significantly more lint yield and percentage than all other entries, except KSA112 and KSA153, but outyielded only UKA59/240 and BPA68 significantly in seed cotton (Table 2). Among the strains and multiline, KSA112 had more seed cotton, lint and lint percentage. However, only KSA20, for lint yield, and KSA81M, for lint percentage, differed significantly.

The significant improvement in lint percentage in the strains and multiline over UKA59/240 indicates potential variability created as a result of continued recycling of planting seeds. Similarly, improvement in seed cotton and lint yields showed the effectiveness of using lint percentage for indirect selection of yield. Meredith and Bridge (1972) found additive gene action as most important in the inheritance of lint percentage, an indication of possible improvement in this character if selected. This was supported by Opondo (1979) who reported high heritability. Scholl and Miller (1976) reported lint yield response to indirect selection through lint percentage, and their results support the findings of this study.

The earliest maturing (measured as number of days to 50% boll opening) entry over all environments was an introduced cultivar, REBA B50 (Table 2). The three new strains did not differ significantly from UKA59/240 in maturity, with KSA20 taking longest to mature. The multiline, KSA81M matured significantly earlier (P<0.05) than UKA59/240 and KSA20. Fibre quality characteristics are presented in Table 4 (for reference purposes only).

TABLE 4. Mean values of lint quality traits of cotton gentoypes tested over 21 environments in western Kenya, 1981-1986

Genotype    Fibre length   Fibre strength    Fibre fineness   Fibre
elongation     inches     -----------------    micronaire       %
                          P.S.I    Gram/Tex            
KSA20         1.05a       86000c    20.0c          3.9d        8.0b
KSA153        1.04abc     85000cd   19.7cd         3.8d        7.9b
KSA112        1.04abc     83000d    19.4d          3.8d        8.9a
KSA81M        1.03bc      83000d    19.2d          3.9cd       8.8a
BPA75         1.05ab      89000b    20.8b          4.1ab       8.7a
UKA59/240     1.04abc     86000c    20.1c          3.9d        8.5a
BPA68         1.04abc     86000c    20.0c          3.9cd       7.9b
HARNA (73)44  1.06a       92000a    21.5a          4.2a        8.7a
REBA B50      1.02c       83000d    19.4d          4.1bc       8.0b
S.E.+/-       0.007        0.86      0.203         0.05        0.17
LSD (P<0.05)  0.02         2414      0.570         0.14        0.48
CV%           3.22         4.25      4.09          5.32        8.67
Means within a column followed by the same letter are not significantly different at P<0.05 (Duncan's Multiple Range Test)

Occurrence of significant genotype X environment interaction allowed for evaluation of the entries for stability of yield across environments. However, since lint yield is a factor of seed cotton yield, stability analysis was performed for the latter only.

The combined regression analysis for seed cotton yield (Table 3) indicated highly significant (P < 0.01) environmental (linear) and genotype X environment (linear) effects, indicating that the entries responded differently to the various environments. Pooled deviations from regression were also highly significant, indicating that a high degree of non-linearity existed in the entry/environment relationship.

TABLE 3. Mean squares in analysis of variance for four traits of cotton and combined analysis of variance and regression for seed cotton yield in 21 environments in western Kenya, 1981-1986

Source             Df   Seed cotton   Lint  yield   Days to   Lint percent 

                          yield     kg ha^-1x10^-2  50%boll       (%)
                      kg ha^-1x10^-3               opening           
Genotype (Gen)      8    0.072**        9.893**      253.52**    13.48**
Environments (Env) 20    4.738**      270.769**    18008.84**    11.80**
Replicates/Env     84    0.071         4.483         110.83        -
Gen X Env         160    0.022**        1.316**       50.83       0.78
Env (linear)        1    94.760**        -              -          -
Gen X Env (linear)  8    0.064**         -              -          -
Pooled deviation  171    0.018**         -              -          -
Pooled error      672    0.015          0.932         44.19        -
CV %                    18.103
** F-statistic sigificant at P < 0.01

Yield stability of a genotype is evaluated from estimates of stability parameters. A desirable stable cultivar is one with mean yield higher than the overall average of all genotypes under test, regression coefficient (bi) close to unity, and small deviations from regression (S2di), possibly close to zero (Eberhart and Russel, 1966). The deviations from regression measures the predictability of genotype reaction to environments and, therefore, has been regarded as the most appropriate criterion of stability in an agronomic sense (Paroda and Hayes, 1971). The regression coefficient should better be treated as an indicator for type of genotypic response to varying environments (Paroda and Hayes, 1971).

The stability parameters estimated for seed cotton yield are shown in Table 5. Two of the new strains, KSA20 and KSA112, were found to be unstable with deviations greater than zero (P < 0.05). KSA153 and REBA B50 were stable. UKA59/240 and BPA68 showed adaptation to unfavourable conditions, whereas BPA75 and HARNA(73)44 were adaptated to favourable environments. The multiline KSA81M was stable and showed good adaptation in all environments, which could be attributed to population buffering as discussed by Allard and Bradshaw (1964).

Coefficients of determination (r2) were suggested (Bilbro and Ray, 1976) to be useful in measuring dispersion around the regression line, and therefore related to the predictability and repeatability of performance within environments. In the current study, coefficients of determination for all genotypes, as shown in Table 5, were highly significant (P < 0.01).

TABLE 5. Regression coefficients (bi), deviations from regression (S2di) and coefficients of determination (r2) for seed cotton yield over 21 environments in western Kenya, 1981-1986

Genotype        bi      S.E.(bi)        S2di             r2
                                  (kg ha^-1x10^-3)
KSA20         1.037     0.0518          0.014*         0.861**
KSA153        1.039     0.0412          0.003          0.918**
KSA112        1.064     0.0512          0.013*         0.803**
KSA81M        1.021     0.0340          0.002          0.841**
BPA75         1.492**   0.0389          0.001          0.890**
UKA59/240     0.892**   0.0325         -0.004          0.907**
BPA68         0.876**   0.0346         -0.002          0.881**
HARNA(73)44   1.110*    0.0419          0.004          0.878**
REBA B50      1.002     0.0318         -0.004          0.782**
*,** Significant at P < 0.05 and 0.01, respectively


Selection for higher lint percentages in commercial cultivar UKA59/240 resulted in improved seed cotton and lint yields. Most cotton grown in Kenya is under rainfed conditions in diverse environments. Diversity here includes amount and duration of rainfall, soils differing in physical and chemical properties, temperature and humidity. Cultural practices, such as sowing dates, pure or mixed cropping, protection from insect pests and weed control, also provide environmental differences. The cotton cultivar to be recommended for commercial production in Kenya, therefore, needs to be one adapted to these diverse environmental conditions, with appreciable yields. The high seed cotton yield, stability and adaptability to all environments of the multiline KSA81M makes it suitable for recommendation as a commercial cultivar in all areas of western Kenya. The other new strains, KSA20 and KSA112 are suited for specific favourable conditions. The introduced cultivar, REBA B50, provides potential for incorporation in the hybridisation and selection programme for higher yields and stability.


We are grateful to the Kenya Agricultural Research Institute (KARI) for financial support on this study, and Director, KARI, for permission to publish this paper.


Allard, R.W. and Bradshaw,A.D. 1964. Implications of genotype-environment interaction in applied plant breeding. Crop Science 4:503-508.

Arnold, M.H. 1970. Cotton improvement in East Africa. In: Crop Improvement in East Africa. Leakey, C.L.A.(Ed.), pp. 178-208. Technical Communication 19 of the Commonwealth Bureau of Plant Breeding and Genetics. Commonwealth Agricultural Bureaux, England.

Becker, H.C. 1981. Correlations among some statistical measures of phenotypic stability. Euphytica 30:835-840.

Bilbro, J.D. and Ray, L.L. 1976. Environmental stability and adaptation of several cotton cultivars. Crop Science 16:821-824.

Eberhart, S.A. and Russell,W.A. 1966. Stability parameters for comparing varieties. Crop Science 6:36-40.

Finlay, K.W. and Wilkinson, G.N. 1963. The analysis of adaptation in a plant breeding programme. Australian Journal of Agricultural Research 14:742-754.

Lin, C.S., Binns, M.R. and Lefkovitch,L.P. 1986. Stability analysis: Where do we stand? Crop Science 26:894-900.

Liu, L.X. and Sun,Q.X. 1993. A comparative study on different statistical measures of stability. Agronomy Abstracts. American Society of Agronomy 85th Annual Meeting. pp. 92.

Manning, H.L. 1963. Realized yield improvement from twelve generations of progeny selection in a variety of upland cotton. Empire Cotton Growing Corporation. Research Memoirs.

McIntosh, M.S. 1983. Analysis of combined experiments. Agronomy Journal 75:153-155.

Meredith Jr., R.W. and Bridge, R.R. 1972. Heterosis and gene action in cotton, Gossypium hirsutum L. Crop Science 12:304-310.

Miller, P.A. and Rawlings, J.O. 1967. Selection for increased yield and correlated responses in upland cotton, Gossypium hirsutum L. Crop Science 7:637-640.

Miller, P.A., Williams, J.C. and Robinson, H.F. 1959. Variety X environment interactions in cotton variety tests and their implications on testing methods. Agronomy Journal 51:132-134.

Opondo, R.M. 1979. Genetic studies of earliness, yield, and fibre length in upland cotton, Gossypium hirsutum L. M.Sc. Thesis. University of Nairobi.

Opondo, R.M., Pathak, R.S. and Ombakho, G.A. 1993. Registration of 'KSA81M' cotton. Crop Science 33:212.

Paroda, R.S. and Hayes, J.D. 1971. Investigation of genotype-environment interactions for rate of ear emergence in spring barley. Heredity 26:157-176.

Scholl, R.L. and Miller, P.A. 1976. Genetic association between yield and fibre strength in upland cotton. Crop Science 16:780-783.

Snedecor, G.W. and Cochran, W.G. 1967. Statistical Methods. Iowa State University Press, Ames.

Copyright 1997 The African Crop Science Society

© Bioline International, 1989 - 2004, Site last up-dated on 03-Ago-2006.