African Crop Science Journal, Vol. 11. No. I, pp. 9-15, 2003 ISSN 1021-9730/2003 $ 4.

00
Printed in Uganda. All rights reserved ©2003, African Crop Science Society

GENOTYPE X ENVIRONMENT INTERACTION STUDIES ON YIELDS OF

SELECTED POTATO GENOTYPES IN UGANDA

G. ABALO, J. J. HAKIZAI , R. EL-BEDEWYz and E. ADIPALA

Department of Crop Science, Makerere University, P.O. Box 7062, Kampala, Uganda
'Kalengyere Research Station, P.O. Box 722, Kabale, Uganda
ZInternational Potato Centre, Sub-Saharan Africa Region, P.O. Box 25171, Nairobi, Kenya

(Received 15 January, 2001; accepted 5 February, 2003)

ABSTRACT

High yielding potato (Solanum tuberosum) genotypes with good levels of resistance to late blight were identified
at KalengyereResearch Station in southwestern Uganda. The yield stability of these genotypes, however, remains
uncertain. For instance, the performance of these genotypes varied over 4 seasons of evaluation and when tested
at Kachwekano site. Genotype x environment interaction study was therefore required to determine their
adaptation in the different agroecologies where the genotypes would be grown. Five sites in Uganda at different
elevations were selected for the study and these were Kalengyere (2450 mas!), Bulegeni (1670 mas!), Mbarara
(15OOmasl),Tororo (1250masl) and Namulonge (1150masl). The trials were conducted for three seasons: namely
2000 (A and B) and 2001 (C) being the first (A) and second (B) and third (C) seasons ofthe trials. Each location
for a particular season was regarded as one environment. Genotypes 389484.20, 389685.2, 389698.12, 389584.22
and 389701.34 performed better than the other genotypes in all the five locations. The G x E analysis was done
using Additive Main effects and Multiplicative Interaction (AMMI) and the biplot identified genotypes
389484.20,391558.16, Victoria, 391558.5, 389584.22, Kisoro and 391558.13 as adapted to Bulegen C, Mbarara
B, Namulonge Band C and Tororo A, Band C. Genotypes 391558.11, 389685.2 and 391558.1 were adapted to
Kalengyere A. Namulonge Band C and Tororo A, B and C were identified as similar environments basing on their
responses.

Key Words: Adaptation, AMMI analysis, Phytophthora infestans, Solanum tuberosum, yield stability

RESUME

Les genotypes de patates a haut rendement (Solanum tuberosum) avec des bons niveaux de resistance a la
brunissure tardive etaient identifies ala station de recherche de Kalengyere en sud-ouest de I' Ouganda. La stabilite
du rendement de ces genotypes, cependant, reste incertaine. Par exemple, la performance de ces genotypes a varie
au dela de 4 saisons d'evaluation et au moment de l'epreuve au site de Kachwekano. L'etude d'interaction
genotype x environnement etait ainsi done necessaire pour determiner leur adaptation dans les differentes zones
agro-ecologiques oil les genotypes seront germes. Cinq sites en Ouganda aux elevations differentes etaient
selectionnes pour l'etude notamment Kalengyere (2450 masl), Bulegeni (1670 masl), Mbarara (1500 mas!),
Tororo (1250 mas!) et Namulonge (1150 masl). Les essais etaient conduits pour trois saisons : nommement 2000
(A et B) et 2001 (C) etant la premiere (A) et la seconde (B) et la troisieme (C) saisons d'essais. Chaque
emplacement pour une saison particuliere etait regardee comme un environnement. Les genotypes 389484,20 ;
389685,2 ; 389698,12 ; 389584,22 et 38970 I,34 ont eu une bonne performance que les autres genotypes dans tous
les cinq emplacements. L'analyse de GxE etait faite utilisant les principaux effets additifs et multiplicatifs
d' interaction (AMMI) et les genotypes double identifies 389484,20 ; 391558,16 ; Victoria, 391558,5 ;389584,22,
10 G. ABALO et al.

Kisoro et 391558,13 comme adaptes a Bulegen C, Mbarara B, Namulonge B et C et Tororo A, B ct C, Lcs

genotypes 391558,I J: 389685,2 et 391558,I ctaient adaptes a Kalengyere A. Namulonge B et C et Tororo A, B
et C etaient identifies comrne des environnements sirnilaires en se basant sur leurs reponses.

Mots.Cles: Adaptation, analyse AMMI, Phvtoplithora infestans. Solanum tuberosum, stahilite de rendement

INTRODUCTION were twofold. Firstly, to determine the effect ofG

x E interaction on the yields (t ha') of elite potato
Selection for potato genotypes with high yield genotypes. Secondly, to identify stable and adapted
and resistance to late blight (Phytophthora genotypes for the different agroecologies In
infestans) continues to be a priority objective of Uganda.
most National Potato Programs in sub-Saharan
Africa, Uganda inclusive. Uganda participates in MATERIALS AND METHODS
the International Late Blight and Coordinated
Regional Yield Trials in a bid to identify high The multilocational trials were carried out at five
yielding and late blight resistant varieties for the locations that differed with respect to soil fertility,
sub Saharan Africa region (Hakiza et al., 1997). soil texture and elevations (Table I). The trials
In 1992 and 1993, the Uganda National Potato were done for three seasons; namely 2000 (A and
Program received two cohorts of diverse potato B) and 2001 (C) being the first, second and third
genotypes from the International Potato Centre seasons of the trials. At each location the genotypes
(CIP), Lima, Peru. Only five promising genotypes were planted on ridges in a randomised complete
were selected from each ofthe cohorts for advanced block design with three replications. A spacing of
yield trials at Kalengyere in Southwestern Uganda 0.7 m x 0.3 min 4-row-plots were adopted at all
in 1998 and 1999. The 10 potato genotypes were locations. The agronomic practices carried out
identified as high yielders (average of 20.2 t ha') included regular weeding, earthing up, three
with good resistance to late blight (average re lative applications of Dithane M45 (Mancozeb 80%
areas under disease progress curves, AUDPC of WP) at a rate of 2.5 kg ha' and dehaulming at
12 %) (Abalo et al., 200 I). The performance of maturity. Although the genotypes were selected
genotypes varied significantly from season to on the basis of high yields and good levels of
season and location to location when tested at resistance to late blight. the genotypes had to be
Kachwekano and Kalengyere (Abalo et al., 200 I), subject to applications ofdithane M45. The spray
which suggested the presence of genotype x was done because three of the five locations of the
environment (G x E) interaction. The G x E trials do not favour late blight epidemics and this
studies are ofparamount importance in the speci fie could lead to a biased comparison of the genotypes
environments in which the genotypes are to be from the different locations. Late blight severity
grown (Ortiz and lise de Cauwer, 1999). The (Henfling, 1987) and bacterial wilt incidence were
potato genotypes were therefore subjected to recorded at Kalengyere and Narnulong e,
multilocational trials. The objectives of the study respectively. At harvest, tuber yields were assessed

TABLE 1. The elevation, soil fertility and texture of the five locations for the trials

Location pH O.M N P K Na Ca Mg % sand %clay % silt Altitude

ppm rn.a.s.l.
% Me/100g of soil Texture

Kalengyere 5.7 5.9 0.33 22.34 1.03 0.32 4.7 2.1 46.0 28.0 26 2450
Mbarara 5.9 2.9 0.16 18.23 1.52 0.17 4.2 1.9 56.0 24.0 20 1500
Namulonge 5.8 1.27 0.19 6.72 0.75 0.07 5.9 1.6 52.0 36.0 12 1150
Tororo 6.0 0.47 0.1 9.97 0.35 0.03 6.8 2.2 76.0 12.0 12 1250
Bulegeni 6.3 3.87 0.18 4.48 2.3 0.23 8.2 2.9 36.0 38.0 26 1670

'Organic matter
 Genotype x environment interaction studies 11

according to numb er of tubers per plant and mean In many situations, the major mi cro -
tuber weight; these values together with plot size environmental variables are not known exactly
(4.5 m x 2.8 m) were used to calculate the overall but they often include location and season effects
yield (t ha"). Temperature and rai nfall data were in addition to the conti nual improvement in field
recorded during the experimental period (Fig . I) . fertility and change in plant handling techniques
Late blight severity data were used to calculate (Kearsey and Pooni , 1996). Because there was no
areas under disease progress curve s (AUDPC), physical measure of some of the climatic and
which were subsequently standardized to give edap hic fac tors, the approach used was to let the
relative AUDPC (Campbell and Madden, 1990). genotypes them selves access the environment.
AUDPC values, bacterial wilt incidence and yield Therefore, a season in a partic ular location was
data were subjected to ana lysis of variance considered as one environment. TheAMMI biplot
(ANOVA) using the MSTATC com puter package was developed by plac ing both genotype and
(Ru ssell, D. Freed, Mic higan Stat e University, environment means on thex-axis orabscissa, and
USA) . Where the "F" stati stics ind ica ted the respective eige nvectors or scores (IPCA I) on
significance, the mea ns were separated using the y-axi s or ordinate (Zobel, 1990). Genotypes
Fisher' s protected Least Sig nifica nce Difference or (env ironments) which appear almo st on a
test (LSD) at P= 0.05 . To determ ine the effects of perpendicular line have similar means, those
genotype x environment interactions on yields, falling on a horizontal line have similar interaction
the data were subjected to Additive Main effects patterns. Geno types (or environments) with large
and Mu ltiplicative Interaction (AMMI) analysis first IPCA scor es (either plus or minus) have large
using Matmodel Ver sion 2.0 computer package interaction; those with values close to zero have
(Gau ch, 1990) and the bip lot drawn. AM MI small interaction (Hill et al., 1998; Stey n et al.,
analy sis was preferred in this study because of the 199 3), and are considered stable.
advantage accrued to it over the usua l breeders
way of apportio ning the variance of G x E. RESU LTS AND DISCUSSION
Regression analysis assu mes linearity in the
respo nse of genotypes to the environments but Yield performance of th e gen otyp es. The stud y
AMMI ana lysis takes into account the non-linear was based on mainly yields while other agro nomic
response of genotypesto environmental variances trait s such earliness of maturi ty was not included
maki ng it an appropriate tool especia lly where the in the study beca use the genotypes were all of
characteristic response of the environments are similar mat urity group. The ana lysis of variance
not known. (not show n) showe d that the genotypes were

400 35
350
W-
• •W- • . :s,.. - . ... . . x.
• .• . ~ .~ _~.~ .4 .~ ,.
}'- .. ... . ..... 30
.E 300 • 'J{ .. ... . .x:
~ 250

.
~200
•
S 150
.... ~

......... . . ~ ..•.. .. . . . . . . . - -e. . -e.. . .. ~

25~
2O ~
15

10 ~
iE
{;. 100

I i ~ I II .
50 5

o
.. [\.1 o
/ : / #'~ I' #' ,~~ ,:§o ... t . .#'/ ,//'/"it<!- .,,1' -:/ ...<f' ~'" ,~~ ,:§o
r:!f- I ~ <:j ~ _ Bulegenl RJF
[==::J Kalengye~ RIF
Months _ Namulonga R1F
- 6 ' - Bulegeni Temp
• • • • · Ka"'ng~ Temp
- . . - NamuIcngeTemp.

Figure 1. Monthly rainfall and average temperatures from January 2000 to July 2001 at 3 locations in Uganda.
12 G. ABALO et al.

significantly (P = 0.01) different in tuber yield (Aba10 et al., 2001; Adipalaetai., 2001 a; Adipala
over seasons and locations. There were also eta!.,200lb).
significant (P = 0.001) interactions for genotypes For Tororo site, yields were low during all the
x location, location x season (P = 0.001) and seasons, however, the best genotypes were
season x genotype x location (P = 0.01). The 389698.12 and 389584.22each with average yields
highest yields were recorded at Bulegeni and of lOt ha . Potato yield during 2000B and 200 IA
Kalengyere with average yields of 22 and 19.6 t were better than during the 2000A. This was
ha', respectively (Table 2). Lower yields were because first season was affected by the long
obtained at Namulonge and Tororo where yields drought and late rainfall. High temperature and
averaged only 9.0 and 8.8 t ha', respectively. little rainfall amount are known to limit tuber
Genotypes 389484.20, 389685.2, 389698. I2, bulking (Steyn et al., 1993; Abalo et al., 2001).
389584.22 and 389701.34 performed better than Tuber yields for the genotypes at Namulonge
the other genotypes in all the five locations. were low during 2000A. This was partly due to
The combined means over locations identified bacterial wilt infestation. which hit the trial during
389685.2 (16.7 tha') as the best yielding genotype the season (Table 3). The inoculum was probably
and 391558.16 (11.5 t ha') as poorest genotype in the soil since the seed tubers were obtained
for 2000A. Genotypes 389698.12 (16.9 t ha'), from Kalengyere Research Station and were
389701.34 (16.7tha'l) and 391558.13 (16.0t bacteria wilt-free. Besides, wilt only attacked the
ha') ranked best in 2000B and least yields were trial at Namulonge site and no wilting was observed
observed from genotypes 391558.16 (13.0 t ha') at other locations.
and 391557.1 (13.1 t ha").
The results of the study satisfied one of the Stability and adaptation of the genotypes.
breeders' goals of selecting for such varieties Analysis of stability and adaptation of genotypes
with acceptable yield. Yield performance at for yields were done using Additive Main effect
Kalengyere and Bulegeni in all the seasons ranged and Multiplicative Interaction (AMMI) model.
from 13.6 to 30.3 t ha' and 19.4 to 29.3 t ha', The AMMI analysis of variance for tuber yield
respectively. These yield ranges are above the (Table 4) indicated that genotypes and
Uganda national average yields of 7.0 t ha' environments were significantly (P < 0.01)

TABLE 2. Mean yields of 12 potato genotypes at 5 locations in Uganda for 3 seasons of 2000 - 2001

Genotype Location-' Season b

Mbarara Kalengyere Bulegeni Namulonge Tororo Means Means Means

2000A 2000B 2001A

389484.20 13.4 20.0 29.3 7.2 7.8 16.3 14.7 14.5

391558.11 16.3 19.3 21.7 8.0 5.9 12.9 15.5 13.3
389685.2 15.0 24.8 20.6 8.3 9.9 16.7 14.7 15.4
391558.16 11.5 13.6 19.4 7.6 9.2 11.5 13.0 11.7
VICTORIA 11.1 14.9 25.9 8.8 9.5 14.0 14.0 13.8
391558.5 12.3 16.0 20.5 10.7 8.2 13.1 14.0 13.0
389698.12 12.0 30.3 20.7 8.1 10.0 15.5 16.9 15.8
389584.22 14.1 16.4 22.2 11.5 10.0 13.8 15.9 13.4
KISORO 12.2 17.0 21.3 11.1 9.8 14.4 14.1 13.3
391557.1 13.8 19.4 19.7 8.0 7.5 14.2 13.1 12.6
389701.34 11.1 26.1 19.4 10.0 7.9 13.1 16.7 14.1
391558.13 14.5 17.4 23.1 8.2 9.9 13.2 16.0 13.1

C.V (%) 14.3 18.9 16.4 14.4 13.8 14.1 14.9 13.7
LSD (P =0.05) 1.7 6.7 4.7 2.1 1.7 3.3 3.3 1.8

a=Means for each location are based on 3 seasons, b= Means for each season are based on five locations A and
B are first and second seasons of in a year
 Genotype x environment interaction studies 13

different and G x E interaction was also very significant (P < 0.05) for tuber yield and accounted
significant (P < 0.001). The AMMI2 model for 72.1 %, 3.2 %and24.7%ofthe treatment sum
supplied an adequate fit to the data since only two ofsquares, respectively. This result indicated that
interaction principal componentaxes (IPCA) were the proportion of environmental and G x E
significant (P < 0.001). The sum of squares for interaction variation for tuber yield was much
genotypes, environment and IPCA I and 2 larger than that due to genotypes main effects.
provided 92% of treatment sum of squares These results concurred with results of many
indicating that AMM12 model effectively studies on G x E (UTA, 1993; Gauch and Zobel,
partitioned treatments sum of squares (Steyn et 1996; Dixon and Nukenine, 1997), which have
al., 1993). Hence the treatment sum of squares shown that the proportion of sum of squares due
contains 92% pattern related to treatment design to difference among sites ranged from 40 - 90 %
and 8% noise related to experimental design. and G x E was usually larger than genotypes main
The AMMI analysis also revealed that the effects. It also implied that yield was affected by
environment (obtained as location by season both the environment and the G x E interaction
combined), genotypes and G x E were highly effects (Ntawuruhunga et al., 2001).

TABLE3. RelativeAUDPC*valuesfor lateblightand percentage incidence of bacterial wilton 12 potatogenotypes

during 2000A*'

Genotype Kalengyere Namulonge

Late blight reiative AUDPC Bacterial wilt percentage incidence

389484.20 9 14
389685.2 15 13
389584.22 14 15
KISORO 25 14
391558.5 13 16
VICTORIA 14 16
389698.12 10 6
389701.34 16 4
389558.11 14 2
391558.13 14 19
391558.16 14 18
391557.1 22 28

C.V (%) 34.9 18

LSD (P = 0.05) 8.9 9.4

*Area under disease progresscurves; **Thefirst seasonof experimental trial

TABLE4.AMMI-2ANOVAtableof potatotuberyieldfor 12potatogenotypes at five locations in Ugandaduring2000

-2001

Source df 55 MS

Total 503 24227.48 48.17

Treatment 167 19174.71 114.82***
Genotype 11 618.17 56.20***
Environment 13 13825.79 1063.52***
GXE 143 4730.75 33.08***
IPCA 1a 23 2427.45 105.54'"
IPCA2b 21 747.65 30.77**
Residual 99 1895.93 45.07
Error 336 5052.77

a =InteractivePrincipalComponent Axis1; b =Interactive Principal Component Axis 2; **P = 0.01, ***P < 0.001
14 G. ABALO et at.

The analysis of the biplot (Fig. 2) revealed that Bulegeni were classi fied as similarenvironments.
genotypes 389484.20 and 391558.5 were high Tororo A, Band C, Namulonge Band C and
yielding and positively interactive. Likewise, Mbarara B were grouped together by AMMI. Site
Bulegeni A and Band Mbarara A were identified selection according to similarity of response helps
as high yielding environments with positi ve IPCA in elimination of expensive and inaccessible sites
I values. Conversely, Kalengyere A, Band C from multilocational trial sites (Hill et al., 1998).
were identified as high yielding but with negative The biplot for tuber yields facilitated the
interactions. The biplot also ranked Mbarara B, identification of the stable genotypes and those
Namulonge Band C, Tororo A, Band C and adapted to specific locations. The study identified
Bulegeni C as low yielding environments but two stable genotypes, 391558.11 and 391557.1.
with positive IPCA scores. The results identified No high yielding genotype showed stability and
genotypes 389484.20, 391558.16, Victoria, this collaborates earlier findings of Hill et al.
391558.5,389584.22, Kisoro and 391558.13 as (1998) that yield stability can only be expected
adapted to Bulegeni C, Mbarara B, Namulonge B from low yielding genotypes that do not benefit
and Tororo A and B as they had positi ve values for from favourable environments.
IPCA 1 (Steynetal., 1993).Genotypes391558.11,
389685.2, and 391557.1 were observed to be CONCLUSION
adapted to Kalengyere A.
Soil fertility levels (Table 1), average The study showed that the proportion of
temperatures and rainfall distribution (Fig. I) environmental variance and the G x E interaction
were probably the main factors that affected the were greater than genotypic variance indicating
genotypes. The environments with cool that the genotypes were closely related but
temperatures and high rainfall amount (Kalengyere responded differently to the differences in
A and B) favoured late blight infection causing a environments. It is apparent that one could rely
large negative contribution to G x E effects (Table more on crop management and suitability of the
3). Environment Namulonge A contributed environment to attain high yields rather than the
negatively to the interaction because of warm genotypic differences. The similarities identified
temperature and presence of bacterial wilt in the in the trial sites will be useful for future regional
soil, which affected the yield. and even international multilocational trial sites.
The AMMI analysis helped in classifying the The best genotypes identified in the study should
test sites. Both first and second seasons of 2000 at be tested on-farm under farmers' own practices.

-1

• Kalen~C

-2

.308ll70134

-3

• 36Q6ge.12

...
-5 ~---------

5 to 20
Gsnotypes and -environmentQlIrn~eln" (Mti ha)

Figure 2. AMM I biplot of 12 potato genotypes evaluated for tuber yield in five locations of Uganda during 2000 (A
and B) and 2001 (c) (A).
 Genotype x environment interaction studies 15

This is to assess their performance and compare Gauch, H.G. and Zobel, RW., 1996. AMMI
the genotypes with the local cultivars. They could analysis of yield trials. In: Kang, M.S. and
perhaps be grown under traditional cropping Gauch, H.G. (Eds.). Genotype-by-
system so that farmers can select varieties that environment interaction, CPC Press, USA.
would satisfy their needs and expectations. Hakiza, J.1., Kakuhenzire, R.M., Alacho, F.O.
and Mukalazi, J. 1997. The International Late
ACKNOWLEDGEMENT Blight and Coordinated Regional Yield
evaluation of potato genotypes in Uganda
The study was funded by The Rockefeller African CropScience Conference Proceedings
Foundation (Grant RF 99006#145) and the 3:981-987.
International Potato Centre (CIP). Thanks are due Henfling, J.W. 1987. Late blight of potato.
to NARD-Uganda, P. Ntawuruhunga and B. Technical Information Bulletin 4. CIP, Lima,
Mateeka for their technical support. Peru. 25pp.
Hill, J., Becker, H.C. and Tigertedt, P.M.A., 1998.
REFERENCES Quantitative and ecological aspects of plant
breeding. First edition, Chapman and Hall.
Abalo, G., Hakiza, r.r, Kakuhenzire, RM., El- pp.275.
Bedewy, Rand Adipala, E. 2001. Agronomic International Institute Of Tropical Agricuture
performance of twelve elite potato genotypes (IITA)., 1993. Archival report (1989-1993).
in southwestern Uganda.African CropScience Part 1. Cassava breeding, cytogenetic and
Journal 9: 17 -23. histology. Gerplasm enhancement. Crop
Adipala, E., Namanda, S., Mukalazi, L, Abalo, Improvement Division. Ibadan, Nigeria. Vol.
G., Kimoone, G. and Hakiza, U. zoots. 2.
Farmer potato varietal selection criteria in Kearsey, MJ. and Pooni, H.S. 1996. The genetical
Uganda. In: Proceedings oftheFifth Triennial analysis of quantitative traits. Chapman and
African Potato Association Conference 29 Hall. The University of Burningham, U.K.
May -2 June 2000, Kampala Uganda African pp.241-265.
Potato Association Conference Proceedings Ntawuruhunga, P., Rubaihayo, P., Whyte, J.B.A.,
5:439 - 443. Dixon, A.G.G. and Osiru, D.S.O. 2001.
Adipala, E., Namanda, S., Mukalazi, J., Abalo, Additive main effects and multiplicative
G., Kimoone, G. and Hakiza, U. 2001b. interaction anaysis for storage root yield of
Understanding fanners perception of potato cassava genotypes evaluated in Uganda.
production constraints and responses to yield African CropScienceJournal 9:591 ~ 958.
decline in Uganda. In: Proceedings of the Ortiz, Rand Ilse de Cauwer, 1999. Genotype x
Fifth Triennial African Potato Association environment interaction and testing
Conference 29 May - 2 June 2000, Kampala environments for plantain and banana (Musa
Uganda. African Potato Association spp. L.) breeding in West Africa. Tropicultura
Conference Proceedings 5:429 - 437. 3:98-102.
Campbell, C.L. and Madden, L.V., 1990. Steyn, PJ., Visser, A.F., Smith, M.P. and
Introduction to Plant Disease Epidemiology. Schoeman, lL., 1993. AMMI analysis of
John Wiley and Sons, New York. 532 pp. potato cultivar yield trials. South African
Dixon, A.G.O. and Nukenine, E.N., 1997. Journal of Plantand Soil 10:28-34.
Statistical analysis of cassava yields with the Zobel, R.W. 1990. A powerful statistical model
additive main effects and multiplicative for understanding genotype-by-environment
interaction (AMMI) model. AfricanJournal interaction. In: Kang, M.S. (Ed.). Genotype-
of Root and Tuber Crops 3:46-80. by-environment Interaction and Plant
Gauch, H.G., 1990. Matmodel Version 2 Breeding. Louisiana State University, Baton
Microcomputer Power, III Clover Lane Rouge, Louisiana. pp.126-141.
Ithaca.