Buscar

Genotype x environment interactions. Plant breeding and cultivar recommendations. FAO174, 2002

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você viu 3, do total de 126 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você viu 6, do total de 126 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Você viu 9, do total de 126 páginas

Faça como milhares de estudantes: teste grátis o Passei Direto

Esse e outros conteúdos desbloqueados

16 milhões de materiais de várias disciplinas

Impressão de materiais

Agora você pode testar o

Passei Direto grátis

Prévia do material em texto

Genotype x environment 
interactions 
Challenges and opportunities for plant 
breeding and cultivar recommendations 
ISSN 0259-2517 
Geno pe x environment 
interaction 
Challenges and opportunities for plant 
breeding and cultivar recommendations 
by 
Paolo Annicchiarico 
Senior Scientist 
lstituto Sperimentale per le Colture Foraggere 
Lodi , ltaly 
FOOO ANO AGRICULTURE ORGANIZATION OF THE UNITEO NATIONS 
Rome, 2002 
FAO 
PLANT 
PRODUCTION 
ANO PROTECTION 
PAPER 
T h :l.s One 
1111 ~ 1 ~11 111 ~~ llll ílll 111111 1~ 
7CFY- FRE- FOJC 
Material com-.direil~u~1& 
The designations employed and lhe presentation oi material in this 
inlormalion product do not imply the expression oi any opinion 
whatsoever on lhe part oi the Food and Agricultura Organlzation 
oi lhe United Nations concerning lhe legal status oi any country, 
territory. city or area or oi íts authon'ttes, or concerning the 
delimilalíon oi its lronliers or boundaries. 
Thc n:fcrcnce 10 1hé differcnt co1nputcr soft"an: pro~run~' includcd 1n 
'Cttions 4.5. 5.9 and 7.3 of thi<. publiea11on i~ for info1mn1ion purposcs 
onl)•. Tbe cboicc or i1c~ does o()( c:on,1i1u1c an cnd01scmcn1 by FAO or 
thc 1nanufacturer~ Of tbese co1npu1er software produ b or OÍ lhe J>'OOUCI 
1hemse1' es. 
ISBN 92·5-104870·3 
Ali righls reserved. Reproduction and dissemination oi material in this 
lnlormalion product lor educational or other non·commerclal purposes are 
authorized without any prior written permission from lhe copyright holders 
provided the source is fully acknowledged. Reproduction oi material in lhis 
inlormation producl for resaleorothercommercial purposes is prohibiled w1thout 
\Ynllen permission of lhe copyright holders. Applicatlons for such permísslon 
should be addressect to lhe Chiei, Publishing Management Service. lnlormatlon 
Division. FAO. Viale delle Ter medi Caracalla, 00100 Roma, llaly or by e-mail 
to copyright@lao.org 
© FAO 2002 
Ma1er ai com dire tos au•orais 
Foreword 
'fhe introducti.on of irnprovcd varieties is one of the n1ost po\vertul and cost-efficient mcans 
of enbancing crop productivity and farmers ' incomes. Efficicncy in varietal development 
itself aod in lhe proce.ss of matching varicties to production areas in1plies an understanding of 
plant responses to diverse environrnents and cropping syste1ns in a target production zone. 
M.ultiJocation testing remains the n1ain tool for underscanding varietal responses to environments, 
but t11e process is both tin1e-consuming and expensive. 1·he efficiency of this analytical 
process can be enhanced using recently developed statistical methods. 1'his publication airns 
to support plant brccders by cxa1nining the opponunities offered by such n1ethods. FAO 
hopes that this publication v.1ilJ be useful to a 'vide variety of persons intcrcsted in efticient, 
sustainable use of plant genetic resources, especially those focusing on the improvemeot of 
agriculture in food-deficit. developing countries. 
f ollo\ving introductory ren1arks on the impact of genotype x environ1nent interaction on 
agricultura! production and plant breeding (Chapter 1 ), adaptaüon and yield stability concepts 
are discussed in relation to breeding and the utilization of crop varieties (Chapter 2). The 
potential usefulness of multi-environment yield triais is also exan1il1ed (Chapters 2 and 3). 
·recllniques re.lating to analysis of variance (Chaptcr 4) and n1odelling of adaptation 
patten1s (Chapter 5) are considered for oplimizing variety recommendation and for defining 
the adaptation strategy and yield stability targets in breeding progr-dn11ncs. Attention is paid to 
limits and opportunities for the scaling-up of results tron1 test sites to the target region. ·rhe 
application of selection theory to t11e analysis of 1nulti-environ1nent data as \vell as additional 
indications that rnay be obtained fro111 adaptive traits are also considered (Chapter 6). Concepts 
and n1easurcs of yit.:ld stability and yield reliabi liry, and their utilization for selection and 
recomn1endation of plant varieties, are highlightcd in Chaptcr 7. 
Lnfonnation on useful soft,vare for data analysis is provided throughout the book with 
special e1nphasis on IRRISTAT, a freely-available software program developcd by thc 
lntematio.nal Rice Research Institute. The book also presents a case study (Chapter 8), in 
\Vhich a large n1ulti-cnvironn1ent data set is used for exen1plifying the different analytical 
procedures, as \vcll as the LRRISTA'f con11nands needed for the analysis. 
Eric A. Kuene1nan 
Chief 
Crop and Grassland Service 
Plant Protcction Division 
iii 
Material com direitos autorais 
Acknowledgements 
The revision of some parts of the text by Dr Kayc E. Basford (University of Queensland), 
Dr Hugh O. Gauch (Comell University) and Dr Mike 'íalbot (Biomathematics & Statistics 
Scotland), and the suggestions made by Dr Manjit S. Kang (Louisiana State University) and 
Dr Bruce Walsh (University of Arizona) on specific aspects of the \VOrk, are appreciated. 
Dr Kamel Feliachi {lnsrin1t Technique des Grandes Cultures, Alger) and Dr Alice Pcrlini 
(lstituto Agronornico per l'Oltremare, Florence) are thanked for pem1itting the use of data 
from the AJgerian-ltalian cooperation project "Ainélioration et renforcement du systerne 
nationale d' adaptation variétale du bié dur en Algérie" in the case study. 
The lntemational Rice Researc.h lnstitute (IRRl) is thanked for perrnitting the use of tbe 
software IRRISTAT and its tutorial material and for allO"-'jng its distribution in thc CD attached 
to this publication. Dr C. Graham McLaren and Ms Violeta Bartolon1e (IRRI) provided 
infonnation related to the soflware. 
The excellent work ofMs Ruth Duffy in the editing and formatting ofthe text is greatly 
appreciated. 
Dr Helena Gómez Macpherson (Cereais Officcr, Crop and Grassland Service, f AO) 
made possible the prepardtion and release of this publication. 
IV 
Material com direttos autorais 
Table of contents 
List of abbreviations ., Vil 
l .ist of t a b les ••• V I U 
List of figures X 
1 . l nt r oduction 1 
1. 1 Backw-ound l 
l.2 Genorype and cnvirorunent 2 
1.3 Ele1nents of a breeding strategy 3 
2. Adaptation and yield stabiJity 5 
2. 1 Analysis of adaptation 5 
2.2 Definition of adaptation strategies 7 
2.3 Wide vs. specific adaptation in breeding programmes 9 
2.4 Targeting varieties 10 
2.5 Assessment of yield stability and rcliability 11 
2.6 Two anal}'1ical flov•-charts 12 
3. Multi-environment yield triais 17 
3. 1 Types of triais and rcquirements of the generated infonnation 17 
3.2 Additional infonnation on environmental factors and genotypic traits 19 
4. Analysis of variance (ANOVA) and estimatioo of ' 'ariance components 21 
4.1 ModelsofANOVA 21 
4.2 Estimation of individual effects and con1parison of mcans 25 
4.3 Estimation of variànce cornponents 26 
4.4 Data transfom1ation 27 
4.5 Computer sofuvare 29 
s. Analysis of ada~tation and idcntification of subr~ions 31 
5 .1 Objcctives of the analysis 31 
5.2 Joint linear regression modelling and complementary analyses 32 
5.3 AMMl modelling and complementary analyses 35 
5.4 Factorial regression modelling and comple1ncntary analyses 44 
5.5 Panem analysis 49 
5.6 Data transfonnation 51 
5.7 Unba)anced data sets 55 
5.8 Characterization of subregions and scaling·uE of results 56 
5.9 Computer software 61 
V 
Material com direitos autorais 
6. Definition of adaptation strategy, selection environmcnts, genetic resources 67 
aod adaptive traits 
6.1 Adaptation srrateg)· 67 
6.2 Seleccion environments 71 
6.3 Gcnl!tic resources and adaptive traits 73 
7. l\•leasures of yield stabilit)· and yield reliability 79 
7 . 1 Yield stability 79 
7 .2 Yiel<l reliability 83 
7 .3 Con1puter soft'"'are 86 
8. Case study:durun1 wheat in Algeria 89 
8. 1 Experirncnt data 89 
8.2 Adaptation stratcgy and yield stability targets for breeding 91 
8.3 Cultivar rccommcndation 96 
Referenccs 105 
vi 
Material com direitos autorais 
A~1MI 
ANO\ 'A 
BLUP 
CG.IAR 
CINIMY1' 
CP 
DF 
GIS 
GE 
GL 
Gl.-Y 
GY 
lCARDA 
IRRI 
MS 
PC 
R.EML 
ss 
List of abbreviations 
additive n1ain effects and multiplicative interaction 
ana lysis of variance 
Best Linear Unbiased Prediction 
Consultative Group on lntemational Agricultura! Research 
lnternational Centre for Maizc and Whear l1nprovc1i1ent 
main crossover point 
degrees of freedorn 
Geographic Jnfom1ation System 
genotype x environ1nent 
genotype x location 
genotype x location x ye.ar 
genotype x year 
lnten1ational Centre for Agricultura! Research in the Dry Arcas 
lnternational Rice Research lnstitute 
1nean square 
principal con1poncnt 
Restricted Maxi1nu1n Likelihood 
sum of squares 
' 
vii 
Material com direitos autorais 
List of tables 
2. 1 Co1nplexity of adaptation pattems as depicted by the number of signiticant PC axes in 7 
lhe analysis of GE and GL data n1atriccs of six data scts 
3. 1 Mcan yield of brcad \Vheat varietics ovcr two years and over four otber years in triais 18 
repeated across a fixed set of locations. for hvo independent data sets 
4. 1 ANOVA n1odels including the factors G = genotypc and L = location or environrnent, and 23 
destimation of variance con1ponents, for triais in a randon1ized con1plete block design 
4.2 ANOVA 1nodels including thc factors G = genotypc, L = location and Y = year, and 23 
estin1ation of varíance components, for triais in a randornized con1plete block design 
repeated in sa1ne years (i.e. L and Y crossed factors) 
4.3 ANOVA rnodcls including rhe factors G = genotype, L = location and Y = year, and 23 
esti1nation of variance co1nponents, for triais in a randomized con1plete block design 
repcated in different years in cach location 
4.4 Analysis of variance for 18 brcad \Vheat varieties grO\\tn for three years in 31 ltalian 24 
locations, \Vith partitioning of G L, interaction by: J) joint regression analysis; 2) AMMl 
analysis; 3) definition of four subregions 
4.5 Calculation of genotype and location 1nain effects, and GL interaction effects, frorn 25 
n)ean values of genotypes at each location 
4.6 Values of t · for calculation of Du11nett 's one-tailed ( or Gupta 's) n1ultiple co1nparisons of 26 
the top-ranking genotype \Vith the re1nainiag eurtries 
4. 7 Relationships of environrnent tnean yield \Vith experin1ental error, and of location 1nean 28 
yield \Vith with~in-location phcnotypic variance or standard dcviation of gcnot)1pe valucs, 
'''ithin-location G Y intcraction mean squarc, and avcragc ""ithin-location phenotypic 
variance of annual yield values for individual genotypes, in different data sets 
5. l Major differences bef\veen f\vo possible objectives tor analysis of adaptation 31 
5.2 Mean yield of four genotypes in four locations and across Jocations, rnean yield of locations 52 
and within-location phenotypic standard deviation of genotype rnean yields 
5.3 Squared Euclidean distance betv.ieen four locations based on genotype means in each site 52 
or gcnotypc x location intcraction cffccts 
6. 1 Predicted yield gai11 over the r.egion of lo\vland northern ltaly fron1 selection of luceme 69 
populations pecifically adapted to three or t\vo subregions relative co selection for wide 
adaptation 
viii 
Material com direttos autorais 
6.2 Mean yield of barley lines previously selected for specific adaptation to either of l\VO 70 
subregions or for wide adaptation to thc region of northern Syria, yield ratio of specific 
to \Vide adaptation strategies, and actual yield gajn relative to mean valuc of six control 
cultivars 
7 .1 Criticai values ofl-Iartley's ( 1950) test for con1parison of severa! variances, applicablc to 83 
cornparison of yield stability 1neasures of Type 4 
8.1 Code, altitude, mean yield, sca.led score on the first GL interaction PC axis for original 90 
and log10-transtorrned yield data, \ViJJter mean temperature and rainfall amount of test 
sites 
8.2 ANOVA results and estimate of variance components, for original and transfonned yield 92 
of 24 genotypes gro\vn at 14 locations for ~·o years 
8.3 Comparison of analytical n1ethods for definition of t\vO subregions for durum wheat 94 
breeding in Algeria. Predicted yield gain over the region rrom a specific adaptation strtltegy 
relative to wide adaptatíon 
8.4 Mean yield. environmental variance and Kataoka's índex of yield reliability across 97 
environments, Type 4 stability variance slope of regression on site mean yield, scaled 
score on thc first. GL interaction PC axis, and factorial regression equation for estin1ating 
GL effects, for eight genotypes 
ix 
Material com direitos autorais 
List of figures 
2.1 Lirnited and large extcnt of \i/itJ1in-location variation relative to bet\A.•een-location variation 6 
for a n1ajor environn1ental factor related to the occurrence of GE interaction, and its 
implication on the extent of GE interaction co1nponents of variance 
2.2 .A.. specific adaptation strategy irnplying plant brecding at one national research centre, 9 
sclect ion of novel gem1plasn1 and evaluation of genetic resources in distinct subregions, 
and subsequent phases of farn1er.s' selection in each subregion 
2.3 Flo\v chart of stcps for definit ion of the adaptation strategy and yield stabiLity targets of 13 
breeding progran1mes fron1 analysis of multilocation yield triais repeated in tin1e 
2.4 Flo\v chart of steps for 1naking variery recon1n1cndatio11s from analysis of rnultilocation 15 
yield triais repeated intime 
5.1 Genotype yiclds 1nodelled by joint rcgression and nominal yields u1odclled by factorial 34 
regression oo rainfall amount. for dun1n1 \vheat varieties across ltalian locations 
-
5.2 AMMI analysis of the genotype-locarion data 1natrix in Tablc 4.4 36 
5.3 Scores on the first ru•o GL interaction PC axes of 18 bre-ad \Vheat genotypes and four 37 
ltalian subrcgions 
5.4 No1ninal vields of Jucerne varieties modelled as a function of lhe score on thc first GL 40 
• 
interactjon PC axis of ten ltalian loc~1cions, or the first GE interaction PC axis of four 
artificial environments 
5.5 Scores on thc first two Gf. interaction PC axcs of 20 Jtalian locations, dcfinition of four 41 
subregions for brcad \vheac variety rccommcndation by grouping sites \Vith the sarne 
expected \\1inning genotypc, and classification of locations into five groups based on 
cluster analysis of site scores on PC 1 and PC 2 
5.6 Cluster analysis of test locations performed on site scores on lhe first GL interaction PC 43 
axis, using the lack of significant G L i11teraction \Vithin group of locations as the truncation 
criterion for definition of groups 
5. 7 Geographic position oftest locations for lucemc in northcm ltaly, and provisional definition 43 
of three subregions for a specific adaptation strategy bascd on AMMI analysis 
cornplcn1cnted by clusrcr analysis and additional infornlation on relevant environmental 
variables of sites 
5.8 Yield response of two bypothetical genotypes modelled as a function of site mean yield 53 
for original. log-transformed anel within-location standardized data under the assu1nption 
of \vithin-locat ion phenotypic standard deviation of genotype yields proportional to site 
rnean yield 
X 
Material com direttos autorais 
5.9 Non1inal yields of genotypes 1nodelled as a function of the first GL interaction PC axis of 54 
locations for original and log-transforrned data, \Vith indication of scaled PC 1 scores of 
sites a11d cstirnation or the main crossover point 
5. 1 O Expectedpair of top-ranking cultivars for yield at each combination of J O annual rainfall 58 
by 10 rnean \vinter ten1perature values of locations. according to an AMMI- 1 rnodel in 
\Vhich site score on PC 1 is predicted by 111ultiple regression as a ftinction of the tv.io 
environ1ncntal variables 
6. 1 Hypothetical selection locations for thrcc adaptation strategies cornpared in ter1ns of 68 
predicted yield gains 
7 .1 Yield rcsponses across fíve environmcnts of hypothetical stable-yielding genotypes 80 
according to m•o concepts of stability, using non-regression or regression stability measurcs 
7 .2 Frequency of yield ( or relative yield) values across environn1ents of t\vo genotypes having 84 
sarne mean yield and contrasting stability as mcasurcd by the variance of yield values, 
and estin1ation of a yield reliability index equal lo the lo\vest yield that is expected in 75% 
of cases 
8.1 Geographic position of te-st locations and definition of t\vO subregions for a specific 89 
adaptation strategy base-O on patten1 analysis and AMMJ + cluster analysis results 
8.2 Non1inal yield and non1inal yield reliability as Joy,1est yield expected in 80% of cases, of 98 
eight cultivars modelled as a function of the scores on the first GL interaction PC axis of 
14 locations 
8.3 Expected pair oftop-ranking cultivars for yield at each con1bination of 10 annual rainfa ll 99 
by l O n1ean \Vinter ten1perattu·e values of locations, according to a rv.10-covariate factoria l 
regression 1nodel 
8.4 Expected pair of top-ranking cultivars for yielcl reliability at each combination of 1 O 103 
annual rainfa ll by l O mean \vinter te1nperature values of locations, according LO a tv.10-
covariate tactorial regression rnodel 
xi 
Material com direitos autorais 
1. lntroduction 
1.1 BACKGROUNO 
United Nations projections estimate that the world 
population will continue to gro\v from thc current 6 billion 
to about 1 O billion by 2050 (FAO, 1996). Thc incrcasc in 
population and the subsequent rise in the demand for 
agricuJtural produce are expected to be greater in regions 
where production is already iosuJlicient, in particular in 
sub-Saharan Africa and south Asia (Pinstrup-Andersen 
et ai., 1999). The necessary increase in agricultura! 
production represents a huge challenge to local tànning 
systems and must come mainly from increased yield per 
unit area, given the limited scope for extension of cultivated 
land \VOrldwide (Evans, 1998). 
The so-callcd Green Revolution - based on the 
introduction of in1proved varieties with high yield potentiaJ, 
togetJ1er \Vith technological packages (mineral fertilizers, 
pesticides, inigation etc.) designe<! to significantly improve 
the cropping envirorunent - has greatly contributed to the 
increase in agricultura! production in severa! rcgions 
world\vide. Further expartsion of this high-input model of 
agriculture is not sustainable, ho,vever, due to the high 
costs entailed and the negative impact on natural rcsources 
(Conv,1ay, 1998; Singh. 2000). ln less favourable areas 
with poor ecological potential for crop production, \Vhere 
food insecurity also depends on the marked climatic 
fluctuations from year to year, the strategy has actually 
produced only Jov• and unstable economic rcturns 
(McCo,vn et ai., 1992). Local gennplasn1 is, therefore, 
often preferred to improved varieties because of its greater 
tolcrance 10 sevcre biotic and abiotic stresses (Byerlee and 
Husain, 1993; Bycrlee and Morris .. 1993; Eyzaguirre and 
l\vanaga, 1996; Tesen1ma and Becbere, 1998; Alinekinders 
et ai. , 1994). Furthennore, ongoing clin1atic changes may 
cause tJ1ese arcas to expand in tropical and subtropical 
regions (Rosenzvveig and Hillel, 1998). 
Plant breeding can be expected to assun1e a pivotal 
role in iucreasing tbe availabi lity and stabi lity of 
agricultura! production in the future, particularly insofar 
ac; increasing attention will be paid to: 
• tJ1e sustainability of agricultura] S)'Stems; and 
• d1e develop1nent of fanning systerns in less fuvourable 
arcas (Sleper ct ai .• 1991 ; Ceccarelli et ai. , 1992). 
Ho\vever, national breeding programmes, \Vhich are 
primarily concemcd \vith thesc objectives, rnay need to 
modify some elements of their Green RcvoJution strategy 
to produce (\vhen socio-econo1n ically convenient) 
improved germplasm capable of maxirnizing the 
agricultura! potentiaJ of speci fie areas and fanning systems, 
and of minin1izjng tbe occurrence of crop failures or very 
low yields in unfavourable years. Jntegration of fanners 
in the selection process, co1nmonly detined as participatory 
plant breeding, rnay help fulfil these objectives, in addition 
to facilitating Lhe adoption of novel gennplasrn (Eyzaguirre 
aod 1 \Vanaga, 1996; McGuire et ai. , J 999; Weltzien et 
ai., 1999). ln the long tenn, bioteclmology - if combined 
v.•ith f onns of international cooperation and intellectual 
property rights legislation promoting such techniques for 
national programrnes (f AO, 1999) - may help achieve 
adaptation and yield stabílity targets, especially by using 
plant material with increased tolerance to prevailing bioric 
and abiotic stresses. 
lmproved adaptation and yield stability may derive 
in the long tenn from the definition of an appropriatc 
breeding strategy, and in the sbort tern1 fron1 the 
appropríate choice of cultivars (vvhether indigenous or 
foreign, and either traditional or released from public or 
private breeding institutions). Jdeally, decisions conceming 
the breeding strategy and crop varieties should be based 
on scientific knowledge of the plant 1naterial and its 
relationsl1ip \Vith cropping environments \Vithin the target 
region. This publication highlights the cootribution that 
data from n1ulti-environment triais can provide in this 
respect. Results of studies relative to similar regions 
world,vide, as \VCll as common sense and practical 
considerations, may also contribute to the decision-making 
process, especially '''here there is a lack of more objective 
infonnation based on experimental data. 
It is well documented (Cooper and Byth, 1996) that 
the investigation in breeding progran1mes of adaptation 
and yield stability have been modest Th.is shortfall exists, 
Material com dire11os autorais 
Genotype x e11viró11111e11t i11tere1ctions: challe11[!.es ánd op/)Ortunities for pln11t breetli11g an(/ culrh·nr reco111111entlarions 
despite the irnportance of these issues, the substantial 
investment by public and private institutions in 111ulti-
envirorunent testing, and the \Vide range of statistical 
methods available. Furthern1ore, varieties (or other 
techniques) are rarely recommcnded on thc basis of a 
thorough assess1nent of adaptation and yield stability 
characteristics. An inversion of tl1is trend probably requires 
that ordinary breeders and agronomists be sought as the 
main users of thcse 1nethods. Relatively sirnple tecluliques 
applicable through friendJy, inexpcnsive sot1,vare havc an 
obvious appeal in this context. 
1.2 GENOTYPE ANO ENVIRONMENT 
With regard lo the comparison of plant material in a set of 
n1ulti-environmcnt yicld triais. tl1e tenn genotype refers to 
a culti var (i.e. with rnaterial genetically homogcneous, 
such as pure lines or clones, or heterogeneous, such as 
open-pollinated populations) rat11er than to an individual's 
genetic make-up. ·rhe tenn envirorunent relates to the set 
of clin1atic. soi l, biotic (pests and diseases) and 
1nanagen1ent conditions in an individual triai carried out 
ata given location in one year (in the case of annual crops) 
or over several years (in the case of perennia.ls). ln 
particular, an environ111ent identifies a given location-year 
(annuals) or location-crop cycle (perennials) con1bination 
in tbe analysis of triais repeated over ti1ne. 
Purely environn1entaletfects, reflecting the difterent 
ccological potential of sites and 1nanagernent conditions, 
are nol of dircct conccrn for the breeding or 
recommendation of plant varieties. Genotypic 1nain effects 
(i.e. differences in mean yield bet\veen genotypes) provide 
U1c only rclcvant infonnation \vhen genotype x environn1ent 
(GE) interaction eflccts are abscnt or igt1orcd. However, 
differences between genotypcs may vary widcly among 
environrneuts ia the presence of GE i11teraction effects as 
large as those reported in extensive investigations (e.g. 
DeLacy et ai., 1990; An11icchiarico, l 997a). ln general, 
OE intcractions are considered a hindrance to crop 
improvement in a t.arget regíon (Kang, l 998). Morcover, 
such cffects rr1ay contribute , togethcr \\rith purely 
environmental effects, to thc temporal and spalial 
instability of crop yields. ·ren1poral instability, .in particular, 
lias a negative effect on famJers' income and, in the case 
of staple crops, contributes to food insecurity at natjonal 
and household levei. On the other hand, OE interactions 
may offer opportunities, especially in the selection and 
adoption of genotypes sho• .. ving positive interaction 'vith 
the location and its prevailing environmental condiLions 
( exploitation of speci fie adapt.arion) or of genolypes \VÍth 
- - .----
2 
low frequency of poor yield or crop failure ( exploitation 
of yield stability) (Siinrn,onds, 199 1; Ceccarelli, 1996). 
Gro,ving awareness of the importance of GE 
interactions has led crop genotypes to be ordinarily 
assessed in 111ulti-environment, regional triais for cultivar 
recommendation or for the final stages of elite breeding 
rnaterial selection. GE effects should not be ignored, mther 
analysed using appropriate techniques, in order to explore 
the potential opportttnities and disadvantages. Provided 
the infom1ation frorn thcse triai s responds to certain 
prerequisitcs, it can help brceding progrrun1nes to: 
• better understand the type and sizc of the G E 
interactioas expected in a given region. and d1e reasons 
for tJ1eir occurrence; and 
• define, if necessary, a strategy to succcssfully copc 
\\iith the effects of interactions. 
·rhe most important OE efTects for targcting culrivars 
or for selection of 1naterial are the crossover type affecting 
top-yielding gcnotypcs. Such effccts in1ply a change of 
ranks bet,veen environments rather than a simple variation 
in Lhe extent of the ditference between genotypes (Baker, 
1988). J-lo\vever, ali GE interaction effects arising from 
lack of gcnctic corrclation a1nong environn1ents (including 
those relating to lov.i-yielding material and not ncccssarily 
of the crossover type) can be relevant if the results for a 
given data set are extrapolated to produce infon11ation on 
the GE eftects that are likely to be rnet in breeding for a 
target region (f\11uir et ai., 1992; Cooper et ai., 1996a). 
Reasons for the occurrence of GE interactions are 
thoroughly discussed elsewhere ( e.g. Bidinger et ai., 1996; 
Kang, 1998). Major interaction can bc cxpected \vhen there 
is: on the one hand, 'vide variation betwecn genotypes for 
n1orphophysiological characters conferring resistance to 
( or avoidance of) one or 111ore stresses, and, on the otl1er, 
\Vide variation bet\vecn cnvirontnents for incidence of the 
san1e stress(es) (as detennined by climatic, soil, biotic and 
1nanagement factors). for example, this siruation 1nay arise 
\vhere there is \Vide variation bet\veen material in tenns 
of intrinsic drought resistance or earliness of crop cycle, 
and bet\veen environments in tbe levei ofterminal drougbt 
-
stress. Other pertinent examples 1nay concern the 
differential response of genotypes to variable leveis of 
stress, such as lo'v te1nperature, soil salinity, nutrient 
deficiency, pests, diseases, lodgíng, grazing or interspecific 
competition in 1nixed cropping, as a consequence of: 
genetic variatjon in tolerance to biotic or abioti.c stresses; 
ability to capture and use nutrieot resources; competi tive 
abi 1 ity etc. 
Largc GE interactions have frequently been reported 
Material co1n direitos autorais 
bet\veen pairs of environments \Vith contrasting leveis of 
one n1ajor stress (Ceccarelli, L 989; Brrunel-Cox, 1996), 
detincd as " favourable'' \vhcn charactcrizcd by low stress 
and high mean yield and "unfuvourable" with high stress 
and lo\v yield. H.o\vever, large interactions rnay also occur 
bct,veen pairs of unfavourablc environrncnts and cven 
betv.1een pairs of moderately favourable environn1ents 
possessing s irnilar rnean yield but \Vith differing 
combinations of stresses or patterns of one major stress 
(Annicchiarico, l 997a). For exatnple, an cnvironmcntal 
fact.or such as soil texturc n1ay produce sirnjlar n1can yicld 
of n1aterial but sizeable genotype x soil texture interaction 
(Koutsos e t ai., 1992). The levei of rnatching of 
gcnetical ly-based dcterminants of phenologícal 
developn1ent (e.g. photoperiod and verna li zat ion 
requirements) \Vith site characteristics related LO the length 
of the grov.1ing season (e.g. day-length and temperature 
patterns) is a11other deter1ninant of ren1arkable GE 
ínteraction, cspecially across relatively large regions 
(Wallace e/ ai .• J 993a, l 993b ). 
1·he genetic structure of plll11t material n1ay also have 
a bearing on thc cxtent of GE ínteraction. Variety types 
characterized by lov,• levels ofheterogeneity (e.g. pure tine, 
clone, si11gle-cross hybrid) or heterozigosity (e.g. pure line) 
tcnd to intcract witJ1. thc eovironmcnt more thll11 types \Vith 
opposite features (e.g. open-pollinated population, 1nixture 
of pure tines), because lhe lo\vcr richncss in adaptive genes 
implied by their genetic structure makes the1n rnore 
susceptible to variation in enviro11n1ental conditions 
(Bccker and Léon, 1988; Brancourr-Huln1el et ai., 1997). 
Indced. harvest security is associated v,rith diversity 
bct\veen and within traditiona l varieties for fam1ers 
producing near subsistence levei (Cla\\1son, 1985). 
A detailed description and discussion of various 
aspects ofGE interaction analysis is available in nun1erous 
revie\v articles (Freernan, 1973; 1-lill, l 975; Denis and 
Vincourt, 1982; Westcott, 1986; l .in er ai., 1986; Becker 
and Léon~ l 988; Crossa, 1990; Romagosa and Fox, l 993; 
. 
Cooper and DeLacy, 1994; van Eeu,vijk~ 1995; Brancourt-
Huhnel et ai., 1997; Kang, 1998), in papers included in 
the books edited by Wiltjams ( l976a), Kang ( J 990), Kang 
and Gauch ( 1996), Cooper and 1-Jammer ( l 996a) and Kang 
(2002), an<l in thc n1onographs by Gauch ( 1992), 
Prabhakaran and Jaín ( 1994) and 8 asford and 'f ukey 
(2000). This publication focuses mainly on a limited 
number of anaJ)itical techruques that. O\ving to the quality 
of inforrnation, the ease of applicarion {also in relation to 
the recommended software) and the lirnitcd ru11ount of input 
data required can bc considcred of major interest for: 
• brceding prograrnmes. especially narional programmes 
3 
lntroducrion 
in less developed countries. in order to increase 
kJ10\\rledge of GE interactions and, possibly, n1odify 
thc stratcgy accordingly; 
• breeding progran1n1es, as well as national institutions 
cornmitted to testing and reconunending crop varieties, 
ín order to asscss adaptation and yield srability pan.en1s 
of the available gennplas1n and exploit the infonnation 
for 1nore etfective selection and targeting of 1naterial. 
The second point 1nay be extended to the assessrnent 
of thc adaptability and yield stabílity of diffcrent 
agricultural techniques (e.g. soil tillage, 'veed control, 
SO\ving date) con1pared on a 1nulti-environrnent basis 
(Gauch and Zobel, 1996a; Picpho, 1998). 
1.3 ELEMENTS OF A BREEDING STRATEGY 
Genetic improvement is a basic con1ponent of national 
policiesfor raising agricultural production in many 
countries. Although breeding program n1es run at 
intemational centres of the CGIAR (Consultative Group 
on lntcmational Agricultura! Research) syste1n have rnade 
an important contribuOon to major lood crops, national 
breeding prograrn1nes (public or private) rnaintain a 
fundamental role for crop i1nprovement in tbeir target 
region (usually appl icable to the whole country). 
particularly insofar as the exploitation of speci fic 
adaptation and yield stability characteristics are concen1ed, 
O\VÍng to their better kl10\\•ledge of. and easier access to, 
local gcnnplasn1 and cropping environn1ents. Thesc 
programrnes aim to make correct decísions on a number 
of issues \Vhich comprise a breeding strategy (Simmonds, 
19791). Decisions may concen1, i11 particular: 
• adaptation strategies, yield stability and other (e.g. 
crop quality) targets; 
• gcnctic resources fonning the genetic base (indigenous 
or exotic, fro1n traditional or in1proved varieties); 
• tecl1niques for the recombination and introgression of 
useful genetic variation; 
• variety type ( e.g. síngle-cross hybrid. double-cross 
hybrid, u11proved population or synthetic variety, with 
regard to outbred species); and 
• brccding plan and selccríon procedures (selection 
environn1ents. indirectselection criteria, presence and 
extent of participatory breeding, experin1ental designs 
etc.). 
, lbid .. p. 186. 
Material com direttos autorais 
Genotype x e11viron111e111 i111erac1io11s: challenges and ºJ' portunities for plan/ hreediJ1g and cultivar reco111111e11da tio11s 
The definition of a strategy with respect to GE 
interactions may require decisions on 1nost of these 
elen1ents, nrunely: adaptation strategJ1 and stability targets, 
gcnetic resources, variety type, brecding plan ru1d selection 
procedures. lnitial decisions may change with ti1n.e as a 
consequence of ne\v opportun.ities otfered by scientific 
progress. experimental cvidence, available funding, food 
security policies, changcs in national sced systcrns, 
international cooperation etc., but they should ren1ain 
consistent \vith the breeding objective. For exa1nple, the 
inconsistency bet,veen targeting also unfavourable areas 
and adopting genetic resources and selection procedures 
producing material s.pecifically adapted to favourable 
environrnents has contributed to the partia! failure of a 
nun1ber of breeding progrrunmes carried out in the Green 
Revolutíon context (Simn1onds, 19792; Ceccarelli, 1994). 
lt is worth noting that the potentially extensive 
application of gcnctic engincering techniques does not 
elitninate the need for breeding programmes to copc with 
GE interactíons, because ahnost no cultivar can asse1nble 
., lbid., p . 358. 
4 
-
genes conferring superior perfonnance in ali environ1nent 
types within a relatively large region. This derives fro1n 
genetically based trade-offs bctween yield potcntial and 
tolerance to major stresses, e.g. drought (Ludlow and 
Mucho\v, 1990; Acevedo and Fereres, 1993), as \Vell as 
fro1n the need to choose betv.reen incompatible leveis of a 
key adaptive trait, such as earlíness of flo\vering (WaUace 
et ai. , l 993a). Also the possible selection for yield based 
oa molecular markerS may require a prelintinary definition 
of adaptation and yield stabilíty targets, sínce a remarkable 
portion of uscful n1arkers are cnvironrnent-specific 
(Paterson er ai., 1991; Hayes et ai. , J 993). Jndee<l, tbe 
potential benefits of exploitarion of environ1nent-specí fie 
n1arkers may jus ti Í)", ín the long run, grealer emphasis on 
specific adaptation strategies. ln a \vide adaptation 
prospect, marker-assisted selectioo may prove distinctly 
less effucti ve tban multi-environrnent, phenotypic selection 
for yield in the presence of relatively large GE interactions, 
especially when cpistatic effects contribute significantly 
to genetic control (Cooper et ai., 1999). 
Material com direitos autorais 
2. Adaptation and yield stability 
2.1 ANALYSIS OF ADAPTATION 
Objective and analysed information 
ln an evolutionary biology context, adaptation is a process, 
adaptedness is the levei of adaptation of plant material to 
a given environment, and adaptability is the ability to sho\v 
good adaptcdness in a \vide range of environn1ents 
(Tigerstedt, 1994). ln a plant breeding context, the first 
two terms relate to a condition rather than a process. 
indicating the ability of the material to be high-yielding 
\Vith respect to a given environment or given condicions 
(to v.1hich it is adapted) (Gallais. L992; Cooper and B)1h, 
l 996). ln breeding for wide adaptation (i.e. adaptability), 
the aim is to obtaiJ1 a variety which perforrns v.relJ iJ1 nearly 
ali environments; in brecdiog for spcci fic adaptation, thc 
ain1 is to obtain a variet)< whicb perfonns \Vell in a definite 
subset of environments with.in a target region. 1ne adaptive 
response of a variety is assessed \Vith respect to othcr 
genotypes and tends to undergo n1odi11cation when better-
perfon11ing gennplasn1 becomes available. 
Breeding for \Vide adaptation and for high yield 
stabiLity and reliability have so1netimes been considered 
one and the sa1ne, insofar as the latter t-.vo tem1s indicate 
a consistently good yield response across environ1nents. 
Some authors, ho\vcver, havc applicd the yicld stabiliry 
concept v.1ith respect to consistency in ti1nc of genotype 
performance, using the adaptation concept in relation to 
consistency in space (Barah et ai., 1981 ; Lin and Binns, 
1988; Evans, 1993). lt has also been v.•idely ackno,~rledged 
(Ghaderi et ai., 1980; Becker, 1984; l.~in and Butler. 1988; 
Bo\vman, 1989; Annicchiarico, J 992, I 997b: Romagosa 
and Fox, 1993; Piepho er ai., 1998) tJ1at only genot)1pe x 
location (GL) interacúon, rather than ali kinds of GE 
intttraction, is use fui for depicting adaptai ion pattems, as 
only this interaction can be exploited by selecting for 
specitic adaptation or by gro,ving specitically adaptcd 
genorypes. For exa1nple, the kno\vledge of specific 
adaptation to past years, as sho,vn by positive genotype 
x year (GY) interaction eflects, cannot be exploited in 
future years, since the climatic conditions that generate 
year-to-year environmental variation are not knov.in in 
5 
advancc. This vic\v implies that analysis of adaptation -
and its i1nplications for the definirion of adaptation 
strategies for breeding programn1es and don1ains of 
cultivar recon11nendation for extcnsion services - 1nay 
concern only responses to locations, geographic areas, 
fanning practices or othcr factors that car1 be controlled 
or predicted prior to sowing. ln particular, the analysis 
of 1nulti-environment yield triais sbould f ocus pri1narily 
on GL interaction, \\'Íth tbe characteristics ofthe locat ions 
depending on climatic, soil, biorjc (pests and diseases} 
and crop managen1ent factors. The remaining i11teractious 
of genotype ""ith the tin1e factor (year for annual crop, 
crop cycle for perennials) sbould be dealt \Vit11 in tenns 
of yield stability. ln son1e cases, the conccpt and tl1c 
analysis of adaptation rnay concern lhe genotype 
r~sponses to a set of n1anagernent practices that have a 
crucial i1npact on GE effects, rather than the responses 
to locations ( e.g. Annicchiarico and Piano, 1994). 
·rhe above definition of GL interacrion i1nplies that 
only GL effects rcpeatable in time are of practica l 
itnportance. These effects can be eitl1er: 
• exploitcd, by gro,ving specificatly adapted gennplas1n: 
or 
• n1inin1ized, by gt'O\Ving widely adapted material. 
Non-repeatable GL effects contribute to the second-order 
int.eraction (genotype x locaJio.n x year [GL Y] for annuals, 
and genotype x location x crop cycle for perennials). 'fhey 
represent the error lem1 for (repeatable)GL interaction in 
analysis of variance (ANO\'A) 1nodels \Vbere lhe tin1e 
factor (considered random) is crossed \vith the location 
factor. [n A NOVA models holding the time factor nested 
into location, non-repeatable GL effects are included in 
the GY interaction ,..,ilhin site, \vhich acts as che crror 
term for GL. interaction (see Section 4.1 ). Thc sizc of the 
G L e ff cct s is a lso relevant : Ln pari icu lar, i f the G l, 
mteraction variance con1ponent (although statistically 
significant) is s1nall co1npared to other con1ponents, 
particularly the genotypic one, it reduces U\e possible 
advantage of breeding for spccific adaptation. 
For annual crops in particular. rnultilocation triais 
Material com direttos autorais 
Genotype x environn1e111 ínteractions: cholle11ges anel opportunities for pla111 breetling an<I cultivar recon1111e11darions 
-
FIGURE 2.1 
Limited (situation 1) and large (situation 2) extent of wlthln-locat.lon varlation relative to between-location varlatlon for a major 
environmental factor related to the occurrence of GE interaction, and its lmplicatlon on the extent of GE lnteraction components of 
variance 
Year-to-year variation 
between/within two 
sites for a major 
environmental factor 
( e.g. rainfall) 
Situation 1 
Situation 2 
300 
Situation 1 =substantial GL interaction 
350 
[] 
400 450 500 550 600 
Situatíon 2 =limited GL interaction (large GY and GL Y effects / large GY effects within location) 
should be repeated in Lime 10 distinguish bet,veen 
repeatable and non-repeatablc GL interaction eff ects. A 
nun1ber of reports from various countries (Talbot, 1984; 
Léon and Bccker, 1988; Atl in and McRae, 1994; \Veber 
and \.\lestern1ann, 1994; Sneller and Oombek, 1995; 
Annicchiarico, l 997a), providing estin1ates of variance 
con1ponents or investigating the consistency of GL 
interaction pattcn1s across individual years, have sho\vn 
that a reliable assessn1ent of GL eftects is not possiblc 
'''ithjust one year's data, because the estimation is inflated 
by non-repeatable eftects. The lack of repeatabi lity is due 
mainJy to che year-to-ycar variation in cli rnatic factors 
v.ri thin locations. 
The emphasis on GL intcraction cíl'ccts is justificd 
even \Vhen analysis of adaptation relates directly to 
genotype responses to environmenta l factors (as \Vith the 
statisticaJ n1odels considered in Section 5.4). Sp~ci fica lly 
adapted genol)~pes can be targeted to respond \VelJ under 
the environmcntal condjtions prevailing in a given area, 
provided that these conditions are not highly variable from 
year to ycar. lmportant ir1fonnation rnay not be obtained if 
the focus is on GE ralhcr than GL effccts, becausc the 
inipact on the GL interaction of an environn1cntal factor 
that strongly affects t11e GE interaction (impact v.1hich 
determines thc scope for specific adaptation) may be large 
or lirnitcd, depending on the extent of lhe bet,veen-location 
6 
variation rclative lo thc \Vithin-location, year-to year 
variation for the factor (Fig. 2. l ). For exa111ple, gcnotype 
variatioo in response to rain&1H IIHl) ' appear ímporta11t when 
analysing OE interaction and negligible \Vhen analysing 
GL intcraction, ~1hen test locations have similar n1ean 
value and large year-to-yea r variat ion for the 
environ111ental variable. 
By concentratJng on GL interaction in analysis of 
adaptation, lhe process is a lso s irnpli fied, because 
adapta1ion pattems \Vhich are rernarkably cornplex \Vhen 
evaluated on a GE basis (requiring three or more 
dirnensions for a convenient n1ultivariate representation) 
beco1ne relatively s i1nple on a GL basis (requiring only 
one or l\\'O dirnensions), n1ainly as a consequencc of lhe 
largcr sizc of the appropriate error tcn1l. 'fhis is shovm in 
Tablc 2. 1 for difTercnt data sets, in \Vhich the nu1nbcr of 
din1cnsions are reprcsent.ed by the number of significant 
G E or OI~ inleraclion principal component axes (according 
to an analytical rnodel described in Section 5.3). Reducing 
tbe nuaiber of dimensions has practical implications, 
because one- or t\vo-din1ensional rnodels offer additional 
and simplilied procedures for thc identification of best-
yielding 1naterial for variely recon1n1endatjon and to help 
breeding progran1mes define adaptation strategies (see 
Sections 5.3, 5.4 and 5.8). 
While carlier studies on OE interaction c,onccntrated 
Material co1n direitos autorais 
on GL interaction in the assessment of genotype 
adaptation and zoning of locations (Yates and Cochran, 
1938; Homer and Frcy, 1957; Abou-El-Fiuouh et ai., 
1969), recent developments and applications of statistical 
1nethods have frequently focused on the interaction of 
genotypes v.iith generic environrnents (Crossa et Cl! .. 
1995; DcLacy et ai. , l 996a; Gauch and Zobcl, 1997). 
Ho,vever, most rnethods proposed for GE interaction 
analysis can easily be adapted to the investigalion of GL 
effects. 
Subregions 
Adaptation patterns \Vith respcct to individual locations 
are of 1 i1nited interest per se. as the san1ple of sites is very 
s1nall cornpared to the large nu1nber of locations in any 
target region. Specific breeding, in particular, can only be 
directed to areas and cannot realistically be so fine-tuncd 
as to cxploit positive intcraction cffccts of gcnotypes with 
indjvidual locations. Ho,vever. sites that are siJnilar in 
tern1s of genotype response can be grouped by dift'erent 
n1ethods (discusscd in Chapter 5), and each group 1nay 
identify a croppingarea that is relatively unifortn because 
GL interaction eftects are lintited or negligible. Such areas 
(possibly the object of spcci fie breeding) havc becn tenned 
by dif:l'erent authors as subregions, subzones, subareas, 
n1acro-environments or mega-cnviron1nents (Homer and 
frey, 1957; Seif et ai. , 1979; Ceccarell i, 1989; ClMMYT. 
1989). Subrcgions may also be dcf1ncd for variety 
reco1nrnendation: each subregion Lhen coincides v.iith a 
recon11nendation domain, grouping those sites \Vith the 
sarne best-perfon11ing genotype(s) (Gauch and Zobel, 
1997). Tilc dcftnition of subregions is not justgeographical, 
but may also enco1npass fanning practices (e.g. irrigated 
or rainted croppíng). 
Subregions have son1etin1es becn defincd on thc basis 
of site sim i larity for cnvironn1ental factors that are 
supposcd to be itnportant but are, in fact, chosen \Vithout 
a definite assessmcnt of their impact on GI. inreraction 
(e.g. Pollak and Corbert, 1993). 1·he arbitrariness of this 
procedurc 1nakes its results less reliable than thosc for 
site sin1ilarity for GL eOects (Gauch, 19921). Ho\vever, 
addition.al intonnation on the clin1atic, soil, biotic and crop 
n1anagcn1ent variablcs closcly rclatcd to Lhe occurrencc 
ofGL intcraction n1ay help locate geographic boundaries 
for subregions (see Seccion 5.8), besides contribucing to 
thc understanding of causal factors for thc inreraction 
(Eisemann et ai., 1990; Bidinger e/ ai., 1996; van Ecuwijk 
1 lbid .. p. 220. 
7 
il<la1Jta1io11 and yieltl stabilíty 
TABLE 2.1 
Complexlty of adaptation patterns as depicted by the number 
of slgnlflcant PC axes in the analysls of GE and GL data 
matrices of slx data sets 
Data SCI No. No. No. Significant PC axesº 
i1e ycars genotypes OE GL 
Bread \vheat 31 3 18 5 2 
Ourum wheat 1 6 3 9 4 2 
Ourum \vheat 2 5 2 15 4 1 
Oururn \vheat 3 6 2 12 4 1 
"'laize 1 11 3 13 4 o 
Maize 2 11 3 11 3 1 
• P s O.OI atcording to F011~ te.si (Cornclius, 1993). in an Al'\ll~ll 
analysis. 
So11rce: Annicchiarico. l 9<)7a. 
et ai., 1996). The zoning process should produce 
subregions tJ1at can be defined on a geographical basis 
or by other rneans, such as cli1natic factors or 
1nanagen1ent practices, in order to be useful for brceding 
or cultivar reco1n1nendation. ln reality, this is not al\vaysthe case: for exan1ple. subregion defrnjtion for maize 
variet)' reconunendation in Jtaly revealed a leopard skin 
patt ern as a result of the inconsistency bet\veen 
geographic proximity and similarity for GL interaction 
effects of the sites, \vhich could not be accounted for by 
any of the availnble enviroru11ental variables (Annicchiarico 
et ai., 1995). ln tl1is case, rccom1nendations concen1ed 
thc \vhole region. ln other ínstances, individual locations 
or small groups of sites (appareat ly distinct, in tenns of 
GL eftects, fron1 a larger group of geographically close 
locarions) 1nay bc added to thc larger group to fonn a 
uniquc subrcgion for breeding or recom1nendation ( e.g. 
Annicchiarico and Perenzin, 1994). 
2.2 DEFINITION OF ADAPTATION STRATEGIES 
Sctting adaptation strategies for breeding progra1n1nes and 
deJining recomrnendation domains for cultivars are distinct 
objectives. As such, they n1ay require partly diffcrcnc 
analytical approaches and provide different results \Vith 
regard to the definition of subregions. 111e sa1ne data set 
1nay be analysed with both objectíves in r11ind. Howevcr, 
lhe adaptation stratcgy objcctivc focuses on the responses 
of a sct of gcnotypcs to obtain indícatíons and generate 
predictions rclative to future breeding material that may 
bc produccd from the gcnetic base of v.1hich the tested 
genotypes are assu1ned to be a representative satnple. 
Assessing the value of a speci fie adaptation strategy, 
implying a ilistinct selection progrru11111e for each subregion 
rather than a unique selection progra1nn1e for the whole 
Material com direitos autorais 
Ge1101ype x e11viro11me11t interactions: cha/le11ges and opportunities .for pla111 breeding an<I c11/tivar recotT1111e11da1íons 
target region, is of obvious interest to globally-oriented 
breeding programa1es of large seed companies or 
international research centres, where the target region 
may include more than one country and very di fferent 
environments. ln this case, each subregion may include 
several countries. Specific adaptation, ho\vever, may also 
prove a valuablc target for national breeding programmes, 
for \\rhich the yicld gain derived from exploitation of GL 
interaction effects \Vithin tbe country can also help face 
the iJ1creasing con1petition exerted on local seed markeLS 
by intemational seed companies. For public mstitutions. 
the brceding of cliversiiied, specifically adapted gennplasm 
can be a major element of a research policy enforcing 
sustainable agricu lture ( Brarnel-Cox et ai., 1991; 
Ceccarelli, 1996) by: 
• maximizing the potential of diffcrent areas by fitting 
cultivars to an environment instead of altering the 
environment (possibly with costly or environrnent-
unfriendly inputs, such as pesticides, fertilizers and 
irrigation) to fit widely adapted cultivars; and 
• sateguarding crop biodiversity by increas ing tbe 
number of varieties under cuhivation, \Yith positive 
implicarions tor the stability of production at national 
levei. 
Furthermore, speci fie brecding may faci 1 itate thc 
t.ecbnological adaptation of varietics by fixi ng 
characteristics of specific interest to subregions (for srnall-
grain cereais, shon st.ra\Y for intensive cereal fanning and 
long straw for extensive cereal-livestock syslcms; for cercai 
or food legume crops, different grain quality characteristics 
etc.). 
Jn general, breeding for spccific adaptation tcnds to 
imply greater genetic gains associated \Vith increased costs 
relative to those for a \Vide adaptation strategy. The genetic 
gains are derived from exploitation of GL interaction effects 
via useful adaptive traits (Bidinger et ai., 1996), as \vell 
as increased heritability of yield as a consequence of 
decreased GL interaction (Kang, 1998). The relatively 
lúgh costs may be due to increased tield testing rather 
tl1an to dupHcation ofbreeding stations, because crossing 
and hybridization operations can be centralized in a single 
national station providing each subregion v.•ith novel 
gertnplasm for local selection (and, possibly, genetic 
resourccs for local testing to identify parent material of 
specific interest). Figure 2.2 presents a hypothetical 
situation including a pbase of farmers' selection in a 
participatory plant breeding sche1ne (v.•hich may or may 
not be prescnt}. 
A comparison of 'vide vs. speciflc adaptation 
8 
stratcgies rnay be based on: 
• yieJd gains predicted for the di ff erent options according 
to sclcction theory, using thc sarne yi.eld data already 
used i.n the preliminary definition of subregions; or 
• actual yield gains provided by different strategies, 
following further research \York ( e.g. Singh et ai. , 
1992; Ce.ccarellj et ai., J 998). 
ln fact, n1ore con1plex scenarios (including indirect 
selection gains in one subregion, derived from direct 
selection in another subregion) can also be envisaged (see 
Section 6.1 ). The con1parison bet\vecn adaptatjon 
strategies based on predicted yie ld ga ins may 
underestirnate the potential advantage of a specificstrdtegy, 
\vhen 1naterial \Vith n1arkedly specific adaptation is Wlder-
represcnted among thc t.estcd varieties or breeding lines 
for various reasons ( e.g. previous selection for wide 
adaptation by local breeding or \Vide representation in a 
sample of foreign, 'videly adapted materiaJ). Like,vise, the 
potential gain of specific breeding for ttnfavourable arcas 
may be underestimated \vhen most tested genotypes have 
been selected in favourable environments. Specific 
adaptation could be more advantageous if it also implied 
the use of a distinct genetjc base for eacb subregion. 
Although not negligible, the positive effccts of specific 
breeding on tl1e biodiversity of cultivated. rnaterial are 
difficult to quantify. 
For the definition of adaptat.ion strategies, the 
con1parison of yield responses across environn1ents 
managcd for contrasting leveis of a given stress (e.g. 
drought or nutrient deficiency, dependent on lhe levei of 
irrigation or fertilization, respectively) is a valid alten1ative 
to lhe current approach (based on thc analysis of regional 
triais managed according to ordinary practices in each 
area), provided that the managed environments can 
represent actual geographic arcas or fam1ing systems 
within the target region. The use of n1anaged environments 
is currently considered 1nai11ly for genotype selection 
within a given adaptatjon strategy (see Section 6.2). 
Even if a \Vide adaptation strategy is preferred, the 
provisional identi fication of subregions can hclp locate 
crucial test sites for gennplasm selection (Abou-El-Finoub 
et ai., l969; Lin and Butler, 1988; Annicchiarico, 1992). 
ln th.is case, a fe\v sites representative of different 
subregions and capable of reproducing lhe n1ean responses 
of genotypcs across thc region can be used for selection of 
material eithcr conten1porarily (Brennan et ai., 198 l ) or 
in tum (Calhoun er ai., l994). Jdentifying crucial test sites 
can also be a valuable objective for the routiue evaluation 
of genotypes carried out by public institutions, such as 
Material com direttos autorais 
.4(/áptarion <Znd yie/(/ stability 
FIGURE 2.2 
A speciflc adaptatlon strategy lmplying plant breedlng at one natlonal research centro, selectlon of novel germplasm and 
evaluatlon of genet lc resources ln dlstlnct subreglons, and subsequent phases of fanners' selectlon ln each subreglon 
Natlonal Research Centre 
(acquisition of genetic resources; recombination of geoetic variatioo and generation of novel gelTTlplasm; coordination of 
test work; possible selectioo for unifOITTlity and muttiplication in isolation) 
information 
on novef 
germpl asm 
• 
• 
• 
• 
Test site 
1 
novel germplasm 
• ' 
genellc resources 
• 
' 
Test site Test siteinformalíoo 
on useíul 
nls pare 
1 
• 
• • 
• 
• 
in subregion A in subreglon B ln subreglon e 
• • 
• 
• 
• 
• • 
• 
• 
• 
• 
·--· • 
(preliminary testing of novel gennplasm; testing of genetic resources) 
--.. l 
• 
• 
• 
• 
• 
• 
• 
• 
• novel gennplasm novel gennplasm novel gennplasm ! 
• 
• 
• 
• 
• 
' 
• ' 
, , J 
F armer 1,2, ••• n Farmer 1,2, ... n Farmer 1,2, ... n 
• 
• 
• 
• 
• 
• 
---
. ln subreglon A ln subreglon B ln subregion e 
' ··-
(cycles of farmers' selection) 
, , 
' 
r , , 
Seed systems Seed systems Seed systems 
in subreglon A ln subreglon B ln subreglon e 
(mulliplication/dístribution of novel cultivars) 
• 
, 
.. ~ 
Farmers Farmers Farmers 
ln subreglon A ln subreglon B ln subreglon e 
(cultivation of novel cultíVars) 
t11ose cornrnitted to the defin ition of list.5 of recom1nended 
varieties, or tbose responsible for the assessment of the 
Value fo r Cultivation and Use of ne,v ly released 
gem1plas111. 
Decisions on the adaptation strategy, \vhich can havc 
a considerable and lasting effect on the organii.ation of a 
plru1t breeding programn1e, sbould be based on the analysis 
of rnore data sets if available, and veritied after a 
rcasonable period oftin1e on thc basis of ncv.• data (DeLacy 
et ai., 1994). 
.2.3 WIDE VS. SPECIFIC ADAPTATION IN 
BREEDING PROGRAMMES 
.A..lready in the l 920s, studies of plant material (Turesson, 
1922) leading to tl1e introduction of the tenn "ecot}rpe" 
highlighted the occurrencc of specific adaptation to certain 
areas and environrnental conditions. During the sarne 
9 
period, pioneer breeders in Great Britain (Engledow, 1925) 
and Strarnpelli, ltaly (Lorenzetti, 2000) advocated the 
importance of understanding and exploiting specific 
adaptation etTects io order to raise crop yields in their 
rcspcctive COllntr ics. Ho,vcver, breeding programmcs in 
the second half of lhe century mostly concentrated on the 
improven1ent of yield potential (Bra1nel-Cox et ai., J 991; 
Evans,, 1993) using - also in lcss developed countrics -
favourable environments for selection (Simrnonds, 1991 ). 
Also the concept of plant ideotype to select for has been 
developed for dilTerent species to increase yields undcr 
favourable cropping conditions (Donald, 1968; Mock and 
Pearce, 1975). This trend has been favoured by numerous 
factors: 
• tJ1e perspective of rapid yield gain offered by high input 
leveis; 
• lhe greatcr profitability of target:ing sccd markcts in 
1nore productive arcas; and 
Material com dire11os autorais 
Genntype x environn1en1 í11terac1iu11s: c:halle11Kes anel opporl1111ities for /Jfont breeding and cultivar r eco1111ne11datio11s 
- ~ -
• the con1111on belicf that select ion in favourable areas 
produces a substantial yield increase also in lcss 
favourable arcas. 
·rhis beliefhas been challenged by theoretical \VOrk (Jinks 
and Connolly, 1973; f'a lconer, 1990; Sin1monds, 199 1) 
and by n1ounting experin1ental evidence concerning 
improvemcnt of crops for environn1ents sufterjng fro1n 
drought stress (Nages,vara Rao et ai., 1989; Ceccarelli 
and Grando, 199 1; Ud-Din et ai., 1992; Byn1e et oi. , 1995; 
Mu11oz et ai. , 1998; Ceccarcll i et ai., 1998), nutrienl 
deficiency (Atlin and Frey, 1989; Banzinger et oi., 1997) 
or both (Cooper et ai., 1997). ·rhese works have 
highlighted that selecting under IO\V stress conditions 
i1nplies a specific adaptation target to,vards favourablc 
areas. 
ln reality, favourable, high-yielding environ1nents for 
selection are not necessarily so because of natural cli1natic 
or soil factors. Optimal management at research stations, 
including abundant fertilizaüon, accurate soil preparation, 
ti1nely so,ving, irrigation and che1nical control of \Veeds. 
pests and diseases, can often result in higher mean yields 
tJ1an in neighbouring fanners fields. Such differences in 
yield may also i1nply remarkable GE interaction bet,veen 
these envi ronn1ents (Pederson and Rathjen, 198 1; 
Ceccare l 1 i, 1994 ). Spaced planting or lo''' densi ty 
procedures adopted fo.r selection n1ay also rcsult in 
i1nproved cropping environments, because they i11crease 
the availability of resources, such as light, \Vater and 
nutrients (Rotili and Zannone, 1975). 
ln recent decades, various progran1mes run at 
international research centres havc n1odíficd their breeding 
stratcgy to produce germplasm suitable for cropping in 
less favourable areas. A specific adaptation strategy has 
been pursued for barley i111provement at the lnternational 
Centre for Agricultura! Research in Dry Areas (lCARDA), 
'vitl1 selcction undcr conditions si1ni lar to those in the target 
environ1nent {favourable or drought-stressed). aod witb 
ímplications ext.ending also to elen1cnts of lhe breeding 
strat.egy, such as the choice of genetic resources and variety 
type (Ceccarelli, 1994). A ' 'shuttle breeding" procedure 
- altcmate seleclion in drought-stresscd (unfavourablc) 
and irrigated (favour-dble) environments - \vas established 
for the selection of \Videly adapted gem1plasm at CI MM YT 
(Edtneades et c1/., 1989; Calhoun et ai., 1994). l f o\vever, 
the possíbility of C;E intcraction bet\vcc11 environn1ents of 
si1nilar ccological potcntial (bascd on crop 1nean yield) 
n1ay lcad to tl1e definition of di fTercnt subregions also 
within unfavourable and n1oderately favourable areas. For 
this reason, barley breeding at ICA_RDA began producing 
10 
specific n1ateria l for diffcrent sets of drought-prone 
countries (Ceccarelli, 1996); rice breeding at IRRJ defined 
speci fie plant ideOl)'pcs for severa] diffcrent ecosyste1ns 
(Fischer. 1996 ); and 'vheat breeding at CitvfMYT is 
attempting a compro1nise bet\veen a \vide adaptation 
prospect and d1e opponunity to breed specifrcally for 12 
difTcrenl 111ega-environn1ents (Braun et ai., 1996). 
Di fferent subregions n1ay be identified notonly within 
Jarge or transnational rcgions (e.g. Crossa et ai., 1991; 
DeLacy er ai., 1994) but also ~· ithin relatively srnall 
regions, as suggested by results for: 
• nor1hcn1 Syria (Ceccarelli. 1996); 
• Italy (Annicchiarico, l 997a) and northern ílaly 
(Annicchiarico, 1992, 2002); 
• Nc• .. v South Walcs (Seif et ctl., 1979; Basford and 
Cooper, 1998); 
• Queensland (DeLacy et ai., J 996c); 
• south~·est Canada (Saindon and Schaalje, 1993); and 
• Ontario (Yan et ai., 2000). 
'l'herefore, lhe choice bet\veen a \Vide and a specific 
adaptalion stratcgy rnay be a kcy question for national 
breeding prog.rammes. 
The spec ific adapcation strategy is receiving 
increasing attcntion in dcveloping countries. son1etimes 
in co111bination ~1ith participatory plant breeding schernes. 
·r hc participation of farmers can: i) support the 
n1ultilocational selection \vork; ii) allo\v for exploiring 
possible spccific adaptation effects cven \Vithin subregions; 
i ii) contributc to enhance thc biodivcrsity of material under 
cuhivation, thereby itnproving production stability; and 
iv) fac ilitate the seed supply to farmers via local seed 
systems (McGuire et ai .. 1999; \Vel.tzien et al., 1999; 
Ceccarelli et ai., 2000). For sn1all farrners in relatively 
poor countrics, such systc1ns 1nay be far rnore important 
than fon11al seed systcn1s (Almckinders et ai., J 994). 
2.4 TARGETING VARIETIES 
Agricultura! techniques, to llo,ving an assessn1enl of thcir 
econo1nic perfonnance in regional yicld triais, can be 
recommended: either \Videly over the targel region or 
specifically for one subregion (Perrin et ai. , l 976; Shaner 
et ai. , 1982). For speci fic reco1n1nendation, the 
reco1nn1cndation don1ain of a given technique 1nay be 
dcíined on thc basis of gcography alone, or also on the 
basis of fa r1n ing practices (e.g. irrigated or rainfed 
cropping) or socio-economic consrraints. lna li cases, tbe 
int-Onnation obtained fron1 previous testing is exploited 
for predicting yield responses in corning years and, n1ost 
Material com direitos autorais 
frequently. in ne''' locations. 
The definition of reco1nntendations tor varieties is 
generally sirnpler than for other techniques. Usuall)' there 
is little difference in secd pricc among cultivars. fn this 
case, variable costs do not need to be taken into account 
in the assessn1ent. Statistical analysis concen1s yield 
response anel, \vhcrc thc genorypes diffcr for crop value. 
gross benefit. ln either case, di fferent domains (i.e. 
sub regions 'vhich are the object o f di stinct 
reco1nmendation) may be ídentificd on the basis of 
repeatable GL interaccion e!Tects \vhen varieties have 
diffcrenr adaptation patterns. 'f he dornains are 
charac terized by different top-yielding, .. ,vinning" 
genotypes (Gauch and Zobel, 1997; Ebdon and Gauch, 
2002). For cach location, chc advantage of spccific vs. 
v.1ide recommendation can be assessed in tenns of the 
di fterence in predicted )iield bet .. veen: i) tlte top-ranking 
gcnotype(s) in the rclevant subrcgion; and ii) thc top-
ra1lking genot)•pe(s) over lhe region (Gauch and Zobel, 
1997). 
More tlian one recomrnended genotype can be a 
sensible choice for either scenario, particularly for ~ride 
reco1n1nendation, as it n1ay lin1it the risk of disasters arising 
due co t11e unforeseen susceptibility of the only culti var 
reco1nn1ended in a vast. area to a biotic or abiotic stress. 
This is particularly the case for the cultívation of 
gcnetically homogcneous variery types (pure lines, clones) 
(Simn1onds. 19792). Possible differences in yield stability 
bet\veen cuJtivars can also be taken into account tnaking 
reco1nn1endarions (see sections bcfO\\' and Chapter 7). 
Favouring the cultivation of specifica lly adapted 
gennplasm is generally convenient tor n1aximizing regional 
yields and increasing chc biodiversity of cult ivated 
material. However, subregions \VÍlh very lin1ited extension 
or negligible advantage of specitically adapted cultivars 
1nay be 111erged wich larger, relatively si1nilar subregions 
v.1hen thc additional costs of 1nultiplying and marketing 
speci tically adapted gennplas1n are likely to out v.1eigb the 
expected benefits (Gauch, 19923). 
Targeting of gcnotypes is also a concern of public 
and private seed cornpanies \Vho 'vísh to veri fy the area of 
adaptation and the agronomic value of novel gem1plasn1. 
This information, ncccssary for planning propcr rnarkcting 
and advisory scherncs, is cspccially 'varrantcd when 
breeding has not cont.emplated a defini te adapt.ation target. 
ln this case, just one top-yielding genOl)'pe, across the 
2 lblá .• p. 267 and p. 361. 
l lbid .. p. 220. 
11 
1t<laptatio11 (lnd .11ield s1abilir)1 
region or in a distinct subregion, is usually identified for 
pron1otion to con1n1ercial variety status. Tech11iques 
envisagcd herein for defining variety recornmendations 
can also be usetul for this purpose. 
2.5 ASSESSMENT OF YIELD STABILITY ANO 
RELIABILITY 
Yield stability 
High yield stabi lity usually refers to a genotype's ability 
to perfonn consistently, \Vhetlter at high or low yield leveis, 
. 
across a \Vide range of cnvironrncnts. As discusscd rnorc 
thoroughly in Section 7. 1, most stability 1neasures relate 
to eicher of t\vo contrasring concepts of stability: "static" 
(Type l) and "dyna1nic" (1ype 2) (Becker and [,,éon, 1988; 
Lin et ai., 1986). Static stability is analogous to the 
biological concept ofhorneosrasis: a stable genorype tends 
to maintain a constant yield across environmcnts.1'he tenn. 
"environrnental sensiüvity" has also been used in this 
respect. \Vhere greater sensitivity corresponds to lov.··er 
stability (Falconer, 1990; Oyke et ai. , 1995). Oynamic 
stability in1plies tor a stable genotype a yield response in 
each environ1nent that is al\l.iays parallel to the ntean 
rcsponse of ihe tested genotypes. i.e. zero GE interaétion. 
TI1e n1easure of dynarnic stability depends on the specific 
set of tested genotypes, unlike the n1easure of static stability 
(Lin e1 ai., 1986). Lin and Bu1ns (199 1) defmed a i ·ype 4 
. 
concept of stability that is strictly related 10 thê stacic 
concept. Type 4 stability relates to consistency of yield 
exclusively in ti1ne, i.e. across years ( or crop cycles) \vithin 
locations, \Vhereas Type 1 stabí lity relates to consistency 
both in ti1nc and in space, i.e. across environrnents 
belonging to the sarne or different sites. 
As \\'ith GL effects, the GE effects contributing to 
yield s1ability can be either: 
• exploited, by breediJ1g and grO\VÜlg genotypes that are 
stable according to the static concept (i.e. \Vith a better 
response in unfavourable environments or years); or 
• ntinin1 ized. by using 111atcrial that is stable according 
to the dynan1ic concept. 
Static scabilit)' may be n1ore useful than dynarnic in a 'vide 
range of situations, cspecially in developing countries 
(Simmonds. l 99 1 ). 
From a fam1er 's point of vie,v, location is a constant 
- not variable - fac1or, and yield consistency over cime is 
thc only rclcvant componcnt of a genotype's yicld stahility. 
lt 'vas therefore proposed (Barah et ai., 1981; Lin and 
Binns, 1988) to evaluate yield stability \Vith regard to GY 
interaction effects \vithi n locations. Jn reality, yield 
Material com dire11os autorais 
Ge11oty1Je x enviro11n1e111 interactions: cli<1lle11ges ond opport1111ilies for fJ/ant breedíng and cultivar reco111111e11<lotions 
- -
c-0nsistency in space also dcscrvcs consideration in tJ1e 
presence of sizcable GL intera.ction, since a selected or 
re<:ommended genotype should be stable-yielding both 
across years and across locations in its area of adaptation 
or recon1n1endation (Picpho. 1998). This is particularly 
so \Vhén there is a prospect of 'vide adaptarion or 
reco1nn1endation, because in the contcx1. of a specific 
adaptation or recommendation the G L effects are 
tninimized by the division of the target region into 
subregions. ln a wide prospect, the assess1nent of yield 
stability in relation to genotype responses to environn1ents 
provides a simple n1eans for considering all possibly 
relevant GE interaction effecrs. ln a specitic prospect, 
assessn1ent bascd on GY interaction e[fects \.Vithin 
Jocations can be recomn1ended, especially for genotypc 
reco1n1nendation (sec Sect ion 7 .2). Wha tever the 
adaptation strategy, breeding for high yield stability can 
be considered a useful targe1 \vhen thc rclevant Gt. 
interaction variation is wide. 
Yield reliabllity 
li igh )' ield stabi 1 ity 1nay be a.'!sociatcd \Vith lo'v 1nean yield 
(or IO\V stability v.iith high mean yicld), v.rhich con1plic,ates 
genotype selection or recommeodation. As an extreme 
exan1ple of high stabiliry associated \Vith low yicld, 
considera hypothetic genotype that yields jus1 above zero 
in ali environments (greatest stability according to its static 
concept), or that is consistently the least-yielding (greatest 
stabi lity according to the dynamic conccpt). Obviously, a 
less stable, higher-yielding genotype \vouJd be preferable. 
·rhe practical interest of co1nbining l1igh leveis of rnean 
yield artd yield stability has led to the develop1nent of the 
yield rcliability concept (Eskridge, 1990; .Kang and Pham, 
199 1; Evans, 1993). A reliable genot)rpe is charact.erized 
by consistently high yield across environments. The use 
of a yield reliability índex faci litates genotypc sclcction 
or reco1n1nendarion, as the n1can yield and the yield 
stability traits are combined into a unique measure of 
genotype merit. Considering yield stability in conjunction 
\Vith mean yield may also provide a more sensiti\re 
c-0n1parison of gcnotypes than \vhen using only n1ean yield, 
O\vingto lhe reduction in ·rype 2 error rates (Ka11g, 1993). 
The stability 1neasure contributing to lhe index of yield 
reliability may relate to the static or the dynan1ic concept 
of stability (see Section 7 .2). 
Genetic improvement 
Despite its potcntial interest, increased yield stability has 
tended to bc a n1inor objectivc in brccding progra1nroes 
\Vorld,vide (Rotnagosa a11d fox, 1993). A nurnber of 
12 
studies revie\\'ed by Becker and Léon ( 1988) and 
Brancourt-liuh11el e1 ai. ( 1997) confimied the early 
indicatioo by Allard and Bradsbaw (J 964) that variety 
types \vhere rhe genetic strucrure implies high leveis of 
heterozigosiry and/or heterogeneity are less sensitive to 
environ1nental variation and are, therefore, more stable-
yielding. Unfortunately, such types 1nay somelimes offer 
fe\ver opportunities for n1axi1nizing the yield potential. 
Within a given varicty type, breeding success.fully 
for this tra it relies on thc adoption of a heritab le or 
repeatable stability rneasurc as a sclection criterion. ·rhere 
is extensivc evidence ( see Section 7. l) that heritab.ility or 
repeatability values tend to: 
• vary fron1 1noderate to low, depending on the situation; 
• be higher when lhe assessn1ent is done over two or 
more years or crop cycles; and 
• be somc\vhat higher for measures that relate to lhe 
static ( fype l or ·rype 4) concept of stability. 
Given tl1e high santpling crror. thc assessrnent of yield 
stability rcquires numcrous tesl environments (at least 
eight) to !:,ruarantee reliability (Kang, 1998; Piepho, 1998). 
Therefore, direct selecrion for yicld stability 1nay be limited 
by high costs and can bc rccommended. evcn whcn it has 
higb priority, only for elite n1aterial in lhe final testing 
stages. ·rhe eh o ice of parental gen11plasrn \vitl1 recognized 
yield stability and of, if possible, a convenientvariety type, 
can play a major role in breeding for more stable crop 
yields. Jn addition, indírect selection for higher yield 
stabi lity n1ay be atternpted by select ing for 
n1orphophysiological traits thathavc provcd to be strictly 
associated \vith this character (see Section 6.3). 
2.6 TWO ANALYTICAL FLOW CHARTS 
Adaptation strategy and ylefd stability targets 
111e main analyticaJ steps involved in tJ1e definition of an 
adaptation strategy and yield stability targets on the basis 
of regional triai data are stun1narized in Figure 2.3 for lhe 
general case of experirnents repeated also in tirne. There 
are six possible conclusions, i1nplying a \.vide or specifíc 
adaptation stratcgy and, in both cases, the inclusion or 
exclusion of incrcased yicld stability as a breeding target. 
WiU1in lhe wide adaptation strategy, indications rnay or 
n1ay not urge the choice of selection locations that contrast 
for GL interaction effects. 
As a preli1ninary srep, it is uscf'ul to estin1ate the 
variancc components for: genotype; GE interaction across 
environments ( location-year or locacion-crop cyc le 
combinations); and the t\vo determinants of GE interaction 
Material co1n direitos autorais 
Adaptation and yield stability 
FIGURE 2.3 
Flow chart of stops for deflnition of the adaptation strategy and yield stabillty targots of broedlng programmes from analysls of 
multllocatlon yield triais repeated intime 
Start 
FAJRLY 
HIGH 
FAIRLY 
HIGH r------. 
lack of 
genetic 
correlation 
across 
environments 
variance 
of Gl 
interaction 
>-i.i analysis of 1-t.i provisional H~ 
specific 
adaplation 
advantage adaptation definition of YES 
subregions 
NO 
LOY/ 
variance of 
• vanance variance otherGE 
of olherGE of Olhe< GE interactions 
interactions ínteractions in subregion 
LOW LOW 
1 1 3 LOW HIGH 5 HIGH HIGH 
selection for wide 
adaptation On few environments 
across reçiion) 
selectioo fOf wide adaptation selection for specífic adaptatioo 
(in few envirooments in 
subreçiion) 
(on cootrasting sites in region) 
2 
4 
selection for ...nde adaptatioo 
and yield stability (in severa! 
envtronmeots across fe9lon) 
selectioo for wide adaptation and yield 
stability (in several environments on 
contrasting sites across region) 
selection for specific adaptation 
aod yield stabil~y [m several 
environments in su.bregioo) 
Note: Envirorunent as location-year or location-crop cycle cornbination (see Section 2.6 for co1nn1ent of steps). 
variance, i.c. thc lack of genctic correlation a1nong 
environ1nents for .genotype values, and thc hetcrogeneity 
of genotypic variance among environrnents (see Section 
4.3). Adaptive responses and yield stability of genotypes 
as affected by heterogeneity of genotypic variance have 
no practical i1nportance for brccding, sincc they relate to 
GL and other GE effects originating fro1n a scale effect of 
the environrnent ar1d i1nplying no change in relative 
rcsponsc of gcnotypes. The larger variance of this 
dctcmlinant relative to the lack of genetic correlation 
~ 
variance detinitely suggests its re.duction tlirough a suitable 
data transforrnation (as discussed, in particular, with 
respect to analysis ofadaptation - see Section 5.6). ln any 
case, i f the lack of genetic correlation variance con1ponent 
is lo\v (e.g. belo,.,. 25-30%) con1pared to tlie gcnotypic 
variance component, it reveals the limited extent of GE 
interaction etfects relevant to breeding, and supports 
13 
(wi thout fu rther analyses) the selection for wide 
adaptation \Vith no regard for yield stabi lity (Fig. 2.3). 
To sirnplify lhe analysis (see Section 5.6). this preliminary 
step n1ay be replaced by the estimation of the genotypic 
and GE interaction con1ponents of variance only, together 
with the verification of the reJationship between location 
n1ean yield and within-location phenotypic standard 
deviation of genotype yields. If no correlation is found 
and tJ1e OE i.nteraction variance con1ponent is larger than 
or about as large as lhe gcnotypic varian.ce c-0mponent 
(e .g. > 80-1 OOo/o ), the nex t analytical step can be 
considered. 
The next step involves the estunation of different 
genotypic and genotype-envi ronmental variance 
components through an appropriate model of cornbined 
A NOVA (see Scction 4. 1 ). An analysis of adaptation may 
already bc justified if thc GL interaction variance 
Material com direitos autorais 
Genot_vpe x environ111en1 i11ter(lctiv11s: cha/lenges an<I opportunitics for 11f(l11/ breetling (lnd cultivar reco111111enfiations 
con1ponent is signi ficant and only moderatcly large (e.g. 
> 30-35o/o rclative to the genotypic variance) (Fig. 2.3), 
also because the variancc of sorne of its co1nponcn1s as 
defined in the analysis of adaptation ( c.g. thc heterogeneity 
of genotypc regressions con1poncnt) may prove larger (see 
case study in Chapter 8). Following an appropriate 
n1odclli11g of the GL cff ects or the site classi fícation based 
on similarity for these efTects (see Chapter 5), subregions 
1nay be provisionally identificd lending thcrnselvcs to a 
practica.J definition based on geography, environn1ental 
factors or farming practices. Those that cannot be 
charactcrized as distinct f'ron1 each other n1ay be rnerged 
at this stage. Likc,visc, subregions \vhich are too sn1all to 
bc of practical intcrcst may be merged • .. vith larger ones. 
\\' ide and speci fie adaptation strategies can be 
compared in tcrms of yicld gains pred.ictcd from original 
yicld data of tJle sa1nc data sct (sec Section 6.1 ). ln che 
case of severa! candidate subregions, speci fie adaptation 
1nay contemplate othcr possibilities besides targcting cach 
subregion (c.g. n1crging of some subregions. neglccting 
subregions of n1inor ilnportance). ln n1ost cases, specific 
breeding "''Ould not contemplate n1ore lhan t.\VO or three 
subregions. A final dccision on the adaptation strategy

Outros materiais