Euphytica 92 : 147-153, 1996 .

147© 1996 Kluwer Academic Publishers. Printed in the Netherlands .

CIMMYT's approach to breed for drought tolerance

Sanjaya Rajaram', Hans-Joachim Braun 2 & Maarten van Ginkel'1 International Maize and Wheat Improvement Center (CIMMYT), Lisboa 27, Mexico D .F. 06600, Mexico ;2 CIMMYT, P.K. 39 Emek, 06511, Ankara, Turkey

Key words : breeding methodology, drought patterns, input-efficiency, wheat, Triticum aestivum

Summary

About 32% of the 99 million ha wheat grown in developing countries experiences varying levels of drought stress .Three major drought types have been identified: Late drought (LD) is common in the Mediterranean region, earlydrought (ED) is found in Latin America and wheat is produced on residual soil moisture (RM) in the Indiansubcontinent and part of Australia . Until 1983, CIMMYT selected all germplasm under near optimum conditionsfor its yield potential and tested only advanced lines under drought . In spite of many critics, this approach proved tobe successful, since in the mid 80's CIMMYT germplasm was grown on 45% of the wheat area in LC with annualrainfall from 300-500 mm and on 21 % in areas with less than 300 mm . Since 1983, CIMMYT's drought breedingmethodology is to alternate segregating populations between drought stressed and fully irrigated conditions (FI)and to test advanced lines under a line source irrigation system . To compare the efficiency of these approach, yieldof four, mostly leading varieties, from each of the regions with LD, ED, RM, and FI and twelve recent CIMMYTcultivars selected for high yield under FI and RM conditions (ALT) were compared under four different moistureregimes (FI, LD, ED, and RM) in 89-90 and 90-91 in Yaqui Valley, Mexico. Genotypic correlation between yieldand days to flowering, days to maturity, height, grains m -2 , TKW, test weight and grain fill period were calculated .

Mean grain yield of the four best lines in the ALT group was highest under all moisture stress regimes, followedby the Fl-group. However, the highest yielding cultivar within each moisture regime was from the Fl-group underFI, from the LD-group under LD, and from the ALT group under ED and RM conditions . Estimates for geneticadvance suggest that FI is the best environment for increasing grain yield even in all three drought environments .This indicates that yield potential per se is beneficial also in drought environments . The highest yield in droughtenvironments was realized by the CIM cultivars selected under FI and RM . Simultaneous evaluation of thegermplasm under near optimum conditions, to utilize high heritabilities and identify lines with high yield potential,and under stress conditions to preserve alleles for drought tolerance seem at present the best strategy .

Introduction

There has been a large transformation in the produc-tivity of wheat due to the application of Green Revo-lution technology. This has resulted in a doubling andtripling of wheat production in many environments,but especially in irrigated areas . The high yieldingvarieties of semi-dwarf statured wheats have continu-ously replaced the older tall types at a rate of 2 millionhectares per year since 1977. The overall value of thisnew germplasm has been estimated at an additional 3billion dollars in 1990, which have most benefited the

agricultural community in developing countries (Byer-lee & Moya, 1993 ; Table 1 ; Figure 1) .

There is a growing recognition that the dissemina-tion, application and adoption of this technology has,however, been slower in marginal environments, espe-cially in the semiarid environments affected by poordistribution of water and drought. The annual gainin genetic yield potential in drought environments isonly about half that obtained in irrigated, optimumconditions . Many investigators have attempted to pro-duce wheat varieties adapted to these semi-arid envi-ronments, however, with limited success . Others have

1 48

Percentage of all varieties

Table 1 . Estimated effects of spring wheat breeding research on production by region in the post-GreenRevolution period, 1977-90 (Byerlee & Moya, 1993)

in ∎

a Excludes winter/facultative wheats .6 Varieties released since 1972 are weighted as follows : CIMMYT cross, 0.85, NARSs cross with CIMMYTparent, 0 .50.

Figure 1 . Trends in the origin of spring bread wheat varieties indeveloping countries, 1966-90 .

criticized the Green Revolution technology for failingto adequately address productivity constraints in semi-arid environments, although their own recommendedtechnology has had limited impact, in particular infarmer's fields (Ceccarelli et al ., 1987). This criticismis in clear contrast to the actual acceptance of semidwarf wheat cultivars in rainfed areas, since most ofthe 16 million ha increase in the area sown to Mexicansemi dwarf wheats in the mid 80's occurred in rainfedareas and in 1990, more than 60% of the dryland areain developing countries are planted with semi dwarfs(Byerlee & Moya, 1993) .

In this paper, we wish to give a presentation ofwhy CIMMYT wheat germplasm has had consider-able adaptive success in semi-arid environments . Wealso wish to draw conclusions regarding an effec-tive methodology for a breeding program addressingdrought prone areas. While doing so, we do not intendto belittle any other methodology or approach fol-lowed elsewhere, but do wish to forward the adoptionby farmers as the decisive criteria of success for anymethodology.

Definition of semiarid environments anddescription of distinct drought patterns

In Table 2, the major global drought patterns observedin wheat production is presented (Rajaram et al ., 1993 ;Edmeades et al ., 1989). Through respectively deal-ing with spring (ME4A), facultative (ME9), and win-ter wheat (ME 12), these three mega-environments arecharacterized by sufficient rainfall prior to anthesis,followed by drought during the grain-filling period .The South America, the Southern Cone type of drought(ME4B) is characterized by moisture stress early inthe crop season, with rainfall occurring during thepost anthesis phase . In the Indian Subcontinent type ofdrought stress (ME4C), the wheat crop utilizes waterreserves left from the monsoon rains during the previ-ous summer season . In the Subcontinent also irrigatedwheat crop (ME 1) may suffer drought due to a reducedor less than optimum number of irrigations .

Sub-SaharanAfrica

West Asia/North Africaa

SouthAsia

LatinAmerica

All

Total production increase in1990 (million t) 0.15 2 .45 9.34 3 .40 15.34Average wheat price(US$ 1990/t) 210 210 195 195 198Total value of production increasein 1990 (US$ 1990 millions)' 31 515 1822 662 3030Percent germplasm ofCIMMYT originb 39 52 44 60 49Value of production increaseattributed to CIMMYT(US$ 1990 millions)b 12 268 802 397 1485

Table 2 . Megaenvironments (ME) involving drought conditions (< 500 mm rainfall)

Traditional methodology of breeding for droughtstress

The traditional methodology, which has been practicedfor many years in varying forms, is typified by handlingof all segregating populations under target conditionsof drought, and recommends the use of local landracesin the breeding process (Ceccarelli et al ., 1987) . Whatis not particularly evidenced by this methodology isany impact on yield, farmers' adoption or final nation-al production . This traditional methodology is basedon the assumption that the agro-ecological situationfacing the farmer does not vary in its expression overtime. It assumes that responsiveness of varieties toimproved growing conditions will not be needed . Alsoit presumes that there always occurs a crossover belowa certain yield level under dry conditions, where mod-em high yielding varieties of a responsive nature wouldalways yield less than traditional land race based geno-types. Such crossovers may occur for selected geno-types and one should always be open to the possibilitythat there are real `drought tolerance' traits operatingat the 1 t/ha and below yield level, that adversely affecthigh yield potential at the 4 t/ha and higher yield levels .So far at CIMMYT such traits were not identified . Inany case, cross over would be restricted to such harshconditions, where in fact farmers choose - rightfullyso - not to grow wheat at all, but rather other knownmore drought tolerant crops such as barley or sorghum,or resort to grazing practices .

Source : Rajaram et al ., 1995. In: Abstracts 8th IWGS, Beijing, China, p. 262 .

Yield (t/hal1

6

4

2

1 3

5

7Environments (t/ha)

Figure 2 . Yield of Veery'S' in the 73 environments of the 15thISWYN (Pfeiffer, 1984).

Alternative methodology of combining yieldresponsiveness and adaptation to drought

At CIMMYT we advocate an `open-ended system' ofbreeding in which yield responsiveness is combinedwith adaptation to drought conditions . Most semi-aridenvironments differ significantly across years in their

9

149

ISWYN 150

fJVeery'S'

VIA•Average

I

yield

Fpps ~ •

r

ME Main rainfall Temperate Wheat Sown Exampleregime types locations

ME4A Winter Temperate Spring Autumn Moroccodominant Settat

ME 4B Summer Temperate Spring Autumn Argentinadominant M. Juarez

ME4C Monsoon Hot Spring Autumn Indiaresidual Dharward

ME6 Summer Temperate Spring Spring Russiadominant Saratov

ME9 Winter Moderate cold Facult . Autumn Algeriadominant Sidi Bel Abbes

ME 1 2 Winter Severe cold Winter Autumn Turkeydominant Ankara

150

Table 3. The effects of the 1BL/IRS translocation on yield char-acteristics of 28 random F2-derived F6 lines (14 IBL/IRS and 141 B) from the cross of Triticum aestivum L. cvs Nacosari 76/Seri 82,under reduced irrigated conditions during the 1991-92 and 1992-93crop cycles . Cd. Obregon, Sonora, Mexico

Source : Villareal et al ., 1994; NS : Not significant,* : Significant at 0.05 level .

water availability and distribution pattern . Hence it isprudent to construct a genetic system in which plantresponsiveness provides a bonus wherever environ-mental situations improve due to higher rainfall . Withsuch a system, improved moisture conditions immedi-ately translate into greater gain to the farmer. Why dowe believe this can be done?

The tale of the VEERY's

In the early 1980's when the advanced lines derivedfrom the spring x winter cross Kavkaz/Buho//Kal/BB(CM33027) was tested in 73 global environments of the15th International Wheat Yield Nursery (15th ISWYN)(Figure 2), their performance was quite untypical com-pared to any previously known high yielding vari-eties. In later tests, we found that these lines, calledVEERY's, carry the 1B/1R translocation from rye,and that general performance of such germplasm wassuperior not only in high yielding environments butparticularly under drought conditions (Villareal et al .,1995 ; Table 3) . From the Veery cross 43 varieties werereleased, excluding those released in Europe .

However, in addition to the creation of a new classof superior germplasm, there is an important lessonin breeding to be learned here . The VEERY's repre-sent a genetic system in which high yield performancein favorable environments and adaptation to droughtcould be combined in one genotype. The two genet-ic systems are apparently not always incompatible,although others have claimed that their combinationwould not be possible . Based on this revelation, it ispossible to hypothesize a plant system in which effi-cient input use and responsiveness to improved levels

of external inputs (in this case available water) canbe combined to produce germplasm for marginal (inthis case semi-arid) environments, that at least main-tain minimum traditional yields and express dramaticincreases whenever the environment improves .

Evidence supporting promotion of thismethodology

1 . Contrary to the condemnation by critics (Ceccarel-li et al ., 1987), by the mid 1980's CIMMYT bredgermplasm, occupied 45% of the semiarid wheatareas with rainfall between 300-500 mm, and 21 %of the area less than 300 mm (Morris & Belaid,1991), including large tracks in West Asia/NorthAfrica (WANA) . By 1990 63% of the dryland areas,in especially ME4A and ME4B, was planted withsemi-dwarf wheats (Byerlee & Moya, 1993), manycarrying the 1B/1R translocation . This representsclear acceptance by farmers, who widely adopt-ed the new responsive germplasm over their tradi-tional varieties . The positive trend among the finalusers of our products can not be ignored . Indirect-ly, it supports our view that the modern genotypeshave adaptation to ME4A and ME4B drought areaswhile expressing high yields in improved condi-tions.

2. To support above assumptions, an experiment wasconducted (Calhoun et al ., 1994; Tables 4, 5, 6)to determine how the most modern and widely(spatially) adapted germplasm compared to com-mercial germplasm from countries representing theMediterranean region (ME4A), the Southern Coneof South America (ME4B) and the Indian Subcon-tinent (ME4C), under conditions artificially simu-lating those three Mega-environments. The mostwidely (spatially) adapted CIMMYT lines out-yielded the commercial varieties in all artificiallysimulated environments. The recent adoption trendof CIMMYT germplasm in these difficult marginalenvironments supports the model of input efficien-cy/input responsiveness .

3 . The story of Nesser. Nesser is an advanced linewith superior performance in drought conditionsbred at CIMMYT/Mexico and identified at ICAR-DA/Syria. The cross combines a high yieldingCIMMYT variety Jupateco and a drought tolerantAustralian variety W3918A. The performance ofNesser in WANA's ME4A environments has beenwidely publicized (ICARDA, 1993), and the line

Plant characteristics 1BL/IRS 113 Mean dif.

Grain yield (kg/ha) 4945 4743 202*Above-ground biomassat maturity (t/ha) 12600 12100 500*Grains/m2 14074 13922 152NSGrains/spike 43.5 40.6 2.9*1000-grain weight (g) 37.05 36.53 0.52*

Table 4 . Wheat genotypes representing adaptation to different moisture environments

ME N

Irrigation

Super Kauz, Pavon 76, Genaro 81, Opata 85ME4A (Mediterranean)

Almansor, Nesser, Sitta, Siete CerrosME4B (Southern Cone) Cruz Alta, Prointa Don Alberto, LAP1376, PSN/BOW CM69560ME4

(Subcontinent)

C306, Sonalika, Punjab 81, Barani

Source : Calhoun et al ., 1994 .

Table 5 . Grain yields of selected wheat genotypes grouped by adaptation (Table 4) andtested under 4 moisture regimes in the Yaqui Valley, Mexico, 1989-90 and 1990-91

Adaptation Group

Full

Late

Early

Residualirrigation'

drought2

drought3

moisture4

ME, Irrigation 6636 a 4198 a 4576 a 3032 aME4C Mediterranean 6342 b 3990 ab 4390 b 2883 bME4B Southern Cone 5028 c 3148 be 4224 b 2359 cME4C

Subcontinent

4778 c

3245 be

3657 c

2704 b

Source : Calhoun et al., 1994.1 , received 5 irrigations ; 2 , received 2 irrigations early before heading ; 3 , receivedone irrigation for germination and two in post heading ; 4 , received one irrigation forgermination only .: Means in the same column followed by the same letter are not significantly differentat 0.05 probability level .

is considered by ICARDA to represent a uniquelydrought tolerant genotype . However, it was select-ed at CIMMYT/Mexico under favorable environ-ments, and carries a combination of input efficiencyand high yield responsiveness . It performs similar-ly to the VEERY lines in the absence of rust (Figure2) .

Based on the above evidence, our proposed operationalmethodology is to actively combine input efficiencyand input responsiveness .

Application

A breeding scheme is described below we use toachieve the combination of the two genetic systems .Two contrasting selection environments are alternated,allowing alternate selection for input efficiency andinput responsiveness .Fl . Crosses involving spatially widely adapted

germplasm representing yield stability and yield,with lines with proven drought tolerance inthe specific setting of either ME4A, ME4B orME4C. Winter wheats and synthetic germplasmare emphasized .

F2. The individual plants are raised under irrigatedand optimally fertilized conditions, and inoculated

1 5 1

with a wide spectrum of rust virulence . Only robustand (horizontally) resistant plants are. These mayrepresent adaptation to favorable environments .

F3, F4. The selected F2 plants are evaluated in modi-fied pedigree/bulk breeding system (Rajaram & vanGinkel, 1995) under rainfed conditions or very lowwater availability. The selection is based on indi-vidual lines rather than on individual plants . Theprogenies are selected based on such criteria asspike density, biomass/vigor, grains/m 2, and oth-ers (Van Ginkel et al ., 1995 ; Table 5) . This indexhelps identify lines which may adapt to low watersituations .

F5, F6. The selected lines from F4 are further evaluat-ed under optimum conditions .

F7, F8. Simultaneous evaluations under optimum andlow water environments . Selection of those linesshowing outstanding performance under both con-ditions . Further evaluation in international environ-ments is carried out for purposes of verification .

Conclusions

The proposed breeding methodology is supported inresearch published in recent years by others, not onlyon wheat (Bramel-Cox et al ., 1991 ; Cooper et al.,

152

Table 6 . Genotypic correlation (rg) between agronom-ic traits and final grain yield, for optimum environment(full irrigations) and reduced water regime (late drought,Mediterranean type) in wheat

Trait

Moisture regimeFull

Lateirrigation drought

Days to heading 0.40 0.19Days to maturity 0.29 0.27Grain fill period -0.32 0.36Height -0.39 0.05Peduncle length -0.46 0.22Relative peduncle extrusion - 0 .51 * 0.25Spike length -0.28 - 0 .50*Spike M -2 -0.12 0.64**Grains/spike

0.62*

-0.42Grains M-2

0.74**

0.68**Yield/spike 0 .55* - 0 .64**1000 grain weight 0 .08 -0.45Test weight 0 .13 0.05Harvest index (HI) 0 .83** -0.39Biomass 0 .90** 0.94**Straw yield 0 .52* 0.86**Yield/day (planting) 0.99** 0.57*Yield/day (heading) 0 .94** 0.44Biomass/day (planting) 0.86** 0.69**Biomass/day (heading) 0.74** 0.63**Vegetative growth rate 0.32 0.63**Spike growth rate 0.62** - 0 .58*Grain growth rate

0.17

-0.44

* ** indicate significance at the 0 .05 and 0 .01 probabilitylevel, respectively.Source : van Ginkel, M. CIMMYT, 1995 .

1994; Duvick, 1990,1992 ; Edhaie et al ., 1988; Uddinet al ., 1992 ; Zavala-Garcia et al., 1992), where theimportance of testing and selecting in a range of envi-ronments including well-irrigated ones, has shown toidentify superior genotypes for stressed conditions .The methodology aims at combining input efficien-cy with input responsiveness, by alternating selec-tion environments during the breeding process . Thisapproach results in germplasm that is accepted byfarmers because it translates improved environmentalconditions into yield grains. The traditional method-ology of only selecting under drought conditions, andnarrowly relying on the landrace genotypes, does notmove yield levels significantly beyond those tradition-ally obtained, and does not provide the farmer with abonus yield in the `fat years' .

References

Bramel-Cox, P.J., T. Barker, F. Zavala-Garcia & J .D . Eastin, 1991 .Selection and testing environments for improved performanceunder reduced-input conditions . p. 29-56. In : Plant Breedingand Sustainable Agriculture : Considerations for objectives andmethods. CSSA Special Publication No. 18 . CSSA and ASA,Madison, USA .

Byerlee, D. & P. Moya, 1993 . Impacts of international wheat breed-ing research in the developing world, 1966-1990. CIMMYT,Mexico, D.F. 135 pp.

Calhoun, D .S ., G. Gebeyehu, A . Miranda, S . Rajaram & M. vanGinkel, 1994. Choosing evaluation environments to increasewheat grain yield under drought conditions . Crop Sci 34: 673-678 .

Ceccarelli, S ., M. M . Nachit, G.O. Ferrara, M.S . Mekni, M . Tahir, J.Van Leur & J.P. Srivastava, 1987. Breeding strategies for improv-ing cereal yield and stability under drought . p. 101-114. In: J.P.Srivastava, E . Porceddu, E . Acevedo & S . Varma (Eds) . DroughtTolerance in Winter Cereals . John Wiley & Sons, New York . 387PP.

Cooper, M ., D.E. Byth & D .R. Woodruff, 1994. An investigationof the grain yield adaptation of CIMMYT wheat lines to waterstress environments in Queensland . II . Classification analysis.Agric Res 45 : 985-1002 .

Duvick, D.N ., 1990. Ideotype evolution of hybrid maize in the USA,1930-1990 . p. 557-570 . In : ATTI Proceedings, Vol . II, II Nation-al Maize Conference : research, economy, environment. Grado(GO) - Italy September 19-20-21, 1990 . centro regionale perla sperimentazione agraria, pozzuolo del friuli edagrocole s .p.a.,Bologna, Italy.

Duvick, D .N. 1992 . Genetic contributions to advances in yield ofU.S. maize . Maydica 37 : 69-79 .

Ethaie, B ., J.G. Waines & A .E . Hall, 1988 . Differential respons-es of landrace and improved spring wheat genotypes to stressenvironments . Crop Sci 28 : 838-842.

Edmeades, G.O ., J . Bolanos, H .R . Lafitte, S . Rajaram, W. Pfeiffer& R.A. Fischer, 1989 . Traditional approaches to breeding fordrought resistance in cereals. p. 27-52. In: F.W.G. Baker (Ed) .Drought Resistance in Cereals . ICSU, Paris and C.A.B . Interna-tional, Wallingford, England .

ICARDA, 1993. Cereal Program . Annual Report for 1992 . ICARDA.Morris, M .L ., A. Belaid & D . Byrelee, 1991 . Wheat and barley pro-

duction in rainfed marginal environments of developing world .Part I of 1990-91 CIMMYT world wheat facts and trends . CIM-MYT, Mexico, D.F. 51 pp .

Rajaram, S ., M. van Ginkel & R.A. Fischer, 1994. CIMMYT'swheat breeding mega-environments (ME) . In : Proceedings of the8th International Wheat Genetic Symposium, July 19-24, 1993 .Beijing, China (in press) .

Rajaram, S . & M . van Ginkel, 1995 . Wheat breeding methodology:international perspectives. Proceedings 20th Hard Red WinterWheat Workers Workshop. Oklahoma City, USA (in press) .

Uddin, N ., B.F. Carver & A.C . Clutter, 1992 . Genetic analysis andselection for wheat yield in drought-stressed and irrigated envi-ronments . Euphytica 62 : 89-96 .

Van Ginkel, M., D .S . Calhoun, G. Gebeyehou, A. Miranda, C. Tian-you, R. Pargas Lara, R.M. Trethowan, K. Sayre, J . Crossa & S .Rajaram, 1995 . Plant traits related to yield of wheat in early, late,or continuous drought conditions (in press) .

153

Villareal, R.L., E. Del Toro, A. Mujeeb-Kazi & S . Rajaram. The Zavala-Garcia, F., P.J . Bramel-Cox, J.D. Eastin, M.D. Witt & D .J .1BL/IRS chromosome translocation effect on yield characteriza- Andrews, 1992. Increasing the efficiency of crop selection forlion in a Triticum aestivum L. cross. Plant Breeding (in press) .

unpredictable environments. Crop Sci 32: 51-57 .