Genotype by environment studies demonstrate the criticalrole of phenology in adaptation of chickpea (Cicer arietinum L.)to high and low yielding environments of IndiaJ.D. Berger a,*, M. Ali b, P.S. Basu b, B.D. Chaudhary c, S.K. Chaturvedi b, P.S. Deshmukh d,P.S. Dharmaraj e, S.K. Dwivedi f, G.C. Gangadhar e, P.M. Gaur f,g,J. Kumar d, R.K. Pannu c, K.H.M. Siddique a, D.N. Singh b, D.P. Singh c,f,S.J. Singh h, N.C. Turner a,i, H.S. Yadava j, S.S. Yadav daCentre for Legumes in Mediterranean Agriculture, Faculty of Natural and Agricultural Sciences,The University of Western Australia, 35 Stirling Highway, Crawley, WA 6009, Australiab Indian Institute of Pulses Research, Kanpur, U.P. 208024, IndiacCCS Haryana Agricultural University, Hisar, Haryana 125004, Indiad Indian Agricultural Research Institute, New Delhi 110012, Indiae Institute for Pulse and Oilseeds Research, Gulbarga, Karnataka 585101, Indiaf Jawaharlal Nehru Krishi Vishwa Vidyalaya, Jabalpur, M.P. 482004, Indiag International Crops Research Institute for the Semi-Arid Tropics (ICRISAT), Patancheru, A.P. 502324, IndiahDurgapura ARS, Jaipur, Rajasthan 302018, IndiaiCSIRO Plant Industry, Private Bag 5, Wembley, WA 6913, Australiaj Jawaharlal Nehru Krishi Vishwa Vidyalaya, Sehore, M.P. 466001, IndiaReceived 3 February 2006; accepted 20 February 2006AbstractIn order to investigate specific and general adaptation of chickpea in India, a wide range of sub-continental, Australian and Mediterraneangenotypes were grown across seven sites characterizing the major chickpea growing areas over 3 years, and extensive data on plant stand,early vigour, phenology, productivity and yield components collected. High and low yielding sites were clearly separated by a range ofphysical and biological characters, low yield being associated with low latitude and pre-season rainfall, high temperature, early phenology,short crop duration, low biomass and fecundity. Genotype by environment interactions for yield were highly significant (P < 0.001), andaccounted for more variance than that attributed to genotypes alone. Ward’s hierarchical clustering indicated that the genotypes could beseparated into discrete groups, comprising material specifically adapted to the north (Clusters 2 and 3) or south (Cluster 5), widely orconsistently poorly adapted germplasm (Clusters 1 and 4, respectively).Cluster 5, comprising germplasm from southern and central India, was characterized by early phenology, confirming the role of droughtescape in southern India. With increasing latitude Cluster 5 genotypes remained early, but had the capacity to delay maturity considerably,resulting in average, and occasionally above average yields. However, compared to well-adapted material in the north, Cluster 5 biomass waslow, and the time interval between flowering and podding up to 50 days, representing repeated cycles of flowering and subsequent abortion.Clusters 2 and 3, dominated by northern Indian genotypes, were characterized by later phenology, and were able to delay the onset offlowering significantly more than the remaining germplasm at late flowering northern sites. In Cluster 3, the second highest yielding groupoverall, this increased both source and sink potential at productive northern sites. Cluster 2 was uniformly later than Cluster 3, and loweryielding at most sites. Cluster 1 was characterized by intermediate flowering and relatively early, responsive maturity, a phenologicalcompromise responsible for wide adaptation, by providing sufficient drought escape in the south, and enough biomass in the north to produceabove average yields in these contrasting environments. ICCV 10 from the International Crops Research Institute for the Semi-Arid Tropics(ICRISAT), and 2 Indian Agricultural Research Institute (IARI) lines, BG 391 and BG 1006, were the most consistently high yielding, rankingwww.elsevier.com/locate/fcrField Crops Research 98 (2006) 230–244* Corresponding author. Tel.: +61 8 9333 6623; fax: +61 8 9387 8991.E-mail address: Jens.Berger@csiro.au (J.D. Berger).0378-4290/$ – see front matter. Crown Copyright # 2006 Published by Elsevier B.V. All rights reserved.doi:10.1016/j.fcr.2006.02.007J.D. Berger et al. / Field Crops Research 98 (2006) 230–244 231in the top 10 at 10 and 8 sites, respectively. Cluster 4, comprising largely Australian cultivars, was characterized by late, unresponsivephenology and the lowest yield at each site.Crown Copyright # 2006 Published by Elsevier B.V. All rights reserved.Keywords: Chickpea; Adaptation; G E; Phenology; Yield; India; Pattern analysis1. IntroductionChickpea (Cicer arietinum L.) ranks second among theworld’s food legumes in terms of area, and is a particularlyimportant crop in South Asia, with large areas in India (5.8–6.1 106 ha) and Pakistan (0.9–1.7 106 ha), responsiblefor 71–73% of global production (FAO, 2004b). The cropwas domesticated in West Asia some 10,000 years ago, andfirst appeared in India between 5000 and 7000 years ago(Zohary and Hopf, 2000). Underlying the long crop historyand extensive cultivation is a considerable phenotypicdiversity among South Asian germplasm, particularly interms of phenology, plant architecture, fecundity and seedFig. 1. Indian trial sites used for the investigation of G E interaction in chickpeaAndhra Pradesh; B, Bihar; G, Gujarat; H, Haryana; K, Karnataka; M, Maharashtra;and WB,West Bengal.colour (Upadhyaya, 2003). The combination of long crophistory, diverse germplasm and significance in terms ofglobal production make India a compelling country in whichto investigate the adaptation of chickpea.Chickpea is grown over a wide range of environmentswithin India, from Karnataka in the south (14.5–18.48N) toPunjab in the north (29.5–31.68N) (Table 1, Fig. 1) (Ali andKumar, 2003). In the south, crop duration is short, typicallyaround 100 days at Hyderabad (Saxena, 1984), and thegrowing season finishes in late January or early February.Minimum and maximum temperatures vary between 15 and30 8C, and there is little change after flowering (Table 1).Crop duration in the north is far longer, between 150 and 160yield. States in which chickpea is grown are identified by abbreviation: AP,MP, Madhya Pradesh; O, Orissa; P, Punjab; R, Rajasthan; UP, Uttar PradeshJ.D.Bergeretal./FieldCropsResearch98(2006)230–244232Table 1Production (Ali and Kumar, 2003) and climatic conditions of the chickpea growing regions of India based on long-term monthly climate averages compiled from weather stations within each state (FAO, 2004a)State/trial site Area (% totalproduction)Latitude(decimal)Longitude (decimal) Season sowing-flowering-maturity Mean temperature (8C) Rainfall (mm)Pre-anthesis(min–max)Post-anthesis(min–max)Pre-season SeasonKarnataka 3.6 14.5–18.4 74.4–77.5 October–December–January 19.3–30.2 16.0–29.3 742 126Site: Gulbarga 1998 17.4 76.9 26th October–15th December–28th January 13.3–31.3 13.5–30.8 461 68Site: Gulbarga 2000 17.4 76.9 18th October–7th December–22nd January 18.7–32.8 20.7–31.2 205 175Andhra Pradesh 1.9 15.5–20.0 76.9–81.1 October–December–January 19.3–30.1 16.3–29.0 821 166Maharastra 11.4 15.9–21.7 72.9–80.0 October–December–February 16.2–30.5 14.0–29.9 809 78Orissa 0.3 18.0–22.3 81.5–86.0 October–December–February 16.6–28.3 15.5–28.1 1515 75Gujarat 0.8 20.7–24.6 69.5–74.9 October–January–February 14.9–30.5 13.0–28.4 677 14Madhya Pradesh 49.1 18.7–25.3 74.2–84.5 October–January–February 11.8–27.1 11.7–26.9 1308 56Site: Jabalpur 1998 23.2 80.0 26th October–5th January–12th March 11.0–27.1 11.2–27.4 969 251Site: Jabalpur 1999 23.2 80.0 5th November–11th January–29th February 9.5–26.7 9.9–26.5 1537 129Site: Jabalpur 2000 23.2 80.0 24th October–29th December–27th February 11.9–29.8 9.3–26.1 1163 89Site: Sehore 1999 23.2 77.1 27th October–6th January–26th February 9.6–29.9 8.3–27.9 1258 133Site: Sehore 2000 23.2 77.1 1st November–8th January–22nd February 9.3–29.9 8.6–27.8 719 65West Bengal 0.6 22.7–24.7 86.0–88.0 October–January–February 14.9–27.5 13.2–27.1 1419 62Bihar 2.2 22.0–26.7 83.4–88.0 October–January–March 12.1–26.2 11.3–25.7 1210 54Rajasthan 13.4 23.1–30.0 72.2–78.0 October–January–March 10.3–26.3 11.3–26.7 592 25Site: Durgapura 1999 26.8 75.8 29th October–16th January–8th March 17.0–31.6 8.8–23.6 276 179Site: Durgapura 2000 26.8 75.8 26th October–14th January–N/A 12.3–28.2 10.7–26.5 428 69Uttar Pradesh 15.4 25.3–30.4 77.1–84.2 October–January–March 10.1–25.7 11.9–27.3 949 59Site: Kanpur 1998 26.8 80.4 29th October–23rd January–26th March 10.2–24.3 12.7–27.6 1341 79Site: Kanpur 1999 26.8 80.4 27th October–1st January–10th March 10.5–28.4 7.9–23.5 897 213Site: Kanpur 2000 26.8 80.4 2nd November–31st December–8th March 9.4–27.4 8.1–23.8 811 60Haryana 1.1 27.6–30.7 74.6–77.6 October/November–January–March 7.1–25.1 10.7–27.9 383 64Site: Hisar 1999 29.0 75.7 3rd November–3rd February–9th April 6.3–24.5 8.8–26.9 306 129Site: New Delhi 1999 28.6 77.2 18th November–6th February–1st April 8.0–21.2 10.7–25.2 508 136Site: New Delhi 2000 28.6 77.2 7th November–4th February–3rd April 7.5–23.4 11.8–27.2 761 99Punjab 0.1 29.5–31.6 74.9–76.8 October/November–January/February–March/April 7.2–23.4 10.3–26.0 661 118States are sorted by ascending latitude of chickpea areas. Trial site data are tabled within the appropriate state (except for New Delhi, which is listed under Haryana) and are based on weekly averages recorded ateach research station.J.D. Berger et al. / Field Crops Research 98 (2006) 230–244 233days at Hisar (Saxena, 1984), and the season finishes inMarch or April. Vegetative phase temperatures are 5–10 8Clower in the north, but increase considerably after flowering,with maxima only 2–3 8C lower than in the south (Table 1).Nevertheless pod set in northern regions is often delayeduntil February because of low temperatures at flowering(Saxena, 1984). While chickpea is grown on stored soilmoisture throughout India, there is geographic variation forboth monsoonal and within-season rain. Eastern states suchas Orissa, Madya Pradesh, West Bengal and Bihar receivethe largest monsoon, and southern states such as Karnatakaand Andhra Pradesh the largest within-season rainfall(Table 1). The north-western states of Rajasthan and Gujaratare particularly dry, receiving little monsoonal, and verylittle within-season rain on average (Table 1). Supportingthis broad environmental range is a widespread plantbreeding program based on 50 breeders at 22 locations (M.Ali personal communication).The study of chickpea adaptation to the Indianenvironment is dominated by detailed research on theexpression of traits such as yield, harvest index, phenology,pod set, nitrogen uptake, leaf area index, relative growthrate, dry matter production and partitioning over time in asmall number of genotypes (usually 2) in southern andnorthern locations (usually Hyderabad and Hisar or Delhi)(see citations in Saxena (1984) and Khanna-Chopra andSinha (1987)). Typically these studies are descriptive, usingspecifically adapted genotypes, usually JG 62 for the southand G 130 for the north, often without reciprocation (i.e.where JG 62 was not tested in the north and vice versa for G130). This work suggests that in northern India long cropdurations coupled with higher growth rates and longerperiods of N uptake provide a larger photosynthetic area toact as a C source, as well as a higher sink potential due to thegreater numbers of flowering nodes, and therefore yields arerelatively high, despite a lower harvest index (Sinha et al.,1983; Saxena, 1984; Khanna-Chopra and Sinha, 1987).Conversely in the south, peak crop growth rates, leaf areaindices, and N uptake occur much earlier, and decline morerapidly during pod filling (Saxena and Sheldrake, 1980;Saxena, 1984). Consequently, adapted plants in the south(i.e. typically Hyderabad) escape drought through earliness(Saxena and Sheldrake, 1980; Saxena, 1984).There is a disconnect between this detailed physiologicalwork and the larger scale field studies based on growinglarge numbers of genotypes over a wide range ofenvironments. As a result it is difficult to extrapolate thedescriptive physiology carried out in Hyderabad and Hisar inJG 62 and G 130 across wider environments and germplasmassemblages. The literature describing traits associated withyield using larger numbers of genotypes is generallyunhelpful. There is a plethora of reports across the last 20years positively correlating yield with: (a) fecundity (pod orbranch number per plant) (Haloi and Baldev, 1984; Jiraliet al., 1988; Yadav, 1991; Singh et al., 1997; Qureshi et al.,2004), (b) biomass (Chaudhary et al., 1988; Jirali et al.,1994; Yadav et al., 2003) and (c) harvest index (Khan andMalik, 1989; Rao, 1996; Qureshi et al., 2004). The numberof publications relating yield to phenology is much smaller:Yaqoob et al. (1990) and Qureshi et al. (2004) bothsuggested yield was negatively correlated to maturity, whileBhambota et al. (1994) suggested there was no relationshipacross four test environments. In general these studies wereconducted in a single environment, often only across a singleyear, and the results are not presented in an environmental orclimatic context. As a consequence these studies contributelittle to an improved understanding of chickpea adaptation tosub-continental environments.This studyaddresses this shortcomingbymeasuring awiderange of traits in diverse sub-continental, Australian andMediterranean genotypes grown across sites which char-acterize the major chickpea growing environments of India.The primary objective was to identify specific or wideadaptation based on yield, and determine which traits wereassociated with this. The secondary objectivewas to examineIndian breeding programmes from the point of view ofspecific or general adaptation based on the performance of theIndian germplasm in the trial. In contrast to countries such asAustralia, where plant breeding is becoming increasinglycentralized (Berger et al., 2004), Indian breeding programmesare located throughout the chickpea growing zone. If specificadaptation is regional, are regionally developed genotypesalways better in their target environments? Are particularlocations and breeding strategies better for developingspecific or general adaptation?2. Materials and methods2.1. Germplasm and experimental sitesThe study was based on an extensive genotype byenvironment trial conducted over 3 years at seven sitescovering the major Indian chickpea growing areas (Fig. 1).Forty-six genotypes were evaluated, comprising 41 ofIndian, 3 of Australian and 2 of Mediterranean-basin origin.Indian material was chosen on the basis of putative droughtresistance, as opined by Indian chickpea breeders, andoriginated from southern (Andhra Pradesh, Karnataka: naccessions = 7), central (Madhya Pradesh, Uttar Pradesh:n = 5) and northern chickpea growing areas (Delhi, Haryana,northern Uttar Pradesh: n = 29). The Indian germplasm wasa mixture of landraces (n = 8), advanced breeding materialand cultivars, released largely in India (n = 17), but also inAustralia (n = 2), and Bangladesh (n = 2). Australian-bredcultivars were developed in northern NSW (Amethyst,Barwon) and Queensland (Norwin), where chickpea isgrown as a cool-season legume, largely on stored soilmoisture with a high probability of rain near maturity(Berger et al., 2004). Two cultivars released in Australia(Dooen and Gully (T 1315)), but originating fromAzerbaijan and Iran, respectively, were also included.J.D. Berger et al. / Field Crops Research 98 (2006) 230–2442342.2. Trial protocolTrials were spatially optimized randomised block designswith three replicates created using SpaDes (Coombes,2002). Annegeri 1, a southern Indian landrace, and longstanding variety, and ICC 4958, a central Indian line releasedas a variety in Bangladesh, were used as checks because oftheir reputed drought resistance, and replicated six times.Seeds were hand-sown in four-row plots 1.2 m wide and 4 mlong, and a uniform density of 33 plants/m2 targeted at allsites. Seeds were pretreated with Bavistin1 to minimize theprobability of soil borne diseases such as Fusarium wilt (F.oxysporum Schlechtend) and root rot (Rhizoctonia batati-cola Taubenhaus), and inoculated with Group N rhizobiaimmediately prior to sowing.Trials were sown in mid October in southern and centralIndia, and late October to early November in northern India.If residual moisture at planting was considered to beinsufficient to allow even germination, a pre-sowingirrigation of approximately 60 mm was applied. Di-ammonium phosphate (DAP) fertilizer and gypsum wereapplied at 100 and 200 kg/ha, respectively. Pod borer(Helicoverpa armigera Hu¨bner) was controlled from lateflowering onwards by endosulfan application.A wide range of data on plant stand, early vigour,productivity, yield components and phenology was collectedat each site. Early vigour was estimated by harvesting 0.5 m2sub-samples at 600 degree days after sowing (assuming abase temperature of 0), drying in a forced-draught oven andweighed. The number of plants harvested was recorded, sothat early dry matter could be expressed either per plant orper unit area. Yield and biomass were measured similarly atphysiological maturity by harvesting 2.5 m of the twocentral rows (equivalent to 1.5 m2) to avoid edge effects.Harvest index was calculated using these data. Standing cropheight was measured in the field at maturity using fiverandom points per plot. Plant length was determined at thesame time by measuring the longest branch in five randomlyselected plants. Seed and pod weight and numbers per plantwere estimated from bulked 10 plant sub-samples harvestedadjacent to the yield quadrats. Dates of complete emergence,50% flowering and podding, end of flowering andphysiological maturity were recorded and expressed asdays after sowing (DAS). These data were used to calculatethe lengths of the vegetative phase (50% flowering minusemergence), flowering phase (end flowering minus 50%flowering), podding phase (maturity minus 50% podding)and season length (maturity minus emergence).2.3. StatisticsANOVA was performed individually at each site toidentify entry error and outliers using residual plots(Genstat, 2002). Subsequently, genotype by environment(G E) analysis was performed on a balanced, outlier-freesubset of 39 genotypes and 15 sites. (Note that six genotypeswere excluded from the analysis to maintain balance: twonew accessions were introduced in 1999, while fouraccessions highly susceptible to dry root rot (Amethyst,Barwon, JG 62, Gully) did not produce seed in Jabalpur in1999 despite the prophylactic measures employed.) Resi-dual plots indicated that error variance and yield werecorrelated, and therefore the raw data was log-transformedto ensure common variance across sites. After transforma-tion variance was random: there was no relationship betweenresiduals and predicted values (data not presented) indicat-ing that ANOVAwas appropriate for G E analysis. Blockswere taken out within sites, and a hierarchical ANOVAmodel (SS 1) was used when factors, such as variety, werefurther sub-divided into clusters or agro-ecosystems.Orthogonal contrasts were used to test the significance ofthese sub-divisions within and between sites.Restricted maximum likelihood (REML) analysis wasused to calculate variance components by treating alltreatment factors as random effects (Genstat, 2002).Finlay and Wilkinson (1963) analysis was used toquantify genotype responsiveness to favourable conditionsby regressing genotype against site mean yields to generateslope coefficients. Genotype responses were strongly linear,with the correlation coefficient (r) ranging from 0.65 to 0.91,with an average of 0.79. A similar approach was used torelate genotype phenology to site latitude.G E interaction was visualized using multivariateapproaches. Ward’s hierarchical clustering (DeLacy et al.,1996) was used to identify discrete groups of genotypes inthe G E mean log yields matrix using SPSS v.11.5 (SPSS,2002). Principal components analysis (PCA) based on thecovariance matrix was used to construct a biplot ofgenotypes (PC scores) and environments (PC factorloadings, shown as biplot vectors). Covariance/variancematrix-based PCA centres the data by subtracting columnmeans, which is equivalent to removing the main effects ofenvironment in this case (Fox and Rosielle, 1982). Becausethe data is not standardized (as in correlation matrix-basedPCA) genotype yield differences are allowed to play a largerrole in pattern formation (Berger et al., 2004).PCA based on the correlation matrix was used to examinethe relationships among continuous plant traits and physicalsite descriptors between sites, and presented as biplots ofsites (PC scores) and traits/descriptors (PC factor loadings).3. Results3.1. SitesANOVA revealed that the largest treatment differences inyield were between sites: there was a greater than 10-folddifference between Sehore in 2000 (0.25 trial/ha) and Hisarin 1999 (2.59 trial/ha), reflecting the range of environmentssampled. Principal components analysis, based on bothphysical and biological site data, clearly discriminatedJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244 235Fig. 2. Principal components biplot (based on the correlation matrix) of trial site physical and biological data. (Note that biomass was not included in the PCAbecause of missing values at some sites, but was strongly correlated with yield: r = 0.96, P > 0.001.) Arrows represent vectors defined by factor loadings ofvariables in PC1 and PC2, markers represent site scores for PC1 and PC2, respectively. Sites are classified into low (Z score<1), medium (1 < Z score<1)and high yielding (Z score >1), and identified by name: Delhi, New Delhi; Durg, Durgapura; Gul, Gulbarga; His, Hisar; Jab, Jabalpur; Kan, Kanpur and She,Sehore, and year abbreviations: 98, 1998; 99, 1999 and 00, 2000.between low and high yielding sites (Fig. 2). Low yieldingsites (Sehore 2000, Gulbarga 1998 and 2000) largelyoccurred on the lower-left of PC1 (Fig. 2), and characterizedby low latitude, early sowing (18th October–1st Novem-ber), low pre-season rainfall, high temperature (post-anthesis mean maxima: 27.8–31.2 8C), long days pre-anthesis and a slow rate of daylength change (Table 1).Biologically, these sites were characterized by earlyphenology (50% flowering: 51–69 days), short seasons(maturity: 95–113 days), low biomass (1.3–2.2 trial/ha) andfecundity (16–31 pods per plant). Medium and highyielding sites were located on the right of Fig. 2, andcharacterized by the opposite: late sown (24th October–18th November), northern locations with cooler tempera-tures throughout the season (post-anthesis mean maxima:23.5–27.9 8C), shorter days, and more rapid daylengthchanges (Table 1), later phenology (flowering 60–93 days),longer seasons (maturity: 116–159 days), higher biomass(2.8–9.4 trial/ha) and fecundity (31–99 pods per plant).Medium and high yielding sites could also be classified bylongitude (PC2, Fig. 2), being associated with either highmonsoonal pre-season rainfall in the east (Jabalpur,Kanpur), or high pre-anthesis rainfall within the growingseason in the west (Delhi, Durgapura, Hisar).3.2. G E interactionG E interactions for seed yield between the 39chickpea genotypes and 15 trial/year combinations werehighly significant (P < 0.001), and accounted for 12.7% ofvariance according to REML, more than that attributed togenotypes alone (11.0%). In order to reveal the patternunderlying this interaction, the matrix of log-transformedgenotype mean yields at the 15 sites was further analysedusing multivariate methods. Ward’s hierarchical clustering(DeLacy et al., 1996) indicated that the 39 accessions couldbe divided into 5 discrete groups (Fig. 3). HierarchicalANOVA (sums of squares 1 model) demonstrated that theinteraction behaviour of the five clusters was highlysignificant (P < 0.001) and explained 42.5% of the totalinteraction sum of squares (data not presented, confirmed byREML variance components distribution).J.D. Berger et al. / Field Crops Research 98 (2006) 230–244236Fig. 3. Hierarchical cluster analysis of 39 chickpea genotypes based onWard’s method using a genotype by site matrix of log-transformed means. The top 33%of genotypes based on average seed yields over all sites are identified by rank order from 1 to 13. Germplasm habitat of origin is given in code: SSM, spring-sownMediterranean; ASMTA, autumn-sown Mediterranean-type (Australia); ASSC, autumn-sown sub-continental (Indian); N, northern India; C, central India andS, southern India. Released varieties are given in bold with a superscripted initial to indicate the country of release: A, Australia; B, Bangladesh and I, India. Thesuperscript L indicates landrace.J.D. Berger et al. / Field Crops Research 98 (2006) 230–244 237Fig. 4. Principal components analysis (based on the variance/covariance matrix) of 39 chickpea genotypes using a genotype by site matrix of log-transformedseed yield means. Biplot vectors are site factor loadings, points are genotype scores with cluster membership (see Fig. 3) superimposed as different markerpatterns. Site/year abbreviations are as indicated in Fig. 2.An ordination was produced by principal componentsanalysis based on the covariance matrix to allow differencesin scale to play a role in pattern formation. Factor loadingsfor all sites were positive on PC1 (Fig. 4). As a result therewas a strong correlation between genotype PC1 scores andseed yield averaged over all sites (r = 0.98, P < 0.001)indicating that genotype mean yield increases from left toright in Fig. 4. PC1 factor loadings in the G E ordination(Fig. 4) were negatively correlated (r = 0.54, P < 0.05)with PC1 loadings in the sites ordination presented earlier(Fig. 2). Stressful, early, largely low yielding sites such asSehore and Gulbarga, dominate the interaction behaviourmodelled by PC1 (Fig. 4), and performance at these sites is agood indicator of productivity overall. Conversely, PC2loadings in the G E ordination (Fig. 4) were stronglypositively correlated (r = 0.79, P < 0.001) with PC1loadings in the previous site physical and biologicalcharacteristics ordination (Fig. 2). Genotypes with highPC2 scores performed better under the less stressfulconditions experienced at the longer season, later, northernlocations such as Delhi, Kanpur and Hisar.Fig. 4 shows that the five Ward’s clusters were clearlyseparated by different PC1 and PC2 scores. Cluster 1, on thefar right of PC1 (Fig. 4) was the most widely adapted,ranking first at eight sites, and second at six sites, with aboveaverage performance at the stressful southern sites (Sehore2000, Gulbarga 1998 and 2000), and consistently highproductivity in the medium to high yielding central andnorthern Indian sites (Table 2). This was reflected by Finlayand Wilkinson (1963) analysis which demonstrated thatCluster 1 was relatively responsive to favourable conditions(more than Clusters 4 and 5 (P < 0.001)), and generallycharacterized by positive y intercepts (Table 3). Cluster 1includes 7 of the top 13 genotypes averaged across all sites(Fig. 3), and is comprised of material from the north (n = 9,mainly from IARI, New Delhi), the centre (n = 1) and thesouth (n = 3). ICCV 10, a released cultivar in both India andBangladesh (bred by ICRISAT in Andhra Pradesh), was themost consistently productive genotype, ranking in the top 10at 10 sites. BG 391 and BG 1006 (the prefix BG identifiesmaterial from IARI) ranked in the top 10 at 8 sites.Cluster 3, in the upper right quadrant of Fig. 4, was thesecond most productive overall (Table 1), and performedbest in the longer season, higher yielding northern sites,reflecting its position high on PC2. In fact orthogonalcontrasts demonstrated that there were no yield differencesbetween Clusters 1 and 3 at all sites yielding more than1.2 trial/ha. However, Cluster 3 was less productive than 1 atall sites below this threshold with the single exception ofSehore 2000. Consequently, Cluster 3 was the mostresponsive of all (P < 0.05), with all genotype regressionslopes between 1.1 and 1.3, and negative y intercepts(Table 3). Cluster 3 was comprised exclusively of genotypesdeveloped in northern India (Fig. 4), in a variety of breedingprogrammes in the late 1970s and early 1980s (GovindBallabh Pant University of Agriculture and Technology,Pantnagar: PANT G 114; Chandra Shekar Azad Universityof Agriculture and Technology, Kanpur: PDG 84-16, K 850;J.D. Berger et al. / Field Crops Research 98 (2006) 230–244238Table 2Cluster productivity (mean log seed yields) at 15 Indian trial sites used for the investigation of G E interactionSite Latitude Longitude Cluster 1 Cluster 2 Cluster 3 Cluster 4 Cluster 5 Mean LSDSehore 2000 23.2 77.1 0.42 (0.4) 0.70 (0.2) 0.16 (0.7) 1.22 (0.1) 0.49 (0.3) 0.60 (0.3) 0.18Gulbarga 2000 17.4 76.9 0.24 (0.6) 0.53 (0.3) 0.35 (0.4) 0.79 (0.2) 0.05 (0.9) 0.39 (0.4) 0.10Gulbarga 1998 17.4 76.9 0.08 (1.2) 0.24 (0.6) 0.23 (0.6) 0.42 (0.4) 0.06 (1.1) 0.15 (0.7) 0.07Sehore 1999 23.2 77.1 0.22 (1.7) 0.07 (1.2) 0.13 (1.3) 0.38 (0.4) 0.20 (1.6) 0.05 (1.1) 0.10Delhi 1999 28.6 77.2 0.16 (1.5) 0.07 (1.2) 0.11 (1.3) 0.08 (0.8) 0.06 (1.2) 0.07 (1.2) 0.06Durgapura 1999 26.8 75.8 0.29 (2.0) 0.21 (1.6) 0.26 (1.8) 0.11 (0.8) 0.19 (1.5) 0.17 (1.5) 0.06Delhi 2000 28.6 77.2 0.34 (2.2) 0.22 (1.7) 0.30 (2.0) 0.07 (0.8) 0.06 (1.2) 0.17 (1.5) 0.12Durgapura 2000 26.8 75.8 0.24 (1.8) 0.13 (1.4) 0.22 (1.7) 0.06 (1.2) 0.20 (1.6) 0.17 (1.5) 0.09Jabalpur 1998 23.2 80.0 0.30 (2.0) 0.19 (1.5) 0.27 (1.9) 0.03 (1.1) 0.32 (2.1) 0.22 (1.7) 0.05Kanpur 1998 26.8 80.4 0.39 (2.5) 0.28 (1.9) 0.39 (2.4) 0.05 (0.9) 0.20 (1.6) 0.24 (1.7) 0.11Jabalpur 2000 23.2 80.0 0.37 (2.3) 0.24 (1.8) 0.33 (2.1) 0.12 (1.3) 0.39 (2.5) 0.29 (1.9) 0.05Kanpur 1999 26.8 80.4 0.38 (2.4) 0.33 (2.2) 0.38 (2.4) 0.09 (1.2) 0.27 (1.8) 0.29 (1.9) 0.07Jabalpur 1999 23.2 80.0 0.42 (2.6) 0.36 (2.3) 0.46 (2.9) 0.05 (0.9) 0.44 (2.8) 0.33 (2.1) 0.10Kanpur 2000 26.8 80.4 0.45 (2.8) 0.37 (2.4) 0.53 (3.4) 0.17 (1.5) 0.43 (2.7) 0.39 (2.4) 0.14Hisar 1999 29.0 75.7 0.52 (3.3) 0.44 (2.8) 0.51 (3.3) 0.18 (1.5) 0.41 (2.6) 0.41 (2.6) 0.07Mean 0.23 (1.7) 0.10 (1.2) 0.21 (1.6) 0.17 (0.7) 0.18 (1.5) 0.03Sites are sorted bymean yield. Values in parentheses are back transformed seed yields in trial/ha. The mean LSD is calculated using the average standard error ofthe difference across all clusters.Indian Institute of Pulses Research (IIPR), Kanpur: IPC 92-39; IARI: BG 396; Punjab Agricultural University,Ludhiana: Tyson (selected from C 235 in Queensland)).Cluster 2, on the left of Fig. 4, also performed better at thehigher yielding northern sites (Table 2), but was significantlyless productive (P < 0.001 to <0.057) than Cluster 3 in allenvironments, with the exception of the low yieldingsouthern sites, Gulbarga 1998 and 2000. Finlay–Wilkinson yintercepts were negative, and slopes significantly lower thanCluster 3. Like Cluster 3, Cluster 2 was also dominated bynorthern germplasm from a variety of origins, but the Indiancultivars in Cluster 2 were generally older, having beenreleased in the 1960s. Re-analysis of the dataset excludingJabalpur 1999, where four varieties were eliminated by dryroot rot, shows that Cluster 2 would also have included theAustralian variety Amethyst, and the Iranian landrace Gully(data not presented).In contrast to Clusters 2 and 3, Cluster 5, located alongthe negative of PC2 (Fig. 3), performed best in central andsouthern sites, yielding significantly above average(P < 0.001) at all sites equal to, or below 23.28S (Gulbarga,Sehore, Jabalpur, Table 2). Cluster 5 was as productive asCluster 1 at Sehore 2000 and Gulbarga 1998, andsignificantly more so at Gulbarga 2000 (P < 0.001), alllow yielding, southern sites. However, orthogonal contrastsdemonstrate that, with the exception of Durgapura 2000,Cluster 5 was significantly outperformed by Cluster 3(P < 0.001 to <0.089) at all northern sites (Table 2).Accordingly, Cluster 5 was relatively unresponsive (lessthan Clusters 1–3 (P < 0.05)), with 9 out of 10 genotyperegression slopes ranging from 0.7 to 1.0, while y interceptswere positive, and significantly larger than all except Cluster1 (Table 3). In contrast to Clusters 2 and 3, Cluster 5 wasdominated bymaterial of southern (n = 4) and central (n = 3)Indian origin (Fig. 4), including only two accessions fromthe north (Uttar Pradesh). Re-analysis without Jabalpur 1999revealed that the central Indian variety, JG 62, would alsohave been included in Cluster 5.Cluster 4, on the extreme negative of PC1 (Fig. 4) wassignificantly below average at all sites (P < 0.002),particularly in the low yielding central and southernenvironments of Sehore and Gulbarga, where only 24–53% of site mean yield was produced by this cluster(Table 2). Cluster 4 comprises the Australian varietiesNorwin and Dooen (Fig. 3), and also Barwon, if Jabalpur1999 is excluded from the cluster analysis. The yieldresponsiveness of these three varieties was very poor, withregression slopes from 0.5 to 0.8, and y intercepts werenegative (Table 3).3.3. Cluster phenologyThere were consistent phenological differences betweenthe five clusters generated from the yield data. Cluster 5,which was specifically adapted to stressful southern sites,was the earliest to flower, set pods and mature (P < 0.001),whereas the uniformly poorly adapted Cluster 4 had thelatest phenology (Fig. 5). Both Clusters 4 and 5 respondedsimilarly to later-flowering environments, and thus theirregression lines formed parallel boundaries containing theremaining clusters (Fig. 5a). The widely adapted Cluster 1was characterized by intermediate flowering and respon-siveness, located centrally between Clusters 4 and 5. Incontrast, Clusters 2 and 3, specifically adapted to higheryielding northern sites, were able to delay flowering at latersites significantly more than the remaining clusters(P < 0.05) (Fig. 5a). Cluster 2 was later flowering at allsites than Clusters 1 and 3 (Fig. 5a).Regression patterns for podding and maturity weredifferent to those for flowering. Cluster 4 was lessresponsive, while Cluster 5 was more responsive than allremaining groups, and therefore their regression slopesJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244 239Table 3Genotype and cluster log seed yield (averaged over all sites) and yieldresponsiveness as defined by the Finlay and Wilkinson (1963) genotype vs.site mean regressionGenotype MeanyieldFW intercept(trial/ha)FW slopecoefficient(t/t site mean)Cluster 1Annegeri 1 0.21 (1.6) 0.3 0.9BG 1006 0.27 (1.9) 0.3 1.0BG 212 0.20 (1.6) 0.1 1.2BG 256 0.25 (1.8) 0.7 0.8BG 361 0.21 (1.6) 0.1 1.1BG 362 0.24 (1.7) 0.2 1.0BG 364 0.21 (1.6) 0.4 1.4BG 372 0.24 (1.7) 0.4 0.9BG 391 0.27 (1.9) 0.3 1.0ICC 10426 0.22 (1.6) 0.0 1.1ICC 8412 0.21 (1.6) 0.1 1.1ICCV 10 0.30 (2.0) 0.1 1.3IPC 92-1 0.19 (1.5) 0.2 1.2Mean 0.23 (1.7) 0.1 1.1Cluster 2Amethyst 0.05 (1.1) 0.1 0.9BG 261 0.04 (1.1) 0.0 0.8BG 276 0.08 (1.2) 0.3 0.6C 214 0.17 (1.5) 0.1 1.1C 235 0.14 (1.4) 0.4 1.2G 130 0.12 (1.3) 0.6 1.3H 208 0.12 (1.3) 0.2 1.0H 75-35 0.13 (1.4) 0.4 1.3HIMA 0.04 (1.1) 0.2 0.9ICC 14880 0.07 (1.2) 0.0 0.8T 1315 0.02 (0.9) 0.2 0.8Mean 0.10 (1.2) 0.2 1.0Cluster 3BG 396 0.23 (1.7) 0.1 1.2IPC 92-39 0.21 (1.6) 0.2 1.2K 850 0.23 (1.7) 0.1 1.2PANT G 114 0.22 (1.6) 0.3 1.3PDG 84-16 0.24 (1.7) 0.0 1.1Tyson 0.14 (1.4) 0.2 1.1Mean 0.21 (1.6) 0.1 1.2Cluster 4Barwon 0.18 (0.7) 0.4 0.8Dooen 0.18 (0.7) 0.0 0.5Norwin 0.18 (0.7) 0.2 0.6Mean 0.17 (0.7) 0.2 0.6Cluster 5ICC 10406 0.18 (1.5) 0.4 0.7ICC 10459 0.17 (1.5) 0.3 0.8ICC 4958 0.15 (1.4) 0.3 0.7ICC 5335 0.17 (1.5) 0.1 1.0ICC 5742 0.18 (1.5) 0.2 0.9ICC 5829 0.18 (1.5) 0.3 0.8ICC 7692 0.18 (1.5) 0.3 0.8ICCC 37 0.23 (1.7) 0.0 1.2IPC 92-2 0.19 (1.5) 0.1 1.0JG 62 0.19 (1.6) 0.1 1.0Mean 0.18 (1.5) 0.2 0.9UnclassifiedBG 365 0.22 (1.7) 0.1 1.1IPC 94-132 0.23 (1.7) 0.4 1.4IPC 94-94 0.19 (1.5) 0.4 1.4Genotype LSD (P < 0.05) 0.05 0.4 0.2Cluster LSD (P < 0.05) 0.02 0.2 0.1Values in parentheses are back transformed seed yields in trial/ha. Genotype and clustermean LSDs are calculated using average standard errors of the difference across all clusters.formed a wedge shape, with large differences at early,southern sites, which decreased as sites became later (Fig. 5band c). Thus at Gulbarga in 2000, Cluster 5 podded 27 daysearlier, and matured 18 days earlier than Cluster 4, whereas inHisar the difference was only 8 and 2 days, respectively(Fig. 5b and c). Clusters 1 and 3were intermediate in podding,significantly later than Cluster 5, but earlier than Cluster 2 atmost sites earlier than 110 days. There were no regressionslopedifferences for poddingbetweenClusters 1–3.However,Cluster 1 was the second earliest maturing (P < 0.001 to<0.019), with large differences at early sites (Fig. 5c),reflected in a regression slope coefficient significantly largerthan that of Clusters 2 and 4 (P < 0.05). Cluster 3 wasintermediate in maturity, being significantly (P < 0.05)earlier and more responsive than Clusters 2 and 4.Plotting genotype responses to latitude confirmed thesetrends and demonstrated the role of germplasm origin.Outliers excepted, the flowering response to latitude waspositively correlated to the genotype mean (r2 = 0.59), andboth the clusters based on yield and centres of origin formeddiscrete groups along the regression curve (Fig. 6a).Southern and central Indian germplasm was earlier flower-ing, and less responsive than that from the north, andtherefore limited to the left of Fig. 6a (with the singleexception of ICC 7692, from Gujarat in northwest India).Accordingly, Cluster 5 was located on the lower leftquadrant of Fig. 6a, Cluster 1 was central, while Clusters 2and 3 were largely located in the upper right quadrant.Norwin and Dooen were both late flowering, unresponsivemembers of Cluster 4, located as outliers in the lower right ofFig. 6a.In contrast to flowering, the maturity response to latitudewas strongly negatively correlated to the genotype mean(r2 = 0.74), and there were no outliers in this relationship(Fig. 6b). Early maturing genotypes (such as those in Cluster5, or of southern and central origin) were able to delay theirmaturity date much more than their later, predominantlynorthern counterparts as trial site latitude increased(Fig. 6b). As a result, the difference between early andlate maturing genotypes became progressively smaller assite mean maturity increased (Fig. 6), and this is responsiblefor the wedge shaped regression pattern in Fig. 5c.3.4. Other traitsOrthogonal contrasts between Clusters 3 and 5 performedwithin sites highlighted other traits associated with specificadaptation to north and south (Table 4). Cluster 3accumulated more biomass (largely vegetative) by maturitythan Cluster 5 at all northern sites (except Delhi in 1999),and three of the five central sites, whereas the opposite wasthe case at Gulbarga (Table 4). In fact, Clusters 1 and 3accumulated the highest biomass overall, whereas Cluster 5was the second lowest, followed by Cluster 4 (data notpresented). Similarly, Cluster 3 was significantly taller thanCluster 5 at most sites, and often appeared to branch moreJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244240Fig. 5. Phenology of the five Ward’s yield clusters regressed against site mean: (a) flowering, (b) podding and (c) maturity. Linear regression models fittingseparate lines for each cluster accounted for 79, 85 and 94% of total variance, respectively.profusely (Table 4). Moreover, Cluster 3 had a highernumber of seeds per pod at many central and northern sites,despite generally having a larger seed size than Cluster 5(Table 4). However, Cluster 5 had a significantly higherharvest index and greater number of pods per plant thanCluster 3 in 6, and 5 of the 7 central and southern sites,respectively (Table 4). Clusters 1 and 5 had the highestharvest indices overall (data not presented). Floweringduration and the interval between flowering and poddingwas significantly longer in Cluster 5 at most sites (Table 4),with particularly large differences in the latter at Hisar (48days versus 21 days).4. DiscussionThis study has provided clear evidence for both generaland specific adaptation to northern and southern/centralJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244 241Fig. 6. Genotype phenological responsiveness to changes in latitude vs. genotype mean flowering (a), and maturity (b), as defined by averaging over all trialsites. (Note that the linear regression in (a) excludes the named outliers, Norwin, Dooen and H 208.) Genotype responsiveness was defined by slopes producedby linear regression against trial site latitude. Genotype-latitude curves were strongly linear, with average correlation coefficients of 0.75 and 0.87 for floweringand maturity, respectively. Cluster membership (see Fig. 3) is superimposed on genotype scores as different marker patterns, and origin is indicated byabbreviation: A, Australia; M, Mediterranean; S, south India; C, central India; N, north India and NW, northwest India (Gujarat).chickpea growing regions of India, and demonstrated thepivotal role of phenology. We confirm the importance ofhigh harvest index and drought escape through earlyflowering, podding and maturity at stressful locations insouthern and central regions like Gulbarga and Sehore(Saxena and Sheldrake, 1980; Saxena, 1984), and demon-strate that specifically adapted germplasm largely originatesfrom these regions. ICC 7692, from Gujarat was aninteresting exception to this trend, being both early andvery high yielding at Gulbarga. However, given that Gujaratis one of the drier and warmer chickpea growing areas inIndia (Table 1), this is perhaps not surprising. Conversely,material specifically adapted to northern sites originatedalmost exclusively from northern breeding programmes, andwas characterized by an intermediate to late phenology, highbiomass and crop height. However, even in northernlatitudes there is a limit to how much flowering andmaturity can be delayed without penalizing yield. ThusCluster 2, and particularly Cluster 4, comprising manyAustralian cultivars, were not as high yielding as the moreintermediate Cluster 3 even in the productive, long seasonnorthern sites such as New Delhi, Kanpur and Hisar.The present study adds insight to the detailed physiologyconducted previously (Saxena and Sheldrake, 1980; Saxena,1984; Khanna-Chopra and Sinha, 1987) by highlighting thedynamic role phenology plays in adaptation. Genotypesspecifically adapted to the north are able to delay floweringat later flowering sites or higher latitudes more than non-adapted material. A similar finding has been reported inyield-responsive Mediterranean Vicia species (Berger et al.,2002). Flowering later under unstressful conditionsincreases both source and sink potential (Saxena, 1984;Khanna-Chopra and Sinha, 1987), and importantly inchickpea, reduces the time interval between flowering andJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244242Table4ProbabilityvaluesfortheorthogonalcontrastofCluster3vs.5forarangeofphenologicalandproductivityassociatedtraitsrecordedateachsiteSiteZoneBiomassVeg.matterSeedyieldHarvestIndexCropheight18branchnumber28branchnumberPods/plantSeedsizeSeeds/podFloweringdurationFlower-podintervalGulbarga2000Southern:17.48N0.002C50.447<0.001C5<0.001C5<0.001C30.030C30.188<0.001C5<0.001C3Gulbarga1998<0.001C5<0.001C5<0.001C50.006C5<0.001C5<0.001C30.527<0.001C30.313Sehore2000Central:23.2oN<0.001C3<0.001C3<0.001C30.6590.04C30.978<0.001C50.1550.09C5<0.001C30.008C50.281Sehore19990.051C30.002C30.283<0.001C5<0.001C30.024C3<0.001C3<0.001C5<0.001C30.392<0.001C5<0.001C5Jabalpur19980.7390.2670.006C50.002C50.9290.395<0.001C3<0.001C5<0.001C5Jabalpur20000.069C50.988<0.001C5<0.001C50.022C5<0.001C30.086C3<0.001C5<0.001C3<0.001C30.017C50.003C5Jabalpur19990.021C3<0.001C30.3730.002C5<0.001C30.970.042C50.015C5<0.001C30.053C5Kanpur1998Northern:26.8-29.08N<0.001C3<0.001C30.1410.046C3<0.001C3<0.001C5<0.001C3Kanpur19990.003C30.6460.6770.048C30.9690.479<0.001C3Kanpur20000.046C30.043C30.078C30.851<0.001C30.770.470.5870.8250.003C3<0.001C50.499Durgapura20000.020C3<0.001C30.596<0.001C5<0.001C30.028C30.3280.126<0.001C30.7310.2530.486Durgapura19990.004C30.044C30.017C30.952<0.001C30.603<0.001C5<0.001C30.008C3<0.001C50.018C5NewDelhi19990.1380.3940.062C30.166<0.001C30.2690.4950.4330.005C30.018C3<0.001C5<0.001C5NewDelhi2000<0.001C3<0.001C3<0.001C30.005C3<0.001C30.190.5360.4460.876<0.001C3<0.001C5<0.001C5Hisar1999<0.001C3<0.001C3<0.001C30.597<0.001C30.003C3<0.001C30.4320.2760.587<0.001C5<0.001C5Significantcontrastsarehighlightedinbold,andtheclusterwithlargervaluesidentified:C3,Cluster3;C5,Cluster5.podding. Because chickpea is sensitive to mean tempera-tures<16 8C at flowering, pod set can be delayed by up to 70days under extreme conditions, manifested by repeatedcycles of flowering and subsequent abortion (Berger et al.,2004, 2005). In the present study this was evident at Hisar,where pod set occurred almost 50 days after flowering inCluster 1 genotypes. The opportunity cost of repeated cyclesof flower set and abortion is the diminution of resourcesallocated to vegetative growth, with the attendant reductionin source/sink potential (Saxena, 1984; Khanna-Chopra andSinha, 1987). This is confirmed by the relatively lowbiomass of Cluster 5 genotypes at all sites except forGulbarga. The precise mechanism behind the capacity ofspecifically adapted material to flower relatively later athigher latitudes remains to be elucidated. In chickpea therate of progress to flowering is determined by the response toincreasing day length or temperature, or more commonly,additive combinations of the two (Roberts et al., 1985).While Roberts et al. (1985) demonstrated that Indian desitypes were more temperature responsive than Mediterraneankabulis, their Indian sample size was too small to drawconclusions about specific adaptation within India. There-fore, whether germplasm specifically adapted to the south ismore temperature responsive, and northern adapted materialmore photoperiod responsive, remains an open question.The issue is important because an understanding of thephotothermal drivers underlying crop phenology simplifiesthe selection of adapted germplasm for new environments.In lentil, where flowering responses are well understood(Erskine et al., 1994), this approach was used to selectappropriate parental material to expand the crop into WestAsian highland regions (Keatinge et al., 1996) and a range ofAustralian environments (Materne, 2003).The indeterminate growth of chickpea (Fig. 6b) allowsgermplasm specifically adapted to the south to partiallycompensate for excessively early flowering in the north, andexplains why yield differences between adapted and non-adapted germplasm are smaller at Hisar than at Hyderabad(Saxena and Sheldrake, 1980; Saxena, 1984). The combina-tion of intermediate flowering and relatively early,responsive maturity, as typified by Cluster 1, is aphenological compromise that leads to wide adaptation,with high yields both in the north and south. In the south,intermediate flowering and early maturity in Cluster 1provides sufficient drought escape to match Cluster 5 yieldsat all but one site, whereas a relatively delayed maturity inthe north gives rise to a similar yield as Cluster 3 at all higheryielding sites. Averaged across all sites, Cluster 1 combinedthe equal highest biomass (shared with Cluster 3) with thehighest harvest index (shared with Cluster 5). ICCV 10 is thebest example of this phenological compromise, consistentlyranking in the top 10 at 10 sites, a result which supportsprevious studies in southern and central Peninjsula India(van Rheenen, 1991) and Mediterranean Australia (Bergeret al., 2004). ICCV 10 notwithstanding, the ultimatephenological package for wide adaptation in India wouldJ.D. Berger et al. / Field Crops Research 98 (2006) 230–244 243combine early, highly responsive flowering and maturitybecause this combination allows for drought escape in thesouth, and sufficient time to develop both source and sinkpotential in the north. This combination did not exist in thepresent study, and may be hard to find given the positivecorrelation between flowering time and flowering respon-siveness to increasing latitude (Fig. 6a).The consistent association between germplasm originand specific adaptation to northern and southern Indiasuggests that the state-based breeding programmes aretargeting their local environments well. This is supported bythe increasing yield and yield responsiveness of the newernorthern varieties (Cluster 3) compared to the older cultivars(Cluster 2). With the exception of the IARI breedingprogramme in New Delhi, there is less evidence for theproduction of widely adapted varieties. Eight of the 13members of Cluster 1 were produced by IARI, and IARIgermplasm was also widely adapted in Australia (Bergeret al., 2004). The IARI chickpea breeding programme isbased on wide intra- and inter-specific crosses usinggenetically diverse parental material from a range oforigins, usually with more than two parents in each cross(Yadav et al., 2004). A shuttle breeding approach isemployed, in which early generation material is first grownin the field in New Delhi and then transferred to Dharwad(Karnataka) in southern India in the subsequent generation,and so on. This appears to select for the intermediatephenological compromise outlined above, and also exposesthe material to a wider range of biotic stresses than areexperienced in a single environment. The use of southernand northern evaluation sites is essential for producingwidely adapted material in India. Southern sites arenecessary to readily identify differences in maturity(Fig. 5), while both southern and northern sites are requiredquantify flowering temperature and photoperiod respon-siveness in order to target new material to matchingenvironments.5. ConclusionsThis study has identified both specific and wideadaptation in chickpeas to low and high yielding environ-ments of southern and northern India, and demonstrated thecentral role of phenology, biomass and harvest index.Drought escape through early phenology and high harvestindex are critical traits for yield in southern and centralIndia. In the north later flowering is necessary to maximizebiomass accumulation, and delay pod set until temperaturesrise sufficiently to prevent abortion. Germplasm specificallyadapted to the north is able to delay flowering significantlymore at later sites than unadapted material. The role oftemperature and photoperiod in specific adaptation tonorthern and southern India will be investigated in acompanion paper.Widely adapted genotypes combine intermediate flower-ing and relatively early, responsivematurity to produce highbiomass and harvest index. Most state-based breedingprogrammes are producing material specifically adapted totheir region. The IARI programme in New Delhi is anexception to this trend, producing widely adapted genotypesby making wide intra- and inter-specific crosses, andevaluating the progeny in both northern and southern India.AcknowledgementsThe authors acknowledge the Australian Centre forInternational Agricultural Research (ACIAR), the Centrefor Legumes in Mediterranean Agriculture (CLIMA), theCommonwealth Scientific and Industrial Research Organiza-tion (CSIRO) and the IndianCouncil ofAgricultural Research(ICAR) for their generous research support. The ICAR, theCGIAR system, and plant breeders in Australia and India arethanked for the use of their germplasm. This work could nothave been attempted without considerable technical assis-tance at each of our trial sites. In this context the authorswouldparticularly like to thank Drs. S.R. Kushwaha, Tej Pal Singh,SureshKumarYadav, andPankajKumar,Mr.N.S.Hooda,Mr.Bishnath, and Mr. Bharat Singh from the Indian AgriculturalResearch Institute (IARI) and Dr. P.K. Tyagi from HaryanaAgriculturalUniversity (HAU). Finally,Ms. Jane Speijers andProf. Peter Clarke (DAWA) are thanked for their statisticalinput, particularly for the ongoing discussion on the use ofmultivariate analysis for visualizing G E interaction.ReferencesAli, M., Kumar, S., 2003. Chickpea research in India: an overview. In: Ali,M., Kumar, S., Singh, N.B. (Eds.), Chickpea Research in India. IndianInstitute of Pulses Research, Kanpur, pp. 1–14.Berger, J.D., Buck, R.P., Henzell, J.M., Turner, N.C., 2005. Evolution in thegenus Cicer – vernalization response and low temperature pod set inchickpea (C. arietinum L.) and its annual wild relatives. Aust. J. Agric.Res. 56, 1191–1200.Berger, J.D., Robertson, L.D., Cocks, P.S., 2002. Genotype environmentinteraction for yield and other plant attributes among undomesticatedMediterranean Vicia species. Euphytica 126, 421–435.Berger, J.D., Turner, N.C., Siddique, K.H.M., Knights, E.J., Brinsmead,R.B., Mock, I., Edmondson, C., Khan, T.N., 2004. Genotype byenvironment studies across Australia reveal the importance of phenol-ogy for chickpea (Cicer arietinum L.) improvement. Aust. J. Agric. Res.55, 1–14.Bhambota, S.K., Sood, B.C., Gartan, S.L., 1994. Contribution of differentcharacters towards seed yield in chickpea (Cicer arietinum L.) Indian J.Genet. Plant Breed. 54, 381–388.Chaudhary, B.D., Singh, P., Kumar, A., Singh, D.P., 1988. Correlation andmultiple regression analysis in chickpea. Crop Improv. 15, 180–186.Coombes, N., 2002. Excavating for designs: SpaDes to DiGGer, spatialdesign search. In: Clarke, P. (Ed.), Australasian Genstat Conference,Busselton, West Australia.DeLacy, I.H., Basford, K.E., Cooper, M., Bull, J.K., McLaren, C.G., 1996.Analysis of multi-environment trials – an historical perspective. In:Cooper, M., Hammer, G.L. (Eds.), Plant Adaptation and Crop Improve-ment. CAB International, Wallingford, pp. 39–124.J.D. Berger et al. / Field Crops Research 98 (2006) 230–244244Erskine, W., Hussain, A., Tahir, M., Bahksh, A., Ellis, R.H., Summerfield,R.J., Robberts, E.H., 1994. Field evaluation of a model of photothermalflowering responses in a world lentil collection. Theor. Appl. Genet. 88,423–428.FAO, 2004a. CLIMWAT. Food and Agriculture Organization of the UnitedNations http://www.fao.org/ag/agl/aglw/climwat.htm.FAO, 2004b. FAOSTAT. Food and Agriculture Organization of the UnitedNations http://www.apps.fao.org/default.htm.Finlay, K.W., Wilkinson, G.N., 1963. The analysis of adaptation in a plantbreeding programme. Aust. J. Agric. Res. 14, 742–754.Fox, P.N., Rosielle, A.A., 1982. Reducing the influence of environmentalmain-effects on pattern analysis of plant breeding environments. Euphy-tica 31, 645–656.Genstat, 2002. Genstat for Windows-6th Edition. VSN International http://www.vsn-intl.com/.Haloi, B., Baldev, B., 1984. Association among physiological traits andplant ideotype in chickpea Cicer arietinum L. as influenced by differentdates of sowing. J. Res. Assam Agric. Univ. 5, 84–89.Jirali, D.I., Panchal, Y.C., Janagoudar, B.S., Patil, B.C., 1994. Studies on thegrowth pattern and yield in chickpea (Cicer arietinum L.) genotypesunder receding soil moisture conditions. Indian J. Plant Physiol. 37,275–276.Jirali, D.I., Panchal, Y.C., Janardhan, K.V., Janagoudar, B.S., Patil, B.C.,1988. Studies on flowering pattern and yield of chickpea (Cicerarietinum L.) under receding soil moisture conditions. Indian J. PlantPhysiol. 31, 293–296.Keatinge, J.D.H., Qi, A.M., Kusmenoglu, I., Ellis, R.H., Summerfield, R.J.,Erskine, W., Beniwal, S.P.S., 1996. Using genotypic variation in flower-ing responses to temperature and photoperiod to select lentil for the westAsian highlands. Agric. Forest Meteorol. 78, 53–65.Khan, I.A., Malik, B.A., 1989. Grain and biological yield association inchickpea. Sarhad J. Agric. 5, 373–376.Khanna-Chopra, R., Sinha, S.K., 1987. Chickpea: physiological aspectsof growth and yield. In: Saxena, M.C., Singh, K.B. (Eds.), TheChickpea. CAB International, Wallingford, Oxon, UK, pp. 163–190.Materne, M.A., 2003. Importance of Phenology and Other Key Factors inImproving the Adaptation of Lentil (Lens culinaris Medikus) in Aus-tralia. University of Western Australia.Qureshi, A.S., Shaukat, A., Bakhsh, A., Arshad, M., Ghafoor, A., 2004. Anassessment of variability for economically important traits in chickpea(Cicer arietinum L.) Pakistan J. Bot. 36, 779–785.Rao, S.K., 1996. Variation for canopy characteristics and their inter relation-ship with seed yield in bold seeded chickpeas (Cicer arietinum L.)Agric. Sci. Digest 16, 230–232.Roberts, E.H., Hadley, P., Summerfield, R.J., 1985. Effects of temperatureand photoperiod on flowering in chick peas (Cicer arietinum). Ann. Bot.55, 881–892.Saxena, N.P., 1984. Chickpea. In: Goldsworthy, P.R., Fisher, N.M. (Eds.),The Physiology of Tropical Field Crops. John Wiley & Sons Ltd., pp.419–452.Saxena, N.P., Sheldrake, A.R., 1980. Physiology of growth, development,and yield in chickpeas in India. In: Proceedings of the InternationalWorkshop on Chickpea Improvement, ICRISAT, Patancheru, India, pp.106–120.Singh, D., Sharma, P.C., Kumar, R., 1997. Correlation and path coefficientanalysis in chickpea. Crop Res. 13, 625–629.Sinha, S.K., Lal, B., Koundal, K.R., Khanna-Chopra, R., 1983. Nitrogenfixation does not limit yield in pulses. Curr. Sci. 52, 916–918.SPSS, 2002. SPSS for Windows. http://www.spss.com/.Upadhyaya, H.D., 2003. Geographical patterns of variation for morpholo-gical and agronomic characteristics in the chickpea germplasm collec-tion. Euphytica 132, 343–352.van Rheenen, H.A., 1991. Chickpea breeding – progress and prospects.Plant Breed. Abstr. 61, 998–1009.Yadav, J.K., Singh, H.L., Kumar, R., 2003. Perusing selection parameters inchickpea (Cicer arietinum L.) New Agriculturist 14, 75–80.Yadav, R.K., 1991. Studies on variability and correlations in chickpea. Adv.Plant Sci. 4, 197–207.Yadav, S.S., Kumar, J., Turner, N.C., Berger, J.D., Redden, R., McNeil, D.,Materne, M., Knights, E.J., Bahl, P.N., 2004. Breeding for improvedproductivity, multiple resistance and wide adaptation in chickpea (Cicerarietinum L.) Plant Genet. Resour. 2, 1–8.Yaqoob, M., Jan, H., Ahmad, M.B., 1990. Interrelationship between grainyield and other important characters in chickpea (Cicer arietinum L.)Sarhad J. Agric. 6, 159–164.Zohary, D., Hopf, M., 2000. Domestication of Plants in the Old World.Clarendon Press, Oxford, pp. 108–111.

a

Abstract

Extracted data

a

Abstract

Extracted data

Related items

Related items