Relationships among Bread Wheat International Yield Testing Locations in Dry Areas.
CIMMYT's key drought evaluation site is located at the Centro de Investigaciones Agricolas del Noroeste (CIANO) in northwestern Mexico (27 [degrees] 20'N and elevation 38 m above sea level). Understanding the relationship between CIANO and key dry locations around the world is critical if we are to properly assess the effectiveness of this type of selection and evaluation. It is also important, particularly for CIMMYT's regional cooperators, to link the performance of different dry locations and regions from around the world with their own environments. Other authors have stated the importance of targeting germplasm to specific environments (Peterson and Pfeiffer, 1989) and increasing the efficiency of yield evaluation through the identification of key locations (Abdalla et al., 1996). Regions with similar dominant moisture stress patterns are the Southern Cone of South America, North Africa--West Asia--Southern Africa, and dry areas in South Asia (Rajaram et al., 1994; Calhoun et al., 1994).
Two types of multiplicative models have been used for studying genotype x environment interaction (GEI) and for developing methods for clustering sites or cultivars into groups without crossover interaction (COI) (Cornelius et al., 1992, 1993; Crossa et al., 1993, 1995, 1996; Crossa and Cornelius, 1993, 1997; Osman et al., 1997). These are the shifted multiplicative model (SHMM) in which [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (Seyedsadr and Cornelius, 1992) and the site regression model (SREG) in which [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII]. (Cornelius et al., 1996). The variable [[bar]y.sub.ij.] is the mean of the [i.sup.th] cultivar (i = 1,2, ..., g) in the [j.sup.th] environment (j = 1,2, ..., e); [Beta] is the shift parameter; [[Mu].sub.j] is the site mean; [[Lambda].sub.k] ([Lambda.sub.1] [is greater than or equal to] ([Lambda.sub.2] [is greater than or equal to] ... [is greater than or equal to] [[Lambda].sub.t]) are singular values that allow the imposition of orthonormality constraints on the singular vectors for cultivars, [[Alpha].sub.ik] = ([[Alpha].sub.1k], ... [Alpha].sub.gk] and sites, [[Gamma].sub.1k] ..., [[Gamma].sub.ek] , such that [MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] and [Sigma].sub.i][[Alpha].sub.ik][[Alpha].sub.ik'] = [[Sigma].sub.j][[Gamma].sub.jk][[Gamma].sub.jk'] = 0 for k [is not equal to] k'; [[bar][Epsilon].sub.ij.] is the residual error.
If SHMM and SREG models with one multiplicative component ([SHMM.sub.1] and [SREG.sub.1]) are adequate for fitting the data and primary effects of the sites, [[Gamma].sub.j1], all of like sign, then [SHMM.sub.1] and [SREG.sub.1] predict non-COI. Thus all cultivars should have consistent patterns of response across all sites included in the analysis (Crossa and Cornelius, 1997). On the contrary, if [[Gamma].sub.j1] are of different signs, then [SHMM.sub.1] and [SREG.sub.1] models predict COI, that is, cultivar ranking in the sites with negative [[Gamma].sub.j1] are the reverse of the cultivar ranking in the sites with positive [[Gamma.sub.j1].
This analysis has been used to determine environmental subgroups of large numbers of sites sown to the same set of cultivars (Fox et al., 1985, 1990). However, trials conducted over many years frequently contain unbalanced sets of cultivars as breeders constantly replace lines with newer materials. In this instance pattern analysis, a combination of classification and ordination analyses has been successfully employed (DeLacy and Lawrence, 1988; Peterson and Pfeiffer, 1989; Abdalla et al., 1996). These techniques have been used to examine the association of locations to CIMMYT spring bread wheat germplasm (DeLacy et al., 1994). However, all these cultivars were developed for irrigated conditions, and site associations were determined across both irrigated and low rainfall conditions. There has been no such attempt to classify global drought locations sown to cultivars specifically developed for performance under moisture limiting conditions.
The aim of this paper is to (i) examine the relevance of selection under terminal moisture stress at CIANO, Mexico compared to the primary drought affected target areas around the world and (ii) examine the association among international testing locations where the SAWYT nursery is planted.
MATERIALS AND METHODS
Locations and Cultivars
Yield data from a total of 156 locations were returned from the SAWYT between 1992 and 1997. A total of six yield nurseries (SAWYTs 1-6), each comprised of 30 to 50 cultivars, were sown. Although most cultivars varied from year to year, a local check cultivar representing the best locally adapted germplasm was included at each site each year. The local check cultivar varied among locations and in some instances changed between years at the same location. All trials were sown as two replicate alpha-lattice designs (Barreto et al., 1997). Yield data from each trial were analyzed by SAS (SAS, 1988). Genotypes were considered fixed effects and replicates and subblocks within replicates as random effects. Adjusted means were calculated for subsequent SHMM and pattern analyzes. Trials were sown under a range of different moisture conditions. A site is defined as a location/year occurrence. High yielding irrigated locations (arbitrarily defined as those with means above 6 Mg [ha.sup.-1]) were removed to ensure that the remaining sites were representative of the potential yield range in most water limited locations. To ensure clusters among locations were biologically based and not artefactual, only those sites indicating significant differences among genotypes, regardless of the size of the coefficient of variation, were retained giving a total number of 122 sites (Table 1). Diseases scored were stem, leaf, and stripe rust (caused by Puccinia striiformis f. sp. hordeii) and Septoria tritici Roberge in Desmaz.
Table 1. Locations returning yield data for the 1st through 6th SAWYTs in each geographic region. SAWYT nursery Region Site and country number Latitude Southern Africa 1 Small Grain Inst. 1,5,6 28 [degrees] 12'S (S. Africa) 2 Langgewens (S. Africa) 3 33 [degrees] 17'S 3 Moredou (S. Africa) 2 33 [degrees] 18'S 4 Gwebi (Zimbabwe) 2 17 [degrees] 41'S North Africa 5 Gezira (Sudan) 1 14 [degrees] 24'N 6 El Khroub (Algeria) 1 36 [degrees] 7'N 7 Sidi-Bel-Abbes 3 35 [degrees] 17'N (Algeria) 8 Zakaria (Algeria) 1 34 [degrees] 36'N 9 Khemis-Miliana 2 36 [degrees] 15'N (Algeria) 10 Ain El Hadjar (Algeria) 2 11 Beni-Slimane (Algeria) 3 36 [degrees] 14'N 12 Ismailia (Egypt) 1,3 30 [degrees] 36'N 13 El Qasser (Egypt) 2 31 [degrees] 20'N 14 Boulifa (Tunisia) 1 36 [degrees] 38'N East Africa 15 Kisozi (Burundi) 5 3 [degrees] 33'S 16 NPBRC-Njoro (Kenya) 5 0 [degrees] 25'S 17 Bembeke (Malawi) 6 14 [degrees] 10'S 18 Holetta (Ethiopia) 6 9 [degrees] 3'N 19 Simba (Tanzania) 6 3 [degrees] 13'S 20 Lyamungo (Tanzania) 3 3 [degrees] 14'S West Asia 21 SPII Cereal Res. 2 35 [degrees] 50'N Inst. (Iran) 22 Zahak-zabol (Iran) 2 30 [degrees] 53'N 23 Ahwaz (Iran) 2 31 [degrees] 17'N 24 Gonbad (Iran) 2 37 [degrees] 16'N 25 Ramtha (Jordan) 1 32 [degrees] 34'N 26 Rawdat Harma (Qatar) 1 25 [degrees] 48'N 27 ICARDA Tel-Hadya 1,5 36 [degrees] 1'N (Syria) 28 Shesham Bagd 2 34 [degrees] 25'N (Afghanistan) 29 Al Kharj (Saudi Arabia) 2 24 [degrees] 15'N Central Asia 31 Akmola([paragraph]) 4,5 51 [degrees] 10'N (Kazakstan) South Asia 32 Dinajpur W.R.C 1,5 25 [degrees] 38'N (Bangladesh) 33 Rajshahi (Bangladesh) 2,3 24 [degrees] 22'N 34 Bhairahawa (Nepal) 1,2,3,4,5 27 [degrees] 6'N 35 Islamabad (Pakistan) 1,4 33 [degrees] 45'N 36 Sariab (Pakistan) 2,5 30 [degrees] 12'N 37 Pirsabak (Pakistan) 5 33 [degrees] 59'N 38 Wheat Res. Inst. 4 31 [degrees] 25'N (Pakistan) 39 Barani (Pakistan) 4,5 32 [degrees] 56'N 40 Dera Ismail Khan 2,3,4,5,6 31 [degrees] 50'N (Pakistan) 41 Durgapura (India) 3 26 [degrees] 58'N 42 DWR-Karnal (India) 3 29 [degrees] 40'N 43 Vijapur (India) 3 23 [degrees] 35'N 44 PAU-Ludhiana (India) 5 30 [degrees] 56'N East Asia 45 San-Pa-Tong (Thailand) 3,4 18 [degrees] 30'N 46 Suwan Farm (Thailand) 3 14 [degrees] 40'N 47 Samoeng (Thailand) 4 18 [degrees] 17'N 48 Pang Ma Pha (Thailand) 4 19 [degrees] 28'N North America 49 CIANO (Mexico) 1,3,4,5,6 27 [degrees] 20'N 50 El Batan (Mexico) 5 19 [degrees] 31'N 51 Mixteca Oaxaca (Mexico) 3,4 17 [degrees] 33'N 52 Tecamac (Mexico) 6 19 [degrees] 43'N 53 Tiacaque (Mexico) 6 19 [degrees] 45'N 54 San Franc. Atizapan 5 19 [degrees] 16'N (Mexico) 55 Kernen Res. Farm 2,3,5,6 52 [degrees] 9'N (Canada) 56 Swift Current (Canada) 2 50 [degrees] 17'N Southern Cone 57 Bela Vista (Brazil) 1 23 [degrees] 0'S 58 Londrina (Brazil) 2 23 [degrees] 22'S 59 CNP-Soja (Brazil) 2 23 [degrees] 12'S 60 Pergamino (Argentina) 1,3,6 33 [degrees] 56'S 61 Marcos Juarez 1,3,4,5,6 32 [degrees] 42'S (Argentina) 62 Tucuman-Obispo 1,6 26 [degrees] 48'S (Argentina) 63 Bordenave (Argentina) 1 37 [degrees] 51'S 64 Parana (Argentina) 1 31 [degrees] 50'S 65 Cordoba (Argentina) 2 31 [degrees] 30'S 66 La Tijereta (Argentina) 5 32 [degrees] 8'S Andean Region 67 San Benito (Bolivia) 1,2,3,4,5,6 17 [degrees] 30'S 68 Sta. Catalina (Ecuador) 1,5 0 [degrees] 22'S Central America 69 Zamorano (Honduras) 4 14 [degrees] 0'N Southern Europe 70 Gimenells (Spain) 5 41 [degrees] 35'N 71 La Mojonera (Spain) 1 39 [degrees] 58'N 72 Cent. de Inv. Agrario 2 37 [degrees] 21'N (Spain) 73 Torregrossa/Belloc 2,4 41 [degrees] 35'N (Spain) 74 Cortijo Torrevuelo 2 37 [degrees] 38'N (Spain) 75 La Orden (Spain) 4 38 [degrees] 49'N 76 Tobiscal (Spain) 4 78 Kentziko Thermi 4 40 [degrees] 38'N (Greece) Eastern Europe 79 Szeged (Hungary) 1 46 [degrees] 0'N 80 Odessa (Ukraine) 3 46 [degrees] 27'N 81 Kharkiv (Ukraine) 4 50 [degrees] 0'N 82 Spring Wheat Lab. 3,4 53 [degrees] 1'N (Russia) Total sites 122 SAWYTs Standard reporting Yield deviation significant mean ([double disease Region Altitude ([dagger]) dagger]) ([sections]) m Mg [ha. sup.-1] Southern Africa 1 1687 2.42-5.02 0.52-0.66 1,5 2 91 4.49 0.73 3 82 2.88 0.68 4 1448 4.88 0.59 North Africa 5 411 2.98 0.66 6 640 3.56 0.42 7 483 4.29 0.81 8 980 0.60 0.17 9 289 3.32 0.48 10 1.37 0.34 11 600 2.59 0.43 12 10 2.89, 3.72 0.77-0.84 3 13 2 1.31 0.41 14 350 4.12 0.31 East Africa 15 2090 1.34 0.37 16 2165 4.18 0.61 5 17 1560 0.48 0.10 18 2400 2.94 0.85 19 1750 2.78 0.40 20 1280 2.08 0.33 West Asia 21 1321 5.56 0.75 22 493 4.84 0.87 23 20 5.07 0.89 24 76 4.19 0.71 25 520 1.61 0.24 26 50 4.13 0.95 27 282 2.26, 4.19 0.57, 0.61 1 28 552 4.68 0.44 29 540 3.88 0.58 Central Asia 31 300 1.34, 1.04 0.21, 0.33 South Asia 32 30 4.64, 3.29 0.88, 0.47 1 33 18 3.42, 3.36 0.28, 0.51 34 105 1.39-2.25 0.42-0.35 35 683 4.41, 3.62 0.47, 0.60 36 1600 0.94, 1.52 0.16, 0.32 37 340 3.36 0.56 38 117 3.00 0.87 39 490 2.36 0.25 40 290 0.52-2.75 0.24-0.52 41 450 2.00 0.43 42 300 5.65 0.69 43 126 2.59 0.52 44 247 4.68 0.66 East Asia 45 300 0.88, 0.97 0.14, 0.34 46 300 3.84 0.57 47 820 3.45 0.86 48 560 4.88 0.49 North America 49 38 2.60-4.14 0.41-0.23 50 2249 3.65 0.64 51 2250 2.71, 1.80 0.51, 0.44 52 2260 2.64 0.82 53 2300 3.66 0.52 54 2640 3.41 0.45 55 497 4.24-5.93 0.41-0.63 56 825 3.16 0.25 Southern Cone 57 618 2.07 0.38 58 540 3.21 0.41 59 620 4.06 0.56 60 65 2.77-5.01 0.52-0.23 3 61 110 1.19-2.79 0.34-0.38 62 460 1.23, 2.47 0.21, 0.42 63 212 5.12 0.99 64 110 2.03 0.12 1 65 425 0.38 0.15 66 200 2.26 0.56 5 Andean Region 67 2730 0.88-3.58 0.14-0.72 4 68 3050 3.35, 2.95 0.44, 0.33 1 Central America 69 805 1.16 0.27 Southern Europe 70 290 4.16 0.67 71 220 5.28 0.49 72 650 4.35 0.43 73 200 4.88, 3.41 0.69, 0.75 74 72 0.96 0.14 75 200 3.38 0.69 76 4.19 0.59 78 38 3.52 0.46 Eastern Europe 79 80 4.30 0.48 80 34 3.48 0.83 81 170 1.15 0.31 82 47 1.54, 2.70 0.37, 0.43 Total sites ([dagger]) When more than two SAWYTs are sown at the same site in different years a yield range is presented. ([double dagger]) When more than two SAWYT are sown at the same site the standard deviations of the lowest and highest yielding sites are presented. ([sections]) Diseases reported included stem, leaf and stripe rust and Septoria tritici. ([paragraph]) Akmola has recently been renamed Astana.
The sites were grouped into seven regions representing northern, southern and eastern Africa, West Asia, South Asia, the Southern Cone of South America, and Mexico. These groupings represent regions suffering somewhat different stress patterns as determined on the basis of long-term weather records. North Africa, West Asia, and Southern Africa generally experience Mediterranean or postanthesis moisture stress; South Asia experiences residual moisture stress; and the Southern Cone predominantly preanthesis stress (Rajaram et al., 1994). Rainfall records were incomplete for a majority of locations during the years in which the SAWYT was grown, for this reason we relied on long term regional rainfall averages to determine into which broad category a particular site fell. To provide separate comparisons of all other regions with the Mexican evaluation sites, the latter were classified as a distinct region. Many locations, particularly in West Asia and North Africa, were sown to the SAWYT trial only once during the 6-yr period. For this reason individual trials within these regions were considered collectively in comparisons with other regions and locations.
Germplasm entering the SAWYT was developed in Mexico by shuttling segregating materials between two contrasting moisture regimes (Rajaram et al., 1994). At CIANO severe terminal moisture stress was generated during the winter crop cycle by gravity irrigating preformed beds 14 d prior to sowing. Segregating and advanced lines were sown in November on a receding moisture profile with no subsequent irrigation. Twenty-year average annual rainfall for the cropping period November to April is 48.2 mm. Materials were harvested in April and sown in May at Toluca in the central Mexican highlands (19 [degrees] 16'N and 2640 m above sea level) which receive approximately 800 mm of annual precipitation. Under this environment, materials are selected for responsiveness to moisture, nutrient inputs, and resistance to disease.
Analysis and Grouping of Locations
Multiplicative Models for Clustering Sites Without Crossover Interaction
In various site-clustering procedures developed on the basis of SHMM or SREG (Cornelius et al., 1992; Crossa et al., 1993; Crossa and Cornelius, 1997), the measure of distance (i.e., dissimilarity) between a pair of sites is the residual sum of squares (RSS) after fitting [SHMM.sub.1] or [SREG.sub.1], RSS([SHMM.sub.1]) or RSS ([SREG.sub.1]), respectively. The dichotomous splitting procedure used on the dendrogram obtained from SHMM cluster analysis facilitates finding groups with negligible COI within clusters. Computations are facilitated because the site regression model with one multiplicative term can be reparameterized as a shifted multiplicative model with one multiplicative component. In this study, the SHMM clustering procedure for grouping sites without COI (Crossa et al., 1993) was applied to each of the six SAWYTs, and clusters of sites with negligible COI were found.
Pattern analysis is the clustering and ordination of sites (or/ and cultivars) in the two-way data table of cultivars x sites. In this study, the data used were the three-way table of cultivar x site x year. It was assumed that cultivars in any given year were a representative sample of the germplasm under evaluation. Sites (individual location/year occurrences) were judged on the basis of their ability to discriminate among cultivars. Only sites that occurred in two or more SAWYTs were included in the overall pattern analysis. Since some sites were sown to more than two SAWYTs in different years, their comparisons had different levels of precision. The clustering strategy used is that recommended by (DeLacy and Cooper, 1990) and used by Abdalla et al. (1996). Dissimilarities between sites in each year and averaged across years were measured by squared Euclidean distances (SED). Since different years had different numbers of cultivars, the averages were weighted by the number of cultivars in each year. The incremental sum of squares criterion and the agglomerative hierarchical strategy procedure with SED as the dissimilarity measure were used for classification.
RESULTS AND DISCUSSION
Associations among all Locations and Regions
Average site yields ranged from 0.38 to 8.48 Mg [ha.sup.-1] during the 6-yr period. Significant disease incidence was reported at 11 of the 122 sites included in the analysis and no sites reported insect damage. Dendrograms developed from the SHMM cluster analysis were used to examine the association of various sites with key regions that frequently experience drought and with sites within those regions. A summary of dendrogram results is presented in Table 2. The number of clusters among various sites located within the seven key regions is expressed as a fraction of the total number of possible groupings or clusters. Site clusters were determined at the third fusion or third group level of the SHMM cluster analysis. A total value for the region is also calculated and expressed as a fraction. For example, comparisons of India with the Mexican region had a total value of 5/9 (56%). This was calculated by adding the fractions of the individual groups CIANO (2/4), Mixteca (2/3), Atizapan (0/1) and El Batan (1/1). Similarly, each location within each of the seven key regions is totaled across locations that clustered at least once with locations in each key region. For example, CIANO, which appears in the Mexican region, clustered 28 times out of 111 possible groupings with 13 different global locations or regions.
Table 2. Summary of regional associations from dendrograms generated for each SAWYT. Number of groupings with the region/ Total number of possible groupings Mexican Region Sites([dagger]) CIANO Semillas Mixteca Atizapan Bangladesh 4/6 0/1 1/1 1/1 India 2/4 -- 2/3 0/1 South Asia 9/25 0/3 4/11 2/7 Brazil 1/4 0/1 -- -- South Africa 1/3 0/1 1/1 0/1 Nepal 1/4 0/1 0/2 1/1 Mexico 2/7 0/1 1/2 1/2 N. Africa 1/8 2/3 0/3 -- Bolivia 0/5 1/1 0/2 0/1 Argentina 3/24 2/5 0/6 0/2 Spain 1/6 0/2 1/3 0/1 Pakistan 2/11 0/1 1/5 0/4 West Asia 1/4 0/3 -- 0/1 Total 28/111 5/23 11/37 5/22 South Asian Region Sites Banglad. Pakistan Nepal India Kazakstan 0/2 7/12 1/3 0/2 South Africa 0/2 5/9 0/4 2/4 Bangladesh -- 0/9 3/4 2/4 Mexico 4/6 3/23 1/8 5/8 West Asia 4/10 3/28 4/11 0/1 India 2/4 1/7 0/4 2/6 Argentina 1/10 9/25 1/11 2/8 Canada 0/3 4/9 0/3 0/4 South Asia 5/17 7/56 3/21 5/21 N. Africa 2/11 2/17 4/11 0/9 Bolivia 0/4 3/14 0/5 1/4 Spain 1/5 4/24 1/9 0/1 Nepal 3/4 0/13 -- 0/4 Pakistan 0/9 6/27 0/13 1/7 Total 22/87 54/273 18/107 20/83 West Asian Region Sites Iran Jordan Qatar Syria Bangladesh 0/4 1/1 0/1 1/2 Nepal 0/4 1/1 0/1 1/2 South Asia 2/20 2/3 0/3 3/10 West Asia 0/3 1/2 0/2 1/2 N. Africa 1/12 1/5 1/5 1/5 Algeria 1/8 0/2 0/2 0/2 Mexico -- 0/2 1/2 0/4 Pakistan 2/12 0/1 0/1 1/5 South Africa 1/4 0/1 0/1 0/2 Argentina 0/4 0/5 0/5 0/7 Spain 0/12 0/2 0/2 0/3 Brazil 0/8 0/1 1/1 0/1 Total 7/91 6/26 3/26 8/45 North African Region Sites Sudan Algeria Egypt Tunisia Nepal 0/1 1/6 1/3 1/1 Spain 0/2 3/10 2/5 0/2 Argentina 0/5 5/16 2/8 0/5 Bangladesh 0/1 1/6 0/3 1/1 Bolivia 0/1 2/6 0/3 0/1 Spain 0/2 2/10 2/8 0/2 Brazil 1/1 0/6 1/3 0/1 West Asia 1/3 3/20 0/17 2/3 South Asia 0/3 2/28 3/18 2/3 N. Africa 0/4 3/16 1/10 0/4 Pakistan 0/1 0/10 2/9 0/1 South Africa 0/1 0/6 1/3 0/1 Mexico 1/2 0/8 0/4 0/2 Total 3/27 22/148 15/94 6/27 Southern African Region Sites S. Africa Zimbab. Kazakstan 2/2 -- Canada 3/4 -- Argentina 6/13 0/1 India 2/4 -- Pakistan 5/10 0/3 South Asia 8/22 0/5 Mexico 2/9 -- Bangladesh 1/4 0/1 Bolivia 0/5 1/1 Spain 1/6 0/3 West Asia 1/11 0/7 Total 31/90 1/21 Eastern African Region Sites Tanzania Burundi Kenya Malawi West Asia -- 1/1 1/1 -- N. Africa 2/3 -- -- -- E. Africa 0/2 1/1 1/1 0/2 Pakistan 1/2 1/4 1/4 0/1 Nepal 1/1 0/1 0/1 -- Brazil 0/1 -- -- 0/1 Argentina 1/5 0/2 0/2 0/3 South Asia 2/8 1/7 1/7 0/1 Mexico 1/5 0/2 0/2 1/3 Bolivia 0/2 0/1 0/1 0/1 Total 8/29 4/19 4/19 1/12 Southern Cone of South America Sites Brazil Argentin Kazakstan -- 2/3 Canada 1/3 3/8 Argentina 3/10 8/26 Algeria 1/6 5/14 Bolivia 0/4 5/13 Pakistan 0/8 9/24 Spain 2/6 5/21 N. Africa 3/11 7/29 South Africa 0/4 3/9 Brazil -- 2/9 South Asia 2/15 10/53 Bangladesh 1/3 1/9 West Asia 3/17 2/20 Mexico 1/5 1/25 Nepal 1/4 0/13 Total 18/96 63/276 Number of groupings with the region/ Total number of possible groupings Mexican Region Sites([dagger]) El Batan Temacac Rancho Total Bangladesh 0/1 -- -- 6/10 India 1/1 -- -- 5/9 South Asia 1/7 0/1 0/1 16/55 Brazil -- 0/1 1/1 2/7 South Africa 0/1 -- -- 2/7 Nepal 0/1 -- -- 2/9 Mexico 0/2 0/2 0/2 4/18 N. Africa -- -- -- 3/14 Bolivia 0/1 0/1 1/1 2/12 Argentina 0/2 0/3 2/3 7/45 Spain 0/1 -- -- 2/13 Pakistan 0/4 0/1 0/1 3/27 West Asia 0/1 -- -- 1/9 Total 2/22 0/9 3/9 South Asian Region Sites Total Kazakstan 8/19 South Africa 7/19 Bangladesh 5/17 Mexico 13/45 West Asia 13/50 India 5/21 Argentina 13/54 Canada 4/19 South Asia 20/115 N. Africa 8/48 Bolivia 4/27 Spain 6/39 Nepal 3/21 Pakistan 7/56 Total West Asian Region Sites Afghani Saudi A. Total Bangladesh 1/1 1/1 4/10 Nepal 0/1 0/1 2/10 South Asia 1/5 1/5 9/46 West Asia 1/6 1/6 4/21 N. Africa 1/3 1/3 6/33 Algeria 1/2 1/2 3/18 Mexico -- -- 1/8 Pakistan 0/3 0/3 3/25 South Africa 0/1 0/1 1/10 Argentina 1/1 1/1 2/23 Spain 1/3 1/3 2/25 Brazil 0/2 0/2 1/15 Total 7/28 7/28 North African Region Sites Total Nepal 3/11 Spain 5/19 Argentina 7/34 Bangladesh 2/11 Bolivia 2/11 Spain 4/22 Brazil 2/11 West Asia 6/43 South Asia 7/52 N. Africa 4/34 Pakistan 2/21 South Africa 1/11 Mexico 1/16 Total Southern African Region Sites Total Kazakstan 2/2 Canada 3/4 Argentina 7/14 India 2/4 Pakistan 5/13 South Asia 8/27 Mexico 2/9 Bangladesh 1/5 Bolivia 1/6 Spain 1/9 West Asia 1/18 Total Eastern African Region Sites Ethiopia Total West Asia -- 2/2 N. Africa -- 2/3 E. Africa 0/2 3/8 Pakistan 0/1 4/12 Nepal -- 1/3 Brazil 0/1 1/3 Argentina 0/3 4/15 South Asia 0/1 5/24 Mexico 0/3 3/15 Bolivia 0/1 1/6 Total 0/12 Southern Cone of South America Sites Total Kazakstan 2/3 Canada 4/11 Argentina 11/36 Algeria 6/20 Bolivia 5/17 Pakistan 9/32 Spain 7/27 N. Africa 10/40 South Africa 3/13 Brazil 2/9 South Asia 12/68 Bangladesh 2/12 West Asia 5/37 Mexico 2/30 Nepal 1/17 Total ([dagger]) Sites include countries and geographic regions.
The South Asian Region
Rainfall, soil type, and farming practice in this region are diverse. Nepalese sites cluster 14% (three of 21 possible groupings) of the time with other sites across the South Asian Region. However, within this region, Nepal and Bangladesh are closely associated, while no relationship exists between Nepal and Pakistan. Kazakstan, South Africa, Canada, and Argentina also associate with Pakistan but less so with other countries within the South Asian region. West Asian locations cluster better with Bangladesh and Nepal than with Pakistan. Mexican locations, including CIANO, are the best predictors of environments in India and Bangladesh; however, they associate poorly with most sites in Pakistan and Nepal (Table 2). Even though the variation in latitude within the region is not large (24-33 [degrees] N), stress patterns across the region are very different. In Pakistan, dry areas are less influenced by monsoonal rain than are equivalent areas in central India, while in Nepal and some parts of India and Bangladesh stress patterns are influenced by availability of irrigation. Stress patterns in Pakistan are similar to those in higher latitude areas such as Kazakstan and Canada (Table 2). This association reflects the lack of photoperiod response in the CIMMYT materials included in the SAWYT nurseries. The grouping of South Africa and Pakistan, both dry rainfed areas of equivalent latitude, indicates a significant degree of association between these areas. Nepal and Bangladesh cluster together because of similar limited irrigation production systems. Within the South Asian region, Bangladesh and India clustered best with 14 different global sites and regions (Table 2).
The West Asian Region
Four Iranian sites returned yield data for the 2nd SAWYT, making it the best represented country in the West Asian region. Unfortunately, the 2nd SAWYT was not sown in Mexico, so comparisons between these sites and Mexican locations cannot be made. Iranian locations cluster least with other sites in the West Asian region and show very little association with other global locations and regions (Table 2). The two locations in Pakistan giving the closest, although still weak association with Iran, Sariab, and Barani are located in the dry, northern areas. If Iran is eliminated, then Bangladesh becomes the best predictor of West Asian locations (4/6), clustering with Jordan, Syria, Afghanistan, and Saudi Arabia followed by the combined South Asian region. Those sites clustering with Bangladesh are rainfed locations, with the exception of Jordan and Saudi Arabia, where the trial was sown under limited irrigation. The limited irrigation regimes generated in Bangladesh, therefore, appear to mimic those in the terminal moisture stress environments of West Asia. Lack of reliable rainfall records from these West Asian locations in the years the trials were sown make it difficult to draw firm conclusions. With removal of the Iranian sites, the next best predictors of West Asia are West Asian locations themselves (4/18) and North African sites (5/21), all of which have, based on long-term rainfall averages, similar Mediterranean type stress patterns. Mexican and South American locations clustered poorly with West Asia, ranging from 1/15 in Brazil to 1/8 in Mexico. Afghanistan and Saudi Arabian locations within the West Asian region clustered best across global locations (Table 2).
The North African Region
Clusters of North African sites with global sites and regions indicated that Nepal gave the best association, followed by Spain and the South American sites in Argentina, Bolivia, and Brazil (Table 2). Mexican sites did not predict this region well (1/16). The Sudanese site expressed little if any association with any of the SAWYT sites. This may be due severe heat stress late in the growing cycle, which is often experienced in Sudan and other North African locations. The best associations within the region are between Algeria and Argentina and Algeria and Bolivia (Table 2). These sites experience early season, preanthesis drought stress. Within the North African region Tunisia clustered best with other global sites and regions.
The Southern African Region
This region is comprised of the two countries South Africa and Zimbabwe. The association of sites in this region may be inflated because only six locations were used. Zimbabwe showed a different pattern of association compared with the South African sites (Table 2). This reflects the different latitude of the Zimbabwean site (17 [degrees] S) compared with the South African locations (28-33 [degrees] S). The grouping of high latitude locations in Kazakstan and Canada with South Africa reflects similar stress patterns. The grouping of Pakistan, India and Argentina suggested that stress patterns were similar among these regions. It is expected that West Asian and North African sites, which experience Mediterranean type drought stress, would group with South Africa. However, only one out of 22 possible groupings occurred between southern Africa and regions with similar stress patterns. However, the strong association of sites in South Africa with 11 different locations (31/90) from around the world suggests that South Africa could be utilized in global wheat breeding efforts.
The Eastern African Region
In this region, four locations returned data from one year only and a fifth, Tanzania, reported data from two years. Stress patterns in Eastern Africa tend to be similar to those in West Asia and North Africa (4/5), indicating the presence of terminal or late season drought stress. Tanzania was the only eastern African location to cluster with CIANO (1/6, not shown in Table 2). South Asian locations where farmers plant on residual moisture following the monsoon or use limited irrigation and Southern Cone locations, typified by preanthesis stress did not associate well with eastern Africa. Ethiopia (0/12) and Malawi (1/12) did not associate well with other global locations and regions. Rainfall records are not available for the Eastern African sites in the years the trials were sown making it difficult to assess whether the growing conditions were different from the long-term average. Among the eastern African locations, Tanzania was most closely associated with other global locations followed by Burundi and Kenya (Table 2).
The Southern Cone of South America
This region is represented by locations in Argentina and Brazil. Their association with high latitude sites in Kazakstan and Canada would indicate similar patterns of adaptation (Table 2). The grouping of Pakistan, Bolivia, Algeria, and Spain with Argentina indicates that patterns of adaptation in Argentina are similar to many parts of the world, where sites experience preanthesis drought stress. Limited rainfall records were available for some sites in Argentina during the years covering this study. While these records indicate preanthesis drought stress predominated, significant rainfall variation was observed at many sites; this is reflected in the relatively weak grouping of locations in Argentina with each other (8/26). Other South Asian sites Nepal, India, and Bangladesh most of which are sown on residual moisture or under limited irrigation, did not associate with Argentina (1/29). Mexico, where development of the genotypes and most of the yield trials were sown using a single preseeding irrigation showed very little association with this region (2/30). The clustering of Argentinean locations with many different global sites and regions (63/276) highlights the strategic value of these locations in differentiating germplasm in a global breeding program.
Association of Locations Repeated in More than One Year
Pattern analysis was used to examine the association among sites repeated in more than one year (Fig. 1). At the first fusion level two groupings resulted. Group 1 contained three locations from Argentina and one from Santa Catalina in Ecuador. South Africa, Egypt, and one location in Pakistan made up the remaining sites. Group 2 contained three South Asian locations, Bangladesh, Nepal, and Pakistan. In addition, Syria, Bolivia, and CIANO (Mexico) also clustered in this group.
Pattern analysis confirmed the conclusions of individual SHMM analyses by indicating an association between CIANO and South Asia (Nepal, Pakistan, and Bangladesh). As no Indian site was sown to SAWYT for more than one year, we could not include India in the comparison. Most South American locations, with the exception of Bolivia, clustered together in a separate group, also supporting SHMM conclusions. Interestingly, Syria, the only West Asian location reporting data for more than one year, also grouped with CIANO, indicating a degree of similarity between stress patterns at these two sites.
Association of Global Locations with CIANO
Locations grouping with CIANO on the basis of yield at the third fusion level of the SHMM cluster analysis were determined. Many of these locations were sown only once to the SAWYT and are indicated in Table 3. Sites in Sudan, Brazil, Qatar, India, Bangladesh, Portugal, Ukraine, Mexico (Atizapan), and Tanzania reported SAWYT data only once and clustered with CIANO. Other sites in Mexico (Oaxaca), Pakistan (NARC), Bangladesh (Dinajpur), and Argentina appeared twice, clustering once with CIANO, while two locations in Nepal and Pakistan (Dera Ismail Khan) occurred four times, clustering only once with the CIANO site.
Table 3. Sites grouping at the third fusion level in the SHMM cluster analysis with CIANO, Mexico, for yield from the 1st, 3rd, 4th, 5th, and 6th SAWYTs. No. of times grouping Sites clustering with CIANO with CIANO 1st SAWYT Sudan; Gezira Res. Station 1/1 Brazil; Bela Vista Do Paraiso 1/1 Qatar; Rawdat Harma 1/1 3rd SAWYT India; Durgapura 1/1 India; Vijapur 1/1 Mexico; Mixteca Oaxaca Bangladesh; Rajshahi 1/1 4th SAWYT Portugal; PBS Alentejo 1/1 Ukraine; Odessa 1/1 Pakistan; NARC Islamabad 1/2 5th SAWYT Mexico; San Franc. Atizapan 1/1 Bangladesh; Dinajpur 1/2 Nepal; NWRP, Bhairahawa 1/4 6th SAWYT Tanzania; Simba 1/1 Pakistan; Dera Ismail Khan 1/4 Argentina; Tucuman-Obispo 1/2
Because many locations clustering with CIANO from the SHMM analyses were sown only once to the SAWYT during the 6-yr period, we can only conclude that an association exists in the year in which the head-to-head comparison occurred. The repeatability of these associations in many cases cannot be determined. The grouping of Indian and Bangladesh sites with CIANO could be explained by similarities in latitude (Table 1) and moisture stress patterns. The drought stress generated at CIANO under limited irrigation is similar to that experienced in many parts of South Asia.
At CIANO, late season severe terminal drought stress is generated by the application of a single preseeding irrigation. This screening method is designed to mimic the terminal moisture stress experienced in the South Asian region following the monsoon. In these regions farmers plant after the monsoonal rains on a receding moisture profile: very little if any rain falls after sowing. The slightly higher latitude and altitude of most Pakistani sites (between 4 and 7 [degrees] farther north and between 79 and 1562 m higher) may explain the reduced level of association between these areas and CIANO. The site most similar in latitude and altitude, Nepal, clustered only 25% of the time with CIANO, however, pattern analysis based on repeated sites outlined in the previous section does confirm a relationship between CIANO and Nepal.
Not surprisingly, only two other Mexican locations clustered with CIANO out of a total of seven possible comparisons. Apart from CIANO, all these sites are located at 2249 to 2640 m above sea level, compared with CIANO's altitude of 38 m, and are at least 8 [degrees] latitude closer to the equator. Two higher latitude sites, Portugal (38 [degrees] N) and Ukraine (46 [degrees] N) also clustered with CIANO.
The low level of grouping between CIANO and Southern Cone locations in Brazil and Argentina (Table 2) can be explained by different prevailing moisture conditions. These sites experienced preanthesis drought stress throughout the duration of the study. Therefore it is not surprising that SAWYT genotypes, developed under moderate to severe terminal moisture stress, differentiated differently for yield in the Southern Cone. This is borne out by the much stronger association between CIANO and India and Bangladesh (Table 2). Sites in India and Bangladesh under limited irrigation generally apply all the available water prior to anthesis. However, the stress generated at CIANO did not associate well with West Asian and North African sites. A possible explanation is that many of the sites in these regions are generally cooler than CIANO and genotype ranking may be influenced by the longer growing season.
Association Between the Same Location Sown to the Same Genotypes in Different Years
A small number of genotypes, ranging from 5 to 10, were in common between years in comparisons between specific SAWYT trials (Table 4). Dendrograms (not shown) developed from SHMM cluster analysis indicated that CIANO, CIMMYT's primary drought testing location, clustered with itself on 5 of 7 occasions at the third fusion level. Sites in Nepal (3/6) and Canada (3/4) also indicated a relatively high degree of association between years. However, Bolivian, Argentinean, and Pakistani locations did not cluster to a significant degree with themselves in the paired comparisons among different SAWYTs.
Table 4. The number of times CIANO and other locations sown to different SAWYTs in different years cluster with themselves at the third fusion level in the SHMM analysis based on genotypes common to each comparison. Number of groupings/Total number of possible groupings SAWYT compa- Argen- Argen- rison Nepal Bolivia tina tina Pakistan ([dag- Mexico Bhai- San Canada M. Perga- Dera ger]) CIANO rahwa Benito Kernen Juarez mino Ismall 1 v 2 -- 0/1 0/1 -- -- -- -- 2 v 3 -- 0/1 0/1 1/1 -- -- 1/1 3 v 4 1/1 0/1 0/1 -- 0/1 -- 0/1 4 v 5 0/1 1/1 1/1 -- 1/1 -- 1/1 5 v 6 1/1 -- 0/1 1/1 0/1 -- 0/1 1 v 3 1/1 1/1 0/1 -- 0/1 0/1 -- 3 v 5 1/1 1/1 0/1 1/1 0/1 -- 0/1 3 v 6 0/1 -- 0/1 0/1 1/1 0/1 0/1 4 v 6 1/1 -- 0/1 -- 0/1 -- 0/1 Total 5/7 3/6 1/9 3/4 2/7 0/2 2/7 ([dagger]) Genotypes in common for each comparison ranged from 5 to 10.
The high degree of association between years of CIANO indicates this location has high repeatability in discriminating germplasm. However, the weak association between the Bolivian, Argentinean, and Pakistani locations reflects the inherent variability characteristic of most rainfed, drought prone environments. Disease was not a major factor influencing genotypic ranking and subsequent location clustering as only 11 sites out of 122 included in the analysis reported significant disease incidence (Table 1). The high degree of association observed between years for CIANO reflects the controlled irrigation conditions under which materials are grown.
As suggested by Osman et al. (1997), the SHMM analysis can be applied routinely to identify subsets of locations without COI and thus find key locations. Furthermore, results obtained by yearly SHMM analyses can always be confirmed by pattern analysis on long-term multilocation trial data. The primary aim of our study was to examine the relevance of selecting germplasm under controlled irrigation at CIANO, Mexico, compared with performance under global drought stressed environments. The relatively low degree of repeatability of these locations, heavily influenced by rainfall and other seasonal factors, make it difficult to find high levels of association between CIANO and many global locations year to year. However, the application of SHMM and pattern analyzes indicated that CIANO is similar to sites in India and Bangladesh. These are not typically rainfed areas and crops are frequently drought stressed through lack of irrigation water.
The gravity-fed residual moisture stress generated at CIANO may need to be modified to improve the relevance of materials selected in Mexico to locations in the Southern Cone, West Asia, and Africa. Generation of a combination of both terminal and preanthesis stress scenarios may increase the frequency of elite materials adapted to these regions. The secondary aim of our study was to examine relationships among international testing locations with a view to identifying key locations for drought screening. Selection using a combination of environments such as CIANO, South Africa, and Argentina, all of which correlated well across many different environments, may provide the platform for greater rates of progress in breeding for dry environments globally.
Abdalla, O.S., J. Crossa, E. Autrique, and I.H. DeLacy. 1996. Relationships among International testing sites of spring durum wheat. Crop Sci. 36:33-40.
Barreto, H.J., G.O. Edmeades, S.C. Chapman, and J. Crossa. 1997. The alpha lattice design in plant breeding and agronomy: Generation and analysis, p. 544-551. In G.O. Edmeades et al. (ed.) Developing drought- and low N-tolerant maize. Mexico, D.F., Mexico.
Calhoun, D.S., G. Gebeyehu, A. Miranda, S. Rajaram, and M. van Ginkel. 1994. Choosing evaluation environments to increase wheat grain yield under drought. Crop Sci. 34:673-678.
Cornelius, P.L., J. Crossa, and M.S. Seyedsadr. 1996. Statistical tests and estimators for multiplicative models for cultivar trials, p. 199-234. In M.S. Kang et al. (ed.) Genotype-by-environment interaction. CRC Press, Boca Raton, FL.
Cornelius, P.L., M.S. Seyedsadr, and J. Crossa. 1992. Using the shifted multiplicative model to search for "separability" in crop cultivar trials. Theor. Appl. Genet. 84:161-172.
Cornelius, P.L., D.A.van Sanford, and M.S. Seyedsadr. 1993. Clustering cultivars into groups without rank-change interactions. Crop Sci. 33:1193-1200.
Crossa, J., and P.L. Cornelius. 1993. Recent developments in multiplicative models for cultivar trials, p. 571-577. In D.R. Buxton et al. (ed.) International crop science I. CSSA, Madison, WI.
Crossa, J., and P.L. Cornelius. 1997. Site regression and shifted multiplicative models clustering of cultivar trials sites under heterogeneity of error variances. Crop Sci. 37:406-415.
Crossa, J., P.L. Cornelius, K. Sayre, and J.I. Ortiz-Monasterio. 1995. A shifted multiplicative model fusion method, for grouping environments without cultivar rank change. Crop Sci. 35:54-62.
Crossa, J., P.L. Cornelius, and M.S. Seyedsadr. 1996. Using the shifted multiplicative model cluster methods for crossover genotype-by-environment interaction, p. 175-198. In M.S. Kang et al. (ed.)
Genotype-by-environment interaction. CRC Press, Boca Raton, FL.
Crossa, J., P.L. Cornelius, M. Seyedsadr, and P. Byrne. 1993. A shifted multiplicative model cluster analysis for grouping environments without genotypic rank change. Theor. Appl. Genet. 85:577-586.
DeLacy, I.H., and M. Cooper. 1990. Pattern analysis for the analysis of regional variety trials, p. 301-334. In M.S. Kang (ed.) Genotype-by-environment interaction in plant breeding. Louisiana State Univ., Baton Rouge, LA.
DeLacy, I.H., P.N. Fox, J.D. Corbett, J. Crossa, S. Rajaram, R.A. Fischer, and M. van Ginkel. 1994. Long-term association of locations for testing spring bread wheat. Euphytica 72:95-106.
DeLacy, I.H., and P. Lawrence. 1988. Combining pattern analysis over years--Classification of locations, p. 175-176. In K.S. McWhirter et al. (ed.) Proc. of the Ninth Australian Plant Breeding Conference. Agricultural Research Institute, Wagga Wagga, New South Wales, Australia.
Fox, P.N., A.A. Rosielle, and W.J.R. Boyd. 1985. The nature of genotype x environment interactions for wheat yield in Western Australia. Field Crops Res. 11:387-398.
Fox, P.N., B. Skovmand, B.K.Thompson, H.J. Braun, and R. Cormier. 1990. Yield and adaptation of hexaploid spring triticale. Euphytica 47:57-64.
Osman, S.A., J. Crossa, and P.L. Cornelius. 1997. Results and biological interpretation of shifted multiplicative model clustering of durum wheat cultivars and testing sites. Crop Sci. 37:88-97.
Peterson, C.J., and W.H. Pfeiffer. 1989. International winter wheat evaluation: Relationships among test sites based on cultivar performance. Crop Sci. 29:276-282.
Rajaram, S., M. van Ginkel, and R.A. Fischer. 1994. CIMMYT's wheat breeding mega-environments (ME). p. 1101-1106. In Proceedings of the 8th International Wheat Genetics Symposium, Beijing, China.
Seyedsadr, M., and P.L. Cornelius. 1992. Shifted multiplicative model for nonadditive two-way tables. Comm. Stat. B. Simulation and Computation 21:807-822.
SAS. 1988. SAS Institute Inc., Cary, NC.
Abbreviations: CIMMYT, International Maize and Wheat Improvement Center; SAWYT, Semi Arid Wheat Yield Trial; CIANO, Centro de Investigaciones Agricolas del Noroeste; GEI, genotype x environment interaction; COI, crossover interaction; SHMM, shifted multiplicative model; SREG, site regression model; SED, squared Euclidean distances.
Richard M. Trethowan,(*) Jose Crossa, Maarten van Ginkel, and Sanjaya Rajaram
Wheat Program, International Maize and Wheat Improvement Center (CIMMYT) Apdo. Postal 6-641, 06600 Mexico DF, Mexico; J. Crossa, Biometrics and Statistics Unit, CIMMYT. Received 7 Sept. 2000.
(*) Corresponding author (email@example.com).
|Printer friendly Cite/link Email Feedback|
|Author:||Trethowan, Richard M.; Crossa, Jose; van Ginkel, Maarten; Rajaram, Sanjaya|
|Date:||Sep 1, 2001|
|Previous Article:||Latitudinal and Longitudinal Adaptation of Smooth Bromegrass Populations.|
|Next Article:||Predicting Progeny Means and Variances of Winter Wheat Crosses from Phenotypic Values of Their Parents.|