Crop Science
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 31 May 2007
Published in Crop Sci 47:1071-1081 (2007)
© 2007 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Agricola
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Related Collections
Right arrow Biometrics
Right arrow Plant and Environment Interactions

CROP BREEDING & GENETICS

Associations of Environments in South Asia Based on Spot Blotch Disease of Wheat Caused by Cochliobolus sativus

A. K. Joshia,*, G. Ortiz-Ferrarab, J. Crossac, G. Singhd, G. Alvaradoc, M. R. Bhattae, E. Duveillerb, R. C. Sharmaf, D. B. Panditg, A. B. Siddiqueh, S. Y. Dasi, R. N. Sharmaj and R. Chandk

a Dep. of Genetics and Plant Breeding, Institute of Agricultural Sciences, Banaras Hindu Univ., Varanasi 221005, India
b CIMMYT South Asia, Regional Office, P.O. Box 5186, Kathmandu, Nepal
c International Maize and Wheat Improvement Center (CIMMYT), Apdo. Postal 6-641, C.P. 06600, D.F. Mexico
d Directorate of Wheat Research, Karnal, Haryana, India
e National Agricultural Research Council, Bhairahawa, Nepal
f Institute of Agriculture and Animal Science, Rampur, Nepal
g Wheat Research Centre, BARI, Dinajpur, Bangladesh
h Regional Agricultural Research Station, Jessore, Bangladesh
i Assam Agriculture Univ., Shillongani, Assam, India
j Bihar Agriculture College, Sabour, Bihar, India
k Dep. of Mycology and Plant Pathology, Institute of Agricultural Sciences, Banaras Hindu Univ., Varanasi 221005, India

* Corresponding author (joshi_vns{at}yahoo.co.in).


   ABSTRACT
TOP
 NOTES
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
Spot blotch is an important disease of wheat (Triticum aestivum L.) in South Asia. Division of test sites for this disease into homogenous subregions is expected to contribute to more efficient evaluation and better differentiation of cultivars. Data from a collaborative regional program of South Asia conducted by CIMMYT were analyzed to group testing sites into relatively homogenous subregions for spot blotch area under the disease progress curve (AUDPC). Five-year data of eight locations from Eastern Gangetic Plains Nursery (EGPSN) and five locations of the Eastern Gangetic Plains Yield Trial (EGPYT) conducted in three countries (India, Nepal, and Bangladesh) of South Asia were used. A hierarchical cluster analysis was used to group locations on the basis of genotype x location interaction effects for spot blotch AUDPC. Cluster analysis divided South Asia into two broad regions and four subregions. This classification was not entirely consistent with the geographic distribution of locations, but clusters mostly followed general geographic-climatic locations. The locations Varanasi (India) and Bhairahawa (Nepal) were identified as the most suitable sites for evaluation of spot blotch, followed by Rampur (Nepal). The major determinant for the clustering was mean temperature. The results suggest that the major wheat region of South Asia can be divided into subregions, which may reduce the cost of resistance evaluation and aid in developing wheat with resistance to this disease.

Abbreviations: AUDPC, area under disease progress curve • CIMMYT, International Maize and Wheat Improvement Centre • EGP, eastern Gangetic plains • EGPSN, Eastern Gangetic Plains Nursery • EGPYT, Eastern Gangetic Plains Yield Trial • G x E, genotype x environment • G x L, genotype x location • METs, multi-environment trials • NWRP, National Wheat Research Program • SED, squared Euclidean distance • SREG, sites regression model.


   INTRODUCTION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
THE IMPORTANCE OF sustained increases in wheat production and productivity for food security is well recognized in South Asia, where wheat is a major staple crop and human population is increasing. After significant expansion during the initial years of the Green Revolution, the wheat production area in the region appears to have stabilized in the last decade (Evenson et al., 1999; Joshi et al., 2006) and is expected to remain at current levels in the coming decades. Adaptable and higher-yielding genotypes will play a key role in meeting the regional demand for grain, and in addition to an increased yield potential, such genotypes will require resistance or tolerance to diverse biotic and abiotic stresses. In the case of crop diseases, resistant cultivars offer the most economical and environmentally safe means for control.

Many biotic stresses constrain wheat productivity in the eastern regions of South Asia (Joshi et al., 2006), but spot blotch of wheat caused by Cochliobolus sativus (Ito and Kurib.) Drechsler ex Dastur [anamorph: Bipolaris sorokiniana (Sacc.) Shoem.] is considered the most important disease (Saari, 1998; Joshi and Chand, 2002; Joshi et al., 2002, 2004a, 2004b; Arun et al., 2003; Sharma et al., 2004; Pandey et al., 2005; Duveiller et al., 2005a). Spot blotch causes substantial economic loss to wheat growers in the region (Saari, 1998; Sharma and Duveiller, 2004), which can be especially devastating for growers in the eastern Gangetic plains, who frequently have small holdings with little land or profitability (Joshi et al., 2007b). Spot blotch is of high importance in the vast eastern Gangetic plains (EGP) of South Asia, including parts of India, Nepal, and Bangladesh, environments in this region vary greatly, however, in moisture supply, temperature, soil type, and abiotic stresses (Joshi et al., 2007a). Under such conditions, genotype x location (G x L) interaction is expected to be large and may not permit differentiation of the performance of genotypes across environments.

The subdivision of regional trials could reduce G x L interactions, thereby allowing better genotype differentiation and reducing costs (Collaku et al., 2002). This is particularly important for factors such as the severity of spot blotch, which is highly influenced by environmental conditions, especially temperature and humidity (Chaurasia et al., 1999; Chand et al., 2003; Duveiller et al., 2005a; Pandey et al., 2005). Proper characterization and association of locations are very important for screening superior breeding lines capable of reducing harvest losses to the disease. The importance of using suitable sites to evaluate germplasm for particular traits is well recognized. Many studies have demonstrated the use of cluster analysis to classify locations on the basis of a single trait (Campbell and Lafever, 1980; Ghaderi et al., 1980; Fox and Rosielle, 1982; Baenziger et al., 1985; Yau et al., 1991; Collaku, 1991; Van Oosterom et al., 1993; Abdalla et al., 1996; Trethowan et al., 2001). Characterization of locations is considered more important in breeding for disease resistance than most other traits, including yield traits especially for data from large and heterogeneous areas where wheat is grown and tested (Campbell and Lafever, 1980; Ghaderi et al., 1980; Fox and Rosielle, 1982; Yau et al., 1991).

Many tools and techniques have been suggested for characterizing and grouping environments, with pattern and biplot analysis considered the most valuable. Horner and Frey (1957) divided oat (Avena sativa L.) test areas into subareas within which the G x L component of variance was substantially reduced. Because genotype responses are measured in different environments, multivariate approaches are usually effective in explaining genotype x environment (G x E) or G x L interactions (Lin et al., 1986; Zobel et al., 1988; Nachit et al., 1992). Among multivariate techniques, cluster analysis based on differences in the response of genotypes across environments is the most widely used. Abou-El-Fittouh et al. (1969) used this technique to classify cotton (Gossypium hirsutum L.) test sites. A number of studies have been conducted in wheat to classify locations using cluster analysis (Campbell and Lafever, 1980; Ghaderi et al., 1980; Fox and Rosielle, 1982; Collaku et al., 2002).

DeLacy et al. (1994) used pattern analysis to study mega-environments in bread wheat comprising different regions and subregions of the world on the basis of 26 yr of grain yield data of the wheat program of CIMMYT. Likewise, DeLacy et al. (1996) described pattern analysis for the analysis of multi-environment trials. Yau et al. (1991) used a hierarchical agglomerative and polythetic clustering technique to analyze ICARDA/CIMMYT Regional Bread Wheat Yield Trial data. Van Oosterom et al. (1993) used cluster analysis to study relationships among barley (Hordeum vulgare L.) environments in the Mediterranean region. Peterson and Pfeiffer (1989) and Peterson (1992) used principal factor analysis to describe wheat location relationships and determine specific production zones for wheat cultivars. Hanson (1994) developed distance statistics based on the concept of genotypic stability to interpret regional soybean [Glycine max (L.) Merr.] tests. Braun et al. (1992) determined locations that correlated most strongly with genotypic means across locations in CIMMYT's International spring wheat yield nursery (ISWYN).

In the analyses used by DeLacy and Lawrence (1988), Peterson and Pfeiffer (1989), Peterson (1992), and Braun et al. (1992), it was assumed that phenotypic correlation for yields of trial genotypes among locations was a measure of similarity of these locations for breeding purposes. Whereas Peterson and Pfeiffer (1989) and Peterson (1992) pooled such correlations across years and applied factor analysis to simplify the resultant long-term matrix, DeLacy and Lawrence (1988) pooled squared Euclidean distances (SEDs) among locations across years and presented them using cluster analysis. Abdalla et al. (1996) studied location relationships from 5 yr of CIMMYT international durum wheat trials using pattern analysis and found that previously defined mega-environment classification of locations needed further investigation.

Interpretation of performance of a number of genotypes in a broad range of environments is generally affected by large G x E interactions (Gauch and Zobel, 1996). Analysis of variance describes only the main effects; it tests the significance of the G x E interaction but provides no insight into the particular patterns of genotypes or environments that give rise to the G x E interaction. Multiplicative models for multi-environment trials have been used for studying G x E interaction, examining genotypic yield stability and adaptation, and developing methods for clustering sites or cultivars into groups with statistically negligible crossover G x E interaction (Crossa and Cornelius, 1997; Crossa et al., 2002). Multiplicative models have an additive (linear) component (i.e., intercept, main effects of sites and/or genotypes) and a multiplicative (bilinear) component (G x E interaction) and, thus, are also named linear-bilinear models. A type of linear-bilinear model suitable for grouping sites and cultivars without cultivar rank change is the sites regression model (SREG). This model is also named GGE (Yan et al., 2001) because it includes the effects of genotype plus G x E interaction. Biplots obtained from graphing the first two components of the multiplicative part of SREG (genotype plus G x E interaction) are useful for summarizing and approximating patterns of response that exist in the original data (Gabriel, 1971, 1978). Crossa et al. (2002) showed that, for SREG models, the biplot of the first two multiplicative components graphs the interaction variation due to noncrossover genotype plus G x E interaction versus the interaction variation due to crossover genotype plus G x E interaction explainable by a second bilinear term if and only if the primary effects of sites are all of the same sign. They further demonstrated how the SREG biplots could be used for identifying subsets of sites and also for genotypes with noncrossover G x E interaction.

Few attempts have been made to classify and characterize international South Asian locations for spot blotch genotypic response. In this study, we have examined 5 yr of spot blotch data for two types of multi-environment trials (METs) and used pattern analysis and the SREG to (i) evaluate the magnitude and nature of genotype, location, and G x L interaction effects for severity to spot blotch disease in the South Asian locations; (ii) study relationships among locations of South Asia for spot blotch evaluation; and (iii) group locations that represent similar selection environments.


   MATERIALS AND METHODS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS
 DISCUSSION
 REFERENCES
 
Experimental Data
Five-year data from two regional, multi-environment trials targeting India, Nepal, and Bangladesh were used: the Eastern Gangetic Plains Screening Nursery (EGPSN) and the Eastern Gangetic Plains Yield Trial (EGPYT), developed jointly by the National Wheat Research Program (NWRP) of Nepal and the regional office of CIMMYT in Kathmandu, Nepal, as a regional effort of CIMMYT to strengthen identification and dissemination of adaptable genotypes in South Asia.

The EGPSN Nursery Multi-Environment Trial
The EGPSN consisted of 150 entries with seven checks, viz., Bhrikuti (an improved check from Nepal), Sonalika or RR-21 (a long-term check from India, spot blotch susceptible), Kanchan (an improved check from Bangladesh), Chriya-3 (a spot blotch–resistant check from CIMMYT, Mexico), PBW 343 (an improved check from India), Achyut (an improved check from Nepal), and a local check of each location. The 143 lines under evaluation changed from year to year, but the seven checks were common in all 5 yr of evaluation. The lines used in the EGPSN were either selected from CIMMYT breeding material or derived using diverse genetic backgrounds for their suitability to the warm, humid environments of South Asia. Breeding programs from Bangladesh, Nepal, and India contributed elite wheat lines developed and/or identified by their own programs. The EGPSN was grown at eight locations from 2000 to 2005: four (Varanasi, Karnal, Shillongani, and Sabour) in India, two in Bangladesh (Dinajpur and Jessore), and two in Nepal (Rampur and Bhairahawa) (Table 1).


View this table:
[in this window]
[in a new window]

  		Table 1. Geographical distribution and meteorological data of eight locations of South Asia used in planting Eastern Gangetic Plains Nursery and Eastern Gangetic Plains Yield Trial during five years of testing.

 
All locations were favorable for spot blotch development and, except for Karnal (India), were characterized by high temperatures and relative humidity. For each EGPSN line, 10 g of seed was used to plant two rows 2 m long with row spacing of 25 cm. Each line was hand-sown in an unreplicated trial. Identical fertilization practices (120 kg N–60 kg P2O5–40 kg K2O ha–1) were followed at all locations in all years. Full doses of K2O and P2O5 were applied at sowing; N was supplied in split applications, with 60 kg N ha–1 at sowing, 30 kg N ha–1 at the first irrigation (21 d after sowing), and 30 kg N ha–1 at the second irrigation (45 d after sowing).

The EGPYT Multi-Environment Trial
Entries in EGPYT were 21 superior EGPSN lines plus four common checks from the EGPSN: Bhrikuti, Sonalika, Kanchan, and PBW 343. The EGPYT was grown at five locations in South Asia during 5 yr (2000–2005). The five locations used in EGPYT were also used in the EGPSN: one in India (Varanasi), two in Nepal (Bhairahawa and Rampur), and two in Bangladesh (Jessore and Dinajpur). Trial design was a 5 by 5 square lattice (alpha lattice) with two replications. Only the four checks were retained for all five locations and years. For each entry, there were six envelopes with approximately 60 g of seed to allow cooperators to plant six rows, 3.0 m long and 1.5 m wide, with a row spacing of about 25 cm.

Area under the Disease Progress Curve
Observations were recorded for spot blotch severity on three different dates, and area under disease progress curve (AUDPC) was determined. To record spot blotch infection levels, we used the double digit (DD, 00–99), modified Saari and Prescott severity scale for foliar diseases in wheat (Saari and Prescott, 1975; Eyal et al., 1987). The first digit (D1) indicates vertical disease progress on the plant and the second digit (D2) indicates severity measured in diseased leaf area. For example, for a score of 59, the 5 represents the height (from ground up) on the diseased plant and 9 represents the average severity up to that level, as percentage of leaf area infected (i.e., 90% spot blotch lesions on the leaves). For each score, the disease severity percentage was based on the following formula:

Formula 1[1]
The AUDPC was calculated using the percent severity estimations corresponding to the disease ratings, as outlined by Shaner and Finney (1977) and Roelfs et al. (1992):

Formula 2[2]
where, Yi = disease level at time ti; ti + 1 – ti = time (days) between two disease scores; and n = number of dates on which spot blotch was recorded. For proper comparison, AUDPC values were standardized by maturity duration of genotypes recorded at each of the locations to make it AUDPC percent days (Reynolds and Neher, 1997).

Statistical Analyses
Pattern Analysis for Associations among EGPSN and EGPYT Locations
Pattern analysis is the approach that jointly uses classification (cluster analysis) and ordination (principal coordinate analysis) to study long-term relationships among locations (DeLacy et al., 1996). In this study, pattern analysis was used on location-standardized data obtained by subtracting the location mean from each value and then dividing by the location standard deviation.

In pattern analysis, locations are evaluated based on their ability to discriminate among genotypes, such that locations that rank genotypes similarly are more similar than locations that rank genotypes differently. Dissimilarity among locations is measured by the SED and used in classification, whereas similarity is used in ordination. For each year and across years, SEDs were calculated to produce a complete location-by-location proximity matrix. Cluster analysis using a weighted average SED across 5 yr was employed to classify, on the basis of AUDPC, the locations into more homogeneous groups. Pattern analyses for EGPSN and EGPYT were performed to detect possible repeatability for some location clusters in both analyses.

A hierarchical cluster analysis using Ward's algorithm (Ward, 1963) was performed using SAS Proc CLUSTER and TREE (SAS Institute, 2003). The hierarchical clustering was truncated at the stage corresponding to the initial sharp decline of R2 (where R2 = squared multiple correlation, which is the sum of squares between all the clusters divided by the corrected total sum of squares).

Ordination using the similarity matrix obtained from the distance matrix was conducted using principal coordinates analysis. The plot of the first two principal coordinate axes should show proximities on the locations similar to those found in the dendrogram obtained from cluster analysis.

Site Regression Model and Its Biplot to Study Genotype Plus G x E in EGPSN and EGPYT
The SREG was used to determine whether the locations clustered in the pattern analyses also tended to cluster in SREG biplot across years. However, the SREG was fitted to only the few genotypes that were common at all locations and in both trials.

We used the SREG for the response of the seven common EGPSN genotypes on the combination of the eight sites and 5 yr for spot blotch AUDPC percent days. The combination of sites and years generated 40 environments that we named by the first two letters of the site followed by a number identifying the name of the nursery. The environments for years 1 to 5 were EGPSN4, EGPSN5, EGPSN6, EGPSN7, and EGPSN8. Thus, the names combining the location and year of testing were, for example, Bh4 = site Bhairhawa for EGPSN4; Di4 = Dinajpur for EGPSN4; Je4 = Jessore for EGPSN4; Ka4 = Karnal for EGPSN4; Ra4 = Rampur for EGPSN4; Sh4 = Shillongani for EGPSN4; Va4 = Varanasi for EGPSN4. Similarly, names for other four EGPSN nurseries were assigned.

A similar approach was followed for 25 environments of EGPYT that resulted from the combination of five locations and 5 yr. The EGPYT trials used during 5 yr (2000–2005) of testing were EGPYT2, EGPYT3, EGPYT4, EGPYT5, and EGPYT6. Hence, the names for first year of testing were: Bh2 = site Bhairahawa for EGPYT2; Di4 = Dinajpur for EGPYT2; Je2 = Jessore for EGPYT2; Ra2 = Rampur for EGPYT2; Va2 = Varanasi for EGPYT2. Similarly, names for other four EGPYT trials were given. Furthermore, stability analysis using the SREG model was performed combining the five sites and 5 yr into 25 environments as described for EGPSN and using four checks common to all environments.

The SREG is given by

Formula 3[3]
where, ij. is the mean of the ith cultivar in the jth environment for g cultivars and e sites (i = 1, 2, ..., g and j = 1, 2, ..., e); µj is the site mean; {lambda}k({lambda}1 ≥ {lambda}2 ≥ ... {lambda}t) are scaling constants (singular values) that allow the imposition of orthonormality constraints on the singular vectors for cultivars, {alpha}k = ({alpha}1k, ..., {alpha}gk)' and sites, {gamma}k = ({gamma}1k, ..., {gamma}ek)', such that

Formula 4[4]
and

Formula 5[5]
for k != k'; {alpha}ik and {gamma}jk, for k = 1, 2, 3, ..., are called primary, secondary, tertiary, etc. effects of ith cultivar and the jth site, respectively; Formula 5ij is the residual error assumed to be normally and independent distributed with 0 means and variance {sigma}2/r (where {sigma}2 is the pooled error variance and r is the number of replicates). The number of bilinear terms is t ≤ min (g, e). Estimates of the multiplicative parameters in the kth bilinear term are obtained as the kth component of the deviations from the additive part of the model. In the SREG model, only the main effects of cultivars plus the G x E interaction are absorbed into the bilinear terms.

The biplots of the SREG had the primary and secondary effects of genotypes and locations plotted. Useful conclusions can be drawn from the biplot about relationships among locations, genotypes, and G x L interaction. For example, locations located in the same direction of the biplot equally discriminate genotypes, whereas locations in the opposite direction ranked the genotypes differently. In this study, we were only interested in the distribution of the locations in each year and how they clustered with other locations. Crossa et al. (2002) pointed out that if the primary effects of sites were all of the same sign, the first component in biplots of SREG would be related to noncrossover genotype plus G x E interaction variability, whereas the second component accounted for crossover genotype plus G x E interaction variability, such that the ideal test location should have a large first primary effect and a near-zero secondary effect (Yan et al., 2001). Thus, to simplify interpretation of the biplot, we have not included the genotypes and concentrate only on the distribution of the locations in the SREG biplot.

Pattern analysis was performed on location-standardized data. Thus its proximity plot should produce a similar location pattern as those shown by the SREG biplot.


   RESULTS
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
DISCUSSION
 REFERENCES
 
Genotypic effect was significant for AUDPC in both the EGPSN and EGPYT tested across locations and years (data not shown). Genotypes accounted for 30% of total variability for AUDPC percent days of spot blotch in the EGPSN, and 34% in the EGPYT. Environmental components (both years and locations) were highly significant (P < 0.01) and had a strong effect on AUDPC. Years accounted for 21 and 28% of the total variability in the EGPSN and EGPYT, respectively (data not shown). Genotype x environment interaction was highly significant in both METs.

Pattern Analyses
In the EGPSN, cluster analysis across years of pooled SEDs among the eight locations based on AUDPC indicated two broad groups (Fig. 1a): (I) Bhairahawa, Varanasi, Rampur, and Shillongani, and (II) Dinajpur, Jessore, Karnal, and Sabour. However, these two groups could be subdivided into four groups of locations (Fig. 1a): (I) Bhairahawa, Varanasi, Rampur; (II) Shillongani by itself; (III) Dinajpur, Jessore, Karnal; and (IV) Sabour by itself. In the other trial (EGPYT) (Fig. 1b), where data of only five locations were available, clustering groups appeared to be two: (I) Bhairahawa, Varanasi, Rampur, and Dinajpur; and (II) Jessore by itself. However, the EGPYT locations could also be subdivided into (I) Bhairahawa, Varanasi, and Rampur; (II) Dinajpur by itself; and (III) Jessore by itself. The two locations of Cluster I (Bhairahawa in Nepal and Varanasi in India) were consistent in both EGPSN and EGPYT (Fig. 1a, 1b). The other location (Rampur in Nepal) also appeared to cluster with Bhairahawa and Varanasi, but was slightly away from it (Fig. 1a, 1b).


Figure 1A
View larger version (12K):
[in this window]
[in a new window]

  		Figure 1a. Dendrogram from the classification of eight locations in EGPSN over five years (2000–2001 to 2004–2005) of testing in South Asia for AUDPC of spot blotch disease of wheat using weighted environment standardized squared Euclidean distance as the dissimilarity measure. A hierarchical agglomerative classification procedure using squared Euclidean distance as the dissimilarity measure and incremental sum of squares as the clustering strategy was used.

 

Figure 1B
View larger version (10K):
[in this window]
[in a new window]

  		Figure 1b. Dendrogram from the classification of five locations in Eastern Gangetic Plains Yield Trial (EGPYT), respectively over five years of testing (2000–2001 to 2004–2005) in South Asia for AUDPC of spot blotch disease of wheat using weighted environment standardized squared Euclidean distance as the dissimilarity measure. A hierarchical agglomerative classification procedure using squared Euclidean distance as the dissimilarity measure and incremental sum of squares as the clustering strategy was used.

 
In the EGPSN and EGPYT, the distribution of the two locations (Bhairahawa and Varanasi) within Cluster I corresponded to their geographic and climatic characteristics (Fig. 2). The location of Rampur is quite close to these two locations and, thus, logically appeared in the same cluster. The mean latitude for the two locations of Cluster I, Bhairahawa and Varanasi, was 25°30' N lat and 25°18' N lat, respectively (Table 1). These two locations are also at the same longitude (83°E long). Bhairahawa and Varanasi are known for rice–wheat cropping systems with same type of soil and climate (Tables 1 and 2). However, other locations (Sabour and Dinajpur), falling under similar latitude (25°15' N lat and 25°38' N lat, respectively) (Table 1), performed very differently and have high rainfall environments. Likewise, two locations, Jessore (Bangladesh) and Karnal (India), belonging to different latitudes (23°11' N lat and 29°43' N lat) (Table 1) performed very similarly (Cluster III in EGPSN), demonstrating that latitude was not the only factor influencing the division of locations for AUDPC.


Figure 2
View larger version (31K):
[in this window]
[in a new window]

  		Figure 2. Eight locations of three countries in South Asia Region used in the pattern analysis for AUDPC of spot blotch of wheat.

 

View this table:
[in this window]
[in a new window]

  		Table 2. Mean temperature of three months of crop growth during five years (2000–2001 to 2004–2005) of testing in eight locations of Eastern Gangetic Plains Nursery in South Asia.

 
In the ordination (principal coordinate analysis) part of pattern analysis of the EGPSN, the first three principal axes explained 41, 36, and 14% of the genotype plus G x L interaction, respectively, whereas in EGPYT those axes accounted for 32, 28, and 23% of the G x L interaction, respectively (Table 3). Figure 3a displays the distribution of all eight locations of EGPSN based on principal coordinates analyses. The first principal coordinate axis separates Bhairahawa, Varanasi, Rampur, and Shillongani from Dinajpur, Jessore, Karnal, and Sabour. The groups formed by the cluster analysis are apparently clear when an ordination technique was used and the locations are displayed in the plot of the principal coordinate analysis. The three environments (Bhairahawa, Varanasi, and Rampur) that were closest to the center of the plot contributed less to the total G x L interaction and therefore can be considered to discriminate among genotypes more similarly and less than other environments. The location Shillongani formed a cluster by itself and does not appear to be highly correlated with the previous three stable environments. Its angle with respect to the line depicting most stable environment is quite large, indicating that it discriminated differently among genotypes. Similarly, Karnal, Dinajpur, and Jessore formed a compact group and Sabour is farther and formed a cluster by itself.


View this table:
[in this window]
[in a new window]

  		Table 3. Scores on the first three principal vectors from the principal coordinate analysis of eight locations used for five years of testing of Eastern Gangetic Plains Nursery (EGPSN) and Eastern Gangetic Plains Yield Trial (EGPYT) in South Asia. The ordination was based on location-standardized grain yields using squared Euclidean distances averaged across years as the dissimilarity measure.

 

Figure 3A
View larger version (19K):
[in this window]
[in a new window]

  		Figure 3a. Proximity plot of the first and second principal coordinate axes of eight locations used in Eastern Gangetic Plains Nursery (EGPSN) over five years of testing using weighted environment standardized squared Euclidean distance as the dissimilarity measure. The four group level obtained from the location classification is superimposed on the ordination.

 
The principal coordinate analysis for EGPYT (Fig. 3b) also demonstrated the closeness of three locations: Bhairahawa, Varanasi, and Rampur for screening spot blotch severities of wheat germplasm lines, as opposed to Dinajpur and Jessore, located further apart in the proximity plots.


Figure 3B
View larger version (19K):
[in this window]
[in a new window]

  		Figure 3b. Proximity plot of the first and second principal coordinate axes of five locations used in Eastern Gangetic Plains Yield Trial (EGPYT) nursery over five years of testing using weighted environment standardized squared Euclidean distance as the dissimilarity measure. The three group level obtained from the location classification is superimposed on the ordination.

 
Site Regression Model
The SREG biplot for eight locations across 5 yr (40 environments) of EGPSN is depicted in Fig. 4a (for clarity genotypes are not included in the figure). Except for 1 yr in Varanasi and 2 yr in Bhairahawa, the other environments are located toward the right side of the biplot. The first SREG component explained 59% of the genotype plus G x E interaction, whereas the second component accounted for 14% of the variability. Results indicated that Varanasi, Bhairahawa, and Rampur were good testing locations because they had large values for primary effects and low values for secondary effects. Furthermore, within these three locations, Varanasi had the highest values for primary effects for two consecutive years, which would make it the best testing location among the three. On the other hand, and as expected from the cluster analysis, Sabour formed an isolated group for 4 yr in the left-hand side of the biplot and appeared to be a very different location. Karnal seemed to be a very unstable location for the expression of spot blotch disease, with all five points scattered from the left to right side of the biplots and not forming a clear pattern with Jessore, as suggested by the cluster dendrogram. A stable location demonstrates a large first primary effect (noncrossover G x E variability) and near zero secondary effect (crossover G x E variability) in the biplot (Crossa et al., 2002).


Figure 4A
View larger version (19K):
[in this window]
[in a new window]

  		Figure 4a. Site regression model (SREG) biplot of number of locations and years on the performance of six common check lines in 40 environments (eight locations and five years) of Eastern Gangetic Plains Nursery (EGPSN) of South Asia. Bh4, Bhairahawa EGPSN4; Bh5, Bhairahawa EGPSN5; Bh6, Bhairahawa EGPSN6; Bh7, Bhairahawa EGPSN7; Bh8, Bhairahawa EGPSN8; Di4, Dinjapur EGPSN4; Di5, Dinjapur EGPSN5; Di6, Dinjapur EGPSN6; Di7, Dinjapur EGPSN7; Di8, Dinjapur EGPSN8; Je4, Jessore EGPSN4; Je5, Jessore EGPSN5; Je6, Jessore EGPSN6; Je7, Jessore EGPSN7; Je8, Jessore EGPSN8; Ka4, Karnal EGPSN4; Ka5, Karnal EGPSN5; Ka6, Karnal EGPSN6; Ka7, Karnal EGPSN7; Ka8, Karnal EGPSN8; Ra4, Rampur EGPSN4; Ra5, Rampur EGPSN5; Ra6, Rampur EGPSN6; Ra7, Rampur EGPSN7; Ra8, Rampur EGPSN8; Sa4, Sabour EGPSN4; Sa5, Sabour EGPSN5; Sa6, Sabour EGPSN6; Sa7, Sabour EGPSN7; Sa8, Sabour EGPSN8; Sh4, Shillongani EGPSN4; Sh5, Shillongani EGPSN5; Sh6, Shillongani EGPSN6; Sh7, Shillongani EGPSN7; Sh8, Shillongani EGPSN8; Va4, Varanasi EGPSN4; Va5, Varanasi EGPSN5; Va6, Varanasi EGPSN6; Va7, Varanasi EGPSN7; Va8, Varanasi EGPSN8.

 
In summary, the SREG biplot for EGPSN (Fig. 4a) shows that Varanasi had disease development that was able to differentiate best among the genotypes. Bhairahawa and Rampur in Nepal also showed good performance with regard to discrimination among genotypes. Sabour and Karnal were not good locations. The two Bangladesh locations (Jessore and Dinajpur) were more stable than Karnal and Sabour, but disease development was less than in the three good locations (Varanasi, Bhairahawa, and Rampur). The SREG biplot (Fig. 4a) tended to discriminate locations more similarly than the proximity plot obtained from principal coordinate analysis (Fig. 3a). These similarities were expected because principal coordinate analysis is based on location-centered and location-standardized data, whereas SREG is based on location-centered data.

The SREG biplot for five locations across 5 yr (25 environments) of EGPYT (Fig. 4b) showed a much clearer response pattern than the previous biplot, with Varanasi emerging as the best site, followed by Rampur and Bhairahawa. The group of locations, Varanasi, Rampur, and Bhairahawa, were located toward the right of the biplots, forming fairly consistent subgroups for the 5 yr, whereas the years of Jessore and Dinajpur were located toward the left side of the biplots, indicating a different pattern of response. The SREG biplot of Fig. 4b discriminated among locations similar to the proximity plot obtained from principal coordinate analysis (Fig. 3b).


Figure 4B
View larger version (15K):
[in this window]
[in a new window]

  		Figure 4b. Site regression model (SREG) biplot of number of locations and years on the performance of four common check lines in 25 environments (five locations and five years) of Eastern Gangetic Plains Yield Trial (EGPYT) nursery of South Asia. Bh2, Bhairahawa EGPYT2; Bh3, Bhairahawa EGPYT3; Bh4, Bhairahawa EGPYT4; Bh5, Bhairahawa EGPYT5; Bh6, Bhairahawa EGPYT6; Di2, Dinjapur EGPYT2; Di3, Dinjapur EGPYT3; Di4, Dinjapur EGPYT4; Di5, Dinjapur EGPYT5; Di6, Dinjapur EGPYT6; Je2, Jessore EGPYT2; Je3, Jessore EGPYT3; Je4, Jessore EGPYT4; Je5, Jessore EGPYT5; Je6, Jessore EGPYT6; Ra2, Rampur EGPYT2; Ra3, Rampur EGPYT3; Ra4, Rampur EGPYT4; Ra5, Rampur EGPYT5; Ra6, Rampur EGPYT6; Va2, Varanasi EGPYT2; Va3, Varanasi EGPYT3; Va4, Varanasi EGPYT4; Va5, Varanasi EGPYT5; Va6, Varanasi EGPYT6.

 

   DISCUSSION
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
REFERENCES
 
The clustering of locations into different groups (Fig. 1a, 1b) for AUDPC of spot blotch disease implied that the eastern part of South Asia, where spot blotch is most prevalent (Duveiller et al., 2005a; Joshi et al., 2007c), can be divided into more homogeneous subregions for more efficient evaluation of resistance in the experimental lines. This result was expected because South Asia includes diverse moisture regimes, temperatures, soil types and micronutrient levels, abiotic stresses, and cropping systems. Effective selection of test sites with representative locations from each subregion should not only reduce the cost of evaluation but should also result in the identification of more dependable resistant sources.

Cluster analysis divided the eastern part of South Asia (India, Nepal, and Bangladesh) into subregions with similar locations (Fig. 2). This classification followed general geographic-climatic regions for at least three locations: Varanasi (India) and Bhairahawa and Rampur (Nepal). In an earlier study using only 3-yr (2000–2002) data with 17 entries from Helminthsosporium Monitoring Nursery, Duveiller et al. (2005b) also showed a close association between these three locations based on AUDPC data using the SREG biplot approach. Additionally, Varanasi was noted as one of the most stable locations across the EGP for expression of spot blotch disease in the genotypes tested.

Environmental variation due to weather conditions (mainly temperature) is often considered a major factor influencing spot blotch disease in wheat (Chaurasia et al., 1999, 2000; Pandey et al., 2005), and this seemed to be partially true in this study. However, disease severity and epidemic expression also depend on stress conditions and crop management parameters (Sharma and Duveiller, 2004; Pandey et al., 2005; Joshi et al., 2006), which may explain inconsistencies regarding geographical associations noted in the present study. Among the four clusters, Cluster I and Cluster IV in Fig. 1a, 1b were more distinct. Cluster I included major wheat-growing sites in the eastern Indo-Gangetic Plains, characterized by warmer temperatures and similar rainfall, humidity, soils, and cropping sequences (mainly rice–wheat). However, although the sites have relatively similar growing times, the duration of cropping season would be longer from east to west due to variations in temperature. Therefore, the sum of temperature might also be of significance in the standardization of AUDPC values (Reynolds and Neher, 1997) and could have contributed to the inconsistencies in geographical associations.

Many of the region's wheat cultivars have been quite similar. For about three decades—from the 1970s to the early 1990s—the region was dominated by Sonalika (RR 21), now regarded as highly susceptible to spot blotch. Presently, another variety (HUW 234) is grown on about two million hectares. The climates, soils, and cropping patterns for Varanasi, Bhairahawa, and Rampur are very similar and could explain why they are clustered for the expression of spot blotch development. The third and fourth clusters of locations observed in EGPSN (Fig. 1a) cannot be considered a complete subregion, because one location (Karnal, India) is in the western part of the region, whereas the other three (Sabour, India; Jessore and Dinajpur, Bangladesh) are in the eastern part. Although Karnal (India) and Jessore (Bangladesh) were found in one cluster (Fig. 1a) of EGPSN, the biplots (Fig. 4a) indicated Karnal as an unstable location in the EGPSN, with all five points scattered from the left to right sides of the biplots and not forming a clear pattern with Jessore.

In the EGPYT (Fig. 1b), the two locations (Jessore and Dinajpur) of Bangladesh separated into Clusters II and III. The locations Jessore, Dinajpur, and Shillongani were high-rainfall sites. Although high humidity is required for the development of spot blotch, water-logging in the previous crop season due to heavy monsoon rains in low-lying fields may cause anaerobic conditions detrimental to the survival of B. sorokiniana (Pandey et al., 2005). Thus, it appears that the low incidence of spot blotch in the heavy rainfall regions (Jessore, Dinajpur, and Shillongani) led to their nonclustering with Varanasi, Bhairahawa, and Rampur. The location Sabour is a part of the eastern Indo-Gangetic Plains of India, but it did not respond like Varanasi, Bhairahawa, and Rampur. This suggests that other factors might have contributed to Sabour's nonclustering with the three locations. In the case of Karnal, which falls under the Northwestern Plains Zone of India, spot blotch development could be low due to relatively cool temperatures.

Spot blotch severity on wheat is known to be higher under late seeding in the eastern Indo-Gangetic Plains, due to higher temperatures in the crucial postanthesis stage (Chaurasia et al., 1999; Joshi and Chand, 2002; Sharma and Duveiller, 2004). Chaurasia et al. (2000) reported that B. sorokiniana development was best in the postanthesis period when mean temperatures reached approximately 26°C. The two best locations (Varanasi and Bhairahawa) for spot blotch in this study displayed a mean maximum temperature in March of around 26°C. In March, almost all wheat in South Asia is in the grain-filling stage. Ideal temperatures and humidity, along with abundant inoculum in the soil, lead to a high incidence of spot blotch. Other locations (such as Karnal, India) either lack the optimum temperatures or do not possess sufficient inoculum in the soil. The temperatures of Jessore, Dinajpur, and Sabour appeared to be higher than those of sites of Cluster I. In the EGPSN, the first principal coordinate vector was significantly correlated with mean minimum temperature r = 0.81 (P ≤ 0.01). The individual correlations with mean minimum temperatures of the 3 months, January (r = 0.81), February (r = 0.82), and March (r = 0.75), were highly significant (P ≤ 0.01). However, the mean maximum temperature was significant (P ≤ 0.01) only for January (r = 0.58). Similarly, in the EGPYT, the first vector was significantly correlated with mean maximum temperature in January, r = 0.52 (P ≤ 0.05). Thus, it appeared that mean temperature in early postanthesis stages, which aggravates spot blotch development, played a significant role in the clustering of different locations of South Asia.

In this study, a hierarchical cluster analysis based on G x L interaction of AUDPC for spot blotch appeared to be effective for subdividing the diverse South Asian region into more uniform subregions for better germplasm evaluation. The impact of abiotic stresses on response patterns in genotypes across different environments has been reported in studies in Australia (Cooper et al., 1994), West Africa–North Asia (Sivapalan et al., 2001), and in winter yield trials in Ontario, Canada (Yan and Hunt, 2001). If wide adaptability is the main breeding objective, representative locations from the eastern Indo-Gangetic plains of India and Nepal (Cluster I) should be chosen. This information could allow a better distribution of resources. On the other hand, if specific adaptability is the primary goal, then resources and efforts can be concentrated within the subregion of interest, as is also being pursued by NARS of different countries. Reducing the number of locations risks losing information, so one should carefully consider only those similar locations that are close in the clustering stages (Collaku et al., 2002). In this study, the best three locations were in India and Nepal, which is useful knowledge because until the late 1990s, wheat cultivars grown in Nepal were the primary cultivars released for eastern Gangetic plains of India and, presently, about half of the advanced lines adapted in Bangladesh come from Nepal.


   ACKNOWLEDGMENTS
 
We are grateful to the national programs that carried out the various trials reported in this study. We thank three anonymous reviewers for making useful suggestions that improved the quality of the article. The authors acknowledge Mr. Mike Listman of CIMMYT for his editing.


   NOTES
TOP
 NOTES
ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 
All rights reserved. No part of this periodical may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying, recording, or any information storage and retrieval system, without permission in writing from the publisher. Permission for printing and for reprinting the material contained herein has been obtained by the publisher.

Received for publication July 18, 2006.


   REFERENCES
TOP
 NOTES
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS
 DISCUSSION
 REFERENCES
 





This Article
Right arrow Abstract Freely available
Right arrow Figures Only
Right arrow Full Text (PDF) Free
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Similar articles in this journal
Right arrow Alert me to new issues of the journal
Right arrow Download to citation manager
Right arrow reprints & permissions
Citing Articles
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Agricola
Right arrow Articles by Joshi, A. K.
Right arrow Articles by Chand, R.
Related Collections
Right arrow Biometrics
Right arrow Plant and Environment Interactions

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Agronomy Journal Vadose Zone Journal
Journal of Natural Resources
and Life Sciences Education		 Soil Science Society of America Journal
Journal of Plant Registrations Journal of
Environmental Quality		 The Plant Genome