Agronomy Journal Journal of Natural Resources and Life Sciences Education
HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
QUICK SEARCH:   [advanced]


     

Published online 1 January 2007
Published in Agron J 99:220-228 (2007)
DOI: 10.2134/agronj2006.0144
© 2007 American Society of Agronomy
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 Similar articles in ISI Web of Science
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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Agricola
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Related Collections
Right arrow Statistics
Right arrow Field evaluation techniques
Right arrow Maize
Right arrow Biometrics
Right arrow Plant and Environment Interactions

Statistics

Yield Stability of Maize Hybrids Evaluated in Multi-Environment Trials in Yunnan, China

Xing-Ming Fana, Manjit S. Kangb,*, Hongmei Chenc, Yudong Zhangd, Jing Tanc and Chuxia Xuc

a College of Agronomy and Biotechnology, China Agriculture Univ., No. 2 Yuanmingyuan Xi Lu, Haidian, Beijing 100094, China
b Dep. of Agronomy & Environ. Management, Louisiana State Univ. Agric. Center, Baton Rouge, LA 70803-2110
c Yunnan Academy of Agricultural Science, Maize Research Center, Long Tou St., Kunming, Yunnan, 650205 China
d Agri-Service, John Deere, 3665 JFK Bldg. 1, Suite 310, Fort Collins, CO 80525

* Corresponding author (mkang{at}agcenter.lsu.edu)

Received for publication May 9, 2006.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Stable performance of maize (Zea mays L.) hybrids in multi-environment trials (MET) is important. The objectives of this investigation were (i) to evaluate grain yield stability of 13 Chinese hybrids tested across 10 locations in 2002 and 2003 via GGE biplot analysis and Kang's yield-stability statistic (YSi) and (ii) to identify nonrepresentative and/or nondiscriminating locations. Within years, cultivars and cultivar-by-location (C x L) interactions were significant. Heterogeneity caused by environmental index did not contribute appreciably to C x L interactions. The YSi identified, among the top five hybrids, LD10, Hai He, and YR1 as common between years. The GGE biplot analysis ranked hybrids with above-average yield across years as Hai He > LD10 > YR1 > Tun004 and for stability of performance as LD10, Hai He, Tun004, and YR1. The GGE biplots revealed that Hai He had the highest yield in seven and LD10 in 10 environments. Three different locations were identified in 2002 and 2003 as the least discriminating. Three common locations in 2002 and 2003 were the least representative of test locations. GGE biplot and YSi identified QC3, XHD892ck, and R313 as the least desirable hybrids. The YSi indicated ZZY6 and SB21-3 to be the most unstable hybrids between years. The only hybrid showing stable performance across locations was Tun004 in 2002. Overall, YSi versus GGE distance correlation (r) = –0.92**. YSi should be useful in selecting superior hybrids in the absence of GGEbiplot software. This information should help streamline the maize testing program in Yunnan.

Abbreviations: C x L, cultivar x location interaction • G, genotype main effect • GEI, genotype x environment interaction • GGE, genotype (G) + genotype x environment (GE) interaction • MET, multi-environment trials • YSi, Kang's yield-stability statistic


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
GENOTYPE x ENVIRONMENT INTERACTION (GEI) is a universal phenomenon when different genotypes (cultivars) are evaluated across diverse environments, as indicated by the voluminous literature on the subject (Kang, 1990, 1998; Snijders and Van Eeuwijk, 1991; Cooper and Hammer, 1996; Brancourt-Hulmel and Lecomte, 2003; Yan and Kang, 2003). GEI refers to differential responses of genotypes or cultivars across a range of environments (Kang, 1998, 2004). The GEI complicates the selection of superior genotypes (Magari and Kang, 1993; Ebdon and Gauch, 2002) and reduces correlation between phenotypic and genotypic values, thereby reducing progress from selection (Comstock and Moll, 1963). Ignoring GEI is problematic when it is significant and larger than the genotype main effect, which is a common scenario in yield trials (Gauch and Zobel, 1996). The GEI is considered to be one of the main reasons for the failure of formal breeding to serve small, resource-poor farmers (Ceccarelli et al., 2006).

Although the lack of consistency in performance across environments complicates cultivar selection, it can provide useful information to the researcher (Busey, 1983; Kang, 1998). For example, it can help justify the need for additional broad-based testing in different environments and predict the variability expected among farms (Busey, 1983). The GEI can be properly exploited to advantage through various approaches (Gauch and Zobel, 1996; Kang, 1998; Annicchiarico, 2002; Yan and Kang, 2003).

Most agronomically and economically important traits, such as grain yield, are quantitative in nature and routinely exhibit GEI. This necessitates genotype evaluations across multiple environments (called multi-environment trials [MET]) in the advanced stages of selection (Annicchiarico, 2002; Kang et al., 2004). In MET, differential genotypic responses to variable environmental conditions, especially when associated with changes in genotypic ranking, hinder the identification of superior, stable hybrids (Epinat-Le Signor et al., 2001). Other undesirable effects of GEI may include masking of the potential utility of exotic material (Giauffret et al., 2000). When GEI for a trait is significant, the use of appropriate breeding/management strategies is essential because the usefulness of overall genotype means is reduced or the use of overall genotype means is questionable (Kang, 1998, 2002; Annicchiarico, 2002).

By growing cultivars in different environments, the highest yielding and most stable cultivars can be identified (Lu'quez et al., 2002). When selecting genotypes for wide adaptation, plant breeders look for a noncrossover GEI or preferably the absence of GEI (Matus-Cádiz et al., 2003). Thus, the estimation of stability of performance becomes important to identify consistent-performing and high-yielding genotypes (Kang, 1998). Various concepts of stability are detailed elsewhere (Becker, 1981; Becker and Leon, 1988; Lin et al., 1986). From the farmers' standpoint, the agronomic concept (Becker, 1981) or Type 2 stability (Lin et al., 1986) is important.

Many stability statistics have been used to determine whether or not cultivars evaluated in MET are stable (Lin et al., 1986; Hühn, 1996; Flores et al., 1998; Hussein et al., 2000; Robert, 2002; Sabaghnia et al., 2006). Because the most stable genotype(s) may not be the highest yielding, the use of methods that integrate yield performance and stability to select superior genotypes becomes important (Kang, 1988; Pham and Kang, 1988; Kang and Pham, 1991; Kang, 1993; Kang and Magari, 1996).

Shukla (1972) presented a statistic called stability variance ({sigma}i2), which is identical to Wricke's (1962) ecovalence in ranking genotypes for stability (Kang et al., 1987). The {sigma}i2 partitions GEI and assigns it to individual genotypes. Shukla (1972) also allowed for the use of environmental covariates, such as environmental index and weather factors (e.g., temperature, humidity, rainfall), to remove heterogeneity from the GEI and compute residual GEI. This method has been used in many studies (Kang and Gorman, 1989; Helms, 1993; Kang, 1993; Magari and Kang, 1993). Covariates can be genotypic and environmental (Biarnes-Dumoulin et al., 1996; Van Eeuwijk et al., 1996), but in most studies, only environmental covariates have been used (Kang and Gorman, 1989; Magari and Kang, 1993; Kang et al., 2006). The use of covariates can improve sample variance estimator when sample size is small (Piepho and McCulloch, 2002). A considerable gain in accuracy can be achieved by using a covariate if there is sufficient correlation between the covariate and the variable under consideration (Piepho and McCulloch, 2002).

The yield-stability statistic (YSi) (Kang, 1993) and the GGE distance (i.e., the distance from the markers of individual genotypes to the ideal genotype) (ideal genotype has the highest yield and is absolutely stable) in GGE biplot analysis (Yan, 2001; Yan and Kang, 2003) help select for yield and stability. The YSi is based on Shukla's {sigma}i2 (Shukla, 1972), which belongs to Type 2 stability (Lin et al., 1986). The GGE biplot analysis is based on singular-value decomposition or principal component analysis (Yan and Kang, 2003). Ranking of genotypes, based on GGE distance, was found to be highly correlated with that based on YSi (r = –0.97**; ** = significant at the 0.01 probability level) (Yan and Kang, 2003).

The yield-stability statistic and GGE biplot analysis have been extensively used in MET analyses. The YSi was first applied in maize (Magari and Kang, 1993) and soybean trials (Pazdernik et al., 1997). The availability of a QBASIC program called STABLE (Kang and Magari, 1995) furthered the use of YSi. For example, Waldron et al. (2002) used it to study stability of yield of cool-season pasture grasses, Gravois and Bernhardt (2000) investigated stability of performance of rice (Oryza sativa L.) cultivars, Upadhya and Cabello (2000a,b) evaluated the stability of potato (Solanum tuberosum L.) seed families, and Priyadarshan (2003) investigated contributions of weather variables for specific adaptation of rubber tree (Hevea brasiliensis Muell.-Arg.) clones. In addition, Bertoia et al. (2002) identified maize inbred lines capable of improving ear and stover yield and quality of silage maize hybrids.

The GGE biplot methodology has been used to evaluate test environments in soybean (Yan and Rajcan, 2002), cotton (Blanche and Myers, 2006), and common bean (Kang et al., 2006); to characterize end-use quality in wheat (Morris et al., 2004); and to target cultivars to specific environments in rice (Samonte et al., 2005). Using GGE biplot analysis, Ober et al. (2005) successfully evaluated physiologic traits as indirect selection criteria for drought tolerance.

In China, MET are widely used for hybrid evaluation. Seed companies wishing to sell their hybrids in a province must get them tested in an MET for 2 yr and recommended by an evaluation board. The evaluation boards recommend cultivars almost exclusively on the basis of mean yield, with little or no consideration for stability of performance. We found one study on sweet corn (Zea mays L.) from China where the additive main effects and multiplicative interaction analysis was used (Wu et al., 2003). Thus, our objectives were to investigate stability of performance of several Chinese hybrids tested across a number of diverse Chinese environments via the use of the YSi (Kang, 1993) as computed via the program called STABLE (Kang and Magari, 1995) and via the GGEbiplot software (Yan, 2001; Yan and Kang, 2003). The information generated should be useful to agronomists in fine-tuning the testing program by targeting appropriate hybrids to different locations and by identifying redundant locations to conserve limited resources, such as time, space, and money.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 
Maize Hybrids and Experiment Implementation
Thirteen maize hybrids (cultivars), contributed by nine seed companies in China, were used in this investigation (Table 1). All hybrids belonged to the full-season maturity group (mean = 123.9 ± 2.5 d) (Table 1). Preliminary observations indicated that the selected cultivars had good yielding ability and other desirable agronomic attributes across the Yunnan province (unpublished data). A randomized complete-block design with three replications at 10 locations was used for trials conducted in 2002 and 2003. The location–year combinations were regarded as environments (20 environments). Soil was sandy loam at all locations, which is typical for Yunnan. Information on locations relative to longitude, latitude, and altitude and on planting dates, plant density, and previous crop grown is given in Table 2. Plots, consisting of five rows (20 m2 with row spacing at 80 cm and plant spacing at 20 cm), were hand-planted. Plots were initially over-planted and later thinned to two plants per hill. Plant density varied from 60030 plants ha–1 in one environment to 62505 plants ha–1 in 12 environments to 67500 plants ha–1 in seven environments (Table 2). Field conditions (i.e., soil type, fertility level, and irrigation system) were similar for all experimental locations. Field management procedures, such as application of fertilizers and insecticides, were standard across locations. Three middle rows of each plot were harvested from 15 September to 5 October, and yield was adjusted to 130 g kg–1 grain moisture content.


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

  		Table 1. Hybrid (cultivar) name and maturity and mean grain yield (Mg ha–1) from multi-environment trials conducted at 10 locations in 2002 and 2003 in Yunan, China.

 

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

  		Table 2. Longitude, latitude, altitude, plant density, and previous crop for the 10 locations used in 2002 and 2003 tests.

 
Statistical Analyses
The yield data were subjected to an ANOVA as follows:
(i) Combinations of years (2002 and 2003) and 10 locations were treated as 20 environments. The PROC GLM of SAS version 9 (SAS Institute, 2002) was used to partition yield variation into environments, cultivars, and cultivars x environment interaction. Environmental means of cultivars were subjected to stability analysis using Kang and Magari's (1995) program ‘STABLE’ to obtain estimates of YSi (Kang, 1993) and {sigma}i2 (Shukla, 1972). The equation for computing {sigma}i2 is as follows:

Formula 1[1]
where s = number of environments, t = number of cultivars, uij = Xij – Formula 1, Xij = observed yield for the ith cultivar in the jth environment, Formula 1 = mean of all cultivars in environment j, and Formula 1 = {sum}juij/s.
The STABLE program also provided a partition of the total cultivars x environments (C x E) interaction into heterogeneity (nonadditivity or the linear effect of environmental index (EI); EI = Formula 1 – Formula 1, where Formula 1 is mean of all cultivars across all environments) and residual C x E interaction.

(ii) Within-year analyses were conducted using PROC GLM of SAS (SAS Institute, 2002) to partition yield variation into locations, cultivars, and C x L interaction. Location means of cultivars were subjected to stability analysis according to Kang and Magari (1995). Total C x L interaction was partitioned into heterogeneity and residual C x L interaction.

The GGEbiplot software (Yan, 2001) was used to generate graphs showing (i) "which-won-where" pattern, (ii) ranking of cultivars on the basis of yield and stability, (iii) location vectors, and (iv) comparison of locations to ideal location (Yan and Kang, 2003). Angles between environment vectors were used to judge correlations (similarities/dissimilarities) between pairs of environments (Yan and Kang, 2003). A GGE distance was computed and correlated with YSi.


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
SUMMARY AND CONCLUSIONS
 REFERENCES
 
The combined ANOVA for grain yield (Table 3) revealed that environments (year–location combinations), cultivars, and C x E interaction accounted for 69, 8.5, and 16% of the total sum of squares, respectively. Of the 69% E variation, 68% was attributable to locations (L), 32% to location x year (L x Y) interaction, and 0% to years (Y). The E portion in MET has been known to be the largest among all sources of variation, but it is regarded as irrelevant for cultivar evaluation (Yan and Kang, 2003). This is the reason E is removed from the observed phenotypic data, which helps concentrate on C and C x E or CE, which are relevant for cultivar evaluation (Yan and Kang, 2003). Both YSi and GGE biplot methodology are based on the C+CE (because C = genotype or G, the GGE contraction is usually used rather than CCE).


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

  		Table 3. Analysis of variance for yield data (Mg ha–1) obtained from maize trials conducted in China at 10 locations in 2002 and 2003 (environments constitute year–location combinations).

 
The significant L x Y interaction warranted separate ANOVA for each year (Table 4). Within each year, locations accounted for 67% (in 2002) and 71% (in 2003) of the total sum of squares for grain yield variation. Of the total sum of squares, C accounted for 8% (in 2002) and 11% (in 2003), whereas C x L interaction accounted for 18% (in 2002) and 13% (in 2003). In these by-year ANOVA, C+CL or G+GL are the relevant sources of variation for cultivar evaluation.


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

  		Table 4. By-year ANOVA for yield data (Mg ha–1) obtained from maize trials conducted in China at 10 locations in 2002 and 2003.

 
Heterogeneity Attributable to Environmental Index
Mean yield levels of different cultivars at different locations for 2002 and 2003 are shown in Table 1, as are the values for environmental index (the covariate used to remove heterogeneity from C x E and C x L interactions) for all 20 environments. The partitioning of the C x E interaction (Table 3) and the C x L interaction (Table 4) into heterogeneity (or nonadditivity) (i.e., linear effect of environmental index) and residual interaction revealed in both cases that environmental index was not an appreciable contributor to these interactions. Of the total C x E interaction sum of squares, heterogeneity accounted for a nonsignificant 8.2% (Table 3). For the by-year ANOVA (Table 4), heterogeneity was nonsignificant in both years and accounted for 11.4% and 9.6% of the total C x L interaction sum of squares for 2002 and 2003, respectively. This suggested that the cause of these interactions must be factors other than those captured by the linear effect of environmental index. Kang and Gorman (1989) found heterogeneity caused by environmental index to be significant in a Louisiana maize multi-environment yield trial, but individual weather factors (e.g., rainfall, humidity, temperature) did not cause significant heterogeneity. Using the factorial regression approach, Epinat-Le Signor et al. (2001) provided biological interpretations of GEI. They found earliness of flowering, water balance, and mean temperature in the second part of the cycle to be contributors to GEI for maize yield. However, they did not use environmental index as a covariate.

In the case of C x E interaction (Table 3), C x L, C x Y, and C x L x Y interactions were significant. The causes of these interactions might be differential influences of stress factors encountered by cultivars in different years and/or locations. Thus, a thorough characterization of environments (locations and years) is needed to better understand the C x L and C x Y interactions. Environmental index is a nonspecific covariate and encompasses all differential factors across environments (e.g., weather factors). Detailed weather data were not available for most of locations used in this study.

Cultivar Stability
The YSi, which is an integrated measure of yield and stability of cultivars evaluated in performance trials, helped identify cultivars that were high-yielding and relatively stable. The YSi values computed across 20 environments (Table 5) identified Hai He, LD10, YR1, SB21-3, and Tun004 as the top five cultivars and YY196, 5307, XHD892ck, R313, and QC3 as the bottom five cultivars. In this combined analysis, all cultivars had highly significant {sigma}i2 (i.e., all cultivars were penalized for instability equally), and that is why YSi ranks paralleled the yield ranks. Under such circumstances, cultivar selection is based solely on yield.


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

  		Table 5. Overall stability analysis of yield data (Mg ha–1) obtained from 13 maize cultivars evaluated across 20 environments (environments constitute year–location combinations) in China via yield-stability statistic (YSi{dagger}).

 
When YSi analyses were conducted for individual years (Table 6), {sigma}i2 estimates obtained from the total C x L interaction showed differential levels of significance for some cultivars in 2002 and 2003. According to YSi, the top five cultivars in 2002 were LD10, Hai He, Tun004, SB21-3, and YR1, which were the same as identified in the combined analysis (Table 3), whereas in 2003, they were LD10, Hai He, ZZY6, YR1, and R314. In the latter year, SB21-3 was replaced by ZZY6. In both years, there were three cultivars (LD10, Hai He, and YR1) in common. The bottom five cultivars in 2002 were QC3, XHD892ck, R314, ZZY6, and R313; the bottom five cultivars in 2003 were QC3, R313, 5307, XHD892ck, and SB21-3. Among the bottom five cultivars, the two years shared three cultivars. Cultivars ZZY6 and SB21-3 were the most unstable cultivars between years, as the former ranked among the top five cultivars in 2003 and the latter ranked among the top five cultivars in 2002, but neither one was selected in both years.


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

  		Table 6. By-year stability analyses of yield data (Mg ha–1) obtained from 13 maize cultivars evaluated in 2002 and 2003 at 10 locations in China via the yield-stability statistic (YSi{dagger}).

 
The only cultivar that exhibited stable performance (nonsignificant {sigma}i2) across locations was Tun004 in 2002 (Table 6). It was among the top five cultivars in 2002 but not in 2003. All other cultivars were judged as unstable in both years. Growers would prefer cultivars that perform consistently across years.

Genotype + Genotype x Environment Interaction Biplot Analyses
Figure 1 represents a polygon view, which indicates that Hai He was the highest-yielding cultivar in the CX3, YX3, QJ3, ZT3, BS2, YX2, and ZT2 locations (the numbers 2 and 3 represent years 2002 and 2003, respectively). Cultivar LD10 was the highest-yielding cultivar in the CX2, LJ2, KM2, SL2, YJ2, BS3, LP3, KM3, YJ3, and SL3 locations. Cultivar 314 was the highest yielding in the LP2 and QJ2 locations. No cultivar yielded the highest in environment LJ3. For the remaining 10 cultivars, highest yield was not recorded in any environment. The cultivars Hai He and DL10 seemed to be widely adapted across several environments. When an "ideal" cultivar view was drawn (Fig. 2 ), Hai He and DL10 were the closest to the ideal cultivar. An ideal genotype is defined as one that is the highest yielding across test environments and is absolutely stable in performance (i.e., one that ranks the highest in all test environments) (Yan and Kang, 2003).


Figure 1
View larger version (18K):
[in this window]
[in a new window]

  		Fig. 1. A genotype + genotype x environment interaction bi-plot showing cultivar performance in each environment. Environments are shown in upper case letters; the numbers 2 and 3 represent years 2002 and 2003, respectively.

 

Figure 2
View larger version (22K):
[in this window]
[in a new window]

  		Fig. 2. A genotype + genotype x environment interaction bi-plot showing mean yield (x axis) and stability (y axis) of cultivars. An ideal cultivar is at the center of the innermost circle.

 
Figures 3 and 4 relate to test environments across 2002 and 2003 (individual-year graphs were made but are not shown here). Figure 3 represents vectors of all 20 environments; the linear map to the right of the graph (in degrees) helps indicate relationships between environments and helps decipher the names of environments when they are too crowded in the graph. When the biplot fits the data perfectly, the cosine of the angle between two vectors represents the correlation between them. The vector length corresponds to discriminating ability (Yan and Kang, 2003). The environments BS3, BS2, LJ3, LJ2, YX2, ZT2 had longer vectors than other environments. Thus, they were the best environments for genetic differentiation of cultivars. The most nondiscriminating locations were YJ in both years and SL in 2002. The location ZT was the least representative of other locations in both years. Overall, the poorest test environments relative to the ideal environment were YJ2, YJ3, QJ2, and SL2. Should limited resources dictate that certain locations be dropped from the testing program, location YJ would be a strong candidate.


Figure 3
View larger version (21K):
[in this window]
[in a new window]

  		Fig. 3. A genotype + genotype x environment interaction bi-plot showing relationships among 20 environments (upper case letters followed by 2 or 3; 2 = 2002 and 3 = 2003). Cosine of an angle between vectors of any two environments represents correlation between the two environments. The unit of the linear map to the right of the graph is in degrees (the angle between the two extreme environmental vectors is about 140 degrees). The smaller the angle between any two vectors, the greater the correlation between them (see Yan and Kang, 2003).

 

Figure 4
View larger version (20K):
[in this window]
[in a new window]

  		Fig. 4. A genotype + genotype x environment interaction bi-plot showing comparisons of 20 environments with ideal environment (the ideal environment is at the center of the innermost circle). The closer an environment is to the ideal environment is, the more desirable it is; the ideal environment is the one with the greatest discriminating ability and is the most representative of all environments.

 
Figure 3 shows that the angles between the environmental vectors of BS in 2002 and BS in 2003 were not very similar. On the other hand, SL in 2002 and 2003, ZT in 2002 and 2003, and KM in 2002 and 2003 were quite similar. Locations QJ and LP in 2002 and LP and SL in 2003 were similar as suggested by individual year graphs (not shown). Because of inconsistency of results between years, this information would not be helpful in dropping any location.

Figure 4 represents discriminating ability and representativeness of environments. An ideal environment is one that is most discriminating for genotypes and is representative of all other environments (Yan and Kang, 2003). Environments BS3 and LJ2, followed by SL3, CX2, CX3, and LP3, were the best environments. The least representative environments were ZT3, ZT2, YX2, LJ3, and LP2.

Relationship between YSi and GGE Distance
The overall YSi showed a relatively high correlation with GGE distance (r = –0.92**). The negative correlation means that according to YSi, genotypes with high YSi values are desirable and that genotypes with shorter distances from the ideal genotype are considered desirable. Yan and Kang (2003) reported a rank correlation coefficient of –0.97** between rankings of genotypes based on mean performance and stability, as measured by the GGE distance, and genotype ranking based on YSi. This correlation is important because in the absence of GGEbiplot software, the STABLE program that computes the YSi statistic could be a good substitute for GGE distance for selecting superior cultivars.

The YSi identified, among the top five hybrids, LD10, Hai He, Tun004, and YR1 as common selections between years. GGE biplot analysis ranked hybrids with above-average yield across years for yield as Hai He > LD10 > YR1 > Tun004 and for stability of performance as LD10, Hai He, Tun004, and YR1. GGE biplot analysis and YSi identified QC3, XHD892ck, and R313 as the least desirable hybrids. The YSi indicated R314, ZZY6, and SB21-3 to be unstable between years (Table 5). The only hybrid with a nonsignificant {sigma}i2 across locations was Tun004 in 2002. All other hybrids showed a significant {sigma}i2 in both years. Overall, in the absence of GGEbiplot software, YSi may be reliably used for selecting hybrids with high yield and stable performance.


   SUMMARY AND CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
REFERENCES
 
The GGE biplot and the YSi statistic identified QC3, XHD892ck, and R313 as the least desirable hybrids. Hybrids ZZY6 and SB21-3 were identified as the most unstable hybrids between years via the YSi statistic. Only one hybrid (Tun004) in 1 yr had stable performance (a nonsignificant {sigma}i2) across locations; all other hybrids showed a significant {sigma}i2 in both years. Because the YSi and GGE distance were highly correlated (r = –0.92**), YSi can be reliably used for selecting high-yielding and stable hybrids in the absence of GGEbilot software. The GGE biplot analyses revealed two hybrids to be highly adapted to several environments (LD10 had the highest yield in 10 of the 20 environments, followed by Hai He, with the highest yield in seven environments). The GGEbiplot software helped identify some of the least discriminating locations and those that were the least representative of test locations. Thus, the GGE biplot methodology was a useful tool for identifying locations that optimized cultivar performance and for making better use of limited resources available for the testing program. Causes of GEI should be ascertained by using one or more environmental covariates (e.g., rainfall, sunshine, temperature, humidity) via the YSi statistic.


   REFERENCES
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 SUMMARY AND CONCLUSIONS
 REFERENCES
 




This article has been cited by other articles:


Home page
 Agron. J.Home page
N. Sabaghnia, H. Dehghani, and S. H. Sabaghpour
Graphic Analysis of Genotype by Environment Interaction for Lentil Yield in Iran
Agron. J., May 7, 2008; 100(3): 760 - 764.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
B. Glaz and M. S. Kang
Location Contributions Determined via GGE Biplot Analysis of Multienvironment Sugarcane Genotype-Performance Trials
Crop Sci., May 1, 2008; 48(3): 941 - 950.
[Abstract] [Full Text] [PDF]

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 Similar articles in ISI Web of Science
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 HighWire
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Agricola
Right arrow Articles by Fan, X.-M.
Right arrow Articles by Xu, C.
Related Collections
Right arrow Statistics
Right arrow Field evaluation techniques
Right arrow Maize
Right arrow Biometrics
Right arrow Plant and Environment Interactions

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