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


     

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 ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Agricola
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Related Collections
Right arrow Best Management Practices
Right arrow Other Crop Management
Right arrow Wheat
Published in Agron. J. 96:169-180 (2004).
© American Society of Agronomy
677 S. Segoe Rd., Madison, WI 53711 USA

WHEAT

Graphic Analysis of Genotype, Environment, Nitrogen Fertilizer, and Their Interactions on Spring Wheat Yield

B. L. Ma*,a, W. Yana, L. M. Dwyera, J. Frégeau-Reida, H. D. Voldenga, Y. Dionb and H. Nassc

a Eastern Cereal and Oilseed Res. Cent. (ECORC), Agric. and Agri-Food Canada, 960 Carling Ave., Ottawa, ON, Canada K1A 0C6
b Centre de Recherche sur les Grains Inc. (CÉROM), 335, Chemin des Vingt-cinq Est, Saint-Bruno-de-Montarville, QC, Canada J3V 4P6
c Crops and Livestock Res. Cent. (CLRC), Agric. and Agri-Food Canada, 440 Univ. Ave., Charlottetown, PE, Canada C1A 4N6

* Corresponding author (mab{at}agr.gc.ca).

Received for publication February 10, 2003.

   ABSTRACT
TOP
 ABSTRACT
INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Interest in growing hard red spring wheat (Triticum aestivum L.) in eastern Canada is increasing due to its potential returns relative to other small-grain cereals and oilseed crops. The objectives of this study were to determine the effects of year, site, genotype, N application, and their interactions on the yield of hard red spring wheat (HRSW) and to demonstrate the application of the recently developed biplot methodology in visualizing agronomic research data. Ten HRSW cultivars were grown in five locations across three provinces from 1998 to 2000, constituting a total of 11 year–site combinations. In each environment, four levels of fertilizer N (50, 100, 150, and 200 kg ha–1) were applied. The N main effect, N x environment interaction, and N x genotype interaction were not significant. However, biplot analysis did reveal crossover N x environment interactions: Although higher N rates generally led to higher yield, the opposite was true in some environments. This was attributed to heavy Fusarium head blight (Fusarium graminearum Schwabe) and/or foliar diseases in these environments, which was exacerbated by higher N rates. The strong genotype x environment interactions were mainly associated with two cultivars that yielded well in most environments but very poorly in two environments in which Fusarium head blight was severe. This study thus highlighted the importance of Fusarium head blight resistance in HRSW production in eastern Canada. An environment x factor biplot was described for the first time, which was highly effective in revealing the interrelationship among environmental factors and in revealing the weather and soil patterns of the environments.

Abbreviations: AMMI, additive main effect and multiplicative interaction • CEF, Central Experimental Farm • GBF, Greenbelt Research Farm • GGE, genotype main effect plus genotype x environment interaction • HRSW, hard red spring wheat • IPC1, first-interaction principal component • 99BRU, St. Bruno in 1999 • 00BRU, St. Bruno in 2000 • 98CEF, Central Experimental Farm in 1998 • 99CEF, Central Experimental Farm in 1999 • 00CEF, Central Experimental Farm in 2000 • 99GBF, Greenbelt Research Farm in 1999 • 00GBF, Greenbelt Research Farm in 2000 • 00HYA, St. Hyacinthe in 2000 • 99PEI, Prince Edward Island in 1999 • 00PEI, Prince Edward Island in 2000 • 99ROS, St. Rosalie in 1999


   INTRODUCTION
TOP
 ABSTRACT
 INTRODUCTION
MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
IN CANADA, bread wheat is traditionally grown in the Prairie Provinces of western Canada where superior bread-making quality is obtained under dryland conditions with relatively low yield. In eastern Canada, grain yield of spring wheat can be two to three times greater than that in western Canada, but bread-making quality is often limited due to low and inconsistent grain protein concentration (Ayoub et al., 1994). Consequently, spring wheat in eastern Canada is used mostly for blending with high quality wheat from western Canada or as livestock feed. Most wheat produced in eastern Canada has been of the winter type and of the soft red or white pastry class. In recent years, producers and millers have shown increasing interest in growing HRSW due to the removal of transportation subsidies on wheat shipped from western Canada.

Nitrogen has been one of the most investigated factors in wheat production. Numerous studies indicate that N fertilization can increase both wheat grain yield and grain protein concentration (Fowler et al., 1990; Gauer et al., 1992; Rasmussen et al., 1997; Rawluk et al., 2000; Ehdaie and Waines, 2001; Cooper et al., 2001; Lafond et al., 2001; Lopes-Bellido et al., 2001; Selles and Zentner, 2001). A lag period in N response exists between grain yield and grain protein concentration. That is, grain protein concentration responds to higher levels of N application than does grain yield (Fowler et al., 1990; Selles and Zentner, 2001). It has also become common knowledge that early-season N application influences both yield and grain protein concentration while N application near or after anthesis influences only grain protein concentration (Wuest and Cassman, 1992a, 1992b; Ayoub et al., 1995; Rawluk et al., 2000; Woolfolk et al., 2002).

Little information exists on the N response of spring wheat in eastern Canada. In a study to evaluate the response of spring wheat cultivars to N application, Ayoub et al. (1995) found that all tested cultivars produced grain suitable for bread wheat quality, but the N management required to achieve this varied: Some cultivars such as Katepwa and Columbus produced high quality grain at all N rates whereas Max required split N applications to achieve acceptable protein concentration. Cultivar, N application level, and N application strategy were shown to be important to bread wheat production in Quebec, Canada (Smith et al., 2003). It was hypothesized that differential responses of yield and protein content to fertilizer N rates existed among genotypes. Comprehensive testing across various environments would be useful in identifying most limiting factors for specific cultivars suitable for individual environments and thus designing appropriate fertilizer technologies to meet the quality requirement.

Agronomic data typically contain complex interactions that are difficult to understand without the aid of some graphical display. The biplot technique of Gabriel (1971) has evolved into a powerful tool for visualizing large data sets with complex interactions (Yan et al., 2000; Yan, 2001, 2002). However, its application has been limited to visual analysis of breeding and genetics research data (Yan and Kang, 2003).

The objectives of this study were twofold: (i) to determine the effects of year, site, genotype, N application, and their interactions on the yield of HRSW in eastern Canada and (ii) to demonstrate the application of the biplot technique in visual analysis of agronomic research data.


   MATERIALS AND METHODS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
RESULTS AND DISCUSSION
 CONCLUSIONS
 REFERENCES
 
Field Experiments
A factorial experiment including four levels of N (50, 100, 150, and 200 kg N ha–1) and 10 cultivars of HRSW was conducted at 11 environments (year–site combinations). The 10 wheat cultivars were Aquino, AC Barrie, SS Blomidon, AC Brio, Celtic, AC Drummond, Grandin, Quantum, AC Voyageur, and AC Walton. Cultivars SS Blomidon, AC Brio, and AC Drummond were not tested in 1998. The 11 environments were

The PEI experiment sites both were a Charlottetown fine sandy loam, classified as Typic Humicryods soil in the USDA soil classification system and Orthic Podzol in the Canadian classification system. The two locations in Ontario were geographically close; they differed only in soil type: soil at the CEF site is a Kars sandy loam, classified as a Hapludalf Udalf Alfisol in the USDA classification system and a Gleyed Gray Brown Luvisol in the Canadian classification system; soil at the GBF site is a Dalhousie clay loam, classified as Eutrocryept Cryept Inceptisol in the USDA system or the Gleyed Melanic Brunisol for the Canadian system. The three locations in Quebec were also close but of different soil types: All of the soils were classified as Typic Humaquept soil in the USDA soil classification system and named as Orthic Humic Gleysol Ste-Rosalie clay loam at St. Bruno, Orthic Humic Gleysol Ste-Rosalie sandy loam for both St. Hyacinthe and St. Rosalie sites in the Canadian classification system. In each year-site, a split-plot design with four replications was used, with N treatments as main plots and cultivars as subplots. The harvested subplot area was 3 by 5 m. The year and location combinations are considered to be random and referred to as the environment factor as described in detail below.

In all the 11 environments (all sites each year), soil samples were taken before sowing and analyzed for the following characteristics: organic matter, total N concentration, available P and K, pH, clay, and sand content. Available P was extracted with the Bray-1 solution and determined with Technicon Autoanalyzer (Technicon, New York) using the ammonium molybdate and ascorbic acid colorimetric method (Murphy and Riley, 1962). Soil test K was extracted with 1 M ammonium acetate (pH 7.0) and determined using the flame photometer (Quikchem FIA+, Zellweger Analytics, Milwaukee, WI). Adequate P and K fertilizer was applied according to the provincial soil test recommendations before sowing (e.g., Ontario Field Crop Publication 296, The Ministry of Agriculture, Food, and Rural Affairs, Toronto, ON, Canada). Fertilizer N as NH4NO3 was broadcast at the planned rates and incorporated to a depth of approximately 5 cm immediately after application. In all site-years, sowing was made during the last week of April and the first week of May. Plots were combine-harvested and grain yield reported at uniform moisture content of 135 g kg–1. In year 2000, harvest indices, percentage tombstone seeds (a measure of Fusarium head blight infestation), and percentage black point seeds (an indication of foliar diseases) were determined using a 1-m row sample from each plot of the CEF and GBF experiments. Monthly mean temperature and monthly total precipitation data during the spring wheat-growing period (from April to August) were obtained from Environment Canada (http://sis.gis.agr.ca/cansis/nsdb/climate/wthr_query.html; verified 11 Sept. 2003). The weather stations were located within 0.5 to 10 km of the experimental sites.

Data Analysis
Analysis of Variance
The SAS procedure GLM (SAS Inst., 1996) was used to estimate the main effects of N, environment (including year, site, year x site interaction), and the N x environment interaction; these effects were tested against the replication within environment as the main-plot error (Error A). The genotype main effect and all other possible interactions among these factors were tested against the residual or subplot error (Error B). All effects but the block within environment were assumed fixed. Treatment and interaction effects were considered to be significant if P ≤ 0.05.

Using the SAS procedure MIXED (Littell et al., 1996), best linear unbiased predictors were estimated to generate a N x environment two-way table, which was further analyzed using the biplot technique (described below). Similarly, a genotype x environment two-way table was generated from the original data for biplot analysis. For convenience, both will be generalized as treatment x environment two-way data.

Biplot Analysis
Two types of biplots, AMMI1 biplot (Zobel et al., 1988) and GGE biplot (Yan et al., 2000), were used to visualize the N x environment two-way data and the genotype x environment two-way data. AMMI stands for the statistical model of additive main effect and multiplicative interaction (Gauch, 1988), and GGE stands for genotype main effect plus genotype x environment interaction (Yan et al., 2000). Both types of biplots displayed treatment x environment interactions, but each had its unique functions. The AMMI1 biplot allows visualization of the main effects of the treatments and of the environments, in addition to the most important treatment x environment interactions. The GGE biplot allows visualization of any crossover treatment x environment interactions, relationships among treatments, and relationships among environments. The joint use of both types of biplot should allow a comprehensive understanding of the data.

AMMI1 Biplot
The AMMI1 biplot is constructed by plotting the main effects of treatments and environments against their respective interaction scores, which are symmetrically scaled scores of the first-interaction principal component (IPC1) resulting from subjecting the double-centered data (i.e., the interaction matrix) to singular-value decomposition. The construction and interpretation of an AMMI1 biplot was described in detail in Zobel et al. (1988). Previously, the two axes in an AMMI1 biplot have different units (original units for main effects and square root of the original unit for interaction, e.g., Zobel et al., 1988). The AMMI1 biplots presented here are slightly different from the conventional AMMI1 biplots in that both of their axes are in square root of the original units. One advantage of this modification is that the shape of the biplot becomes objective; it is determined only by the data if the axes are drawn such that they have the same physical scale.

GGE Biplot
A GGE biplot was constructed using the first two principal components (PC1 and PC2) derived from subjecting the environment-centered data to singular-value decomposition. A GGE biplot does not display the main effects of the environments but has many visual interpretations that an AMMI1 biplot does not have: (i) The polygon view of a GGE biplot allows visualization of the which-won-where pattern (i.e., which variety or N application had the highest yield in which environment); (ii) the average environment coordination view allows simultaneous visualization of the mean performance and stability of the treatments, the discriminating ability vs. representativeness of the environments; and (iii) the environment vector view allows visualization of the interrelationship among environments (Yan, 2001, 2002; Yan and Kang, 2003). For appropriate visualization of both the relationship among the environments and the crossover treatment x environment interactions, the singular values were entirely partitioned into the environment eigenvectors (Yan, 2002).

In addition, attempting to characterize the environments and to relate the mean yield of the environments to various weather and soil factors, a biplot based on an environment x factor two-way table was constructed, which was similar to that based on a genotype x trait two-way table described by Yan and Rajcan (2002). The biplot was based on the first two principal components derived from subjecting the data scaled by the factor-centered and within-factor standard deviation. The singular values were entirely partitioned into the factor eigenvectors so that the biplot was suitable for visualizing the relationship among factors as well as the weather and soil pattern in each of the environments.

All biplots presented in this paper were generated using the software GGEbiplot package that runs in a Windows environment, an earlier version of which was described in Yan (2001). Up-to-date information on GGE biplot is available at http://www.ggebiplot.com (verified 9 Sept. 2003).


   RESULTS AND DISCUSSION
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
CONCLUSIONS
 REFERENCES
 
Analysis of Variance and Variance Components
Analysis of variance was conducted to determine the effects of N, environment (consisting of year, site, and year x site interaction), genotype, and all possible interactions among these factors, on grain yield (Table 1). Nitrogen, environment, and interaction between them were tested against block (replication) within environment (main-plot error), and all other sources were tested against the residual (subplot error) (Table 1).


View this table:
[in this window]
[in a new window]
  		Table 1. Analysis of variance on grain yield for the 11 site-year environments and all possible interactions.

 
Across sites and years, grain yields of genotypes varied from 1.1 to 5.3 Mg ha–1 while differences in grain yield among N rates were small, and the effect was often mixed (Table 2). In general, positive N treatment effect on yield occurred at the higher-than-average yield sites, such as 99GBF, 99ROS, 00HYA, and 00BRU, and was associated with sufficient precipitation during the growing period, especially in June and July in these environments. Although only one site was tested in 1998 (98CEF), the severe moisture stress from heading to physiological maturity (i.e., total precipitation was only 45% within 2 wk around heading and 69% in July of the 40-yr averages of the site) resulted in the lowest yield among the 11 site-years. In contrast, grain filling for 00CEF and 00GBF sites was adversely affected by Fusarium head blight and/or other foliar diseases due to excessively wet conditions (total precipitation within 2 wk around heading was 201% of the 40-yr averages of the site), as discussed in more detail below. The most significant source of total variation was the genotype x environment interaction. The most significant components of the environment effect were the year main effect and the site main effect. The most significant components of the genotype x environment interaction were genotype x site, followed by genotype x year x site and genotype x year. However, N main effect, genotype x N interaction, and genotype x environment x N interaction were nonsignificant (Table 1). This is in agreement with studies conducted in eastern Canada (Ayoub et al., 1995) and in Australia (Cooper et al., 2001). In the latter study, a genotype x management (genotype selection, N level and timing, preceding crops, and irrigation) interaction was found to be the dominant factor in the genotype x environment interaction. Lack of N main effect, and N x environment interaction in this study was probably due to either the high N-supplying power of soils in eastern Canada, such that an application of 50 to 100 kg N ha–1 fertilizer was adequate for maximum yield, or to the inability of genotypes to reach their yield potential because of diseases and other stresses in some of the site-years. Nonetheless, higher N rates did increase grain protein concentrations for all varieties, and it is necessary to apply higher N rates for certain varieties to reach the minimum protein level for bread-making quality (Ma et al., unpublished, 2003).


View this table:
[in this window]
[in a new window]
  		Table 2. Mean grain yield (kg ha–1), expressed as the best linear unbiased predictors (Littell et al., 1996) of spring wheat listed in N x environment two-way data.{dagger}

 
Further discussion will be focused on biplot analysis of the N x environment two-way data and the genotype x environment two-way data. Although the N x environment interaction was not significant, further biplot analysis of the N x environment two-way data is justified for two reasons. First, even though N x environment was not significant as a whole, some individual N x environment interactions may be important. Second, N x environment interactions represent a large category of possible management x environment interactions in agronomic research, and the method described here may be useful for similar studies.

Visualization of the Nitrogen x Environment Two-Way Data
Yield mean, expressed as the best linear unbiased predictors of yield (Littell et al., 1996), for each of the four N rates in each of the 11 environments (year–site combinations) are presented in Table 2. The environment main effect explained 98% of the total variation. Nitrogen rates explained <0.5% of the variation, and the rest was due to N x environment interaction. Therefore, almost all information contained in Table 2 is graphically displayed in an AMMI1 biplot (Fig. 1), in which the main effects of N rate and environment are displayed by the x-axis and the IPC1 scores of N and environment are displayed by the y-axis. The unit for both axes is square root of kilograms per hectare.



View larger version (14K):
[in this window]
[in a new window]
  		Fig. 1. AMMI1 biplot of the N x environment two-way data. Both axes are in square root of kilograms per hectare. IPC1, first-interaction principal component; E, environment; G, genotype.

 
Figure 1 shows that there was little variation among the four N rates in mean yield, but the 11 environments fell clearly into two groups: the above-mean group, including 99ROS, 00HYC, 00BRU, 99GBF, and 99CEF, and the below-mean group, including 98CEF, 00CEF, 00GBF, 00BRU, 00PEI, and 99PEI. The y-axis of Fig. 1 displays the interactions between the N rates and the environments and reveals that the 50 kg ha–1 N rate interacted positively with environments with positive IPC1 scores, namely, 00CEF, 00GBF, and 99BRU. In other words, the 50 kg ha–1 N rate produced higher yields in these environments than what can be expected from their respective mean yields across all N rates and the mean yield of 50 kg ha–1 N rate across all environments. On the contrary, 50 kg ha–1 N rate interacted negatively with environments with negative IPC1 scores such as 00BRU and 99ROS. The 200 kg ha–1 N rate interacted with the environments in exactly the opposite way. Under these environments, grain set and filling were negatively affected by the moisture stress (i.e., severe shortage of precipitation), and N uptake and crop demand for N were probably both limited. As reported elsewhere (Lafond et al., 2001), the extent of crop yield response to fertilizer N application depended on the moisture status.

Although the AMMI1 biplot (Fig. 1) allows visualization of the main effects of N rates and the environments, it does not show which N rate led to the highest yield in each of the environments. The which-won-where pattern can be visualized only by the polygon view of a GGE biplot. The GGE biplot based on data in Table 2 explained 97% of the yield variation of the environment-centered data (i.e., variation due to N main effects plus N x environment interaction, Fig. 2). The polygon is formed through connecting the markers of the four N rates. Starting from the biplot origin, perpendicular lines are drawn to each side of the polygon, which divide the biplot into four sectors. The which-won-where pattern is examined as follows. The N rate at the vertex of the polygon in any sector is the nominal N rate that produces the highest yields in all environments that fall in that sector. Thus, N-200 produced the highest yields in five environments: 00BRU, 99ROS, 99GBF, 99PEI, and 00HYC; N-150 produced the highest yields in two environments: 00PEI and 98CEF; and N-50 resulted in the highest yields in four environments: 00CEF, 00GBF, 99BRU, and 99CEF. The N-100 rate was not the best N rate for any of the environments. All statements based on Fig. 2 can be verified from the data (Table 2). Little yield variation can be seen among N rates in environments that were located near the biplot origin (e.g., 98CEF and 99CEF, Fig. 2 and Table 2).



View larger version (20K):
[in this window]
[in a new window]
  		Fig. 2. A polygon view of the GGE biplot of N x environment two-way data, showing which N level had the highest yields in each of the environments. PC1 and PC2 are first and second principal components, respectively.

 
That higher N application rates led to lower yields in four environments, especially at 00CEF, was most likely associated with the occurrence of Fusarium head blight and/or other foliar diseases as measured in two of the four environments (Table 3). At both sites, severity of Fusarium head blight and foliar diseases was significantly different among cultivars and N rates. Two cultivars (AC Barrie and AC Brio) had consistently lower Fusarium head blight scores while SS Blomidon had the highest disease score. Fertilizer N application exacerbated the disease severity under the wet situation: The higher level of N increased the incidence of Fusarium head blight and foliar diseases, and thereby reduced grain yield, as also reported by others (Gilbert and Tekauz, 2000; Reid et al., 2001). Harvest indices in 00CEF were lower (with an average value of 0.15) than other site-years (e.g., 99CEF had average harvest index value of 0.37). When 00CEF was excluded from analysis, the regression coefficient of yield against N application rates increased from 0.5 kg of wheat grain per kg of N (P = 0.223) to 1.3 kg of grain per kg of N (P = 0.002) even though it is still questionable whether this magnitude of N response can justify N fertilization of HRSW (the gain from increased wheat yield may not offset the fertilizer cost) in eastern Canada. However, N application may have a greater effect on grain protein concentration as reported previously (Fowler et al., 1990; Selles and Zentner, 2001; Ma, unpublished, 2003). To meet the minimum protein levels for bread-making quality, a greater amount of fertilizer N than required for economic yield may be needed for certain varieties, as demonstrated previously in Quebec (Ayoub et al., 1995).


View this table:
[in this window]
[in a new window]
  		Table 3. Average tombstone seeds (%), black point seeds (%), and harvest index (HI) of 10 cultivars and four N rates at Central Experimental Farm (CEF) and Greenbelt Research Farm (GBF) in 2000.

 
Visualization of the Genotype x Environment Two-Way Data
Yield mean, expressed as the best linear unbiased predictor of yield (Littell et al., 1996), for each of the 10 cultivars in each of the 11 environments is presented in Table 4. The environment main effect, genotype main effect, and genotype x environment interaction explained 86, 3, and 11%, respectively, of the total variation. The AMMI1 biplot explained 95% of the total variation (Fig. 3). For the environment main effects, Fig. 3 is arranged exactly the same as Fig. 1. Differences in mean grain yield (i.e., main effects) among genotypes were generally small, with AC Brio and Quantum being the highest- and Celtic the lowest-yielding cultivars. Apparent genotype x environment interactions existed, as represented by the y-axis (IPC1). The most important interactions were positive interactions between genotype AC Walton (and to a lesser extent, SS Blomidon) and environments 00PEI and 99PEI and the negative interactions between these two genotypes and environments 00CEF and 00GBF (Fig. 3).


View this table:
[in this window]
[in a new window]
  		Table 4. Mean grain yield (kg ha–1), expressed as the best linear unbiased predictors (Littell et al., 1996) of spring wheat listed in genotype x environment two-way data.{dagger}

 


View larger version (18K):
[in this window]
[in a new window]
  		Fig. 3. AMMI1 biplot of the genotype x environment two-way data. Both axes are in square root of kilograms per hectare. IPC1, first-interaction principal component; E, environment; G, genotype.

 
Although an AMMI1 biplot displayed 95% of the total variation for the N x environment two-way data, it does not explicitly reveal the which-won-where pattern. The GGE biplot (Fig. 4 and 5) based on the genotype x environment two-way data (Table 4) explained 76% of the total variation of the environment-centered data (i.e., GGE). A GGE biplot could be visualized from various aspects (Yan, 2001, 2002; Yan et al. 2000; Yan and Kang, 2003), some of which are detailed below.



View larger version (29K):
[in this window]
[in a new window]
  		Fig. 4. Environment vector view of the GGE biplot for the genotype x environment two-way data, showing interrelationships among the environments in terms of their discrimination of the genotypes. PC1 and PC2 are first and second principal components, respectively. Numbers 1–5 in the squares indicate five groups of environments.

 


View larger version (32K):
[in this window]
[in a new window]
  		Fig. 5. Polygon view of the GGE biplot for the genotype x environment two-way data, showing which genotypes yielded most in each of the environments. PC1 and PC2 are first and second principal components, respectively. Numbers 1–4 in the squares indicate four groups of genotypes.

 
Five Groups of Environments
The much larger genotype x environment interaction (11%) relative to the genotype main effect (3%) led to strong crossover genotype x environment interactions, as evidenced by the fact that environmental PC1 scores took different signs and the environments fell in all quadrants (Fig. 4). The 11 environments fell into five apparent groups in terms of their discrimination of the genotypes (Fig. 4). Group 1 consists of 00CEF and 00GBF; Group 2 consists of 99BRU, 99ROS, and 99GBF. Group 1 and Group 2 were negatively correlated as evidenced by the large angle between them. As a result, genotypes that yielded well in Group 1 would yield poorly in Group 2, or vice versa. Group 3 includes a single environment, 00PEI; Group 4 includes a single environment, 99PEI; and Group 5 consists of 98CEF, 99CEF, 00HYC, and 00BRU. Group 5 was located near the biplot origin, indicating that this group of environments was not discriminating of the genotypes, i.e., all genotypes yielded similarly in these environments. Group 3 (00PEI) was more or less independent of other groups except Group 4 (99PEI). Due to the complexity of the environments, no genotype performed well in all environments, with the result that the genotype main effects were much smaller than the genotype x environment interaction.

Four Groups of Genotypes
The GGE biplot revealed four apparent groups of genotypes in terms of their response to the environments (Fig. 5). Group 1 consisted of six genotypes, namely, Aquino, Grandin, AC Barrie, Celtic, AC Brio, and AC Voyageur; Group 2 consisted of two genotypes, AC Walton and SS Blomidon; and Groups 3 and 4 each consisted of a single genotype, Quantum and AC Drummond, respectively. Genotypes in Groups 1 and 2 were located opposite to each other, with reference to the biplot origin; they, therefore, responded to the environments in opposite directions. The same can be said for Quantum and AC Drummond.

The Highest-Yielding Cultivars in Each Environment
Figure 5 indicates that genotype Celtic (genotype Group 1) yielded the best in Group 1 environments (00CEF and 00GBF); Group 2 genotypes (AC Walton and SS Blomidon) were the best or near the best in 6 of the 11 environments, namely, 00PEI, 99PEI, 99BRU, 99BRO, 99GBF, and 98CEF, which consisted of environment Groups 2, 3, and 4; Quantum (genotype Group 3) was, or was close to being, the best in 00CEF and 00GBF (environment Group 1) and in 00PEI (environment Group 3); and AC Drummond was the best or near the best in environment Group 2 (99BRU, 99BRO, and 99GBF).

Although AC Walton and SS Blomidon were the highest-yielding cultivars in six environments, they were the poorest in 00CEF and 00GBF (Fig. 5). These negative interactions constituted the most important sources of genotype x environment interaction, which can be attributed to the fact that heavy Fusarium head blight epidemics and other foliar diseases occurred in 00CEF and 00GBF and that AC Walton and SS Blomidon were highly susceptible to Fusarium head blight and foliar diseases. When 00CEF and 00GBF were excluded from analysis, the relative magnitude of the genotype main effects increased whereas that of genotype x environment interaction decreased (data not shown).

Characterizing the Environments Using an Environment x Factor Biplot
To explain the overwhelming environment main effect and the interactions, monthly mean temperature, monthly total precipitation, and basic soil properties in each of the year–site combinations were collected (Table 5). An environment x factor biplot based on data in Table 5 is presented in Fig. 6. The environment vectors are drawn to facilitate visualization of the interrelationship among factors. When well represented by the biplot, the cosine of the angle between the vectors of two factors approximates the correlation coefficient between them. Thus, an angle of 90° means independent, an angle of >90° indicates negative correlation, and an angle of <90° suggests a positive correlation. Several groups of closely associated factors can be identified: mean monthly temperatures from April to July (T4 to T7) were closely associated; monthly precipitation in April and August (P4 and P8) were closely associated; and these two groups of factors are negatively associated. All soil properties except Bray-1 soil P were closely associated, reflecting the fact that fine-textured soils (with more clay and silt) were richer in organic matter, N, and K and had a higher pH. Soil sand percentage was obviously the inverse of soil clay and silt percentages. Figure 6 also shows that the weather factors and the soil factors were largely independent (orthogonal) as expected. The seemingly close correlation between sand percentage and precipitation in July (P7) may not show an apparent cause–effect relationship. However, such a correlation may be associated with the limited moisture-holding capacity of sandy soils, compared with those of higher silt and clay content, and with the relatively greater biological value of July rain.


View this table:
[in this window]
[in a new window]
  		Table 5. Grain yield, monthly mean temperature, monthly total precipitation, and soil characterization in each site-year combination.

 


View larger version (27K):
[in this window]
[in a new window]
  		Fig. 6. Vector view of an environment x factor biplot summarizing the interrelationship among the weather and soil factors presented in Table 4. PC1 and PC2 are first and second principal components, respectively. T4 to T8, monthly mean temperature in April to August; P4 to P8, monthly precipitation in April to August; K, soil test K (mm kg–1); N, soil total N concentration (%); OM, soil organic matter (%); soil-P, soil Bray-1 P (mm kg–1).

 
Environmental mean yield was not closely associated with any of the factors as yield has a short vector (Fig. 6). Based on the angles, yield may be positively associated with temperatures from April to July (T4 to T7) and negatively associated with precipitation in April (P4) and August (P8). Yield may also be positively associated with soil clay content, which was closely associated with soil K concentration. These observations were verified by the correlation coefficients among the factors (data not shown) although none of the correlations was significant, possibly due to the small number of environments. The trend displayed in the biplot, however, deserves further investigation.

The polygon view of the biplot (Fig. 7) revealed two sets of contrasts. The first was the contrast between the 1999 environments and the 2000 environments. All environments in 1999 except 99PEI were located on one polygon side; located on the perpendicular line to this polygon side were monthly temperatures from April to July (T4 to T7). This indicates that in 1999, all locations except PEI had equally high temperatures from April to July. All environments in 2000 were located on another polygon side, which was on the other side of the biplot origin. Precipitation in April and August (P4 and P8) was located near the perpendicular line to this polygon side, indicating that in 2000, all locations had more precipitation in April and August. Since the two polygon sides were almost parallel and on opposite sides of the biplot origin, the 1999 environments and the 2000 environments were sharply contrasted in terms of temperature from April and July and precipitation in April and August. The second contrast was between the location St. Bruno (BRU) and the locations in PEI and CEF. The former had a clay loam and the latter had a sandy loam.



View larger version (26K):
[in this window]
[in a new window]
  		Fig. 7. Polygon view of an environment x factor biplot to display the weather and soil patterns of the environments, based on data in Table 3. PC1 and PC2 are first and second principal components, respectively. T4 to T8, monthly mean temperature in April to August; P4 to P8, monthly precipitation in April to August; K, soil test K (mm kg–1); N, soil total N concentration (%); OM, soil organic matter (%); soil-P, soil Bray-1 P (mm kg–1).

 

   CONCLUSIONS
TOP
 ABSTRACT
 INTRODUCTION
 MATERIALS AND METHODS
 RESULTS AND DISCUSSION
 CONCLUSIONS
REFERENCES
 
This paper reports a N fertilizer study of HRSW over locations and years in eastern Canada. The yield variation was comprised mainly of the environment (year–site combination) main effect, a common observation in multienvironment trials for most field crops. Although neither N main effect nor the N x environment interaction was significant, interpretable crossover N x environment interactions were revealed through biplot analysis. The general trend was that higher N applications produced higher yield, but in two environments, higher N rates produced lower yields and reduced harvest index, associated with increased incidence of Fusarium head blight and other foliar diseases. Significant genotype main effect and genotype x environment interactions were detected. Quantum and Celtic were the highest- and lowest-yielding cultivars, respectively. The genotype x environment (interaction) was primarily due to the high-yielding, but Fusarium-susceptible, cultivars SS Blomidon and AC Walton. These two cultivars produced the highest yields in more than half of the environments but the poorest yields in 00CEF, in which Fusarium head blight and other foliar diseases interfered with dry matter allocation (low harvest index) due to excess precipitation around heading. This finding highlights the importance of Fusarium and foliar diseases in HRSW production in eastern Canada. Based on the yield, it appears that a 50 kg ha–1 N rate is optimum for most environments for majority cultivars tested. However, to ensure minimum protein levels for bread-making quality, at least 100 kg ha–1 N rate was required for some of the varieties (Ma et al., unpublished, 2003).

We demonstrated that biplots are useful tools for understanding complex agronomic data. The AMMI1 biplots allowed visual assessment of the treatment (N or genotype) and the environment (year–site combination) main effects. This biplot also displayed the treatment x environment interactions, but a GGE biplot was more effective in revealing the relationship among treatments in terms of their responses to the environment, the relationship among environments in terms of their discrimination of the treatments, and moreover, the crossover treatment x environment interactions. The environment x factor biplot provided a tool in characterizing the environments (revealing the most important patterns). This aspect of biplot application has not been reported in the literature. Although the weather and soil factors collected in this study were not able to explain the environment main effects and the treatment x environment interactions, the methodology described here should be useful to similar agronomic research data.


   ACKNOWLEDGMENTS
 
We wish to thank Dr. J. Yang and Dr. G. Butler for their careful and constructive reviews of this manuscript. The excellent technical assistance of V. Deslauriers, D. Balchin, L. Evenson, P. Matthew, and B. Wilson of the Eastern Cereal and Oilseed Research Centre (ECORC), Agriculture and Agri-Food Canada, is gratefully acknowledged. ECORC Contribution no. 02-103.


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




This article has been cited by other articles:


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]

Home page
 Crop Sci.Home page
K. L. Roozeboom, W. T. Schapaugh, M. R. Tuinstra, R. L. Vanderlip, and G. A. Milliken
Testing Wheat in Variable Environments: Genotype, Environment, Interaction Effects, and Grouping Test Locations
Crop Sci., January 16, 2008; 48(1): 317 - 330.
[Abstract] [Full Text] [PDF]

Home page
 Agron. J.Home page
K. Girma, S. L. Holtz, D. B. Arnall, L. M. Fultz, T. L. Hanks, K. D. Lawles, C. J. Mack, K. W. Owen, S. D. Reed, J. Santillano, et al.
Weather, Fertilizer, Previous Year Yield, and Fertilizer Levels Affect Ensuing Year Fertilizer Response of Wheat
Agron. J., November 6, 2007; 99(6): 1607 - 1614.
[Abstract] [Full Text] [PDF]

Home page
 Crop Sci.Home page
H. G. Gauch Jr.
Statistical Analysis of Yield Trials by AMMI and GGE
Crop Sci., May 18, 2006; 46(4): 1488 - 1500.
[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 ISI Web of Science (8)
Right arrow Citing Articles via Google Scholar
Google Scholar
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Right arrow Search for Related Content
PubMed
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Agricola
Right arrow Articles by Ma, B. L.
Right arrow Articles by Nass, H.
Related Collections
Right arrow Best Management Practices
Right arrow Other Crop Management
Right arrow Wheat

HOME HELP FEEDBACK SUBSCRIPTIONS ARCHIVE SEARCH TABLE OF CONTENTS
The SCI Journals Crop Science 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