| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a Dep. of Plant Science, Nova Scotia Agricultural College, P.O. Box 550, Truro, NS, B2N 5E3 Canada
b Atlantic Food and Horticulture Research Centre, Agric. and Agri-Food Canada, 32 Main Street, Kentville, NS, B4N 1J5 Canada
c SCPFRC, Agric. and Agri-Food Canada, 43 McGilvray St., Univ. of Guelph, Guelph, ON, N1G 2W1 Canada
gatlin{at}cadmin.nsac.ns.ca
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
variance was subdivided into a genotype x region (or subregion) variance 
, and a within-region or -subregion 
2GL. No 2GS was observed within the eastern or western regions, and genotypic correlations across subregions within regions approached 1.0. Significant 2GS was observed for eastern versus western Canada, but the correlation between genotypic effects across these regions was 0.83. In a selection experiment, subdivision of the eastern or western regions did not increase response. Selection in the east produced greater yields in both the east and west. The same genotype ranked first for yield in both regions. There was little specific adaptation to subregions, and two-row barley genotypes were broadly adapted across northern North America. Further subdivision of the regions is unwarranted, and selection in either region is likely to result in response in the other. The lack of local adaptation indicates that breeding programs that test broadly are likely to outperform ones that are narrowly targeted.
Abbreviations: DH, doubled haploid lines • GE, genotype x environment • GEI, genotype x environment interaction • GLY, genotype x location x year • GS, genotype x subregion • QTL, quantitative trait loci
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
, and (iii) the relative precision with which means are estimated within and across subregions.
Atlin et al. (2000) showed that subdivision will only increase response if genotype x subregion interaction is large relative to 2G. Under this circumstance, rG' is low, and the increase in 2G
due to subdivision is likely to be sufficient, relative to the genetic variance in the undivided region, to counterbalance the loss of precision resulting from reduced within-subregion testing. If there is little 2GS, subdivision will reduce response.
This model can be used to assess the delineation of target regions. Canadian barley breeding is organized on a regional basis to produce cultivars with specific adaptation to major production areas. Most programs recognize eastern and western Canada as separate regions. Some programs also treat subregions within the east and west as separate targets. There is evidence that some of these divisions may be unnecessary. Some cultivars perform well in both east and west, although the regions differ substantially in rainfall, diseases, and soil type (Kong et al., 1994). Little 2GL has been reported within the Maritimes (Atlin and McRae, 1994) and Quebec (Bernier-Cardou et al., 1983). The magnitude of 2GL for cereal yields within western Canada has not been extensively studied. In a study of hard red spring wheat (Triticum aestivum L.) cultivars tested over nine sites and 5 yr, Baker (1969) reported that 2GL was small relative to 2GLY and 2G. May and Kozub (1993), in a study of 11 genotypes tested over nine sites in two 3-yr periods, found that GL interaction was significant in only one of these periods, whereas genotype x location x year (GLY) effects were highly significant in both. These results are consistent with the hypothesis that there is limited local adaptation to specific areas within eastern and western Canada.
The apparently limited 2GL in barley, and the observation by Kong et al. (1994) that cultivars can be identified that perform well in both the east and west, indicate that germplasm used in Canadian barley breeding programs is broadly adapted. Because 2GL appears to be limited, 2GS is also likely to be relatively small over large regions. As a result, subdividing Canada's major barley-growing areas into small subregions may not result in an increase in 2G that is large enough to offset the reduced precision expected to result from a subdivision of testing resources. Breeding programs that test at many sites over a broad area may thus be more successful than programs that attempt to finely calibrate cultivars for local adaptation. To test this hypothesis, we studied the magnitude of 2GS relative to other components of variance in a set of doubled haploid (DH) lines from the cross Harrington/TR306 for mapping barley quantitative trait loci (QTL). The QTL analysis of this data set was originally reported by Tinker et al. (1996). The data are valuable for the study of GEI because they pertain to a large population of random lines derived without selection from an F1 between elite parents, and because the wide geographical range of sites at which evaluation was undertaken permits the testing of hypotheses concerning regional groupings of locations.
The specific objectives of this study were to: (i) assess the magnitude of 2GS for barley in Canada and measure the extent to which genotype performance is correlated across subregions, and (ii) determine if subdivision of eastern and western Canada into additional subregions is likely to increase response to selection.
|
Materials and methods |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
Grouping of Locations into Regions and Subregions
Trial locations in southern Ontario, Quebec, and PEI were grouped in an aggregate eastern region; all other trials were grouped in an aggregate western region. Locations were further grouped into two eastern and three western subregions on the basis of geographical proximity. Within four of the five subregions, individual locations were separated by no more than two degrees of latitude and longitude. Charlottetown, PEI, and Ste-Anne-de-Bellevue, Quebec, were placed in the same subregion although they are further apart because previous experiments have shown that barley cultivar means at these two sites tend to be highly correlated (unpublished data). Trial locations, listed by regional grouping, are presented in Table 1
. The number of trials varied among subregions and among years, ranging from three in Manitoba–North Dakota to seven in Saskatchewan.
|
 |
 |
 | (1) |
This differs from the conventional model for the analysis of multiple-environment trials (METs) in that the location term and its interactions are subdivided into components caused by the effect of regions or subregions and by the effect of locations within regions or subregions. The effect of a region or subregion is considered fixed because inferences are to be made about the specific partition being studied; regions or subregions are not considered to be random samples of the target area because they have been specifically constituted on the basis of environmental differences or genotypic response. All other terms are random. Locations are nested within regions or subregions because they are a random sample of farmers' fields within the area. In this study, subregions are widely distributed across the continent. A common year effect across both eastern and western Canada is not plausible; years are therefore nested within regions or subregions because it is assumed that annual weather patterns at sites within a region will be more similar than at sites in different regions.
For both the within-subregion analyses and the genotype x region or subregion analyses, variance components were estimated by the RE ML option of the SAS MIXED procedure (Littell et al., 1996). Because variance components were estimated from unweighted means of one-replicate trials in 1992 and two-replicate trials in 1993, the residual variance contains both 2GLY and a portion of 2, the within-trial plot variance.
Genotypic Correlations among Regions or Subregions, and between Genotypic Effects Estimated in Undivided Target Regions and Their Constituent Subregions
Correlations of genotypic effects among regions or Subregions i and i' (rG(ii')) were estimated as:
 | (2) |
G(i) and G(i') are genotypic standard deviations estimated from the within-subregion analyses for regions or Subregions i and i', respectively. The covariance of line means across regions or subregions is an estimator of the genotypic covariance if environment and genotype x environment effects in different regions or subregions are independent. Some of these correlation estimates slightly exceeded 1.0 because of sampling error; these have been presented as 1.0 in Table 4
.
|
between genotypic effects in undivided regions and in their constituent subregions was estimated according to the method of Atlin et al. (2000):
 | (3) |
2G and 2GS are the genotypic and genotype x subregion variances, respectively, from the analysis across subregions. Variance components and their standard errors were estimated by the RE ML option of the SAS MIXED procedure.
The Effect of Subdivision of Regions on Response to Selection
The effect of subdivision of eastern and western regions on within-subregion response to selection was evaluated through the use of an empirical selection experiment, in which the 1992 trials were considered selection environments and response was evaluated in the 1993 trials. In each of the five subregions, in the aggregate eastern and western regions, and in the entire data set, the 14 highest-yielding lines [selection intensity (i) = 0.097] were selected in 1992. The mean yield of each of these eight selected groups was determined within each subregion, and in the aggregate eastern and western regions, in 1993. In any subregion, direct response was considered to be the mean yield in 1993 of the 14 lines selected in that subregion in 1992 minus the overall mean of the 145 lines in that subregion. Within each of the five subregions, direct response was compared with indirect response to selection in the aggregate eastern and western regions, and to indirect response to selection based on the mean yield over all sites in 1992. Direct response in each of the aggregate regions was compared with indirect response to selection in the other region, and to indirect response to selection based on the mean yield over all sites. The standard error (SE) of mean selection response was estimated as:
 |
2GL, 2GY, and 2GLY are the genotype x location, genotype x year, and genotype x location x year variance components from the within-region or subregion analysis, and l and y are the numbers of locations and years (1 yr in all cases) upon which estimates of response are based. |
Results and discussion |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
|
2GY and 2GL were smaller than 2G (Table 2). In all regions, the random residual component accounted for the bulk of the phenotypic variance. The division of the entire set of locations into eastern and western regions may be warranted from the standpoint of increasing response to selection in the east; 2G in the east was approximately twice as large as 2G estimated for the entire data set.
Cockerham (1963) noted that there is a flux between the genotypic and genotype x environment variances that is dependent on the size of the target region. In very large and diverse regions, genotype x location interaction is expected to be large relative to the genotypic variance. In the other extreme case, for a target region consisting of a single location, the estimate of genotypic variance contains the entire genotype x location interaction. Comstock and Moll (1963) pointed out that conversion of genotype x location interaction into genotypic variance that can contribute to selection response is the main reason for the subdivision of large breeding targets. If genotype x subregion interaction is large, then subdivision will be effective in increasing genetic variance in the subregions relative to the original undivided area. Subdivision of the entire country into eastern and western regions did appear to marginally reduce within-region 2GL (Table 2), although, within western Canada, this source of variation remained large relative to the genotypic variance . Further subdivision of the eastern and western regions did not usually result in increased within-subregion genotypic variances; only the Manitoba–North Dakota subregion, in which only three environments were sampled, had a genotypic variance that was greater than that of the region of which it is a part (Table 2). Nor did subdivision reduce the pooled within-region genotype x location interaction (2GL) (Table 3) . On this basis, subdivision of the eastern and western regions appears unhelpful.
|
2GS was relatively small, and can be largely explained by the difference in the magnitude of genotypic variances in the two regions. The magnitude of V(G(env)), the component of 2GS accounted for by heterogeneity of genotypic variance, was calculated according to Cockerham (1963) as 58.4, or about 60% of 2GS. Thus, most of the genotype x region interaction for yield in eastern versus western regions is due to heterogeneity of genotypic variance, rather than to genotypic rank change. These estimates do not strongly support the division of Canadian barley-producing regions into broad eastern and western regions, and show that there is little to be gained by further subdivision.
Correlations among Regions and among Subregions
Significantly positive phenotypic correlations were observed among all pairs of subregions, except Ontario and Alberta (Table 4). Genotypic correlations were close to 1.0 for the two eastern subregions and for both eastern subregions with Manitoba–North Dakota. In the west, a high value for rG(ii') was also observed between Saskatchewan and Alberta and between Saskatchewan and Manitoba–North Dakota. The size of rG(ii') between pairs of subregions within the east and west supports the conclusion drawn from the variance component analysis, namely, that there is no advantage from further subdivision of the eastern and western regions. The overall rG(ii') between eastern and western means was 0.83, indicating that there is a strong positive relationship between yield potential in eastern and western Canada in this population of lines.
The pattern of correlations indicates that for at least the two years sampled, performance in Manitoba–North Dakota was more closely related to performance in eastern Canada than to Alberta and Saskatchewan. Grouping Manitoba–North Dakota with the eastern regions left the east-west rG(ii') essentially unchanged (data not shown). In general, the patterns of correlation are only moderately supportive of the traditional breakdown of Canadian barley breeding and cultivar evaluation programs into eastern and western domains separated at the border between Manitoba and Ontario. Relatively large genetic correlations were observed among all subregions except southern Ontario, which was more strongly linked to eastern than western subregions. The high levels of rG(ii') indicate that selection in any subregion, if adequately replicated, is likely to produce a favorable response in the others, and that it may be possible to select cultivars that are well adapted across Canada. This hypothesis is supported by the fact that the same genotype, based on mean yield over all sites and both years, ranked first in both the eastern and western regions (data not shown).
Correlations Between Genotypic Effects Estimated in Large Areas and in Their Constituent Subregions (rG')
Estimates of rG' derived from Eq. [3] were 1.0 for both the eastern and western regions with respect to their constituent subregions. The perfect genetic correlation estimated between means in the larger regions and in the subregions makes it unlikely that further subdivision of the eastern and western regions will result in increased selection response. This could only occur if subdivision increased heritability (Atlin et al., 2000). The absence of 2GS within east and west, and the reduced replication over locations that accompanies subdivision of a testing region, make this improbable.
For genotypic effects estimated for all of Canada relative to genotypic effects in the east or west only, rG' was 0.94 if estimated according to the variance components reported in Table 3 corrected for heterogeneity of within-region 2G (Cockerham, 1963), or 0.91 if estimated according to Atlin et al. (2000) as the square root of rG(East.West) (Table 4). These correlations indicate that most response to selection based on means over all locations will be expressed within both the eastern and western regions.
The Effect of Subdivision of Target Regions on Response to Selection
In general, subdivision of the eastern and western target regions into smaller subregions did not result in increased within-subregion selection response (Table 5)
. Direct response was greater than twice its standard error within only two (Ontario and Manitoba–North Dakota) of the five individual subregions. Selection based on eastern means resulted in a numerically greater response than did direct selection in all subregions except Ontario. Selection based on western means produced a numerically greater response than direct selection in three of five subregions. Interestingly, selection based on eastern means produced a substantially greater response in the western region as a whole than did direct selection based on western means, providing striking confirmation that it is sometimes more effective to select outside a target region than within it. The inferiority of direct selection in the west to indirect selection in the east was presumably because genotypic variances were lower in the west than in the east, and random genotype x environment variation was greater.
|
|
Conclusions |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
These results demonstrate the necessity of considering both precision of estimation of genotype means and the magnitude of 2GS when deciding whether or not to subdivide a target region. This variance, which is the only component of GEI that can be exploited to increase selection response, is often likely to be a relatively small fraction of 2GL. This is because 2GL may be due to heterogeneity of within-subregion genotypic variances (Cockerham, 1963) and/or to random location-to-location variation in genotype ranks within subregions (Atlin et al., 2000) rather than to local adaptation. GEI analyses that do not disaggregate these constituents of 2GL may overstate the benefits of selecting for local adaptation.
Our findings are strong evidence that the broad geographical adaptation of spring wheat (Braun et al., 1992; Cooper et al., 1993) is also characteristic of two-row barley. The same line was the highest-yielding genotype in both eastern and western regions, indicating that it is feasible to select cultivars with continent-wide adaptation. Many plant breeding companies and international agricultural research centers have recognized the relatively limited importance of genotype x region interaction in cereals and other annual crop species, and have structured their breeding programs to exploit the economies of scale that result from breeding for broad adaptation. This breeding strategy has the potential to accelerate genetic erosion, leading farmers to abandon indigenous cultivars for broadly adapted high-yield varieties (HYV). Small-scale, decentralized, farmer-participatory breeding and variety selection programs have been advocated to reduce genetic erosion by improving indigenous germplasm and exploiting local adaptation (e.g., Maurya et al., 1988). In some cropping systems, such programs can increase crop genetic diversity when localized environmental constraints result in a failure of HYVs to perform well (Witcombe et al., 1996; Sthapit et al., 1996). This study indicates that in many other cropping systems, the degree of local adaptation may be limited. In these cases, small-scale breeding programs seeking local adaptation may require extensive within-subregion testing networks if their products are to be competitive with those from large-scale programs that select for broad adaptation.
|
ACKNOWLEDGMENTS |
|---|
Received for publication July 1, 1998.
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMaterials and methodsResults and discussionConclusionsREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  S. Ceretta and F. van Eeuwijk Grain Yield Variation in Malting Barley Cultivars in Uruguay and Its Consequences for the Design of a Trials Network Crop Sci., January 16, 2008; 48(1): 167 - 180. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. B. Blanche and G. O. Myers Identifying Discriminating Locations for Cultivar Selection in Louisiana Crop Sci., February 24, 2006; 46(2): 946 - 949. [Abstract] [Full Text] [PDF]
|
|
|
|
 |
|
 D. Soleri and D. A. Cleveland Scenarios as a Tool for Eliciting and Understanding Farmers' Biological Knowledge Field Methods, August 1, 2005; 17(3): 283 - 301. [Abstract] [PDF]
|
|
|
|
|
|
W. Yan and N. A. Tinker An Integrated Biplot Analysis System for Displaying, Interpreting, and Exploring Genotype x Environment Interaction Crop Sci., May 6, 2005; 45(3): 1004 - 1016. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
G.N. Atlin, R.J. Baker, K.B. McRae, and X. Lu Selection Response in Subdivided Target Regions Crop Sci., January 1, 2000; 40(1): 7 - 13. [Abstract] [Full Text]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 | |||
