Published in Crop Sci. 44:1519-1526 (2004).
© 2004 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
CROP BREEDING, GENETICS & CYTOLOGY
Selection Environments for Maize in the U.S. Western High Plains
Fernando R. Guillen-Portala,*,
W. Ken Russellb,
Kent M. Eskridgec,
David D. Baltenspergerd,
Lenis A. Nelsonb,
Nora E. D'Croz-Masonb and
Blaine E. Johnsone
a Northwestern Agricultural Research Center, 4570 MT 35, Kalispell, MT 59901
b Dep. of Agronomy and Horticulture, Univ. of Nebraska, Lincoln, NE 68583
c Dep. of Biometry, Univ. of Nebraska, Lincoln, NE 68583
d Panhandle Research and Extension Center, Univ. of Nebraska, Scottsbluff, NE 69361
e Pioneer Hi-Bred Int., A DuPont Company, 19456 State Hwy. 22, Mankato, MN 56001
* Corresponding author (fguillen{at}montana.edu).
 |
ABSTRACT
|
|---|
Dryland maize (Zea mays L.) production in the U.S. western High Plains is hampered by variable yields because of substantial environmental variation in this region. This study was conducted to determine the degree to which the ranking of superior maize hybrids for dryland production in the western High Plains was predictable from performance of the same hybrids in highly productive, irrigated environments in the same region. Forty-five maize hybrids were evaluated for grain yield performance under different water regimes in western Nebraska, eastern Wyoming, and northeastern Colorado in 1998 and 1999. The value of genotypic variance was by far larger in fully irrigated test environments (0.70) than in nonirrigated test environments (0.01–0.17). The genotypic mean repeatability in fully irrigated test environments (0.63) compared with that in nonirrigated test environments (0.18–0.69, respectively), and it showed correspondence with yield performance. The genetic correlation between fully and nonirrigated environments (0.72) was lower than that observed between all-nonirrigated environments (0.78–1.02). Thus, the proportion of direct advance in the former case (0.63) was generally lower than in the latter (0.41–0.97). However, an environmental similarity ratio (ESR) derived from crossover interaction indicated that water-contrasting environments were as similar (ESR = 0.53) as nonirrigated environments (ESR = 0.49) in ranking the maize hybrids. Selective identification of maize hybrids in irrigated environments for production under nonirrigated environments in the western High Plains might be a useful surrogate to direct selection in the latter environments.
Abbreviations: COI, crossover interaction • Dj, pairwise differences in grain yield between genotypes in environment j • ESR, environmental similarity ratio • GEI, genotype-by-environment interaction • PDA, proportion of direct advance
|
INTRODUCTION
|
|---|
DRYLAND MAIZE PRODUCTION in the western High Plains of the USA increased dramatically from 1994 to 2001. In 2001, approximately 31 000 ha of nonirrigated land in the Nebraska Panhandle were planted to maize, 14 times greater than in 1994 (Nebraska Agricultural Statistics Service, 2002). Insufficient moisture during late July, when early maturing maize hybrids reach the flowering stage, is usually the primary limitation to grain yield in this region of the USA. Also, variable amounts and an uneven distribution of precipitation from May through August result in large variations in average grain yield, thereby increasing the uncertainty of profitable maize production in the area. The use of maize cultivars with specific adaptation to growing seasons with marginal water might lead to somewhat greater yields when these seasons do occur.
Success in breeding programs requires evaluation environments that are representative of the target population of environments (Allen et al., 1978). However, the problem with testing in marginally productive environments when breeding for good performance in these environments is that the ratio of genetic to nongenetic variance frequently is less than in highly productive environments (Atlin and Frey, 1989; Bänziger et al., 1997; Rosielle and Hamblin, 1981). In fact, when productivity is extremely low, it is not even possible to discriminate selectively among genotypes. Because of this, and the often observed moderate-to-high correlation of genotypic yield performance across a wide range of seasonal water amounts, some researchers have recommended that breeding for stress tolerance should be performed under optimal conditions (Byrne et al., 1995; Rajaram et al., 1996; Specht et al., 2001). It also has been asserted that breeding for stress tolerance under optimal conditions permits an efficient allocation of the resources available (Allen et al., 1978; Glaz et al., 1985). Others have contended that breeding for stress should be performed under conditions that are representative of the target environment (Van Oosterom et al., 1993; Ceccarelli and Grando, 1996).
Typically, these contrasting recommendations arise from how one views the strength of the genetic correlation between stress and nonstress environments and the comparative degree of heritability for environments of high vs. low productivity. The direction of correlated response is evidently determined by the value of the genetic correlation between the test and target environments (Allen et al., 1978; Rosielle and Hamblin, 1981). Rosielle and Hamblin (1981) assessed the impact of selection for per se mean yield over test and target environments as opposed to selection for a minimal difference between the two environmental types. They concluded that the former strategy would increase mean yields in both types of environments if the genetic variance in the target environments is small compared with that in the test environments and also if the genetic correlation between them is not highly negative.
Crossover interaction (COI) is that part of the genotype x environment interaction (GEI) that is attributable to changes in genotypic rank among environments. Baker (1988a)(1990), Crossa (1990), and others recognized that COI is the most intricate type of GEI with respect to identifying the best genotypes in a selection program. Gail and Simon (1985) developed a statistical test for COI between two treatments evaluated in a number of independent trials. Baker (1988a) considered this method to be particularly appropriate for GEI analysis. Russell et al. (2003) used a measure of distance between two environments, on the basis of the Gail-Simon definition of COI, to cluster environments such that COI within clusters was minimized. This distance measure was a more direct means (vis-à-vis the genetic correlation) of determining the degree to which genotypic rank in one set of environments predicted the genotypic rank in a second set of environments.
To our knowledge, no studies on environment selection for the improvement of dryland maize production in the U.S. western High Plains have been reported. The objective of our study was to determine the degree to which the ranking of superior maize hybrids for dryland production in the western High Plains was predictable from performance of those hybrids in the highly productive, irrigated environments in the same area. This issue was addressed not only by considering the genetic correlation between dryland and irrigated environments and the repeatability within each of these types of environments, but also by considering the amount of COI between these environmental types.
|
MATERIALS AND METHODS
|
|---|
Nine yellow dent commercial single-cross maize hybrids with desirable maturity for the Nebraska Panhandle [‘Cargill 3677’ (H1), ‘Cargill 4111’ (H2), ‘DeKalb DK449’ (H3), ‘DeKalb DK442’ (H4), ‘Mycogen 2500’ (H5), ‘Mycogen 2275’ (H6), ‘Mycogen 2395’ (H7), ‘Ottilie 2431’ (H8), and ‘Ottilie 2439’ (H9)] and their corresponding 36 double crosses (reciprocals excluded) were evaluated in this study. These 45 hybrids plus four nonrelated commercial hybrids were evaluated in 1998 and 1999 in five different locations of western Nebraska, eastern Wyoming, and northeastern Colorado under fully irrigated, partially irrigated, and nonirrigated water regimes (Table 1). The fully irrigated trial was placed at the Scottsbluff Agricultural Laboratory at Mitchell, NE, given that this facility has been historically involved in irrigated-maize evaluation activities for the western part of Nebraska.
View this table:
[in this window]
[in a new window]
|
Table 1. Key features of 12 production environments used for the evaluation of 45 single-cross and double-cross maize hybrids.
|
|
At each location, the hybrids were evaluated using a 7 x 7 partially balanced lattice design with three replications. The experimental unit was a two-row plot 7.6 m long with a row spacing of 0.76 m. The plant density for the fully irrigated trials was equivalent to 74 000 plants ha–1, and for the partially and nonirrigated trials was equivalent to 37 000 plants ha–1. In each trial, chemical fertilizer was applied at the recommended local rates. Weeds and pests were controlled using labeled pesticides. Data were collected on numerous traits (Guillen-Portal, 2000), but only grain yield, which was measured on a per plot basis and converted to megagrams per hectare after adjusting to 155 g kg–1 water content, was used in the analyses we report here. The yield data for the nonrelated hybrids were excluded. In each water regime, all effects except genotypes were considered random.
Predicted Correlated Response to Selection
Predicted correlated response to selection was assessed considering fully irrigated–nonirrigated, and nonirrigated–nonirrigated scenarios. The latter was considered a control scenario for comparison purposes. For the first scenario, the fully irrigated trial at Mitchell (averaged across years) was considered a test environment (X) and the trials located at Sidney (PI), Sidney (NI), Redington, PineBluffs, and Akron (each averaged across years) were considered a sample of the target population of environments (Y). For the second scenario, the trial at the partially irrigated location (averaged across years) and the trials at each of the nonirrigated locations (averaged across years) were considered test environments, and for each case the remainder of the locations (averaged across years) were considered a sample of the target population of environments. Fully irrigated trials were not included in the second scenario.
For each scenario and each environmental group, variance component estimates were obtained by equating observed mean squares to expected mean squares from the combined analyses of variance. Coefficients of the expected mean squares were estimated by the RANDOM option of the GLM procedure from SAS (SAS Institute, 1988). Standard errors of the variance component estimates were obtained following Anderson and Bancroft (1952). Repeatability of genotypic mean performance (i.e., heritability) was calculated on an entry mean basis following Cooper et al. (1993):
where K2G = mean square for genotypic effects, 
2GE = variance component for GEI, 2 = variance component for pooled experimental error, nE = number of environments, and nR = number of replications in each environment.
Standard errors of the mean repeatability were obtained following Gordon et al. (1972). Estimates of the genetic correlation between test and target environments 
were calculated following Falconer (1989) under the assumption that covariance attributable to environmental errors was negligible; i.e.,
where rPX,Y = phenotypic correlation between test and target environments, and hX and hY = the square roots of the genotypic mean repeatabilities in test and target environments, respectively. The proportion of direct advance (PDA) from indirect selection in test environments was defined following Falconer (1989) as:
Crossover Interaction Analysis
Tests for significance of COI were obtained by the approach of Gail and Simon (1985). Pairwise differences (Dj) in grain yield between genotypes in each environment (990 pairwise genotypic differences) and estimates of their standard errors (2j) were calculated by the LSMEAN/PDIFF statement/option from PROC MIXED of SAS (Littell et al., 1996). The test statistic for COI is min (Q+, Q–), where Q+ = 
for all Dj < 0 and Q– = for all Dj 
0. A likelihood-ratio test rejects the null hypothesis if min (Q+, Q–) > C, a critical value. Values of C (df = 12, 
= 0.05) were obtained from Table 1 of Gail and Simon (1985).
In each environment, only the grain yield differences between those pairs of genotypes displaying a statistically significant COI (P < 0.05) were identified. For these, if the ith genotype was significantly superior to the i'th genotype, an environmental score of "+" was assigned; otherwise, an environmental score of "–" was assigned. An ESR was then calculated for each pair of environments as ESR = (number of hybrid pairs for which the environmental score in each environment was a + or –)/the total number of hybrid pairs with a significant COI across all environments. Values of ESR could thus range from 1 (complete similarity in genotypic rank between the two environments) to 0 (complete dissimilarity).
|
RESULTS
|
|---|
Mean grain yield ranged from 9.63 to 12.01 Mg ha–1 in fully irrigated environments and from 2.10 to 6.42 Mg ha–1 in nonirrigated environments (Table 1). Mean grain yield in each partially irrigated environment fell in the range of the mean grain yields at the nonirrigated environments and was significantly less than the mean grain yield in either fully irrigated environment. Therefore, for all subsequent analyses, partially irrigated environments were treated as nonirrigated environments (coded as Sidney+).
Across all environments, the R2 value for the linear regression of mean environmental grain yield on the amount of environmental seasonal water was 53%. The seasonal water amount was the greatest at the two fully irrigated environments, which had the highest grain yields, and the least at Akron 1998, the environment with the lowest yield (Table 1). However, much of the variation in grain yield within the nonirrigated environments was not associated with total water. Among nonirrigated environments, the largest mean grain yield (Sidney 1998) was obtained under a fairly low amount of seasonal rainfall (280 mm), whereas mean grain yield at PineBluffs 1999, the second lowest grain yielding nonirrigated environment, was obtained under a relatively large amount of rainfall (420 mm). The incidence of pale western cutworm (Agrotis orthogonia Morrison) during the initial stages of growth affected the stand of several plots at Akron 1998 and 1999, and PineBluffs 1999. Also, a hard-pan soil layer at Akron 1999 caused irregular plant growth in several plots.
Predicted Correlated Response to Selection
Significant variation in grain yield among environments, hybrids, and GEI was found for both the fully irrigated and nonirrigated groups of environments (Table 2). In both environmental groups, environments were the largest source of variation (largest mean square). In the nonirrigated group, the year effect was minimal, with the location effect and the year-by-location interaction constituting most of the environmental variation. Most of the GEI variation arose from location x hybrid, and year x location x hybrid interactions with little, if any, year x hybrid interaction. The relative variation due to GEI, measured by the corresponding F-ratios between GEI and error mean squares, was 1.5 times larger for nonirrigated environments than for fully irrigated environments, justifying further examination of the GEI in the discrimination of genotypes across environments.
View this table:
[in this window]
[in a new window]
|
Table 2. Mean squares from the analysis of variance for grain yields produced by 45 maize hybrids in two fully irrigated (FI) and 10 nonirrigated (NI) environments.
|
|
The ratio of GEI variance to genotypic variance in fully irrigated test environments was 0.07, which compared with 0.18, 0.28, 0.79, and 3.8 in Sidney+, Sidney, Redington, and PineBluffs, respectively, of the nonirrigated test environments (Table 3). In general, the GEI variance was smaller in the test environments than in the target environments. Only in Redington and Akron was the opposite observed. Thus, Mitchell, the fully irrigated test environment, and Sidney+, and Sidney, two of the nonirrigated test environments appeared to show promise as selection environments because they allow maximum expression of genetic variability among hybrids and minimum interference because of variations in the environmental conditions.
View this table:
[in this window]
[in a new window]
|
Table 3. Variance component estimates (±SE) in fully irrigated (FI) test and nonirrigated (NI) test environments, and in corresponding target environments. Test environments are each location averaged across years (1998, 1999), target environments are 10 NI location x year combinations for the FI test environment, and eight NI location x year combinations for each of the NI test environments.
|
|
The genotypic mean repeatability in the fully irrigated test environment was 0.63, which was slightly smaller than that observed in the nonirrigated test environments Sidney+ and Sidney (0.67 and 0.69), but larger than the observed at Redington, PineBluffs, and Akron (0.53, 0.18, and 0.60, respectively) (Table 4). There appeared to be a positive correspondence between the mean repeatability values and the mean environmental yield values in the test environments. On the basis of these parameters as environment selection criteria, the best selection environments would be Mitchell, Sidney+, and Sidney. The genotypic mean repeatability in target environments was consistently larger than that observed in the test environments, ranging from 0.74 to 0.83.
View this table:
[in this window]
[in a new window]
|
Table 4. Average environmental yield and genotypic mean repeatability in fully irrigated (FI) test and nonirrigated (NI) test environments, and in corresponding target environments, and phenotypic and genetic correlation coefficients between grain yield means and proportion of direct advance (PDA) in target environments based on selections made in test environments.
|
|
The value of the genetic correlation between test and target environments for the fully irrigated–nonirrigated scenario was relatively high (0.72), but not higher than any of the values observed in the nonirrigated scenarios (ranging from 0.78–1.02) (Table 4). The PDA value observed for the fully irrigated–nonirrigated scenario (0.63) was higher only to the nonirrigated–nonirrigated scenario in which PineBluffs was the test environment (0.41). The highest PDA value of 0.97 observed in the latter scenario corresponded to Sidney+ as test environment. These results suggested that correlated response to selection in fully irrigated conditions for selection in nonirrigated conditions in the target area might not be as efficient as correlated response to selection practiced in nonirrigated conditions.
Crossover Analysis
Across all environments, COI was significant (P < 0.05) in 43 out of 990 possible genotypic pairwise differences (Table 5). With a 5% comparison-wise Type I error rate, the expected number of significant COI was 49.5. Because the observed number of significant COI was less than the number expected by chance alone, no strong evidence for COI in this data was found.
View this table:
[in this window]
[in a new window]
|
Table 5. Differences between 43 pairs of genotypes involved in a significant crossover interaction within environments. Differences greater than zero are termed "+", and differences smaller than zero are termed "–".
|
|
The lowest environmental similarity (ESR = 0.14) was observed between Sidney 1998 and Akron 1998, and the highest (ESR = 0.91) between the 1998 and 1999 year in Akron (Table 6). Nonirrigated environments were more similar to each other in 1999 than in 1998 with ESR values of 0.50 and 0.42. Among locations, PineBluffs and Akron, which were at the highest and lowest elevation outside the Nebraska Panhandle, showed the highest similarity (ESR = 0.64), and Sidney and Akron displayed the lowest similarity (ESR = 0.18). The fully irrigated and nonirrigated environmental classes were more similar in 1999 than in 1998 (ESR = 0.64 and ESR = 0.50). For specific fully irrigated vs. nonirrigated pairs, the highest similarity was between Mitchell and Redington (ESR = 0.70), and the lowest between Mitchell and PineBluffs (ESR = 0.43).
The ESR had an overall mean value of 0.50 with a standard deviation of 0.18. The statistical properties of this parameter are unknown, but, in this data set, the ESR presented a normal distribution (P > 0.46, Shapiro-Wilk test). The average ESR value between specific pairs of fully irrigated and nonirrigated environments was 0.53 (ranging from 0.16–0.86). This value compared with the average ESR of 0.49 between specific pairs of nonirrigated environments (ranging from 0.14–0.91) (Table 6), since they were not significantly different than the overall mean, based on a Z-test statistic (data not shown). These comparisons indicated that a random pair of nonirrigated environments was no more closely related than a random pair of fully irrigated and nonirrigated environments.
The average ESR values between each of the test environments and their corresponding target environments for both of the scenarios considered were also not significantly different from the ESR overall mean, except when Sidney was the test environment (ESR = 0.39) (Table 7). The correlation between the ESR and the average yield difference across genotypes between a pair of environments was only –0.11 when all pairs of environments were considered (Fig. 1A)
. However, when only fully irrigated- nonirrigated pairs of environments were considered, this correlation was –0.43, which was significant (Fig. 1B). These results suggested that the ranking of hybrids, and consequently the possibility of correctly extracting the best among them, would be similar between these scenarios, except for the case in which Sidney would be used as a test environment. Also, they indicated that the higher yielding nonirrigated environments were more similar to fully irrigated environments than were the lower yielding nonirrigated environments.
View this table:
[in this window]
[in a new window]
|
Table 7. Average environmental similarity ratio (ESR) derived from significant crossover interaction between fully irrigated (FI) test environments, nonirrigated (NI) test environments and corresponding target environments, and degree of commonalty between them.
|
|
View larger version (16K):
[in this window]
[in a new window]
|
Fig. 1. Scatter plot of environmental similarity ratio (ESR) vs. yield differences for maize of environmental pairs across 12 environments—(A) all possible pairs of environments and (B) only fully irrigated–nonirrigated pairs of environments. * Significant at the 0.05 probability level.
|
|
The number of common hybrids within the select group in the fully irrigated–nonirrigated scenario was five, whereas in the alternative scenario the number ranged from three (PineBluffs test environment) to six (Sidney+ and Sidney test environments) (Table 7). An examination of the select group in the best scenarios based on the ESR (Mitchell and Sidney+ test environments) showed four common hybrids (H1, H3, H6, and H9) within the select groups. However, in contrast to what would be expected, the examination of the fully irrigated–nonirrigated and the worst nonirrigated–nonirrigated scenarios (Sidney test environment in the latter case) showed five common hybrids (H1, H2, H3, H6, and H9) within the select groups. It is presumed that, at least in this study, the amount of COI observed among the environments did not significantly affect the ranking of those hybrids in the select group.
|
DISCUSSION
|
|---|
Genotypic variance component estimates were generally large in the trials proposed as test environments, and small in those proposed as target environments, which is in accordance with previous studies (Allen et al., 1978; Atlin and Frey, 1989). The positive correspondence between genotypic mean performance and genotypic mean repeatability found in this data set is also in agreement with results of previous studies (Atlin and Frey, 1989; Zavala-Garcia et al., 1992), and points to the Mitchell, fully irrigated environment and the Sidney+ and Sidney nonirrigated environments as potential selection environments for identification of the best hybrids in the target region.
Although our results on correlated response suggested that maximum response to selection would be obtained when hybrids are evaluated in environments similar to those in which they are to be produced, it should be noted that the number of trials in the nonirrigated target environments was five times larger than the number of trials in the fully irrigated test environments. The higher genotypic mean repeatability observed in target environments was not the result of an increased genotypic variation in these environmental groups but of the large difference in the number of trials within each environmental group. Atlin and Frey (1989) demonstrated that, in indirect selection studies, both the number of trial repetitions and the number of replications per trial within the test vs. those within the target environments might significantly affect the computed value of the genotypic mean repeatability and of the genetic correlation coefficient between test and target environments.
In our analysis of correlated response, the proportion of test environments to target environments was 1:5. Given that the fully irrigated environments were considered a random sample of the population of fully irrigated environments, we can assume that the variance components in this environmental group would not change largely if more environments were considered. If five fully irrigated environments had been sampled, then the value of the genotypic mean repeatability in this environmental group would be 0.82 and the DPA value would rise from 0.63 to 0.72. Regardless of the variation in the magnitude of correlated response as a result of changes in the number of trials in the test and target environmental groups, studies on indirect selection show that more important than the magnitude of correlated response to selection is the consistency in the identification of the best genotypes across environments (Baker, 1988; Gregorius and Namkoong, 1986; Shabana et al., 1980).
Although only 4% of the COI interactions were significant, the distribution of the observed COI was not consistent with an inference of the absence of COI. If the inclusion of a hybrid in a significant COI was simply a chance event, then the number of inclusions would ordinarily be distributed as a binomial variable, where the probability of occurrence was 49.5/990 and the maximal number of occurrences was 44. The number of times a given hybrid would be expected to be included in a significant COI by chance alone is given by the product of these two values, which equaled 2.2. However, several hybrids were involved in many more than this expected number of COI. Single-cross H8 and double-cross H3 x H9 each were involved in nine significant COI. The probability of this being a chance event was approximately 1 in 10 000. This unlikely event suggested that the observed COI were not entirely random events.
It is likely that the types of materials included (single and derived double crosses) have affected the manifestation of significant, qualitative GEI. It is known that genetically heterogeneous entries frequently display complex patterns of response to environmental fluctuations compared with genetically homogeneous entries (Bramel-Cox, 1996; Schnell and Becker, 1986). In this data set, the select group across environments based on grain yield consisted almost entirely of single-cross hybrids, as reported by Guillen-Portal et al. (2003). Thus, it is inferred that single-cross hybrids form a genotypic group characterized by exhibiting not only superior performance but also broad adaptability to the environmental conditions encountered in the U.S. western High Plains. Nevertheless, in a set composed entirely of materials with the same genetic structure, whether the ranking among entries with high productivity (i.e., those within the select group) is unaffected by COI is unclear and merits further investigation. The degree of environmental similarity differed slightly across years. Consistently, the similarities among locations were greater in 1999 than in 1998. Lack of repeatability is an issue of major concern when using GEI for breeding purposes (Baker, 1988b). As pointed out by DeLacy et al. (1996), the use of GEI in correlated responses will be of practical relevance only if the correspondence between environments is repeatable on a temporal basis. This aspect should be investigated in greater detail to identify the causal factors behind the patterns of environmental response found.
Our results on the change in the ranking of maize hybrids across highly contrasting environments indicated that in the given target area, pairs of fully irrigated and nonirrigated environments were as similar as pairs of nonirrigated environments. Thus, the selective identification of maize hybrids with superior yield performance in optimal, fully irrigated target area conditions would be an alternative to direct selection in dryland target area conditions.
|
ACKNOWLEDGMENTS
|
|---|
We want to express our gratitude to Mr. Lane Darnall at Redington, NE; Mr. Lance Theobold at PineBluffs, WY; and Dr. Randy Anderson from the USDA-Colorado State University Research and Extension Station at Akron, CO, for allowing the use of their premises for the implementation of some of the experiments; to Daryl Travniceck for the assistance in the development of the SAS algorithm for the analysis of crossover interaction; to Glen Frickel for conducting field research; to Christina Cox for collecting field data; and to the following seed companies for providing seed of the single-cross maize hybrids: Cargill Seed Company (Cargill hybrids); Monsanto Company (DeKalb hybrids); Dow Agro Sciences (Mycogen hybrids); and Ottilie RO Seeds (Ottilie hybrids). We also are grateful to Dr. James Specht for his valuable comments and discussions on the manuscript.
|
NOTES
|
|---|
Research partially funded by grant number 87301 from Pioneer Hi-Bred International, Inc. A contribution of the University of Nebraska Agricultural Research Division, Lincoln, NE 68583. Journal Series No. 14127.
Received for publication June 19, 2003.
|
REFERENCES
|
|---|
- Allen, F.L., R.E. Comstock, and D.C. Rasmusson. 1978. Optimal environments for yield testing. Crop Sci. 18:747–751.[Abstract/Free Full Text]
- Anderson, R.L., and T.A. Bancroft. 1952. Statistical theory in research. McGraw-Hill, New York.
- Atlin, G.N., and K.J. Frey. 1989. Predicting the relative effectiveness of direct vs. indirect selection for oat yield in three types of stress environments. Euphytica 44:137–142.[Web of Science]
- Baker, R.J. 1988a. Test for crossover genotype-by-environment interactions. Can. J. Plant Sci. 68:405–410.
- Baker, R.J. 1988b. Differential response to environmental stress. p. 492–504. In B.S. Weir et al. (ed.) Proceedings of the Second International Conference on Quantitative Genetics. Sinauer Associates Inc., Sunderland, MA.
- Baker, R.J. 1990. Crossover genotype-by-environment interaction in wheat. p. 42–51. In M.S. Kang (ed.) Genotype-by-environment interaction and plant breeding. Louisiana State University Agricultural Center, Baton Rouge, LA.
- Bänziger, M., F.J. Betran, and H.R. Lafitte. 1997. Efficiency of high-nitrogen selection environments for improving maize for low-nitrogen target environments. Crop Sci. 37:1103–1109.[Abstract/Free Full Text]
- Bramel-Cox, P.J. 1996. Breeding for reliability of performance across unpredictable environments. p. 309–339. In: M.S. Kang and H.G. Gauch, Jr. (ed.) Genotype-by-environment interaction. CRC Press, Boca Raton, FL.
- Byrne, P.F., J. Bolaños, G.O. Edmeades, and D.L. Eaton. 1995. Gains from selection under drought vs. multilocation testing in related tropical maize populations. Crop Sci. 35:63–69.[Abstract/Free Full Text]
- Ceccarelli, S., and S. Grando. 1996. Drought as a challenge for the plant breeder. Plant Growth Regul. 20:149–155.[Web of Science]
- Cooper, M., D.E. Byth, I.H. DeLacy, and D.R. Woodruff. 1993. Predicting grain yield in Australian environments using data from CIMMYT international wheat performance trials. 1. Potential for exploiting correlated response to selection. Field Crops Res. 32:305–322.
- Crossa, J. 1990. Statistical analysis of multilocation trials. Adv. Agron. 44:55–58.
- DeLacy, I.H., K.E. Basford, M. Cooper, J.K. Bull, and C.G. McLaren. 1996. Analysis of multi-environment trials–An historical perspective. p. 39–124. In M. Cooper and G.L. Hammer (ed.) Plant adaptation and crop improvement. CAB International, Oxon, UK, International Crops Research Institute for the SemiArid Tropics, India; and International Rice Research Institute, Philipines.
- Falconer, D. S. 1989. Introduction to quantitative genetics, 3rd ed. Longman Scientific and Technical, Essex, UK.
- Gail, M., and R. Simon. 1985. Testing for qualitative interactions between treatment effects and patient subsets. Biometrics 41:361–372.[Web of Science][Medline]
- Glaz, B., J.D. Miller, and M.S. Kang. 1985. Evaluation of cultivar testing locations in sugarcane. Theor. Appl. Genet. 71:22–25.
- Gordon, I.L., D.E. Byth, and L.N. Balaam. 1972. Variance of heritability ratios estimated from phenotypic variance components. Biometrics 28:401–415.[Medline]
- Gregorius, H.R., and G. Namkoong. 1986. Joint analysis of genotype and environmental effects. Theor. Appl. Genet. 72:413–422.[Web of Science]
- Guillen-Portal, F.R. 2000. Hybrid maize adaptability to drought stress assessed with combining ability and genotype-by-environment effects. Ph.D. diss. (Diss. Abstr. AA19991989) (Diss. Abstr. Int. 61B:5065). Univ. of Nebraska, Lincoln.
- Guillen-Portal, F.R., W.K. Russell, D.D. Baltensperger, K.M. Eskridge, N.E. D'Croz-Mason, and L.A. Nelson. 2003. Best types of maize hybrids for the western High Plains of the USA. Crop Sci. 43:2065–2070.[Abstract/Free Full Text]
- Baker, R.J. 1988. Tests for crossover genotype-environmental interactions. Can. J. Plant Sci. 68:405–410.
- Littell, R.C., G.A. Milliken, W.W. Stroup, and R.D. Wolfinger. 1996. SAS system for mixed models. SAS Institute, Cary, NC.
- Nebraska Agricultural Statistics Service. 2002. [Online]. Available at http://www.nass.usda.gov/ne/; verified 3 May 2004.
- Rajaram, S., H. Braun, and M. van Ginkel. 1996. CIMMYT's approach to breed for drought tolerance. Euphytica 92:147–153.
- Ritchie, S.W., J.J. Hanway, and G.O. Benson. 1993. How a corn plant develops. Special report No. 48, Iowa State Univ. Coop. Ext. Serv., Ames, IA.
- Rosielle, A.A., and J. Hamblin. 1981. Theoretical aspects of selection for yield in stress and non-stress environments. Crop Sci. 21:943–946.[Abstract/Free Full Text]
- Russell, W.K., K.M. Eskridge, D.A. Travniceck, and F.R. Guillen-Portal. 2003. Clustering environments to minimize change in rank of cultivars. Crop Sci. 43:858–864.[Abstract/Free Full Text]
- SAS Institute. 1988. SAS User's guide: Statistics. SAS Institute, Cary, NC.
- Shabana, R., T. Bailey, and K.J. Frey. 1980. Production traits of oats selected under low, medium, and high productivity. Crop Sci. 20:739–744.[Abstract/Free Full Text]
- Specht, J.E., K. Chase, M. Macrander, G.L. Graef, J. Chung, J.P. Markwell, M. Germann, J.H. Orf, and K.G. Lark. 2001. Soybean response to water: A QTL analysis of drought tolerance. Crop Sci. 41:493–509.[Abstract/Free Full Text]
- Schnell, F.W., and H.C. Becker. 1986. Yield and yield stability in a balanced system of widely differing population structures in Zea mays L. Plant Breed. 97:30–38.
- Van Oosterom, E.J., D. Klejin, S. Ceccarelli, and M.M. Nachit. 1993. Genotype-by-environment interactions of barley in the Mediterranean region. Crop Sci. 33:669–674.[Abstract/Free Full Text]
- Zavala-Garcia, F., P.J. Bramel-Cox, J.D. Eastin, M.D. Witt, and D.J. Andrews. 1992. Increasing the efficiency of crop selection for unpredictable environments. Crop Sci. 32:51–57.
This article has been cited by other articles:
|
|
|
|
 
D. Baxevanos, C. Goulas, J. Rossi, and E. Braojos
Separation of Cotton Cultivar Testing Sites based on Representativeness and Discriminating Ability Using GGE Biplots
Agron. J.,
August 11, 2008;
100(5):
1230 - 1236.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
