Published in Crop Sci 39:1671-1675 (1999)
© 1999 Crop Science Society of America
677 S. Segoe Rd., Madison, WI 53711 USA
Crop Science 39:1671-1675 (1999)
© 1999 Crop Science Society of America
CROP BREEDING, GENETICS & CYTOLOGY
Potential of Using Plant Row Yield Trials to Predict Soybean Yield
J.M. Hegstada,
G. Bollerob and
C.D. Nickellb
a Pioneer Hi-Bred Intl., 7230 NW 70th Ave., P.O. Box 177, Johnston, IA 50131-0177 USA
b Univ. of Illinois at Urbana-Champaign, Dep. of Crop Sciences, 1102 S. Goodwin, Urbana, IL 61801 USA
hegstajeff{at}phibred.com
 |
ABSTRACT
|
|---|
Plant row yield trials (PRYT) are measured as an indicator of yield potential for many private and public soybean breeding programs. To identify elite lines and develop new cultivars, the highest yielding PRYT lines are usually advanced to multiple replications in different environments. Early generation PRYT testing has the advantage of identifying elite lines in the initial phases of the selection process. The objective of this study was to determine if single row PRYT testing is a reliable predictor of yield in multiple environment advanced yield testing. In 1996, five F2:3 populations of crosses between different elite cultivars were grown as single row PRYT. After the 1996 PRYT test, the top 10, middle 10, and bottom 10 yielding PRYT lines from each population were selected for advanced testing in 1997 and 1998. Yield and yield rank correlations between selected 1996 PRYT and advanced yield testing were highest for the `Jack' x `Resnik' and `Asgrow A3733' x `Burlison' populations, respectively. Matrix analysis indicated that the lines selected from the Asgrow A3733 x Burlison population were the most stable when 1996 PRYT data are compared with advanced yield test data. Approximately five lines can be identified from each population that are the highest yielding in PRYT testing and advanced yield testing. Additionally, approximately five lines could be identified from each population that are the lowest yielding in PRYT testing and advanced yield tests. In the populations examined it can be concluded that early generation PRYT testing would allow for progress to be made in identifying elite soybean lines with high yield potential.
Abbreviations: PRYT, plant row yield trials
|
INTRODUCTION
|
|---|
THE LIMITED AMOUNT OF SEED HARVESTED
on a single F2 plant of soybean [Glycine max (L.) Merr.] often restricts F3 yield evaluation in unreplicated plots. Short-row plots were a valuable method of determining yield potential in early generations of testing (St. Martin et al., 1990). The objective of early generation testing is to obtain information on yield potential at the beginning of the selection process (Thorne, 1974). The success of early generation testing depends on the ability to distinguish differences between genotypes in early generations, and that those differences will persist during later generations of selection. Several reports on early generation testing in different crop species have conflicted in the success rate, and no consensus has been reached as to the effectiveness of the procedure. Working with wheat lines (Triticum aestivum L.), Knott and Kumar (1975) did not find significant differences for yield using an early generation selection scheme rather than a single-seed-descent procedure. Ntare et al. (1984) discovered that early generation testing was as effective as a single-seed-descent program in identifying cowpea lines [Vigna unguiculata (L.) Walp.] that were not significantly different for yield. However, other reports in several self-pollinated crop species have concluded that from one generation to the next, the yield correlation in early generation testing was lower than in other selection methods (Hamblin and Evans, 1976; O'Brien et al., 1978; Rahman and Bahl, 1986; Singh et al., 1990).
Early generation yield testing of soybean has the potential to identify lines that will produce high yield in later generations of advancement. Reports regarding the effectiveness of early generation testing to predict future yield in soybean have been inconsistent. Voight and Weber (1960) determined that lines selected by early generation testing were superior in yield to lines selected in bulk and pedigree methods. Leudders et al. (1973) discovered that lines selected from populations advanced by pedigree, early generation testing, and bulk methods were not significantly different. Early generation yield testing of bulk progeny was superior to pure line family method for identifying the top F2 families for advancement (Boerma and Cooper, 1975a; Ivers and Fehr, 1978). A different study by Boerma and Cooper (1975b) found single-seed-descent was more effective and efficient than early generation testing.
The above reports have concentrated on determining the effectiveness of early generation testing in comparison to other selection schemes. It is uncertain if early generation testing of PRYT would correspond to advanced yield testing. Plant row yield trials are measured as an indicator of yield potential for many private and public soybean breeding programs. Each PRYT line represents a unique genetic combination of a cross. A considerable amount of time and effort is expended each year in the maintenance and evaluation of PRYT plots. The typical practice is to maintain and harvest several thousand PRYT rows to identify top yielding lines. A small percentage of the highest yielding PRYT are advanced for further testing, while the remainder are discarded. It is therefore of interest to determine if PRYT that are the highest, middle, and lowest yielding in one year correspond to the highest, middle, and lowest yielding lines of advanced tests in subsequent years. The objective of this study was to determine the potential of using PRYT testing to predict yield in later generations of multiple environment advanced yield evaluations.
|
Materials and methods
|
|---|
Cultivars Asgrow A3733, Burlison (Nickell et al., 1990b), Edison (McBlain et al., 1991), Jack (Nickell et al., 1990a), Resnik (McBlain et al., 1990), and Thorne (McBlain et al., 1993) were crossed to generate five F2 populations (Table 1)
. The F2 populations were grown in 1995 at the University of Illinois Crop Sciences Research and Education Center, Urbana-Champaign, IL. From each population, approximately 100 individual plants were tagged and harvested separately. The F2:3, F2:4, and F2:5 generation were grown in 1996, 1997, and 1998, respectively, as single replicate PRYT in a randomized complete block design with parental cultivars as checks. PRYT rows consisted of 45 seeds, were 1.3 m long, and had 76 cm between row spacing. Cultivar Resnik was grown as a common border adjacent to each PRYT row to minimize neighbor row environmental effect. The agronomic traits measured were: yield (kg ha-1; adjusted to 130 kg ha-1 moisture), maturity (date when approximately 95% of the plants had mature pod color), moisture (%), and seed quality (score: 1 = good quality to 5 = poor quality).
View this table:
[in this window]
[in a new window]
|
Table 1 Variance component and broad sense heritability estimates for different populations of F2:3 soybean lines from the 1996 plant row yield trials (PRYT)
|
|
From the 1996 PRYT test, the top 10, middle 10, and lowest 10 yielding lines were selected within each population. In 1997 and 1998, the 150 selected lines and parental cultivars were grown in advanced yield tests with two replications in two different environments at the University of Illinois Crop Sciences Research and Education Center, Urbana-Champaign, IL. An environment is defined as the year and field location where the test was grown. Environment 1 was in 1997 at the south farm at Urbana, a Flanagan silt loam (fine, montmorillonitic, mesic Aquic Argiudoll) soil. Environment 2 was in 1997 at the Cruse farm at Champaign, a Flanagan silt loam soil. Environment 3 was in 1998 at the South Grein farm, in a field that was half Dana silt loam (fine-silty, mixed, mesic Typic Argiudoll) and half Elburz silt loam (fine-silty, mixed, mesic Aquic Argiudoll) soil. Environment 4 was in 1998 at the Cruse farm, a Flanagan silt loam soil. Advanced yield tests were four row plots, 3.5 m long, with 76-cm row spacing. The agronomic traits measured from harvesting the center two rows of each plot were: yield, maturity, height to top node (cm), lodging (score: 1 = all plants erect to 5 = all plants prostrate), moisture, and seed quality.
Agronomic data were subject to simple correlation analysis to determine correlation between PRYT and advanced yield testing. Variance component estimates for the 1996 environment were calculated to determine the potential effectiveness of selecting within each population. Variance components (VP, VE, VG) were calculated on a per-plot basis with based yield data from the entries and check cultivars for each population;
. VP is the phenotypic variance based upon yield variance for each population. VE is estimated by calculating the average yield variance of the check cultivars for each population. The VE will equal the VP because each check cultivar is genetically identical;
. VG for the population is estimated with the formula
. Broad-sense heritability (HB) was estimated for each population by the relationship
. Confidence intervals for heritability were calculated as described by Knapp et al. (1985).
Agronomic data from advanced tests were subject to analysis of variance (ANOVA) by PROC GLM, PROC MIXED and VARCOMP functions of SAS (Littell et al., 1996). In the materials examined, environments, replications (nested within environments), and populations were considered as random effects, and entries (nested within populations) were considered as a fixed effect. Significance was determined by a F-test as described by McIntosh (1983) and the numerator and denominator degrees of freedom were approximated as described by Satterthwaite (1946).
|
Results and discussion
|
|---|
The materials used in this study are public cultivars that were released for high yield and/or resistance characteristics. The cultivars are group II to group III in maturity and although they are from a narrow genetic base, they are variable for several different agronomic characteristics. In this study, cultivar Resnik was grown as a common border row between PRYT in each of the 3 yr. Because of a lack of space, this practice is usually not applied in private breeding programs that analyze thousands of PRYT rows. Because of the use of a border row, there may be a reduction in precision due to the use of a larger land area that may not be homogeneous across the field. However, it was hypothesized that the error associated with using a larger field would be less that the environmental error that would occur due to neighboring plot effects.
The interpretation of the data that would most closely resemble a cultivar development program would focus on using the 1996 PRYT data for comparison to the 1997 and 1998 advanced yield tests. Variance components for the 1996 PRYT were calculated to estimate the genetic (VG) and environmental (VE) components of the phenotypic variance (VP). Two populations, Jack x Resnik and Asgrow A3733 x Burlison, had estimates of VG that were greater than VE (Table 1). The estimates of broad sense heritability (HB) were 0.63 for the Jack x Resnik population, and 0.55 for the A3733 x Burlison population (Table 1). The other three populations had VE values which were greater than the VG estimates, resulting in lower HB estimates (Table 1). The variance estimates would suggest that there would be more success in distinguishing and selecting stable lines from the A3733 x Burlison and Jack x Resnik populations. It would be expected that the A3733 x Burlison and Jack x Resnik populations would have higher correlations between 1996 PRYT and the advanced yield testing because these populations had the highest HB estimates.
To determine the relative stability of the PRYT entries across environments, simple correlation analysis was completed from 1996 to 1998 for all single row PRYT
. In general, the 1997 PRYT data did not correlate as highly to the 1998 PRYT data as the 1996 PRYT data correlated to the 1998 PRYT data (Table 2)
. When PRYT performance is compared across the 3 yr of testing, the overall correlations for yield and yield rank were the highest for the A3733 x Burlison population (Table 2). In this population, the correlations for yield and yield rank across the three PRYT evaluations were significant at the 0.05 level. The A3733 x Burlison population would most likely have similar yield results across different environments of PRYT testing. The other populations had more variability in yield and yield rank correlation across the three environments of PRYT testing (Table 2). Although some of the correlation values are low, they are positive and would suggest that for the populations examined, selection of the highest, middle, and lowest yielding lines would be consistently similar in different environments.
View this table:
[in this window]
[in a new window]
|
Table 2 Simple correlation (r) analysis for yield and yield rank of plant row yield trials (PRYT) from 1996, 1997, and 1998
|
|
Mean squares were estimated for the advanced yield tests grown in 1997 and 1998. Mean square estimates show that the environments were significantly different at the 0.01 level for maturity, lodging, and seed quality at the 0.05 level for height (Table 3) . In this study, the environments were not statistically different for yield. The populations examined are significantly different at the 0.01 level for all the agronomic traits measured (Table 3). These populations should represent a random sample of soybean populations available, although some of the parental cultivars are in common. The entry (nested within population) and environment x population interaction is significantly different at the 0.01 level for all agronomic traits measured (Table 3). These results are not unexpected because the 30 entries from each population were selected on the basis of significant differences for yield.
View this table:
[in this window]
[in a new window]
|
Table 3 Mean square estimates for advanced yield tests (150 entries) evaluated in four environments in 1997 and 1998
|
|
Variance component estimates confirm the results of the mean square analysis by displaying which components account for the variance of each agronomic trait (Table 4)
. For yield, the population, environment x population, and entry (nested within population) account for more of the variance compared with the environment (Table 4). The data presented should therefore be applicable to yield testing results that would be obtained with different soybean populations.
View this table:
[in this window]
[in a new window]
|
Table 4 Variance component estimates for advanced yield tests (150 entries) evaluated in four environments in 1997 and 1998
|
|
Correlations between the selected single row 1996 PRYT and advanced yield testing were variable, dependant upon the population. The highest yield correlation between selected 1996 PRYT and the mean of the 1997 and 1998 advanced yield tests was for the Jack x Resnik population (Table 5)
. The yield correlation value for the Jack x Resnik population was significant at the 0.05 level (Table 5). The highest yield rank correlation between the 1996 PRYT and advanced yield tests was for the A3733 x Burlison population (Table 5). The yield rank correlation value for the A3733 x Burlison population was significant at the 0.05 level (Table 5). These data confirm what would be expected from the variance component and HB estimates from the 1996 PRYT environment (Table 1). Lines selected from PRYT data from the A3733 x Burlison and Jack x Resnik populations were more stable across the sampled environments compared with the other three populations. Across all 150 lines, the yield correlation between 1996 PRYT data and a mean of 1997 +1998 advanced testing data was 0.07 (Table 5). The yield rank correlation between the selected PRYT in 1996 and a mean of the 1997 + 1998 advanced tests was 0.15 (Table 5). These overall correlation values are low and not significant, most likely because the two populations with Edison as a parent were negatively correlated.
View this table:
[in this window]
[in a new window]
|
Table 5 Simple correlations (r) for yield and yield rank between selected 1996 plant row yield trials (PRYT) and multiple environment advanced tests
|
|
Yield means for the top, middle, and bottom 10 1996 PRYT were compared with a yield mean of the 1997 + 1998 advanced yield test. The mean values for the top, middle, and bottom 10 1996 PRYT rows are significantly different for all populations at LSD = 0.05 (Table 6)
. In addition, for all populations, the mean values of the top, middle, and bottom 10 lines in the advanced test are all significantly different at LSD = 0.05 (Table 6). Across the five different populations, the mean for the top 50 lines of the 1996 PRYT was significantly higher
than the mean for the middle 50 and lower 50 of the advanced test (Table 6).
Matrices for each population were created to determine the change in ranking of the lines from PRYT to advanced yield tests. Matrices were aligned with the entries for the top 10, middle 10, and bottom 10 1996 PRYT, and their respective order when compared with a mean for yield of the four environments in the 1997 + 1998 advanced tests. The relative stability of the lines is shown in the matrix by displaying the categorical changes of the 30 entries for each population. The total number of stable lines is calculated by adding the diagonal values from the upper left to the lower right of the matrix. Examining the matrices shows the A3733 x Burlison population had the most stability when 1996 PRYT yield data was compared against the yield mean of advanced testing (Table 7)
. Of the 30 lines selected in 1996, 17 did not change yield classification in this population and were the most stable across the different environments. The other four populations were more variable, with average ranges in stability from six of 30 for the Burlison x Edison population to 12 of 30 for the Jack x Resnik population (Table 7). Across all populations, 37% (55/150) of the lines did not change yield classification from 1996 PRYT testing to advanced yield testing (Table 7). There was a 44% (22/50) observed probability that the PRYT lines selected in the top 10% of each population were in the top third of the advanced yield tests (Table 7). Additionally, the probability that lowest 10 PRYT lines were in the lower third of the advanced tests was 36% (18/50) (Table 7).
View this table:
[in this window]
[in a new window]
|
Table 7 Stability matrices of the selected top ten, middle ten, and bottom ten entries from each population comparing 1996 plant row yield trials (PRYT) with a mean of 1997 + 1998 advanced yield testing
|
|
In conclusion, the data from the populations examined reveals lines selected for advanced testing from the Asgrow A3733 x Burlison and Jack x Resnik populations were the most stable when PRYT testing is compared to advanced yield testing. These results confirm the variance broad-sense heritability estimates associated with the 1996 PRYT environment. The other populations were not as stable, but progress to identify elite lines would be possible. Across the different populations sampled, PRYT testing was useful to identify several lines that were the highest yielding in advanced testing.
|
NOTES
|
|---|
Contribution from the Illinois Agric. Exp. Stn., Urbana, IL. Research supported by the Illinois Soybean Program Operating Board.
Received for publication January 25, 1999.
|
REFERENCES
|
|---|
- Boerma H.R., Cooper R.L. Comparison of three selection procedures for yield in soybean. Crop Sci. 1975;15:225-229 a.[Abstract/Free Full Text]
- Boerma H.R., Cooper R.L. Performance of pure lines obtained from superior-yielding heterogeneous lines in soybean. Crop Sci. 1975;15:300-302 b.[Abstract/Free Full Text]
- Hamblin J., Evans A.M. The estimation of cross yield using early generation and parental yields in dry beans (Phaseolus vulgaris L.). Euphytica 1976;25:515-520.
- Ivers D.R., Fehr W.R. Evaluation of the pure-line family method for cultivar development. Crop Sci. 1978;18:541-544.
- Knapp S.J., Stroup W.W., Ross W.M. Exact confidence intervals for heritability on a progeny mean basis. Crop Sci. 1985;25:192-194.[Abstract/Free Full Text]
- Knott D.R., Kumar J. Comparison of early generation yield testing and a single seed descent procedure in wheat breeding. Crop Sci. 1975;15:295-299.[Abstract/Free Full Text]
- Leudders V.D., Duclos L.A., Matson A.L. Bulk, pedigree, and early generation testing breeding methods compared in soybeans. Crop Sci. 1973;13:363-364.
- Littell R.C., Milliken G.A., Stroup W.W., Wolfinger R.D. SAS System for mixed models. Cary, NC: SAS Institute, 1996.
- McBlain B.A., Fioritto R.J., St. Martin S.K., Calip-Dubois A., Schmitthenner A.F., Cooper R.L., Martin R.J. Registration of `Resnik' soybean. Crop Sci. 1990;30:424-425.[Free Full Text]
- McBlain B.A., Fioritto R.J., St. Martin S.K., Calip-Dubois A., Schmitthenner A.F., Cooper R.L., Martin R.J. Registration of `Edison' soybean. Crop Sci. 1991;31:488-489.[Free Full Text]
- McBlain B.A., Fioritto R.J., St. Martin S.K., Calip-Dubois A., Schmitthenner A.F., Cooper R.L., Martin R.J. Registration of `Thorne' soybean. Crop Sci. 1993;33:1406.[Free Full Text]
- McIntosh M.S. Analysis of combined experiments. Agron. J. 1983;75:153-155.[Abstract/Free Full Text]
- Nickell C.D., Noel G.R., Thomas D.J., Waller R. Registration of `Jack' soybean. Crop Sci. 1990;30:1365 a.[Free Full Text]
- Nickell C.D., Thomas D.J., Gray L.R., Hanson P.M. Registration of `Burlison' soybean. Crop Sci. 1990;30:232 b.[Free Full Text]
- Ntare B.R., Akenova M.E., Redden R.J., Singh B.B. The effectiveness of early generation (F3) yield testing and the single seed descent procedures in two cowpea (Vigna unguiculata (L.) Walp.) crosses. Euphytica 1984;33:539-547.
- O'Brien L., Baker R.J., Evans L.E. Response to selection for yield in F3 of four wheat crosses. Crop Sci. 1978;18:1029-1033.[Abstract/Free Full Text]
- Rahman M.A., Bahl P.N. Evaluation of early generation testing in chickpea. Plant Breeding 1986;97:82-85.
- Satterthwaite F.E. An approximate distribution of estimates of variance components. Biom. Bull. 1946;2:110-114.
- Singh S.P., Lepiz R., Gutierrez A., Urrea C., Molina A., Teran H. Yield testing of early generation populations of common bean. Crop Sci. 1990;30:874-878.[Abstract/Free Full Text]
- St. Martin S.K., Dye B.W., McBlain B.A. Use of hill and short row plots for selection of soybean genotypes. Crop Sci. 1990;30:74-79.[Abstract/Free Full Text]
- Thorne J.C. Early generation testing and selection in soybeans: association of yields in F3 and F5 derived lines. Crop Sci. 1974;14:898-900.[Abstract/Free Full Text]
- Voight R.L., Weber C.R. Effectiveness of selection methods for yield in soybean crosses. Agron. J. 1960;52:527-530.[Abstract/Free Full Text]
This article has been cited by other articles:
|
|
|
|
 
F. K. Padi and J. D. Ehlers
Effectiveness of Early Generation Selection in Cowpea for Grain Yield and Agronomic Characteristics in Semiarid West Africa
Crop Sci.,
March 19, 2008;
48(2):
533 - 540.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
D. G. Jones and C. W. Smith
Early Generation Testing in Upland Cotton
Crop Sci.,
December 2, 2005;
46(1):
1 - 5.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
T. C. Helms, J. H. Orf, and J. T. Terpstra
Resource Allocation to Select for Yield in Soybean
Crop Sci.,
September 1, 2002;
42(5):
1493 - 1497.
[Abstract]
[Full Text]
[PDF]
|
|
|
|
|
|
|
L. G. Streit, W. R. Fehr, and G. A. Welke
Family and Line Selection for Seed Yield of Soybean
Crop Sci.,
March 1, 2001;
41(2):
358 - 362.
[Abstract]
[Full Text]
[PDF]
|
|
|
TOP
ABSTRACT
INTRODUCTION
