| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
a CSIRO Plant Industry, Davies Laboratory, PMB, PO Aitkenvale, Qld. 4814. Australia
b Bureau of Sugar Experiment Stations, PMB57 Mackay Mail Centre, Qld. 4741
Corresponding author (Phillip.Jackson{at}pi.csiro.au)
 |
ABSTRACT |
|---|
 TOPNOTES ABSTRACT INTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Abbreviations: CCS, commercial cane sugar • REV, relative economic value
|
INTRODUCTION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
There are few published reports of competition in sugarcane. Skinner (1961) and Skinner and Hogarth (1978) examined competition in Australia. Results from both studies highlighted that variance due to competition was potentially large in sugarcane variety trials, and could seriously bias selection trial results. However, interpretation of results in these reports was limited by the methods used. First, these studies involved evaluating clones derived from previous stages of selection. If competition effects were large among seedling clones, either as individual seedlings in the first selection stage, or in subsequent small plots, prior selection pressure would be expected to discard uncompetitive clones. Such studies may underestimate the importance of competition in original populations, and are of unknown relevance to the earliest stages of selection. Second, these studies only examined trials with plots of a single plot size (e.g., three-row plots) and estimated competition effects using certain assumptions about the relative level of competition expressed in different rows in multirow plots. It was assumed that the outside rows of a three- or four-row plot would express half the competition effect expressed in a single-row plot, and that the middle row(s) in a three or four row plot would be free of any effects due to competition. However, this may not be the case: if growth in an outside row(s) of a three- or four-row plot was strongly affected by inter-plot competition (adversely or favorably), this would result in further inter-row competition effects (in the opposite direction) passed on to the adjacent rows in the same plot.
If the primary objective of variety trials is to select genotypes that maximize economic value in a pure commercial stand, then selection based on any measurement in small plots can be regarded as indirect selection. Using the framework described by Falconer and Mackay (1996), performance in small plots would be regarded as a secondary character to which selection may be applied, with the aim of improving the primary character, performance in pure stand. Accordingly, response to selection in small plots may be predicted on the basis of simple models describing correlated response to selection (Falconer and Mackay, 1996) provided appropriate statistical and genetic parameters are determined from field experimentation for different plot sizes. This approach, also makes it possible to predict the effectiveness of a wide range of different methods of selection, based on different selection criteria and plot size x replicate configurations.
The aim of the research reported here was to obtain estimates of such parameters and then to use these to predict effective approaches to selection among relatively unselected clonal populations in sugarcane breeding programs. The clones used in this study were representative of those directly derived from hybridization in two different sugarcane breeding programs, and were unbiased by any previous selection. As such, the genetic parameters are useful for other studies that may simulate and assess different options for selection from the first stages of selection in sugarcane breeding programs.
The study was conducted in the Burdekin region in Australia. Sugarcane crops in this region are fully irrigated, usually have high yields (up to 210 Mg ha-1; average 130 Mg ha-1) and may lodge up to 7 mo prior to harvest (Muchow et al., 1994). It has been suggested that lodging may be detrimental to achieving higher sugar yields under these types of environments (Muchow et al., 1994). If this is the case, resistance to lodging may be one feature of the "ideal" genotype for such environments. Canes that grow tall and thin may be more prone to lodging (Amaya et al., 1996). However, they are also likely to have a significant competitive advantage in small plots because they capture more solar radiation than shorter neighboring plots (Tovey et al., 1973). Even after lodging, such clones may experience a competitive advantage because they tend to fall on top of neighboring plots. While speculative, it is possible that selection in small plots may bias selection against short, thick stalked genotypes that resist lodging and may have greatest potential in pure stands under high yielding conditions. Therefore, the role of selection in small plots under very high yielding and lodged conditions was also of special interest. Previous reported studies on competition in sugarcane have not focused on such environments.
|
MATERIALS AND METHODS |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Both experiments were established on the alluvial plain of the lower Burdekin River in Australia (19.0° S lat; 147.3° E long). Experiment 1 was planted on a commercial farm on 7 April 1995. Experiment 2 was planted on 17 May 1995 on the BSES experiment station at Brandon. Both experiments were planted and grown according to accepted commercial practices for the Burdekin region, which includes regular furrow irrigation. However, the yield of the first ratoon crop of Exp. 2 was significantly lower than that normally obtained in the Burdekin region (trial mean yield was 74 Mg ha-1 compared with region average of approximately 130 Mg ha-1). The reason for the low yield was at least partly due to the short growing duration between the harvest of the plant crop and the harvest of the first ratoon crop.
In each experiment, each clone was planted to three plot shapes. In the first experiment, plot shapes were 1 row by 10 m, 2 rows by 10 m and 6 rows by 20 m. Each clone by plot shape combination was randomized within each of two blocks, so that the design used was equivalent to a randomized complete block design with two blocks and treatments comprising plot shape x clone combinations. Within each block there were two replicates of each clone in the 1 row by 10 m plot shape and one replicate of each clone for the other two plot sizes. Randomization of clones and plot shapes within each block was done with the constraint that no clone was positioned adjacent to a plot containing the same clone, either laterally or at plot ends.
For the second experiment, plot shapes were 1 row by 10.6 m, 2 rows by 10.6 m and 6 rows by 21.2 m. The design in this experiment was a 7 by 7 simple lattice (Cochran and Cox, 1957) with two complete blocks. Each incomplete block contained all three-plot shapes of seven varieties. As with the first experiment, constrained randomization was used to avoid different plot shapes of the same variety competing as neighbors.
Measurements
Harvesting of the plant cane crop occurred on 15 July and 8 October for Exp. 1 and 2, respectively. The first ratoon crops were harvested on 16 September and 16 June for Exp. 1 and 2, respectively. The second ratoon crop for Exp. 2 was harvested on 14 Nov. 1998. The latter harvest was done later in the season than normal commercial practice and under very wet conditions. Significant growth of late tillers (suckers) was observed in most plots at harvest.
The plant cane crops of both trials were burnt before harvest, and mechanically harvested. In the ratoon crops, Exp. 1 was burnt before harvest, while Exp. 2 was harvested without burning. Cane from individual rows of the plots was weighed at harvest using mobile weighing machines. For Exp. 1, only the middle two rows of the six row plots were weighed. For Exp. 2, a 2-m buffer was left at the ends of the 6 row by 21.2-m plots, so that only 17.2 m was weighed.
At harvest, a hand-cut sample of four stalks was removed from each harvested row for determination of commercially recoverable sugar content (commercial cane sugar, CCS, BSES, 1984). Sugar yield (Mg ha-1) on a per plot basis was calculated from the product of cane yield and CCS. Relative economic value (REV, $A/ha) on a per plot basis was determined from revenue that would be realized from sugar produced minus variable costs of production. Because the major costs of harvesting, cane transport and milling, are approximately proportional to cane yield, clones that produce high sugar yields via high CCS rather than high cane yield are more profitable. The following formula was used for estimating REV:
 |
 |
 |
While it was assumed that REV ought to be the primary selection criterion, some sugarcane breeding programs also regard improvement of sugar yield in pure stands as the primary selection criterion. Therefore, in this study the effectiveness of selection in small plots was also assessed assuming sugar yield as the primary selection criterion.
The difference between performance in small plots and performance in the bordered rows of the six-row plots for any character was defined as competition for that character. Competition effects estimated for each clone in each block in each experiment, therefore, could be subjected to the same analyses as any other character.
In Exp. 1 some measurements additional to CCS and cane yield in each plot in the plant crop were also taken. Measurements were made in the plant crop on 28 Sept. 1995 of the height from the ground to the top of the canopy. The mean of two measurements per plot was recorded, with each measurement being taken at random locations within each plot. Visual ratings were given of overall appearance (1 worst, 9 best), based on apparent cane mass present. Ratings were made by two experienced technical officers independently on 19 Dec. 1995, and averaged for analysis.
Statistical Analysis
Results from each experiment were analyzed using the SAS statistical package (SAS Institute Inc, NC, USA). For each experiment, analyses of variance were firstly done for each attribute for each plot shape in each crop-year. For analyses of individual attribute x plot shape x crop-year combination, the following model was assumed:
 |
2g. Variance components for genotypes and error were estimated from the following expectations of mean squares:
 |
Analyses of covariance were done for selected pairs of characters within the same experiment. These included characters measured on the same plots or on different plot shapes. This was done in the same manner as the analyses of variance except that sums of cross products and mean cross products were determined, with appropriate covariance components and mean cross products substituted for variance components and mean squares.
Broad sense heritabilities (h2) for each trait were determined from (Falconer and Mackay, 1996):
 |
2g = genetic variance, and 2p = phenotypic variance. Phenotypic variance was determined from:
 |
2e = error variance and r = number of replicates. Genetic correlations between characters X and Yr were determined from:
 |
 |
 |
Standard errors of genetic correlations were estimated using the procedure of Tallis (1959).
Performance for different characters in one- or two-row plots was considered to represent possible secondary indirect selection criteria for predicting performance in the bordered rows in large plots. If performance in the middle two rows of the six row plots is designated character Y, and performance in small plots is designated character X, then the correlated response in character Y from indirect selection based on character X (CRy) was determined as follows (Falconer and Mackay, 1996):
 |
gy is the genotypic standard deviation for character Y. |
RESULTS AND DISCUSSION |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
|
As expected, error variances were also inflated in most cases in small plots compared with the large plots. This effect was greatest for cane yield, where error variances were around 4 to 12 times those in the largest plots (Table 1). Heritability estimated for different plot sizes (data not shown) differed little between plot sizes because both genetic and error variances varied similarly across different plot sizes.
Correlations between Performance in Different Plot Sizes
Genetic correlations between characters measured in unbordered and bordered plots in plant and first ratoon crops are given in Table 2. For each character measured in single-row plots and two-row plots in each crop-year, genetic correlations with the same trait measured in large plots in each crop-year are shown. Also shown are genetic correlations between the same characters in unbordered plots and both sugar yield and REV measured in each crop-year in bordered plots. For CCS, genetic correlations between the different plot sizes for the same crop-year were all close to unity. This indicates that measurement of this character in small plots is unbiased. Therefore, providing adequate precision in measuring CCS is obtained, selection in single-row plots would be effective for improving performance of CCS in near pure stands.
|
Visual appearance grades in one-row plots had a genetic correlation with cane yield in the large plots of 0.57 ± 0.14 in the environment in which this was recorded. This was at least as large as the genetic correlation for cane yield measured directly in single-row plots (rg = 0.48 ± 0.13, Table 2). The advantage of visual appearance grade over cane yield in this respect would appear to be due to competition effects not affecting estimates (at least at the time it was done), as indicated by the high genetic correlation between visual appearance grades in the one-row plots and the large plots (1.07 ± 0.12). Visual appearance grades in small plots also had a similar heritability to cane yield in the same plots (Table 1—see ratio of genetic variance to error variance). These results suggest that, for one-row plots, visual appearance grade would be at least as effective as cane yield as a selection criterion for improving cane yield in large plots. For the plant crop in Exp. 1, which was the only crop where visual grades were recorded, predicted gains in cane yield in large plots from selection (not shown) based on visual grades were greater than those determined with measured cane yield as a selection criterion.
Genetic variance for competition effects for cane yield were highly significant (P < 0.01) and this trait had moderate heritability (Table 3). Competition effects for cane yield showed no consistent correlation with cane yield in the bordered rows of the large plots (Table 3), ranging from -0.29 to 0.63 in the different experiments and crop-years. This is consistent with the findings of Skinner and Hogarth (1978) who showed that among highly selected clones that correlation between competition and "true yield" was highly variable in different environments, with a mean of -0.02 in trials where significant competition effects were apparent.
|
In all cases for cane yield in Table 2, correlations involving the two row plots were slightly greater than for single row plots (generally by about 0.05–0.20), but were still significantly less than 1.0. This indicates that selection for cane yield would be more effective in two row plots, but in some cases would still be seriously limited because of bias associated with competition effects.
Collectively, the genetic correlations obtained for cane yield in different plot sizes suggest that selection for cane yield in unbordered plots would result in improved performance in pure stands, because of positive genetic correlations in all cases. However, gains would be limited in most situations, especially in single row plots, because of genetic correlations significantly less than 1.0.
For sugar yield and REV, the genetic correlations between performance in small plots and bordered plots were also variable, but in all cases were significantly less than 1.0. This was particularly the case for sugar yield, reflecting the strong effect of cane yield in determining this attribute. For example, in the plant crop, the genetic correlation between sugar yield in single-row plots and sugar yield in the bordered plots were 0.65 and 0.48 for Exp. 1 and Exp. 2, respectively.
It may be assumed that the major objective in selection is to improve performance for REV in pure stands. The genetic correlation between a particular character in small plots and REV in large plots is of interest in assessing the potential effectiveness of using that trait in small plots for improving the primary selection criterion. An important result in Table 2 is that the genetic correlation between CCS measured in small plots and REV in large plots was as great or greater than for any other character in small plots, including REV itself. For example, for Exp. 1, the genetic correlation between CCS in the single row plots in the plant crop and REV in the 6-row plots was 0.82 (±0.07), while the correlation for REV itself was less at 0.72. The higher genetic correlations associated with CCS compared with other characters in this context was the case for both experiments, although it was most marked for Exp. 2.
By contrast, the genetic correlations between cane yield in the small plots and REV in the large plots were quite low (e.g., rg = 0.38, 0.13 for single-row plots in the plant crop-year for Exp. 1 and 2, respectively). This indicates that gains from selection on the basis of cane yield in small plots for improving REV would be quite limited, even if precise data were obtained. Overall, the high genetic correlations for CCS suggest that effective selection in small plots could be based on selection for CCS alone, which would offer significant practical advantages over selection criteria relying on estimating cane yield, as discussed below.
Practical Selection Options in Breeding Programs
As with all selection trials, there is a trade-off between increasing replication, plot size, and number of genotypes that can be evaluated, given limited resources (Gauch and Zobel, 1996). Increasing replication will increase precision of estimates of clone means; increasing plot size will increase genetic correlation with yield in pure stand (Table 2) and decrease error variance (Table 1), but decrease genetic variance (Table 1); while increasing number of genotypes being evaluated will allow increased selection intensity. In deciding on an optimal selection system, the trade-offs associated with all factors need to be considered and different options compared for expected gains from selection. Further, in a multi-stage selection system, the optimization of one stage of selection cannot be done in isolation from other stages. Changes in one stage of selection (e.g., number of clones to be evaluated, selection intensity used) will affect the design and performance of selection stages following or prior to that stage. Proper development of an optimal selection system based on results from this study and other knowledge is a major exercise, and beyond the scope of this paper. However, gains from selection from a limited set of comparisons for selection among sugarcane clones are considered here, as a guide for further investigations. In addition, these comparisons are used to show that more effective approaches than currently used in sugarcane breeding programs can probably be identified.
In practice, the first stage of selection in sugarcane breeding programs involves evaluating clones as single plants grown from true seed, either in family plots, or as individual clones. The second stage involves clonal propagation of cane cut from the individual plants and further testing in plots grown from this cane. The selection options compared here relate to those available immediately following the first seedling stage of selection or propagation, i.e., Stage 2 in the selection system. Under good growing conditions, individual seedling clones may produce up to about 20 m of cane that can be planted to the next selection stage. Therefore, options in the second stage of selection are limited by planting material: total row length in any plot size x replication configuration cannot practically exceed 20 m in most situations. The current selection system in breeding programs in Australia generally involves planting single-row plots with either one or two replicates. For a given amount of resources, trials with two replicates of single-row plots, or single replicates of two-row plots would only allow evaluation of half the number of clones than if single-row plots with single replicates were used. Halving the number of clones reduces selection intensity and gains from selection. The results from this study (in which available planting material was increased through additional years of propagation) allowed different configurations of plot size and replication to be examined, as well as consider impacts of different selection criteria.
Table 4 shows the predicted response from selection based on different traits, based on selection in (i) one-row plots x one replicate per clone, (ii) one-row plots x two replicates per clone,and (iii) two-row plots x one replicate per clone. A selection intensity of 5% is assumed for option (i), which corresponds approximately to past practice in Australian sugarcane breeding programs. For the other two options, given equivalent resources, only half the number of clones would be able to be evaluated so that the selection intensity would be 10%, assuming an equivalent number of clones were selected for the next stage of selection. For each character considered for selection in Table 4, correlated responses to selection for performance in the large plots are shown for sugar yield, and for REV. It may be assumed that the key objective in selection under most situations would be to maximize gain in REV per unit cost of selection.
|
In Australia, selection of clones in trials has frequently been based on sugar yield, or other indices related to sugar yield with additional weighting toward CCS. The results presented here suggest that selection based solely on CCS may be effective, and that very little weighting to cane yield as a character for selection would be appropriate. This result leads to the question of whether measuring cane yield in early stage selection trials represents optimal allocation of resources, or whether reliance on CCS data for selection may be more appropriate in early stages. Measuring cane yield not only is associated with some cost, but it may also add an extra year to selection if harvesting and weighing is practiced, as in Australian breeding programs. When a trial is harvested and weighed in the plant crop, the seed source of selected clones in the trial is unavailable until the ratoon crop is grown. Therefore, either a further year is required so that planting material to establish trials for the next stage of selection, or otherwise expensive additional plots of all clones need to be concurrently propagated if this time delay is to be avoided. A high correlation is usually found between performance in plant and ratoon crops in early stage trials (Jackson, 1992; Mirzawan et al., 1993) so there is usually little advantage in waiting for ratoon crops in terms of obtaining additional data to use in selection.
If cane yield results from a trial with unbordered plots are obtained, an optimal selection index (Baker, 1986; Smith, 1936) involving cane yield and CCS could be identified that would result in greater gains than just using CCS alone. The results obtained in this study indicate that such an index is likely to have a very heavy weighting on CCS and little on cane yield. It is also likely that visual gradings for cane yield could be as effective in a combined selection index with CCS compared with use of cane yield. As indicated previously, visual grades measured in single row plots appear to be as effective as measurements of cane yield for predicting performance in bordered plots. Given that such ratings are cheap and do not involve crop destruction, these grades could be used to discard clones which are likely to have extremely low cane yield, without the need to harvest and weigh all plots. However, the use of visual gradings requires further research in order to optimize their application.
Further research is required to examine alternative selection systems and to identify optimal systems. However, on the basis of results reported here, it would appear that an optimal system may involve the following.
1. Evaluation of a large number of clones in small, single row plots for CCS in the plant crop.
2. Selection of clones based largely on CCS, but with discard of material with particularly poor visual grades, and progression of selections to the next stage of selection in the same season as evaluation.
3. Evaluation of the high CCS selections in multi-row plots (resembling commercially grown pure stands) for accurate assessment of cane yield and further CCS evaluation.
|
ACKNOWLEDGMENTS |
|---|
|
NOTES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
Received for publication September 14, 1999.
|
REFERENCES |
|---|
|
TOPNOTESABSTRACTINTRODUCTIONMATERIALS AND METHODSRESULTS AND DISCUSSIONREFERENCES |
|---|
This article has been cited by other articles:
|
|
 |
|
  L.-P. Wang, P. A. Jackson, X. Lu, Y.-H. Fan, J. W. Foreman, X.-K. Chen, H.-H. Deng, C. Fu, L. Ma, and K. S. Aitken Evaluation of Sugarcane x Saccharum spontaneum Progeny for Biomass Composition and Yield Components Crop Sci., May 1, 2008; 48(3): 951 - 961. [Abstract] [Full Text] [PDF]
|
|
|
|
|
|
S. B. Milligan, M. Balzarini, K. A. Gravois, and K. P. Bischoff Early Stage Sugarcane Selection Using Different Plot Sizes Crop Sci., September 1, 2007; 47(5): 1859 - 1864. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| 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 | |||
