3. Opportunities for improved plant performance
3.1 Breeding/breeding efficiency
Sugarcane: Research Towards Efficient and Sustainable Production. Wilson JR, Hogarth D M , C a m p b e l l JA and G a r s i d e AL (Eds).
C S I R O Division of Tropical C r o p s and Pastures, B r i s b a n e . 1996. p p . 3 7 - 3 8 37 INCREASE IN SUGAR YIELD FROM PLANT BREEDING FROM 1946 TO 1994
C H A P M A N LS
BSES, Private Mail Bag 57, Mackay Mail Centre, Q 4741, Australia
A B S T R A C T
Sugar yield was increased by 0.12 and 0.15 t/ha/yr rainfed and irrigated sugarcane culture respectively, by the release of new varieties over the period 1946 to 1994. This increase in yield gave a 135 % return on investment to the Australian Sugar Industry from plant breeding expenditure, in 1994.
These estimates were calculated from the results of two variety trials grown at the Sugar Experiment Station, Mackay over 8 seasons, 1987-95.
The six varieties used in the experiments were the dominant commercial varieties for the period in central Queensland, thus enhancing the reliability of the predicted gain in yield.
I N T R O D U C T I O N
The Bureau of Sugar Experiment Stations is the main provider of new sugarcane varieties for the Australian Sugar Industry. The majority of these varieties are bred at Sugar Experiment Stations located at Meringa, Ayr, Mackay and Bundaberg along the Queensland east coast. There is an exchange of varieties with other sugarcane breeding countries and some foreign varieties are grown. Currently, 9 0 % of sugar production in Australia is from locally bred varieties. When new varieties are introduced they have superior attributes which may be associated with specific disease and insect resistance, improved sugar quality, and favourable agronomic attributes. Usually, they are also selected to have a higher yield than the varieties they replace.
The trend in sugar yield over time associated with the introduction of new varieties is confounded with changes in cultural and management practices. Consequently, it is usually impossible to isolate the magnitude of the i n d i v i d u a l effects. E x a m p l e s of c h a n g e in c u l t u r a l and management practices over the last 48 years include increased irrigation and fertiliser use, the adoption of mechanical harvesting, the use of herbicides and reduced cultivation, to name but a few.
This paper used data from recent experiments in the Mackay district to assess the gain in yield and industry benefit from important commercial varieties released over the last 48 years. The experiments were a series of ratoon trials on a range of varieties, released between 1946 and 1994, w h i c h w e r e g r o w n u n d e r c o m m o n c u l t u r a l a n d e n v i r o n m e n t a l conditions. T h e six varieties were grown in rainfed and irrigated experiments for eight crops on the Sugar Experiment Station, Mackay.
The benefit to the Australian Sugar Industry on expenditure on plant breeding was estimated by extending these results to the whole industry.
M E T H O D S
Full details of the field experiments were published by Chapman et (1992). Briefly, the varieties Q 5 0 , Q68, Q87, Q 1 2 4 , Q 1 3 8 and N C o 3 1 0 were grown in two experiments on the Sugar E x p e r i m e n t Station, Mackay between 1987-95. One experiment was rainfed and the other furrow irrigated. Scheduling of irrigation was based on estimating soil water deficit from 'Class A' pan evaporation. Plots were irrigated at a soil water deficit of 64 mm, calculated by using pan evaporation, canopy development and a pan factor of 0.8 (McGuire 1991). Experiments had three replicates, and a plant crop and seven ratoons were grown.
Notable features of the seasonal conditions were good moisture for plant, first, fifth, sixth and seventh ratoons, dry conditions for early growth of second, third and fourth ratoons and late growth of third ratoons.
Plant and third ratoon growth was adversely affected by high cyclonic rainfall. Cane yield was measured in whole plots by weighing harvested- cut-billets with a truck-mounted scale. CCS content was measured by the standard method (Anon 1984) in j u i c e crushed from 10-stalk samples from each plot at harvest.
The yield level of varieties was calculated as the mean production over 8 crops for cane yield, CCS and sugar yield. These were then regressed over the year that the varieties were first grown commercially in central Queensland: Q50, 1 9 4 6 ; N C o 3 1 0 , 1958; Q 6 8 , 1957; Q87, 1968; Q 1 2 4 , 1984; Q 1 3 8 , 1994.
RESULTS
The newer varieties Q124 and Q138 consistently had higher cane yield man the older varieties (Table 1). Cane yield generally declined from first to third ratoon, increased again from fourth to sixth ratoon and declined in seventh ratoon. Irrigation increased cane yield, but there was a seasonal variation, even in the irrigated yields, not entirely related to water stress effects on crop growth. Season variation also occurred in CCS with N C o 3 1 0 and Q124 having high levels in most seasons, and Q87 having high levels under irrigation (data not presented).
Table 1 Yield of cane (t/ha) six varieties grown to seventh ratoons in rainfed and irrigated experiments.
Crop Variety Class Q50 NCo310 Q68 Q 8 7 Q124 Q 1 3 8 Rainfed
86 92 58 48 65 57 53 51 64
88 117 86 73 90 91 94 63 88
89 108 65 46 61 49 49 43 64
85 109 86 75 96 100 110 77 92
114 112 81 77 100 109 128 80 100
105 121 89 74 99 104 123 91 101 Irrigated
Regressions of cane and sugar yield on time of first commercial planting were significant for both the rainfed and irrigated crops (Fig. 1). These regressions indicated that introducing new varieties increased cane yield by 0.75 and 1.00 t/ha/yr, or 0.12 and 0.15 t sugar/ha/yr for rainfed and irrigated situations respectively. There were no significant trends for C C S . This contrasts with the results of Cox & Hansen (1995) w h o
Fig. 1 Regressions of mean (a) cane yield, (b) CCS and (C) sugar yield on year variety was first planted commercially in central Queensland.
attributed recent high CCS to new varieties in the central and southern Queensland regions.
D I S C U S S I O N
In assessing the reliability of these estimates of the yield benefit from plant breeding there are a number of positive features of this study. The positive attributes are the cultivars selected for the experiments have been or will in the near future be significant for the central Queensland area. Q50 and NCo310 have both, in their time, produced over 90% of the crop in one year. Q68 and Q87 were also widely popular and successful varieties.
Q124 contributed over 60% of the crop in 1994 and is increasing in popularity. Q138 is not yet widely grown, because it was released in 1994.
The experimental data were not confounded with management or cultural effects, as the varieties were all grown together at the same site in a properly randomised experiment and all received the same treatments under rainfed
and irrigated conditions. Current commercial practice is to use a crop cycle of plant and four ratoons. However, older ratoons are regularly grown. Provided yields can be maintained, profitability of cane growing is favoured by having a longer ratooning cycle, as the high cost of planting can be amortised over a longer crop cycle. Mean yield of sugar for varieties was therefore calculated over plant and seven ratoons rather than plant and four ratoons.
However, in contradiction to these various positive features, the data are s o m e w h a t l i m i t e d in scope b e c a u s e they are only from t w o experiments at one site over one eight year period. Notwithstanding this above limitation, a cost/benefit analysis was conducted to compare the return to the Australian Sugar Industry from improved varieties against the cost of plant breeding, using data for the 1994 season.
Several assumptions were made: (1) a crop cycle of plant and seven ratoons; (2) the increase in sugar yield of 0.12 and 0.15 t/ha/yr was applied across all rainfed and irrigated canegrowing areas respectively;
(3) the irrigated area was 0.4 of the total area and (4) the sugar price was $350/t.
The returns to the Australian Sugar Industry from the use of new varieties, both imported and locally bred, was calculated as: 365 000 ha x 0.4 irrigated x 0.15 t/ha of sugar x $350/t = S7.7M; plus 365 000 ha x 0.6 rainfed x 0.12 t/ha of sugar x S350A = S9.2M; giving a total of S16.9M.
The cost of plant breeding activities in 1994 was estimated at S7.2M.
These costs include $5.3M by BSES (BSES 1994), $0.5M by CSR (A Wood, personal communication). $1.4M from S R D C ( S R D C . 1994/
95).
C O N C L U S I O N S
Plant breeding in 1994. delivered a return of 1 3 5 % on the year's investment using this simple cost/benefit analysis. This return was reduced to 100% if a crop cycle of plant and 4 ratoons was assumed.
There is a 10-12 year delay from the time of original crossing until new varieties are delivered. As investment on plant breeding is ongoing, this analysis is adequate to indicate yield increases are occurring and that the high costs are justified. The cost input included not just the funds used for direct crossing, selection and production of new varieties, but also those used for research into plant breeding technologies such as disease resistance, molecular markers, flowering control and so on.
Not included in the returns from new varieties are benefits to the industry of disease and insect control, and of sugar quality, which are often associated with new varieties.
Investing in plant breeding is profitable for the Australian Sugar Industry as plant breeders are producing higher yielding varieties. The investment in research into new technologies are likely to enhance returns from plant breeding in the future.
A C K N O W L E D G M E N T S
T h e a u t h o r w i s h e s t o t h a n k t h e B S E S staff for a s s i s t i n g with experiments, in particular Rita Kupke. Kay Harris and James Currie and staff from CSIRO Tropical Crops and Pastures. Funding for the project was provided by Sugar Research and Development Corporation and BSES Board.
R E F E R E N C E S
Anon (1984) laboratory Manual for Australian Sugar Mills, Vol 1. BSES publications.
Chapman LS, Ferraris R, Ludlow MM (1992) Ratooning ability of cane varieties, variation in yield and yield components. Pros. Aust. Soc. Sugar Cane Technol. Conf. 14:130-138
Cox MC, Hansen PB (1995) Productivity trends in southern and central region and the impact of new varieties. Proc. Aust. Soc. Sugar Cane Technol.
Conf. 17:1-7
Jones PN, Ferraris R, Chapman LS (1993) A technique for minimising confounding of genotype x year and genotype x crop type effects in sugarcane. Euphytica 76:199-204
McGuirc PJ (1991) Irrigation of Sugarcane, pp23-29, BSES, SRDC, Gov.
Printer Qld.
Bureau of Sugar Experiment Station 14th Annual Report (1994) Sugar Research and Development Corporation (1994/5) Annual Research and
Development Program 38
Sugarcane: Research Towards Efficient and Sustainable Production. Wilson JR, H o g a r t h D M , C a m p b e l l JA and G a r s i d e AL ( E d s ) .
C S I R O Division of Tropical C r o p s and Pastures, Brisbane. 1996. pp. 3 9 - 4 1 39 BEST LINEAR UNBIASED PREDICTION AS A METHOD OF ESTIMATING BREEDING VALUE IN
SUGARCANE
STRINGER JK1, M c R A E TA2 and COX M C5
'Bureau of Sugar Experiment Stations, PO Box 86, Indooroopilly Q 4068 Australia
2Bureau of Sugar Experiment Stations, PMB 57 Mackay Q 4741 Australia
3Bureau of Sugar Experiment Stations, PO Box 651, Bundaberg, Q 4670 Australia
A B S T R A C T
Bureau of Sugar Experiment Stations (BSES) currently assess the breeding potential of sugarcane parents by combining several years of agronomic performance data, breeding information and disease ratings into an index. Although this method is comprehensive, it takes many years to estimate reliably the breeding value of a parent.
Family selection trials are typically highly unbalanced and data analysis cannot be undertaken by ordinary least squares approaches.
Statistical techniques such as Best Linear Unbiased Prediction (BLUP) were specifically developed to allow prediction of breeding values from unbalanced animal data sets. Although the theory could be adapted to other breeding programs, there have been few applications in plants.
BLUP analyses undertaken on family selection data provided BSES plant breeders with a simple and rapid means of combining data from a wide range of sources to identify superior parents. Preliminary results suggest that BLUP is as effective as the current BSES method for identifying superior parents and these parents are rapidly identified with fewer years of information. The use of BLUPs is likely to increase the rate of population improvement by better choice of parents and crosses.
I N T R O D U C T I O N
Identification of s u p e r i o r g e n o t y p e s for use as p a r e n t s is a key component of any plant breeding program. Sugarcane breeders from BSES use a formula for assessing the breeding value of a parental clone which combines several years of agronomic performance data, breeding information and disease ratings into an index. Although this method is comprehensive, it takes many years to estimate reliably the breeding value of a parent.
Each year about 300 new clones enter the parental collection. Crossing is expensive and only a limited number of crosses can be made. A method is needed which can be applied to early stage family selection trials, so mat inferior parents can be rapidly identified by a progeny test and removed from the collection. This would result in a more efficient breeding program and thus increase the rate of population improvement.
Data from early stage family selection trials are typically highly unbalanced and so classical statistical approaches such as ordinary least squares are inappropriate. Best Linear Unbiased Prediction (BLUP) is a proven technique in animal breeding for obtaining precise estimates of breeding value from highly unbalanced data sets. B L U P allows data from a diverse range of mating designs, relatives, traits and precisions to be combined into a single breeding value for each trait and genotype (White & Hodge 1989).
BLUP has been used extensively in animal breeding but there are few applications in plants. There are only two published applications of BLUP to sugarcane data. Using a balanced data set, Chang & Milligan ( 1 9 9 2 a , b ) e v a l u a t e d c r o s s e s a t o n e l o c a t i o n a n d s o f a m i l y b y environment interactions were not considered. It is when data are highly unbalanced that BLUP usually exhibits superiority over other techniques and is of relevance to BSES.
The objectives of this study were to determine the effectiveness of BLUP as a method for estimating the breeding potential of sugarcane parental clones. Comparisons were made between B L U P and the current BSES method of estimating breeding value.
C U R R E N T BSES M E T H O D O F E S T I M A T I N G B R E E D I N G VALUE
The traditional method used by BSES in Australia to estimate the breeding potential of parental clones incorporates the following information.
Clones are initially planted in the parental collection if they exhibit superior performance in yield trials conducted over a number of years and locations. Clones are assessed relative to commercial standards and the results are combined in a selection index called net merit grade (NMG) (Skinner 1965).
Disease resistance status
Disease ratings for several major diseases are assigned to a clone on a 1-9 scale where 1 is resistant and 9 is susceptible. In the empirical formula, disease ratings are in the form of an adjustment factor. The form of the adjustment depends on the region from which the parental clone comes and thus more important diseases in a particular area are given greater weighting.
Breeding performance
Several hundred experimental crosses are evaluated each year at five Sugar Experiment Stations spread across the major sugar growing regions. T h e aim is to identify those crosses with specific combining ability (SCA) coupled with high general combining ability (GCA). These superior crosses are called proven which means more seedlings from that cross should be planted. The system for identifying proven parents
(i) Selection rate - the percentage of original seedlings that are selected and replanted in later stages of testing (Hogarth &
Skinner 1986).
(ii) Family selection - whole families are rejected or selected based on mean performance (Falconer 1960).
Inbreeding
Before a cross is made, the level of inbreeding (F) is determined.
Depending on the level of inbreeding, a proposed cross may not be made or it may be penalised. Four levels of inbreeding are recognised:
(i) Selfs - Female and male clones are identical, F = 1/2 (ii) Line breeding or parent/offspring - Female clone is same as
either parent of male clone or vice versa, F =1/4 (iii) Half sibs - Female and male clones have one parent in common,
F = 1/8
(iv) Full sibs - Female and male parents of both clones are identical, F = 1/4
For selfs, line breeding and full sibs the crosses are avoided and half sibs are penalised.
A g r o n o m i c performance
40
The agronomic performance data, breeding information and disease ratings are used to calculate a breeding value estimate of a clone for a particular breeding program. This procedure is detailed in Hogarth &
Skinner 1986. The breeding values range from 0-10.
Difficulties in making improvements
BSES plant breeders have been dissatisfied with the current method of c a l c u l a t i n g b r e e d i n g value but have found it difficult to m a k e improvements. A highly unbalanced mating design caused by unreliable and sparse flowering at the main breeding station in North Queensland (Berding & Skinner 1987) coupled with visual assessment of yield for families had precluded the use of a statistical approach. Since the introduction of mobile weighing machines in the late 1980s an objective evaluation of families can be obtained. These data facilitated an investigation into statistical techniques such as BLUP.
MATERIALS A N D M E T H O D S Trial details
In the current assessment of the usefulness of BLUP, plant crop (first harvest) NMG data for the 1988-1994 series seedlings from the Mackay and Bundaberg breeding programs were used for analysis. Each family plot contained 20 clones planted as single seedlings in 12-13m plots with each clone 0.6m apart. Families were planted in several blocks with v a r y i n g r e p l i c a t i o n on the M a c k a y and B u n d a b e r g S u g a r Experiment Stations.
Model
BSES family selection data have an incomplete diallel mating design (Griffing 1956). The linear model is:
where
Analysis
The statistical program, G A R E M L (Huber 1993). was used to obtain BLUP estimates of parental general and specific combining abilities.
G A R E M L applies the algorithm developed by Giesbrecht (1983) to estimate REML variance components (Patterson & Thompson 1971) and uses the theory developed by Henderson (1973) to obtain BLUP estimates of breeding value.
Although G A R E M L is computationally efficient, the inversion of matrices in the BLUP procedure is memory intensive especially with large data sets. To simplify the analysis due to memory and time constraints, the year effect was removed from the NMG family selection data prior to analysis. For each year in the data sets, this involved subtracting the mean and divided by the variance.
As the BLUP analyses were undertaken on standardised NMG data, the resulting breeding values ranged from -1 to 1. A data transformation was applied so the breeding values ranged from 0 - 10.
RESULTS A N D DISCUSSION
In the 1994 and 1995 crossing seasons, G A R E M L was used to determine B L U P estimates of breeding value for parental clones from the Central and Southern Queensland breeding p r o g r a m s . This provided plant breeders with a rapid method of identifying superior parents based on family information. BSES plant breeders noted that BLUP could identify many similar superior parents based on fewer years of data (maximum of 6 years) in comparison to the current method (10 years).
C h a n g e s were m a d e to c o m p u t e r p r o g r a m s to e n a b l e a d e t a i l e d examination of the crosses made in the 1995 season. Of 1140 crosses made. 127 were designated as proven crosses. Approximately 6 0 % of these provens had B L U P ratings for both parents (73 out of 127). Of these 7 3 . 54 had average B L U P breeding values above 5 (good crosses) while 19 were designated as inferior (3 or less). This indicated that B L U P is effective for identifying superior parents.
The most accurate method to compare breeding value estimates based on the current B S E S method and B L U P s is to contrast the agronomic performance of crosses selected from parents c h o s e n by the t w o t e c h n i q u e s in a p r o g e n y test. Research to m a k e this c o m p a r i s o n c o m m e n c e d in 1994 when 40 BLUP, 40 empirical and 20 r a n d o m crosses for the Central and Southern Q u e e n s l a n d b r e e d i n g p r o g r a m s w e r e m a d e . P r o g e n y p e r f o r m a n c e d a t a w i l l b e a v a i l a b l e for a s s e s s m e n t in late 1996.
The power of B L U P as a predictive tool was determined using family selection data from Central and Southern Queensland. B L U P estimates based on all available family N M G information up to 1993 were calculated. Predicted performance from the BLUPs based on some 200- 4 0 0 clones was correlated with actual trial p e r f o r m a n c e in 1994.
Predicted performance using the current method based on the 1993 series breeding clones was also correlated with actual N M G in 1994. Breeding values were updated using 1994 information and predicted performance for the B L U P s and c u r r e n t m e t h o d were c o r r e l a t e d with actual performance in 1995. The results are given in Table 1.
Table 1 Correlation coefficients (r) to allow a comparison of the predictive power of BLUP with that of the current BSES method using family selection NMG data from Southern and Central Queensland.
[n is the sample size of the data set used to calculate each correlation coefficient.]
Year
1994 1995 BLUP vs NMG 0.62 (n = 81) 0.63 (n = 97) Southern
Queensland Empirical formula 0.45 (n = 81) 0.50 (n = 97) vsNMG
BLUPvsNMG 0.64 (n = 71) 0.38 (n = 125) Central
Queensland Empirical formula 0.49 (n = 71) 0.12 (n = 125) vsNMG
The correlation coefficient for BLUP vs N M G was always greater than for the current method vs N M G . These results are encouraging as the current method incorporates information for up to ten years whereas B L U P results are based on a m a x i m u m of 6 y e a r s . Both sets of correlations were low for Central Queensland in 1995. This may be due to problems experienced with moisture stress in die propagation of me seedlings. On the benches, all seedlings from one family are planted together. Given that the irrigation system only failed on one part of the bench, then some families would have been suffered severely from moisture stress and shown uncharacteristic results.
C O N C L U S I O N S
Preliminary results suggest that the B L U P method is as effective as the current BSES method for identifying superior parents. As the B L U P method becomes more refined by including information on relatives its superiority will increase. Economic savings associated with using BLUP and a reduced generation interval make it a potentially effective method for estimating the breeding value of parental clones.
R E F E R E N C E S
Berding N, Skinner JC (1987) Traditional b r e e d i n g m e t h o d s . In:
Copersucar International Sugarcane Breeding Workshop, pp. 269- 320. Copersucar: Piracicaba, Brazil.
Chang Y S , Milligan SB (1992a) Estimating the potential of sugarcane