Near infrared analysis of sugarcane (Saccharum spp hybrid) bud scales to predict resistance to Eldana stalk borer (Eldana saccharina Walker).
The eldana stalk borer (Eldana saccharina Walker) is the most serious pest of the Southern African sugarcane industry, and it is imperative that effective control measures are available to minimize economic damage. Because conventional control methods have had limited success, cultivar resistance is seen as the most viable method of controlling infestation. However, due to the space- and time-consuming nature of the present screening methods, only small numbers of cultivars can be tested relatively late in the Plant Breeding selection programme. Increased resistance in breeding and selection populations is therefore slow. Buds are a preferred entry point of eldana larvae as they are softer than the rind that is present on the rest of the stalk surface. Preliminary results by other workers suggested that near infrared spectroscopy (NIRS) could provide a rapid screening method for the chemical profile in bud scales, the outer coating of buds and therefore the first contact point of an invading larva. If feasible, analysis of samples using this method could be done in the South African Sugar Experiment Station's (SASEX) stage two selection trials, providing an early indication of eldana resistance on large numbers of cultivars, without the necessity of separate trials. However, knowledge of how environments, position of bud scales on the stalk and age affect NIRS is required in order to determine the feasibility of the method. Planting of a trial with an identical set of genotypes across a range of environments, sampled at a number of ages, would provide the necessary information on environmental effects, whilst simultaneously providing the necessary range of samples to develop a calibration between bud scale chemical profiles and eldana resistance ratings. Inheritance patterns of the characteristics being measured is also required if they are to be used in a breeding programme. The original work by Rutherford (1993) was carried out on only five calibration sets (a set of standard clones with relatively well-known eldana resistance ratings), and different sets were not comparable due to what was assumed to be environmental differences between calibration sets. One aspect of the current experiment was to examine more closely the effect of genotype x environment interaction (G x E) on the performance of the NIRS technique under a range of conditions. Two sites were chosen to represent the conditions encountered in trials carried out by SASEX. The crops were sampled at three ages, representing the range of ages at which sugarcane is harvested in South Africa. Two locations on the stalk were also examined, top and bottom, for removal of bud scales, based on the assumption that aging of bud scales may affect chemical composition. A new NIRSystems 6500 instrument was acquired during the course of this study. Data from the new instrument indicated that there were no longer differences between the different calibration sets, and therefore no longer differences between environments. Spectra for different samples were very close, the differences being of the same scale as those recorded with repeated measures of the same samples, or between the readings for the standard solvent solution. This led to the conclusion that the differences observed on the original NIRSystems 5000 instrument were due to instrument error, not environmental differences. More importantly, the different calibration sets were not comparable despite being similar to each other. Prediction from one calibration set to another was low. These observations led to the conclusion that NIRS was not a suitable method for determining chemical compounds associated with tolerance of sugarcane genotypes to eldana borer. The original NIRS instrument was subject to error, and the small number of calibration sets included in the study led to the erroneous conclusion that NIRS was suitable for the prediction of varietal tolerance to eldana. With the acquisition of the new instrument, the errors generated by the old instrument became apparent. With the increase in number of calibration sets included in the study, it also became apparent that a global calibration covering all environments was not possible. An analysis of the heritability of the chemical compounds associated with eldana resistance was also included in this study. A biparental progeny design of 24 crosses with 33 unselected offspring per cross was used. This trial would have been analysed once the calibration had been developed using the environmental trial, and it would have provided knowledge of the breeding behaviour of the chemical compounds associated with tolerance to eldana. Because the NIRS technique proved to be unsuitable for detection of chemical compounds associated with eldana resistance, the heritability of these chemical compounds could not be studied. As the NIRS study did not produce data, the G x E interaction analysis and determination of heritability was applied to the bud scale mass data set. This study showed a relatively low positive correlation between bud scale mass and resistance to eldana. The broad sense heritability estimate for bud scale mass from the G x E interaction analysis was 0.45, and the narrow sense heritability estimate from parent-offspring regression analysis was approximately 0.27, suggesting a low degree of genetic determination in bud scale mass. The G x E interaction analyses gave varying results depending on the method used. The ANOVA analysis suggested that ages, sites and years had an effect on bud scale mass, while deviation from maximum plot showed no significance for G x E interactions. The number and choice of genotypes selected as unstable also varied with the method used to determine the stability of individual genotypes. Regression analysis and rank order analysis revealed a number of unstable genotypes, whilst stability variance and ecovalence, which produced similar results, detected only two unstable genotypes. In the rank order analysis correction of data to remove genotype effect, reduced the number of unstable genotypes, suggesting that the G x E interaction effect was partially confounded with the bud scale mass of the genotypes. This was a more reliable method than the uncorrected rank order analysis, and would be the preferred analysis type of all those tried.