OPERATIONAL VALIDATION OF THE EFFICACY OF SPECTRACANEâ„¢, A HIGH-SPEED ANALYTICAL SYSTEM FOR SUGARCANE QUALITY COMPONENTS

INSTALLATION of SpectraCane™ systems at four BSES Limited experiment stations enabled data for all quality components to be acquired cost-effectively for all selection stages. The research objectives were to validate the analytical accuracy of SpectraCane™ and compare the precision of data obtained from near infra-red spectroscopic prediction (NIS) and routine laboratory analyses (RLA). Random samples from the SpectraCane’s™ 2009 throughput at BSES Meringa and duplicate subsamples from a final assessment trial (FAT) provided an assessment population (n = 841). Accuracy for four primary and four derived quality components was determined using regression analysis and mass balance summation. The FAT data (n = 288) provided estimates of the precision for NIS and RLA analyses. All except two regressions of NIS vs RLA values were skewed, with b values being highly significantly different from b = 1.0. Slopes (b) ranged from 0.891 to 1.011 and from 0.802 to 0.991 for the primary and derived components, respectively. Regression accounted for a highly significant proportion of variation between the two analyses for all components. Fibre and purity were the poorest in each set. Duplicate sub-sampling was an adequate strategy but error ratios for NIS data were markedly lower than the respective RLA data. Phenotypic and genotypic correlations of NIS predicted vs RLA values were excellent for all components, with those for purity the weakest. Mass balance summation indicated overall analytical accuracy was excellent. Estimates of broad sense heritability and genetic coefficient of variation differed little between NIS and RLA data. Little loss in gains from selection will result from use of NIS data. Rather, application of NIS analysis will enhance selection gains considerably, through increased objectivity and markedly reduced personnel inputs. On average, NIS analysis is half as precise as RLAs with sampling standard deviation ratios ranging from 1.55 to 2.74.