Submitted to: Field Crops Research
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 8/26/2015
Publication Date: 11/6/2015
Citation: Piaskowski, J., Brown, D., Garland Campbell, K.A. 2015. NIR calibration of soluble stem carbohydrates for predicting drought tolerance in spring wheat. Field Crops Research. 108:285-293. Interpretive Summary: The problem is that selection for drought resistance in wheat can be done by assaying stem carbohydrates but the wet chemistry method to assay for them is time consuming and uses hazardous chemicals. Near-infrared spectroscopy can be used as an alternative method but this techniques needs to be calibrated to data from the field. This research was initiated to develop a robust calibration between NIR spectra and soluble carbohydrate concentration in wheat and to determine whether the soluble stem carbohydrates were correlated to performance under drought. Wheat stems were harvested from plots grown in six environments under a range of moisture regimes in the Pacific Northwest. Soluble stem carbohydrates were assayed using wet chemistry and NIR and the data were calibrated using a series of mathematical models. Partial least squares regression provided the most accurate and reliable predictions for soluble carbohydrates and the data were correlated with performance in dry environments. This research provides an additional method to assay wheat for tolerance to drought.
Technical Abstract: Soluble stem carbohydrates are a component of drought response in wheat (Triticum aestivum L.) and other grasses. Near-infrared spectroscopy (NIR) can rapidly assay for soluble carbohydrates indirectly, but this requires a statistical model for calibration. The objectives of this study were: (i) to build a robust calibration between the NIR spectra and soluble carbohydrate concentration of ground wheat stems; and (ii) to determine whether soluble stem carbohydrates are correlated with yield rankings of drought-stricken wheat grown in the northwestern United States. Five spring wheat cultivars were grown in field trials conducted at six environments in the state of Washington varying in annual precipitation from 212 to 474 mm. Wheat stems were harvested from all plots at the onset of grain fill and assayed for NIR reflectance. Soluble stem carbohydrates were determined on a subset of the samples. The NIR data were calibrated to soluble stem carbohydrates using multiple linear regression, partial least squares regression, ridge regression with best linear unbiased prediction, random forest, least absolute shrinkage and selection operator (lasso), elastic net, and Bayesian lasso regression. Partial least squares regression provided the most accurate and reliable predictions for soluble carbohydrates. Correlations between soluble stem carbohydrates and grain yield were consistent across environments (r = 0.904 and ' = 0.80). The effect of environment on the variation in response variables was lower for soluble carbohydrates than yield (5.93 and 71.7%, respectively) across environments. These data provide evidence that stem carbohydrates can aid in selecting cultivars with enhanced drought resilience.