CSIRO Land and Water, Private Bag 12, Hobart, Tasmania, Australia.
Central Science Laboratory, University of Tasmania, Private Bag 74, Hobart, Tasmania , Australia.
Tree Physiol. 2017 Jan 31;37(1):131-141. doi: 10.1093/treephys/tpw083.
Near-infrared reflectance spectroscopy (NIRS) is frequently used for the assessment of key nutrients of forage or crops but remains underused in ecological and physiological studies, especially to quantify non-structural carbohydrates. The aim of this study was to develop calibration models to assess the content in soluble sugars (fructose, glucose, sucrose) and starch in foliar material of Eucalyptus globulus. A partial least squares (PLS) regression was used on the sample spectral data and was compared to the contents measured using standard wet chemistry methods. The calibration models were validated using a completely independent set of samples. We used key indicators such as the ratio of prediction to deviation (RPD) and the range error ratio to give an assessment of the performance of the calibration models. Accurate calibration models were obtained for fructose and sucrose content (R2 > 0.85, root mean square error of prediction (RMSEP) of 0.95%–1.26% in the validation models), followed by sucrose and total soluble sugar content (R2 ~ 0.70 and RMSEP > 2.3%). In comparison to the others, calibration of the starch model performed very poorly with RPD = 1.70. This study establishes the ability of the NIRS calibration model to infer soluble sugar content in foliar samples of E. globulus in a rapid and cost-effective way. We suggest a complete redevelopment of the starch analysis using more specific quantification such as an HPLC-based technique to reach higher performance in the starch model. Overall, NIRS could serve as a high-throughput phenotyping tool to study plant response to stress factors.
近红外反射光谱（NIRS）常用于评估饲料或作物的关键养分，但在生态和生理研究中仍未得到充分应用，特别是在定量非结构性碳水化合物方面。本研究旨在开发校准模型，以评估辐射松叶片材料中可溶性糖（果糖、葡萄糖、蔗糖）和淀粉的含量。偏最小二乘法（PLS）回归用于对样品光谱数据进行分析，并与使用标准湿化学方法测量的含量进行比较。使用完全独立的样本集对校准模型进行验证。我们使用预测偏差比（RPD）和范围误差比等关键指标来评估校准模型的性能。我们获得了果糖和蔗糖含量的准确校准模型（验证模型中 R2>0.85，预测均方根误差（RMSEP）为 0.95%–1.26%），其次是蔗糖和总可溶性糖含量（R2~0.70，RMSEP>2.3%）。与其他模型相比，淀粉模型的校准效果非常差，RPD=1.70。本研究确立了 NIRS 校准模型能够以快速且经济有效的方式推断辐射松叶片样本中可溶性糖的含量。我们建议使用更具体的定量方法（如基于 HPLC 的技术）对淀粉分析进行全面重新开发，以提高淀粉模型的性能。总体而言，NIRS 可以作为一种高通量表型工具，用于研究植物对胁迫因素的反应。
