Development of near‐infrared spectroscopy calibrations to quantify starch and soluble sugar content in the roots of Rumex obtusifolius

Abstract

Decruyenaere V, Clément C, Agneessens R, Losseau C & Stilmant D (2011) Development of near‐infrared spectroscopy calibrations to quantify starch and soluble sugar content in the roots of Rumex obtusifolius. Weed Research 52, 1–5.SummaryThe long‐term objective of non‐chemical measures applied to control a perennial weed is to limit its vigour and regrowth capacity by reducing its acquisition of resources and subsequent allocation to reserve organs. In this context, we tested the possibility of developing near‐infrared spectroscopy (NIRS) calibrations to quantify non‐structural carbohydrates in Rumex obtusifolius roots. The database used to develop the initial calibration contained 25 reference samples in which starch and total soluble sugar content ranged from 10.4 to 47.9% and from 12.4 to 22.2% of the dry matter (DM) respectively. The calibration performances were good, with an R² higher than 0.85 for both the tested parameters and the standard error of cross‐validation (SECV) of 1.32% and 1.03% for the starch and total soluble sugar estimates. However, validation with an independent data set of samples collected later led to a high standard error of prediction (SEP) and poor performances for total soluble sugar quantification, because the range of the validation samples was not represented in the calibration data set. A global calibration covering the diversity of the initial calibration and validation data sets was developed, and validation was performed on 10 samples discarded from this global data set. With ratios of standard deviation of validation set to SEP (RPD) of 8.0 and 5.4 for total soluble sugar and starch, respectively, we concluded that both calibrations could be used to analyse R. obtusifolius roots. The necessity for the calibration to cover sample variation reinforces the need to conduct reference analyses on 5–10% of the samples of each new set analysed, to improve calibration performances.