A comprehensive dataset of near infrared spectroscopy measurements to predict nitrogen and carbon contents in a wide range of tissues from Brassica napus plants grown under contrasted environments

Abstract

Winter oilseed rape (WOSR, Brassica napus L.) is the third largest oil crop worldwide that also provides a source of high quality plant-based proteins. Nitrogen (N) and carbon (C) play a key role in plant growth. Determination of N and C contents of plant tissues throughout the growth cycle is crucial in assessing plant nutritional status and allowing precise input management. In the dataset presented in this article, 2427 WOSR samples arising from a large diversity of tissues collected on WOSR diversity were analyzed by near infrared spectroscopy from 4000 to 12,000 cm^-1. At the same time, reference chemical data for the N and C contents of the same samples were determined by elemental analysis using the Dumas method. Partial least squares regression has been used to develop predictive models linking spectral and chemical data, so that new samples can be characterized without the need for reference methods. This dataset could be used to test new calculation algorithms in order to enhance prediction performance or for training purposes. These models can be used as a rapid method for determining N and/or C content, adding to decision-support tools for fertilizer application throughout the plant developmental cycle.

Keywords: Abiotic stress; Brassica napus; Calibration model; Dataset; N and C contents; Near infrared spectroscopy; Prediction.

Fig. 1

Distribution of N and C contents (% of dry matter) within the whole sample collection (n= 2427).

Fig. 2

Near infrared absorbance spectra of 2427 oilseed rape samples. Spectra were acquired on wavenumbers from 4000 to 12,000 cm⁻¹ and shown before pretreatment (A), after a pretreatment using standard normal variate (SNV) (designated by the OPUS8.1 software as the optimal pre-treatment for the N predictive model) (B), and after a combined pretreatment using SNV and first derivate (designated by the OPUS8.1 software as the optimal pre-treatment for the C predictive model) (C).

Fig. 3

Projection of the first two principal components (CP1 and CP2) of a PCA of raw spectral data after removing the outliers. The scatterplot has been colored according to tissue (A), crop season (B) or growing conditions (C).

Fig. 4

Histogram of N content (A) and C content (B) in the samples used for model calibration (Cal) or validation (Val) after removing the outliers.

Fig. 5

Prediction models for N content (A) or C content (B). Data were depicted regarding calibration (Cal, black circles) or validation (Val, grey circles) datasets. The linear regression curve between measured reference values and the values predicted from near infrared spectra are given with the corresponding coefficients of determination (R²) for each dataset.