The Development of Hyperspectral Distribution Maps to Predict the Content and Distribution of Nitrogen and Water in Wheat (Triticum aestivum)

Abstract

Quantifying plant water content and nitrogen levels and determining water and nitrogen phenotypes is important for crop management and achieving optimal yield and quality. Hyperspectral methods have the potential to advance high throughput phenotyping efforts by providing a rapid, accurate, and nondestructive alternative for estimating biochemical and physiological plant traits. Our study (i) acquired hyperspectral images of wheat plants using a high throughput phenotyping system, (ii) developed regression models capable of predicting water and nitrogen levels of wheat plants, and (iii) applied the regression coefficients from the best-performing models to hyperspectral images in order to develop prediction maps to visualize nitrogen and water distribution within plants. Hyperspectral images were collected of four wheat (Triticum aestivum) genotypes grown in nine soil nutrient conditions and under two water treatments. Five multivariate regression methods in combination with 10 spectral preprocessing techniques were employed to find a model with strong predictive performance. Visible and near infrared wavelengths (VNIR: 400–1,000nm) alone were not sufficient to accurately predict water and nitrogen content (validation R2 = 0.56 and R2 = 0.59, respectively) but model accuracy was improved when shortwave-infrared wavelengths (SWIR: 1,000–2,500nm) were incorporated (validation R2 = 0.63 and R2 = 0.66, respectively). Wavelength reduction produced equivalent model accuracies while reducing model size and complexity (validation R2 = 0.69 and R2 = 0.66 for water and nitrogen, respectively). Developed distribution maps provided a visual representation of the concentration and distribution of water within plants while nitrogen maps seemed to suffer from noise. The findings and methods from this study demonstrate the high potential of high-throughput hyperspectral imagery for estimating and visualizing the distribution of plant chemical properties.

Multivariate Analysis Models Based on Full Spectra Range and Effective Wavelengths Using Different Transformation Techniques for Rapid Estimation of Leaf Nitrogen Concentration in Winter Wheat

“…However, the sensitive absorption wavelength of N lies in short-wave-infrared, which is easily obscured by water-vapor absorption characteristics (Curran, 1989;Fourty et al, 1996;Feng et al, 2008). Moreover, the spectral sensing information obtained in this study was from field-based spectral radiometers; research on aerial-based hyperspectral imagery for crop N status estimation is an important tendency in precision agriculture (Nigon et al, 2015;Bruning et al, 2019). Unlike conventional field-based spectrometers that only collect spectral information for a single point, a hyperspectral imaging system can obtain images of the whole target for each wavelength recorded (Wu and Sun, 2013).…”

“…Multiple linear regression is a common approach used to calibrate the relationship between multiple independent variables and a dependent variable, which was successfully used for the evaluation of in situ canopy spectra and involved stretching the results of a simple linear regression analysis from a single dimension into multiple dimensions (Bruning et al, 2019).…”

“…Nonetheless, only a limited number of studies made an attempt to evaluate the aforementioned effective wavelength selection methods in combination with spectral transformation techniques (CR and FDR) for a comparative and comprehensive estimation of leaf N concentration of winter wheat. Moreover, limited previous research, along with its recommended algorithms and spectral indices, has been conducted to estimate the leaf N concentration in the same ecological locations (Montes et al, 2011;Cao et al, 2017;Zhao et al, 2018;Bruning et al, 2019), neglecting the problem of unsynchronized winter wheat growth stages under different conditions. Therefore, further work is needed to systematically analyze the performance of multiple methods for predicting leaf N concentration in winter wheat under different ecological areas, unsynchronized growth stages, cultivars, and N supply.…”

Leaf Nitrogen Concentration and Plant Height Prediction for Maize Using UAV-Based Multispectral Imagery and Machine Learning Techniques

“…RF is one of the most powerful methods in the current literature related to machine learning tasks [57][58][59][60][61][62]. The increase in data dimensionality is often seen as a problem for most traditional methods.…”

A Machine Learning Framework to Predict Nutrient Content in Valencia-Orange Leaf Hyperspectral Measurements

“…N deficiency is linked to a characteristic chlorosis symptom, which is observable at the visible spectra [21,25,28]. Still, considerable research was also able to identify spectral bands and wavelengths in the near and short-wave infrared regions related to this nutrient [2,3,25,28,36]. Regardless, even though N is a pretty standard nutrient to be evaluated by remote sensing systems, the same cannot be said about others.…”