In vitro plant spectral response reveals dust stress

Abstract

Early-stage plant stress detection is a key measure for sustainable agriculture management. Mineral dust as an abiotic stressor affects the physical, chemical, and physiological characteristics of plants, which are linked to the plant's visible and near-infrared (VNIR) reflectance. However, considering the intensity of plant exposure to dust and associated spectral feedback remain unclear. This study investigates the effects of dust particles on the spectral properties of 11 plant species over the growing season by conducting an in-vitro experiment based on VNIR spectroscopy. The capabilities of machine learning algorithms based on VNIR data, including partial least-squares regression (PLSR) and support vector machine (SVM), were also evaluated for dust stress detection. Analyses show that increases in dust concentration lead to (i) reduction of leaf chlorophyll and water contents; (ii) increase of spectral reflectance at 450–490, 640–660, 1370–1450, and 1820–1940 nm; (iii) decrease of spectral reflectance at 530–590, 740–1200 nm; (iv) decrease the slope and height of the red edge; (v) red absorption feature (AF) became smaller and shifted towards shorter wavelength; (vi) reduction of area, width, and depth of AFs at 400–740, 1350–1450, and 1800–1900 nm; and (vii) shift of AF position at 400–740 nm towards shorter wavelength. The results show that, PLSR estimates dust concentration with an R² ranging from 0.83 to 0.95. Additionally, the SVM successfully distinguishes between dust-exposed and non-dust-exposed samples, achieving an overall accuracy of 80–96 %. The research reveals how mineral dust affects the spectral behavior of plants, providing a basis for early-stage dust stress detection through the combination of VNIR spectroscopy and machine learning. Leveraging the research findings, transition from laboratory spectroscopy to hyperspectral remote sensing imagery enables cost-effective and extensive spatiotemporal monitoring, facilitating timely protective measures to mitigate dust-induced damage to plants.

Plant phenotyping relevance

VNIR分光と機械学習により、植物のダストストレス状態をスペクトルから検出・推定する方法を評価しており、植物表現型取得が研究の中心である。

Code and data availability

公開状態または取得可能な本文経路を確認できませんでした。

No evidence-backed public reproduction asset is currently recorded.