Hyperspectral indicators and characterization of glyphosate-induced stress in common lambsquarters (Chenopodium album L.)

Authors: SOTO, MARIO - University Of Arkansas; PONCET, AURELIE - University Of Arkansas; FRANCE, WESLEY - University Of Arkansas; VELASQUEZ, JUAN - University Of Arkansas; BURGOS, NILDA - University Of Arkansas; Ashworth, Amanda; BRYE, KRISTOFER - University Of Arkansas; KOPARAN, CENGIZ - University Of Arkansas

Submitted to: Smart Agricultural Technology
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: 3/7/2025
Publication Date: 3/14/2025
Citation: Soto, M., Poncet, A.M., France, W., Velasquez, J., Burgos, N., Ashworth, A.J., Brye, K.R., Koparan, C. 2025. Hyperspectral indicators and characterization of glyphosate-induced stress in common lambsquarters (Chenopodium album L.). Smart Agricultural Technology. 11. Article 100890. https://doi.org/10.1016/j.atech.2025.100890.
DOI: https://doi.org/10.1016/j.atech.2025.100890

Interpretive Summary: Crop sensing uses digital technologies to assess crop development and health. However, processing is necessary to generate quantitative information that supports data-driven decision-making. Practical applications of crop sensing vary with spatial, temporal, radiometric, and spectral sensor data resolutions. Finer spatial, temporal, and radiometric resolutions are needed to capture smaller-scale field processes and detect more subtle changes in crop health. Therefore, researchers set out to demonstrate the practical application of hyperspectral sensing for the assessment of weed response to herbicide injury. This study created vegetation indexes for predicting weed damage based on herbicide applications for precision weed management. Once identified, this knowledge could be used to enhance the research process through high-throughput phenotyping or improve the efficiency and timeliness of grower decision-making with regards to weed control and scouting for herbicide-resistant biotypes.

Technical Abstract: Hyperspectral sensors are increasingly used to develop optimized vegetation indices (VIs) that capture plant spectral response to specific stressors, such as nutrient deficiencies or diseases. This work applied hyperspectral sensing to quantify weed response to herbicide treatments to model the spectral response of common lambsquarters (Chenopodium album L., CHEAL) to glyphosate application. Thirteen glyphosate treatments, including one untreated control, were applied to 13 x 4 replicates = 52 CHEAL seedlings cultivated in a greenhouse. Visual injury ratings and non-imaging hyperspectral data were collected 14 days after treatment application. Three signatures were collected from different leaves from each CHEAL plant. Hyperspectral data collected outside the 400 to 1,000 nm range were excluded. Clean spectral signatures were then normalized, smoothed, and averaged per plant. The spectral data resolution were reduced from 0.25 to 5 nm. Visual injury ratings demonstrated treatments created a significant gradient of injury ranging from 0 to 98%. The gradient was associated with visibly manifest differences in leaf spectral signatures. Three dimensionality reduction techniques were used to select 31 wavelengths identified as the best indicators of CHEAL spectral response to glyphosate-induced injury and create 45,732 vegetation indices. Four vegetation indices were then selected to develop a random forest model that accurately predicted injury ratings. The random forest residual mean absolute error was 7.7%. Results demonstrated hyperspectral sensing could be used to accurately assess real-time herbicide efficacy to weed populations for improved precision weed management and enhanced crop health and yield.