UAV-based hyperspectral imaging to evaluate plant moisture and desiccant response in lentil (Lens culinaris)

Many farmers rely on herbicides as desiccants to dry crops before harvest and aid in harvest efficiency. Optimizing desiccant selection, application rates, and timing is critical for maintaining lentil yield and quality while minimizing herbicide residues. This study utilize unmanned aerial vehicle based hyperspectral imaging to assess desiccant efficacy in lentil crops across field trials at Saskatoon (2019, 2020) and Lethbridge (2023) research farms in Canada. Five conventional herbicide treatments, two application rates of an organic desiccant, and an untreated control were evaluated. Desiccation progress was monitored using visual dry-down ratings and plant moisture content at 0, 3–4, 7, 10, 14, and 24 days after treatment. Spectral data were collected using an aerial hyperspectral system and analyzed with Partial Least squares regression and Support Vector Regression models to predict desiccation response. Year-wise regression analyses yielded R² values of 47–68% for visual ratings and 27–63% for moisture predictions, while combined analyses showed R² values of 65% and 67%, respectively, with Support vector regression consistently outperforming Partial Least squares regression. The area under the desiccation progress curve for visual ratings, moisture , and normalized difference vegetation index quantified desiccant performance. Ammonium nonanoate (32%) exhibited the fastest desiccation, occasionally surpassing conventional herbicides. These results highlight the potential of unmanned aerial vehicle-based hyperspectral and machine learning for evaluating desiccation efficiency and plant moisture content. This approach advances precision crop management and enhances desiccant screening methodologies in modern agriculture.