High-Throughput Switchgrass Phenotyping and Biomass Modeling by UAV
Fei Li, C. PIASECKI, Reginald J. Millwood, Benjamin Wolfe, Mitra Mazarei, C. Neal Stewart
Abstract
Unmanned aerial vehicle (UAV) technology is an emerging powerful approach for high-throughput plant phenotyping field-grown crops. Switchgrass (Panicum virgatum L.) is a lignocellulosic bioenergy crop for which studies on yield, sustainability, and biofuel traits are performed. In this study, we exploited unmanned aerial vehicle (UAV)-based imagery (LiDAR and multispectral approaches) to measure plant height, perimeter, and biomass yield in field-grown switchgrass in order to make predictions on bioenergy traits. Manual ground truth measurements validated the automated UAV results. We found UAV-based plant height and perimeter measurements were highly correlated and consistent with the manual measurements (r = 0.93, p < 0.001). Furthermore, we found that phenotyping parameters can significantly improve the natural saturation of the spectral index of the optical image for detecting high-density plantings. Combining plant canopy height (CH) and canopy perimeter (CP) parameters with spectral index (SI), we developed a robust and standardized biomass yield model [biomass = (m × SI) × CP×CH] where the m is an SI-sensitive coefficient linearly varying with the plant phenological changing stage. The biomass yield estimates obtained from this model was strongly correlated with manual measurements (r = 0.90, p < 0.001). Taking together, our results provide insights into the capacity of UAV-based remote sensing for switchgrass high-throughput phenotyping in the field, which will be useful for breeding and cultivar development