Estimating Switchgrass Biomass Yield and Lignocellulose Composition from UAV-Based Indices
2025
Daniel Wasonga | Chunhwa Jang | Jung Woo Lee | Kayla Vittore | Muhammad Umer Arshad | Nictor Namoi | Colleen Zumpf | DoKyoung Lee
Innovative methods for estimating commercial-scale switchgrass yields and feedstock quality are essential to optimize harvest logistics and biorefinery efficiency for sustainable aviation fuel production. This study utilized vegetation indices (VIs) derived from multispectral images to predict biomass yield and lignocellulose concentrations of advanced bioenergy-type switchgrass cultivars (&ldquo:Liberty&rdquo: and &ldquo:Independence&rdquo:) under two N rates (28 and 56 kg N ha&minus:1). Field-scale plots were arranged in a randomized complete block design (RCBD) and replicated three times at Urbana, IL. Multispectral images captured during the 2021&ndash:2023 growing seasons were used to extract VIs. The results show that linear and exponential models outperformed partial least square and random forest models, with mid-August imagery providing the best predictions for biomass, cellulose, and hemicellulose. The green normalized difference vegetation index (GNDVI) was the best univariate predictor for biomass yield (R2 = 0.86), while a multivariate combination of the GNDVI and normalized difference red-edge index (NDRE) enhanced prediction accuracy (R2 = 0.88). Cellulose was best predicted using the NDRE (R2 = 0.53), whereas hemicellulose prediction was most effective with a multivariate model combining the GNDVI, NDRE, NDVI, and green ratio vegetation index (GRVI) (R2 = 0.44). These findings demonstrate the potential of UAV-based VIs for the in-season estimation of biomass yield and cellulose concentration.
显示更多 [+] 显示较少 [-]