Perennial ryegrass biomass retrieval through multispectral UAV data

Frequent biomass measurement is a key activity for optimal perennial ryegrass (Lolium perenne) management in intensive forage-based dairy operations. Due to the necessary high-frequency (i.e., weekly or monthly) pasture monitoring and continuous trend of larger farms, such perceived as an operational bottleneck. Consequently, substantial effort directed development accurate automated technological solutions assessment. The popularization unmanned aerial vehicles (UAVs) combined with multispectral cameras should allow observational system able deploy machine learning algorithms near real-time dry-matter (DM) mapping. For successful operation, these systems deliver radiometrically orthomosaics robust models generalize across different periods. Nevertheless, accuracy radiometric calibration generalization ability seldom evaluated. Also, pipelines require minimum processing power fast deployment. This study has established two-year experiment comparing reflectance measurements between handheld spectrometer commercial UAV camera. Different based on regression-tree architecture were contrasted regarding accuracy, speed, model size. Model performances validated, providing error-metrics baseline temporal validation. results have shown that standard procedure imagery sub-optimal, requiring further post-processing presenting low correlation spectral bands dates. after post-calibration, use presented better error than point-based sensors, respectively displaying average 397.3 464.2 kg DM/ha when employed alongside best performing algorithm Cubist). When trained validated years, performance was largely reduced deemed unfit purposes. Cubist/M5 family exhibited advantageous characteristics compact structure, allowing higher level interpretability, while smaller size faster deployment Random Forest, Boosted, Bagged Regression Trees algorithms.