Estimation of nitrogen uptake, biomass, and nitrogen concentration, in cover crop monocultures and mixtures from optical UAV images

Cover crops (CC) immobilize mineral soil N in their biomass, preventing N losses during crop rotation intervals. As the CC biomass is incorporated into the soil and decomposes, N is released for the following main crop. The efficiency of CC N uptake and release depends on CC quantity and quality, which can be enhanced in mixtures. Traditional N uptake measurements are labour-intensive and limited in capturing spatial variability. We calibrated relationships between traditional measurements and multispectral data from an Unmanned Aerial Vehicle (UAV) to quantify CC traits with minimal disturbance and high spatial resolution in both monocultures and mixtures. This innovative approach combined vegetation indices, textural features, and a photogrammetry-derived canopy surface model to predict CC traits. Linear models were trained for biomass, N uptake, and C:N predictions, while a K-Nearest-Neighbour model was trained for N concentration. When evaluated on the test set, the calibrated remote sensing models accurately predicted CC aboveground biomass (R2: 0.71, RMSE: 287.1 kg/ha, NRMSE: 11.74 %), N concentration (R2: 0.80, RMSE: 1.77 gN /kg, NRMSE: 6.96 %), N uptake (R2: 0.56, RMSE: 9.38 kgN /ha, NRMSE: 15.08 %), and C:N ratio (R2: 0.62, RMSE: 1.86, NRMSE: 10.98 %). The field experiment included monocultures, bi-, and tri-species mixtures of common vetch (Vicia sativa), black oat (Avena strigosa), and fodder radish (Raphanus sativus). N uptake was similar between treatments, yet the CC species differed in strategies, producing high biomass with low N concentration or vice versa. This study provides a basis for spatially predicting key CC traits using UAV optical data.