Automated precise seeding with drones and artificial intelligence: a workflow

Abstract

Aerial seeding with drones has great potential in forest restoration but faces enormous challenges to be efficient and scalable. Current protocols use blanket seeding throughout the area to be restored, meaning a high demand for seed since many seeds arrive in sites unsuitable for establishment. High precision seeding directed to safe microsites at submeter scale could reduce seed use per hectare, reducing economic and ecological costs, while increasing establishment success. Here, we propose an alternative, precision approach to make drone seeding more successful and efficient. This requires (1) submeter-scale selection of target microsites for seeding founded in ecological knowledge; (2) high-resolution remote sensing imagery to train artificial intelligence (AI) systems in target microsite recognition; and (3) process automation by transferring target microsite coordinates from the AI system to the drone. This will reduce seed inputs per unit area, seedling establishment failure risks, and drone operation costs.