Integrating Drones into Weed Management on the Palouse: Final Report

This final report evaluates the feasibility of integrating drone-based multispectral imaging into weed management on the Palouse. Over 130 drone missions were conducted across wheat, barley, and garbanzo fields using RGB, multispectral, thermal, and LiDAR sensors. Field-mapped validation points were collected to support development of classification models capable of distinguishing crops, weeds, and bare ground.

Random Forest classifiers were developed using spectral reflectance values, vegetation indices, and plant height metrics. Results show that weed discrimination is crop-dependent. Garbanzo fields offered clear spectral separation between crops and weeds, while wheat and barley systems proved more challenging due to structural and spectral similarities. Timing of flights was critical, with a narrow June window providing optimal discrimination and management opportunity in cereal systems.

The project establishes mission protocols, sensor strategies, and workflow considerations necessary for scaling drone-based weed mapping in large dryland production systems.