Assessing the Performance of Multispectral Sensors Mounted on Unmanned Aerial Vehicle for the Prediction of Soil Organic Carbon Levels at Field-Scale
The quantification of soil organic carbon (SOC) is critical for sustainable agricultural production. Conventional field measurements for assessing SOC content are time-consuming, costly and require large soil sampling efforts. The remote monitoring of SOC using unmanned aerial vehicles (UAVs) possesses the capability to be faster and more economically advantageous when compared to conventional soil sampling methods. This research sought to examine the potential of UAV-mounted multispectral (400-800nm) sensors for SOC prediction at the sub-field scale. To do so, UAV-based imagery was acquired over agricultural fields under bare soil conditions. A total of 806 georeferenced soil samples were collected at 20m intervals for each study site. We used multivariate regression analysis to assess the relationship between SOC and reflectance. The R2 and RMSE were calculated between estimated and observed SOC. Laboratory and UAV reflectance were combined to explore the potential of transferrable models that could estimate SOC across various platforms.