Unmanned Aerial Remote Sensing for Cotton Yield Estimation.

Crop yield is an important metric which aids producers and researchers in making informed management decisions. Yield may be measured at limited spatial resolution and important information about in-field spatial variability is missing. Technology such as unmanned aerial vehicles (UAVs) should be leveraged to estimate yield with minimal labor. Research heretofore rarely addresses the use of multispectral UAV imagery in cotton yield estimation. The objective of this study was to determine whether multispectral UAV imagery could quantify in-field yield variability in irrigated cotton production. This study comprised a randomized split plot design with four replications. The main plot factor was three irrigation rates applied as a percentage of the estimated crop evapotranspiration (ET_c) requirement: 0, 40, and 80 percent. The sub-plot factor was eight cultivars. Multispectral UAV imagery was acquired at weekly intervals from 30 meters above ground level and it was processed to produce high resolution orthomosaics. Derivatives from the orthomosaics were analyzed to assess whether imagery acquired at several growth stages was representative of seed cotton yield. Normalized Difference Vegetation Index (NDVI) had a linear relationship with yield, which was strongest at approximately peak bloom (r² = 0.78). A novel method, Lint Pixel Count, had an equally strong polynomial relationship (r² = 0.78) on the day of harvest. Results suggest that both methods could help quantify seed cotton yield at different time points in the growing season, which warrants continued exploration.