Application of Unmanned Aircraft System (UAS)-based RGB and Multispectral Data to Monitor Winter Wheat During the Growing Season

Abstract:

Traditional crop monitoring practices are time-consuming, costly, labor-intensive, and lack appropriate spatial and temporal resolution necessary for efficient monitoring. This study assesses the application of unmanned aircraft systems (UASs) equipped with low-cost RGB and red-edge multispectral sensors to monitor winter wheat over the growing season. Specifically, we used UAS-based data to monitor a biochemical trait (nitrogen content), structural traits (leaf area index (LAI), plant height, and biomass), and yield, all under different pre-plant nitrogen rates (0, 40, 80, 120, and 160 lbs. N/ac). We collected UAS data at emergence, tillering, heading, and maturity growth stages of winter wheat during the growing season (between January and June 2020). We used five vegetation indices (VIs; NDVI, NDRE, ExG, EVI, and CLRE) and UAS-derived plant height to estimate winter wheat nitrogen content, structural traits as well as yield. To validate our results, we measured plant nitrogen content, LAI, plant height, and fresh and dry biomass for each flight campaign. Among all remote sensing metrics, red edge-based VIs including NDRE and CLRE showed significant relationship with nitrogen content (R² = 0.37 and p-val = 0.008 for NDRE; R² = 0.33 and p-val = 0.005 for CLRE) only at tillering stage. For structural traits, among UAS-derived metrics, NDVI was found to be best for estimating plant height (R² = 0.68 and p-val = 0.000 at tillering stage; R² = 0.69 and p-val = 0.000 at heading stage), LAI (R² = 0.52 and p-val=0.000 at tillering stage; R² = 0.30 and p-val = 0.013 at heading stage), and fresh biomass (R² = 0.39 and p-val = 0.003 at tillering stage; R² = 0.42 and p-val=0.002 at heading stage). At harvest, insignificant relationships between VIs and yield were observed. This insignificant relationship was expected due to senescence of winter wheat at harvest (i.e. very low greenness). However, variability in UAS-derived plant height (expressed as standard deviation of height) was strongly associated with yield (R² = 0.44 and p-val = 0.001). This study demonstrates that UAS-based sensing can provide an efficient tool for monitoring various crop traits (e.g. biochemical and structural traits) as well as yield during key growing stages and assist farmers in making best management decisions.