Nithya Rajan, J. Alex Thmoasson, Amrit Shrestha, Karem Meza Capcha, & Jeffrey Siegfried
Abstract
Nitrogen (N) is an essential macronutrient that controls metabolism during the vegetative growth stages and affect the quantity and quality of cotton yield. Timely detection of N levels in cotton is required to maintain optimal harvest outputs. The use of UAV platforms for data collection facilitate rapid coverage of large field areas. In this study, the spectral and morphological changes in cotton plants in response to four N treatments (0 lb./ac, 50 lb./ac, 100 lb./ac, 150 lb./ac) were observed from sequentially captured UAV-based multispectral images. The structure from motion (SfM) algorithm converted the overlapping 2D images into 3D models that had spectral and morphological information. A temporal analysis of the products derived from the 3D model (multispectral orthomosaic and digital elevation models) were used to study the variations in plot characteristics throughout the growing season. Vegetation indices (NDRE, NDVI) and plant height were tested for sensitivity to the different N treatments. The NDRE index differentiated well between low and high nitrogen treatments early (10%). It was also sensitive to treatments that were given side dressing N application during squaring(5 - 9 %). The NDVI index and plant height were sensitive for the low and high N treatments after the squaring stage, but were unable to detect changes from the side dressing application. The NDRE was also observed to decrease during senescence, thereby proving to be a good indicator for maturity tracking. The total N in biomass samples from each plot were compared the the plot averaged NDRE and plant height values. A good correlation (R2 =0.809 and R2=0.876 respectively) was observed during the mid-season estimates.
Comments