Making the Cotton Replant Decision: A Novel and Simplistic Method to Estimate Cotton Plant Population from UAS-calculated NDVI

Abstract

One proposed use of unmanned aerial systems (UAS) in crop production is to produce quantitative data to support replant decisions by assessing plant stands. Theoretically, analysis of UAS imagery could quickly determine plant populations across large areas. The objective of this research was to investigate the ability of UAS to quantify accurately varying plant populations of cotton (Gossypium hirsutum L.). Field studies were conducted in Jackson, Milan, and Grand Junction, Tennessee in three consecutive growing seasons. Treatments included five seeding rates ranging from 8,500 to 118,970 seed ha-1. After emergence, cotton plants were manually counted and images were collected in 2016 and 2017 with a MicaSense RedEdge multispectral sensor and in 2018 with a Sentera Double 4K multispectral sensor. Sensors were mounted to a quad-copter UAS flying at altitudes of 30, 60, 75, and 120 m above ground level. Spectral properties were assessed to generate normalized difference vegetation index (NDVI) thresholds that were used to limit the analysis to only plant material. Images were processed and analyzed to estimate number of plants and compared to actual plant populations within each plot. Images obtained from lower altitudes proved to be more accurate, with greatest correlations to actual ground-truthed plant populations from data collected at an altitude of 30 m. The utilization of the described novel method of estimating cotton plant population from NDVI-calculated UAS imagery might improve upon spatial and temporal efficiency in comparison to current methodology of estimation.