Utilizing an Aerial Platform to Assess Cotton Plant Population.

In cotton production today, plant stand counts are the most common method used to determine plant population across a given area. This method consists of measuring a selected linear distance of plant row, counting the number of plants within this selected distance, and repeating the counts in a few locations to estimate plant population for a given field. However, this approach is reliant upon a highly uniform plant population across the entire field and can be influenced by human bias. One proposed use of unmanned aerial systems (UAS) is to produce quantitative data to support replant decisions by assessing plant stands. Theoretically, an aerial approach could provide spatially dense information on plant populations across large areas quickly and remove human bias. Therefore, the objective of this research was to investigate the ability of a UAS system to accurately and precisely determine varying plant populations of cotton (Gossypum hirsutum, L.). Field studies were conducted in Grand Junction, TN in 2016 and 2017 and Milan, TN in 2017. Treatments were replicated four times and included seeding rates of 118970, 76480, 33990, 17000, and 8500 seeds ha^-1 in order to produce a range of plant populations. After emergence, cotton plant stands were manually counted and images were obtained from a MicaSense Red Edge (MicaSense, Seattle, WA) multi-spectral sensor mounted beneath a custom quad-copter flying at an altitude of 30 m. Red and NIR spectral band images were collected and stitched using Atlas. A Python coded program was constructed to calculate NDVI and count plants within each plot using the ArcPy plugin of ArcMap 10.5 (ESRI Redland, CA). Simple linear regression was run in JMP Pro 13 (SAS Institute, Cary, NC), correlating estimated number of plants to ground-truthed counts. Based on initial results, the utilization of aerial imagery may be a sufficient tool to improve accuracy and efficiency of plant stand assessment.