Skip to main content
ARS Home » Southeast Area » Stoneville, Mississippi » Crop Production Systems Research » Research » Research Project #440191

Research Project: Predicting Nitrogen and Sulfur Deficiency in Corn using Optical Sensors and Weed Mapping with UAV Multispectral Sensors

Location: Crop Production Systems Research

Project Number: 6066-22000-089-001-S
Project Type: Non-Assistance Cooperative Agreement

Start Date: Jun 1, 2021
End Date: May 31, 2026

The Department of Plant and Soil Sciences at Mississippi State University (MSU) and the USDA-ARS Crop Production Systems Research Unit (CPSRU) will evaluate optimal nitrogen (N) and sulfur (S) rate for corn production in Mississippi, testing the feasibility of optical sensors in N deficiency prediction under varying rates of S, and testing the feasibility of optical sensors in S deficiency prediction under varying rates of N. The results from this study will provide a fast-diagnostic tool to predict S deficiency in corn. This project will train a team of future precision agriculture (PA) agronomists in modern digital agriculture techniques to tackle agricultural and environmental issues far into the future. Infield weed mapping with UAV multispectral sensor: quadcopter vs. fixed wing study, MSU and USDA-ARS-CPSRU scientists will 1) Identify the specific altitude for both platforms to detect weeds, 2) Evaluate the maximum speed for both platforms to detect weeds, and 3) explore potential for variable herbicide application based on weed mapping.

Experiments will be established at two locations in 2021, consisting of 13 treatments within a randomized complete block design with four replications. The treatments will consist of a check plot (0 N and 0 S) and interactions of N application at rates 100, 200, 300 lb/acre and S application at 0, 20, 40, 60 lb/acre. Weather variables, such as rainfall and temperature, will be collected from the local weather stations. Biophysical data including plant population, plant to plant distance variation, plant height, ear height, stalk diameter, leaf area index, total shoot biomass, N and S content in biomass, N and S content in grain, and final grain yield will be collected at the appropriate growth stages. Throughout the growing season, Optical sensor readings will be collected using Trimble Green Seeker (Tipp City, Ohio), Holland Scientific Crop Circle sensor (Lincoln, Nebraska), and sUAS using MicaSense RE-MX (Renton, Washington) multispectral sensors. Reflectance bands will be used to calculate VIs that best detects N and S deficiency. The efficiency of different VIs will be estimated by calculating the sensitivity equivalents (SEq). In developing models, the VIs, weather, and biophysical variables will be considered. In brief, a Pearson correlation test will be conducted, and highly correlated variables are removed. The best subset selection procedure is used to select the best variables to build the models. The best model for each location is chosen using cross-validation prediction error (Cp), Akaike information criteria (AIC), Bayesian information (BIC), and adjusted R2. All data analysis will be done using RStudio® (Boston, MA) statistical software. Models will be independently validated before being recommended to producers. A field experiment will be conducted on about 10 acre field in Stoneville, MS, USA (latitude: 33.445062°, longitude: 90.869967°) at the USDA-ARS, CPSRU research farm. Corn and soybean will be planted on raised beds at 101 cm spacing. Palmer amaranth (Amaranthus palmeri S. Wats.) and Italian ryegrass (Lolium perenne ssp. multiflorum) seeds will be broadcasted at three different densities (low, medium, and high). Nine combinations of both weeds will be replicated four times in a randomized complete block design. Previously identified regions of the spectrum to best detect palmer amaranth (Reddy et al., 2014) and Italian ryegrass (Huang et al., 2018) will be used. The quadcopter will be flown at 10 and 20 mph and the fixed-wing at 20 and 40 mph. Both UAVs will be flown at four different altitudes (10, 30, 60, and 90 m) to detect the ideal altitude for weed mapping. Furthermore, a variable rate spray map will be generated based on the best detection method, and herbicide applications will be made. Post application UAV data will be collected to check the efficacy of the variable rate herbicide application.