Date of Graduation


Document Type


Degree Name

Master of Science in Crop, Soil & Environmental Sciences (MS)

Degree Level



Crop, Soil & Environmental Sciences


Aurelie M. Poncet

Committee Member

Brye, Kris R.

Second Committee Member

Daniels, Mike B.

Third Committee Member

Ross, W. Jeremy

Fourth Committee Member

Henry, Chris G.


management zones, precision agriculture, prescription map, variable rate seeding, within-field variability, yield prediction


Soybean seeding rate (SR) is selected according to planting date, maturity group, soil properties, and yield goal. Even though one SR is applied in most fields, site-specific adjustments with variable-rate seeding (VRS) technology can help optimize production. The project goal was to determine if VRS could be beneficial to soybean producers in Arkansas. The project objectives were to i) assess planter performance and soybean yield response to five seeding rate treatments, ii) characterize the drivers of yield variability, and iii) create a-posteriori prescription maps for VRS. Five SR treatments (185,000, 247,000, 309,000, 370,000 and 432,000 seeds ha-1) were applied using a randomized complete block strip design in two site-years referred to as production fields A (2021) and B (2022) in Lincoln County in Arkansas. As-applied SR were collected from the planter. Plant population was calculated from stand counts collected at the four to five trifoliate stages. Yield monitor data were collected at harvest. As-applied SR, plant population, and yield data were averaged by strip and analyzed using mixed-effect models. While significant differences in as-applied SR and plant population (P=0.05) were observed between treatments, there was no difference in yield. Yet, the within treatment yield variability calculated as a coefficient of variation ranged from 4% to 14% and further analysis was computed to identify the drivers of in-field variability. Soil mapping unit (SMU) information was downloaded from the Soil Survey Geographic database. Soil samples were collected in 91 and 80 locations in fields A and B to characterize soil pH, potassium (K), phosphorus (P), and soil texture defined by both the textural class and percent sand, silt, and clay. Digital elevation models (DEM) were downloaded from the United States Geological Survey public data repository and used to compute flow accumulation. A total of 3,586 and 2,153 grid points were created in field A and B. Soil texture, pH, K, and P, SMU, elevation, and flow accumulation were estimated in each grid point using interpolation or extraction. A 100-fold cross-validation with a 10% calibration/90%validation data split was computed to identify the model that best describes site-specific relationships between yield, SR, and soil properties in each field. The best model described yield as a function of the interaction between SR and both soil pH and soil K in field A. The best model described yield as a function of SMU, percent sand – both linear and quadratic effects-, percent clay, and the interaction between SR and percent clay in field B. As the interaction between SR and other parameters were significant in both fields, a-posteriori prescription maps were created using the following approach. Yield was predicted for each SR treatment and grid point using the best models. Grid points were organized in groups of 4 to create 896 and 538 management zones (MZ) in fields A and B, respectively. Analysis of variance and post-hoc analyses were computed to identify the optimum SR in each MZ. Results were summarized into prescription maps. Future research may include comparison of results between growing seasons, economic analysis, implementation, and on-farm validation.