PHD, Kent State University, 2011, College of Arts and Sciences / Department of Geography

Corn and soybean yield maps derived from yield monitors can be applied for precision agricultural practices by using them to develop or help develop management zones (field areas managed homogeneously). Applying variable fertilizer rates to zones based on need has been shown to increase profits, in part, due to less fertilizer being used than with uniform application. This can have environmental benefits by resulting in less run-off or leaching of fertilizer into the hydrologic system. Many corn and soybean farmers do not have yield monitors to produce yield maps. To help resolve this problem, this research focuses on predicting corn and soybean yield at the field scale. Corn and soybean yield monitor data were acquired and cleaned by different methods to develop better data to base predictions on. Correlations between different Landsat-derived values and corn or soy yield at different growth stages were made. Artificial neural networks (ANN) models based on four independent variables were developed to predict yield and results were compared to multiple linear regression (MLR). Yield cleaning methods that included median neighborhood statistics processing produced better data. Landsat correlations with soybean yield were most reliably high when solely using band 4 during much of the reproductive stage (R²=0.63) while corn yield was better predicted during later vegetative stages. Many different indices proved useful to predict corn, with soil-adjusted vegetation indices having the highest correlations (R² ranging from 0.60 to 0.62). Overall, it was shown that Landsat can predict yield better and, hence, sense crop condition better at distinctly different times of the season for corn and soybeans. ANN predicted yield slightly better than MLR, having an R² value 0.03 higher and increased the R² value with the Landsat crop condition variable by 0.115. Additionally, a Landsat-based county corn yield prediction model that included imagery from the end of July to the latt (open full item for complete abstract)

Committee: Mandy Munro-Stasiuk (Advisor); Scott Sheridan (Committee Member); Emariana Taylor (Committee Member); Joseph Ortiz (Committee Member); Murali Shanker (Committee Member) Subjects: Agriculture; Geographic Information Science; Remote Sensing