Understanding crop yield variability as affected by spatial field characteristics and varying weather conditions is critical in the development of site-specific management systems. The objective of this project was to assess the temporal and spatial effects of soil, landscape, and weather covariates on corn yield for a Missouri field. We used Bayesian hierarchical general linear modeling for this analysis. The model included a mean structure of covariates, a random spatial effect and a fixed temporal effect. Corn yield maps from a 36-ha claypan soil field were obtained using combine yield monitoring for 1997, 1999, 2001 and 2003. Yield and soil property maps were aggregated to a 30 meter spatial scale. The covariates included field elevation and slope, soil electrical conductivity (EC), mean daily maximum temperature (mTEMP) and cumulative precipitation (cPREP) for July and August. The conditional autoregressive (CAR) model was used to model the spatial association among grids. The computations were performed through Gibbs sampling using a free package, WinBugs. Results showed that the CAR model was able to capture a great deal of spatial yield association among grids. In general, slope and soil EC had a negative effect on corn yield. Further, cPREP had a positive effect while mTEMP showed a negative effect on corn yield. Temporally, yield slightly decreased over the 1997 to 2003 time period. However we do not attribute this to any long-term change in climate but simply to random occurrence.

Keywords: Crop yield variability, yield map, hierarchical Bayes, general linear model, WinBugs.