Spatial Analysis of Production Data for Precision Agriculture.

Advances in precision agriculture, particularly with increasing availability of geomatics-based technologies such as yield monitoring, electrical conductivity (EC) and variable rate application now allow agronomists and farmers to generate massive geo-referenced data on yield, agronomic performance and soil properties. However the analysis of such massive geo-referenced data is challenging because commonly used methods (mixed-model analysis or Kriging) are computationally demanding. For the large–scale production data from farm fields, the covariance matrix describing spatial variability is too large to be handled by most existing statistical softwares. In this presentation, we will evaluate different geostatistical methods for spatial prediction for their prediction accuracy and computational efficiencies. The evaluation will be done with simulated data and yield map data collected from a farm in Alberta.