Identifying potential soybean management zones from multi-year yield data

One approach for developing potential management zones for a variable-rate precision-agriculture system is to identify areas within a field exhibiting similar yield behavior. In this study, we applied cluster analysis of multi-year soybean (Glycine max [L.] Merr.) yield to partition a field into a few E groups or clusters with similar temporal yield patterns and investigated the relationships between these yield clusters and the easily measured properties elevation (and the simple terrain attributes derived from elevation) and apparent soil electrical conductivity (ECa). The analysis was applied to 5 years of soybean yield data collected from 224 plots arranged along eight transects spanning a 16-ha field. The partitioning phase of cluster analysis revealed that the 224 locations were best grouped into five clusters. These clusters were roughly aligned with landscape position and were characterized by the yield response to growing season precipitation above or below the 40-year average. Canonical discriminant functions constructed from the simple terrain attributes and ECa predicted correct cluster membership for 80% of the plots. While not perfect, the discriminant functions were able to capture the major characteristics of the yield cluster distribution across the field, indicating that these easily measured variables are strongly related to soybean yield. Applying the functions with high-resolution terrain and ECa attributes, we mapped soybean yield zones within the 16-ha field and an adjacent 16-ha field where multi-year yield data were not available. Cluster analysis of multi-year yield data and easily measured terrain and soil date may be useful in constructing effective management zones within fields and once developed can be applied to similar fields lacking detailed spatial yield data. Published by Elsevier B.V.