Submitted to: Computers in Agriculture
Publication Type: Peer Reviewed Journal
Publication Acceptance Date: November 14, 2003
Publication Date: March 1, 2005
Citation: Jaynes, D.B., Kaspar, T.C., Colvin, T.S., 2005. Identifying potential soybean management zones from multi-year yield data. Computers in Agriculture. 46(1-3):309-327. Interpretive Summary: Varying the rate of agricultural inputs across a field according to the site-specific needs of a crop holds promise for producing crops with fewer inputs and thus improved profitability for farmers and reduced impacts of farming on the environment. To be successful, variable rate applications require that production fields be divided into smaller areas or management zones where inputs such as fertilizer can be tailor applied. It has become obvious over the past decade that identifying management zones within non-irrigated soybean fields is not an easy task because of the great seasonal variability in yield patterns within fields. In this research, we demonstrate that precise records of soybean grain yield from several years can be used to classify the field into a few zones that reflect stable yield behavior within a field. We also demonstrate that easily measured auxiliary field information such as elevation, slope, and soil electrical conductivity can be used to predict the distribution of these stable yield zones within a field. Using these auxiliary data should make it relatively inexpensive to identify stable yield zones within other fields. The next step in the research is to quantify whether these yield response zones can be used as management units for adjusting the application rate of inputs such as fertilizer. These findings will be of value to researchers, farmers, agricultural consultants, and others interested in applying variable rate techniques to crop production.
Technical Abstract: 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 groups or clusters with similar temporal yield patterns and investigated the relationships between these yield clusters and the easily measured field 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 percent 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 yield data may be useful in constructing effective management zones within fields with and without detailed spatial yield data.