Since its inception and acceptance as a benchmarking tool within the economics literature, data envelopment analysis (DEA) has been used primarily as a means of calculating and ranking whole-farm entities marked as decision making units (DMU) against one another. Within this study, instead of ranking the entire farm operation against similar peers that encompass the study, individual data points from within the field are evaluated to analyze the site-specific technical efficiencies estimated at sub-field locations. A hypothetical grid superimposed upon a field creates the DMU’s so that scale efficiency can be visually assessed in a map and spatially analyzed. Input variables include as-applied inputs, geospatial data on soil characteristics, and aerial remotely-sensed imagery. Output variables were based upon yield monitor sensors from harvest equipment from one or more years and therefore one or more crops grown in rotation. Both bio-physical agronomic relationships and economic characteristics were evaluated. Analysis can be conducted on either physical units or on the dollar values of these inputs and outputs. The data here are analyzed by superimposing a grid over a production field in Kansas. Once technical efficiencies were calculated for each site-specific grid cell, the results were spatially mapped across the field to form what looks quite similar to a yield map, only instead of yield, the map now represents the site-specific technical efficiency of that particular field. From this point, tests for global and local spatial autocorrelation indicated the presence of spatial effects, further providing true economic insights into the variability generated either by nature or by the farmer.
These results are useful for the agricultural industry as they represent the first new techniques evaluating efficiency and economics applied to precision agriculture in many years. This initial study can easily be extended to include a farmer’s field with a deliberate intervention, i.e. on-farm experiment; where the technical efficiency of the experiment and in particular regression residuals can be assessed. Additional extensions to this technique can be applied to a community of farmers’ fields in a big data analysis.
