An Attempt to Optimize Yield Maps by Comparing Yield Data from aPlot Combine and from a Combine Harvester

This paper deals with yield maps derived from yield monitor data of a combine harvester. Every value of the yield map is determined by robust fitting of a paraboloid surface over a spatial neighborhood around the point to be mapped, so that the influence of outliers is bounded or canceled completely. The neighborhood used looks like the top view of the shape of the wings of a butterfly gliding along the harvest tracks. To determine the optimal size and shape of the neighborhood, an experiment was conducted with yield data measured by a plot combine (called as true values) as well as yield monitor data from a commercial combine with a wider head, which then harvested the same stretch behind the plot combine. The commercial combine was equipped with two monitors. The yield maps are considered to have been optimized if the mean squared deviations (mean squared error) between the true and mapped values has been minimized. A large neighborhood proved necessary for both yield monitors, with the result that the best yield maps obtained appear to be very smooth. Both yield maps could be optimized further by rescaling the yield map values so that their dispersion increases.