THE CONTROL OF ERRORS IN MOMENTARY YIELD DATA FROM COMBINE HARVESTERS

The installation of grain flow sensors and positioning equipment on co mbine harvesters has provided the means to identify spatial variation in crop yield. Changes in the rate of grain flow at a grain flow senso r sometimes do not reflect changes in crop intake giving rise to erron eous momentary yield data. This study examines the effect of data scre ening on the statistical quality of momentary yield data. A data scree n based on the modelling of the pattern of grain flow at the start of harvest runs is the focus of particular attention. Momentary yield dat a were subjected to geostatistical analysis following error screening, Three screening methods, with the most rigorous based on a model of t he grain flow in the first 60 s following the lowering of the cutterba r, were applied to the data. The average of the momentary yield data p assing this screen was similar to the average yield calculated from th e total grain quantity and the area harvested. Geostatistical analysis showed that this screen also gave the highest quality data. The effec ts of different sampling intensities from 1-10 s on the statistical qu ality of the data were minor. Geostatistical techniques for examining and interpolating momentary yield data are outlined and proposals for the automatic control of yield data logging are presented. With this a pproach, the data logging system can be programmed to delete data on t he basis of their association with harvester operating patterns that c ause these errors. (C) 1996 Silsoe Research Institute