Cotton Yield Monitor Values Drift over Time

On-farm experimentation (OFE) using precision agriculture tools is an important resource to understand the spatial variation of crop response to management practices and thus improve agronomic decisions. High-quality yield data is fundamental for getting real insights from this type of experiments. Previous researches have demonstrated that yield data quality affects the outcome of OFE. Many reports have also shown that cotton yield monitors are sufficiently affected by varietal properties and lint turnout to alter the inferences made from data with multiple varieties. Recent researches observed that the errors were correlated with the time of harvest, which can introduce artificial trends in the yield maps. Thus, there is great interest in finding new methods to improve yield monitor information quality. The aim of this work was to investigate the existence of temporal trends affecting cotton yield monitor errors and develop methodologies to improve yield data quality. A data-fusion technique was developed using high-frequency data (1 Hz) from the cotton mass flow sensors and the low frequency (1/500 Hz) information available for each module in the new cotton harvester models. Errors up to 30% were observed comparing the extreme conditions of the day. The errors were correlated with cotton moisture, air relative humidity and time of harvest. For the new machine models, moisture and weight of each module can be used to correct the data and increase the quality of the yield maps. This data fusion could be implemented on the real-time yield mapping software or in post-processing. For older machines, in which this data is not available, post-processing using quadratic regression with the time of day can be used with a similar performance. This procedure represents an important step to improve the use of sensor-based yield data to evaluate on-farm trials.