Exploiting yield maps and soil management zones

Abstract

Combine harvesters often have yield monitors fitted as standard, and many farmers use them to create yield maps for their fields. These maps contain important information about the spatial and temporal variation of fields. Understanding the variation of a field helps the farmer to make more informed management decisions on how he or she might vary inputs such as fertilizer. The objectives of this project were to establish robust and accessible protocols for the production of reliable yield maps from yield monitor data and create management zones, compare the merits of measuring soil nutrients (a) at the field scale, (b) using management zones and (c) using grids, and to assess the extent to which yield maps can be used to inform the management of soil variation. In the first instance, we reviewed some of the existing software available for cleaning the yield maps and evaluated those including the one we have developed (ROTH-YE). A novel statistical method developed to filter the values associated with flow delay was included in the ROTH-YE software. ROTH-YE performed as well or better than any of the other yield cleaning software we have compared in this study. We then went on to use the cleaned monitor data to create management zones using a spatially smoothed version of a fuzzy k-means classification. This method identifies areas of the field that vary similarly to one another across seasons. In practice, identifying these zones is useful for the farmer as it highlights and quantifies differences in yield that should be explored further.

Farmers sample their fields periodically to assess the nutrient concentration and pH to formulate the fertilizer application rates to the crop. We compared the cost effectiveness of three soil sampling schemes: field-based, management zone-based and grid-based. The advantages of using grid- and zone-based sampling strategies over field-based ones varied from field to field in our study, although for most of the fields grid-based sampling performed better than zone-based sampling. We found that the spatial variation of the yield monitor data (quantified by estimating a variogram) could be used to predict which sampling scheme was likely to be most profitable. We also explored the use of metrics derived from the variogram to help farmers to decide on whether to use uniform or variable rate management within a given field. We reviewed two methods from the literature and, based on our findings, proposed a new method for farmers to rank their fields for the potential for variable management based on these metrics.