Problems using intensive sampling of soil moisture and nitrates to predict yield variation: inferences for site-specific farming

Non-Peer ReviewedSeveral conventional agronomic experiments in the Brown and Dark Brown Soil Zones, with emphasis on an experiment conducted on a knob and kettle landscape, were used to determine if intensive sampling for soil nitrates and water was useful for predicting yield variation and thereby represent a potential strategy for variable rate application of N fertilizer. Although there was a large range in soil water and, thus, crop water use in each of three years on the knob and kettle landscape, the durum grain yield was not strongly correlated with soil water at seeding. This was attributed to negative correlation between yield and soil water for areas having significant unexploited soil water from habitually poor crop growth combined with the generally positive correlation between soil water and yields for productive areas of the field. There was a poor relationship between soil nitrate in the previous fall and grain yields. The response to N and water varied widely within the field and, thus. the response function for each small area needs to known to successfully use a map of soil nitrates and water to predict optimum N fertilizer rates. Although the N supplying capability of the soil is an important factor determining the crop response to N for a specific field area, for medium-textured soils on glaciolacustrine landscapes, there was not a positive correlation between potentially mineralizable N and crop yield. On rolling landscapes, crop yields were well correlated to surface organic carbon concentrations but this correlation was poor on gently undulating landscapes. Grain yields were relatively well correlated with yields the previous year on all landscapes providing drought years were excluded (during which yields were often negatively correlated with past yields). This suggests that a field yield map from a combine yield monitor and global positioning system will be useful for identifying areas of the field that should be managed similarly because they have the same general historical response to weather and fertilizer. Since lower yielding areas within a field often do not have lower crop water use than higher yielding areas, a potentially effective site specific management system would involve identifying and correcting (if possible) the yield-limiting factors in the less productive areas of the field. Intensive soil sampling to map water and N was not useful for predicting yield variation so that practice does not appear to be worthwhile for site-specific farming systems in the Brown Soil Zone of Saskatchewan