I don’t work in a hotbed of variable-rate fertilizer applications. For my clients, 75% of their nutrient programs are based on liquid dairy manure. This greatly reduces the need for dry phosphorus and potassium on most fields. The bottom line: dairy manure takes care of most of the corrective P and K applications, and the dry fertilizer that we use is intended to maintain soil test levels.

The one soil amendment that might make sense for us to apply at a variable rate would be lime. Fortunately, most of our soils in eastern Wisconsin have a pH in the range of 7.3 to 7.7, so we rarely need to apply lime.

Time for a Change

Even though I seldom design a soil sampling protocol for variable-rate fertilizer, I have always questioned the science behind point-sampling on a 2.5-acre grid and kriging data from specific points to produce a map that supposedly shows the variation of soil test values within a field. My skepticism was validated a few years ago when I saw Raj Khosla, precision ag professor at Colorado State University, present on field variability of soil test values. He explained that one of his study fields was point-sampled on a 2.5-acre grid. Then, the grids were further divided progressively down to less than a quarter of an acre to determine the correct spatial intensity needed to produce an accurate map of soil nitrate variability. His study concluded that individual points of soil sample locations need to be in relative close proximity in order to be spatially dependent. Each soil nutrient will have a different distance upon which two points can be considered to be related. Also, this distance will vary somewhat between fields and regions. The moral of the story is that it is probably not logical to point-sample on a 2.5- or even a 1-acre grid and then allow a computer to interpolate values between the points to produce a scientifically valid map of variability. In such situations (i.e. spatially independent data), our best estimate is a local average within each grid cell.

I recently saw a presentation by University of Kentucky soil scientist Joshua McGrath that reinforced this concept. His presentation dealt with variable-rate phosphorus and a similar study of point-sampling using different grid sizes down to a quarter of an acre. He concluded that “interpolated soil sample maps (>¼ acre grid) are unreliable at best.” He also stated that the value of any point in the grid was better predicted by averaging all of the sample points than by using nearby values to interpolate a predicted value for the point.

Something To Consider

I am sure you share my desire to give clients the best possible recommendations. This information needs to be considered when designing a protocol for precision soil sampling and data interpretation. Nobody knows each of your client’s fields better than you and your client. We owe clients the consideration of having our beliefs challenged, so we can either reaffirm or modify them to make the best possible recommendations with confidence. When asked to provide a map of soil test variability, I sample and treat each 2.5-acre grid as if it were a separate field by pulling 12 cores in a zigzag pattern across the grid. Other consultants have turned to zone sampling to improve the accuracy of their maps.