I agree with you on N.

I've been thinking along those lines for if and when I try to do a variable rate N prescription for the field.

I was however more interested in how to divide the field into zones for soil sampling. I've done some further analysis and I think I've answered my own question.

As I indicated in my first port, this field has already been grid sampled on 2.5 acre grids. I have since calculated the standard deviation and the coefficient of variance for the samples based upon their relationship to soil type. I looked at those samples that fell within soils 109, 12, and 218. These soils had the most sample points, thus I could be more confident in the results.

For soil 109:

P average = 42         K average = 170

P STDEV = 7.6          K STDEV = 10.8

P Variance = 18%      K Variance = 6%

For soil 12:

P average = 42         K average = 167

P STDEV = 17           K STDEV = 42.9

P Variance = 41%      K Variance = 26%

For Soil 218:

P average = 78          K average = 187

P STDEV = 32.3         K STDEV = 43.3

P Variance = 41%      K Variance = 23%

Based on these numbers, it looks like soil 109 would be a good candidate to sample as it's own zone - especially if one were considering K only - 6% variability is very good. However, soils 12 and 218 look to have too much variability to sample as their own zones. 41% variance in P readings is high and I don't think I could justify that level of error when sampling as a zone.