A landscape-scale experiment on the changes in available potassium over a winter wheat cropping season

We present an approach for analyzing experiments which uses cokriging to map treatment responses as coregionalized variables. This geostatistical approach enables us to predict treatment responses or contrasts at points or regular blocks across a landscape, or estimate averages over the whole landscape. We use cokriging as it gives coherent estimates of the individual responses and their linear combinations, or contrasts.The method is illustrated with an experiment performed across 4 arable fields (42.1 ha) in southern England where the response is the change in exchangeable potassium over a winter wheat season. The three treatments are different rates of K-fertilization, a control where no fertilizer was applied, a standard agronomic rate, and a high rate.We found that the response varied spatially for all treatments which implies that uniform K-fertilization may be inappropriate. Furthermore, the results for the standard treatment indicate that over time exchangeable potassium will decrease if current management practices are continued. In terms of differences between treatments, for most locations in the landscape, the standard and high rate resulted in a significantly larger increase in potassium than the control treatment. 82.2% of the experimental area showed no significant difference in the response between the standard and high treatments. We believed this mainly due to the large nugget effect in the auto-semivariograms which results in a large kriging variance of the contrast relative to kriged estimate of the contrast.