Abstract
Variable‐rate nutrient management relies on soil fertility maps, yet the associated uncertainty is typically ignored in developing recommendations. In this article, ordinary kriging (OK) and several alternatives that rely on local estimates of uncertainty derived via probability kriging (PK) are evaluated for developing phosphorus (P) and potassium (K) fertilizer recommendations, using soil data collected at two different intensities. A decision support framework that incorporates economic and agronomic criteria to derive block‐average optimal quantile estimates (Opt Q) maximized net return for an 8‐year fertilization program. The economic results show an increase of $7.31 ha−1 and $1.04 ha−1 for P and K fertilization, respectively, using Opt Q rather than OK with the 0.2‐ha sampling grid and $14.79 ha−1 and $8.93 ha−1 for P and K fertilization, respectively, using the 0.5‐ha sampling grid. These results illustrate the importance of accounting for estimation uncertainty in developing variable‐rate fertilizer recommendations.