Subsoil K availability may be an important factor limiting the precision of K fertilizer recommendations, but it is too expensive to routinely sample and analyze subsoils. Our objective was to develop and validate simple empirical equations for predicting indexes of K supply capability of subsoils based on physical and chemical properties commonly reported by Soil Survey. Ten Minnesota soils were sampled by genetic horizons to a depth of 1 m and were analyzed for exchangeable K from air-dried (EKD) and moist (EKW) samples, nonexchangeable K (NK), potential K buffering capacity (PBCK), water-soluble K (SK), and 39 physical and chemical characteristics. The best regression equations for each K index had the indicated coefficients of determination (in parentheses): EKD (2 = 0.85), EKW (r2 = 0.77), NK (r2 = 0.86), and PBCK (r2 = 0.88). No equation could be developed for SK. Validation of the EKD equation with a total of 88 independent soil samples and Soil Survey data from Minnesota had a standard error of prediction of $49 mg K kg-1. The EKD equation had a larger error of prediction ($63 mg K kg-1) when data from seven midwestern states (n = 512) were included. This equation tended to predict greater amounts of exchangeable K than were measured, especially when EKD < 200 mg K kg-1. EKW could be estimated from EKD (r2 = 0.94). Equations for NK and PBCK could not be validated as thoroughly as that for EKD because the database is not available. Standard error of prediction for NK (n = 6) and PBCK (n = 7) equations were $515 mg K kg-1 and $11 cmol kg-1 mol-0.5, respectively. Depending on the degree of precision desired, these equations could be used for estimating subsoil K supply capability in the presence of adequate Soil Survey data, but the tendency toward overprediction should be recognized.
© Williams & Wilkins 1993. All Rights Reserved.