Predicting and Testing Site-specific Potassium Fertilization of Maize in Soils of the Tropics—an Example from Thailand

Estimating K requirements in an accurate manner has not been easy in soils of the tropics. Challenges include managing soils with low exchange capacity under high rainfall and the consequent leaching losses of K, estimating K requirements when K-fixing clay minerals are present, and adjusting for crops that remove large amounts of K. In addition, issues in K fertilization are similar to that of phosphate in that reactivity with the soil is important; however, K fertilization also has some similarities with N fertilization wherein the crop requirement is so large as to strongly influence the amounts of fertilizer needed. A method of estimating sitespecific K fertilizer recommendations was proposed and tested in five maize (Zea mays, L.)-producing provinces in Thailand. The accuracy of the fertilizer K prediction was field-tested in 2002 in the province of Nakhon Ratchasima. For 2003, the field test took place in the provinces of Lop Buri and Nan; for 2004, the testing took place in Pitsanulok and Loei provinces. The results in 2002 and 2003 indicated that the initial estimate of critical level of extractable K was too high; no crop responses to added K were obtained in these two years. In 2004, experiments conducted on soils containing less extractable K demonstrated a clear response to added fertilizer K. Fertilizer K predictions of the proposed method in 2004 were compared with field-generated K response curves, and the results indicated that the model generally predicted K requirements more accurately than did the previous method based on a Mitscherlich-Bray equation. Two key coefficients of the model, the soil K critical level and the soil K buffer coefficient, were compared for 2002 and 2003. The critical level was obtained from the field experiment in 2004 and increased as yields increased. Comparisons of model predictions with other published studies on soil K indicated that they were within expected variation. The K model was coded to run together with the P module in the Phosphorus Decision Support System software. (Soil Science 2006;171:968–980)