Estimating buffer coefficients for the phosphorus decision support system: preliminary comparisons among field and laboratory estimates from three Asian upland soils.

A phosphorus decision support system (PDSS) has been developed to diagnose P deficiency and predict P fertilizer requirements for Ultisols and Oxisols. Uncertainty analysis has indicated that a major source of uncertainty in prediction was the value of the buffer coefficient (BC, the increase in extractable P per unit of applied P). In the current implementation of PDSS, BC is predicted using an equation that relates BC with soil clay content. We compared PDSS estimates of BC with those obtained by regressing extracted Mehlich 1 P values against amounts of P applied to soil, (a) in field experiments at crop harvest and one year after P addition, and (b) in laboratory wet-to-dry incubations with 1: 1 soil:P solution mixture. In the field, P amounts were added to upland rice-soybean rotations at 0 to 236 mg P dm -3 at Matalom and 0 to 166 at Siniloan, Philippines, and at 0 to 275 at Sitiung, Indonesia, sites in the IRRI Long Term P Experiment (LTPE) network. In the laboratory, P rates of 0 to 272 mg P dm -3 were added and the soil dried at 30 C for 10 days. The PDSS estimates of BC were 0.22 at Matalom (35% clay), 0.10 at Siniloan (54% clay) and 0.07 at Sitiung (63% clay). The field BC values decreased over time as expected from 0.25 at crop harvest to 0.14 after one year at Matalom and from 0.21 to 0.1 at Siniloan but remained unchanged at 0.09 and 0.11 at Sitiung. The laboratory estimates were similar for soil sampled at different times and were close to the one-year field estimates with values averaging 0.16 at Matalom, 0.07 at Siniloan and 0.12 at Sitiung. These estimates, however, differed somewhat from both field estimates at crop harvest and PDSS predictions. While PDSS predictions were generally in the range of the field and laboratory estimates, there was a lack of consistency in the rankings in this limited range of sites and clay content. A much larger set of sites and comparisons would be needed to further refine the prediction of BC.