MAXIMUM PHOSPHORUS ADSORPTION CAPACITY AND BINDING ENERGY CONSTANT OF AN OXISOL FITTING DIFFERENT LANGMUIR MODELS

Maximum phosphorus adsorption capacity (Pmax) is a useful tool for characterizing the phosphorus (P) sorption of soils, and the Langmuir model, which allows this estimate, has widespread use. However, estimated values of Pmax and the binding energy constant (k) may be quite different when fitted to non-linear or linearized Langmuir models, or models with single or multiple regions. The objective of this study was to compare the effect of using different methods to fit the Langmuir model to the estimation of Pmax and k values. A single soil with high Pmax was mixed with increasing amounts of sand, producing artificial soils with increasing P adsorption capacities but with the same clay mineralogy. Soil from the B horizon of an Oxisol containing 800 g kg(-1) clay was mixed with washed sand in amounts necessary to obtain samples with 0, 200, 400, 600, and 800 g kg(-1) of clay. These artificial soils were incubated for 30 days with lime to raise the pH in water level to 6.0. The adsorption isotherms were constructed and the data fitted using the following Langmuir model approaches: NLin - nonlinear with a single region; L-1R - linearization with a single region; L-2RG - linearization with two regions and graphical adjustment; L-3RG - linearization with three regions and graphical adjustment; and L-2RE - linearization with two regions and statistical fitting. The results showed that the Pmax values estimated by all methods were dependent on the clay content of soils and could be used to characterize the P adsorption of the soil. However, when fitted to more than one adsorption region, the Pmax values for the last region were significantly higher than those observed after incubation of the soil with different P levels. Therefore, the Pmax of the last region in multi-region models may not be appropriate for characterizing soil P adsorption capacity. As expected, the k values were dependent on soil clay content in the first (or single) region of the linearized models, but they did not follow this trend in the nonlinear model. Thus, caution should be taken in interpretation of the k constant fitted to nonlinear models.