This study was carried out with the objective of evaluating the phosphate adsorption and desorption processes and relate them to the soils mineralogy. Thirteen profiles were selected from three Brazilian States and their A and B horizons were collected. The samples were submitted to routine physical and chemical analyses and were subjected to intensive mineralogical characterization. Additionally, the soils were analysed by two techniques of recent use in Soil Science: Atomic Force Microscopy and Diffuse Reflectance Spectroscopy. The first technique presented small advantage when compared with the traditional Electron Microscopy, but both technique did not allow a good particles distinction. The Diffuse Reflectance Spectroscopy showed a good soil potential for studies of Soil Science, due to its simplicity and low cost. The color characterization, by this method, allowed an excellent quantification of the iron oxides and also the calculation of the hematite/(hematite + goethite) ratio. Organic matter did not affect the goethite and hematite quantification, but affected the color xi coefficients determination. In the phosphate adsorption study the soils exhibited a large adsorptive capacity and almost all of them presented maximum adsorption values above 1,000 μg/g. The pH in NaF 1 mol/L and gibbsite were directly and, caulinite was inversely correlated with phosphate maximum adsorption. By stepwise regression analysis, goethite was the main mineral individually related with the phenomenon, and when considered together with gibbsite and caulinite explained more than 94% of the maximum adsorption variation. When most of the soils characteristics were considered in the stepwise regression analysis, the P-remainder and redness ratting with an inverse relationships and pH (NaF) and specific surface (N 2 ) with a direct relationship, explained more than 96% of the maximum adsorption variation. The A and B horizons and the yellowish and reddish soils presented similar phosphate maximum adsorption, however it was observed a visual tendency of higher adsorption values associated with B horizon and yellowish samples in the adsorption curves. In the desorption study, the Johnson-Mehl model allowed successfully to fit the kinetic curves, with desorption maximum values ranging from 60 to 1,000 μg/g. It was recovered 25 to 66% of phosphorus previously applied, and the phosphate maximum desorption, by resin extraction, occurred approximatelly with 24 hours of reaction. The significant positive correlation between phosphate desorption and adsorption was caused by the chosen method to apply the phosphorus before incubation. The desorption process was dependent of the amounts previously applied to incubation. The maximum desorption significantly correlated with gibbsite contents and Al 2 O 3 from sulfuric and digestion. The percent desorbed was inversely associated with iron oxides and surface area and, directly associated with caulinite, both statistically significant. There were no statistical differences between, yellow and red soil colors but the B horizons exhibited the highest maximum desorption values and lowest desorption velocity. The percent desorbed was similar in these two horizons.
