The P-adsorption capacities of 13 Danish sands were studied by short-term isotherm batch experiments and related to the physico-chemical characteristics of the sands. The maximum P-adsorption capacities (Q) and P-binding energy constants (b) were calculated using the Langmuir-isotherm model. The Freundlich model was also used, but it was not useful for the description of adsorption phenomena per se since it fitted well P-removal data even if precipitation of P-salts occurred simultaneously. The Langmuir model described the data well (R(2)=0.90-0.99) when precipitation of phosphates did not occur and seems therefore to be useful for describing the adsorption processes per se. The relationships between maximum sorption capacities and physico-chemical characteristics of the sands were investigated using classical univariate and partial least squares regression analyses. Among the physico-chemical properties of the sands, Ca and Mg content, grain size, porosity, bulk density and hydraulic conductivity were significantly related (P<0.1) to the maximum adsorption capacity as estimated by the Langmuir model. Using the maximum P-adsorption capacities, it was estimated how long the P-removal can be sustained with the different sands in subsurface flow constructed reed beds. If the most efficient sand for P-adsorption was used, the adsorption capacity would be used up after about 1 year, while, for the less efficient sands, the P-retention would go on for about 2 months. This suggests that, in order to sustain a long-term P-removal in subsurface flow constructed reed beds, precipitation reactions of insoluble P-salts should be promoted. P-binding energy constants were not significantly related to the physico-chemical properties of the sands, except the Ca content, which showed, however, a low correlation coefficient.