Results from a long-term study on leaching and transformation of the herbicide glufosinate-ammonium (ammonium-DL-homoanalin-4-yl (methyl)phosphinate) in a layered soil column were compared with model calculations. Soil columns with layers of loamy sand and sand were treated with 3kg/ha glufosinate-ammonium, applied to the soil surface. The soils were irrigated and the leachate was monitored for glufosinate-ammonium and the main metabolite 3-methyl-phosphinico-propionic acid during a period of 256 d. Calculations for glufosinate-ammonium and the metabolite were performed with the soil column model EXSOL. Linear equilibrium sorption coefficients (K(d)) and first-order half-lives were used for fitting effluent curves for both the herbicide and the metabolite. Sorption coefficients were compared with those values estimated from a relation to the clay content derived from batch studies. Fitted half-lives were compared with the results from die-away and 14CO2 evolution studies of both the parent and the daughter compound. The observed leaching of glufosinate-ammonium could be well simulated using K(d) (0.5 cm3/g) and half-life (11.2 d). The half-life fitted for the metabolite (10-30 d) was a factor 2 to 3 smaller than that measured with batch degradation experiments. This may be the result of using degradation data estimated by 14CO2 evolution, which only indirectly indicates the disappearance of the compound.