In a regulatory context, numerical models are increasingly employed to quantify leaching of pesticides and their metabolites. Although the ability of these models to accurately simulate leaching of pesticides has been evaluated, little is known about their ability to accurately simulate long-term leaching of metabolites. A Danish study on the dissipation and sorption of metribuzin, involving both monitoring and batch experiments, concluded that desorption and degradation of metribuzin and leaching of its primary metabolite diketometribuzin continued for 5-6 years after application, posing a risk of groundwater contamination. That study provided a unique opportunity for evaluating the ability of the numerical model MACRO to accurately simulate long-term leaching of metribuzin and diketometribuzin. When calibrated and validated with respect to water and bromide balances and applied assuming equilibrium sorption and first-order degradation kinetics as recommended in the European Union pesticide authorization procedure, MACRO was unable to accurately simulate the long-term fate of metribuzin and diketometribuzin; the concentrations in the soil were underestimated by many orders of magnitude. By introducing alternative kinetics (a two-site approach), we captured the observed leaching scenario, thus underlining the necessity of accounting for the long-term sorption and dissipation characteristics when using models to predict the risk of groundwater contamination.